theory of mind and executive function impairments …...theory of mind and executive function...
TRANSCRIPT
Theory of Mind and Executive Function Impairments in Autism Spectrum Disorders and their Broader Phenotype:
Profile, Primacy, and Independence
Dana Wong, B.Sc. (Hons.)
This thesis is presented in partial fulfilment of the degree of
Doctor of Philosophy/Master of Psychology (Clinical Neuropsychology)
of the University of Western Australia
School of Psychology, 2004
ABSTRACT
Impairments in both theory of mind (ToM; the ability to attribute mental states to
oneself and others) and executive function (EF; a group of high-level cognitive
functions which help guide and control goal-directed behaviour) have been
demonstrated in individuals with autism spectrum disorders (ASDs). Both deficits have
been proposed by different groups of researchers as being the single primary cognitive
deficit of autism, which can subsume the other deficit as secondary or artefactual.
However, few studies have examined the nature of the relationship between ToM and
EF in ASDs or conducted a systematic investigation of their relative primacy. This
research principally sought to establish the primacy and independence of impairments in
ToM and EF in ASDs and thereby evaluate the validity of single versus multiple
primary deficit models of autism.
These aims were addressed in two studies, both broad in scope. The first study
was an investigation of the profile, primacy, and independence of ToM and EF
impairments in individuals with ASDs. The sample included 46 participants with ASDs
and 48 control participants matched on age and non-verbal ability. The profile of
impairments was examined by measuring ToM and a range of EF components using
tasks employing, wherever possible, process-pure indices of performance. Primacy was
measured by focussing on i) whether or not the deficits observed were universal among
individuals with ASDs; ii) whether the deficits were able to discriminate individuals
with ASDs from matched controls (i.e., predict group membership); and iii) the ability
of ToM and EF deficits to explain the full range of autistic symptomatology, as
measured by correlating cognitive performances with behavioural indices. The
relationship between ToM and EF impairments was investigated by conducting
correlations between ToM and EF variables as well as analysing the incidence of
dissociations between impairments in the two domains. The ASD group was found to
demonstrate significant impairments in ToM and several components of EF including
planning, verbal inhibition, working memory (in a context where inhibitory control was
required), and both verbal and non-verbal generativity. However, neither ToM nor EF
impairments were able to meet all of the criteria for a primary deficit in ASDs. EF
deficits were found to be more primary, but could not account for ToM as a secondary
deficit, as ToM and EF were found to be independent (i.e., uncorrelated and dissociable)
deficits in the ASD group. This pattern of results suggested that a multiple deficits
model involving at least two independent impairments appeared to best characterise
i
ASDs, but the data were compatible with several variants of such a model (e.g.,
involving distinct subtypes versus a multidimensional spectrum).
The second study was an investigation of ToM and EF impairments in siblings
of individuals with ASDs, who have previously been found to demonstrate a subclinical
“broad autism phenotype”. The main aims of this study were i) to identify whether
ToM or EF deficits could meet criteria for an “endophenotype” or vulnerability marker
for the autism genotype in unaffected relatives, which would have further implications
about the primacy of ToM and EF in ASDs; and ii) to further investigate the validity of
various multiple deficits models of ASDs by examining the pattern of ToM and EF
performance in those showing the broad phenotype. Participants were 108 siblings of
individuals with ASDs and 67 siblings of controls, tested on the same ToM and EF
tasks used in the first study. Confirming the superior primacy of EF deficits found in
Study One, there was no significant difference in ToM performance between ASD and
control siblings, but ASD siblings showed weaknesses on two measures of EF.
Furthermore, there appeared to be different subgroups of siblings demonstrating
different cognitive profiles, consistent with the heterogeneity evident in the first study.
This research indicated that ASDs cannot be explained by a single primary
cognitive deficit. These findings hold important theoretical and empirical implications
and highlight further questions about which type of multiple deficits model might best
explain ASDs.
ii
TABLE OF CONTENTS
ABSTRACT............................................................................................... i LIST OF TABLES...................................................................................... vii LIST OF FIGURES.................................................................................... ix ACKNOWLEDGEMENTS......................................................................... x CHAPTER 1. General Introduction: Explaining Autism...................... 1 1.1 Autism: Diagnosis and epidemiology.................................................. 2 1.2 Explaining autism: The cognitive level of explanation........................ 4 1.3 Overview of the thesis........................................................................ 9 1.3.1 Rationale and aims................................................................... 9 1.3.2 Thesis structure........................................................................ 11 CHAPTER 2. Literature Review: Theory of Mind and Executive Function in Typical Development and in Autism..................................
13
2.1 Theory of mind (ToM) ........................................................................ 14 2.1.1 Defining and measuring ToM.................................................... 14 2.1.2 Models of ToM and its development......................................... 16 2.1.3 ToM in autism........................................................................... 21 2.2 Executive function (EF) ...................................................................... 32 2.2.1 Defining and measuring EF...................................................... 32 2.2.2 Models of EF and its development........................................... 36 2.2.3 EF in autism.............................................................................. 42 2.3 The ToM-EF relationship.................................................................... 54 2.3.1 Models of the ToM-EF relationship........................................... 54 2.3.1.1 Expression accounts.................................................... 55 2.3.1.2 Common conceptual requirements of ToM and EF..... 62 2.3.1.3 Emergence accounts................................................... 66 2.3.1.4 Common neuroanatomical bases for ToM and EF...... 72 2.3.2 The ToM-EF relationship in autism........................................... 78 CHAPTER 3. Selection and Description of Measures......................... 87 3.1 Diagnostic measures.......................................................................... 88 3.1.1 Autism Screening Questionnaire.............................................. 88 3.1.2 Autism Diagnostic Interview – Revised..................................... 89 3.2 IQ measures....................................................................................... 90 3.3 ToM measures.................................................................................... 90 3.3.1 Simple false belief task............................................................. 91 3.3.2 First-order false belief task....................................................... 92 3.3.3 Second-order false belief task.................................................. 93 3.3.4 Dewey stories........................................................................... 94
iii
3.4 EF measures...................................................................................... 95 3.4.1 Tower of London....................................................................... 96 3.4.2 Intra-dimensional, Extra-dimensional Set-shifting task............. 99 3.4.3 Response Inhibition and Load task........................................... 104 3.4.4 Opposite Worlds....................................................................... 106 3.4.5 Relational Complexity............................................................... 107 3.4.6 Pattern Meanings...................................................................... 110 3.4.7 Uses of Objects........................................................................ 113 3.4.8 Stamps task.............................................................................. 114 3.5 Behavioural measures........................................................................ 116 3.5.1 Measures of repetitive behaviour.............................................. 116 3.5.1.1 Repetitive Behaviours Questionnaire.......................... 116 3.5.1.2 Repetitive Behaviours Interview.................................. 117 3.5.2 Measures of social behaviour and communication................... 121 3.5.2.1 Social Behaviour Questionnaire.................................. 121 3.5.2.2 Social and communication ADI-R domains................. 121 CHAPTER 4. Study One: Profile, Primacy, and Independence of Theory of Mind and Executive Function Impairments in Autism Spectrum Disorders................................................................................
123
4.1 Introduction......................................................................................... 124 4.1.1 Aims.......................................................................................... 124 4.1.2 Hypotheses............................................................................... 126 4.2 Method................................................................................................ 131 4.2.1 Participants............................................................................... 131 4.2.2 Procedure................................................................................. 134 4.3 Results................................................................................................ 136 4.3.1 Data screening.......................................................................... 136 4.3.2 Group comparisons on ToM and EF tasks............................... 136 4.3.2.1 False belief tasks......................................................... 139 4.3.2.2 Dewey Stories.............................................................. 141 4.3.2.3 Tower of London.......................................................... 142 4.3.2.4 IDED set-shifting task.................................................. 143 4.3.2.5 Response Inhibition and Load task.............................. 144 4.3.2.6 Opposite Worlds task................................................... 147 4.3.2.7 Relational Complexity.................................................. 149 4.3.2.8 Pattern Meanings......................................................... 149 4.3.2.9 Uses of Objects........................................................... 151 4.3.2.10 Stamps task............................................................... 153 4.3.2.11 Summary and effect sizes of group comparisons...... 154 4.3.3 Universality of ToM and EF deficits.......................................... 157 4.3.4 Ability of ToM and EF variables to predict group membership. 159
iv
4.3.5 Behavioural measures: Group comparisons and derivation of indices used in correlational analyses......................................
162
4.3.5.1 Repetitive Behaviours Interview.................................. 162 4.3.5.2 Social and communicative functioning......................... 164 4.3.6 Correlations between ToM/EF and behavioural measures....... 165 4.3.7 Relationship between ToM and EF........................................... 171 4.3.7.1 Correlations between ToM and EF.............................. 171 4.3.7.2 Dissociations between ToM and EF............................ 175 4.4 Discussion.......................................................................................... 176 4.4.1 Profile of ToM and EF deficits................................................... 177 4.4.2 Primacy of ToM and EF deficits................................................ 186 4.4.3 Independence of ToM and EF deficits...................................... 193 4.4.4 Towards a “multiple primary deficits” model of ToM and EF in ASDs.........................................................................................
199
CHAPTER 5. Literature Review: The Broad Autism Phenotype......... 207 5.1 Autism as a genetic disorder.............................................................. 208 5.2 The broad phenotype.......................................................................... 210 5.2.1 The behavioural phenotype...................................................... 210 5.2.2 The cognitive phenotype........................................................... 212 5.2.2.1 General intellectual ability............................................ 213 5.2.2.2 Specific cognitive deficits............................................. 215 CHAPTER 6. Study Two: Theory of Mind and Executive Function in Siblings of Individuals with Autism Spectrum Disorders....................
221
6.1 Introduction......................................................................................... 222 6.1.1 Aims.......................................................................................... 222 6.1.2 Hypotheses............................................................................... 225 6.2 Method................................................................................................ 226 6.2.1 Participants............................................................................... 226 6.2.2 Procedure................................................................................. 228 6.3 Results................................................................................................ 228 6.3.1 Sibling group comparisons on ToM and EF tasks.................... 228 6.3.1.1 False belief tasks......................................................... 229 6.3.1.2 Tower of London.......................................................... 231 6.3.1.3 IDED Set-shifting task.................................................. 232 6.3.1.4 Response Inhibition and Load task.............................. 233 6.3.1.5 Opposite Worlds task................................................... 235 6.3.1.6 Pattern Meanings......................................................... 238 6.3.1.7 Uses of Objects........................................................... 238 6.3.1.8 Stamps task................................................................. 239 6.3.1.9 Summary of sibling group comparisons....................... 241 6.3.2 Comparisons between ASD siblings and ASD probands......... 242
v
6.3.3 Ability of cognitive variables to predict sibling group membership.............................................................................
243
6.3.4 Proband-sibling relationships within the ASD families............. 244 6.3.4.1 Correlations between proband IQ and siblings’ cognitive performances...............................................
244
6.3.4.2 Correlations between probands’ and siblings’ cognitive performances...............................................
246
6.3.5 Prevalence of deficits in ASD siblings...................................... 246 6.3.6 Correlations between ToM and EF........................................... 246 6.3.7 Dissociations between ToM and EF......................................... 251 6.3.8 Results from behavioural measures......................................... 252 6.4 Discussion.......................................................................................... 254 6.4.1 Endophenotype status of ToM and EF impairments................. 254 6.4.2 Differentiating the multiple deficits models............................... 260 CHAPTER 7. General Discussion: Constructing an Explanatory Model for ASDs........................................................................................
265
7.1 Summary of the findings..................................................................... 266 7.2 Methodological strengths and limitations............................................ 267 7.3 Conclusions on constructing an explanatory model for ASDs............ 269 7.4 Future directions................................................................................. 272 REFERENCES.......................................................................................... 279 APPENDIX A. Repetitive Behaviours Interview – Current Version..... 333 APPENDIX B. Correlations between EF task variables in the control group (Study One)...................................................................................
349
APPENDIX C. Separate ToM-EF correlations for young and old age subgroups within the control sample (Study One)..............................
351
APPENDIX D. Separate group comparisons for young and old age subgroups on EF tasks (Study One)......................................................
353
vi
LIST OF TABLES Table: 1. The five scores computed for each item on the Tower of London............. 99 2. Demographic characteristics of the samples............................................... 133 3. Order of test battery and age range for each test........................................ 135 4. False belief task results: Percentage of participants in each group with
perfect scores [or high scores in the case of the alternative aggregate score] on belief questions, and significance of group comparisons...........
141
5. IDED Set-shifting task results: Percentage of low error scorers in each group for each stage of each task condition, and significance of group comparisons................................................................................................
144
6. RIL task results: Mean (and SD) of each group, and significance of group comparisons, for error and RT difference scores and the shape error score...................................................................................................
146
7. Opposite Worlds results: Mean (and SD) of each group for error/time scores in each condition and difference scores, and significance of group comparisons................................................................................................
148
8. Pattern Meanings results: Mean (and SD) of each subgroup [or the percentage of low error scorers for dichotomous variables], and significance of group comparisons.............................................................
151
9. Uses of Objects results: Mean (and SD) of each group [or the percentage of low error scorers for dichotomous variables], and significance of group comparisons......................................................................................
152
10. Stamps task results: Mean (and SD) of each group [or the percentage of low scorers for dichotomous variables], and significance of group comparisons................................................................................................
154
11. Summary and effect sizes of significant group differences........................ 155 12. Universality of ToM and EF deficits in the ASD group............................. 159 13. Logistic regression analysis of group membership as a function of VIQ,
ToM and EF variables................................................................................. 161
14. Median (and range) of RBI severity summary scores for the ASD and control groups.............................................................................................
162
15. Factor loadings of RBI severity summary scores....................................... 163 16. Raw and partial correlations between cognitive measures and
behavioural factors within the ASD group................................................. 166
17. Raw and partial correlations between cognitive measures and RBI composite scores within the ASD group.....................................................
169
18. Raw and partial correlations between ToM and EF measures within the control group...............................................................................................
173
19. Raw and partial correlations between ToM and EF measures within the ASD group..................................................................................................
174
vii
Table: 20. Summary of significant partial correlations between ToM and EF
variables in the control and ASD groups.................................................... 175
21. The incidence of ToM-EF dissociations in the ASD group........................ 176 22. Demographic characteristics of the sibling samples................................... 227 23. False belief task results: Percentage of siblings in each group with
perfect scores [or high scores for the alternative aggregate] on belief questions, and significance of group comparisons.....................................
231
24. IDED Set-shifting task results: Percentage of low error scorers in each sibling group for each stage of each task condition, and significance of group comparisons......................................................................................
233
25. RIL task results: Mean (and SD) of each sibling group, and significance of group comparisons, for error and RT difference scores and the shape error score...................................................................................................
235
26. Opposite Worlds results: Mean (and SD) and significance of group comparisons for each sibling group for error/time scores in each condition and difference scores, and for each gender for time scores........
237
27. Uses of Objects results: Mean (and SD) of each sibling group, and significance of group comparisons.............................................................
239
28. Stamps task results: Mean (and SD) of each sibling group [or the percentage of low scorers for dichotomous variables], and significance of group comparisons..................................................................................
240
29. Effect sizes, r (and d), of significant group differences between sibling groups and between proband groups..........................................................
242
30. Results of logistic regression analysis of sibling group membership......... 243 31. Raw and partial correlations between proband PIQ and VIQ and
siblings’ scores on ToM and EF measures................................................. 245
32. Raw and partial correlations between ToM and EF variables within control siblings............................................................................................
248
33. Raw and partial correlations between ToM and EF variables within ASD siblings........................................................................................................
250
34. Summary of partial correlations between ToM and EF variables in the control and ASD probands and siblings.....................................................
251
35. The incidence of ToM-EF dissociations in the ASD siblings.................... 252
viii
LIST OF FIGURES Figure: 1. A single primary cognitive deficit model of autism.............................. 5 2. A multiple cognitive deficits model of autism, in which each
cognitive deficit underlies a different domain of symptomatology....... 8
3. An example of a Dewey Story............................................................... 95 4. The starting configuration for the Tower of London stimuli................. 98 5. Stimuli for the Perseveration condition of the IDED set-shifting task.. 102 6. Stimuli for the Learned Irrelevance condition of the IDED set-
shifting task............................................................................................ 103
7. Example of a Relational Complexity item with 1 relational change..... 109 8. Example of a Relational Complexity item with 4 relational changes.... 109 9. Example of a more difficult Relational Complexity item without
consistent relational changes.................................................................. 110
10. One of the five test stimuli for the Pattern Meanings task..................... 111 11. The practice stimulus for the Pattern Meanings task............................. 112
ix
ACKNOWLEDGEMENTS
First and foremost credit clearly goes to my principal supervisor Murray Maybery, who
is a rare treasure in putting his students’ needs before his own. He is unfailingly patient,
encouraging, logical, and sensible. Thanks also to my co-supervisor Joachim
Hallmayer, whose expertise in autism and genetics and constructive feedback on a draft
improved the clarity, accuracy, and coherence of the thesis.
This PhD research formed part of a larger project on the broad autism
phenotype, the Western Australia Family Study of Autistic Spectrum Disorders
(WAFSASD), which was funded by a National Health and Medical Research Council
grant. Alana Maley, research assistant extraordinaire on the WAFSASD, put in
countless hours of recruiting families, driving to opposite ends of the city and state, and
interviewing and testing a seemingly endless number of participants. My hugest
appreciation for all that you contributed. Dorothy Bishop, one of the WAFSASD’s
chief investigators, offered expert guidance throughout the project. Wayne Hill put
together a monstrous database as well as doing a number of the ADI-Rs. Sarah
Davenport, Isabel Fernandez, Kate Fitzpatrick, Elise Mengler, Sarra Miller, Bronny
Morgan, Nicole Petterson and Keira Thomson all helped with testing and/or data entry
for the WAFSASD. Valued assistance was also provided by Matt Huitson, whose task
programming skills saved me a lot of time and frustration, and Herb Jurkiewicz, who
helped with some of the stimuli.
Liz Pellicano shared with me the questions, ideas, and bafflement that go along
with doing autism research, and in doing so managed to help rekindle my enthusiasm
for not only my own research but also research in general, right when it was needed.
My officemates, co-whingers, and distractors Kate Harwood and Mark Woodman
served proficiently as my credibility meters (as well as keeping me up to date on world
affairs). Opinions, grievances, ridicule, coffee, and gossip were also shared with Kate
Frencham, Keira Thomson, Flavie Waters, and Allyson Browne. I was kept fed and
financed by my generous family, particularly in the later stages after my scholarship had
run out. My gorgeous Glen saw me through to the finishing line with a constant supply
of comfort, silliness, and (bad) humour.
Finally, my humble and sincere gratitude to the participants of this research – the
kids both with and without ASDs, their brothers and sisters, and mums and dads – who
gave their time and effort so generously. May this thesis be a step forward in
understanding the puzzle of autism.
x
CHAPTER 1
General Introduction: Explaining Autism
1.1 Autism: Diagnosis and epidemiology
1.2 Explaining autism: The cognitive level of explanation
1.3 Overview of the thesis
1.3.1 Rationale and aims
1.3.2 Thesis structure
1
1.1 Autism: Diagnosis and epidemiology
Autism is classified as a pervasive developmental disorder and is defined and diagnosed
by its clinical symptomatology, rather than biological markers or aetiology. Current
diagnostic criteria, as specified by the Diagnostic and Statistical Manual of Mental
Disorders, 4th edition (DSM-IV; APA, 1994) and the International Classification of
Diseases, 10th edition (ICD-10; WHO, 1992) require the presence of symptoms in three
categories: i) impairment in social interactions, ii) abnormal development of language
and nonverbal communication, and iii) restricted and repetitive patterns of behaviour,
interests and activities. Examples of specific symptoms listed in DSM-IV in the social
domain include impaired use of nonverbal behaviours such as eye contact, facial
expressions and gestures, failure to develop appropriate relationships with peers, and a
lack of spontaneous seeking to share enjoyment and interests; the communication
domain lists features such as a delay in or total lack of language development,
pragmatic difficulties, stereotyped use of language, and impaired pretend play and
imitation; and examples of repetitive behaviours include intense preoccupations, rigid
adherence to routines and rituals, and stereotyped motor mannerisms such as hand
flapping. The DSM-IV criteria specify that six of the twelve symptoms listed must be
present, with at least two from the social domain and one from each of the other two
domains. Delayed or abnormal functioning in at least one of the three domains must
also have been present prior to the age of three years. While it possible for autism to be
identified as young as 18 months (Baron-Cohen, Allen, & Gillberg, 1992; Johnson,
Siddons, Frith, & Morton, 1992), it is more commonly and reliably diagnosed at around
the age of three years or older.
Other pervasive developmental disorders1 such as Asperger syndrome
(individuals with autistic symptomatology who have normal intelligence and adaptive
skills and no delay in the onset of speech) and Pervasive Developmental Disorder Not
Otherwise Specified (PDDNOS; individuals who show significant symptomatology but
who do not meet full criteria for a specific PDD) are generally considered related but
distinct entities on the autism spectrum, although the boundaries and validity of each
diagnosis remain a matter of current debate (Bishop, 2000; Macintosh & Dissanayake,
2004; Miller & Ozonoff, 2000; Ozonoff, South, & Miller, 2000; Rapin, 1997). One of
1 The term “pervasive developmental disorder” refers to the DSM-IV/ICD-10 category which includes autism, Asperger syndrome, Pervasive Developmental Disorder Not Otherwise Specified, Rett’s disorder, and Childhood Disintegrative disorder. Throughout this thesis, the term “autism spectrum disorder” will be used to refer to the former three of these diagnoses.
2
the characteristics of autism is its variability, with symptom severity, intellectual ability,
and degree of language impairment varying widely across individuals. Most studies
estimate that around 70% of individuals with autism are mentally retarded – that is,
have an IQ below 70 (see Fombonne, 2003). When individuals with more broadly
defined autism spectrum disorders (ASDs) are included, the proportion of affected
individuals with comorbid mental retardation decreases substantially; for example,
Chakrabarti and Fombonne (2001) found that less than half of children with ASDs have
Performance IQs less than 70.
Conservative prevalence estimates for autism currently stand at 10/10,000, with
estimates for Asperger syndrome at 2.5/10,000, and at 15/10,000 for PDDNOS, making
a combined prevalence for all ASDs of 27.5/10,000 (Fombonne, 2003). The prevalence
of ASDs has reportedly increased in recent years, with three of the latest surveys
providing estimates around twice as high as the above figures (Baird et al., 2000;
Bertrand et al., 2001; Chakrabarti & Fombonne, 2001). Fombonne (2003) reports that
the median prevalence rate for autism in 16 surveys published between 1966 and 1991
was 4.4/10,000, whereas the median rate for 16 surveys published in the period 1992-
2001 was 12.7/10,000. While this apparent increase has led some to propose various
environmental aetiologies for autism, other possible contributing factors include
changes in diagnostic practice, increased awareness, “diagnostic substitution” (e.g.,
choosing a diagnosis of autism instead of mental retardation for the purposes of
educational placement or funding), earlier diagnosis, and methodological issues (see
Volkmar, Lord, Bailey, Schultz, & Klin, 2004).
Autism is more common in boys than in girls, with a mean sex ratio of 4.3:1
across epidemiological studies; the ratio is higher for non-retarded individuals with
autism, with a median of 5.75:1 across studies (Fombonne, 2003). High socioeconomic
status and immigrant status have been associated with higher rates of autism in some
small samples, but larger, well-designed studies have not supported these associations
(Fombonne, 2003). A number of comorbid medical conditions have also been
commonly associated with autism, with the most prevalent being epilepsy (Fombonne’s
review estimates that 16.8% of individuals with autism also have epilepsy, but this may
be an underestimate given that the median age of the samples is lower than the usual age
of onset of seizures in autism). Proposed associations with other conditions such as
Fragile X, tuberose sclerosis, neurofibromatosis, and phenylketonuria (PKU) are less
well established as many studies do not provide evidence that the prevalence is higher
than predicted by chance (Fombonne, 2003; Volkmar et al., 2004).
3
1.2 Explaining autism: The cognitive level of explanation
The construction of a causal model of autism (and ASDs more broadly) has proven an
extremely complex and challenging task at all levels of explanation: genetics,
neurobiology, cognition, and behaviour2. While we are now confident that autism has a
genetic basis (see Chapter 5), the genetic mechanisms and specific genes involved are
still not understood and non-genetic factors have also been implicated. Attempts to
identify key neuroanatomical and neurobiological abnormalities have resulted in a
variable array of inconsistent findings, with almost all areas of the brain proposed as
being abnormal in autism at one time or another. At the level of behaviour, it remains
unclear whether autism is best conceived of as a unitary syndrome, a set of related but
distinct subtypes, or a continuum or spectrum of abnormalities (Boucher, 1996; this is
discussed further below).
Paralleling this search for convergence at the genetic, neurobiological, and
behavioural levels of explanation has been the pursuit of a core marker or single
primary deficit at the level of cognition. In the absence of a unique biological marker
for autism, the identification of a primary cognitive deficit could help to both define the
boundaries of the disorder and highlight possible neurobiological substrates. The notion
of a single primary cognitive deficit is attractive because it is parsimonious and provides
unity – that is, it is a way of explaining the regular co-occurrence of the triad of
impairments which characterise autism and it justifies the use of a single label,
“autism”. For these reasons, Morton and Frith (1995, 2001; see also Frith, Morton, &
Leslie, 1991) have argued strongly that autism may be explained by a single primary
cognitive deficit which underlies the whole range of autistic symptomatology. The basic
structure of this kind of model is presented in Figure 1.
The notion of a primary or core deficit has been crucial in guiding and
constraining cognitive theories of autism. Michael Rutter was one of the first autism
researchers to promote the idea of a primary deficit, with his treatment of the term
implying that he considered universal manifestation, early appearance, prognostic
significance, and ability to account for performance on a range of tasks to be important
signs of primacy (Rutter, 1968). More recently, a primary cognitive deficit has been
defined as “universal, specific, and necessary and sufficient to cause the symptoms of
2 This distinction of four broad levels of explanation follows Pennington and Welsh (1995), among others, and should be considered provisional. Other divisions are possible; for example, Morton and Frith (1995) collapse genetic and neurobiological factors under one heading, “biological”. Finer divisions are also possible, for example within the level of neurobiology.
4
the disorder...in other words,...the proximal cognitive cause of the behavioural
symptoms of the disorder” (Pennington & Ozonoff, 1996, p. 57). These three criteria of
universality, uniqueness to autism, and ability to explain the behavioural symptoms of
autism have consistently recurred in recent definitions of primacy (e.g., Hughes, 2001;
Ozonoff & McEvoy, 1994; Turner, 1997). An additional criterion commonly used to
assess primacy is that of causal precedence, or the ability of the proposed deficit to
predate and explain the earliest symptoms of autism (Boucher, 1996; Happé, 1994b;
Pennington & Ozonoff, 1991; Pennington & Welsh, 1995; Tager-Flusberg, 2001). Four
key criteria3 for judging the primacy of a cognitive deficit in autism may therefore be
identified as:
1. Its universality among individuals with autism;
2. Its uniqueness to individuals with autism;
3. Its causal precedence, or ability to account for the earliest symptoms of autism; and
4. Its explanatory value, or ability to explain the full range of autistic symptomatology.
Non-genetic factors
Genetic liability
Brain abnormalities
Brain abnormalities
Cognitive deficit
Behavioural
symptom Behavioural
symptom Behavioural
symptom
Figure 1. A single primary cognitive deficit model of autism.
3 These four criteria will be used to evaluate primacy throughout this thesis, although the list is not claimed to be comprehensive or definitive. Other features frequently cited as signifying a primary deficit include persistence or stability throughout development (e.g., Ozonoff & McEvoy, 1994; Pennington & Welsh, 1995; Rutter, 1983) and existence in the broad phenotype of autism (Bailey, Phillips, & Rutter, 1996; Hughes, 2001; see Chapter 5).
5
It could be argued that these criteria for primacy are too stringent, due to the phenotypic
variability which exists in any syndrome (Tager-Flusberg, 1999a) and the possibility of
subgroups within the autism spectrum. However, any single primary cognitive deficit
model of autism should theoretically be able to meet the criterion of universality and be
able to account for the range of symptoms displayed by individuals with autism
(multiple deficits models are discussed further below).
Over the years, cognitive theories of autism have adopted many different forms.
The pioneering work of Hermelin and O’Connor (1970) demonstrated that neither
general mental retardation or peripheral (i.e., sensory or motor) processing could
explain the specific pattern of impairments displayed by individuals with autism, instead
finding evidence of abnormal “central” processes such as sequencing, concept
formation, and abstraction. Around the same time, Rutter (1968) proposed that
language or “coding” deficits were primary to autism. Subsequent hypotheses regarding
the nature of the primary impairment in autism have included aberrant sensory
processing (Ornitz, 1969, 1988), deficits in arousal modulation and attention (Dawson,
1991; Dawson & Lewy, 1989; Hutt & Hutt, 1968), impaired complex information
processing (Minshew, Goldstein, Muenz, & Payton, 1992; Minshew, Johnson, & Luna,
2001), lack of socio-affective or interpersonal relatedness (Hobson, 1989, 1993), and
abnormal social responsiveness or orienting to social information (Klin & Volkmar,
1993; Mundy & Neal, 2001; Mundy & Sigman, 1989). However, difficulties meeting
the various criteria for primacy (particularly the criterion of explanatory value for the
full range of symptoms) have meant that none of these theories has established itself as
a widely accepted candidate for a single primary deficit. Current research is dominated
by three main theories of the primary cognitive deficit in autism: i) lack of theory of
mind (inability to attribute mental states to oneself and others), ii) executive dysfunction
(impairment in high-level cognitive functions which guide and control behaviour toward
attainment of a goal), and iii) weak central coherence (tendency for piecemeal or local
information processing). Significant impairments in these areas have been established
in numerous studies of individuals with ASDs4. Proponents of these theories, in
particular the former two, have strongly asserted that the impairment in question is the
single primary cognitive deficit in autism. Additional impairments in other domains are
usually accounted for as secondary, correlated, or artefactual consequences of the single
primary deficit.
4 Studies of theory of mind and executive function in ASDs are reviewed extensively in Chapter 2. Central coherence studies are briefly discussed in Section 4.4.4 of Chapter 4.
6
The idea of a single primary deficit has been subjected to increasing criticism,
however. Goodman (1989) is often cited as an advocate of the multiple primary deficits
approach, arguing that genetic and environmental insults may act upon several distinct
neural systems which share in common a vulnerability to those insults. These multiple
neurological abnormalities then create simultaneous impairments in several cognitive
domains, and “synergistic interactions” between these impairments result in a distinct
syndrome. In this model, the shared vulnerability of several neural systems (e.g.,
through shared blood supply or neurotransmitters) is the unifying factor in creating a
unitary syndrome. In a similar vein, Pennington et al. (1997) proposed that the unifying
explanation may occur at the level of neurochemistry (e.g., a dopaminergic deficit),
which would result in multiple cognitive impairments that were not necessarily
connected at a cognitive level.
Others in favour of multiple primary cognitive deficits have argued against the
notion that autism is a unitary syndrome which requires a single unifying level of
explanation. As mentioned earlier, at least two alternative conceptions are possible,
which also incorporate ASDs besides autism. One is the notion of related but distinct
subgroups, or a “categorical” system of subtyping. Categorical systems are “intended to
divide populations into subgroups that share a common aetiology, symptom
presentation, and course that is distinct from those of other subgroups” (Beglinger &
Smith, 2001, p. 412). Subgroup divisions in ASDs could be defined in a number of
different ways, such as according to PDD subtype (i.e., autism, Asperger syndrome,
PDDNOS), the domains in which symptoms are present, symptom severity, or level of
(intellectual/adaptive) functioning, with the latter variable appearing to hold the best
discriminative and predictive validity in studies employing cluster analysis (Fein et al.,
1999; Prior et al., 1998; Stevens et al., 2000). If ASDs are conceptualised as a group of
distinct subtypes, then a single primary deficit model would not be plausible, and
instead there would need to be as many primary deficits as there were subgroups (unless
more extreme or severe subgroups were characterised by a larger number of primary
deficits and other milder subgroups were characterised by fewer primary deficits).
Therefore, across ASDs as a whole, primary deficits would not meet the criteria of
universality or explanatory value (although, they should meet these criteria within the
relevant subgroup).
The other major alternative model of ASDs is that of a multidimensional
spectrum, where dimensions such as symptom severity or level of functioning are
conceptualised as a continuum ranging from “normal” to severe or extreme, rather than
7
forming discrete subgroups. The idea of autism as a unitary syndrome is also
compatible with the notion of a spectrum, but in that case it would be unidimensional in
nature. In a multiple primary deficits model, there would be more than one cognitive
deficit, each underlying a different dimension. Again, the various dimensions could be
defined in different ways; for example, each symptom domain could be a dimension, or
there could be one dimension for symptom number and severity, and another for level
of functioning (Szatmari et al., 2002). In the version where the dimensions are
symptom domains, there would need to be as many primary deficits as there were
symptom domains5 (thus, a minimum of three independent cognitive deficits of varying
severity would need to underlie the triad of impairments in autism, whereas individuals
with PDDs showing symptoms in only two domains would show two primary deficits).
Therefore, the criteria of universality and explanatory value across all individuals with
ASDs would not be met by primary deficits in this model either (although these criteria
should be met by anyone displaying the relevant symptom, with differing degrees of
impairment according to the severity of the symptomatology). Figure 2 presents an
example of a multiple primary cognitive deficit model of autism based on the concept of
autism as a continuum with three dimensions, with each dimension corresponding to a
symptom domain.
Genetic origins Non-genetic factors
Behavioural symptom
Cognitive deficit
Behavioural symptom
Cognitive deficit Cognitive deficit
Brain abnormalities
Brain abnormalities
Behavioural symptom
Figure 2. A multiple cognitive deficits model of autism, in which each cognitive deficit
underlies a different domain of symptomatology.
5 This assumes that the symptom domains are dissociable, such that each symptom could potentially be displayed in isolation.
8
While these multiple cognitive deficits models of ASDs represent plausible alternatives
to the notion of autism as a unitary syndrome with a single primary cognitive deficit,
strong claims about singular primacy are still being made by proponents of the major
current cognitive hypotheses. The validity of these claims not only rests on how well
the proposed primary deficit can meet the four criteria for primacy, but also on whether
the deficit can explain or subsume the other cognitive impairments which characterise
ASDs. The construction of an integrated explanatory model of ASDs requires
identification of which cognitive processes are the most primary in ASDs and how they
relate both to each other and to the genetic, neurobiological, and behavioural levels of
explanation6.
1.3 Overview of the thesis
1.3.1 Rationale and aims
The overarching aim of the current research is to contribute to an explanatory model of
ASDs, primarily by investigating the structure of the cognitive level of explanation, but
also by examining its relationships with other levels of explanation (mainly the
behavioural, but also the genetic in an indirect sense) – and thereby to evaluate the
validity of a single versus multiple primary cognitive deficit model of ASDs. More
specifically, this thesis focusses on two of the major current cognitive theories of
primary deficits in ASDs: lack of theory of mind and executive dysfunction. These two
theories represent the most fertile ground for debate regarding the primacy of and
relationship between cognitive deficits in ASDs. This is firstly because proponents of
these theories have made the strongest claims, as well as presenting the most convincing
yet controversial evidence, about the deficit in question being the single primary deficit
in autism (whereas those arguing for weak central coherence have tended to more often
present it as one of multiple deficits); and secondly because the relationship between
theory of mind (ToM) and executive function (EF) has been the subject of considerable
theoretical and empirical scrutiny in typical development, but has been less well studied
6 Of course, this assumes that the cognitive level of analysis is necessary and/or useful in explaining autism. The importance of cognition in constructing causal models for developmental disorders has been justified persuasively by Morton and Frith (2001) and Tager-Flusberg (1999a), who argue that cognition is necessary to bridge the gap between brain and behaviour in a parsimonious and theory-driven manner. Postulating areas of strength and weakness at the mediating level of cognition allows us to form sensible, coherent interpretations of apparently unrelated behavioural and biological observations.
9
in ASDs (although several claims and assumptions have been made about their
relatedness in ASDs). This lack of empirical attention is somewhat surprising, as any
proponent of a single primary deficit model must show that the primary deficit (e.g., in
ToM) causes any other deficit (e.g., in EF) demonstrated by individuals with ASDs.
Moreover, most multiple deficits models would need to show that ToM and EF were
independent impairments (either characterising different subgroups or underlying
different dimensions of ASDs).
The current research consists of two studies, both broad in scope. The first study
examined the profile, primacy, and independence of ToM and EF impairments in
individuals with ASDs. This is only the second study to examine these issues together
in one large investigation, with the first (Ozonoff, Pennington, & Rogers, 1991)
containing several limitations which were addressed in this study (see Chapter 4). The
three central aims of Study One were to determine i) the specific profile of ToM and EF
deficits which characterises ASDs (as a necessary first step before further examining
primacy and independence); ii) whether impairments in ToM and/or EF can meet the
criteria for a primary cognitive deficit in ASDs (as assessed by its universality,
uniqueness, and explanatory value), and, should no impairment meet the criteria fully,
which appears to be the most primary; and iii) whether or not ToM and EF impairments
are related in ASDs, and if so, what the nature of that relationship might be. Several
competing hypotheses about the relative primacy of and relationship between ToM and
EF were tested, with each having different implications for which type of single or
multiple deficit model could best explain ASDs. These aims and hypotheses and the
way in which they were addressed are elaborated in Chapter 4.
The second study attempted to confirm and extend the results of Study One by
investigating ToM and EF impairments in siblings of individuals with ASDs. As ASDs
are genetic disorders (see Chapter 5), examining cognitive weaknesses in relatives of
individuals with ASDs can be a useful method of identifying potential markers of
genetic vulnerability as well as testing models of primary deficits in ASDs. The main
aims of Study Two were i) to identify whether ToM or EF performance can meet
criteria for an “endophenotype” or vulnerability marker for the autism genotype, and
thereby seek confirmation of the results of Study One regarding the relative primacy of
ToM and EF in ASDs; and ii) to further investigate the validity of various
single/multiple deficits models of ASDs by examining the pattern of ToM and EF
performance in individuals showing the broad phenotype. Again, the aims and
10
hypotheses of this second study and its extensions to previous research are further
discussed in Chapter 6.
1.3.2 Thesis structure
In Chapter 2, the constructs of ToM and EF are reviewed with reference to both typical
development and autism. Each ability is defined; its methods of measurement are
discussed; relevant models of its structure and typical development are presented; and
evidence for its impairment in and primacy to autism is critically reviewed. Next, the
various hypotheses about the nature of the relationship between ToM and EF in typical
development are considered, and these hypotheses are then re-examined with regard to
the relationship in autism. This critical analysis of previous research on the nature,
primacy, and independence of ToM and EF in typical development and autism provides
the context for the thesis and for Study One in particular.
A large range of diagnostic, IQ, cognitive, and behavioural measures were used
in both of the studies in the thesis. Chapter 3 is devoted to the description and rationale
for selection of these measures. For each questionnaire, interview, and task, the basis
for its inclusion in the research and a thorough description are both provided. This
reflects a general emphasis on the use of appropriate assessment tools, particularly in
the area of EF, which has suffered from a history of poor measurement precision.
Chapter 4 contains the major study of the thesis. The main aims of Study One
were outlined in the previous section. The broader phenotype of autism is reviewed in
Chapter 5, as a background for second study of the thesis. This briefer review covers
the genetic basis for autism and the behavioural and cognitive characteristics of first-
degree relatives of individuals with autism. Chapter 6 contains Study Two, the central
aims of which were also described in the previous section.
In the General Discussion in Chapter 7, the results of both studies are
summarised and their implications for conceptual models of ASDs are discussed. The
importance of integration between the various levels of explanation is highlighted and
emphasis is placed on the need to consider the process of development in constructing
explanatory models of developmental disorders.
11
12
CHAPTER 2
Literature Review: Theory of Mind and Executive Function in Typical Development and in
Autism
2.1 Theory of mind (ToM)
2.1.1 Defining and measuring ToM
2.1.2 Models of ToM and its development
2.1.3 ToM in autism
2.2 Executive function (EF)
2.2.1 Defining and measuring EF
2.2.2 Models of EF and its development
2.2.3 EF in autism
2.3 The ToM-EF relationship
2.3.1 Models of the ToM-EF relationship
2.3.1.1 Expression accounts
2.3.1.2 Common conceptual requirements of ToM and EF
2.3.1.3 Emergence accounts
2.3.1.4 Common neuroanatomical bases for ToM and EF
2.3.2 The ToM-EF relationship in autism
13
This chapter reviews previous research on the constructs of ToM and EF both in typical
development and in autism, providing a context for three of the central concerns of the
current research – the profile, primacy, and independence of ToM and EF impairments
in ASDs. ToM is discussed in the first section, followed by EF in the second section.
Each of these sections contains i) a solid background on how the construct is defined
and measured, reflecting a general emphasis on measurement precision, particularly in
the area of EF; ii) a review of relevant models of the typical development of ToM/EF,
in order to provide a theoretical context within which both evidence of the impairment
of ToM/EF in ASDs and models of the ToM-EF relationship may be evaluated; and iii)
a review of evidence for the impairment of ToM/EF in autism and the specific profile of
that impairment, followed by a critical analysis of evidence for the primacy of the
ToM/EF impairment to autism. The third section of the review addresses the
relationship between ToM and EF, covering both i) theories of the nature of the
relationship in typical development, which are outlined in detail as each makes different
predictions about the ToM-EF relationship in autism; and ii) evidence for the nature of
the relationship in autism, which not only has implications for the validity of theories of
the relationship based on typical development, but more importantly is relevant for the
question of primacy (i.e., can a primary deficit in ToM explain or subsume a secondary
deficit in EF, or vice versa?) This review of methodology, theory, and evidence in the
fields of ToM and EF is therefore intended as a backdrop against which findings from
the current research may be appraised and interpreted.
2.1 Theory of mind (ToM)
2.1.1 Defining and measuring ToM
The term “theory of mind” refers to the ability to attribute oneself and others with
mental states, such as desires, beliefs, and intentions, in order to explain and predict
actions. The phrase was first used by Premack and Woodruff (1978), who stated that:
In saying that an individual has a theory of mind, we mean that the individual imputes mental
states to himself and to others...A system of inferences of this kind is properly viewed as a
theory, first, because such states are not directly observable, and second, because the system can
be used to make predictions, specifically about the behavior of other organisms (p. 515).
14
After noting flaws in the methodology used by Premack and Woodruff (1978) to
examine whether or not chimpanzees have a ToM, Dennett (1978) pointed out that ToM
could be demonstrated conclusively only by predicting the way another person will
behave on the basis of a false belief (otherwise the actual situation, habitual or regular
aspects of the other person’s behaviour, or the subject’s own true beliefs could be used
to predict the person’s actions, without the need to appeal to mental states). This
proposal was first employed with humans by Wimmer and Perner (1983), who tested
typically developing children on what has now become a classic false belief task,
sometimes called the “unexpected transfer” test. A scenario is presented in which a
story character, Maxi, puts a chocolate in cupboard A before he goes out to play. While
he is gone, his mother moves the chocolate to cupboard B. Maxi then returns, and
participants are asked, “Where will Maxi look for the chocolate?”. Two control
questions testing knowledge of the chocolate’s original and current location ensure that
the child recalls the story and followed the sequence of events. In order to answer the
belief question correctly (that Maxi will look in cupboard A), the child requires an
understanding that Maxi holds a false belief, which is different from the child’s own
knowledge of the actual situation, and which will lead Maxi to behave in a way which
contradicts the actual situation (i.e., Maxi’s behaviour is a product of what he believes
to be true rather than what is really true). An accurate answer on the belief question
therefore suggests that the participant appreciates the distinction between mind (the
internal and mental) and world (events, situations or behaviours).
In another commonly used false belief task, variously called the “Smarties task”,
the “unexpected contents” task, or the “deceptive box test” (Perner, Leekam, &
Wimmer, 1987), the child is shown a box of Smarties and asked what s/he thinks is
inside. After responding “Smarties”, the child is shown that the box actually contains a
pencil. The pencil is then put back in the box and the box is closed. The child is asked
a control question about the actual content of the box (i.e., a pencil). The child’s ability
to attribute false beliefs to others is then assessed by asking what another child (or
family member) would think was in the box. In some versions, the child is also asked
about his/her own previous belief about the content of the box, when s/he was first
shown it. Gopnik and Astington (1988) found that children who fail the false belief
question also incorrectly answer that they thought the box contained pencils when they
themselves first saw it, consistent with the view that the development of ToM pertains
to knowledge of one’s own mind as well as the minds of others.
15
A variation on the Smarties task is a test of the “appearance-reality distinction”,
or the “unexpected identity” task (Flavell, Flavell, & Green, 1983). In this task, the
child is shown an object with a deceptive identity, such as a sponge that looks like a
rock, and is asked what it looks like. The real nature of the object is then demonstrated
to the child (e.g., by squeezing the sponge), and the child is asked what it really is. The
subsequent two questions follow the same structure as the Smarties task, with the child
being asked what s/he thought the object was when first shown it, and what another
child would think the object was.
Such false belief tasks are now central to current developmental research on
social cognition, serving as a marker for ToM in both typically developing and
disordered populations (Wellman, Cross & Watson, 2001). However, ToM has been
measured in many other ways, some of which also exploit the false belief concept, and
others of which measure other types of mentalistic understanding. Baron-Cohen (2000)
reviews 20 kinds of tasks which are purported to measure ToM, including tests of
deception, the mental-physical distinction, recognition and expression of mental-state
words, decoding mental states from the eyes, and understanding the mental functions of
the brain.
2.1.2 Models of ToM and its development
The timing and mechanisms of the normal development of ToM have been studied
extensively (see Astington, Harris & Olson, 1988; Carruthers & Smith, 1996; Lewis &
Mitchell, 1994; Mitchell & Riggs, 2000; Perner, 1991; Wellman, 1990; Wellman et al.,
2001; Whiten, 1991). The large majority of studies demonstrate a definitive
improvement in performance on false belief (and other ToM) tasks between the ages of
3 and 5 years, with 3-year-olds consistently making errors suggesting that they are
unable to separate belief from reality (e.g., in the unexpected transfer task they will
assert that Maxi will look for the chocolate in cupboard B, its actual location). In their
meta-analysis of 178 studies measuring young children’s performance on false belief
tasks, Wellman et al. (2001) found that average false belief performance changes
rapidly between 3 and 4.5 years from significantly incorrect (i.e., below chance) to
significantly correct (above chance). Although ToM development through middle
childhood to adolescence is much less well studied, evidence suggests that advances in
these years include “an understanding that people’s mental states...are often consistent
across situations in the form of personality traits, a greater appreciation of the mind as
16
an active constructor and interpreter of knowledge, and a growing awareness of the
presence, influence, and sources of ongoing thoughts – that is, active mental ideation”
(Wellman & Lagatutta, 2000, p. 31).
Theoretical accounts of ToM development tend to focus on the crucial period
between 3 and 5 years, centreing on the debate as to whether young children fail false
belief tasks because they lack the conceptual understanding required to respond
correctly (“competence”), or because the insufficient development of other cognitive
capacities (e.g., inhibitory control, linguistic comprehension) masks the access to or
expression of understanding (“performance”). Explication of these models is useful
because of their in-depth analysis of what is involved in successful performance on false
belief tasks. This is important both for understanding what underlies the ToM
impairment in autism, and for analysing the relationship between ToM and EF. An
overview of the major relevant accounts of ToM development therefore follows.
Competence accounts vary in their postulated mechanisms of ToM development,
but they share in common the idea that ToM matures continuously in a series of
successive stages of discovery, each of which is developmentally related to the next (as
opposed to an innate, modular ability which comes on-line early). Gopnik, Wellman
and colleagues (Gopnik, 1993; Gopnik & Meltzoff, 1997; Gopnik & Wellman, 1994;
Wellman & Gelman, 1998) favour the “theory theory”, which proposes that children’s
early conceptions of the mind are theory-like, sharing key features with scientific
theories: they are abstract (i.e., framed in a different vocabulary from empirical
observations), hold explanatory and predictive power, lead to distinctive interpretations
of evidence, and are open to revision based on counterevidence. The theory theory
holds that ToM development is a gradual transition from one view of the mind to
another, rather than being a simple all-or-none acquisition of “a” theory of mind.
In line with the theory theory, Wellman and colleagues (Bartsch & Wellman,
1989, 1995; Gopnik & Wellman, 1994; Wellman, 1990; Wellman & Woolley, 1990)
propose a specific developmental sequence in which children’s understanding of the
motivational forces behind people’s actions advances from a “simple desire”
psychology to a more adult “belief-desire” psychology. In Wellman’s framework, 2-
year-olds hold a simplified understanding of desire and perception (where others are
attributed internal dispositions toward or against certain actions or objects), but fail to
17
understand that people have internal mental representations1 of the world such as
beliefs; in an intermediate phase, 3-year-olds develop a nonrepresentational
understanding of belief while beginning to comprehend representational aspects of
desire and perception; and at around age four, children begin to realise that individuals’
beliefs (i.e., their representations of reality rather than reality itself) determine their
actions. According to the theory theorists, 2- and 3-year-olds fail the standard false
belief (unexpected transfer) task because as simple desire psychologists, they do not
attribute a belief to Maxi, but rather they predict that Maxi will act to fulfil his desire for
chocolate and will therefore look where the chocolate actually is.
Perner (1991, 1993, 1995, 2000) has articulated an alternative competence
account of ToM development which, like Wellman, Gopnik and colleagues, focuses on
children’s understanding of mental states as representations. While Perner considers
himself a theory theorist, he states that the use of the word “theory” is meant to signal
the notion that conceptual understanding unfolds as a result of the growth of
interdependent concepts, rather than suggesting that children’s intellectual growth is
analogous to scientists making new discoveries (which is explicitly proposed by Gopnik
& Wellman, 1994). He argues that young children begin with a nonrepresentational
conception of mind, and that it is only when children acquire a general theory of
representations (a Representational Theory of Mind; RTM) that they are able to solve
false belief tasks. In Perner’s model, an RTM involves comprehending that
propositions (e.g., beliefs) are semantically evaluable as being true or false. That is,
propositions are “about” a world against which their truth is evaluated (Perner, 2000).
In claiming that young children do not understand representations, his contention is that
they do not understand that a proposition can be evaluated by someone else as having a
different truth value than the one it has in reality (or the one assigned to it by the child
his/herself). Thus, Perner’s (1995) explanation of young children’s failure on the
unexpected transfer task is that:
...they cannot distinguish between the state of the world that the belief is about and how the
believer conceives of that state of the world, or in other words, children cannot conceive of
belief as misrepresenting where the chocolate really is as being in location A. Without this
understanding children cannot understand why an agent who wants to find the chocolate in its
real world location (B) would act as if the chocolate were in A. (p. 251).
1 Here, a “representation” may be defined as an entity in the mind which represents a state of affairs in the world, like a “picture-in-the-head” (Leslie & Thaiss, 1992). This is distinct from Leslie’s (1987, 1994a) concept of “metarepresentation”, which is discussed later.
18
However, understanding propositions as evaluable as true or false is not enough to
successfully solve false belief tasks (Perner, 2000). The child must additionally realise
that the belief has causal power – it takes precedence (over the world itself) in
determining behaviour. Thus, a false belief about the chocolate’s location, rather than
the chocolate’s actual location, makes Maxi look in the wrong place.
Another distinction between the positions outlined by Perner and Wellman is the
specific nature of the successive theories of mind that children are said to discover, with
Perner rejecting Wellman’s simple desire psychology in favour of his concept of
“prelief”. He argues that young children’s appreciation of pretence (i.e., their ability for
pretend play, which is first employed between 18 and 24 months) implies a realisation
that people do not always act in a way that satisfies their desires objectively. However,
since the young child cannot differentiate between actions based on a false belief which
is held as true and actions based on pretence where what is being pretended is not held
as true, s/he understands these states of belief and pretence as the amalgamated
protoconcept of prelief, which s/he conceptualises as “behaving as if” (Perner, Baker &
Hutton, 1994). Children eventually develop a more adult stage of understanding of
“behaving as is”, whereby people behave according to the beliefs they hold as being
true (this stage in Perner’s model is not distinguishable in obvious ways from the 4-
year-old stage of understanding outlined by Wellman and colleagues).
In contrast to competence accounts such as those of Wellman and Perner,
performance accounts hold that young children fail false belief tasks because processing
limitations2 mask their true ability, as evidenced by demonstrations of earlier
competence when the testing procedure is modified – such as by asking the child
“Where will Maxi look first for his chocolate?” in the unexpected transfer task (e.g.,
Chandler, Fritz, & Hala, 1989; Freeman & Lacohée, 1995; Mitchell & Lacohée, 1991;
Roth & Leslie, 1998; Siegal & Beattie, 1991). This position has been articulated most
thoroughly by Leslie and colleagues, who argue that ToM arises from an attentional
mechanism specialised for selectively attending to mental states (the Theory of Mind
Mechanism; ToMM) which is innate, domain-specific, operates spontaneously from
very early in life without formal instruction, and can be dissociably damaged – in other
words, it is modular (German & Leslie, 2000; Leslie, 1987, 1991, 1994a, 1994b; Leslie
& Roth, 1993; Leslie & Thaiss, 1992; Roth & Leslie, 1998; Scholl & Leslie, 1999,
2 These include executive functions such as inhibition and working memory. Accounts of ToM development based around advances in EF are reviewed in Section 2.3.1.
19
2001; Surian & Leslie, 1999). For Leslie, very young children’s apparent appreciation
for something as abstract and unobservable as others’ mental states is best explained by
an innately specified module.
As for Perner, the very early appearance of pretend play is an important factor in
Leslie’s model, however for Leslie it marks an early capacity for metarepresentation,
which also underlies the concept of belief and indicates the early presence of a ToMM.
Leslie (1987, 1994a), based on Pylyshyn (1978), distinguishes between primary
representations, which are direct, literal representations about a state of affairs in the
world; and metarepresentations3, which may be described as representations of
representations. A metarepresentation describes an agent’s (e.g., mother’s, self’s)
mental state, or provides an “agent-centred” description of a situation, which is
“decoupled” from the primary representation and processed as if it were a copy or report
of the primary representation (Leslie & Roth, 1993). It does this by specifying an
“informational relation” (or “propositional attitude”; e.g., DESIRING, PRETENDING)
between the agent, an aspect of reality (described by a primary representation) and an
imaginary situation (described by the “decoupled” representation). For example, the
metarepresentation mother PRETENDS [of] this banana [that] “it is a telephone”
allows the child to make sense of his/her mother’s behaviour of talking to a banana by
reference to his/her mother’s mental state (i.e., her attitude of pretence towards the
banana), without making the real-world inference that “bananas are telephones”.
Leslie’s assumption is that there is a small core set of innate informational
relations available to the ToMM early on, such as BELIEVING, DESIRING, and
PRETENDING. As these attitudes are all deployed within the same
metarepresentational structure, Leslie’s explanation for why young children are able to
demonstrate understanding of pretence and desire, but not belief, rests on an additional
component of his model termed the “Selection Processor” (SP; Leslie & Thaiss, 1992).
The SP is an inhibitory mechanism which allows the child to select the specific relevant
information that is required for the belief content inference, while disregarding
prepotent competing information (e.g., in the unexpected transfer task, inferring the
correct content of Maxi’s belief requires selecting the situation which Maxi was
3 Perner (1991) has criticised Leslie’s use of the term metarepresentation as suggesting that the young child has a conscious “theory of representation”. However, Leslie has specified (Leslie & Thaiss, 1992; Leslie & Roth, 1993) that he does not intend it in this sense, but rather intends it to denote a kind of data structure computed by our cognitive system, or an information processing mechanism which helps to create conceptual knowledge. He does not mean to imply that children have a conscious theory that mental states are representations in the head (as Perner does to some degree when he proposes the Representational Theory of Mind).
20
exposed to at the beginning of the scenario from memory and resisting basing the
inference on the current situation - a tendency which is prepotent because beliefs are
usually true representations of current reality; Leslie, 1994a). According to Leslie, 3-
and 4-year-old children do not differ fundamentally in their conception of belief, but 3-
year-olds fail the false belief task because the SP is poorly developed. Hence, task
manipulations which decrease the load on the SP often result in an improvement in
young children’s performance on false belief tasks.
A volley of criticisms directed at both theoretical foundations and
methodological approaches continues to shoot back and forth between competence and
performance theorists4 (see, for example, German & Leslie, 2000; Perner, 2000; Roth &
Leslie, 1998; Wellman et al., 2001). Theory theorists have accused modularity accounts
of being “antidevelopmental” (Gopnik & Wellman, 1994; for a defence see Scholl &
Leslie, 1999), while modularity theorists argue that theory theories are purely
descriptive (lacking a specification of the cognitive architecture and mechanisms
underlying theory development), and require that the young child develop explicit
theories about impossibly abstract concepts (Roth & Leslie, 1998; but see Perner, 2000).
Competence theorists claim that studies purporting to show improved performance of 3-
year-olds on simplified false belief tasks have not been consistently replicated and are
open to alternative interpretations (Perner, 2000), while performance theorists maintain
that 3-year-olds failure on standard false belief tasks is a false negative (Leslie, 1994a).
While it is not possible to do full justice to these arguments here, placing ToM in at
least a broad theoretical context is helpful both in evaluating models of the ToM-EF
relationship and in conceptualising the impairment of ToM in autism – the latter of
which we turn to now.
2.1.3 ToM in autism
Complementing the vast literature addressing the typical development of ToM, an
equally large, if not larger, number of studies have investigated ToM in children with
autism. This body of work began with a seminal paper by Baron-Cohen, Leslie and
Frith (1985), in which children with autism were tested on a variation of Wimmer and
Perner’s (1983) unexpected transfer task. Baron-Cohen et al.’s “Sally-Anne” scenario,
which has become the most frequently used version of the unexpected transfer test in
4 For other major accounts of ToM development which have not been reviewed here (e.g., simulation theory, counterfactuality), the reader is referred to Mitchell and Riggs (2000).
21
the autism literature, involves two doll protagonists, Sally and Anne. Sally places a
marble into her basket, then leaves the scene. Anne takes the marble and hides it in her
box. When Sally returns, the child is asked “Where will Sally look for her marble?”
(the Belief Question). Two control questions probe knowledge of the current location
of the marble (the Reality Question) and the marble’s initial location (the Memory
Question). Baron-Cohen et al. found that while all autistic and control children were
able to correctly answer the control questions, only 20% of children with autism passed
the Belief Question, compared with 85% of typically developing children and 86% of
children with Down’s Syndrome (suggesting that the poor performance in children with
autism was not attributable to intellectual disability). They interpreted this result as
evidence for a metarepresentational deficit specific to autism (based on Leslie’s (1987)
theory of ToM), which had the potential to explain autistic symptoms such as social
impairment and lack of pretend play.
Impaired performance of individuals with autism on false belief tasks has since
been replicated in numerous studies (although failures to replicate have also occurred,
as discussed later). These studies have included a range of task variations such as using
real people instead of puppets and a “think” question rather than a “look” question (i.e.,
“Where does Sally think the marble is?”), as well as using alternative false belief
paradigms such as the deceptive box (“Smarties”) test and the unexpected identity
(appearance-reality) task (Baron-Cohen, 1989a; Charman & Baron-Cohen, 1992;
Eisenmajer & Prior, 1991; Leekam & Perner, 1991; Leslie & Frith, 1988; Leslie &
Thaiss, 1992; Perner, Frith, Leslie & Leekam, 1989; Ozonoff et al., 1991; Reed &
Peterson, 1990; Surian & Leslie, 1999). Individuals with autism have also
demonstrated significantly poorer performance than controls on various other tasks
tapping mentalising ability5, such as sequencing of mentalistic picture stories (Baron-
Cohen, Leslie & Frith, 1986); tests of the mental-physical distinction (Baron-Cohen,
1989a; Ozonoff et al., 1991); describing the mental functions of the brain (Baron-
Cohen, 1989a; Ozonoff et al., 1991); recognition, comprehension and expression of
mental state terms (Baron-Cohen et al., 1994; Tager-Flusberg, 1992; Ziatas, Durkin &
Pratt, 1998); inferring the mentalistic significance of the eyes (Baron-Cohen, Campbell,
Karmiloff-Smith, Grant, & Walker, 1995; Baron-Cohen et al., 1999a; Baron-Cohen,
Jolliffe, Mortimore, & Robertson, 1997; Baron-Cohen, Wheelwright, Hill, Raste, &
Plumb, 2001a); attribution of mental states to animated shapes (Castelli, Frith, Happé &
5 The term “mentalising ability” is intended as a synonym for ToM (i.e., the ability to make inferences about mental states).
22
Frith, 2002); conceptual perspective-taking (Dawson & Fernald, 1987); tests of
deception (Baron-Cohen, 1992; Russell, Mauthner, Sharpe, & Tidswell, 1991; Sodian &
Frith, 1992); understanding that “seeing-leads-to-knowing” (Baron-Cohen & Goodhart,
1994; Leslie & Frith, 1988); understanding that beliefs cause emotions (Baron-Cohen,
1991a); and understanding of intentions (Phillips, Baron-Cohen & Rutter, 1998).
On the basis of this kind of evidence, some researchers have proposed that the
whole range of autistic symptomatology may be explained by a single, primary,
cognitive deficit in ToM (e.g., Baron-Cohen, 1988, 1991c; Frith et al., 1991; Leslie,
1987, 1991). Furthermore, the same authors have argued that the apparent domain
specificity of the ToM impairment in autism is existence proof that ToM is a modular
capacity. For example, Leslie (1987, 1991; Leslie & Thaiss, 1992) has argued that the
modular Theory of Mind Mechanism (ToMM), which automatically leads us to interpret
behaviour in terms of an agent’s mental states, is specifically impaired in autistic
individuals and can explain their social and communicative impairments and lack of
pretend play. Baron-Cohen (1994, 1995, 1998) outlined an alternative view whereby
ToMM does not come fully prepackaged as an innate module, but rather is preceded by
several lower-level modular mechanisms which extract relevant social information and
provide critical inputs to the development of ToM. These mechanisms include an Eye
Direction Detector (EDD) which alerts the infant to the eye region and thereby provides
opportunities to learn the mentalistic significance of eye gaze; an Intentionality Detector
(ID) that directs attention to animate actions, enabling the infant to learn about goal-
directedness; and a Shared Attention Mechanism (SAM), which uses inputs from the
other two mechanisms to allow the infant to work out if s/he and another person are
jointly attending to the same thing. In this model, ToMM is conceptualised as being
either a more mature development of SAM, or is triggered by SAM.
Tests of the validity of the ToM hypothesis of autism (the view that autism may
be explained by a primary deficit in a ToM module) have focused on the central criteria
required to uphold the position (outlined in Chapter 1, Section 1.2): that i) a ToM
impairment is universal among individuals with autism; ii) a ToM impairment is unique
to individuals with autism, iii) a ToM impairment can explain the earliest signs of
autism in infants (causal precedence); and iv) a ToM impairment can account for the
entire range of symptoms displayed by individuals with autism (explanatory value). In
addition, the modular ToM hypothesis must meet criterion v), that failure on ToM tasks
is best explained by a domain-specific ToM impairment, and cannot be accounted for in
23
terms of other cognitive constructs. The evidence for each of these claims is reviewed
below.
i) Universality. From the first study of ToM in autism (Baron-Cohen et al.,
1985), it was evident that a proportion of autistic individuals were able to pass false
belief tasks. The percentage of autistic individuals found to pass standard false belief
(unexpected transfer or “Sally-Anne”) tasks in subsequent studies has varied from 15%
(Reed & Peterson, 1990) to 55% (Prior, Dahlstrom, & Squires, 1990), with 90% of
participants with autism passing in one study (Dahlgren & Trillingsgaard, 1996).
Although in most cases the proportion of passers with autism is significantly smaller
than the proportion of successful control participants (usually matched on verbal mental
age), the finding that any child with autism passes false belief tasks poses a challenge to
the ToM hypothesis of autism (although random responding could result in a correct
response). Baron-Cohen (1989b) responded to this challenge with a study
demonstrating that individuals with autism who pass standard first-order false belief
tasks are still unable to make more complex second-order false belief attributions (i.e.,
of the form “Mary thinks that John thinks the icecream van is in the park”; Perner &
Wimmer, 1985). He proposed that autism is characterised by a specific developmental
delay in ToM, such that older and more able participants with autism are able to pass
first-order false belief tasks which are usually mastered by the age of four, but still fail
on more difficult tasks which are usually only passed by the age of six or seven.
However, Baron-Cohen’s (1989b) finding has not been replicated in a number of
subsequent studies, which have found that a subset of participants with high-functioning
autism or with Asperger syndrome also pass second-order false belief tasks (Bauminger
& Kasari, 1999; Bowler, 1992; Dahlgren & Trillingsgaard, 1996; Leekam & Prior,
1994; Ozonoff et al., 1991; Sparrevohn & Howie, 1995). Ozonoff et al. (1991) found
that EF deficits were more universal than ToM impairment among high-functioning
autistic individuals (see Section 2.2.3 for further discussion of this finding).
Furthermore, Tager-Flusberg and Sullivan (1994b) showed that both autistic and control
first-order task passers were able to pass a shorter and less complex second-order task
than the task used in previous studies, suggesting that their failure on traditional second-
order tasks was more likely to be due to the high information processing load than a
lack of conceptual understanding.
Nevertheless, it has been argued that ToM is not measurable only by
performance on false belief tasks (e.g., Tager-Flusberg, 2001). Studies using higher-
level tests of mentalising ability have found that first- and second-order false belief task
24
passers still demonstrate evidence of impairment in ToM. Happé (1994a) found that
individuals with autism who passed both first- and second-order false belief tasks were
significantly poorer than mentally handicapped and normal children and adults at
providing context-appropriate mental state explanations for nonliteral utterances made
by story characters, which she argues is a more “advanced”, naturalistic test of ToM.
First- and second-order passers performed more poorly than controls on a test requiring
inference of complex mental states from expression of the eyes (Baron-Cohen et al.,
1997), and high functioning adults with autism who passed a first-order false belief task
performed significantly worse than controls on tests measuring attribution of mental
states to voices and eyes (Kleinman, Marciano & Ault, 2001). Frith, Happé and
Siddons (1994) also found that first-order passers still showed impairments in everyday
social behaviours which require mentalising.
Studies examining the characteristics of autistic false belief passers have tended
to find that a high verbal mental age or verbal IQ is a necessary but not sufficient
condition for passing false belief tasks (Charman & Baron-Cohen, 1992; Eisenmajer &
Prior, 1991; Leekam & Perner, 1991; Prior et al., 1990; Sparrevohn & Howie, 1995). In
a review of the literature, Happé (1995) found that children with autism require a verbal
mental age more than twice as high as control participants in order to pass false belief
tasks. Other studies have found that chronological age is a significant factor, either in
addition to or instead of verbal mental age (Baron-Cohen, 1992; Prior et al., 1990),
while still others have found no relationship between age and ability variables and false
belief task performance (Baron-Cohen et al., 1985; Perner et al., 1989). The general
finding that passers tend to be of higher verbal ability is consistent with the idea, put
forward by a number of authors, that individuals with autism who pass false belief tasks
do so not by the usual use of ToM, but by using an alternative compensatory route to
success (Eisenmajer & Prior, 1991; Frith et al., 1991; Happé, 1995; Holroyd & Baron-
Cohen, 1993; Ozonoff et al., 1991). For example, Frith et al. (1991) suggested that able
autistic individuals may have learned or extracted explicit rules about certain social
situations, such as “When something in the world changes, people who just happen not
to have seen the change occur behave (for some reason) as if they do not know about
these changes” (p. 436). A study by Happé et al. (1996) provides some support for this
idea, finding that adults with Asperger syndrome showed activation of different areas of
the prefrontal cortex from controls when listening to mentalistic stories. However, more
direct evidence confirming that false belief task passers are using compensatory
strategies to deduce their solution is yet to be obtained.
25
ii) Uniqueness. Proponents of the ToM hypothesis argue that a ToM impairment
is unique to autism, citing evidence that control groups of children with either Down’s
syndrome (e.g., Baron-Cohen et al., 1985), other kinds of mental retardation (e.g.,
Charman & Baron-Cohen, 1992), or specific language impairment (Leslie & Frith,
1988) do not show impaired performance on false belief tasks in comparison with
children with autism. However, these findings have been challenged in a number of
other studies which have either failed to replicate significantly poorer performance of
children with autism on various ToM tasks compared with controls (Carpenter,
Pennington & Rogers, 2001; Charman & Lynggaard, 1998; Dahlgren & Trillingsgaard,
1996; Oswald & Ollendick, 1989; Prior et al.,1990, Tager-Flusberg & Sullivan, 1994a),
or have found ToM impairments in other clinical populations. It has become apparent
that mentally retarded, non-autistic individuals perform more poorly on false belief
tasks than would be expected given their chronological and mental age (Benson,
Abbeduto, Short, Bibler-Nuccio, & Maas, 1993; Yirmiya, Erel, Shaked, & Solomonica-
Levi, 1998; Yirmiya & Shulman, 1996; Yirmiya, Solomonica-Levi, Shulman, &
Pilowsky, 1996; Zelazo, Burack, Benedetto, & Frye, 1996a). Yirmiya et al.’s (1998)
meta-analysis comparing the ToM abilities of individuals with autism, mental
retardation (MR), and typically developing individuals showed that although autistic
individuals were the most severely impaired on ToM tasks, individuals with MR also
performed significantly more poorly than typically developing individuals. This result
led them to conclude that it may be the severity of ToM impairment rather than the
impairment itself that is unique to autism. They also found that the aetiology of the MR
was an important factor, with individuals with Down’s syndrome performing better than
other individuals with MR of unknown aetiologies.
In addition to MR, impairments in ToM have been found in deaf children (de
Villiers, 2000; Peterson, 2002; Peterson & Siegal, 1995), blind children (Brown,
Hobson, Lee, & Stevenson, 1997; Minter, Hobson, & Bishop, 1998), and individuals
with schizophrenia (Corcoran, Mercer, & Frith, 1995; Mazza, De Risio, Surian,
Roncone, & Casacchia, 2001; Pilowsky, Yirmiya, Arbelle, & Mozes, 2000), bipolar
affective disorder (Kerr, Dunbar, & Bentall, 2003), borderline personality disorder
(Fonagy et al., 1995), non-verbal learning disorder (Buitelaar, Swaab, van der Wees,
Wildschut, & van der Gaag, 1996), Parkinson’s disease (Mengelberg & Siegert, 2003;
Saltzman, Strauss, Hunter, & Archibald, 2000), and frontotemporal dementia (Gregory
et al., 2002; Lough & Hodges, 2002). Contrary to the findings of Leslie and Frith
(1988), other studies have found ToM deficits in children with specific language
26
impairment and other communicative disabilities (e.g., Dahlgren, Dahlgren Sandberg, &
Hjelmquist, 2003). These challenges to the uniqueness criterion of the ToM hypothesis
have been refuted by claims that these non-autistic clinical groups do not show as severe
an impairment on ToM tasks as individuals with autism, and that they fail ToM tasks
for different reasons than individuals with autism (i.e., their failure is not due to a
genuine metarepresentational deficit). For example, individuals with MR may fail
because of poor general cognitive and linguistic skills, deaf and blind children may fail
because they lack the necessary perceptual input, such as access to language or facial
information, and individuals with borderline personality disorder may fail because
parental neglect and abuse prevented the normal development of ToM (Baron-Cohen,
2000; Corcoran, 2000; Tager-Flusberg, 2001). However, these claims are yet to be
confirmed empirically. In addition, as discussed further in the domain specificity
section, it must be demonstrated that children with autism do not also fail ToM tasks
because of domain-general cognitive or linguistic difficulties.
iii) Causal precedence. Much of the evidence for the ToM hypothesis has
focused on performance on false belief tasks, on which successful performance
normally develops at around the age of four and is interpreted as evidence of a
metarepresentational capacity (Leslie, 1987) or representational understanding of mind
(Perner, 1991). However, in most cases autism is apparent at a much younger age, with
deficits in social responsiveness and reciprocity, symbolic play, gaze behaviour, joint
attention, and imitation often noticed during infancy or when the child is a toddler (e.g.,
Dawson & Adams, 1984; Klin, Volkmar & Sparrow, 1992; Mundy & Sigman, 1989;
Volkmar et al., 1987). Klin et al. (1992) pointed out that as the crux of the ToM
hypothesis of autism is the inability to represent others’ mental states, then the resulting
prediction would be that social impairment in autism should only become apparent at
the age at which metarepresentational skills appear in typically developing children. It
is unclear exactly when this is, with Leslie’s (1987) original thesis being that pretend
play may be the earliest manifestation at around 18 months, and later authors suggesting
that earlier behaviours such as protodeclarative pointing (11-12 months) and joint
attention (8-12 months) may be the earliest signs (Baron-Cohen, 1989d, 1991b),
although these latter abilities are proposed as “precursors” to ToM rather than signs of
an early ToM itself6. Regardless, Klin et al. (1992) found that six types of social
behaviour from the Vineland Adaptive Behavior Scales which emerged prior to the age
6 Leslie and Happé (1989) have, however, argued that joint attention may also indicate the emergence of an ability to represent mental states, as these behaviours convey the intention to communicate.
27
of eight months successfully discriminated autistic children from controls. The
nonrepresentational nature of these behaviours, such as “shows anticipation of being
picked up by a caregiver” and “reaches for familiar person”, was taken to indicate an
early pre-mentalising social impairment in autism.
These kind of findings of early, apparently non-mentalistic social impairment in
autism have been interpreted as evidence in favour of a more primary affective,
emotional, or intersubjective impairment in autism (e.g., Hobson, 1993; Klin &
Volkmar, 1993; Mundy, Sigman, & Kasari, 1993). However, the very early recognition
of autism in Klin et al.’s (1992) participants is not typical, with most other studies
finding that it is not possible to reliably detect autism until at least the age of 18 months
(e.g., Johnson et al., 1992). In addition, some of Klin et al.’s autistic participants did
show typical social behaviours, raising the possibility of different subgroups within the
autism spectrum. The question of whether the ToM hypothesis can meet the criterion of
causal precedence remains a matter of debate (see Charman, 2000).
iv) Explanatory value. The strongest form of the ToM hypothesis asserts that
impairment in the ToM module can explain the entire range of symptoms displayed by
individuals with autism (Frith et al., 1991), although original accounts of the ToM
hypothesis focussed mainly on the social and communicative impairments characteristic
of autism. For example, Baron-Cohen (1988) proposed that the ToM hypothesis would
predict impairments in social skills requiring an ability to represent mental states, as
well as pragmatic language skills, as conversing requires that the speaker be aware of
the listener’s mental state. While the relationship between ToM and real-life social
skills appears to make intuitive sense, it has not been directly investigated in many
studies. Dawson and Fernald (1987) reported a significant correlation between autistic
children’s conceptual perspective-taking ability and their teachers’ ratings of social
skills. Frith et al. (1994) found that individuals with autism who passed false belief
tasks were more likely to show evidence of “mind-reading” in their everyday social
behaviour and had better communicative abilities, while those who failed false belief
tasks showed few social behaviours requiring understanding of mental states. In their
sample of young French preschoolers with autism or PDDNOS, Hughes, Soares-
Boucaud, Hochmann, and Frith (1997) found significant differences between ToM
“passers” and “failers” in ratings of everyday social behaviours requiring mentalising
abilities, but only when the teacher rather than the parent was the informant. However,
neither Prior et al. (1990) nor Sparrevohn and Howie (1995) found a significant
28
correlation between false belief performance and social skills, as rated by parents and
teachers respectively.
The literature examining the relationship between ToM and language abilities in
autism is much larger, with most studies confirming Baron-Cohen’s (1988) prediction
of a relationship between ToM and pragmatic language skills (e.g., Capps, Kehres, &
Sigman, 1998; Tager-Flusberg & Sullivan, 1995). However, it has also become clear
that individuals with autism also show non-pragmatic language impairments (e.g., in
lexical and grammatical knowledge) which are not likely to be the result of a ToM
deficit (Tager-Flusberg, 1999b), but which do correlate with false belief task
performance (e.g., Happé, 1995; Sparrevohn & Howie, 1995). This raises the question
of the direction of the causal relationship between language and ToM, the answer to
which is “likely to be complex” (Tager-Flusberg, 2000). While longitudinal studies
have shown that joint attention behaviours (arguably “precursors” to ToM) in toddlers
with autism predicted language gains several years later (Sigman & Ruskin, 1999),
suggesting that ToM ability is necessary for adequate language development, the
reverse has also been demonstrated - that structural language skills play a key role in
ToM development (de Villiers & de Villiers, 1999). Regardless, it is clear that there is a
close relationship between ToM and language in autism.
The same cannot be said, however, for the relationship between ToM and the
much-neglected third feature of the autistic triad, the repetitive behaviours and restricted
interests which are part of the DSM-IV criteria for autism. While the ToM hypothesis is
able to account for the lack of pretend play displayed by autistic children, it is less
obvious how it might explain other aspects of the third feature of the triad, such as
obsessional interests or repetitive arm-flapping or toe-walking. Baron-Cohen (1989c)
and Carruthers (1996) both attempted to explain repetitive behaviours in autism by
proposing that they develop as a strategy to cope with and gain control over the
unpredictable and frightening social world that surrounds the child who is unable to
understand others’ mental states. This account predicts that the frequency of repetitive
activities should be higher in social settings, especially those which lack a predictable
structure. However, most studies have reported the converse finding, that rates of
stereotyped behaviour are lowest during periods of social interaction and highest during
periods where no interpersonal demands are made (Clark & Rutter, 1981; Dadds,
Schwartz, Adams, & Rose, 1988; Donnellan, Anderson, & Mesaros, 1984). In the only
study reported in the literature so far to directly investigate the relationship between
ToM and repetitive behaviours in autism, Turner (1996, 1997) found no relationship
29
between false belief task performance and the incidence or severity of a large range of
repetitive behaviours.
Similarly, the ToM hypothesis faces difficulty explaining so-called “non-triad”
features of autism, which appear frequently but are not part of the diagnostic criteria
(Frith & Happé, 1994; Tager-Flusberg, 2001). These include savant abilities,
exceptional visuospatial and visuoperceptual skills, over-selective attention, and
heightened sensory sensitivities. These aspects of autism do not bear an obvious
relation to ToM ability, and may be better explained by the local processing style that
appears to be characteristic of autistic individuals (Frith & Happé, 1994; Happé, 1997,
1999; Plaisted, 2000, 2001). The inability thus far of the ToM hypothesis to adequately
meet the criterion of explanatory value could arguably be considered one of the most
substantial problems to have faced it.
v) Domain specificity. The criterion of domain specificity results from the claim
that ToM reflects an innate module that develops separately from other cognitive
capacities, and is independently impaired in autism (Leslie, 1987, 1991; Baron-Cohen,
1991c). This assertion has been defended by citing evidence that individuals with
autism are able to pass tasks which have equivalent structure and demands to false
belief (or other ToM) tasks, but do not have mentalistic content. This approach of
comparing autistic assets and deficits on tasks which require mentalising and those
which do not has been dubbed the “fine cuts” technique by Frith and Happé (Frith &
Happé, 1994; Happé & Frith, 1995). For example, it has been found that while
individuals with autism fail false belief tasks, they are able to pass tests involving false
photographs, drawings, and models7 (Charman & Baron-Cohen, 1992, 1995; Leekam &
Perner, 1991; Leslie & Thaiss, 1992). Similarly, autistic individuals demonstrate
understanding of behavioural but not mentalistic picture sequences (Baron-Cohen et al.,
1986), understand “see” but not “know” (Perner et al., 1989), and engage in physical
sabotage but not deception (Sodian & Frith, 1992). Additionally, Baron-Cohen (1991c)
found that participants with autism were not impaired in domains of social cognition
which do not require a ToM.
However, the claim of domain specificity has come under increasing criticism,
with recent evidence suggesting that impaired performance on false belief tasks may
7 The false photograph paradigm, for example, runs as follows: a horse puppet takes a photograph of a cat puppet. The cat then moves from the chair to the bed. The child is asked, “In the photograph, where is the cat sitting?”, as well as two control questions probing knowledge of where the cat was when the horse took the photograph and where the cat is now. It is argued that this task is identical in structure to the false belief task, but requires reasoning about outdated physical representations instead of mental representations.
30
still be explained, and even better accounted for, by deficits in more domain general
processes such as EF (see Section 2.3) or language impairment (Bruner & Feldman,
1993; Tager-Flusberg, 2000). Zelazo and colleagues (Zelazo et al., 1996a; Zelazo,
Burack, Boseovski, Jacques, & Frye, 2001) have argued that failure to explicitly test
and meet the assumption that two tasks (such as the false belief and false photograph
tasks) are of the same underlying complexity is problematic, and weakens any
arguments for domain specificity and modularity (this argument, and further criticisms
of the false photograph task, are discussed in Section 2.3). They have found that
children with autism are impaired on another control task without mental content, which
is matched for underlying complexity to the false belief task (e.g., Zelazo et al., 1996a).
Frye (2000) also provides a number of cogent a priori arguments for why ToM should
not be considered a domain specific function. For example, given that ToM refers to
understanding one’s own beliefs as well as others’ (with research confirming self-other
equivalence on the Smarties and appearance-reality tasks), Frye questions where the
domain boundaries in our own beliefs would be, as our beliefs can be about non-
mentalistic things such as physics or biology. He also provides a critique of Baron-
Cohen’s (1994) Intentionality Detector module, pointing out that assigning intention on
the basis of the direction of movements will tend to over-ascribe intentionality to every
change in direction we happen to take, and under-ascribe intentionality to acts which do
not involve movement, such as not preventing something from happening.
While proponents of the ToM hypothesis have constructed some fairly plausible
defences against several of the attacks on its claims of universality, uniqueness, causal
precedence, explanatory value, and domain specificity, converging counterargument
and evidence has resulted in a general retreat from the strong version of the hypothesis -
that autism may be explained by a single primary cognitive deficit in a ToM module.
While some authors still largely adhere to the original strong version of the ToM
hypothesis (e.g., Surian & Leslie, 1999), many of its original proponents now advocate
a weaker version in which ToM is conceptualised as one of multiple cognitive
impairments in autism (Baron-Cohen & Swettenham, 1997; Frith & Happé, 1994;
Happé & Frith, 1996), and/or is not necessarily considered to be a unitary module, but
rather a more multidimensional ability which emerges gradually during development
(Tager-Flusberg, 2001).
31
2.2 Executive function (EF)
2.2.1 Defining and measuring EF
Because of its complex and theoretical nature, defining and operationalising “executive
function” has proven to be a persistent problem. To some extent, the chosen definition
of EF is dependent upon the author’s favoured model of its underlying structure.
However, EF is generally understood to be an umbrella term covering a number of
related but distinct high-level cognitive capacities which help guide and control
purposeful behaviour towards attainment of a goal (e.g., Lezak, 1993; Luria, 1966;
Stuss & Benson, 1986; Welsh & Pennington, 1988). These capacities include planning,
set-shifting (also known as attentional switching or cognitive flexibility), strategy
formation, inhibition, working memory, generativity8, decision-making, and self-
monitoring. In his overview of issues in EF assessment, Rabbitt (1997) proposed that:
“executive control is necessary to deal with novel tasks that require us to formulate a goal, to
plan, and to choose between alternative sequences of behaviour to reach this goal, to compare
these plans in respect of their relative probabilities of success and their relative efficiency in
attaining the chosen goal, to initiate the plan selected and to carry it through, amending it as
necessary, until it is successful or until impending failure is recognised.” (p. 3)
Compounding the difficulty with the precise definition of EF is the regular tendency to
use the term “frontal” (or more precisely, “prefrontal”) as a synonym for “executive”,
thereby confusing neuropsychological and neuroanatomical concepts. This confusion
has arisen because the cognitive construct of EF was originally posed in response to
observations of patients with frontal lobe damage, whose disorganised and disinhibited
behaviours were hypothesised to have their origins in executive dysfunction. This led to
a situation whereby some authors have considered any operation performed by the
frontal lobes (or any behavioural symptom of frontal lobe damage) to be an EF,
including constructs or functions such as emotion regulation and affective
responsiveness, social behaviour and personality, insight, humour appreciation, and
self-awareness. In the view of Zelazo and Müller (2002), EF includes both “cool”,
cognitive aspects and “hot”, affective aspects. However, it is important to make it clear
that in this thesis, EF will be considered a purely cognitive (“cool”) construct, as
8 The term ‘fluency’ is often used for this concept, however ‘generativity’ is preferred within this thesis.
32
defined (albeit broadly) above9, independent from any neuroanatomical basis. While
there is strong support for the notion that EFs are at least partially subserved by frontal
regions of the brain, mounting evidence indicates that the relationship is certainly not
well defined, and many EF measures lack both sensitivity and specificity to frontal
lesions (Reitan & Wolfson, 1994; Stuss & Alexander, 2000; Tranel, Anderson &
Benton, 1994).
Understandably, the measurement of EF in both adults and children has been
just as problematic as its definition. The difficulty with EF measurement was predicted
by Fodor (1983), who proposed the existence of domain-general, non-modular “central
processes” which would be “bad candidates for scientific study” (p.127). Unlike ToM,
where the false belief paradigm has (arguably) become a “gold standard” for its
measurement, a similar gold standard for assessing EF has proven elusive – and perhaps
unfeasible. This is not only because of EF’s complexity, but also because EF is a
theoretical rather than an operational term (Burgess, 1997). To borrow Burgess’
example, one can clearly call a patient dyscalculic if s/he shows impaired performance
on calculation tasks (or prosopagnosic if s/he shows impaired performance on face
recognition tasks), but there is no equivalent way of determining whether or not a
individual may be diagnosed as dysexecutive: there is no prototypical screening
measure. This has meant that unlike any other cognitive domain, the validity of an EF
test is not typically evaluated on psychological grounds, but rather in terms of whether
or not patients with frontal lesions show impaired performance on it. However, the
loose correspondence between the psychological and anatomical make this inference
problematic. For example, one of the most widely used tests of EF is the Wisconsin
Card Sorting Test (WCST; Grant & Berg, 1948), in which participants must work out
rules for sorting cards by certain categories, and then adapt their responses according to
feedback when the rules unexpectedly change. Patients with frontal lobe lesions have
previously been found to achieve less categories and make more perseverative errors
than patients with posterior lesions (Drewe, 1974; Milner, 1963). However, more recent
evidence has shown that non-frontal or diffuse brain damage can produce similar
deficits (e.g., Anderson, Damasio, Jones, & Tranel, 1991; Anderson, Bigler, & Blatter,
9 Clearly, affective factors influence and interact with cognitive processes, however it is possible to distinguish the two conceptually, methodologically and neuroanatomically. Furthermore, in examining the relationship between ToM and EF, considering affective and/or social factors to be part of the EF domain unhelpfully clouds the issue - for example, Zelazo and Müller (2002) actually deem false belief tasks to be tests of EF.
33
1995), and that adequate WCST performance does not exclude frontal pathology (e.g.,
Eslinger & Damasio, 1985).
There has, however, been some attempt to delineate purely cognitive criteria for
a test being a test of EF. For example, Phillips (1997) proposed that any test
hypothesised to measure EF should have the following characteristics: i) the test should
be novel, in order to tap goal identification and strategic planning (as well-practiced
tasks can be performed using previously formulated strategies); ii) the test should be
effortful in terms of task planning and execution, requiring inhibitory control and
monitoring; and iii) the test may involve working memory, in order to coordinate
concurring processing requirements. Similarly, Walsh (1978) proposed that EF tasks
require novelty, complexity, and the need to integrate information. However, these
criteria clearly apply only to multifactorial EF tasks – other more specific tests of
particular EF components may not involve all of these features (the notion of EF
components is discussed further in the next section).
Besides the difficulty in ascertaining the construct validity of EF tests, there are
several other reasons why measuring EF is challenging. Firstly, the psychometrics of
EF tests are notoriously poor. By its very nature, EF is required in novel situations; yet
because tests can only be novel once, the test-retest reliability of EF tasks is
consequently low (Rabbitt, 1997). Secondly, most EF tests lack purity – that is, they tap
multiple underlying processes, making it difficult to discern specific reasons for failure.
For example, the WCST has been commonly described as a test of abstraction and
flexibility, yet it also requires selective attention to relevant dimensions of the stimuli,
generation of a sorting rule, working memory to hold the sorting principle in mind, and
inhibition of the prepotent response to sort the cards according to the rule just used, as
well as non-EF processes such as the use of verbal feedback provided by the examiner,
and appreciation of the category of number (Ozonoff, 1995a; Pennington & Ozonoff,
1996). Furthermore, attempts to control task demands in order to isolate the relevant
abilities may not always be successful, as some EF tasks specifically require the
simultaneous co-ordination of a variety of different processes (Kimberg & Farah, 1993).
A third difficulty with EF measurement is that of low “process-behaviour
correspondence” (Burgess, 1997). In contrast to most other cognitive domains which
are only manifest in circumscribed situations (e.g., calculation abilities are manifest
when one is required to perform a calculation, or face recognition abilities when
presented with a face), EFs manifest themselves across a range of different situations.
As a consequence, there is an imprecise correspondence between behaviour and the
34
underlying process: a specific EF impairment can result in a variety of behaviours, and a
specific behaviour may be caused by a variety of EF (or other cognitive) impairments.
Furthermore, the same behavioural sequence is likely to require fewer EFs over time,
even within one short period, as it becomes more practiced. A final problem with
assessment of EF is that testing situations are for the most part structured and guided by
the examiner, removing much of the load on EF for the examinee. This results in poor
ecological validity of EF tests (Cripe, 1996).
Do these issues with EF assessment apply equally to children and adults? Until
recently, the large majority of EF research has focus on adult populations, and
measuring EF in childhood has only lately become a more popular topic of interest10 as
it becomes clear that EF develops much earlier than previously thought (see Section
2.2.2). Hughes and Graham (2002) argue that while EF measurement in children has its
own set of problems, conversely there are actually some difficulties with adult EF
assessment which are not so problematic in childhood. For example, children may
perceive a new task as novel for longer, possibly leading to greater stability in
underlying processes and overall performance (and therefore improved test reliability
and validity). In addition, as EF tasks need to be simplified in order to be
developmentally appropriate, the problem of task impurity is likely to be reduced.
However, there are also difficulties associated with assessing EF in children, the most
obvious of which, according to Hughes and Graham (2002), is children’s limited
language skills. This leads to a number of problems: i) complex task instructions tax
verbal comprehension, which may influence task performance for non-EF reasons; ii)
because fluent literacy is not an automatic skill until late in development, many adult EF
tasks which depend on written language being over-learned are not appropriate for
children (e.g., the Stroop test, in which reading a colour-word such as “red” is assumed
to be a prepotent response, making it difficult to instead say the different colour that the
word is printed in); and iii) language itself may play a role in EF, by both guiding
behaviour through internal self-talk and by enabling the use of verbal working memory
to rehearse strategies. Clearly, the development of appropriate assessment tools for EF
in children is an important focus for ongoing research.
While reviewing the literature on the definition and measurement of EF appears
to paint a rather negative picture, it should be emphasised that EF has still managed to
retain its utility, relevance and validity as a measurable construct. Tranel et al. (1994)
10 For lists of commonly used EF tests in children, see Anderson (1998) or Zelazo and Müller (2002).
35
argue that despite the difficulties in studying EF, the term provides a useful heuristic or
shorthand for denoting a relatively well-agreed upon set of capacities with several
unique characteristics: i) they are the highest level of human cognition; ii) they are
difficult to operationalise and therefore hard to measure quantitatively; iii) they are
closely intertwined with personality and consciousness; and iv) they have intimate
connections with the prefrontal cortex. In addition, recent research has focused on
improving the measurement of EF by using more fine-grained tests with several
performance measures and multiple control tasks in order to isolate specific components
of EF which may be impaired (Delis, Squire, Bihrle, & Massman, 1992; Godefroy,
Cabaret, Petit-Chenal, Pruvo, & Rousseaux, 1999; Ozonoff, Strayer, McMahon &
Filloux, 1994) as well as attempting to make the task paradigms more ecologically valid
(Manly et al., 2001; Wilson, Evans, Emslie, Alderman, & Burgess, 1998) and child-
friendly (Espy, Kaufmann, Glisky, & McDiarmid, 2001; Gerstadt, Hong, & Diamond,
1994; Hughes, 1998a). It is nevertheless important to acknowledge the difficulties with
the definition and measurement of EF, as it will become evident that these concerns
were influential both in selecting appropriate EF tasks for the current research, and in
interpreting the results gleaned from those tasks.
2.2.2 Models of EF and its development
A burgeoning literature has produced a sizeable number of alternative theoretical
frameworks for conceptualising EF, and, concurrently, the functions of the prefrontal
cortex (see Eslinger, 1996; Grafman, 1994; Stuss & Knight, 2002). As is the case for
EF measurement, models of EF have mostly been based on adults, with theories of EF
development borrowing heavily from adult concepts. Several adult-based models have
utilised the classic distinction between automatic and controlled actions from traditional
cognitive psychology (Atkinson & Shiffrin, 1968; Schneider & Shiffrin, 1977), where
unlike automatic actions, controlled actions involve conscious, effortful processing and
are required in novel, non-routine situations. Based on a similar routine/non-routine
dichotomy, an influential model of EF by Norman and Shallice (Norman & Shallice,
1980, 1986; Shallice, 1988) included two mechanisms for regulating behaviour: the
Contention Scheduler, which operates in routine or overlearned situations via automatic
priming of stored knowledge (analogous to scripts or schemas), which are cued either
by environmental stimuli or conceptual thought; and the Supervisory Attentional System
(SAS), which is activated in non-routine (novel, complex, difficult, and/or conflicting)
36
situations and in which conscious internal knowledge states can override the contention
scheduling mechanism and set the priority for action by creating new action schemata.
This model held an intuitive appeal and accounted for data on attention and action
failures in patients with prefrontal lesions, who were proposed to have intact Contention
Schedulers but an impaired SAS (Shallice & Burgess, 1991). Shallice (2002; Shallice
& Burgess, 1996) recently elaborated upon his model, outlining a number of different
components to the SAS including schema selection (which can occur via three different
methods), schema implementation, and schema checking and monitoring.
While Shallice (1984, 2002) contends that central control processes consist of
multiple components, others have argued for a more unitary control structure or
mechanism. Duncan (Duncan, Burgess, & Emslie, 1995; Duncan, Emslie, Williams,
Johnson, & Freer, 1996) to some extent represented this position when he claimed that
EF is largely synonymous with Spearman’s g, or fluid intelligence. Others have
proposed that the range of EF failures may be attributed to a single process, such as
inhibition (e.g., Dempster, 1992, 1993) or working memory (e.g., Case, 1985;
Goldman-Rakic, 1995). The idea of a single EF system or mechanism has become
increasingly unpopular, however, as evidence and opinion has converged upon the
notion that EF consists of multiple separable components (Baddeley, 1996, 2002;
Godefroy et al., 1999; Miyake, Friedman, Emerson, Witzki, & Howerter, 2000;
Pennington, 1997; Stuss & Alexander, 2000). This conceptualisation accounts better for
data showing weak correlations between various EF tasks (e.g., Boone, Ponton,
Gorsuch, Gonzalez, & Miller, 1998; Hughes, Russell, & Robbins, 1994; Miyake et al.,
2000) and differential impairment on various EF tasks in patients with lesions in
different parts of the prefrontal cortex (see Stuss & Alexander, 2000). In addition, it
allows us to distinguish between the various clinical groups in which executive
dysfunction is found, by examining qualitative differences in profiles of performance on
EF components (reviewed in the next section).
However, there has been little agreement on the appropriate taxonomy for the
components of EF. Lezak (1995) proposed four EF components: i) volition, ii)
planning, iii) purposive action, and iv) effective performance. In their problem-solving
framework of EF, Zelazo, Carter, Reznick and Frye (1997) also outlined four
components: i) problem representation, ii) planning, iii) execution (rule use) and iv)
evaluation (error detection/correction). While these two frameworks present four
sequential stages to the problem-solving process, other conceptualisations focus more
on concurrent or non-time-dependent executive processes. For example, Anderson
37
(1998; Anderson, Levin, & Jacobs, 2002) proposes three EF components: i) attentional
control (including selective and sustained attention and response inhibition), ii) goal
setting (incorporating initiation, planning, and problem solving), and iii) cognitive
flexibility (including working memory, attentional shifting and self-monitoring). One
model which has been particularly influential in the developmental literature is Roberts
and Pennington’s (1996) interactive framework, in which performance on EF tasks is
held to be a product of two separate but interdependent processes: working memory (to
hold the task demands or rules in mind) and inhibitory control (to guide behaviour
according to those rules). These two components compete for limited executive
resources. Support for this model comes from studies showing that on EF tasks,
prepotent response errors increase (indicating poorer inhibitory capacity) as the working
memory demands increase (Roberts, Hager & Heron, 1994).
A number of factor analytic studies have unfortunately produced a range of
different results, without clearly indicating any particular model as superior to others
(see Royall et al., 2002, for a review). For example, Burgess, Alderman, Evans, Emslie,
and Wilson (1998) found three factors which they termed Inhibition, Intentionality and
Executive Memory, whereas Collette, van der Linden and Salmon (1999) found two
factors, Inhibition and Working Memory, while Boone et al. (1998) found only one
Cognitive Flexibility factor (although in the latter study the EF tasks were only
modestly correlated and the authors concluded that the tests tapped somewhat different
abilities). In several of these studies, variables from the same EF task often load on
different factors, and in addition, the same task may load on different factors in different
studies depending upon which tests are included in the analysis. These inconsistencies
are hardly surprising given the “impurity” of EF tasks and their questionable
psychometric properties.
As argued by Hughes and Graham (2002), factor analysis studies using children
may be more fruitful, as more simple, “pure” tests may still tax EF in children and test
performances may be more reliable. Support for the fractionation of EF in children has
indeed been provided by several studies which have generally revealed three or four
distinct EF factors (Espy, Kaufmann, McDiarmid, & Glisky, 1999; Hughes, 1998a;
Levin et al., 1991; Luciana & Nelson, 1998; Pennington, 1997; Welsh, Pennington &
Groissier, 1991). Although named differently by different authors, these factors have
consistently included cognitive flexibility or set-shifting, inhibition, and working
memory, with the addition or substitution of a planning component in some studies. For
example, Hughes (1998a) identified Attentional Flexibility, Inhibitory Control and
38
Working Memory factors, and Pennington (1997) similarly found Set Shifting or
Cognitive Flexibility, Motor Inhibition, and Verbal Working Memory factors; while
Welsh et al. (1991) named their factors Planning, Hypothesis Testing & Impulse
Control, and Fluid & Speeded Response. However, inconsistencies also appear in this
developmental research, where the same task may be clustered with different tasks or be
part of different factors across studies, although it is difficult to tell how much of this
variability is attributable to different performance indices being used in the various
studies.
The stages of EF development have only relatively recently become the subject
of systematic research, as evidence accumulates in opposition to the early influential
notion that the prefrontal cortex was not functional at all until adolescence and did not
reach maturity until around the age of 24 (Golden, 1981). Behavioural and
electroencephalogram (EEG) data as well as case studies of children with early frontal
lesions all now refute this view, indicating prefrontal activity even in infancy. For
example, Diamond and Goldman-Rakic (1989; Diamond, 1985) found that by 12
months of age, human infants achieved errorless performance on classic delayed
response and A-not-B tasks, performance on which they argue requires working
memory and inhibition, and is sensitive to frontal lesions in monkeys. Bell and Fox
(1992) demonstrated changes in frontal EEG recordings during the first year of life
which correlated with improved performance on the A-not-B task. A number of case
studies (Anderson, Bechara, Damasio, Tranel, & Damasio, 1999; Eslinger, Biddle, &
Grattan, 1997; Marlowe, 1992; Price, Daffner, Stowe, & Marsel Mesulam, 1990) have
also demonstrated that very early prefrontal lesions result in immediately noticeable
consequences as well as EF deficits and impaired social and moral behaviour later in
life. Nevertheless, it is clear that although it is certainly not “silent” in infancy, both the
physiological and functional development of the prefrontal cortex follow a particularly
protracted developmental course. Investigations of synaptic density, dendritic growth,
myelination, interhemispheric connectivity, metabolic activity, and electrical (EEG)
activity all show that the prefrontal cortex continues to develop through middle
childhood and adolescence (Diamond, 2002; Huttenlocher & Dabholkar, 1997;
Schwartz, 1997; Thatcher, 1997).
Cognitive studies of the development of EF have focused on mapping
developmental trajectories for the various EF components. An early study by Passler,
Isaac, and Hynd (1985) found that EF development was a multistage process, with a
spurt of development between the ages of 6 and 8 and mastery evident by the age of 12.
39
Similarly, Chelune and Baer (1986) found that WCST performance improved between 6
and 10 years of age, with adult performance achieved by 12 years. More recent studies
incorporating a larger range of EF measures have extended the age range and more
thoroughly articulated the multidimensional nature of EF development. A study by
Levin et al. (1991) supported previous findings that tests of concept formation, set-
shifting and inhibition appear to be mastered by the age of 12, however they also found
additional gains in their adolescent 13-15 year-olds on measures of generativity and
planning. Welsh et al. (1991) found evidence for three distinct developmental stages,
the first beginning at around 6 years, a second commencing at around the age of 10, and
a third during adolescence. Consistent with Levin et al. (1991), they found that some
components of EF (e.g., the ability to resist distraction, impulse control or inhibition)
matured earlier than others (e.g., generativity, planning skills). An investigation of EF
development in late childhood and adolescence by Anderson, Anderson, Northam,
Jacobs & Catroppa (2001) found that while the developmental trajectory for EFs in this
period was generally flatter than during early and middle childhood, differential
developmental trends were observed within the different EF domains, with attentional
control and planning showing the greatest improvements during adolescence, while
cognitive flexibility was already matured by the age of 12. At the other end of the age
spectrum, studies by Zelazo and colleagues (Zelazo & Reznick, 1991; Zelazo, Frye, &
Rapus, 1996b) have demonstrated developments in rule use (the third stage in their
problem-solving framework of EF) between the ages of 2 and 5. Several studies have
also found significant improvements in inhibitory control between the ages of 3 and 6
years (Diamond & Taylor, 1996; Gerstadt et al., 1994; Kochanska, Murray, & Coy,
1997).
Thus, descriptive studies mapping the development of EF have shown fairly
consistently that i) the first emergence of EF occurs early in life, probably around the
end of the first year; ii) EF development appears to follow a multistage process, with
important changes occurring between the ages of 2-5 and 6-10, with adult performance
levels reached by the age of 12 in several domains, and performance in other domains
continuing to develop through adolescence; and iii) the various components of EF
follow different developmental trajectories, with cognitive flexibility and inhibition
tending to develop first and planning and generativity maturing later (see Anderson,
2002, for a slightly different mapping of EF development).
Attempts to characterise the development of EF within a theoretical framework
have tended to emphasise either one or two central constructs which account for EF
40
development as a whole. One view is that age-related changes in EF may be explained
by the construct of inhibition, such that children become increasingly able to resist
interference and keep task-irrelevant information out of working memory (Bjorklund &
Harnishfeger, 1990; Dempster, 1992, 1993; Harnishfeger & Bjorklund, 1993). As for
adult models of the structure of EF, this account is limited by its unidimensionality,
defaulting to the explanation that children find some tasks more difficult than others
simply because they require more inhibition, and being unable to explain developments
in EF tasks with minimal inhibitory requirements (Zelazo et al., 1997; Zelazo & Müller,
2002). A more popular approach has been to argue that EF changes result from both
working memory and inhibition, either as potentially separable components (Diamond,
2002; Diamond & Taylor, 1995; Gerstadt et al., 1994) or interacting processes (Roberts
et al., 1994; Roberts & Pennington, 1996). Results of a recent well designed study by
Beveridge, Jarrold, and Pettit (2002) favoured the view of inhibition and working
memory as independent and additive rather than interacting components of EF. While
they found that increasing the working memory load of an inhibition task did have a
detrimental effect on performance (consistent with Roberts et al., 1994), by using tests
with multiple levels of both inhibitory and working memory requirements they found
that interactions between the two processes were non-significant in both 6- and 8-year-
olds.
A recent alternative theory of EF development is Zelazo and Frye’s Cognitive
Complexity and Control (CCC) theory (e.g., Frye, Zelazo & Palfai, 1995; Zelazo, 2000;
Zelazo & Frye, 1998), which, according to Frye (2000), is most relevant to the EFs of
planning and deliberative action. This account focuses on development in the preschool
years, proposing that within this period there are increases in the complexity of
children’s rule systems (plans formulated in potentially silent self-directed speech, e.g.,
“If I see a mailbox, then I need to mail this letter”). Complexity is measured by the
number of levels of embedding in these rule systems. Embedded rules establish a
hierarchy in which rules are arranged beneath setting conditions (which select or restrict
the application of a rule), and have the form “if s1, then if a1, then c1” in which s is a
setting condition, a is an antecedent, and c is a consequent. Zelazo and Frye (Frye et al.,
1995; Zelazo & Frye, 1998; Zelazo & Reznick, 1991) have shown that 3-year-olds
readily integrate two “if-then” rules (e.g., in the Dimensional Change Card Sorting
(DCCS) Task, modelled on the WCST, they are able to comprehend “If the test card is
red then place it here; if blue then there”) but having difficulty representing a higher-
order “if-if-then” rule that allows them to switch flexibly between incompatible pairs of
41
rules (e.g., “If sorting by colour, then if red then here, if blue then there. If sorting by
shape, then if car then here, if flower then there”). The CCC account is reminiscent of
Halford’s proposal that cognitive development is characterised by the developing ability
to represent increasingly complex relations between items in parallel (Halford, 1993;
Halford, Wilson, & Phillips, 1998), but differs from Halford in the central importance
placed on embedded or hierarchical rule structures (Frye & Zelazo, 1998).
While both the inhibition-working memory accounts and the CCC theory of EF
development are able to account reasonably well for data within their specified task and
age domains, it is doubtful whether either of them could account for the range of
findings from descriptive studies mapping developmental trajectories of EF. Neither
theory explicitly accounts for the multi-stage process of development from infancy to
adulthood or the differential rate of development for the various components of EF. To
some extent, these limitations stem from the widely acknowledged problems with the
definition and measurement and EF, which make it difficult to agree on which
component processes underlie each EF task and which aspects are central to EF
development. Nevertheless, this important cognitive domain remains critical to the
explanation of a wide range of clinical disorders - including autism.
2.2.3 EF in autism
Research on executive dysfunction in autism has gathered momentum in a more gradual
manner than the ToM literature, beginning with early case reports of autistic individuals
documenting what would now be called EF deficits (Scheerer, Rothmann, & Goldstein,
1945; Steel, Gorman, & Flexman, 1984). Using tests of spontaneous colour and tone
sequence production, Frith (1972) found what might be interpreted as a generativity
impairment in children with autism, with the autistic sample producing more rigid,
restricted, and less unique patterns. In 1978, Damasio and Maurer published an
influential paper noting behavioural similarities between individuals with autism and
patients with frontal lobe damage, such as ritualistic and compulsive behaviours and
concreteness in thought and language. They proposed a neurological model of autism
involving the frontal lobes and parts of the temporal lobes, basal ganglia and thalamus.
Following this, Rumsey and colleagues (Rumsey, 1985; Rumsey & Hamburger, 1988)
tested a group of high-functioning male adults with autism on executive and non-
executive neuropsychological tasks. They found that the autistic men performed
significantly more poorly than controls on the WCST as well as measures of cognitive
42
flexibility and problem solving, but showed intact or only mildly impaired performance
in other cognitive domains. This finding was followed up in autistic adolescents by
Prior and Hoffman (1990), who found impaired performance on the WCST and a maze
test; and in individuals with Asperger syndrome by Szatmari, Tuff, Finlayson, and
Bartolucci (1990), who found impaired performance on the WCST.
The current era of EF research in autism was launched by a study by Ozonoff et
al. (1991), which compared the primacy of ToM and EF impairments in a group of
high-functioning children with autism. Contrary to their expectation, Ozonoff et al.
found that in their autism group, EF deficits (as measured by the WCST and the Tower
of Hanoi, a measure of planning) were more universal than ToM deficits and were
better predictors of autism group membership. Such findings have since been
consolidated in a number of studies showing impairment of individuals with autism
compared with age and IQ-matched controls on tasks tapping a range of EF
components, including cognitive flexibility or attentional shifting (Ciesielski & Harris,
1997; Courchesne et al., 1994; Goldstein, Johnson, & Minshew, 2001; Hughes &
Russell, 1993; Hughes et al., 1994; Minshew et al., 1992; Ozonoff & Jensen, 1999;
Ozonoff & McEvoy, 1994; Ozonoff et al., 1994), planning (Hughes, 1996a; Hughes et
al., 1994; Ozonoff & Jensen, 1999; Ozonoff & McEvoy, 1994), and generativity
(Boucher, 1988; Craig & Baron-Cohen, 1999; Lewis & Boucher, 1991; Turner, 1999;
Williams, Moss, Bradshaw, & Rinehart, 2002). In their review of studies on EF in
autism, Pennington and Ozonoff (1996) calculated the average effect size of group
differences on EF tasks to be 0.98 (a large effect according to Cohen, 1988), and as high
as 2.07 on the Tower of Hanoi. These EF deficits do not appear to be attributable to
impairments in more basic attentional processes, such as sustained or selective attention
or basic attentional capacity (Bryson, Landry, & Wainwright, 1997; Garcia-Villamisar
& Della Sala, 2002; Garretson, Fein, & Waterhouse, 1990; Goldstein et al., 2001;
Minshew et al.,1992).
These findings resulted in the hypothesis that EF deficits may be primary in
autism (e.g., Hughes & Russell, 1993; Ozonoff et al., 1991; Russell, 1997a).
Furthermore, prefrontal dysfunction has been posited to be the underlying
neuroanatomical basis for EF impairment in autism (Ozonoff, 1995a; Ozonoff et al.,
1991). In a test of the prefrontal hypothesis, Bennetto, Pennington, and Rogers (1996)
examined the pattern of performance displayed by individuals with autism on various
memory tasks, and found that it was consistent with that typically displayed by frontal
lobe patients. Minshew, Luna, and Sweeney (1999) found that the pattern of
43
performance of autistic individuals on oculomotor tasks suggested a disturbance in
prefrontal circuitry. Consistent with the notion that autistic symptomatology may have
its basis in frontal dysfunction, impairments in social interaction, spontaneous speech
and pragmatic communication, and the production of novel, goal-directed behaviours,
are also displayed by patients with frontal lobe damage, including children who have
sustained early damage to the prefrontal cortex (e.g., Alexander, 2002; Ames,
Cummings, Wirshing, Quinn, & Mahler, 1999; Anderson et al., 2002; Eslinger et al.,
1997; Stuss & Benson, 1984, 1986; Tranel, 2002).
Neuropathological and neuroimaging studies have also provided some evidence
of prefrontal abnormalities in autism, although so far no gross abnormalities have been
consistently identified. Casanova, Buxhoeveden, Switala, and Roy (2002) discovered
minicolumnar abnormalities in the frontal and temporal lobes of children with autism,
and Piven et al. (1990a) found evidence of abnormal neural migration in the frontal
lobes of three autistic individuals, although this was only one-fifth of their sample.
Piven et al. (1995) also found that the frontal lobes were small in comparison with other
cortical areas in subjects with autism. Zilbovicius et al. (1995) demonstrated evidence
that maturation of the frontal cortex (as measured by regional cerebral blood flow,
rCBF) was delayed in autistic individuals. Decreased rCBF in frontal areas has also
been found in a number of other studies (George, Costa, Kouris, Ring, & Ell, 1992;
Ohnishi et al., 2000; Sherman, Nass, & Shapiro, 1984). Functional neuroimaging
studies have shown reduced dorsolateral prefrontal activation during spatial working
memory tasks in autism (Luna et al., 2002) as well as differences in the pattern of
activation in the prefrontal cortex and other brain regions during ToM tasks (Castelli et
al., 2002; Happé et al., 1996).
However, prefrontal changes are only one of many brain abnormalities which
have been documented in autism (see Bauman, 1999; Deb & Thompson, 1998; Koenig,
Tsatsanis, & Volkmar, 2001), with other areas of significance including the cerebellum
(see Courchesne, 1997), corpus callosum (Piven, Bailey, Ranson, & Arndt, 1997a), and
limbic or medial temporal structures (Bachevalier, 1994; Bauman & Kemper, 1994).
Most neurobiological theories of autism in fact do not give prominence to the prefrontal
cortex (e.g., Akshoomoff, Pierce, & Courchesne, 2002; Waterhouse, Fein, & Modahl,
1996). Concluding that prefrontal abnormalities are the most significant in causing the
symptoms of autism would therefore be premature, particularly given the inconsistency
of results across neurobiological studies of autism in general (Ozonoff, 2001). In
addition, a major difficulty with the prefrontal hypothesis is that children who sustain
44
early lesions to the prefrontal cortex do not actually develop autism, but more often
display a syndrome resembling psychopathy or conduct disorder (Anderson et al., 1999;
Eslinger, Grattan, Damasio, & Damasio, 1992; Eslinger et al., 1997). While the
behavioural impairments displayed by children with frontal lesions may be broadly or
categorically similar to those displayed by children with autism, there are obvious
qualitative differences and it would be difficult to mistake one for the other in a clinical
setting. However, it may be that prefrontal dysfunction is a necessary but not sufficient
criterion for the development of autism (Ozonoff, 1995a), or that the timing of the insult
is a crucial variable in behavioural outcome. Nevertheless, it is not clear that prefrontal
dysfunction is the neurobiological basis for EF impairment in autism11.
The central concern of this section, however, is the hypothesis that EF
impairment, regardless of its neuroanatomical underpinnings, may be the primary
cognitive impairment in autism (Hughes & Russell, 1993; Ozonoff et al., 1991;
Ozonoff, 1995a; Pennington et al., 1997; Russell, 1997a). The associated claim of this
hypothesis is that a primary impairment in EF may also explain the ToM deficit
observed in individuals with autism (a claim which is examined in Section 2.3.2). As
for ToM, the EF hypothesis of autism has undergone a number of tests of whether or not
it meets the criteria for primacy, as reviewed below. Unlike ToM, the EF hypothesis is
not required to prove its domain specificity, as EF is not claimed to be modular.
i) Universality. The evaluation of whether or not EF deficits are universal in
autism has not received nearly as much attention as the equivalent question in the ToM
literature. This is probably because of the lack of a “gold standard” of EF performance
on which failure can be unequivocally evaluated – unlike false belief tasks, performance
on EF tasks is usually not a matter of pass or fail. Many studies of EF in autism
interpret the presence of a group difference as evidence of an EF deficit in autism, but
do not look further at what proportion of the autism group showed such a deficit. Those
researchers who have examined the universality of deficits have tended to choose an
arbitrary criterion for what comprises a “fail” or what defines a “deficit”. Ozonoff et al.
(1991) used the proportion of participants scoring below the mean of the control group
as their index of the universality of deficits, and found that 96% of their autism group
showed an EF deficit using this criterion. This finding was (and still is) cited as
11 As Pennington et al. (1997) point out, EF impairment can also result from diffuse structural or metabolic differences in brain development or diffuse brain lesions in adult patients (as discussed in Section 2.2.1), indicating either that disrupting the connectivity of the whole brain may mimic the effects of a focal frontal lesion, or that the neuroanatomical basis of EF tasks is not specific to the prefrontal cortex.
45
evidence that EF deficits were almost universal among individuals with autism.
However, defining any score below the mean as a deficit is a very lenient criterion – in
most rating systems for impairment severity, a score needs to be at least 1 standard
deviation (SD) below the mean to be considered as even mildly impaired (Heaton,
Grant, & Matthews, 1991; Lezak, 1995).
Other studies have been more equivocal than Ozonoff et al.’s (1991) study in
their findings on the universality of EF impairments in autism, usually finding lower
proportions of individuals with autism demonstrating difficulties. Liss et al. (2001)
found that 57% of their group of high-functioning autistic adolescents scored within 1
SD of the mean of the control group on the WCST, and 29% performed better than the
control mean. Teunisse, Cools, van Spaendock, Aerts, and Berger (2001) found that
only 46% of their high-functioning adolescents with autism showed poor cognitive
shifting, defined using the lenient criterion of any positive z score on the sum of two
variables measuring the number of trials required for successful performance. Hughes
et al. (1994) reported that 67% of their autism group failed both the Tower of London
(ToL) and the Intradimensional, Extradimensional (IDED) set-shifting task (using an
arbitrary criterion of failure), with 92% failing the ToL and 75% the IDED task.
Ozonoff and McEvoy (1994) report that 41% of their autistic sample performed within
the normal range on the WCST (but none on the Tower of Hanoi); and Ozonoff and
Jensen (1999) found that up to 36% of their sample scored above the control mean on at
least one EF task, with only half the sample performing below the control mean on all
three tasks used.
So, while universality has been assessed using a variety of different methods
making it difficult to compare results across studies, it is apparent that EF deficits are
not universal among individuals with autism. Those who have examined the
characteristics of individuals who perform in the normal range on EF tasks have usually
found, as for ToM, that they are older (Ozonoff & Jensen, 1999) and/or have higher
verbal IQ (Liss et al., 2001; Ozonoff & McEvoy, 1994). However, unlike ToM, it
would be difficult to argue that these older and higher-functioning individuals have
developed some kind of compensatory strategy to aid their EF performance, as the very
nature of EF tasks is that they are novel and would not have been encountered before,
making strategies difficult to develop in advance (although it is conceivable that general
strategies may have developed or been learned to offset recognised limitations in certain
cognitive abilities, for example by using visual imagery to compensate for a verbal
working memory deficit). An alternative argument might be that these EF “passers”
46
could show difficulties on tasks tapping other components of EF which may not have
been measured in the studies reviewed (e.g., generativity). Hughes (2001) has
suggested that low- and high-functioning individuals with autism may show distinct
types of EF impairment, as indicated by the different types of repetitive behaviour
displayed by these two groups (Turner, 1997), and therefore that group heterogeneity
may prevent universal EF characteristics from being discovered. In addition, it may be
that different aspects of EF are impaired at different points in the development of
children with autism in comparison with age-matched controls, as EF components show
different developmental trajectories. These possibilities await empirical investigation.
ii) Uniqueness. A major challenge to the EF hypothesis of autism (which has
been acknowledged and discussed by all its proponents) is that EF impairments are
displayed in a number of other disorders, including ADHD (e.g., Grodzinsky &
Diamond, 1992; Oosterlaan, Logan, & Sergeant, 1998; Pennington, Groissier, & Welsh,
1993; Shallice et al., 2002) schizophrenia (e.g., Elliott, McKenna, Robbins, & Sahakian,
1995; Pantelis et al., 1997; see Hoff & Kremen, 2003), Tourette’s syndrome (Baron-
Cohen & Robertson, 1995; Channon, Flynn, & Robertson, 1992), obsessive-compulsive
disorder (Christensen, Kim, Dyksen, & Hoover, 1992; Cox, Fedio, & Rapoport, 1989;
Head, Bolton, & Hymas, 1989; Veale, Sahakian, Owen, & Marks, 1996), and early-
treated phenylketonuria (Diamond, Prevor, Callender, & Druin, 1997; Smith, Klim, &
Hanley, 2000; Welsh, Pennington, Ozonoff, Rouse, & McCabe, 1990), not to mention
neurological disorders affecting frontal lobe functions such as frontotemporal dementia
(e.g., Razani, Boone, Miller, Lee, & Sherman, 2001; see Grossman, 2002), Parkinson’s
disease (e.g., Owen et al., 1993), and traumatic brain injury (e.g., Anderson et al., 2002;
Levine et al., 1998). The question of how these symptomatically different disorders
could all share the same cognitive basis has been dubbed the “discriminant validity
problem” (Pennington & Ozonoff, 1996).
Proponents of the EF hypothesis of autism have attempted to circumvent this
difficulty meeting the uniqueness criterion by proposing that different disorders may be
characterised by different severity and age of onset of EF impairment, and different
profiles of impairment on the various components of EF (Ozonoff, 1997b; Ozonoff et
al., 1994; Pennington & Ozonoff, 1996). In particular, Ozonoff (1997b; Ozonoff &
Jensen, 1999) has suggested that autism is characterised by deficits in cognitive
flexibility and planning but spared inhibitory capacity. Studies by Ozonoff et al. (1994)
and Ozonoff and Jensen (1999) demonstrated evidence of differentiation among the EF
profiles of children with autism, ADHD and Tourette’s syndrome, with autistic children
47
showing poor performance on tests of cognitive flexibility and planning but not
inhibition, the ADHD sample characterised by a specific impairment in inhibitory
control, and individuals with Tourette’s showing little evidence of any EF impairment.
The notion of intact inhibition in autism has received support in other studies using a
negative priming paradigm (Brian, Tipper, Weaver, & Bryson, 2003; Ozonoff &
Strayer, 1997). A few studies have suggested that working memory may also be spared
in autism (Ozonoff & Strayer, 2001; Russell, Jarrold, & Henry, 1996).
However, while these studies appear to paint a relatively clean picture of spared
and impaired EFs in autism and other disorders, other findings have not been so
consistent. Some have argued for and/or found evidence of impairments in inhibition
(Hughes, 1996b; Rinehart, Bradshaw, Tonge, Brereton, & Bellgrove, 2002; Williams et
al., 2002) and in working memory (Bennetto et al., 1996; Pennington et al., 1997) in
individuals with autism, although it has been argued that these impairments emerge only
if the task involves both inhibitory and working memory requirements (Russell, 1997b).
Others have found set-shifting to be intact in high-functioning individuals with autism
and/or Asperger syndrome (Ozonoff et al., 2000; Rinehart, Bradshaw, Moss, Brereton,
& Tonge, 2001; Turner, 1997). The role of generativity impairments (e.g., Turner,
1999) has also been under-emphasised. In her review of EF in autism, Hughes (2001)
was able to say only that EF deficits in autism were “high-level and non-spatial” (p.
258), a characterisation which clearly lacks the desirable specificity. Research on EF in
other disorders is characterised by similar inconsistencies. For example, in their review
of EF in ADHD, Sergeant, Geurts, and Oosterlaan (2002) concluded that the pattern of
EF deficits in ADHD was not consistent across studies and did not appear to be specific
to ADHD.
A number of factors may have contributed to the lack of consistency in
identifying unique EF profiles across studies of EF in autism and other disorders.
Firstly, the usual concerns about the measurement precision of EF tasks will inevitably
affect the reliability and interpretation of results (particularly given that most studies
have used multifactorial tasks such as the WCST). Secondly, different research groups
have focused on different aspects of EF depending upon their theoretical framework
(Hughes, 2001), with very few studies including tasks measuring the full range of EF
components. Thirdly, the task modality and/or mode of response (i.e., verbal or non-
verbal) has varied across studies, which may be important as spatial ability is superior to
verbal ability in autism (Happé & Frith, 1996), and this superior spatial ability may
boost performance on non-verbal EF tasks as compared with verbal tasks (Hughes,
48
2001). Fourthly, it has been suggested that some of the inconsistencies across studies
may be related to whether or not the tasks were computerised, and therefore whether
performance may have been affected by the need for social interaction; and also,
whether or not feedback about task performance is given verbally by the examiner or is
provided automatically by the task (Ozonoff, 1995b, 2001). Finally, as discussed
earlier, variability in the age and level of functioning (i.e., intellectual ability) of the
sample may also influence the pattern of results12. Clearly, studies using a wide range
of well-defined EF tasks, using both verbal and non-verbal response modes, and which
may be broken down into component processes, are needed to determine whether
autism is associated with a unique EF profile.
iii) Causal precedence. This criterion of primacy has also posed difficulties for
the EF hypothesis. The first study to test EF in young children with autism (McEvoy,
Rogers, & Pennington, 1993) found that autistic preschoolers (mean age = 5.1 years)
made significantly more perseverative errors on a spatial reversal task tapping inhibition
and set-shifting (but not A-not-B, Delayed Response or Delayed Alternation tasks,
which showed ceiling and floor effects). Similarly, Dawson, Meltzoff, Osterling, and
Rinaldi (1998) found that young children with autism (mean age = 5.4 years) performed
more poorly than controls on a version of the A-not-B task. However, no other studies
have managed to replicate these findings. Using a younger sample than McEvoy et al.
(1993), Wehner and Rogers (1994, cited in Pennington et al., 1997) found no difference
between their autism and control groups on the same spatial reversal task. To examine
whether this may be because perseverative behaviour increases as children with autism
grow older (contrary to typical development), Griffith, Pennington, Wehner, and Rogers
(1999) conducted a longitudinal study investigating the development of EF over the
course of a year. Besides finding no difference in performance between young children
with autism (mean age = 4.3) and developmentally delayed controls on a range of age-
appropriate EF tasks (measuring inhibition, set-shifting, spatial and object working
memory, and action monitoring) upon initial testing, they found that performance on the
spatial reversal task at the second testing period did not change significantly over time,
suggesting that young children with autism do not exhibit a deficit on this task at either
12 Besides the possibility that low- and high-functioning children with autism may actually exhibit different profiles of EF impairment, Russell (1997b) also points out that different findings in these two groups may be due to the different comparison groups used. If the control group consists of developmentally delayed or mentally retarded children who also display EF impairments (in comparison to typically developing children), then low-functioning children with autism may not be impaired compared with their IQ-matched control group. However, high-functioning children with autism may display an impairment in comparison with their more typically developing controls.
49
4 or 5 years of age. More recent studies by Dawson et al. (2002a), Stahl and Pry (2002),
and Rutherford and Rogers (2003) have also failed to find EF deficits in young children
with autism using the spatial reversal task and other measures of inhibition, working
memory, and set-shifting, although Rutherford and Rogers (2003) reported marginal
differences on their measure of generativity.
The apparently intact EF abilities of young children with autism suggest that
executive dysfunction cannot account for the earliest symptoms of autism. However,
two possible explanations for these null findings have been proposed (Griffith et al.,
1999; Pennington et al., 1997). Firstly, most of the above studies have used children
with developmental delays as controls, and have found that these children perform more
poorly on the EF tasks than typically developing children. Hence, it may be that young
children with autism are impaired on the EF tasks, but this impairment is not specific to
autism. A second explanation, which is more favourable to the EF hypothesis of
autism, is that EF impairments have been missed because the studies have not
incorporated the full range of EF components or task modalities in their test batteries.
Because of the age of the children, tasks have been primarily non-verbal (which may
favour children with autism); and in addition, measures of planning and generativity
have not been included, probably because of the lack of age-appropriate measures of
these abilities. Rutherford and Rogers’ (2003) finding of marginal differences in
generativity in young children with autism indicates that this explanation may hold
promise. In addition, Russell and colleagues (Biro & Russell, 2001; Russell, 1997b;
Russell, Jarrold, & Hood, 1999) have proposed that children with autism are only
challenged by EF tasks if they contain arbitrary rules (which require the online rehearsal
of novel strategies), which most EF tests used with very young children (such as the A-
not-B task) do not contain. As usual, however, further studies are needed to investigate
all of these possibilities.
iv) Explanatory value. Unlike the ToM hypothesis, the ability of the EF
hypothesis to account for the range of autistic symptomatology is perhaps its greatest
strength. While executive dysfunction was initially proposed largely to explain the
repetitive behaviours and restricted interests characteristic of autism, it has also fairly
consistently demonstrated links with social and communicative impairments. Two
studies by Berger and colleagues (Berger, Aerts, van Spaendonck, Cools & Teunisse,
2003; Berger et al., 1993) found that set-shifting performance was a significant
predictor of social understanding and social competence in high-functioning adolescents
and young adults with autism, although the same group was not able to replicate that
50
result in a third study (Teunisse et al., 2001). Gilotty, Kenworthy, Sirian, Black, and
Wagner (2002) found significant correlations in their autistic sample between parental
reports of everyday executive abilities (using the Behavior Rating Inventory of
Executive Function) and social and communication skills as measured by the Vineland
Adaptive Behavior Scales, such that impaired EF was associated with poorer adaptive
skills. McEvoy et al. (1993) also found a significant correlation between EF and early
social and communication skills in young children with autism. Liss et al. (2001) found
that relationships between EF and adaptive functioning were no longer significant when
VIQ was partialled out; however, their study was also inconsistent with other studies in
finding that autism versus control group differences on EF tasks disappeared when VIQ
was accounted for.
A number of studies have examined the relationship between EF deficits and
joint attention, as one of the earliest signs of social impairment in young children with
autism13. It has been proposed that difficulties in shifting attention, rather than
mentalising impairment, may underlie problems with joint attention in young autistic
children (Burack, 1994; Courchesne et al., 1994). This proposal received support in
McEvoy et al.’s (1993) study, which found that the frequency of joint attention
behaviours was significantly correlated with cognitive flexibility. However, those
studies which failed to find EF impairments in young children with autism have
generally also not found any correlation between EF and joint attention. Stahl and Pry
(2002) and Rutherford and Rogers (2003) both found no relationship between EF
measures and joint attention in their young autistic sample. Dawson and colleagues
(Dawson et al., 1998, 2002a) found that performance on tasks purportedly tapping
ventromedial prefrontal and medial temporal function (e.g., the delayed nonmatching to
sample task) correlated with joint attention, but not performance on tasks tapping
dorsolateral prefrontal function (i.e., more classic EF tasks such as the A not B and
spatial reversal tasks). These findings of no relationship between EF and early signs of
social impairment are likely to relate to the null findings of group differences in EF at
that young age. If young children with autism do not show impairment on EF tasks, one
can hardly argue that EF deficits underlie their abnormal joint attention behaviours.
Indeed, Swettenham et al. (1998) found that infants with autism (mean age = 20
13 While joint attention is clearly a social behaviour, it is also often interpreted as a precursor to or marker of ToM (Baron-Cohen, 1995). The relationship between joint attention and EF is therefore also relevant to the upcoming section on the relationship between ToM and EF in autism. Similarly, pretend play is thought to reflect metarepresentative capacity and so its relationship with EF is also of relevance to that section.
51
months) showed more difficulty shifting attention between people than between objects,
suggesting that their impairment may lie in social orientation rather than attentional
shifting. Nevertheless, the extent to which EF and joint attention may be related in
autism remains a matter of debate (Hughes, 2001).
Several studies have also examined the link between EF and the absence of
spontaneous pretend play in autism. As for joint attention, these investigations have
been spurred by suggestions that a lack of pretend play may have its basis in EF deficits,
specifically an impairment in generativity (Jarrold, Boucher, & Smith, 1994a, 1996;
Lewis & Boucher, 1995) or the inability to disengage attention from salient external
stimuli to access internal, hypothetical play schemas (Harris, 1993; Harris & Leevers,
2000) rather than an inability to mentalise. Observations that children with autism have
intact comprehension of pretend acts (Jarrold, Smith, Boucher, & Harris, 1994b) and
that they are able to produce structured, elicited, or instructed pretence (Lewis &
Boucher, 1988) are consistent with this view. In a series of studies, Jarrold et al. (1996)
showed that children with autism have the capacity to engage in the mechanics of
pretence, but that they produced significantly less pretence than controls in spontaneous
and weakly structured conditions, suggesting that their difficulty lay in the production
or generation of pretence. Similarly, Lewis and Boucher (1995) found that the
generation of original actions in the play of autistic children was more consistent with a
generativity hypothesis than a metarepresentational deficit. In a more recent study,
Rutherford and Rogers (2003) found that the performance of children with autism on a
generativity task was a significant predictor of pretend play.
Surprisingly, although the usual behavioural consequences of EF or prefrontal
impairment correspond closely with the repetitive behaviours displayed by individuals
with autism, only one published study has directly examined the association between EF
and repetitive behaviours in autism. Turner (1997) found significant correlations in
children with autism between measures of inhibition, set-shifting, and generativity and
the incidence and severity of various aspects of repetitive behaviour (e.g., repetitive
movements, circumscribed interests) as measured by a parental interview. Furthermore,
specific EF components appeared to underlie different types of repetitive behaviour; for
example, repetitive movements were associated with performance on a test of
inhibition, whereas sameness behaviour was correlated with measures of generativity.
The EF hypothesis has therefore demonstrated good explanatory value in terms
of its ability to account for both social and communicative impairments and repetitive
behaviours in autism, with the exception of early signs of social impairment such as
52
joint attention. Like the ToM hypothesis, it is less able to account for non-triad features
of autism such as savant abilities and heightened visuospatial and visuoperceptual skills,
a fact which has been somewhat overlooked in the EF literature. The causal direction of
correlations between EF and behavioural symptoms is another issue to consider, as it
may be that executive dysfunction is a consequence of fewer social interactions or
engagement in high rates of repetitive, restricted activities, rather than vice versa.
However, evidence does not appear to support this possibility. In a number of studies,
increasing the structure of the environment has been found to result in less stereotypic
and more social behaviours (Clark & Rutter, 1981; Dadds et al., 1988), indicating that
reducing EF demands facilitates social interaction and reduces repetitive behaviour,
which would not be expected if EF impairment was caused by the behavioural
symptoms. Also, the results of longitudinal studies have shown that EF performance
predicts later social understanding (Berger et al., 1993, 2003).
So how does the EF hypothesis fare? Like ToM, while it has defended some of its
weaknesses fairly successfully, it does not convincingly meet all of the criteria for a
single primary cognitive deficit of autism. Although it holds good explanatory value for
most of the symptoms of autism (with the exception of some early symptomatology and
non-triad features), the evidence collected so far suggests it lacks causal precedence and
that EF deficits are not universal among individuals with autism. In addition, the
variability of findings among studies of EF in autism is problematic. Methodological
issues of the measurement precision of EF tasks, the different developmental trajectories
of the various EF components, the difficulty in designing age-appropriate EF tasks for
young children which tap the range of EF abilities, the variability among studies in the
age and level of functioning of the sample, and variations in the modality, arbitrariness
of the rules, and mode of presentation, response, and feedback of the tasks used, have
all clouded the definition of the universality, specific profile, and developmental course
of EF deficits in autism. While it is fairly clear that autism is characterised by
significant EF deficits, these methodological issues need to be addressed in order to
determine how primary those deficits may be to autism.
53
2.3 The ToM-EF relationship
2.3.1 Models of the ToM-EF relationship
On the surface, there is no particular reason to propose a link between the constructs of
ToM and EF: why should the ability to attribute mental states to oneself and others
relate to cognitive capacities which aid the control of action? In fact, an accumulating
number of recent studies have consistently demonstrated an empirical relationship in
typical development; for example, there are strong correlations between various types of
ToM and EF tasks which remain significant when age and IQ variables are partialled
out (Carlson & Moses, 2001; Carlson, Moses, & Breton, 2002; Davis & Pratt, 1995;
Frye et al., 1995; Gordon & Olson, 1998; Hala, Hug, & Henderson, 2003; Hughes,
1998a, 1998b; Lang & Perner, 2002; Russell et al., 1991; see also Perner & Lang, 1999,
who report a large average effect size of 1.08 across the studies conducted up until
then). Marked improvements in ToM and in EF (particularly inhibitory control) both
occur around the same age, in the preschool period between 3 and 5 years of age (e.g.,
Gerstadt et al., 1994; Kochanska et al., 1997; Wellman et al., 2001; Zelazo et al.,
1996b). The co-occurrence of ToM and EF deficits not only in autism but also in
schizophrenia (e.g., Corcoran et al., 1995; Elliott et al., 1995), frontal lobe pathologies
(Bach et al., 1998; Channon & Crawford, 2000; Gregory et al., 2002; Rowe, Bullock,
Polkey, & Morris, 2001; Saltzman et al., 2000), and possibly Fragile X syndrome
(Garner, Callias, & Turk, 1999) is also suggestive of a meaningful relationship. A range
of explanations for this observed relationship have been proposed by various authors,
including links based on i) the EF requirements of ToM tasks (“expression” accounts),
ii) a third common conceptual requirement, iii) functional dependence during
development (“emergence” accounts), and iv) shared neuroanatomical bases. These
classes of explanation each have different implications for the nature of the relationship
between ToM and EF impairments in autism (these implications are reviewed in Section
2.3.2 and are important in the interpretation of the results of analyses of the ToM-EF
relationship in Study One). As such, a review of each follows.
54
2.3.1.1 Expression accounts14
This type of account holds that the relationships observed between performances on
ToM and EF tasks are (at least partly) due to the executive requirements of ToM tasks,
and therefore that failure on ToM tasks may be caused by impaired or underdeveloped
EF rather than (or in addition to) poor mentalising ability. In other words, EF might
affect the expression of ToM capacity. Proponents of this account have emphasised
either inhibitory control (Carlson & Moses, 2001; Carlson, Moses, & Hix, 1998; Hala &
Russell, 2001; Leslie & Polizzi, 1998; Roth & Leslie, 1998; Russell et al., 1991),
working memory (Davis & Pratt, 1995; Gordon & Olson, 1998; Keenan, 1998), or a
combination of both inhibition and working memory (Carlson et al., 2002; Hala et al.,
2003) as the crucial EF factors affecting ToM performance. These ideas have been
tested both by manipulating the EF requirements of various ToM tasks and by
examining correlations or predictive relationships between the relevant EF components
and ToM variables.
The idea that ToM tasks require inhibitory control has been advanced by several
authors, although there has been some disagreement regarding exactly what it is that
needs to be inhibited. Across a series of studies, Russell and colleagues (Hala &
Russell, 2001; Russell et al., 1991; Russell, Jarrold, & Potel, 1994) have argued that
knowledge of current physical reality is more salient than knowledge of mental reality,
and that tests of both deception and false belief require children to suppress or inhibit
responding on the basis of their physical knowledge in favour of their less salient mental
knowledge15. For example, it is proposed that in the standard false belief (unexpected
transfer) task, the child is required to disengage from (and inhibit the prepotent response
to report) his/her knowledge about where the object is currently located, and instead
refer to an empty location.
This hypothesis has been tested mainly by using a measure of strategic deception
called the “windows task”. Russell et al. (1991) found that despite prior training (using
14 Perner and Lang (1999, 2000) label this class of explanation slightly differently, as “Executive component in ToM tests”. As the notion of common task requirements refers to the idea that common underlying processes are required in both tasks, the term “expression account” is preferred here, as this more accurately encompasses the notion that EFs may influence the expression of ToM capacity in everyday life (due to the executive requirements of perceiving and inferring others’ mental states) as well as on structured tasks. 15 It should be noted that while Russell (1996, 1997b) argues that most ToM tasks confound EF and mentalising demands, his view of the ToM-EF relationship is not actually that it may be explained entirely because of common performance requirements. His main theory is reviewed in this section under the heading of “emergence accounts”.
55
two opaque boxes) on the rules of the task whereby they had to point to an empty
location to prevent their opponent from winning a chocolate, 3-year-olds typically
pointed to the true location of the chocolate on test trials, where they were able to see
the chocolate but the opponent could not. Furthermore, the majority of the 3-year-olds
persisted in revealing its true location across a series of 20 trials, suggesting that a
failure of inhibition or inability to disengage from a salient stimulus was underlying
their difficulty, rather than a conceptual deficit with deception (and thus ToM). This
interpretation of the results was supported in two further studies. Russell et al. (1994)
found that removing the opponent (and therefore the requirement to deceive) did not
affect 3-year-olds’ performance on the windows task. Conversely, Hala and Russell
(2001) found that the performance of 3-year-olds improved when the inhibitory
demands of the task were reduced, such as by removing the requirement to directly
point to the chocolate and instead using a mechanical pointer to indicate the appropriate
response (as pointing correctly to true locations is likely to be a well-practiced,
reinforced and therefore prepotent response). Using a different approach, Moore et al.
(1995) found that when their own desires were particularly strong or salient, 3-year-olds
performed as poorly on a conflicting desire task as on a false belief task. This suggests
that even though desire is purportedly easier for young children to understand because
of its non-representational nature, 3-year-olds have difficulty judging others’ desires
when EF demands are high. Together, these results suggest that the failure of young
children on ToM tasks may be at least partially attributable to inadequate inhibitory
control rather than (or in addition to) a poorly developed ToM.
Carlson, Moses and colleagues (Carlson & Moses, 2001; Carlson et al., 1998)
have outlined a similar account of the role of inhibition in ToM performance. Like
Russell, while they allow for genuine development in the understanding of mental
concepts, they argue that the inhibitory requirements of ToM tasks affect the expression
of ToM ability in 3-year-olds16. Using a similar deception paradigm as Hala and
Russell (2001), Carlson et al. (1998) found that 3-year-olds showed improved
performance under conditions requiring low inhibitory control (i.e., when pictorial cues
or arrows were used to mislead the opponent rather than pointing), and that they were
equally successful in using arrows to point whether the opponent was present or not. In
a correlative study, Carlson and Moses (2001) found that the link with inhibition
16 Carlson and Moses (2001) also state that their results are equally compatible with an expression account (i.e., that inhibitory dysfunction impedes the expression of ToM ability) as with Russell’s (1996, 1997b) emergence account, to be reviewed later as mentioned in the previous footnote.
56
extended beyond deception, finding that performance on a battery of ToM tasks was
significantly correlated with a number of indices of inhibitory control, and that these
correlations remained robust after the effects of age, gender, number of siblings, verbal
ability, and a number of other cognitive abilities were removed. Other studies have also
indicated that like deception, young children’s performance on the standard false belief
task improves when inhibitory demands are reduced by using a response mode that is
not influenced by a prepotent response history or by reducing the salience of the desired
object’s current location. Examples include tracing out the path a naive character would
take in searching for their desired object (Freeman, Lewis, & Doherty, 1991), giving an
explanation for a protagonist’s wrong search in an empty location (Bartsch & Wellman,
1989), and indicating which of two twin boys, one searching in the actual location and
one in the empty location, had been absent during the transfer of the object (Robinson &
Mitchell, 1995).
In line with their ToMM-SP model (described in Section 2.1.2), a more
unequivocal expression account of the relationship between inhibitory processes and
ToM has been proposed by Leslie and colleagues (Leslie & Polizzi, 1998; Roth &
Leslie, 1998). Besides differing from the above two accounts on the level of ToM
ability attributed to 3-year-olds (Leslie argues that the ToM module is fully active by
this age but that its abilities are usually masked by the processing requirements of ToM
tasks, whereas Russell and Carlson and colleagues favour the view that some conceptual
ToM development does take place between the ages of 3 and 4), Leslie and his
colleagues also have a different view of what creates the salience-related difficulty for
3-year-olds on the standard false belief task. They argue that because beliefs are
typically true, there is a default (or prepotent) assumption that beliefs are true, and
therefore the attribution of a non-default (false) belief requires inhibition of this default
assumption (Leslie, 1994a; Leslie & Polizzi, 1998). Thus, they identify the competition
as being between two belief contents (one of which represents physical reality) rather
than between physical and mental realities (notably, Leslie and colleagues do not offer
an analysis of the inhibitory requirements of other ToM tasks such as tests of
deception). In support of their hypothesis, Leslie and colleagues conducted a series of
cleverly designed experiments which supported previous findings that reducing the
inhibitory demands of false belief tasks improves the performance of 3-year-olds (Roth
& Leslie, 1998), but also showed that increasing the inhibitory requirements of the false
belief task had a significant detrimental effect on the performance of 4-year-olds (Leslie
& Polizzi, 1998).
57
Inhibitory control is not the only executive process that has been implicated in
ToM task performance. Olson (1989) argued that developments in children’s capacity
for holding complex representations in mind may support or underlie their
understanding of false belief. Similarly, Halford (1993) proposed that working memory
capacity may limit young children’s success in situations which require the
simultaneous integration of two representations of a situation (i.e., reality and belief).
In a test of this hypothesis, Davis and Pratt (1995) found that backward digit span
performance significantly predicted scores on the unexpected contents and appearance-
reality tasks over and above age and verbal ability (accounting for around 6% of the
variance), but forward digit span did not. They interpreted this as suggesting that
development in the central executive, but not articulatory loop, component of Baddeley
and Hitch’s (1974) working memory model was a small but significant determinant of
false belief task performance. Using an additional false belief task (the unexpected
transfer test) and a more age-appropriate working memory measure involving dual-task
performance, Keenan, Olson, and Marini (1998) also found that after controlling for
age, working memory capacity was a significant predictor of false belief performance
(accounting for 7.4% of the variance). The influence of working memory capacity on
the expression of ToM ability is supported in other studies showing that reducing the
usual memory demands of ToM tasks has a facilitative effect on performance (e.g.,
Chandler & Hala, 1994; Freeman & Lacohée, 1995; Mitchell & Lacohée, 1991;
although see Hala et al., 2003; Robinson, Riggs, & Samuels, 1996).
While these studies permit the fairly acceptable conclusion that working
memory plays some limiting role in the expression of mentalistic concepts that may
already exist, Gordon and Olson (1998) considered the more contentious possibility that
increasing computational resources may actually allow the formation of those concepts.
They argued that the key capacity required for false belief understanding is the ability to
hold in mind and then update a previously created representation when a new
representation is created by the current perceptual situation. They used two working
memory tasks, both of which required children to perform concurrent mental activities,
but only one of which required them to hold the product of such activity in mind such
that it could be updated on the basis of some new perceptual information. While both
their tasks showed strong correlations with false belief performance after controlling for
age (accounting for up to 40% of the variance), the more complex working memory task
contributed a significant amount of variance to false belief performance over and above
the other more simple working memory task. Gordon and Olson concluded that while
58
primitive concepts such as self, true, and real may be available earlier, “their co-
ordination into a higher-order structure depends upon increased computational
resources” and thus that “conceptual content and conceptual complexity combine not
only in the performance on theory of mind tasks but also for the formation of the
understanding itself” (1998, p. 81)17.
Two studies found that the relationship between ToM and working memory no
longer held after age and verbal ability were controlled for (Hughes, 1998a; Jenkins &
Astington, 1996). Hala et al. (2003) proposed that this discrepancy may be attributable
to the lack of significant inhibitory demands in the working memory tasks used in these
two studies (i.e., they were simple tests of maintenance of information in working
memory over time and did not require dual-task performance or the simultaneous
activation of two concurrent activities). This interpretation is supported by Davis and
Pratt’s (1995) finding that forward digit span was not a significant predictor of false
belief performance, in contrast to backward digit span (which arguably involves not
only rehearsing the sequence of numbers but also inhibiting the tendency to report them
in the order heard).
The idea that EF tasks involving both working memory and inhibitory
components may show the strongest relationship with ToM had been raised earlier by
Carlson and colleagues (Carlson & Moses, 2001; Carlson et al., 2002). They argued
that false belief tasks require both working memory and inhibition in that the child must
hold in mind two representations simultaneously as well as make a response based on
the representation which directly conflicts with his/her own salient perspective.
Although this group earlier favoured the view that the ToM-EF relationship was based
purely on the inhibitory requirements of ToM tasks, their shift in view was prompted by
two separate studies in which they found that of two types of inhibition task, the type
which involved a heavier working memory load (whereby two conflicting alternatives
needed to be held in mind) was the more powerful predictor of ToM performance, and
added extra variance over and above the low working memory load inhibition task
(Carlson & Moses, 2001; Carlson et al., 2002). In addition, Carlson et al. (2002) found
that a working memory task with no inhibitory requirement did not predict false belief
performance independently of age and both verbal and non-verbal intelligence. Perner,
Lang, and Kloo (2002b) also failed to find a significant relationship between ToM and
inhibition when a simple go-nogo inhibition task with low working memory load was
17 Again, this view is therefore probably best conceived of as a combination of expression and emergence accounts of the ToM-EF relationship.
59
used. Consistent with these results, a recent study by Hala et al. (2003) found that
“pure” measures of inhibition and working memory did not predict false belief
performance individually, but tasks combining inhibitory and working memory
requirements were strongly predictive of false belief performance after age and verbal
ability were controlled.
Only a few studies have examined the contribution of EF components other than
inhibition and working memory to ToM performance, with mixed results. Hughes
(1998a) found a significant correlation between tests of attentional flexibility (or set-
shifting) and deception after age and verbal and non-verbal intelligence were partialled
out. However, flexibility did not correlate with false belief performance and the
correlations with deception were not as robust as those with performance on inhibition
tasks. Harris (1993) argued that in both ToM tasks and tests of planning, children must
envisage a hypothetical state of affairs and respond or make a prediction in accordance
with that hypothetical situation, which typically contradicts the response which is
suggested by the true or current state of affairs. According to Harris, children will
exhibit difficulty on both types of task if they are unable to shift or disengage from their
current context to a hypothetical and conflicting context. However, he does not present
any evidence specifically examining this hypothesis. Using a planning test developed
for use with young children, Bischof-Köhler (1998, cited in Perner & Lang, 2000) found
a relationship between planning ability and false belief performance, but the direction of
the relationship was such that false belief understanding appeared to be necessary for
planning success. However, the effect of age or verbal ability on this relationship was
not reported. Moses and Carlson (2000, cited in Carlson et al., 2002) did not find a
significant relationship between planning ability and ToM after age and verbal ability
were partialled out. It therefore remains unclear whether EF components other than
inhibition and working memory are correlated with ToM, and if so, what might underlie
these relationships.
Overall, though, the evidence appears to be consistent with the idea that the EF
requirements of ToM tasks or abilities are a significant factor influencing the
developmental relationship between ToM and EF. However, a number of criticisms of
expression accounts have been advanced by Perner and colleagues (Perner, 1995, 2000;
Perner & Lang, 2000; Perner et al., 2002b), whose central contention is that ToM-EF
correlations are not solely attributable to task requirements (and, relatedly, that the
60
preschool development in ToM is not solely attributable to developments in EF)18.
While they more readily accept the evidence suggesting that tests of deceptive pointing
include a significant executive component, Perner and colleagues have questioned the
methodology of several of the studies purporting to demonstrate earlier competence on
false belief tasks when the EF demands are reduced. For example, Perner (1995; Perner
et al., 2002b) argued that in Bartsch and Wellman’s (1989) study, equal numbers of
children passed the explanation version of the false belief task (which does not include
an obvious inhibitory requirement) as the standard prediction task; and that it was only
after receiving an overly helpful prompt that they displayed additional correct answers
on the explanation version. Other studies using alternative explanation paradigms have
found that explanation tasks are equally as difficult as prediction tasks (Hughes, 1998a;
Moses & Flavell, 1990; Perner et al., 2002b; Wimmer & Mayringer, 1998), although
Russell, Hill, and Franco (2001) have pointed out that the mean age in these studies was
around 4 years (compared with 3 years in Bartsch & Wellman’s study), which may have
masked differences by boosting scores on the prediction task. Perner (1995) also argued
that Robinson and Mitchell’s (1995) finding of significantly improved performance on
their identical twin explanation paradigm compared with a standard prediction version
may be explained by a difference in the baseline performance expected for children with
no understanding, and that there is no difference in difficulty once the data are adjusted
for correct guesses – a post-hoc analysis which was then confirmed by the pattern of
results obtained by Perner et al. (2002b).
While these findings suggest that modified false belief (“explanation”) tasks
which purportedly remove the EF component may be just as difficult as standard false
belief tasks (see also Robinson & Beck, 2000), even more pertinent for Perner are
studies showing that performance on explanation versions correlates just as strongly
with EF scores as performance on standard prediction versions. Hughes (1998a) found
that performance on her explanation version correlated as strongly as performance on a
standard false belief task with scores on tests of inhibitory control, and Perner et al.
(2002b) found that performance on their explanation version correlated as strongly as
performance on a prediction version with scores on the dimensional change card sorting
task (which arguably requires set-shifting and inhibition).
These results suggest that the ToM-EF relationship is not solely due to the EF
requirements of ToM tasks. However, this conclusion rests on the assumption that
18 Perner’s own theory of the ToM-EF relationship is reviewed in Section 2.3.1.3, under the heading of “Emergence accounts”.
61
explanation versions of false belief tasks do not incorporate any EF requirements.
Russell et al. (2001) argue that tasks requiring a linguistic explanation for why an empty
location was visited still require children to set aside or inhibit their knowledge of the
actual location. It could also be argued that other explanation versions still require the
child to hold in mind two conflicting perspectives, thereby taxing working memory.
For example, in the identical twin versions used by Robinson and Mitchell (1995) and
Perner et al. (2002b), the child is still required to hold in mind the sequence of events
that has occurred and consider the different experiences of both twins simultaneously in
order to work out why one twin looks in the wrong location. In addition, in defence of
the expression accounts, it should be noted that the majority of authors who argue that
EF abilities constrain performance on ToM tasks do not hold that the ToM-EF
relationship is solely due to performance-based factors, that there is no deeper
conceptual link, or that the development in mentalistic understanding in the preschool
period is attributable only to increasing EF capacity without any additional conceptual
development (Leslie and colleagues are an obvious exception). Certainly, none of the
authors have made the claim (sometimes attributed to them) that mentalising ability
does not exist and that ToM tasks are simply EF tasks. On the basis of the evidence as a
whole, it seems reasonable to accept that successful performance on some ToM tasks
requires a certain level of capacity in EF (particularly inhibition and working memory)
and that executive difficulties may impact upon ToM performance, and therefore that
the correlations observed between ToM and EF are partially due to performance-based
commonalities.
2.3.1.2 Common conceptual requirements of ToM and EF
Rather than positing that the ToM-EF relationship arises from the EF requirements for
successful ToM performance, this account contends that both ToM and EF share a third
common underlying conceptual requirement. The main account falling in this category
is the CCC theory (see Section 2.2.2; Frye, Zelazo, & Burack, 1998; Frye et al., 1995),
which proposes that false belief and EF tasks both require the use of embedded
conditionals, or if-if-then rules (an example of a task structure involving embedded
conditionals, the Dimensional Change Card Sorting (DCCS) task, is described in
Section 2.2.2). Frye’s (2000) analysis of the logical structure of the standard false belief
task (where the child must predict where Maxi will look for his chocolate) in terms of
if-if-then embedded conditionals runs as follows:
62
IF me (s1), IF looking for chocolate (a1), THEN here (c1).
IF Maxi (s2), IF looking for chocolate (a1), THEN there (c2).
Frye (1999) proposed that while ToM and EF are distinct and neither underlies the
other, they are related in that they depend on different applications of the same set of
reasoning rules, and the development in this reasoning ability underlies the development
of ToM and EF at the same age. The same embedded rules “guide the inferences
necessary for theory of mind and allow the formulation of action that results in
improved executive control” (Frye, 1999, p. 121-122).
Frye et al. (1995) tested the CCC account by comparing preschoolers’
performance on three false belief tasks and two reasoning tasks with an embedded
conditional structure: the DCCS task and a physical causality task where a marble was
rolled down a covered ramp either to a hole directly below its entry point or across to
the opposite side, depending on the setting condition, and children had to predict where
the marble would be found. Frye et al. found similar age-related improvements
(between 3 and 5 years of age) across both types of task. In a further study, Frye,
Zelazo, Brooks, and Samuels (1996) showed that 3-year-olds were able to successfully
perform a simplified version of the physical causality task with a simple if-then
structure. Frye et al. (1995) also found that scores on the reasoning and ToM tasks were
significantly correlated with age partialled out. Furthermore, ToM performance only
correlated with performance on reasoning tasks with an embedded rule structure, and
not with performance on tasks with simple if-then structures.
A number of criticisms of CCC theory and its explanation of the ToM-EF
relationship have been advanced. Carlson and colleagues (Carlson et al., 1998; Carlson
& Moses, 2001) argue that Frye et al.’s (1995) data are also consistent with an
inhibitory control interpretation, as, for example, the DCCS task requires children to
inhibit their previous way of responding and shift to a new response. Zelazo and Frye
(1998) refute the inhibition interpretation by pointing to data showing that 3-year-olds
are able to effectively inhibit a previous way of responding when the task conforms to a
simple if-then structure (Marcovitch, Zelazo, Boseovski, & Cohen, 1997, cited in
Zelazo & Frye, 1998) and that on a task with an embedded rule structure, 3-year-olds
still performed poorly when evaluating the sorting of a puppet – that is, when they were
not themselves required to inhibit a previous response (Jacques, Zelazo, Kirkham, &
Semcesen, 1999). However, conversely, Carlson et al. (1998) highlight the fact that the
deceptive pointing and arrow tasks used in their study had identical rule structures, but
3-year-olds’ performance was significantly better on the arrow task (which had a lower
63
inhibitory requirement). Perner and Lang (2002) also found that 3-year-olds were able
to perform well on variations of the DCCS task which had an embedded rule structure,
but which did not include an extradimensional shift (i.e., the rule reversed rather than
changed dimensions from colour to shape) and did not involve a visual clash between
target and test cards (i.e., had reduced inhibitory requirements). Moreover, Perner
(2000) calls attention to the fact that go-nogo tasks (which require a simple pair of
rules) are as difficult for 3-year-olds as other inhibition and conditional reasoning tasks.
Perhaps even more pertinent is Carlson and Moses’ (2001) finding that one of their
inhibition measures which had a simple if-then structure was one of the strongest
predictors of ToM performance. Similarly, Sabbagh, Moses, and Shiverick (2001, cited
in Carlson & Moses, 2001) found that false belief performance was strongly correlated
with inhibition, but performance on the false photograph task (which has an identical
rule structure) was not.
These data suggest that similarities in the rule structure of ToM and EF tasks
cannot account entirely for the ToM-EF relationship. Perner, Stummer, and Lang
(1999) also present a more a priori argument against the CCC account’s analysis of the
standard false belief task. They point out that in the DCCS and physical causality tasks,
the conditional structures describe rules which the child must know in order to solve the
task. However, in the case of the false belief task, the conditional rules (e.g., “If Maxi,
if looking for chocolate, then here”) are not part of the task’s instructions and cannot be
those explicitly used by the child in solving the task, as such a rule would only be
possible if the child was repeatedly exposed to Maxi going to the empty location.
Perner and colleagues argue that this highlights the arbitrary nature of the rules chosen
to describe the false belief task. They offer the following analysis:
IF I am looking for the chocolate (a1), THEN here (c1).
IF Maxi is looking for the chocolate (a2), THEN there (c2).
This plausible alternative reduces the task to one with a pair of simple if-then rules,
which 3-year-olds should be capable of performing successfully. Zelazo, Jacques,
Burack, and Frye (2002) attempted to refute these criticisms by arguing that their claim
is not that people must learn the rules, but must formulate them in an impromptu
manner in order to solve the task. They argue that their analysis of the rule structure of
the false belief task is not logically necessary, but is an empirical claim which is
confirmed by the correlations observed between ToM and rule-based reasoning tasks.
However, this would mean that any task showing correlations with the rule-based
reasoning tasks could then be considered to have an embedded conditional structure -
64
surely a circular and ad hoc argument. Hence, difficulties with the logical defence of
the conditional structure of false belief tasks, as well as an inability to account for data
regarding both the abilities of young children and correlations between ToM and EF
tasks with simple rule structures, present a significant challenge to the CCC account of
the ToM-EF relationship.
An alternative “common conceptual requirements” account has been presented
by Halford and colleagues (Halford, 1993; Halford et al., 1998), although this has not
been as thoroughly investigated or discussed as the CCC account. Halford’s theory
states that processing capacity (or working memory) is limited by the complexity of the
relations (i.e., the number of related dimensions or sources of variation) that may be
processed in parallel, and that as processing capacity develops, children should be able
to represent concepts of increasingly higher relational complexity. He proposes that
young children’s difficulty on standard false belief and appearance-reality tasks may be
explained by their inability to represent “ternary” relations, or problems with three
related dimensions. His analysis of the standard false belief task is that it requires
representing the relation between an object and two different representations of its
location: one based on knowledge of its actual location and the other on a false belief of
its location. He expresses this situation as the ternary relation:
Find-object (<known-event>, <actual-location>, <believed-location>),
instances of which are:
Find-object (<saw-moved>, <object-in-location-A>, <believe-object-in-location-A>)
and
Find-object (<not-seen-moved>, <object-in-location-A>, <believe-object-in-location-
B>).
Halford argues that young children are able to understand any of the component binary
relations (e.g., Find-object (<not-seen-moved>, <object-in-location-A>)), but that they
cannot integrate two object-percept relations into a single ternary relation. This also
explains their poor performance on other kinds of task, including EF tests such as the
Tower of London, which require the same or a higher degree of relational complexity
(the Tower of London is described in the next chapter, Section 3.4.1).
Halford’s view is therefore similar to the CCC account, but differs in that it
emphasises the number of relations between pieces of information rather than the
presence of a hierarchical or embedded conditional structure. Because of this, Halford’s
proposal may be more resistant to some of the criticisms levelled against the CCC
account on the basis of the purported embedded rule structure of the false belief task.
65
However, his account has yet to be directly tested, although evidence of a relationship
between working memory and ToM (Davis & Pratt, 1995; Keenan et al., 1998) is
consistent with it. In addition, the relational complexity of the EF tasks (for example, of
inhibitory control) which are mastered at the same time as false belief tasks remains to
be determined.
2.3.1.3 Emergence accounts19
In this category falls two main theories of the ToM-EF relationship, one claiming that
EFs are a prerequisite for the development of ToM, and the other claiming that ToM is
necessary for EF to develop.
i) EF is required for ToM. This position is represented mainly by Russell
(1996, 1997b), whose argument is essentially that a sense of “agency” underlies self-
awareness, which in turn underlies the development of ToM. According to Russell,
agency has four main features: i) action-monitoring (the process through which changes
in experience are perceived to have been caused by the self and not the world), ii)
instigation (the ability of agents to determine their own perceptual sequences), iii) non-
observational knowledge of actions (the phenomenon whereby if an agent is in control
of his/her actions, s/he does not have to consciously observe them to know what they
are), and iv) privileged knowledge of goals (whereby in acting in a goal-directed fashion
we know incorrigibly what the goal is, whereas a third person does not). Russell
considers EF to be equivalent to action-monitoring and instigation (or at least that these
are the fundamental aspects of EF), and it is in this sense that he views EF as underlying
the development of ToM20. He asserts that these features allow a sense of ownership –
the perception of experiences as one’s own, and not determined by the world – and
therefore a self-awareness which he calls “pre-theoretical” (i.e., bodily-based and
immediate, requiring no comprehension of psychological concepts). This pre-
theoretical self-awareness is a necessary condition for the development of ToM - a
form of self-awareness which does depend on mental concepts.
19 Perner and Lang (1999, 2000) label these “functional dependence” accounts. The term “emergence” accounts is preferred here as this more directly refers to the developmental aspect of this class of explanation – that is, the notion that one ability depends on the other to develop or “emerge”. 20 Russell (1997b) recognises that EFs include other components such as inhibition, cognitive flexibility and working memory, and goes on to say how monitoring and instigation relate to these components. For example, he says that instigation (defined as the capacity to take actions not driven by habit or the external world) is analogous to the concept of generativity, but also requires inhibition; and flexibility requires adequate monitoring of the outcome of an incorrect response and instigation of a new strategy.
66
Empirical tests of Russell’s (1997b) theory have mostly been conducted on
children with autism (who are purported to have inadequate action-monitoring or
instigation), and these are reviewed in Section 2.3.2. However, a study by Hughes
(1998b) is also relevant. She found that preschoolers’ early EF performance,
particularly on a test of goal-directed action and inhibition, predicted ToM scores one
year later; but that early ToM scores did not predict later EF performance. Although
this study did not directly measure monitoring and instigation, it provides general
support for the notion that EF is required for ToM rather than vice versa.
Perner and Lang (1999, 2000) have critiqued Russell’s theory on conceptual
grounds, claiming that while it can explain how early action-monitoring may
fundamentally enable the early and later development of ToM, it does not adequately
explain why developments in false belief and inhibition in particular should occur at the
same age (around 4 years, later than the development of action-monitoring) or why
ToM and EF should be correlated at this age. Russell does not specifically address this
issue in his writings, tending to focus on expression or performance-based explanations
for the ToM-EF relationship during the preschool period, without relating later
inhibition to earlier action-monitoring and instigation.
Also posing a challenge for Russell’s theory are the existence of disorders where
EF is impaired but ToM is intact. If EF is a prerequisite for ToM development, one
would not expect children with impaired EF to show typical ToM capacity. Three
studies have now shown that children with ADHD or at risk of ADHD have impaired
EF (particularly inhibitory control) but intact performance on ToM tasks (Charman,
Carroll, & Sturge, 2001; Hughes, Dunn, & White, 1998; Perner, Kain, & Barchfeld,
2002a). A study by Tager-Flusberg, Sullivan, and Boshart (1997) demonstrated a
similar dissociation in children with Prader-Willi syndrome and Williams syndrome,
who showed impaired EF and intact ToM with no correlation between EF and ToM
performance. In addition, six children failed both EF tasks but passed both ToM tasks,
while no children passed both EF tasks if they failed both ToM tasks21, inconsistent
with the notion that intact EF is a prerequisite for ToM. Baron-Cohen and Robertson
(1995) reported a case of a child with Tourette’s syndrome who passed all ToM tasks
but failed two of three tests of inhibition, although this study comes with the usual
caveats of a single-case design.
21 This additional data was not contained in the original publication but was reported by Perner and Lang (2000).
67
Although Russell has not directly responded to these challenges to his theory, he
has implied that he subscribes to a multi-componential view of EF (Russell, 1997b) and
therefore may argue that individuals showing a ToM-EF dissociation have intact action-
monitoring and instigation but impairments in other aspects of EF. However, this
would contradict his assertion that monitoring and instigation are the fundamental basis
for EF, underlying its other components. Another important issue in evaluating
evidence of dissociations, highlighted by Perner and Lang (2000), is that the criterion
for failure of EF tasks is arbitrary in many cases, and so evaluating the relative pass/fail
rates of ToM and EF tasks suffers from the absence of absolute standards of
performance. Although Perner’s view takes the opposite form, he concludes that while
Russell’s theory is in need of greater specification regarding the aspects of ToM and EF
it incorporates, there is no firm evidence against it (Perner, 2000; Perner & Lang, 1999,
2000).
ii) ToM is required for EF. This account was first alluded to by Wimmer
(1989, cited in Perner, 1991), Frith (1992) in reference to schizophrenia, and Carruthers
(1996), and was then developed further by Perner (1998; Perner & Lang, 1999, 2000).
The essence of this position is that the metarepresentational capacity which
(purportedly) underlies ToM is necessary for volitional control over action. Wimmer’s
initial idea (as described by Perner) was that a better understanding of our own mind
and mental concepts allows better control over our mental processes and behaviour.
Carruthers (1996) developed this notion by positing that normal human reasoning
routinely involves second-order evaluation of first-order thoughts, beliefs and desires
(e.g., how strong is my desire to do x as opposed to y?), a kind of reflexive,
introspective access to our recent conscious mental events. He argues that the operation
of a ToM module underlies this meta-access to our own beliefs and thoughts, and in
turn, that this meta-access is a necessary condition for the evaluation of recent problem-
solving strategies such as is required on many EF tasks.
Perner (1998) elaborated upon this idea by specifically delineating the
metarepresentational requirements of the contention scheduling and SAS aspects of
Norman and Shallice’s (1980, 1986) model of EF (reviewed in Section 2.2.2). He
argues that contention scheduling does not require “meta-intentional” understanding
(i.e., a declarative, conscious representation of one’s goal), because at this level action
schemas control each other automatically by mutual inhibition and activation of
competing behavioural sequences (such as in trial-and-error learning). On the other
hand, intentional actions such as following verbal instructions or planning a future
68
action sequence (which require control by the SAS) demand a declarative representation
of a goal as desired or intended (i.e., “something the examiner wants me to do” or
“something I want to do”), so that the correct novel action schema can be boosted22.
These representations are called meta-intentional because they involve representing the
intended action sequence as intended. In certain situations, though, boosting the desired
action sequence is not sufficient – the inhibition of competing action schemas is also
required. In these cases, Perner argues, one needs to understand why the particular
competing action schema in question needs to be inhibited, and to do so one needs to
consciously conceptualise the action sequence as a tendency one has – that is, meta-
represent the schema as a representational vehicle (a representation which is not
specified by its content, such as a procedural action sequence). Thus, it is only in
situations where a competing action schema needs to be inhibited that we require
metarepresentational (not just meta-intentional) capacity. The developmental
relationship between tests of inhibitory control and false belief understanding therefore
occurs because they both require metarepresentational capacity - both require the ability
to represent representational vehicles (either action sequences or “pictures-in-the-head”)
which have causal efficacy (i.e., make people act in a certain way). In a sense, then,
Perner’s account is one of a common conceptual requirement of ToM and EF (i.e., they
both rely on metarepresentational capacity) rather than ToM itself being a prerequisite
for EF (although Perner himself places his account under the “functional dependence”
heading, equating metarepresentational ability with ToM and then saying that EF tasks
are applied ToM tasks)23.
The main piece of direct evidence for Perner’s theory comes from a study by
Lang and Perner (2002) which examined the relationship between early EF (as
measured by the DCCS and Luria’s hand game, which requires inhibitory control), false
belief and a knee-jerk reflex task. This latter task required the child to identify whether
or not they intended to move their leg after a reflex movement was elicited by the
examiner. Perner et al. argued that like the false belief and EF tasks, this requires an
understanding of mental states as representations which are causally responsible for
22 Perner (1998) characterises this distinction as being that contention scheduling occurs at the level of the representational vehicle, and the SAS exerts control at the level of representational content (see Perner, 1995 for further explanation). 23 It also resembles a “common conceptual requirements” account in that Perner implies that ToM and EF both depend on metarepresentational capacity throughout the lifespan, and not just during development. However, it was classified as an emergence or functional dependence account here both because that is how it is classified by Perner himself, and because the hypothesis does emphasise that ToM (or metarepresentational capacity) is necessarily for EF to develop.
69
actions (as the child needs to differentiate between intentional and accidental
movements). Consistent with their predictions, they found that the three types of task
were significantly correlated with age and verbal ability partialled out, implying that all
three abilities depend upon a common developmental factor. Perner and Lang (2000)
also report further relevant results from what was presumably a preliminary version of
the study, which showed that the knee-jerk reflex task still explained a significant
amount of variance in false belief performance beyond the EF tasks, suggesting that the
relationship between false belief and knee-jerk reflex understanding could not be
explained by any executive component in the knee-jerk task.
Perner’s account of the ToM-EF relationship has, like all the preceding
accounts, been subjected to a range of critiques. Russell (1997b; Russell et al., 2001)
has argued that the assumption that any behaviour with a second-order character (i.e.,
where the subject is required to represent to itself what it is doing and what needs to be
done) necessarily involves ToM is an unjustified over-stretching of the ToM concept.
He asserts further that action schemas or tendencies are not representations in any useful
sense, or at least, do not necessarily require metarepresentational understanding.
Russell has also criticised Perner’s interpretation of his result with the knee-jerk task,
arguing that the task could be considered to require inhibition of an answer based on
perceived outcome; and that the reason why it explains variance in the false belief task
beyond that explained by the EF tasks is that the response to be inhibited is verbal,
rather than a motor act as in Luria’s hand game (Russell et al., 2001; Russell, Hala, &
Hill, 2003). It could also be argued that the correct rejection of the reflex movement as
intentional requires action-monitoring (i.e., the ability to perceive the difference
between changes in experience caused by the self and the world), and therefore that the
results are also consistent with Russell’s agency theory.
A number of empirically based criticisms of Perner’s theory have been advanced
by Hughes (1998b) and Carlson and colleagues (Carlson & Moses, 2001; Carlson et al.,
2002). Firstly, Hughes’ (1998b) finding that early EF predicted later ToM but not vice
versa is inconsistent with the notion that ToM is a prerequisite for EF. Perner and Lang
(1999, 2000) attempt to reinterpret this finding in their favour by arguing that EF tasks
assess the understanding of mental states as causally efficacious as much as ToM tasks,
and that Hughes’ data can be explained by assuming that this metarepresentational
understanding occurs in reference to one’s own actions first, and in reference to others’
actions later. However, this is not consistent with findings that on the unexpected
contents (Smarties) and unexpected identity (appearance-reality) tasks, correct reporting
70
of one’s own previous belief develops at the same time as, or even after, the correct
prediction of others’ beliefs (e.g., Gopnik & Astington, 1988). Also, this interpretation
would mean that the findings of impaired EF and intact ToM in certain disorders, the
evidence used by Perner against Russell’s theory, would pose an equally difficult
problem for Perner: is it plausible that children could be impaired in a developmentally
precedent ability in comparison with intact performance on an ability which develops
later?
Secondly, Hughes (1998b) points out that significant improvements in inhibition
and goal-directed behaviour occur during infancy (see Section 2.2.2), long before
children acquire Perner’s representational theory of mind. Perner et al. (1999) provide a
more solid defence of this problem by distinguishing between “automatic inhibition”
(when a more highly activated schema naturally inhibits less activated competitors, or
relatively automatic suppression of motor or cognitive responses), which he argues is
tapped by the A-not-B task and other EF tasks used with infants, and “executive
inhibition” (when no alternative schema is automatically activated and a response must
be actively inhibited, or when there is deliberate suppression of a response to achieve an
internally represented goal), which is what is tapped by EF tasks mastered around the
age of 4, and which requires metarepresentational understanding.
A third empirical problem for Perner was discovered by Carlson and colleagues
(Carlson & Moses, 2001; Carlson et al., 2002), who found that their two types of
inhibition task showed differential relationships with ToM, but both required executive
inhibition and therefore according to Perner should have been equally related to ToM.
Perner et al. (2002b) also found themselves that performance on a go-nogo task which
required executive inhibition was not significantly correlated with false belief prediction
or card sorting performance, contrary to their predictions. Fourthly, as highlighted by
Hughes (1998b), young children may correctly verbalise their understanding of the rules
of a task but nevertheless demonstrate perseveration of the incorrect response (Zelazo et
al., 1996b), indicating that meta-intentional ability is not sufficient for strategic
performance on an EF task.
The existence of dissociations between ToM and EF whereby ToM is impaired
but EF is spared poses another difficulty for Perner, although, consistent with Perner’s
account, these cases appear to be rarer than cases of the reverse dissociation. A number
of studies of individuals with brain injuries have demonstrated a ToM-EF dissociation
such that ToM is impaired and EF intact (reviewed in the next section). In addition,
deaf children displaying intact EF performance still demonstrate a ToM impairment
71
(Remmel, 2003). However, it could be argued that the abnormal development of ToM
in deaf children has its basis in a different process to other conditions and is therefore a
poor example of this dissociation, as it appears to be impoverished language
development which underlies the delay in ToM rather than a metarepresentational
deficit (de Villiers, 2000).
Overall, then, while Perner’s theory of the ToM-EF relationship has an
interesting and well-developed conceptual basis, it has not as yet been able to
adequately refute conceptually grounded critiques or account for all the available data.
Perner and Lang (1999) suggest that both emergence accounts may be correct, such that
ToM and EF are interdependent: “an understanding of mental states as causally
efficacious is required for executive inhibition, and executive inhibition is a main
exercise ground for a theory of mind at this stage of development” (p. 342). However,
while both of the emergence accounts are strengthened by evidence suggesting a deep
link between ToM and EF during conceptual development, they are equally weakened
by evidence that ToM and EF may be dissociably impaired (this is discussed further
later).
2.3.1.4 Common neuroanatomical bases for ToM and EF
This category of explanation holds that correlations between ToM and EF may be
coincidental or accidental, occurring because both abilities depend upon the same or
proximal brain regions (Bach et al., 1998; Ozonoff, 1995a; Ozonoff et al., 1991;
Pennington et al., 1997). On this account, the concurrent developments in ToM and EF
which occur around the age of 4 are due to late maturation of these common brain
structures. It can also explain the frequent co-occurrence of ToM and EF impairments,
on the basis that proximal neuroanatomical structures will often be damaged together.
While the notion of common underlying brain regions is not inconsistent with any of the
other theories of the ToM-EF relationship, the idea that this is the only link between the
constructs is a possibility unique to this hypothesis. This account therefore allows for
dissociations between ToM and EF, although of course only if the brain regions in
question are not absolutely identical.
So what are the brain regions in question? Areas within the prefrontal cortex are
obvious suspects. In Sections 2.2.1 and 2.2.2, we saw that while EF tasks are not
necessarily sensitive or specific to the functioning of the prefrontal cortex, the view that
EFs rely upon the prefrontal cortex is fairly well established (e.g., Owen, Downes,
72
Sahakian, Polkey, & Robbins, 1990; see Stuss & Knight, 2002). Neuroimaging studies
and investigations of brain-damaged and psychiatric patients have converged on the
notion that the dorsolateral prefrontal cortex in particular is important for working
memory, problem-solving, attentional flexibility and planning (Burgess, 2000; Cabeza
& Nyberg, 2000; Collette & van der Linden, 2002; Dagher, Owen, Boecker, & Brooks,
1999; Fuster, 2000; Goldman-Rakic & Leung, 2002; Grattan, Bloomer, Archambault, &
Eslinger, 1994; Kane & Engle, 2002; Mega & Cummings, 1994; Weinberger, 2002).
An increasing number of recent neuroimaging studies of the brain regions
involved in ToM (which are generally adult studies comparing activation during
advanced ToM tasks with structurally similar tasks containing no mentalistic content)
have implicated a network of structures including the medial prefrontal cortex, anterior
cingulate cortex, superior aspects of the temporal lobes, and the temporal poles (Baron-
Cohen et al., 1994, 1999a; Brunet, Sarfati, Hardy-Baylé, & Decety, 2000; Castelli et al.,
2002; Fletcher et al., 1995; Gallagher et al., 2000; Goel, Grafman, Sadato, & Hallett,
1995; for reviews, see Abu-Akel, 2003; Frith & Frith, 2000; Gallagher & Frith, 2003;
Kain & Perner, 2003). In their review of neuroimaging studies of ToM, Gallagher and
Frith (2003) argue that the anterior paracingulate cortex (which is part of the medial
frontal cortex, and lies just anterior to the anterior cingulate cortex proper) is the crucial
region dedicated specifically to processing mental states, and the temporal regions
which are commonly activated in ToM tasks have more secondary functions such as the
interpretation of biological motion (which may be necessary to ascribe intentionality to
others) and episodic memory (which may be required to imagine ourselves in the
situation of another person). The orbitofrontal cortex and amygdala have also been
proposed to have a role in social cognition (Baron-Cohen & Ring, 1994; Brothers,
1996), however activation in these areas is not seen in the majority of neuroimaging
studies of ToM. Gallagher and Frith (2003) conclude that while these areas may form
part of the social brain in general (with the amygdala involved in the automatic response
to socially salient stimuli as well as playing a key role in emotion, and the orbitofrontal
cortex in the processing of affective, particularly aversive, social stimuli), they are
unlikely to be directly involved in ToM. However, neuroimaging studies are of limited
utility in investigating the role of the orbitofrontal cortex in ToM as it is difficult to
obtain reliable activation maps for this region (Gregory et al., 2002).
The importance of the prefrontal cortex in ToM has also emerged in studies of
neurological patients, which have demonstrated significant ToM impairments in patients
with prefrontal damage (Bach et al., 1998; Channon & Crawford, 2000; Gregory et al.,
73
2002; Happé, Malhi, & Checkley, 2001; Lough, Gregory, & Hodges, 2001; Lough &
Hodges, 2002; Rowe et al., 2001; Stone, Baron-Cohen, & Knight, 1998; Stuss, Gallup,
& Alexander, 2001). However, the regions of the prefrontal cortex implicated in these
studies have been somewhat more ambiguous than those indicated in neuroimaging
studies of normal adults24. Rowe et al. (2001) found no effect of the site of lesion
within the prefrontal cortex on ToM performance. Some studies have muddied the
issue either by defining damage as being in the “orbitomedial” region, using the terms
ventromedial and orbitofrontal interchangeably, or focussing on the area of overlap
between the orbitofrontal and ventromedial regions (Gregory et al., 2002; Lough et al.,
2001; Lough & Hodges, 2002; Stuss et al., 2001). Studies by Happé et al. (2001) and
Stone et al. (1998) found that ToM is impaired following orbitofrontal lesions, contrary
to the findings of neuroimaging studies. Cicerone and Tanenbaum (1997) also describe
a patient with traumatic orbitofrontal injury who performed poorly on ToM-like tasks
requiring interpretation of social situations. Drawing on additional evidence that
patients with orbitofrontal damage commonly show marked changes in social
behaviour, Stone (2000) concludes that the orbitofrontal cortex is the most crucial
region for ToM. However, Bach, Happé, Fleminger, and Powell (2000) report a case of
an adult male with orbitofrontal damage who showed intact performance on ToM tasks
even though he displayed a disturbance in social behaviour. Eslinger (1998) also
reviews evidence showing that patients with orbitofrontal lesions have impaired
emotional or affective empathic processing, but intact performance on cognitive aspects
of empathic processing25. Importantly, though, studies of neurological patients rarely
implicate the dorsolateral prefrontal cortex in ToM (although see Price et al., 1990).
Regardless of whether ToM relies more heavily on orbitofrontal or medial
frontal regions, more relevant are studies addressing the notion that ToM and EF are
related because they rely on proximal brain regions. Consistent with this hypothesis, a
number of studies have reported co-existing ToM and EF impairments in patients with
prefrontal damage not limited to specific dorsolateral, ventromedial or orbitofrontal
regions (Bach et al., 1998; Channon & Crawford, 2000; Gregory et al., 2002; Rowe et
al., 2001; Saltzman et al., 2000). Gregory et al. and Rowe et al. both found that while
24 The laterality of ToM representation in the brain is also unclear, although it appears that patients with right hemisphere damage show ToM impairments more consistently than patients with left hemisphere damage (see Kain & Perner, 2003). 25 Interestingly, consistent with the notion of a distinction between cognitive and affective aspects of empathy, Blair et al. (1996) found that psychopaths show intact performance on ToM tasks but do not show typical affective responses (as measured by physiological arousal) to images of individuals in distress.
74
their frontal patients scored poorly on both ToM and EF tasks, these deficits were not
significantly correlated, consistent with the idea that the two types of task may rely on
different aspects of the prefrontal cortex. This specialisation within the prefrontal
cortex has also been supported by studies demonstrating dissociations between ToM
and EF in patients with damage to specific prefrontal regions. Case and group studies
have demonstrated specific impairments in ToM in the face of intact EF in patients with
frontotemporal dementia (Lough et al., 2001; Lough & Hodges, 2002). In addition,
Fine, Lumsden, and Blair (2001) reported a case of a patient with congenital amygdala
damage who demonstrated impaired ToM but intact EF. The reverse dissociation, of
intact ToM with impaired EF, was also reported by Bach et al. (2000). A double
dissociation of sorts was demonstrated by Stone et al. (1998), who found that while
patients with orbitofrontal damage failed ToM tasks regardless of their working
memory load, dorsolateral prefrontal patients displayed impaired ToM performance
only under conditions where the working memory load was high (moreover, under these
conditions they made errors on control questions as well as belief questions).
This account of the ToM-EF relationship has therefore been fairly resistant to
criticism, as it is able to explain both the relationships and the dissociations between
ToM and EF. Of course, in addition to the lack of consistency in defining what
constitutes an impairment, the problem of equating ToM and EF tasks for difficulty
should be noted as a caveat in interpreting the dissociations observed in neurological
patients (and any other clinical samples or individuals), although most studies are
careful to note the absence of any floor or ceiling effects. Aside from this, Perner and
Lang (2000) have argued that the theory lacks strong predictive value, as any kind of
task association or dissociation is compatible with it. Perner (2000) adds that while it
accounts for a general developmental relationship between ToM and EF based on
common timing of the maturation of prefrontal structures, it does not specifically
predict that false belief tasks and “executive inhibition” should be mastered at the same
time. He also points out that environmental factors such as number of older siblings
influence the age at which false belief tasks are mastered (Perner, Ruffman, & Leekam,
1994; Ruffman, Perner, Naito, Parkin, & Clements, 1998). However, a number of other
extraneous factors also influence ToM development (e.g., language, visual perception),
but this does not preclude the notion that ToM has a specific neuroanatomical basis
which is proximal to the structures involved in EF.
75
So, what can we conclude overall about the relative strength of the various models of
the ToM-EF relationship? As we have seen, no account has eluded criticism. Leaving
aside the problem of defining “impairment” and equating ToM and EF tasks for
difficulty, one recurrent theme is that the explanation must account for the observed
correlations between ToM and EF and the frequent co-occurrence of deficits in the two
areas, as well as allowing for dissociable impairments in either direction. The
“emergence” accounts in particular are weakened by evidence of dissociations, as they
both imply a fundamental conceptual dependence between the two constructs during
development. Similarly, dissociations are problematic for the “common conceptual
requirements” accounts particularly if the tasks on which dissociations are demonstrated
are both purported to rely on the same third underlying mechanism. While the
“common neuroanatomical bases” explanation accounts best for these data, it lacks
specificity in its predictions regarding the components of EF and types of ToM task
which should be related. Also, it does not specifically account for data showing that
performance on false belief and deception tasks improves when the EF requirements are
reduced. The “expression” accounts do not strictly predict ToM-EF dissociations as one
would expect that performance on ToM tasks would be affected if EF is impaired
(although it would be acceptable for ToM to be impaired while EF is intact), but if the
account is limited to specific aspects of EF such as inhibition or working memory, then
dissociations with other EF components would be allowable. Although this account has
been criticised on the basis of evidence suggesting that the ToM-EF link extends
beyond the level of the EF requirements of ToM tasks, those findings do not exclude the
possibility that there may be both performance-based and deeper conceptual or
functional (and neuroanatomical) links.
It is also possible that ToM and EF may be dependent on each other for their
initial development but then become separable processes when matured (linked only by
the EF requirements of ToM tasks and/or their common neuroanatomical substrates) –
in which case emergence accounts would not be inconsistent with the existence of
dissociations in adults or children over the age of five (in whom ToM and EF had
previously developed normally). Perner et al. (2002a) do not appear to concur with this
possibility, implying that functional dependence should extend across the life span26.
However, Karmiloff-Smith (1992; Karmiloff-Smith, Scerif & Ansari, 2003; Thomas &
Karmiloff-Smith, 2002; see also Bishop, 1997) has argued persuasively that processes
26 If Perner’s account is viewed as a “common conceptual requirements” account, this claim becomes somewhat more defensible. However, he suggests that this is the case for both his and Russell’s accounts.
76
which are dissociable in adulthood cannot be assumed to be so during development and
that “a difference in performance....at any point in development does not permit the
inference of a stable double dissociation at a later or earlier time” (Karmiloff-Smith et
al., 2003, p.162). This argument carries the inference that double dissociations between
ToM and EF observed during middle childhood and adulthood do not necessarily mean
that the two abilities were not interrelated during earlier stages of development27. It is
possible and not implausible that performance-based, conceptual, functional, and
neuroanatomical factors interact and combine to produce the observed relationships
between ToM and EF during different stages of development and in different disorders.
This remains a speculative proposition, however – the nature or existence of the
relationship between ToM and EF beyond the preschool period has been largely
overlooked by all of the main accounts, which have focused in particular on the
relationship between false belief and certain aspects of EF (inhibition, working memory,
conditional reasoning, monitoring) between the ages of 3 and 5, without generating or
testing specific predictions about the ToM-EF relationship in later childhood,
adolescence and adulthood (or addressing the role of components of EF such as
generativity and flexibility, which are still developing during late childhood and
adolescence).
Only three studies provide separate data on correlations between ToM and EF
for non-clinical controls older than 5 years, with inconsistent results. Perner et al.
(2002a) found a number of significant correlations between a second-order false belief
task and a range of EF measures in typically developing 4.5–6.5 year-olds, while
Charman et al. (2001) did not find any significant correlations between advanced ToM
stories and measures of inhibition and planning in their 8-10 year-old typically
developing controls. The only available adult data is from Channon and Crawford
(2000), who report significant correlations for their healthy adults (with a mean age of
43 years) between advanced ToM stories and two measures of generativity, but not
other EF measures of flexibility, inhibition and planning. These data suggest that the
nature of the ToM-EF relationship may not be the same for older children and adults as
for young children, and therefore that ToM-EF dissociations in these age groups may
not be easily interpreted in terms of theories based on the preschool period. Notably,
almost all reported ToM-EF dissociations have occurred in individuals or samples older
27 Another example of this concept is that visuospatial skills are functionally dependent upon basic vision for their typical development (e.g., Vecchi, 1998), but in adult disorders it is possible for visuospatial processing to be impaired without disruption to vision itself, such as in cases of spatial neglect (see Heilman, Watson, & Valenstein, 1993).
77
than 5 years. The only exception is Hughes et al.’s (1998) study on “hard-to-manage”
preschoolers, which found that these children showed largely intact performance on
ToM tasks in comparison with impaired performance on EF tasks, but nevertheless that
ToM and EF were correlated in this group. Evidently, further studies on older age
groups are necessary to delineate the nature of the ToM-EF relationship beyond the
early years and the meaning and implications of ToM-EF dissociations for the various
accounts of the ToM-EF relationship. This is particularly important for the
interpretation of studies on the ToM-EF relationship in autism, which have been
conducted largely on older children.
2.3.2 The ToM-EF relationship in autism
As a developmental disorder characterised by both ToM and EF impairments, autism
provides an interesting test case for the various accounts of the ToM-EF relationship,
each of which generates different predictions about the relationship between ToM and
EF deficits in autism. The nature of the relationship is highly relevant to the evaluation
of hypotheses of autism which propose a primary deficit in either ToM or EF. A single
primary cognitive deficit account of autism needs to demonstrate that one impairment
subsumes or explains the other; and conversely, a multiple cognitive deficits account
would be consistent with evidence suggesting that ToM and EF are (at least partially)
independent deficits in autism.
Surprisingly, only a few studies have directly measured correlations between
ToM and EF in autism, with many authors relying on the mere existence of both ToM
and EF impairments in autism, other indirect evidence, or the theories and evidence
generated from the study of typically developing children to argue for their position.
Those who claim that ToM and EF are related deficits in autism rely heavily on Ozonoff
et al.’s (1991) finding that ToM and EF were correlated in their sample of high-
functioning individuals with autism. However, this was based on single ToM and EF
composite scores, therefore obscuring the specific nature of the relationship; and
furthermore, age was not partialled out of the correlation, leaving open the possibility
that the correlation may have been mediated by age.
The evidence regarding the nature of the ToM-EF relationship in autism will be
reviewed by examining the predictions of each of the accounts reviewed in the previous
section for the ToM-EF relationship for autism, and how well these predictions fit with
the available data.
78
i) Expression accounts. The idea that the failure of children with autism on
ToM tasks may be at least partially due to difficulties with their EF requirements has
been most overtly advocated by Russell and colleagues (Hughes & Russell, 1993;
Russell, 1997b; Russell, Saltmarsh, & Hill, 1999). In favour of this, Hughes and
Russell (1993) found that participants with autism continued to fail a test of strategic
deception (the windows task) when there was no opponent present. Russell et al. (1999)
found that children with autism demonstrated significantly poorer performance than
controls on the conflicting desire task used by Moore et al. (1995), suggesting that their
difficulty with the false belief task is not restricted to a lack of understanding of the
representational nature of belief. Charman and Lynggaard (1998) also found that the
performance of children with autism on the Smarties task was enhanced by the
provision of a photographic cue which (arguably) reduced the working memory and
inhibitory demands of the task, although Bowler and Briskman (2000) were not able to
replicate this effect using the standard Sally-Anne false belief task.
Although this evidence indicates only that children with autism show
impairments on tasks with both ToM and EF requirements as well as on tasks with only
EF requirements, some proponents of the EF hypothesis of autism have nevertheless
suggested that children with autism may fail ToM tasks because of their EF
requirements (e.g., Ozonoff, 1997a; Russell et al., 1999). One problem with this
account, which has been overlooked by all of its critics, is that children with autism do
not tend to demonstrate impairments on tests of inhibitory control or working memory
(see Section 2.2.3), the main EF components implicated in expression accounts of the
ToM-EF relationship. Although Russell and his colleagues do not directly address this,
they implicitly sidestep it by interpreting their findings on the strategic deception and
conflicting desire tasks in terms of a difficulty with mental disengagement rather than
emphasising inhibition, an interpretation which is a little more consistent with the
attentional shifting difficulties more consistently displayed by individuals with autism.
Also, no studies have directly examined the performance of children with autism on
tests combining inhibitory and working memory requirements (in comparison with tests
tapping one or the other), which, as we have seen, appears to be more relevant to false
belief performance.
The most common argument advanced by critics of the view that EF
impairments may explain the poor performance of children with autism on ToM tasks,
however, is that they are able to pass the “false photograph” task (described in Section
2.1.3). The claim is that the false photograph task has an identical task structure to the
79
false belief task, and therefore that their failure on false belief tasks cannot be due to
their EF requirements (Baron-Cohen & Swettenham, 1997; Leslie & Roth, 1993; Leslie
& Thaiss, 1992). However, a number of criticisms of the false photograph task have
been put forth in return. Pennington et al. (1997) argued that the “false” photograph is
not actually false: it does not misrepresent current reality because the nature of
photographs is that they do not portray current reality, and therefore the adequate
performance of children with autism could be explained by their intact understanding of
cameras. Pennington et al. and Hughes et al. (1994) both also maintain that the camera
and photograph are perceptually salient, available and enduring to participants in a way
that inferred beliefs are not. Similarly, Russell (1997b) claims that the inhibitory
demands made by the false photograph task are far weaker than those made by the false
belief task, as the participant is required only to inhibit their current perception of a
three-dimensional representation (i.e., a toy) in order to refer to what is known about a
two-dimensional representation (i.e., a photograph of a different toy). This claim was
tested by Russell et al. (1999) by using a modified version of the false photograph task
in which the initial photograph was taken of a blank wall, designed to increase the
relative salience of the current representation (a three-dimensional doll) in comparison
to the old one (where nothing was present). They found that while children with autism
were able to pass the standard false photograph task, they demonstrated impaired
performance on the modified version, indicating that when the inhibitory demands of
the task matched those required by the false belief task more equally, children with
autism could not sustain intact performance. Russell (1997b; Russell et al., 1999) has
nevertheless made it clear that he is not of the view that the ToM deficits displayed by
individuals with autism are entirely due to EF impairment, or that if the EF demands of
ToM tasks were removed, then autistic individuals would show normal performance.
Although Leslie and colleagues present an expression account of the ToM-EF
relationship in typical development, they of course do not subscribe to this view of ToM
failures in autism. While they propose that 3-year-olds fail false belief tasks because of
an impaired Selection Processor (SP), they argue that children with autism have an
intact SP but instead fail false belief tasks because of impaired metarepresentational
capacity or ToMM (Leslie & Thaiss, 1992; Leslie & Roth, 1993). In support of their
view, Leslie and colleagues have reported evidence that children with autism do not
benefit from helpful task modifications as 3-year-olds do, and that while 3-year-olds
will attribute beliefs to others even if they are incorrect, children with autism will not
attribute any beliefs at all (Leslie & Roth, 1993; Roth & Leslie, 1998; Surian & Leslie,
80
1999). Leslie and colleagues acknowledge the presence of EF impairments in autism,
but in their model, these are independent from the SP and from ToMM. They subscribe
to a view of EF as a fractionated system whereby children with autism are impaired in
some EF components but not those involved in the SP (Leslie & Roth, 1993). If the SP
is considered to be an inhibitory mechanism, this in fact fits quite well with the
literature suggesting that inhibition is intact in autism.
ii) Common conceptual requirements of ToM and EF. In discussing the
applications of their CCC theory for autism, Zelazo, Frye and colleagues have strongly
advocated the role of domain general processes (such as rule-based reasoning) in the
cognitive aetiology of autism and argued against the conception that autism is
characterised by a domain specific impairment in a theory of mind module (Frye et al.,
1998; Zelazo et al., 1996a, 2001). Zelazo et al. (2002) tested the hypothesis that
individuals with autism may fail ToM tasks because of domain-general difficulties in
rule use by examining correlations between performances on two false belief tasks, the
physical causality task, and the DCCS task. They found that the correlation between
ToM and rule use tasks was not significant for severely impaired individuals with
autism (due to floor effects on most tasks), but was significant for their mildly impaired
group. They interpreted this result as indicative of the lack of domain-specificity of
ToM impairments in autism, and furthermore, argued that ToM deficits in autism may
be accounted for by rule-based reasoning impairment.
Besides the study’s small sample size (with only 10 mildly impaired
participants) and the lack of a control group (necessary to ensure that any difficulties
displayed are connected to autism; Colvert, Custance, & Swettenham, 2002), an
important limitation of Zelazo et al.’s (2002) study is that they did not partial out age or
IQ variables in their correlations. A study reported by Colvert et al. (2002) which
addressed these limitations nevertheless replicated the result with 20 high-functioning
children with autism, finding significant correlations between false belief and DCCS
performance with age, verbal and non-verbal ability partialled out. However, as Colvert
et al. point out, further research is needed to investigate what other factors (e.g.,
inhibition, salience of the switch of setting conditions) might account for these
correlations; particularly in light of the criticisms of CCC theory outlined in the
previous section’s review. In addition, as for dissociations observed in other disorders,
the presence of ToM-EF dissociations in autism (reviewed below) challenge the notion
that the two abilities depend upon a common conceptual ability.
81
Halford has not discussed or tested the implications of his relational complexity
account (Halford, 1993; Halford et al., 1998) for autism. His proposal implies that
individuals with autism may demonstrate limited relational complexity, a prediction
awaiting empirical confirmation.
iii) Emergence accounts. In his account of why a sense of internal agency is
a necessary prerequisite for the development of ToM, Russell (1997b) specifically
posited that autism may be a disorder characterised by impaired action monitoring and
instigation, and therefore that these deficits may underlie the abnormal development of
ToM28. Previous studies showing deficits in imitation (Smith & Bryson, 1994) and
motor planning (Hughes, 1996a) in autism are consistent with this hypothesis. Russell’s
first direct investigations of action monitoring in autism were promising, although not
compelling (he has not studied instigation deficits, saying that this has been adequately
covered by others under the guise of “generativity”). Russell and Jarrold (1998) found
that on a task involving the launching of missiles towards targets, children with autism
failed to correct errors based on both external and internal feedback. This was
interpreted as indicating an impairment in constructing visual schemata of motor acts,
which are necessary for action monitoring (although the authors acknowledged that
their data could also be consistent with a deficit in flexibility). Russell and Jarrold
(1999) tested higher-level self-monitoring by using a task requiring children with autism
to recall whether they or another person had performed a certain action. Consistent with
their predictions, they found that children with autism demonstrated impaired
performance on this task, suggesting that they were failing to monitor their actions as
their own. However, they also demonstrated some subtle difficulties on memory-based
control tasks. More recent studies have not been so favourable towards Russell’s
theory. Using a range of tasks including monitoring of basic actions, reporting an
intention when the outcome was unintended but desired, and reporting on intended
actions when the action achieved was unexpected, Russell and Hill (2001) did not find
any strong evidence of monitoring impairments in children with autism. Similarly, Hill
and Russell (2002) did not find evidence for a self-monitoring impairment in autism
using a test of memory for actions which involved a self/other source attribution (i.e., a
judgement of who performed the act), inconsistent with Russell and Jarrold’s (1999)
28 Russell (1997b; Russell & Hill, 2001) also argues that these deficits can account for the range of other EF deficits displayed by individuals with autism. He proposes that action-monitoring and instigation underlie the development of verbal self-regulation (or “inner speech”), which in turn is necessary to hold in mind arbitrary rules. Therefore, individuals with autism are impaired on EF tasks which have arbitrary rules.
82
results. These failures to meet the predictions of Russell’s (1997b) theory have led him
to reconsider his conceptualisation of the core EF deficit in autism. Consistent with
Ozonoff (1997b), Russell and Hill (2001) proposed that set-shifting or flexibility may
instead be the core impairment in autism, and that this deficit may have a “homologous”
rather than a “causal” or functional relationship with ToM. They propose that if one
assumes that cognition is a form of set-shifting between domains, then children who are
mentally inflexible would find it challenging to reflect on mental acts (their own and
other people’s).
Other proponents of the EF hypothesis of autism have presented alternative
emergence accounts of the ToM-EF relationship in autism, although these have not been
as extensively developed as Russell’s either conceptually or empirically. Hughes and
Russell (1993) suggested that a child with an impairment in dealing with novelty and
making decisions due to a damaged SAS (Norman & Shallice, 1980, 1986) would fail to
develop successful social relations, with the developmental outcome being an impaired
ToM. Pennington et al. (1997) proposed that autism is characterised by a severe deficit
in working memory, which results in an early disruption in the planning and execution
of complex behaviour. Because this occurs early in development, it affects the
acquisition and use of concepts that require the integration of information across time
and contexts. Concepts with these requirements include a recognition of one’s own and
others’ intentions and their correspondence or conflict, which is involved in imitation as
well as later ToM abilities. Hence, an early impairment in working memory would
result in the development of a mentalising impairment. An obvious challenge for this
account is the absence of convincing evidence for a working memory deficit in autism
(in the absence of any inhibitory requirements). Ozonoff and McEvoy (1994) suggested
that an early and persistent impairment in the ability to disengage from the external
environment and guide behaviour by internal mental models (see Harris, 1993) would
have significant consequences for the ability to appreciate others’ perspectives (which
requires disengagement from one’s own prepotent thoughts). A problem for all of these
accounts, however, is the lack of convincing evidence of early EF deficits in autism (see
Section 2.2.3), which speaks against the notion of EF impairment as being causally
primary.
The existence of ToM-EF dissociations in individuals with autism, whereby EF
is impaired but ToM intact, presents additional difficulties for these accounts (which in
one way or another all propose that a primary EF impairment underlies the abnormal
development of ToM), although only one study has reported data relevant to this
83
dissociation. Ozonoff et al. (1991) found that in their sample of high-functioning
individuals with autism, EF deficits were almost universal but ToM deficits only
occurred in a subset of the sample, with the implication being that some individuals
failed EF tasks while passing ToM tasks. This finding is inconsistent with the view that
EF is a necessary prerequisite for ToM and that early EF deficits result in later ToM
impairment. Ozonoff et al. conclude that while EF deficits are primary in autism, they
are unlikely to be causally related to ToM (they adopt a “common neuroanatomical
bases” position, as reviewed below). However, it is possible that differences in the level
of difficulty of the two sets of tasks may account for the pattern of results (Perner &
Lang, 2000).
The reverse emergence account proposed by Perner and colleagues has not been
directly examined with autistic individuals, although they have implied that a ToM (or
metarepresentational) impairment may underlie EF deficits in autism just as for 3-year-
olds (Perner, 1998; Perner & Lang, 2000). The most obvious difficulty with the
application of this account to autism is that children with autism have shown intact
performance on tests which may be regarded as measuring Perner’s “executive
inhibition” (e.g., Ozonoff & Strayer, 1997), which his theory would not predict. ToM-
EF dissociations in autism whereby ToM is impaired but EF intact also contradict his
notion that metarepresentational ability is a prerequisite for EF development. Baron-
Cohen and Robertson (1995) reported a case of a child with autism who failed several
ToM tasks but performed successfully on EF tasks, and Baron-Cohen, Wheelwright,
Stone, and Rutherford (1999b) report the same dissociation in three high-functioning
adults with autism. Of course, the small number of individuals for which this
dissociation has been noted limits the generalisability of these findings. Perner’s theory
of the ToM-EF relationship in autism requires a direct and systematic investigation
before it may be adequately evaluated, although the available evidence is not overly
favourable towards it.
iv) Common neuroanatomical bases for ToM and EF. The possibility that
ToM and EF impairments may co-occur in autism because of their proximal
neuroanatomical substrates was first proposed by Ozonoff et al. (1991; see also Bishop,
1993). However, Ozonoff (1997a; Ozonoff & McEvoy, 1994) subsequently put forward
an opinion that there may be both performance-based and conceptual links as well.
Baron-Cohen and Swettenham (1997), on the other hand, clearly expressed the view
that ToM and EF are best conceptualised as independent deficits in autism, which
probably co-occur because of their shared frontal origins.
84
While prefrontal abnormalities have been found in individuals with autism (as
discussed in Section 2.2.3), more convincing evidence for this class of explanation
would involve demonstrating both dorsolateral and medial or orbitofrontal impairment,
as the purported substrates for EF and ToM respectively. Two studies provide indirect
evidence for dorsolateral abnormalities in autism. Luna et al. (2002) found significantly
reduced activation in the dorsolateral prefrontal cortex in individuals with autism during
the performance of a spatial working memory task. Goldberg et al. (2002) also
interpreted the presence of impairments on an eye movement anti-saccade task as
suggestive of dorsolateral prefrontal dysfunction in autism. The idea that autism may
involve medial frontal or orbitofrontal dysfunction has been suggested by a number of
authors (e.g., Bachevalier & Loveland, 2003; Damasio & Maurer, 1978; Mundy, 2003)
on the basis of the region’s apparent role in social behaviour. In a series of studies,
Dawson and colleagues (Dawson et al., 1998, 2002a; Dawson, Osterling, Rinaldi,
Carver, & McPartland, 2001) obtained indirect evidence of ventromedial prefrontal
dysfunction in early autism by demonstrating impairments on tasks previously found to
be linked with ventromedial functioning. Two studies have also found reduced
activation in medial frontal areas during the performance of ToM tasks in individuals
with autism (Castelli et al., 2002; Happé et al., 1996). While these studies do not
provide unequivocal evidence of dorsolateral and medial/orbitofrontal abnormalities in
autism, they are at least consistent with the possibility that ToM and EF impairments
may co-occur in autism because of damage to their proximal neural substrates.
Overall, then, we can conclude only that the nature of the relationship between ToM and
EF in autism remains unclear. The only studies to have conducted direct correlations
between ToM and EF in autism have either failed to partial out the effects of age and IQ
variables (Ozonoff et al., 1991; Zelazo et al., 2002), not examined specific relationships
with components of EF (Ozonoff et al., 1991), or only included one type of EF task
(Colvert et al., 2002). Studies addressing expression accounts of the ToM-EF
relationship have not yet systematically varied the EF requirements of ToM tasks, and
the emergence accounts again struggle with the presence of ToM-EF dissociations in
autism as well as being inconsistent with some of the available data. The accounts
which propose that ToM and EF are related in autism, while intuitively appealing,
therefore remain open to further investigation. Interestingly, these accounts mostly
originate from proponents of the EF hypothesis of autism, who argue that EF deficits
may explain ToM impairments in autism (either because of performance-based or
85
functional/developmental links). Notably, Baron-Cohen and Leslie - the most
prominent proponents of the ToM hypothesis of autism - both adhere to the view that
ToM and EF are independent deficits in autism. The independence and relative primacy
of ToM and EF in autism are clearly important matters awaiting further empirical work.
These matters are addressed in the current research.
86
CHAPTER 3
Selection and Description of Measures 3.1 Diagnostic measures
3.1.1 Autism Screening Questionnaire
3.1.2 Autism Diagnostic Interview – Revised
3.2 IQ measures
3.3 ToM measures
3.3.1 Simple false belief task
3.3.2 First-order false belief task
3.3.3 Second-order false belief task
3.3.4 Dewey Stories
3.4 EF measures
3.4.1 Tower of London
3.4.2 Intra-dimensional, Extra-dimensional Set-shifting task
3.4.3 Response Inhibition and Load task
3.4.4 Opposite Worlds
3.4.5 Relational Complexity
3.4.6 Pattern Meanings
3.4.7 Uses of Objects
3.4.8 Stamps task
3.5 Behavioural measures
3.5.1 Measures of repetitive behaviour
3.5.1.1 Repetitive Behaviours Questionnaire
3.5.1.2 Repetitive Behaviours Interview
3.5.2 Measures of social behaviour and communication
3.5.2.1 Social Behaviour Questionnaire
3.5.2.2 Social and communication ADI-R domains
87
Both of the studies contained in this thesis involve the use of a large range of cognitive,
behavioural, and diagnostic measures. This chapter is devoted to a comprehensive
discussion of each set of measures, including a rationale for the selection of each
measure, a brief overview of its psychometric properties where possible, and a detailed
description of what it entails. This precedes the chapters describing the two studies
partly because of its length (due to the number of measures involved), and partly
because the measures used are common to both studies. The reader is encouraged to
refer back to this chapter when the procedure and results of the studies in Chapters 4
and 6 are being discussed.
3.1 Diagnostic measures
3.2.1 Autism Screening Questionnaire1 (ASQ; Berument, Rutter, Lord, Pickles
& Bailey, 1999)
The ASQ was developed as a screening instrument for autism and other PDDs, based on
current diagnostic criteria and for use with all age groups. It is a 40-item questionnaire
completed by the individual’s primary caregiver. The questions are based on the
Autism Diagnostic Interview-Revised (ADI-R; Lord, Rutter, & Le Couteur, 1994 – see
the following section) but have been modified to be easily understandable in a
questionnaire format. It includes questions on reciprocal social interaction, language
and communication, and repetitive and stereotyped behaviours. There are two versions,
one for individuals under the age of six and the other for those aged six and over. A
score of 1 is assigned for the presence of abnormal behaviour and a score of 0 for its
absence. The total score therefore ranges from 0 to 39 (an item on current language
level is not included in the score). Berument et al. found that a score of 15 or more on
the ASQ was the optimal cutoff point for differentiating ASDs from other diagnoses.
The ASQ shows good diagnostic validity and the correlation between the ASQ total
score and the ADI algorithm score is high (Berument et al., 1999).
In the current research, the ASQ was used mainly as a screening instrument to
determine whether or not individuals in the control group (or siblings of individuals
with ASDs and control siblings in the second study) displayed symptoms of autism. If a
1 The ASQ has now been published as the Social Communication Questionnaire, however as the version of the questionnaire used in this research was obtained from the authors prior to publication, the ASQ was deemed to be a more appropriate descriptor.
88
participant in one of these groups met the cutoff criterion for an ASD on the ASQ, the
ADI-R was then administered. The ASQ cutoff point was lowered to a more
conservative score of 10 (rather than 15) in this research, to ensure that both i) any
controls scoring highly on the ASQ did not meet ADI-R criteria for an ASD, and ii) any
individuals with mild ASD symptomatology met the criterion and were administered the
ADI-R.
3.2.2 Autism Diagnostic Interview – Revised (ADI-R; Lord et al., 1994)
The ADI-R is a modified version of the ADI (Le Couteur et al., 1989), which is a
standardised, semistructured interview for caregivers of individuals for whom autism is
a possible diagnosis. The diagnostic algorithm is based on ICD-10 (WHO, 1992)
criteria for autism, but it can also provide a DSM-IV (APA, 1994) diagnosis as the two
diagnostic systems are very similar. It has demonstrated good reliability and validity
(Lord et al., 1994). The duration of the interview for a practiced administrator is
approximately 1.5 – 2 hours. Special training is required for administrators and
approval for use of the instrument is given after completion of a test interview.
The ADI-R consists of five sections: opening questions, communication (both
early and current), social development and play (early and current), repetitive and
restricted behaviours (early and current), and general behaviour problems. Each item is
scored either 0 (behaviour not present), 1 (behaviour probably present but criteria not
fully met), or 2 (behaviour definitely present), and occasionally a score of 3 is used to
indicate extreme severity. An algorithm cutoff score determines whether an individual
meets diagnostic criteria within each of the three domains of abnormality (i.e.,
communication, social interaction, and repetitive/restricted behaviours). In order to
meet diagnostic criteria for autism, the individual must meet criteria in each of these
three domains, as well as exhibiting some abnormality in at least one area by 36 months
of age.
The application and utility of the ADI-R in diagnosing ASDs other than autism
and in differentiating ASD subtypes has not yet been investigated. However, as
individuals with clinical diagnoses of Asperger syndrome and PDDNOS were included
in the current research, a more lenient ADI-R diagnostic criterion was introduced such
that any individual exceeding the cutoff point in at least one of the three domains was
considered to have met criteria for an ASD. Section 4.3.2 in Chapter 4 describes the
89
results of comparisons between the “full criteria” and “partial criteria” groups in Study
One.
3.2 IQ measures
Two Verbal and two Performance subtests from the Wechsler scales (WPPSI-R, WISC-
III or WAIS-III, depending on the participant’s age) were used to estimate Verbal IQ
(VIQ) and Performance IQ (PIQ) respectively. VIQ and PIQ scores were estimated by
pro-rating sums of scaled scores based on the two subtests for each scale. Verbal
subtests were Vocabulary (providing definitions of words) and Similarities (identifying
the way in which two things are alike). Performance subtests were Picture Completion
(identifying the missing part of a picture) and Object Assembly (assembling pieces of a
puzzle to form a whole object). These subtests were chosen because they are
representative tests of verbal and non-verbal ability2, as well as being the least similar to
other measures in the test battery.
3.3 ToM measures
Three different false belief tasks, varying in level of difficulty, were selected as the
main ToM measures for both studies. Emphasis was placed on measures of false belief
as these have been the main focus of studies of the ToM-EF relationship. The tasks
chosen were all in common usage in the literature (unexpected contents/identity,
standard first-order false belief, and second-order false belief; these are all described in
detail below). A more advanced social cognition measure (Dewey Stories) was also
included because of expected ceiling effects on false belief tasks in older control
participants.
The three false belief tasks were administered in a hierarchy of difficulty, with
different starting points for participants of different ages. This was done mainly to
conserve time within the extensive test battery. The “simple” false belief task
(including unexpected contents and unexpected identity items, as described below) was
2 In the WISC-III (which was used with the most participants), the Vocabulary subtest loads .81 on the VIQ factor and .79 on the Verbal Comprehension (VC) index; and the Similarities subtest loads .75 on the VIQ factor and .72 on the VC factor. The Object Assembly subtest loads .66 on the PIQ factor and .69 on the Perceptual Organisation (PO) index; and the Picture Completion subtest loads .50 on the PIQ factor and .53 on the PO factor. Although the Block Design subtest has the highest loading on the PIQ and PO factors, it was not chosen because it is considered to be a measure of central coherence and therefore would have complicated the interpretation of PIQ scores.
90
considered the easiest of the three tasks, with first-order false belief (or unexpected
transfer) next in the hierarchy and second-order false belief the most difficult. The
more challenging nature of second-order false belief tasks was demonstrated by Perner
and Wimmer (1985) in typically developing children and Baron-Cohen (1989b) in
children with autism. The level of difficulty of “simple” and first-order false belief
tasks has generally been found to be more equal (Wellman et al., 2001), but findings are
consistent with the assumption that an individual who passes the first-order false belief
task is likely to have passed the simple false belief task.
The hierarchy of task administration operated such that participants were
administered either the simple false belief task (for children between 4 and 6 years of
age) or the first-order false belief task (for 7- to 16-year-olds) first, and then only
proceeded to the more difficult task(s) if the initial task was passed3. Pass or failure was
measured by performance on belief questions only, whereby a score of 2/3 or more for
the simple false belief task or 3/6 or more for the first- and second-order tasks was
considered a pass. If the initial (or subsequent) task was failed, failure was also
assumed on the more difficult task(s). If the 7- to 16-year-olds passed the first-order
false belief task, they were assumed to have also passed the simple false belief task;
however, if they failed the first-order false belief task, the simple false belief task was
administered.
Only a few studies have investigated the reliability and validity of false belief
tasks, with somewhat equivocal results. An initial study by Mayes, Klin, Tercyak,
Cicchetti, and Cohen (1996) found poor test-retest reliability for standard first-order
false belief tasks, however Hughes et al. (2000) found fair to moderate reliability across
a wider range of false belief tasks, and high reliability when aggregate scores were used.
Charman and Campbell (1997) found that a range of ToM tasks demonstrated moderate
reliability in individuals with learning disorders. In a sample of children with autism,
Grant, Grayson, and Boucher (2001) found good convergent validity of several false
belief tasks as well as high consistency across task versions.
3.3.1 Simple false belief task (Flavell et al., 1983; Perner et al., 1987)
This task included three items, one of which was an unexpected contents item and two
of which were unexpected identity items. These two types of item resemble each other
3 Of note, most studies which have examined the effect of task order on false belief performance have not found any significant order effects (e.g., Gordon & Olson, 1998; Hala et al., 2003).
91
closely. Because a higher number of trials per task is preferable in general, they were
grouped together and considered as different items of a single task. This has been done
in previous studies (e.g., Gopnik & Astington, 1988), and its validity was confirmed by
Wellman et al.’s (2001) meta-analysis, which showed no difference in the level of
performance on the two item types. It was also confirmed in this research, where strong
correlations were found across the three task items in the control participants of both
studies (on the belief questions, r ranged from .52 to .7, all ps < .01).
In the first task item (the unexpected contents item), the child is shown a box of
Smarties and asked, “What do you think is inside this box?” After s/he responds, the
box is opened and the child is shown that the box actually contains a pencil. The pencil
is then replaced in the box, and the child is asked, “What is really in the box?” (the
Reality question). S/he is then asked, “When you first saw the box, all closed up like
this, what did you think was inside it, Smarties or a pencil?” (the Own Belief question).
Finally, the participant is asked, “If I show X (parent/sibling) the box all closed up just
as I showed you, and I ask X what he/she thinks is in the box, what do you think X will
say, Smarties or a pencil?” (the Others’ Belief question)4. The other two items (the
unexpected identity items) involve the same questions, except the stimulus for the
second trial is a sponge which is spray-painted to look like a rock (after being asked
what s/he thinks it is, the child is then allowed to squeeze it, then asked what it really is,
and so on); and the stimulus for the third item is a black pen which contains red ink
(after being asked what colour s/he thinks the pen is, the experimenter writes with it to
show that it is red, then the child is asked what colour it is really, and so on). The child
is given a score of 1 or 0 for each of the Reality, Own Belief and Others’ Belief
questions, and the scores for the three trials are summed for each question type.
3.3.2 First-order false belief task (Wimmer & Perner, 1983; Baron-Cohen et al.,
1985)
For the current studies, rather than using puppets, a video was filmed in which six
independent scenes are depicted5. The task is introduced to participants by saying
“Now we are going to watch some short videos. Each video tells a story. After we
finish watching each video, I will ask you some questions about what happened in the
4 The order of control and belief questions has been found to have no effect on participants’ responses (Eisenmajer & Prior, 1991; Leslie & Frith, 1988). 5 It should be noted that Wellman et al.’s (2001) meta-analysis demonstrated that the medium in which false belief tasks are presented (e.g., video, puppets, real people) had no effect on performance.
92
story”. In four of the scenes, an object is placed in Location 1 by Actor 1, who then
leaves the room. Actor 2 moves the object to Location 2, and Actor 1 then re-enters the
room. The participant is then asked i) where Actor 1 will look for the object (the Belief
question), ii) where the object actually is (the Reality question), and iii) where Actor 1
placed the object at the beginning (the Memory question). In another one of the scenes,
Actor 1 places Object 1 in a covered bag and leaves the room, and then Actor 2 replaces
Object 1 with Object 2. The participant is asked i) what Actor 1 thinks is in the bag, ii)
what is actually in the bag, and iii) what Actor 1 placed in the bag in the beginning. In
the remaining scene, Actor 1 draws Picture 1 on a board and leaves the room, then
Actor 2 rubs out Picture 1 and draws Picture 2. The participant is asked i) what picture
Actor 1 thinks is on the board, ii) what picture is actually on the board, and iii) what
picture Actor 1 drew in the beginning. Each scene therefore follows the same basic
structure. In each case, both of the locations, objects or pictures are visible on the
screen when the participants are being asked the questions.
Participants are given a score of 1 or 0 for each of the Belief, Reality, and
Memory questions, and scores for each question type are summed over the six scenes.
In pilot testing, it was found that many participants gained a score of 1 on all items and
found the task easy. A discontinue criterion was therefore introduced whereby if
participants gained full marks for the Belief questions for three consecutive scenes, they
were not shown the remaining scenes and gained automatic credit for these. If this
discontinue criterion was met but participants had scored 0 on any of the Reality or
Memory questions, their overall score in these question categories was calculated on a
pro-rata basis – that is, a percentage correct was calculated for those items administered,
and then multiplied by six.
3.3.3 Second-order false belief task (Perner & Wimmer, 1985; Baron-Cohen,
1989b)
This task was also presented in video format, with each of the six scenes followed by
Belief, Reality, and Memory questions. In four of the scenes, Actor 1 places an object in
Location 1, and leaves the room, but spies on Actor 2 without Actor 2 knowing. Actor
2 moves the object to Location 2, and Actor 1 then re-enters the room. The participant
is then asked i) where Actor 2 thinks that Actor 1 will look for the object (the Belief
question), ii) where the object actually is (the Reality question), and iii) where Actor 1
placed the object at the beginning (the Memory question). In another one of the scenes,
93
Actor 2 draws a picture on a sheet of paper while Actor 1 watches, then Actor 1 leaves
the room. While Actor 1 secretly watches, Actor 2 decides to draw a different picture
instead. The participant is asked i) what Actor 2 thinks Actor 1 thinks the picture is, ii)
what the picture actually is, and iii) what Actor 2 drew in the beginning. In the
remaining scene, Actor 1 offers Actor 2 an orange and a banana, which are both placed
in a lunchbox. Actor 2 takes the orange, and then Actor 1 leaves the room. While
Actor 1 secretly watches, Actor 2 replaces the orange and takes the banana instead, and
eats it. Actor 1 then re-enters. The participant is asked i) what Actor 2 thinks that Actor
1 thinks she ate, ii) what she actually ate, and iii) what she took in the beginning.
Again, in each case, both locations, pictures or fruits are visible on the screen when the
participants are being asked the questions.
Participants are given a score of 1 or 0 for each of the Belief, Reality, and
Memory questions, and scores for each question type are summed across the six scenes.
The same discontinue criterion was used for this task as for the first-order false belief
task.
3.3.4 Dewey Stories (Dewey, 1991)
Dewey (1991) states that she composed this task in 1974 as an informal measure of
knowledge of social norms and human relations. It was chosen as a higher-level, more
advanced measure of social cognition than the false belief tasks. While Dewey (1991)
reports qualitative data on the unusual comments made by individuals with autism in
response to the stories, she does not report any quantitative scoring method or any
results from typically developing or other control samples. No other published study
has used the measure, and there have been no published investigations of its reliability
or validity. Its inclusion in this research can therefore be considered somewhat
experimental. While several of the story items appear to tap mentalistic understanding,
its validity as a measure of ToM remains to be investigated (see Section 4.3.2.2 of
Chapter 4), as it could be argued that the task may also be successfully performed
simply by drawing on knowledge of normative or common social behaviours.
The stimuli for the task are 7 stories, each one paragraph in length, which
describe a sequence of events containing certain social scenarios (Figure 3 contains an
example). The stories used in this study were taken directly from Dewey (1991),
however one story was shortened and another was substantially modified to be more
appropriate for an Australian sample. Two or more sections from each story are
94
underlined, and a pair of empty brackets follows each underlined part. Participants are
asked to rate each underlined behaviour according to how they think most people would
judge that behaviour if they witnessed it. They are asked to use the following scale:
Fairly normal behaviour in that situation [ A ]
Behaviour that is a little unusual in that situation [ B ]
Rather strange behaviour in that situation [ C ]
Very eccentric or shocking behaviour in that situation [ D ]
Although there are no set right or wrong answers for each rating, as a way of judging
the typicality of participants’ responses, each response was compared with norms
derived from 30 undergraduate psychology students. Responses of the normative
sample showed an equal split between the frequency of “B” and “C” responses on
several items, and so it was decided to place B and C in the same category of response.
The scoring system worked such that responses closer to the dominant normative
response were assigned a lower score. For items where A was the dominant response (n
= 8), participants who chose A scored 0, B/C scored 1, and D scored 2; and for items
where B/C was the dominant response (n = 9), A scored 1, B/C scored 0, and D scored
1. There were no items where D was the dominant response. Scores were summed
across items to produce a total score, on which lower scores represented a higher social
awareness.
F
B
o
Emily, age nineteen, overslept on the morning of her aeroplane trip. When she woke up, there was
just enough time for her to dress and get to the airport, so she skipped her breakfast. [ ] At noon, the
steward came around with lunch, but Emily was so hungry by then that one portion did not satisfy
her. She watched a little girl across the aisle toy with her food, complaining “I can’t eat it.”
Apparently, the father didn’t want any more, because he told the child to just leave it. Emily leant
across the aisle and said, “If your little girl doesn’t want her tray, can you pass it over for me?” [ ]
igure 3. An example of a Dewey Story.
3.4 EF measures
ecause one of the central aims of this research was to conduct a thorough investigation
f the EF profile characteristic of ASDs, as well as to examine the relationship of
95
various EF components to ToM (these aims are discussed in the next chapter), a strong
emphasis in the test battery was placed on measures of EF. Given the difficulties
following from the task impurity of most widely used EF tests (as discussed in the
previous chapter, Section 2.2.1), specific assessment of component processes using
tasks with high construct validity was given priority in task selection. The component
process approach to EF assessment has been strongly advocated by several authors (e.g.,
Hill, 2004; Ozonoff, 1995a, 1997a, 1997b, 2001). The tasks chosen are relatively
simple and/or include control conditions allowing precise delineation of the underlying
EF process(es) involved, although for some EF domains (e.g., planning), this was not as
easily achievable due to both the nature of the component and the availability of “pure”
tasks. A wide range of EF components was assessed, including planning (measured by
the Tower of London), set-shifting or cognitive flexibility (the Intra-dimensional, Extra-
dimensional set-shifting task), inhibition and its interaction with working memory
(Response Inhibition and Load task and Opposite Worlds), relational reasoning
(Relational Complexity), and generativity (Pattern Meanings, Uses of Objects and the
Stamps task). It was desirable to test each EF domain using both verbal and non-verbal
response modes, which was possible for the inhibition and generativity domains. The
child-friendliness of the tasks was another factor considered in EF task selection. It was
important for the tasks to be appropriate for a fairly large age range, so tasks with a low
floor and high ceiling were regarded as preferable.
3.4.1 Tower of London (Culbertson & Zillmer, 1998b; Shallice, 1982)
The Tower of London (ToL) was first designed as a cognitive measure by Shallice
(1982), who found that it was performed poorly by patients with frontal lobe lesions.
The ToL’s sensitivity to frontal dysfunction has been supported in a number of
subsequent studies using both adult clinical samples (Carlin et al., 2000; Owen et al.,
1990) and head-injured children (Levin et al., 1994, 1996). Shallice proposed that the
ToL specifically measures planning ability, which may be defined as the ability to
generate, select, organise, integrate, and monitor behaviours needed to achieve a future
goal (Culbertson & Zillmer, 1998a; Lezak, 1995). The validity of the ToL as a measure
of planning was supported by Shallice’s (1982) finding that ToL performance did not
covary with measures of visuospatial ability or working memory, although subsequent
studies have found that working memory and inhibition may also be involved in task
performance (e.g., Welsh, Satterlee-Cartmell, & Stine, 1999).
96
A wide range of administration and scoring procedures for the ToL have been
used in different studies. Three groups of researchers have published proposed
standardised versions of the ToL for use with paediatric populations (Anderson,
Anderson, & Lajoie, 1996; Culbertson & Zillmer, 1998b; Krikorian, Bartok, & Gay,
1994). Both Anderson et al. and Krikorian et al.’s versions require the readministration
of failed items, whereas Culbertson and Zillmer’s version only requires the child to do
each problem once, and uses the number of extra moves made as its main dependent
measure. Culbertson and Zillmer (1998b) argue that the readministration of failed items
significantly increases the amount of on-task time, which is a liability when assessing
younger children and clinical populations with limited attentional capacities. It can also
provoke frustration and distress, leading to decreased motivation and co-operation.
Their version has demonstrated adequate test-retest reliability as well as good criterion-
related, diagnostic and construct validity (Culbertson & Zilmer, 1998a, 1998b).
For these reasons, administration and scoring procedures for the version of the
ToL used in this research were based on those outlined by Culbertson and Zillmer
(1998b). The major differences were that the floor was lowered by including 1- and 2-
move items (rather than beginning with 3-move items); there were four items at each
level of difficulty instead of three; the instructions were slightly modified to encourage
participants to plan moves in advance; and scores were adjusted (see below) to account
for participants who completed problems in fewer moves than the minimum number
because they broke the task rules.
Participants are presented with a tower structure consisting of three wooden
posts of descending heights mounted on a wooden base. Three coloured discs (red,
black and white) are placed on the posts in a standard starting position (see Figure 4).
The participant is then required to rearrange the three coloured discs on the posts so that
the new configuration corresponds to the pattern presented on a 21cm x 15cm stimulus
card. The participants are informed that this must be accomplished in the minimum
number of moves, which is told to them verbally as well as being written at the top of
the stimulus card. In addition, they are told that they must adhere to the following four
rules: i) they can only use one hand to move the discs, ii) they can only move one disc at
a time, iii) discs cannot be placed on the board or table - only on the posts, and iv) they
cannot put more discs on a post than it will hold. Examples of breaking a rule are
demonstrated, each time with the experimenter saying “You can’t do this”.
97
Figure 4. The starting configuration for the Tower of London stimuli.
Participants are given one 1-move and two 2-move practice examples, during which if a
rule is broken or extra moves made, the rules are reiterated and the correct solution
demonstrated. The following instructions are then given:
“Now I am going to set up more disc patterns and see if you can make them on your
board in as few moves as possible. You may find that some of the patterns are difficult,
but do the best you can. Each pattern can be solved. You should look carefully at the
pattern and the board and plan the best move to start with. Take your time planning, as
each move you make counts towards the total. If you think you can’t finish it in the
correct number of moves, then just keep going and try and do it in the fewest number of
moves you can.”
Items range in difficulty from 1 move to 7 moves, with four items at each level of
difficulty. Participants aged between 4 and 13 begin with 1-move items, and
participants aged 14 and over begin with 3-move items (and are given automatic credit
for 1- and 2-move items if they complete at least two 3-move items in the minimum
number of moves). If the first item at a new level of difficulty is failed (either by
breaking the rules or using too many moves), the correct solution is demonstrated.
There is a time limit of 2 minutes on each item, after which the item is discontinued.
Testing is discontinued if participants fail all items at a particular level of difficulty.
Remaining items are assumed to have been failed and are assigned the maximum total
number of moves (i.e., 20).
Five scores are computed for each test item. These are listed in Table 1. The
sum of extra move scores and adjusted extra move scores is calculated for each block of
items (i.e., each level of difficulty) as well as overall. The total number of problems
completed in the minimum number of moves is also computed (i.e., the total number of
98
problems with an adjusted extra move score of 0). Although Culbertson and Zillmer
(1998b) also calculated initiation and solution times, these were not used in this research
as not all participants were administered all items (due to the different starting points for
different ages and the discontinue rule), making mean times (either overall or block by
block) difficult to calculate and analyse in a meaningful way.
Table 1. The five scores computed for each item on the Tower of London
1. Total number of rule violations Included moving 2 discs off the posts at the same
time, placing more discs on a post than it would
hold, and placing discs on the board or table
2. Extra moves Total number of moves – minimum moves*
3. Adjusted extra moves Adjusted moves – minimum moves, where adjusted
moves = total moves + (2 x no. of rule violations)
4. Extra move score Designed to avoid excessive inflation of the “extra
moves” index by an extreme number of total
moves, this is calculated as follows:
• If extra moves = 0, extra move score = 0
• If extra moves = 1-5, extra move score = 1
• If extra moves ≥ 6, extra move score = 2
5. Adjusted extra move score Identical to the extra move score except using
adjusted extra moves (3) instead of extra moves (2)
*If the total number of moves exceeds 20, it is reduced to 20 to avoid inflation of the
“extra moves” index by excessive moves on a limited number of items. For example, if
a participant executes 24 moves on a 7-move problem, then the score would be
calculated as follows: 20 - 7 = 13. In addition, the total number of moves is assigned a
value of 20 for any item not solved within 2 minutes.
3.4.2 Intra-dimensional, Extra-dimensional (IDED) Set-shifting task (Owen et
al., 1993)
The original version of this task, which forms part of the CANTAB (Cambridge
Neuropsychological Test Automated Battery), was designed as a WCST-like
computerised measure of attentional set-shifting. In comparison to the WCST, the
IDED set-shifting task is simpler to allow participants of a wider age and ability range
to participate, and involves a series of stages containing a number of internal control
99
conditions to aid the elucidation of the mechanisms involved in successful or
unsuccessful task performance. It has demonstrated fair test-retest reliability (Lowe &
Rabbitt, 1998). Using this task, it was found that patients with Parkinson’s disease and
with frontal lobe damage, but not patients with temporal lobe damage, demonstrated an
inability to shift attention between two perceptual dimensions at the “extra-dimensional
shift” stage (Downes et al., 1989; Owen, Roberts, Polkey, Sahakian, & Robbins, 1991).
However, observations that patients with frontal lobe dysfunction and
Parkinson’s disease may fail the task for different reasons led Owen et al. (1993) to
develop a modified version of the CANTAB procedure, which was designed to
distinguish whether impairments in attentional set-shifting ability are caused by an
inability to release attention from a relevant stimulus dimension (Perseveration), or an
inability to re-engage attention to a previously irrelevant dimension (Learned
Irrelevance). This version (described further below) therefore allows even further
breakdown of the processes involved in set-shifting performance, making it attractive
for inclusion in the current research. Using this modified IDED task, Owen et al. (1993)
found that the difficulty with extra-dimensional shifting demonstrated by patients with
frontal lesions was caused by perseveration to the previously relevant dimension,
whereas patients with Parkinson’s disease tended to show learned irrelevance. When
Turner (1997) used the task with children with autism, she found that low-functioning
participants demonstrated significantly more errors at the extra-dimensional shift stage
of the Perseveration condition, but not the Learned Irrelevance condition.
The modified version of the IDED set-shifting task includes two task conditions,
one intended to assess perseveration and the other to assess learned irrelevance. In both
conditions, each trial consists of two patterns appearing on a computer touchscreen
(their positions randomly alternating between four rectangular boxes to the top, bottom,
left and right of centre of the screen), and the participant is required to choose which
one is “correct” according to an unspecified rule, with feedback provided by the
computer. Participants are given the following instructions:
“This is a game where you have to work out the rule for choosing the right answer. On
the screen you are going to see two patterns. The patterns will appear in any two of four
boxes. One of the patterns is right and the other is wrong, and you must tell the
computer the one you think is right. You do this by touching the pattern on the screen.
There is a rule which you can follow to make sure you make the right choice every time.
The computer will not tell you the rule – you will have to work it out for yourself. To
100
begin with, there is nothing on the screen to tell you which of the patterns is correct so
when you choose your first answer you will just have to guess. However, the computer
will give you a message after each try to tell you whether you are right or wrong. If you
are right, it will come up with the word “CORRECT”, written in green, and if you are
wrong, it will say “INCORRECT”, written in red. The computer will be keeping track
of how well you are doing. When the computer can tell that you know the rule, the
computer will then change the rule, but it will not tell you that the rule has changed.
You will have to work out the new rule. That won’t happen very often. Do you have
any questions before you start?”
Each condition comprises 8 stages presented in the same fixed order: a simple
discrimination (SD) and reversal (SDR), then a compound discrimination (CD) and
reversal (CDR), then an intra-dimensional shift (IDS) and reversal (IDR), and finally an
extra-dimensional shift (EDS) and reversal (EDR). Participants can only proceed to the
next stage after reaching the criterion of 6 consecutive correct responses.
In the Perseveration condition (see Figure 5), the task begins with subjects being
required to learn which of two geometrical shapes is correct (SD condition). The
subject is then required to reverse the learnt rule and respond to the previously incorrect
stimulus in the target stimulus dimension (SDR). The next stage introduces an
additional stimulus dimension, white lines, which are paired with the shapes. At this
stage the same shape remains correct, with the nature of the lines being irrelevant (CD).
Once that is learnt, the subject is again required to reverse the learnt rule and respond to
the other shape (CDR). The next stage is the IDS stage, where the subject is presented
with new exemplars for both of the stimulus dimensions (shapes and lines). Although
the exemplars are different to the two previous stages, the relevant stimulus dimension
(shape) remains the same. In the EDS stage, the previously irrelevant stimulus
dimension (lines) is replaced by a new stimulus dimension which becomes relevant
(solidity), and the previously relevant dimension (shape) becomes irrelevant. Thus,
participants must shift their attention from a previously relevant to a new stimulus
dimension, and failure reflects perseveration to the previously relevant dimension.
The Learned Irrelevance condition (see Figure 6) proceeds as for the
Perseveration condition for the first 6 stages, except that colour is the relevant
dimension and number the irrelevant dimension. However, in the EDS stage, the
relevant dimension (colour) is replaced by a previously irrelevant dimension (number),
and a new dimension (size) becomes the irrelevant dimension. Hence, participants must
101
shift their attention to a previously irrelevant stimulus dimension, and failure reflects
learned irrelevance associated with the previously irrelevant dimension.
Stage Stimuli Relevant dimension
Irrelevant dimension
SD
Shape
-
SDR
Shape
-
CD
Shape
Lines
CDR
Shape
Lines
IDS
Shape
Lines
IDR
Shape
Lines
EDS
Solidity
Shape
EDR
Solidity
Shape
Figure 5. Stimuli for the Perseveration condition of the IDED set-shifting task (see text
for explanation of abbreviations). The correct choice is always displayed on the left.
102
Stage Stimuli Relevant dimension
Irrelevant dimension
SD
Colour
-
SDR
Colour
-
CD
Colour
Number
CDR
Colour
Number
IDS
Colour
Number
IDR
Colour
Number
EDS
Number
Size
EDR
Number
Size
Figure 6. Stimuli for the Learned Irrelevance condition of the IDED set-shifting task
(see text for explanation of abbreviations). The correct choice is always displayed on
the left.
103
Failure to achieve the criterion of 6 consecutive correct responses within 50
trials at any one stage results in discontinuation of the test. There is a 1000ms interval
between successive trials. Each condition lasts approximately 10 minutes and the two
conditions are separated by at least 30 minutes of unrelated tests. Unlike Owen et al.’s
(1993) procedure, in the current study the dimensions used in each condition (i.e.,
shape, lines, solidity, colour, number, and size) were consistent across participants (as
shown in Figures 5 and 6), rather than counterbalancing the dimensions across the two
conditions. The only other difference between the current and Owen et al.’s version
was that the Perseveration condition was presented first for all participants (rather than
the order of conditions being counterbalanced across participants).
So that participants could be effectively compared across conditions, the main
index of performance was the number of “errors to criterion” within each stage of the
task (this was also the main index of performance used by Owen et al., 1993). If the test
was discontinued because the criterion of 6 consecutive correct responses was not met
within 50 trials, a value of 25 (the value expected with random responding) was
assigned for the errors to criterion score for subsequent stages of the task which were
not administered.
3.4.3 Response Inhibition and Load (RIL) task
The basic idea for this non-verbal computerised test of inhibition, which was created by
the author, was taken from Drewe (1975) but with substantial modifications and
additions. Drewe’s study included two types of task, one involving the requirement to
press a button in response to one type of stimulus but not another (otherwise known as a
“Go-Nogo” task) and the other requiring the participant to inhibit the prepotent response
to match stimuli of the same colour by pressing a button which was opposite in colour
to the stimulus. This latter task type had a control condition which required participants
to press a button which was the same colour as the stimulus. The inclusion of a control
condition was desirable for the current research, as subtraction of scores on this
condition from scores on the inhibition condition allows more precise identification of
the level of performance on the inhibition task condition without the confounding
effects of non-inhibitory processes such as speed of processing and motor coordination.
This “non-matching to target” paradigm was therefore adopted for this research
but modified in order to improve aspects of the methodology. In Drewe’s task, the
stimulus and two response buttons (one red and one blue) were always present and
104
visible (with the stimulus button lighting up in either red or blue), whereas in the current
task a touch screen was used so that the stimulus was presented only briefly before the
response buttons were presented. This also allowed the two coloured response buttons
to change sides randomly from trial to trial, ensuring that the participant was responding
on the basis of the colour of the response button rather than its spatial location. The
inhibition task condition was also modified so that the colours of the stimulus and
response buttons were different to the control condition (which preceded it). This was
to avoid confounding inhibition with set-shifting (specifically, reversal), as the use of
exactly the same stimuli in control and inhibition conditions means that the inhibition
condition then requires the participant to reverse the stimulus-response contingencies
and therefore directly “shift set” (Ozonoff et al., 1994).
An important addition to the task was an extra condition which involved an
increased working memory load, thereby allowing evaluation of the interaction between
inhibitory capacity and working memory. This condition was included in order to
examine two hypotheses: i) that children with autism (and/or their siblings) may be
impaired only on tasks which combine inhibitory and working memory demands, and ii)
that false belief measures show correlations with performance on tasks that combine
inhibitory and working memory demands, but not with each capacity individually.
Performance on the working memory load condition was compared with that on the
inhibition condition, to assess the specific effect of the working memory load, and also
with performance on the control condition, to assess the combined effect of inhibition
and working memory requirements. The three task conditions are described below.
Condition 1 (Control condition): In this condition, either a pink or green
stimulus circle (approximately 5cm in diameter) appears at the top of a computer touch
screen for 250ms, and then two smaller response circles (approximately 3.5cm in
diameter), one pink and one green, appear simultaneously at the bottom left and right
corners of the screen. Participants are instructed to touch the response circle which is
the same colour as the stimulus circle. Participants have 4s to respond before the
response circles disappear and the next trial begins. An equal number of pink and green
stimulus circles are presented, and the order of presentation is random. As already
mentioned, the response circles change sides randomly (i.e., the pink circle can appear
on either the right or left), to ensure that participants are responding to the colour of the
circle rather than simply its position on the screen. Participants are required to use one
hand only to respond. Performance indices are the percentage of errors (i.e., responding
to the wrong coloured stimulus), and the median RT for correct trials.
105
Condition 2 (Inhibition condition): This condition is identical to Condition 1
except that the colours of the stimulus and response circles are purple and yellow, and
the participant is required to touch the response circle which is the opposite colour to
the stimulus circle. Hence, if the stimulus circle is purple, participants must touch the
yellow response circle, and vice versa. As for Condition 1, performance indices are the
percentage of errors and the median RT for correct trials.
Condition 3 (Working Memory Load condition): In this condition, instead of the
stimulus being a circle, it is either a square, triangle or cross. As in Condition 2,
participants must touch the response circle which is opposite in colour to the stimulus
shape (the colours in this condition are orange and grey). However at random intervals,
between trials, the three shapes are displayed on the screen and the participant must
touch the shape which was presented in the most recent trial. This occurs for 25% of
the trials. The participant must therefore recall the shape of the stimulus as well as
inhibiting the prepotent tendency to respond to the same colour. Performance indices
for the responses to the colour of the stimulus are identical to those in Conditions 1 and
2. For the questions about the shape of the stimulus, performance is measured by the
percentage of errors.
In each condition, participants perform 7 practice trials, during which any errors
are pointed out verbally and corrected. Following the practice trials, there is a pause
during which the participant may ask any further questions. The 60 critical trials then
proceed, during which every third error is pointed out and the participant reminded of
the task rules. The inter-trial interval is 1000ms in all conditions.
3.4.4 Opposite Worlds (Manly, Robertson, Anderson, and Nimmo-Smith,
1998)
This task was selected as an additional measure of inhibition, where unlike the RIL task,
a verbal response is required. Opposite Worlds is a subtest of the Test of Everyday
Attention for Children (TEA-Ch; Manly et al., 1998), and is similar in design to
Gerstadt et al.’s (1994) Stroop-like day-night task, but instead involves reading the
number “1” as “2” and vice versa. It demonstrates good test-retest reliability (Manly et
al., 2001). It forms part of the “attentional control/switching” factor in the TEA-Ch, but
the naming of this factor reflects the executive nature of the factor rather than accurately
describing the requirements of the tasks that load on it. Manly et al. (1998, 2001)
consider Opposite Worlds to be a test of verbal inhibition, pointing out that as
106
participants are required to switch from the Opposite to the Same World (control)
condition as well as vice versa, performance on the Opposite World condition may be
attributed to the requirement to inhibit a prepotent verbal response rather than the
demands of task switching. The task has displayed good construct and convergent
validity, correlating significantly with other measures of inhibition (Manly et al., 1998).
Opposite Worlds is administered only to children who are able to read the
numbers 1 and 2. The stimuli are yellow squares linked together in an undulating
pattern on a black background, with each square containing either a 1 or a 2. The task is
introduced by saying: “In this test there are two sorts of world. There is the Same
World, where everything is as you would say it here, and the Opposite World, where
you have to say the opposite of what you would say here”. An example page with two
Same World examples at the top and two Opposite World examples at the bottom is
shown to the participant. The experimenter points to the beginning of the first Same
World example and says: “Here I would say “Start, one, one, two, two, one, Stop”. The
child is encouraged to complete the same item and then the other Same World example.
While the child reads the numbers, the experimenter points to each square in turn, and
does not move onto the next square until the child has said the correct number. After
successful completion of the Same World examples, the experimenter then points to the
first Opposite World example and says: “We’re now going to the Opposite World where
we have to say the opposite. Here, when we see a one we have to say “two”, and when
we see a two we have to say “one”. This is how to do it: “Start, one, one, two, one, two,
Stop”. Both examples are then completed by the child.
The participant then completes the four test trials in the order: Same World,
Opposite World, Opposite World, Same World. S/he is reminded of the instructions at
the beginning of each trial. The time taken to complete each trial is recorded from the
time the child says “Start” to the time s/he says “Stop”. The number of errors for each
trial is also recorded, with an error being defined as any occasion upon which the child
says a “1” when required to say “2”, or vice versa. A total time score (summing the
time taken across the two trials) and total error score (summing the errors made across
the two trials) are calculated for each of the Same and Opposite World conditions.
3.4.5 Relational Complexity (Waltz et al., 1999)
This task was included as a measure of relational reasoning, which refers to the ability
to “form and manipulate mental representations of relations between objects and
107
events” (Waltz et al., 1999, p. 119). While relational reasoning is not often included in
lists of EF components, it was assessed mainly to test Halford’s notion that limited
capacity to integrate multiple relations (i.e., relational complexity) may underlie failure
on false belief tasks (Halford, 1993; Halford et al., 1998; see Section 2.3.1.2 in the
previous chapter). Halford et al. (1998) argue that working memory capacity may be
best defined in terms of the complexity of the relations that can be processed in parallel,
and therefore the Relational Complexity task may also be considered a test of working
memory – a domain which is more often considered an aspect of EF. Waltz et al.
(1999) found significant impairments in patients with prefrontal cortical damage on
their Relational Complexity task, and proposed that failures on various types of EF task
could be accounted for by a deficit in relational integration.
The task is similar in format to the Raven Standard Progressive Matrices Test, in
which the missing part of a pattern must be chosen from six alternatives. The current
version is based on Waltz et al.’s (1999) adaptation but has more levels of difficulty,
more pictures within each item, and more alternative answers. In this version, each
problem consists of a 3 x 3 matrix of square-shaped simple geometric pictures, with the
bottom right-hand corner picture missing. Participants are asked to select the missing
picture from eight alternatives (see Figure 7). Problems vary in the number of relational
changes (e.g., in shape, size, rotation), occurring over horizontal and/or vertical
dimensions of the matrix, which must be attended to while selecting the missing picture.
Nonrelational (Level 0 complexity) items consist of identical pictures, with participants
simply having to choose the matching picture from the eight alternatives. The highest
level of difficulty for relational problems are of Level 4 complexity – requiring the
simultaneous processing of 4 relational changes (see Figure 8). In order to raise the
ceiling of the task, some more difficult items were also included where the relevant
stimulus dimensions do not necessarily vary in a consistent way across the vertical or
horizontal dimensions of the matrix (see Figure 9). An additional item at the end of the
task consists of a matrix with four missing pictures, and participants are required to
move four cut-out pictures into their correct places.
There are 4 problems at each level of difficulty. Participants are instructed to
take their time and point to the correct answer when they are sure of it. They are given
a maximum of two minutes for each problem. A score of 1 or 0 is given for each trial.
They are given three minutes to solve the final problem with the four missing pictures.
The task is discontinued if the participant fails all four problems at a particular level of
108
difficulty, with remaining items assumed to have been failed. The sum of correct
responses is calculated for each level of difficulty and overall.
Figure 7. Example of a Relational Complexity item with 1 relational change.
Figure 8. Example of a Relational Complexity item with 4 relational changes.
109
Figure 9. Example of a more difficult Relational Complexity item without consistent
relational changes.
3.4.6 Pattern Meanings (Wallach & Kogan, 1965; Turner, 1999)
Tests of generativity measure the ability to produce multiple novel responses
spontaneously following a single cue or instruction. In this research, the generativity
domain was tested using three different tasks because this aspect of EF has been under-
researched in autism, despite studies demonstrating its potential ability to explain
several symptoms of autism (e.g., Jarrold et al., 1996; Turner, 1997). There are three
basic types of generativity task: word fluency (requiring the participant to generate
words beginning with a certain letter or belonging to a certain category), design fluency
(where participants must produce abstract designs or patterns), and ideational fluency
(requiring generation of uses for objects or interpretations of abstract line drawings).
Word fluency was not tested in this research, mainly because it relies heavily on
vocabulary, making it difficult to disentangle reasons for poor performance (particularly
in autism, where verbal ability is typically impaired). A special emphasis was placed on
ideational fluency, with two tasks of this capacity included, as Turner (1999) found
110
particularly poor performance on ideational fluency tasks in both low- and high-
functioning individuals with autism.
Pattern Meanings is one of the measures of ideational fluency and requires
verbal generativity. The stimuli are five meaningless line drawings, taken from Wallach
and Kogan (1965) and also used by Turner (1999), which were printed on individual
14.3cm x 9.2cm laminated cards (see Figure 10 for an example). An additional drawing
was used for a practice item. Administration procedures were identical to those used by
Turner (1999), except that participants were given 90s instead of 150s to generate
responses for each item. This shorter interval was introduced following pilot testing, in
which it was found that participants tended to produce only a very small number of
responses in the last minute, and would often become restless and impatient or
inattentive. Before presentation of the practice stimulus, participants are told that the
task is one in which they will be shown some different patterns and asked to think of all
the things the pattern looks like, or what it could be. Participants are then shown the
practice stimulus card (see Figure 11) and asked “What could this be?” Any appropriate
response is reinforced and the participant is encouraged to think of other things the
pattern looks like. The experimenter also makes the following suggestions (if they have
not already been provided by the participant): “a hedgehog”, “someone with spiky hair”,
“sparks from a fire cracker”, and “a brush”. Participants are told that they are allowed
to turn the cards around and view them from any orientation. They are then given the
test stimuli one at a time, and for each one asked “What could this be?”. Stimuli are
presented in a random order.
Figure 10. One of the five test stimuli for the Pattern Meanings task.
111
Figure 11. The practice stimulus for the Pattern Meanings task.
Scoring procedures were similar to those used by Turner (1999), but an extra
“uninterpretable response” category was added. This category was introduced because it
was found during scoring that a number of responses could not be classified in any of
the other categories. Each response was therefore classified as belonging to one of the
following five scoring categories, and the number of responses in each category was
summed across the five test items:
1. Correct response: A response which represents a plausible interpretation of the
pattern.
2. Incorrect response: A response that represents an inappropriate or implausible
interpretation of the pattern (e.g., for the pattern displayed in Figure 10: “this could
be a shoe”).
3. Repetition: A response which is a repetition of a previous response (for the current
stimulus or a previous stimulus).
4. Redundant response: A response that varies from a previous response only in terms
of one minor element or feature of the response (e.g., “two hills”, “two mountains”,
“two sand-hills”, etc.)
5. Uninterpretable response: A nonsensical response, which cannot be interpreted as
fitting into any of the above categories (e.g., “up and down”).
As some unusual responses were sometimes difficult to classify, the scoring of Pattern
Meanings (and the Uses of Objects task described below) was more subjective than
112
other tasks in the protocol. Across all types of fluency tasks used in her study, Turner
(1999) reported 85% inter-rater agreement and kappa values in excess of .70, indicating
satisfactory inter-rater reliability. Because the version of Pattern Meanings used in this
study employed slightly different scoring criteria from Turner, inter-rater reliability of
this version was calculated using a subset of data from 22 participants (sampled
randomly from the ASD and control groups in Study One and the ASD sibling and
control sibling groups in Study Two). There was 93.3% agreement between the two
raters and Cohen’s kappa was .81, indicating excellent inter-rater reliability for this
version.
3.4.7 Uses of Objects (Wallach & Kogan, 1965; Turner, 1999)
Uses of Objects also measures ideational fluency and requires verbal responses. In this
task, the stimuli are six different objects. Three objects are “conventional” items with
well-established functions (a pencil, a brick, and a mug), and three are
“nonconventional” items with no clear or established function (a piece of plain navy
blue material measuring 14 x 51 cm, a 50 cm length of dowelling, and a 90 cm long
piece of clothing elastic). As with Pattern Meanings, administration procedures
matched those used by Turner (1999), but again with the shorter 90s interval in which to
provide responses. The task is introduced as one in which participants will be asked to
think of all the ways in which some different objects could be useful. Participants are
then asked “For example, how could we use a newspaper? Tell me something useful we
could do with it”. Any appropriate suggestions made by the participant are praised and
further responses encouraged. The examples “you could use it to start a fire”, “you
could roll it up and swat flies with it”, and “you could use it to wrap a present” are
provided by the experimenter if not already produced by the participant. Participants
are then asked to think of as many uses as they can for the six different objects, one at a
time. For each of the conventional items, the experimenter gives two examples, one
representing the object’s established function (e.g., “you could use a mug to drink
from”), and one that is more imaginative (e.g., “you could use a mug as a vase for
flowers”). For the nonconventional items, the experimenter gives just one imaginative
example (e.g., “you could use a piece of material to wrap up pencils if you wanted to
carry them”). After the examples are provided, the participants are asked to say all the
other ways in which the object could be useful. The objects are presented in the same
order for each participant (pencil, dowel, brick, material, mug, elastic).
113
Scoring procedures were again similar to those used by Turner (1999) but, as
with Pattern Meanings, an extra “Uninterpretable responses” category was added. In
addition, a “Non-Useful responses” category was introduced as it was found during
scoring that many responses were plausible things that could be done with the object,
but that did not serve any useful purpose (e.g., “you could sharpen a pencil”). Hence,
each response was categorised into one of the following six scoring categories:
1. Correct response: A response which represents a plausible use for the object.
2. Incorrect response: A response that represents an inappropriate or implausible use
for the object (e.g., for the brick: “eat it”).
3. Repetition: A response which is a repetition of either one of their own previous
responses (for the current object or previous objects) or one of the examples.
4. Redundant response: A response that varies from a previous response only in terms
of one minor element or feature of the response (e.g., for the brick: “to build a
garage”, “a shed”, “a factory”, etc.)
5. Uninterpretable response: A nonsensical response, which cannot be interpreted as
fitting into any of the above categories (e.g., for the piece of fabric: “blow down”).
6. Non-useful response: A response which describes something plausible that could be
done to or with the object, but which does not include a useful purpose for the object
(e.g., for the piece of elastic: “stretch it”).
The number of responses in each category was summed separately for conventional and
nonconventional items, as well as overall. Inter-rater reliability for Uses of Objects was
calculated using a subset of data from 23 participants (again sampled randomly from the
ASD and control groups in Study One and the ASD sibling and control sibling groups in
Study Two). There was 86.8% agreement between the two raters and Cohen’s kappa
was .76, indicating good inter-rater reliability.
3.4.8 Stamps task (Frith, 1972)
This task was based on one used by Frith (1972) as a measure of the spontaneous self-
generation of underlying rules in patterns. It was considered a test of design fluency in
this research despite the fact that Frith did not label her task in this way, as it is a non-
verbal task requiring participants to produce multiple novel responses. The task differs
from standard design fluency measures in that it involves producing patterns from a set
114
of materials rather than drawing abstract designs. While it has been used far less
frequently than other design fluency tasks, its scoring system allows analysis of a
number of different processes underlying task performance, making it amenable to a
component process approach. In addition, Frith (1972) demonstrated interesting results
with the task in children with autism, who tended to rigidly adhere to the same
underlying pattern rules, used a restricted range of available materials, and did not
generate original patterns.
The task procedure was based on Frith (1972), with some minor procedural and
scoring modifications. Participants are provided with four stamps of different shapes
and colours, and a piece of paper with a line of 16 boxes on it. They are asked to make
whatever pattern they like with the stamps, putting one stamp in each box. There are
eight trials, four using only two of the stamps and four using all four stamps. If the
child does not use all the stamps available during the first eight boxes of a trial (i.e.,
only uses one stamp on the two-stamp trials or less than four stamps on the four-stamp
trials), at that point s/he is reminded that there are more stamps available. The two-
stamp and four-stamp trials are presented alternately. The trials are divided up into two
blocks, separated by at least half an hour, with each block consisting of four trials (two
trials with two stamps, and two trials with four stamps).
Four types of scores are calculated for each trial:
1. Complexity. Rules are defined as consistently recurring sub-units of a fixed number
of elements (e.g., the pattern red/green/red/green etc. has an underlying alternation
rule as two elements are repeated over and over again; the pattern red/red/red/red
etc. has an underlying rule to repeat a single element; and the pattern
red/green/black/blue/red/green/black/blue etc. has the underlying rule to repeat a
group of four elements). Ratings of complexity are based on the number of
elements contained in a sub-unit, using the following scale:
i) repetitions of single elements are given the lowest rating of 1
ii) repetitions of two single elements (i.e., alternations) are given a rating of 2
iii) repetitions of three or four single elements are given 3
iv) on two-stamp trials, if two stamps are used in the sequence, but the pattern
consists of more than just an alternation (e.g., red/red/green/red/green/
green), a score of 3 is given
v) on four-stamp trials, if three or four stamps are used but the pattern consists
of more than just cycling through the three or four elements (e.g.,
115
red/red/green/black/black/blue/red/green/green/black/blue/blue), then a score
of 4 should be given.
In a case where a single rule can not account for the whole sequence of 16 items, but
only for part of it, the rule must account for at least one half of the sequence in order
to receive its score. If this criterion is not met, the pattern is considered
unidentifiable and given a rating of 1.
2. Rule adherence. All sequences which can be completely accounted for by a single
rule (i.e., a repeated sub-unit or a mirror-reversed pattern) are given a score of 1.
All sequences which are irregular in any way, including those with predominant or
unidentified rules, are given a score of 0.
3. Restriction. In the four trials where four stamps are used to build a pattern, a score
of 1 is given (for each trial) if the child uses fewer than the four stamps available. A
score of 1 is also given if the child only uses one stamp on two-stamp trials.
4. Originality. Any sequence that occurs only once in all of the trials is considered
“original” and given a score of 1. This score is only given if the original sequence
follows an identifiable pattern. If the “original” sequence is random, it scores 0.
Scores are summed across the eight trials to produce overall complexity, rule adherence,
restriction and originality scores for each participant.
3.5 Behavioural measures
Autistic symptomatology includes impairments in social interaction and
communication, and repetitive behaviours. Social and communication behaviours were
considered together in the current research, for reasons described below in Section
3.5.2.2. Thorough measurement of repetitive behaviours was emphasised, as discussed
in the introduction to Study One (see Section 4.1.1 in Chapter 4).
3.5.1 Measures of repetitive behaviour
3.5.1.1 Repetitive Behaviours Questionnaire (RBQ)
The RBQ was developed by the author as a screening measure to be completed prior to
administration of the Repetitive Behaviours Interview (RBI; see Section 3.5.1.2). The
RBQ covers the same repetitive behaviours as the RBI, but the questions are answered
116
in a yes/no questionnaire format. The caregiver of the individual completed the
questionnaire. S/he was asked to tick “yes” if his/her child had ever shown the
behaviour under question, regardless of its frequency, and whether it be recently or in
the past. Any questions that were ticked “yes” were then asked again verbally, with
follow-up questions, in the RBI, but questions which were ticked “no” were not
repeated within the RBI. The purpose of this structure of administration was mainly to
conserve time, given the time-consuming nature of the test protocol and other
interviews.
3.5.1.2 Repetitive Behaviours Interview (RBI; Turner, 1996)
The RBI was developed by Michelle Turner as part of her PhD thesis. Neither the full
RBI nor a thorough description of it have been published, so it is described in detail
here, and the current version is contained in full in Appendix A. It was designed to
measure the presence and severity of a large range of repetitive behaviours, including
those typically displayed by individuals with autism as well as those characteristic of
other clinical groups. Turner’s version of the RBI consists of 59 questions covering 10
categories of repetitive behaviour: stereotyped manipulation of objects, object
attachments, stereotyped movements, tic-like behaviours, self-injurious behaviour,
obsessive-compulsive behaviours, insistence on sameness of environment, rigid
adherence to routines and rituals, repetitive use of language, and circumscribed
interests. Each interview question asks whether or not the caregiver’s child displays a
particular type of behaviour, and includes specific examples of behaviours covered by
the question, so that caregivers are clear about the type of behaviour being targeted and
forgetting is minimised. The interview assesses whether or not the target behaviour is
displayed currently (once a week or more over the last three months), as well as whether
it had ever been displayed previously. These Recent and Lifetime behaviours are rated
separately. In the current research, only the Recent behaviours ratings were used,
because relationships between current cognitive and behavioural functioning were the
central concern.
Scoring procedures. For the classes of behaviour which occur in discrete
episodes (i.e., stereotyped manipulation of objects, stereotyped movements, tic-like
behaviours, self-injurious behaviour, and repetitive use of language), information on the
frequency of the behaviour is coded using an 8-point scale. The codes, which refer to
how often each episode of the behaviour occurs, range from (0) “never” to (7) “almost
117
constantly”, with intermediate codes referring to the number of episodes occurring per
week and per day (ranging from 1-2 times per week to more than 30 times per day).
Information on the duration of each episode is also included because individuals may
show a particular repetitive behaviour infrequently, but engage in it for a long period of
time, thus frequency data alone could be misleading. Duration information is coded
using a 5-point scale ranging from (0) “less than one minute” to (5) “30 minutes or
longer”. Caregivers are not given a list of the frequency and duration codes, in order to
prevent response bias. However, if any response is unclear, the caregiver is asked for
the number of times per day the behaviour is shown, or asked to choose between two of
the duration codes. In Turner’s version of the RBI, the circumstances which commonly
lead to each of the discrete-episode type behaviours are also coded in one of eight
categories, including “at no specific time or situation”, “when anxious or tense”, and so
forth.
For other “steady-state” behaviours which do not occur in discrete episodes and
can not be coded in terms of frequency and duration (i.e., object attachments, obsessive-
compulsive behaviours, insistence on sameness of environment, and rigid adherence to
routines and rituals), the severity of the behaviour is coded using a simpler 3-point
scale. In general, a code of (0) is used to indicate the absence of the target behaviour (or
at least, lack of abnormal levels of the target behaviour), (1) denotes mild inflexibility or
mild-moderate behavioural severity, and (2) indicates marked inflexibility or extreme
severity. Codes of (1) and (2) are specifically operationalised for each question.
Because the nature of some of the behaviours is such that they are shown to some
degree in a normal population (e.g., having regular routines, favourite items and so on),
severity is often gauged by the impact of the behaviour on the rest of the family. Each
of the sections on “steady-state” behaviours is followed by a series of questions about
how the child would react if s/he was prevented from indulging in each behaviour that
has been rated.
The two questions on circumscribed interests are structured slightly differently,
being rated in terms of the usual or unusual nature of the interest, the degree of
obsessionality with the interest, the typical or atypical manifestation of the interest, and
the degree to which it prevents the individual from pursuing other interests.
During interview administration, care is taken to ensure that the same behaviour
is not coded twice, under different questions. In cases where the same behaviour arises
twice or seems to fit under two different categories, the behaviour is coded according to
the most notable feature of the behaviour. For example, if a caregiver reports that their
118
child continuously kicks around a ball while walking around the house, this behaviour
would be coded under the object manipulation question, rather than the repetitive
pacing question. Similarly, a behaviour is not always coded under the question which
elicits its description by the informant, if it fits more appropriately under another
question. If a child shows two distinct behaviours which both fall under one question,
both are recorded and scored.
Differences in the current version. The version of the RBI used in the current
research differed from Turner’s in several ways. Firstly, only the questions which were
ticked “yes” on the RBQ were asked within the RBI. Secondly, the questions regarding
the circumstances which commonly lead to the display of the discrete-episode type
behaviours were not included. This was because it was found during initial testing that
parents found these questions quite difficult to answer clearly, and also because it was
felt that data gleaned from these questions were not essential for the current study.
Thirdly, the questions about how the child would react if s/he was prevented from
indulging in each of the “steady-state” behaviours were not asked either, for similar
reasons. Finally, the section on compulsive behaviours was expanded from two to five
questions, covering a larger range of behaviours. As a result of the latter three
modifications, the current version of the RBI includes 52 rather than 59 questions.
None of the questions from the original RBI about the presence and severity of the
repetitive behaviours themselves were removed or changed in the current version.
Summary variables used in statistical analyses. The main measures derived
from the current version of the RBI were the presence of behaviour and severity
summary scores for each behavioural category. The presence of behaviour summary
score was calculated by assigning a score of 1 for each question that received a
frequency rating above 0 (or severity rating above 0 for the “steady-state” behaviours),
and then calculating a sum of scores for all questions in the category.
The severity summary scores were slightly more complex. For the discrete-
episode behaviours, Turner simply used the frequency codes and did not use the
duration codes in her analyses. In the current research, it was decided that a severity
score which included both frequency and duration information would be a more
accurate reflection of the time spent on each behaviour. For each possible combination
of frequency and duration codes, the maximum number of minutes per week spent
doing the behaviour was calculated. For example, for a behaviour coded (2) “3-6 times
per week” for frequency and (3) “4-9 minutes” for duration, the maximum number of
minutes per week would be 54 (6 x 9). Because there was a very large range in the
119
number of minutes per week possible (0 to 10080), each combination of codes was then
ranked in severity, such that the lowest number of minutes per week was given a score
of 0 and the highest was given a score of 32. Thus, each of the possible combinations of
frequency and duration codes corresponded with a score between 0 and 32 inclusive.
Each discrete-episode behaviour rated on the RBI was therefore given a severity score
of between 0 and 32, and the severity summary score for each behavioural category
consisted of the sum of the severity scores for each behaviour in that category. For the
“steady-state” behaviours, the severity scores for each behaviour were simply the same
as the 0, 1 or 2 rating assigned during interview, with the severity summary score being
the sum of these scores across the behaviours in each category. The severity summary
scores for all behavioural categories were converted to t scores (with a mean of 50 and
standard deviation of 10) using the grand mean and standard deviation across the autism
and control groups, thereby enabling comparisons across different categories while
controlling for the fact that the number of items and range of scores is variable across
categories6.
To reduce the number of statistical comparisons required in analyses examining
the relationship between cognitive functioning and repetitive behaviours, Turner further
collapsed the severity summary scores for each behavioural category into four
composite variables: Repetitive Movements, Sameness Behaviour, Repetitive
Language, and Circumscribed Interests. The same composite variables were used in
this research, with the addition of a Compulsive Behaviours variable (due to the
addition of items in this category in the current version). The Repetitive Movements
composite score was the sum of severity summary scores for the stereotyped
manipulation of objects, stereotyped movements, tic-like behaviours, and self-injurious
behaviours categories. The Sameness Behaviour composite score included the severity
summary scores for insistence on sameness of environment, rigid adherence to routines
and rituals, and object attachments. The Compulsive Behaviours, Repetitive Language,
and Circumscribed Interests composite scores were simply the severity summary scores
for those categories.
Reliability and validity. In her PhD, Turner reports test-retest and inter-rater
reliability data for her version of the RBI. In terms of test-retest reliability, she reports
an average of 96% agreement across two administrations with regard to the simple
presence or not of each behaviour. The agreement was 83% for the frequency and
6 From this point onwards, the term “severity summary score” will mean the t score.
120
duration codes for the discrete episode behaviours, and 92% for the severity codes for
the “steady-state” behaviours. Inter-rater reliability was very good, at a mean of 99.5%
agreement for the frequency and duration codes, with a corresponding mean Kappa
value of .99. For the severity codes, there was a mean agreement of 91%, with a Kappa
value of 0.87. Turner (1996) did not explicitly examine the validity of the RBI in her
thesis. However, in the current studies, a high correlation between the
Repetitive/Restricted Behaviours domain of the ADI-R and an overall sum of severity
summary scores across categories on the RBI, which was conducted across all groups in
Studies 1 and 2, r = .73, p < .001, suggested good construct validity. The underlying
factor structure of the RBI was also examined in Study 1, the results of which are
reported in Section 4.3.5.1 of Chapter 4.
3.5.2 Measures of social behaviour and communication
3.5.2.1 Social Behaviour Questionnaire (SBQ; Skuse et al., 1997)
The SBQ is a 12-item questionnaire completed by the individual’s caregiver, which was
originally devised for use with a sample of individuals with Turner’s syndrome (Skuse
et al., 1997). It includes 12 statements primarily relating to the child’s everyday social
awareness and behavioural appropriateness; for example, “not aware of other people’s
feelings”, “does not pick up on body language”, and “does not understand how to
behave when out, e.g., in shops or other people’s houses”. These statements are rated as
0 (not at all true), 1 (quite or sometimes true), or 2 (very or often true). Scores therefore
range from 0 to 24. The questionnaire demonstrates good internal consistency, test-
retest reliability, and validity (Skuse et al., 1997).
3.5.2.2 Social and communication ADI-R domains
As the SBQ is a brief, limited measure of social functioning, questions in the social
domain of the ADI-R which related to current functioning were selected and scores
summed to form an additional measure of social behaviours. Similarly, scores on
questions in the communication domain relating to current functioning were also
summed as measure of communicative ability. Only questions relating to current
functioning were used (rather than all the questions usually used to calculate the
traditional algorithm for social behaviours and communication) because relationships
121
with current cognitive capacity were of central interest, as well as for the sake of
comparability with the RBI, from which measures of current behaviour only were taken.
These two ADI-R summary scores of current social behaviours and communication
correlated quite highly, r = .77, p < .001. A factor analysis conducted with the two
ADI-R summary scores and the SBQ score also demonstrated that all three measures
loaded on the same factor (the results of this factor analysis are described more fully in
Section 4.3.5.2 of the following chapter). It was therefore decided to create a composite
score of all three measures of social/communicative ability (i.e., the SBQ and the
current social and communication scores from the ADI-R). This was achieved by
conducting a factor analysis deriving factor scores for each participant using a
regression equation.
122
CHAPTER 4
Study One: Profile, Primacy, and Independence of Theory of Mind and Executive Function Impairments in Autism Spectrum Disorders
4.1 Introduction 4.1.1 Aims 4.1.2 Hypotheses
4.2 Method
4.2.1 Participants 4.2.2 Procedure
4.3 Results
4.3.1 Data screening 4.3.2 Group comparisons on ToM and EF tasks
4.3.2.1 False belief tasks 4.3.2.2 Dewey Stories 4.3.2.3 Tower of London 4.3.2.4 IDED set-shifting task 4.3.2.5 Response Inhibition and Load task 4.3.2.6 Opposite Worlds task 4.3.2.7 Relational Complexity 4.3.2.8 Pattern Meanings 4.3.2.9 Uses of Objects 4.3.2.10 Stamps task 4.3.2.11 Summary and effect sizes of group comparisons
4.3.3 Universality of ToM and EF deficits 4.3.4 Ability of ToM and EF variables to predict group membership 4.3.5 Behavioural measures: Group comparisons and derivation of indices
used in correlational analyses 4.3.5.1 Repetitive Behaviours Interview 4.3.5.2 Social and communicative functioning
4.3.6 Correlations between ToM/EF and behavioural measures 4.3.7 Relationship between ToM and EF
4.3.7.1 Correlations between ToM and EF 4.3.7.2 Dissociations between ToM and EF
4.4 Discussion 4.4.1 Profile of ToM and EF deficits 4.4.2 Primacy of ToM and EF deficits 4.4.3 Independence of ToM and EF deficits 4.4.4 Towards a “multiple primary deficits” model of ToM and EF in ASDs
123
4.1 Introduction
4.1.1 Aims
Chapter 2’s literature review revealed that individuals with autism consistently display
both ToM and EF deficits, but that the primacy and independence of these two
impairments remain a matter of current debate. The first of the two studies contained in
this thesis was principally aimed at elucidating the profile, primacy, and independence
of ToM and EF deficits in children with ASDs, with the broader aim of clarifying the
structure of the cognitive level of explanation in a causal model of autism. Thus, the
three central aims of Study One were to determine i) the specific profile of ToM and EF
deficits which characterises ASDs; ii) whether impairments in ToM and/or EF can
adequately meet the criteria for a primary cognitive deficit in ASDs, and which appears
to be the most primary; and iii) whether or not ToM and EF impairments are related in
ASDs, and if so, what the nature of that relationship might be (i.e., which theory of the
ToM-EF relationship is best supported by the data). The remainder of this section
describes how these aims were addressed in the current study.
i) Aim 1: Determining the profile of ToM and EF impairments. The specific
profile of ToM and EF impairments in ASDs was examined by comparing the
performance of individuals with ASDs with control participants matched on age and
non-verbal ability on a range of ToM and EF tasks. In particular, emphasis was given to
the precise measurement of a range of EF components. As described in Section 2.2.3 of
Chapter 2, previous studies of EF in autism have been weakened by the use of tasks
which are often impure and/or require non-verbal responses only (which may advantage
individuals with ASDs), and which do not cover the full range of EF components. This
study sought to address those weaknesses, not only in order to provide an accurate map
of the cognitive profile typical of ASDs, but also to help determine whether that profile
may be unique to autism (as, for example, the presence of inhibition deficits would be
inconsistent with the unique EF profile proposed by Ozonoff and colleagues; see
Section 2.2.3) and how each component may relate to ToM ability (discussed further
below). As described in Chapter 3, planning, set-shifting, inhibition, working memory,
relational reasoning, and generativity components were all measured. Both verbal and
non-verbal tests were used where possible. A task involving both inhibitory and
working memory demands was included, following suggestions that only tasks
combining both components are i) impaired in autism and ii) related to ToM. A test of
124
relational reasoning was included in order to examine Halford’s (1993) notion that the
capacity to integrate multiple relations may be a key ability underlying false belief
understanding (this represents a “common conceptual requirements” account of the
ToM-EF relationship, as described below in hypothesis 3). Generativity was also
assessed in detail, in response to indications that generativity deficits may hold strong
explanatory value in terms of the symptoms of autism (e.g., Jarrold et al., 1996; Turner,
1997).
ii) Aim 2: Determining the primacy of ToM and EF impairments. As we have
seen, common criteria used to judge the primacy of a cognitive deficit to a disorder are
its i) universality in individuals with the disorder, ii) uniqueness to the disorder, iii)
causal precedence or ability to account for the earliest symptoms of the disorder, and iv)
explanatory value or ability to account for the whole range of symptoms displayed by
individuals with the disorder. In this study, all of these criteria were tested in some way
for both the ToM and EF hypotheses of autism except for the third criterion of causal
precedence, as children below the age of 5 were not included in the sample. The main
reason for this was that it was important to test the range of EF components, using both
verbal and non-verbal response modes if possible, which is difficult for a young sample
both because tests in some EF domains (e.g., generativity) are not yet available for this
age group and because the limited verbal abilities of young children constrain the tests
appropriate for use1.
The criterion of universality was addressed in this study by calculating the
proportion of participants with ASDs displaying an impairment on the variable in
question, with “impairment” defined as a score worse than one standard deviation from
the control mean – a stricter cutoff for impairment than the lenient criterion of any score
below the control mean, which was used by Ozonoff et al. (1991). The uniqueness
criterion was tested indirectly, by analysing which ToM and EF variables best predicted
membership of the ASD group (this methodology was also used by Ozonoff et al., 1991,
to assess uniqueness). As individuals from other clinical groups were not assessed,
with the exception of a few children in the control group with mild intellectual
handicaps, this test of uniqueness should be considered as addressing whether the ToM
and EF deficits displayed were unique to individuals with ASDs compared with
individuals of equivalent age and non-verbal ability (rather than being unique to ASDs
1 Some 4-year-old autistic and control children were tested as part of the WAFSASD (see Method – Section 4.2) and it was found that many were unable to adequately comprehend or perform several of the EF tasks.
125
compared with all other clinical conditions). Explanatory value was measured by
calculating correlations between ToM/EF variables and behavioural measures of autistic
symptomatology (i.e., social/communicative functioning and repetitive behaviours). A
particular emphasis was placed on a thorough examination of each cognitive
impairment’s relationship with repetitive behaviours and restricted interests, in
comparison with a briefer assessment of social and communicative functioning.
Although this emphasis was not strictly necessary for the exploration of explanatory
value, it was considered important because this third aspect of the autistic triad has been
one of the main grounds for discriminating the ToM and EF hypotheses (that is, both
ToM and EF capabilities show relationships with and appear able to explain social and
communicative impairment, but the ToM hypothesis does not account well for repetitive
behaviours). In addition, these non-social aspects of autistic symptomatology have been
largely neglected in previous research, with only one published study directly
addressing the relationship between ToM/EF and repetitive behaviours (Turner, 1997).
iii) Aim 3: Determining the independence of ToM and EF impairments. The
nature of the relationship between ToM and EF in children with ASDs was investigated
by comparing the pattern of correlations between ToM and EF variables in the ASD
participants with similar correlations in the control group. Thus, the presence of
significant correlations between ToM and EF in individuals with ASDs would be
suggestive of an underlying relationship, and the pattern of correlations would show
which EF components may be important for ToM performance or development. The
incidence and direction of dissociations between ToM and EF deficits in the ASD group
were also examined, by calculating the proportion of ASD participants with impaired
EF who displayed intact ToM, and vice versa (with impairment defined in the same way
as for the universality calculations). This allowed assessment of whether one ability
appeared to be a prerequisite for the other (or whether one impairment ever occurred
without the other), which is relevant for the question of primacy as well as helping to
discriminate between the different theories of the ToM-EF relationship – in particular,
the two emergence accounts (see Section 2.3 in Chapter 2).
4.1.2 Hypotheses
Predictions for the profile of deficits. It was expected that both ToM and EF deficits
would be found in our sample of individuals with ASDs, with poorer performance
expected on higher-level ToM measures. In terms of the specific profile of EF deficits,
126
based on the outcomes of previous research it was predicted that ASD participants
would show impairments in planning, set-shifting, and generativity, but not inhibition or
working memory. However, consistent with Russell’s (1997b) proposal, it was
hypothesised that ASD participants may show impairments on the task combining
inhibition and working memory requirements. It was expected that in domains where
both verbal and non-verbal measures were used, individuals with ASDs would be more
likely to show impairments on verbal tasks. No specific predictions were made with
regard to performance on the relational reasoning task, which has not been used
previously with individuals with autism; however, given previous findings of intact
working memory in ASDs, it was thought possible that this domain may also be intact
(as it tests Halford’s (1993) notion of working memory).
Predictions for the primacy and independence of deficits. In considering the possible
outcomes of analyses of the primacy and independence of ToM and EF deficits in
individuals with ASDs, a number of different hypotheses are conceivable, all of which
hold different implications for theories of the primary cognitive deficit(s) of autism as
well as theories of the ToM-EF relationship. These hypotheses include the following:
1. There is only a single, primary deficit in ASDs, with no secondary impairments.
This hypothesis would be supported if only ToM or only EF impairments are
displayed by the ASD group. This possibility is not likely given fairly consistent
evidence that both ToM and EF impairments are present in children with ASDs.
2. ToM and EF impairments are related in ASDs such that one deficit is primary and
either causes or explains the other, which is secondary. This possibility could be
consistent with expression, emergence, and common neuroanatomical bases
accounts of the ToM-EF relationship. For example:
i) If EF deficits are primary and cause a ToM deficit because of performance-
based factors (i.e., the expression account), this would be revealed in a
pattern of results showing EF deficits as more primary, significant
correlations between ToM and certain EF components (most likely inhibition
and working memory), and no or few dissociations such that those EF
components are impaired but ToM is intact2.
2 Dissociations in the other direction would also be unlikely as EF should not be intact in individuals with ASDs if EF deficits are primary.
127
ii) If a ToM impairment is primary and causes a secondary EF deficit because
of functional dependence during development (i.e., Perner’s emergence
account), this would be reflected in a pattern of results demonstrating a ToM
deficit meeting criteria for primacy, significant correlations between ToM
and EF, and no dissociations in ASD participants such that ToM is impaired
but EF is intact (dissociations in the other direction would also be
inconsistent with Perner’s theory, as discussed in Section 2.3.1.3, and
unlikely as per footnote 2).
iii) If an EF (or ToM) deficit is primary and a secondary ToM (or EF) deficit is a
consequence of its neuroanatomical proximity, then one would expect to see
evidence of the primacy of EF but not ToM (or ToM but not EF), and
correlations between ToM and EF, but dissociations would be acceptable
such that EF (or ToM - i.e., the primary domain) is impaired and
performance in the other domain is intact.
3. ToM and EF impairments are related, but neither is primary; there is a third deficit
which is primary and causes both deficits. This result would not be supportive of
either the ToM hypothesis or the EF hypothesis of autism. It would be consistent
with a “common conceptual requirements” account of the ToM-EF relationship.
This hypothesis would be reflected by results showing neither ToM nor EF deficits
adequately meeting the criteria for primacy (as while they would be caused by the
primary deficit, there would not be as direct a relationship with symptoms),
significant correlations between ToM and EF variables, and no or few dissociations
in either direction (at least on tasks with the common conceptual basis).
4. ToM and EF impairments are independent in ASDs, but only one is primary. In this
hypothesis, the most likely explanation for the co-occurrence of the non-primary
impairment would be its neuroanatomical proximity to the primary impairment, but
unlike version iii) of hypothesis 2, the two deficits are not correlated. This lack of
correlation despite neuroanatomical proximity would suggest something unusual
about the ToM-EF relationship in ASDs as compared with typically developing
children. Results would show primacy of one of the deficits but not the other, and
no correlations between ToM and EF deficits. Dissociations would be allowable
such that performance in the primary domain is impaired but the second impairment
does not always occur.
128
5. ToM and EF impairments are independent in ASDs and both are equally primary.
Like hypothesis 4, the co-occurrence of impairments would be most likely explained
by their neuroanatomical proximity, but unlike hypothesis 4, both are primary.
Results would be expected to demonstrate that both ToM and EF deficits meet
criteria for primacy, but there would be few significant or strong correlations (as
although ToM and EF deficits would have to co-occur in the large majority of ASD
participants, they may not necessarily co-vary in severity). Dissociations would not
be expected to occur if all criteria for primacy were met by both impairments, as
both primary deficits would have to be impaired in each individual with an ASD.
This is a somewhat unlikely outcome as it is improbable that two independent
deficits would both show complete explanatory value for the full range of
symptoms.
6. ToM and EF impairments are independent in ASDs, and neither meets all criteria
for primacy. This represents a more classic “multiple primary deficits” model of
ASDs, in which the deficits both hold causal importance but neither are universal or
can account for the full range of symptoms. Again, the co-occurrence of
independent deficits is most likely to be caused by common neurobiological
substrates. This hypothesis is consistent with at least three different scenarios
regarding cognitive deficits in ASDs, for example:
i) There may be different subgroups of individuals with different primary
deficits (these subgroups may be classified according to level of intellectual
functioning or symptom severity, for example – see Section 1.2 in Chapter
1). In this case, neither ToM nor EF would be universal among the whole
sample, but both may hold good explanatory value within the relevant
subgroup (correlations with symptoms across the whole sample may not be
strong, however). Dissociations in both directions would be expected, such
that one subgroup could show intact ToM but impaired EF, and another
subgroup could show the opposite pattern (there may also be a third group
where both abilities are impaired).
ii) If autism is considered to be a multidimensional spectrum, for example if a
ToM deficit was the basis of one aspect of symptomatology (e.g., social
impairment) and EF deficits were the basis for another aspect (e.g., repetitive
behaviours), then neither deficit would be likely to be universal (if the
129
sample was heterogeneous and not all individuals showed all symptoms),
and each deficit would only hold explanatory value for the relevant symptom
domain. ToM-EF dissociations in either direction may occur in individuals
who do not display all aspects of symptomatology.
iii) There may be a third (or more) cognitive deficit(s), which may be more
primary than or at least equally primary as ToM and EF deficits. This may
actually also be the case for either of the above two scenarios (i.e., there
could be 3 subgroups characterised by different primary deficits, or 3
independent cognitive deficits underlying the three aspects of
symptomatology). This third deficit may be related to either ToM or EF
deficits, but would not explain them both as in hypothesis 3.
Hence, hypotheses 1-4 all represent different versions of a single primary cognitive
deficit model of autism, whereas hypotheses 5 and 6 both represent multiple primary
deficits models.
In the only previous study to directly address the primacy and independence of
ToM and EF deficits in autism in a similar manner to this study, Ozonoff et al. (1991)
found most support for version iii) of hypothesis 2. They found that ToM and EF were
correlated in autism, but that EF deficits were more primary, as judged by their
universality and uniqueness to autism. Although dissociations were not explicitly
examined, they reported that a subset of their ASD sample showed impaired EF but
intact ToM. They interpreted this pattern of results as suggesting a neuroanatomical
link between ToM and EF deficits in autism, such that they were correlated but the
relationship was not causal at a cognitive level. However, as described in Chapter 2,
Ozonoff et al.’s (1991) study was weakened by i) its use of impure EF tasks and the
employment of an EF composite score which obscured the specific nature of both EF
deficits and the ToM-EF relationship; ii) its lenient definition of impairment; and iii) its
failure to partial out age from the ToM-EF correlations. In addition, Ozonoff et al. did
not examine the presence of ToM-EF dissociations in both directions, the outcome of
which is an important discriminator between the six hypotheses outlined above.
Because of these weaknesses, and because other research relevant to the primary
and independence of ToM and EF in autism has been equivocal, no strong predictions
about which one of the above hypotheses was likely to be supported were made prior to
conducting this study. Because of weak plausibility, a low likelihood was placed on
hypotheses 1 and 5, and based on previous studies it was also suspected that neither
130
ToM nor EF deficits may fully meet all criteria for primacy. The current study may
nevertheless be considered an exploratory but extensive investigation of the primacy
and independence of ToM and EF impairments in ASDs. It builds upon Ozonoff et al.’s
(1991) original study and other relevant research by i) utilising a range of EF tasks
designed to tap separate components of EF, the results of which were analysed
separately throughout; ii) adopting a stricter criterion of impairment; and iii) partialling
out age, VIQ and PIQ from all significant correlations. It also explicitly examines the
presence of double dissociations between ToM and EF, and includes investigations of
the explanatory value of ToM and EF deficits as an additional measure of primacy
(which hold particular importance as a way of discriminating between the 3 scenarios
presented in hypothesis 6).
4.2 Method3
4.2.1 Participants
Autism Spectrum Disorders (ASD) Group. There were 48 participants with ASDs
ranging in age from 5 to 18 years. Participants in this group were mainly recruited
through Western Australian autism centres (specialising in assessment and/or therapy
with individuals with ASDs) and support groups, including the Autism Association of
Western Australia, Intervention Services for Autism and Developmental Delay, the WA
Disability Services Commission, and the Asperger Syndrome Support Group.
Participants of a previous study on the genetics of autism conducted through the Centre
for Clinical Research in Neuropsychiatry were also invited to participate in the current
study. The study was advertised using brochures, features in newsletters, and
presentations to professionals and parents. Parents expressed interest by returning a slip
via mail to the research team giving consent to be contacted about the study.
3 Both of the studies in this thesis formed part of the Western Australia Family Study of Autism Spectrum Disorders (WAFSASD), a large-scale project funded by a National Health and Medical Research Council grant awarded to chief investigators Joachim Hallmayer, Murray Maybery, and Dorothy Bishop. The rationale and methodology for this thesis were nevertheless developed largely independently from the broader aims of the WAFSASD. The current author selected and developed all of the cognitive measures used in the thesis (as well as the RBI), was principally responsible for administration of these tasks to participants, and chose and conducted all statistical analyses reported. However, the diagnostic instruments were selected in collaboration with other WAFSASD investigators, and similarly were administered by research assistants for the WAFSASD. In addition, recruitment of families was conducted in collaboration with WAFSASD research assistants. Some of the probands with autism who participated in the WAFSASD were too low-functioning to complete all of the cognitive tasks used in this study, and were not included.
131
All participants had received a clinical diagnosis of autism (n = 28), Asperger
syndrome (n = 13) or PDDNOS (n = 7) from a health professional (e.g., paediatrician,
psychiatrist, psychologist). The presence of autistic symptomatology in at least one
domain was then verified using the Autism Diagnostic Interview – Revised (ADI-R).
Two participants (one with a clinical diagnosis of autism and one with Asperger
syndrome) were excluded as they did not exceed cutoff scores in any of the three ADI-R
domains (i.e., social interaction, communication, restricted/repetitive behaviour). Of the
remaining 46 participants in the ASD group, 34 met criteria in all three domains of the
ADI-R, 10 met criteria in two domains, and 2 met criteria in one domain only.
Other exclusion criteria were the presence of genetic abnormalities or
neurological dysfunction (e.g., head injury, encephalitis, neurofibromatosis, cerebral
palsy), with the exception of epilepsy4. There were four participants in the ASD group
with comorbid diagnoses, as reported by their parents (2 with dyspraxia, 1 with
epilepsy, and 1 with dyspraxia and epilepsy).
Control Group. Forty-nine control children ranging in age from 5 to 17 years were
recruited to participate in the study. Of these, 46 were typically developing children and
3 had mild intellectual disabilities. The control group was selected to match the ASD
group on age and PIQ (reasons for not matching on VIQ are described below).
Recruitment of this group was mainly achieved through Western Australian schools,
again through brochures and newsletters mailed to parents. Because of difficulty
recruiting sufficient numbers of boys with low PIQ, in some schools all boys whose
parents gave consent were tested on IQ measures, and then the parents of those boys
with PIQs in the range of 60-95 were contacted and asked if they would like to
participate in the larger study. The children with mild intellectual disabilities were
recruited through the WA Disability Services Commission (as controls for children in
the ASD group with PIQs between 60 and 70).
Exclusion criteria were a known or suspected ASD, as well as genetic and
neurological abnormalities. Mothers of control participants completed the Autism
Screening Questionnaire (ASQ) in order to screen for symptoms of autism in the control
group. If participants scored above the cutoff point of 10 on the ASQ, the ADI-R was
administered. One participant, who had a mild intellectual disability, met criteria for
4 Although epilepsy is a neurological illness, because it is a common comorbid condition of autism, it was felt that exclusion of participants with epilepsy may result in a sample which was non-representative of autism.
132
autism on the ADI-R and was excluded from further analysis, leaving 48 participants in
the control group. Two participants in the control group had received clinical diagnoses
of ADHD (as reported by their parents).
Demographic characteristics of each group are presented in Table 2. The ASD and
control groups were matched on chronological age, t(92) = 1.74, p = .09, and PIQ, t(92)
= .92, p > .1. Children in the ASD group had significantly lower VIQs than the control
group, t(92) = 3.7, p < .001. Because children with autism typically show a significant
discrepancy between VIQ and PIQ, matching groups on Full-Scale IQ or on both VIQ
and PIQ was not considered appropriate or possible. VIQ was therefore included as an
additional independent variable in group comparisons, as described in Section 4.3.2.
All participants had a PIQ of 60 or above, and a VIQ of 50 or above. The proportion of
girls was slightly higher in the control group than in the ASD group, and chi-square
analysis revealed that the difference approached significance, χ2 (1, N = 94) = 3.65, p =
0.06. However, this was not considered to be a problem as analyses conducted to
compare the performance of boys and girls in the control group on all cognitive tasks
revealed no significant differences. Gender was not introduced as an additional
independent variable (IV) in analyses because the number of girls in the ASD group was
considered to be too small.
Table 2. Demographic characteristics of the samples
ASD group (n = 46) Control group (n = 48)
Age: Mean (SD, range) 10.73 (3.96, 5-18) 9.49 (2.94, 5-17)
Male: Female 40: 6 34: 14
PIQ: Mean (SD, range) 96.07 (18.23, 63-138) 99.42 (16.99, 64-137)
VIQ: Mean (SD, range) 91.76 (21.77, 52-150) 106.58 (16.85, 64-138)
The ASD and control groups were also matched in terms of their families’
socioeconomic status. This was assessed using education data from both mothers and
fathers, which was coded using the following system: 1 = up to year 10 (or equivalent)
of high school; 2 = up to year 12 (or equivalent) of high school; 3 = diploma, trade
certificate, apprenticeship, or other traineeship; and 4 = university degree. A chi-square
analysis comparing the education levels of ASD and control parents (the analysis
included both mothers and fathers) revealed that there was no difference in the
education level of the parents of ASD and control children, χ2 (3, N = 150) = 5.71, p >
133
.1. The difference remained non-significant when only the highest code from each
family was included in the analysis, χ2 (3, N = 89) = 3.35, p > .1.
With an n of 46 in the ASD group and 48 in the control group, the power of the
study to detect medium sized effects (i.e., d = .5) at an alpha level of .05 reached an
acceptable level at .78.
4.2.2 Procedure
All questionnaires, parental interviews, and cognitive tasks are described in detail in
Chapter 3. Initial screening questions regarding medical history (to assess whether
participants met criteria for participation) were asked of the participant’s mother via
telephone. Informed consent was obtained from the mother of each participant, on
behalf of both herself and her child (direct consent was also obtained from participants
over 12 years of age, with the exception of children whose level of understanding of the
research was judged to be insufficient to give informed consent). Questionnaires were
generally sent to participants’ mothers prior to the first testing session. Tests and
parental interviews were usually administered at the participants’ homes, or in testing
rooms at the Centre for Clinical Research in Neuropsychiatry. The ADI-R took
approximately 2 hours to administer, and the RBI an additional 5-30 minutes, depending
on the number of questions asked. The test battery took approximately 2.5 hours in
total to administer5. The order of test administration was fixed, except the order of
Wechsler subtests differed according to whether the WPPSI-R, WISC-III or WAIS-III
was administered (the order of subtest administration specified by each test was
retained). Testing was often divided into two sessions, in order to prevent fatigue and
distractibility. For practical reasons, when testing was conducted across more than one
session, the break was not always at the same point within the battery. Some tests were
administered only to participants within a certain age range. The order of testing (not
including other tests administered for WAFSASD) and the age range for each test is
displayed in Table 3.
5 This includes other tests not reported within this thesis but which were conducted as part of the WAFSASD. The IQ, ToM and EF tests took between 1.5 and 2 hours in total.
134
Table 3. Order of test battery and age range for each test
Test Age range
1. WPPSI-R: i) Object Assembly
ii) Vocabulary
iii) Picture Completion
iv) Similarities
WISC-III: i) Picture Completion
ii) Similarities
iii) Vocabulary
iv) Object Assembly
WAIS-III: i) Picture Completion
ii) Vocabulary
iii) Similarities
iv) Object Assembly
5-6
7-16
17+
2. Stamps task – first 4 trials 5-16
3. Tower of London All ages
4. Dewey Stories 10+
5. Simple false belief task 5-16*
6. First-order false belief task 5-16*
7. Second-order false belief task 5-16*
8. IDED set-shifting: Perseveration condition 7+
9. Response Inhibition and Load task 7+
10. IDED set-shifting: Learned Irrelevance condition 7+
11. Uses of Objects All ages
12. Relational Complexity All ages
13. Stamps task – second 4 trials 5-16
14. Pattern Meanings All ages
15. Opposite Worlds 7-16
*Not all children within this age range were necessarily administered these tasks. The
structure of administration of the false belief tasks is described in Section 3.3 of Chapter
3.
135
4.3 Results
This section includes subsections covering i) data screening; ii) group comparisons on
ToM and EF tasks; iii) analyses addressing the universality of ToM and EF deficits in
the ASD group; iv) logistic regression analyses examining the ability of ToM and EF
task performance to predict ASD/control group membership; v) group comparisons on
and derivation of indices for the behavioural measures; vi) correlations and multiple
regressions examining the relationship between ToM/EF and behavioural measures; and
viii) analyses examining the relationship between ToM and EF. SPSS (Statistical
Package for the Social Sciences) Version 10.0.5 was used for all analyses.
4.3.1 Data screening
Data from all measures were screened for normality and outliers. For variables with
distributions that did not depart substantially from normality, outliers falling more than
3 standard deviations (SD) from the mean of the group (i.e., ASD or control) were
trimmed to 3 SD from the mean. Several variables demonstrated highly skewed
distributions. Square root, logarithm and inverse transformations were attempted for
these variables. If transformation was successful, the transformed variable was used for
all analyses (including correlations and regressions). For some variables where a large
proportion of participants all gained the same score, transformations were ineffective.
For these variables, scores were dichotomised. Again, the dichotomised variable was
used in all analyses. Relevant specific details regarding outliers, transformations and
dichotomising of scores are included within the results section for each measure.
4.3.2 Group comparisons on ToM and EF tasks
For all group comparisons, the performance of the ASD group as a whole was compared
with the control group. For tests administered only to participants within a certain age
range (see Table 3), t-tests were conducted to check that the participants available from
the ASD and control groups for those tests were still matched on age and PIQ. Results
showed that the groups were matched for all tests, with the exception of Dewey Stories.
The way in which this was handled is described in Section 4.3.2.2.
To address concerns that the range of symptom severity within the ASD group
may affect results (i.e., autism versus other PDD subgroups may display different
136
patterns of results), comparisons were also conducted between participants in the ASD
group who exceeded cutoff scores in all three domains of the ADI-R (i.e., met “full
criteria”; n = 34) and those who exceeded cutoff scores in only one or two domains (i.e.,
met “partial criteria”; n = 12). The two subgroups were matched on age and PIQ.
Almost all comparisons on cognitive tasks revealed no significant subgroup
differences6. The only task on which the two subgroups showed different patterns of
performance was Pattern Meanings, and these results are reported in Section 4.3.2.8.
For all other tasks, as there were no significant differences it was thought appropriate to
consider the “full criteria” and “partial criteria” subgroups together as one sample for
group comparisons.
As described in Section 4.2.1, four participants in the ASD group and two in the
control group had a clinical diagnosis other than an ASD (e.g., ADHD, epilepsy,
dyspraxia). To check that the presence of a non-ASD diagnosis was not strongly
influencing results, group comparisons were conducted excluding these participants
from the sample. All significant group differences remained significant, so participants
with non-ASD diagnoses were included in all analyses reported.
A consistent approach to group comparisons on each task was followed,
involving the following steps:
1. T-tests (or chi-square analyses for dichotomous variables) comparing the
performance of the ASD and control groups were conducted for all task variables.
2. Scatterplots between task variables and age, VIQ, and PIQ were examined for any
non-linear relationships. No significantly curvilinear relationships were detected.
The relationship between age and some task variables was slightly curvilinear, but
not to an extent that warranted the use of special analyses.
3. Pearson product-moment correlations7 were conducted between the task variables
and age, VIQ, and PIQ.
6 It should be noted that some caution should be exercised in interpreting these non-significant subgroup differences because the power to detect them may not be adequate. However, the lack of subgroup differences is still likely to mean that including the PDD subgroup in the ASD sample did not have any significant effect on the overall result besides increasing the sample size and therefore power of the main analyses. 7 Although some variables were dichotomous, Cohen and Cohen (1983) state that the formula for the point biserial correlation coefficient is computationally equivalent to the formula for the product-moment correlation coefficient. They assert that the difference in formula is of no significance when computer programs are used for data analysis, because whatever formula the program uses will work when variables are scored 0-1. For ease of reporting, r is used throughout the results section to denote both a Pearson product-moment and a point biserial correlation coefficient (which are computed identically anyway).
137
4. For task variables which were significantly correlated with age and/or PIQ, an
analysis of covariance (ANCOVA) was conducted, mainly to assess whether any
non-significant group differences became significant when extraneous variance
attributable to age and/or PIQ was removed. Miller and Chapman (2001)
recommend the use of ANCOVA in this way as a “noise reduction technique”.
5. If task variables were correlated with VIQ, ANCOVA was not considered to be an
appropriate technique to examine the effect of VIQ on group comparisons, as the
groups were not matched on VIQ. In their review of the use of ANCOVA with
nonrandomly assigned groups, Miller and Chapman (2001) argue that ANCOVA
cannot be used to “control for” group differences on a covariate, which they state is
a highly consistent view in the technical literature. Essentially, this is because when
the covariate and the independent variable (in this case, group) are not independent,
the regression adjustment of the independent variable (IV) may remove part of the
effect of group or produce a spurious effect of group (see Miller & Chapman, 2001,
for further explanation). As an alternative to ANCOVA in this situation, Maxwell
and Delaney (1990) suggest the “blocking” of participants on the covariate, and then
introducing the “blocked” variable as an additional IV in analyses. This strategy
was adopted in the current study. Participants (in the ASD and control groups
combined) were divided into three equal groups according to their VIQ score, and
then VIQ level was used as an IV in a 2-way ANOVA (with group as the other IV)
and the task variable as the dependent variable (DV). In this way, the influence of
VIQ on group comparisons was assessed by examining whether i) the main effect of
group remained significant when VIQ was controlled for by introducing it as an
additional IV, or ii) if any non-significant group differences became significant
when the effect of VIQ was separated out. In addition, any group x VIQ
interactions would be of interest in examining the possibility that group differences
were found for some VIQ levels but not others (i.e., interactions would indicate
heterogeneous regression slopes).
6. If dichotomous variables correlated with age and/or IQ variables, the effect of
age/IQ was assessed by conducting logistic regression analyses with the
dichotomous variable as the outcome variable, and group and age/IQ as the
predictor variables. This allowed assessment of the independent contribution of
group to the outcome variable minus the variance attributable to age/IQ.
138
4.3.2.1 False belief tasks
Examination of distributions of scores from all false belief tasks revealed that a large
proportion of participants (particularly from the control group) gained perfect scores for
both belief and control questions. All variables were therefore recoded as dichotomous
such that a perfect score was coded as 1 and any other score as 0. This fairly strict
scoring criterion was considered appropriate for the age and level of ability of both the
ASD and control groups, as a more lenient scoring system would have produced ceiling
effects. It should be noted, however, that a score of 0 is better interpreted as indicating
an “unstable” false belief performance rather than a true failure on the task.
Five participants (four in the ASD group, one in the control group) were not
administered the First-order and Second-order false belief tasks due to equipment
malfunction. These participants had all passed the Simple false belief task and were
therefore assigned the mean value of other participants in their group who had passed
the Simple false belief task (which was in turn coded dichotomously according to the
criteria described above). As the false belief tasks were administered to a restricted age
range, the overall sample size for all false belief tasks was 89 (n = 43 in the ASD group,
n = 46 in the control group). However, the ns for the memory and reality questions, as
well as the own belief questions in the Simple false belief task, were limited to those
who actually did the task – as these questions were not assumed to be passed or failed
according to performance on other false belief tasks, as was the case for the belief
questions (see Section 3.3 for a description of the structure of administration of the false
belief tasks). Percentages of participants gaining perfect scores (i.e., “perfect scorers”)
in each group for each false belief task (on the belief questions only) are presented in
Table 4.
i) Simple false belief task. A chi-square analysis revealed that there was no
statistically significant difference between the ASD and control groups on the reality
questions, χ2 (1, N = 36) = .82, p > .1. However, on the belief questions referring to the
participant’s own previous belief, significantly fewer children in the ASD group were
perfect scorers, χ2 (1, N = 36) = 4.42, p < .05, indicating that they were more likely to
incorrectly state their own previous beliefs8. On the belief questions referring to others’
beliefs, significantly fewer children in the ASD group were perfect scorers, χ2 (1, N =
8 While there were significant group differences on this variable, it was not included in subsequent analyses (e.g., correlations) because the sample size for the variable was considered to be too small.
139
89) = 7.25, p < .01, indicating that they were less likely to make accurate predictions
about the beliefs of others.
Performance on the others’ belief questions was significantly correlated with
both age, r = .23, p < .05, and VIQ, r = .37, p < .001. A logistic regression with age,
VIQ and group as the predictors showed that according to the Wald criterion, the
independent contribution of group to variance in others’ belief questions performance
became only marginally significant with age and VIQ partialled out, z = 3.61, p = .06.
ii) First-order false belief task. There was no significant difference between
the ASD and control groups on the reality questions, χ2 (1, N = 76) = 1.74, p > .1, or
memory questions, χ2 (1, N = 76) = .001, p > .1. On the belief questions, a significantly
lower proportion of children in the ASD group gained perfect scores, χ2 (1, N = 89) =
11.34, p < .01, indicating that they were less likely to make accurate predictions about
others’ false beliefs.
Performance on belief questions was significantly correlated with both age, r =
.26, p < .05, and VIQ, r = .44, p < .001. In a logistic regression with age, VIQ and
group as the predictors, the independent contribution of group remained significant, z =
7.60, p < .01.
iii) Second-order false belief task. As for the other false belief tasks, there
was no significant difference between the ASD and control groups on the reality
questions, χ2 (1, N = 68) = 1.34, p > .1, or memory questions, χ2 (1, N = 68) = .29, p >
.1. There were significantly fewer perfect scorers in the ASD group on the belief
questions, χ2 (1, N = 89) = 4.93, p < .05.
Scores on belief questions were significantly correlated with age, r = .28, p <
.01, and VIQ, r = .44, p < .001. In a logistic regression with age, VIQ and group as the
predictors, the independent contribution of group was no longer significant, z = 2.16, p
> .1.
iv) Overall false belief performance indices. An aggregate score was also
calculated across the three false belief tasks, for use in other analyses. The sum of
correct responses on belief questions (including only others’ belief questions for the
simple false belief task) was dichotomised in the same way as for each individual task,
such that a perfect score (15/15) was coded as 1 and any other score as 0. Chi-square
analysis revealed that there were significantly fewer perfect scorers overall in the ASD
group than in the control group, χ2 (1, N = 89) = 8.1, p < .01. The aggregate score was
significantly correlated with age, r = .29, p < .01, and VIQ, r = .42 p < .001. Group
140
remained a significant predictor of the aggregate score when age and VIQ were
partialled out in a logistic regression, z = 5.96, p < .05.
As the dichotomous scoring system used for the aggregate score was a fairly
strict one, a more lenient scoring criterion was also used for an alternative aggregate
score. In the alternative system, any participant scoring 13 or more out of 15 (i.e.,
making either 0, 1, or 2 incorrect responses) was given a score of 1 (“high scorers”), and
participants with lower scores were assigned a 0 (“low scorers”). A chi-square analysis
showed that significantly fewer ASD participants were high scorers than control
participants, χ2 (1, N = 89) = 6.25, p < .05. The alternative aggregate score correlated
significantly with age, r = .26, p < .05, PIQ, r = .24, p < .05, and VIQ, r = .50, p < .001.
When a logistic regression was conducted with these age and IQ variables and group as
predictors, the effect of group became only marginally significant, z = 2.81, p = .09.
Table 4. False belief task results: Percentage of participants in each group with perfect
scores [or high scores in the case of the alternative aggregate score] on belief questions,
and significance of group comparisons
ASD group Control group p p with age/
IQ control
Simple false belief:
Own belief 55.0 87.5 *
Others’ belief 72.1 93.5 ** -
First-order false belief 48.8 82.6 ** **
Second-order false belief 51.2 73.9 * -
Aggregate score 39.5 69.6 ** *
Alternative aggregate [55.8] [80.4] * -
* p < .05; ** p < .01; *** p < .001; - p > .05.
4.3.2.2 Dewey Stories
With the Dewey Stories task administered to only a small subset of the sample in the
older age range, the ASD (n = 17) and control (n = 18) participants who completed the
task were not matched on age and PIQ. However, the total score on Dewey Stories was
not significantly correlated with either age, r = -.19, p > .1, or PIQ, r = -.30, p = .08, and
so the non-matching of groups on these variables was not considered to be important.
141
The total score variable did not require transformation. A t-test comparing the
total scores of the ASD group (M = 7.94, SD = 3.91) and control group (M = 5.89, SD =
2.95) revealed a marginally significant group difference in the expected direction, t(33)
= 1.76, p = .09. The total score was significantly correlated with VIQ, r = -.48, p < .01.
Because of the small sample size for this task, VIQ was split into two levels rather than
three. An ANOVA with group and VIQ level as the IVs showed that group differences
in the total score did not remain significant when assessed independently of VIQ,
F(1,31) = .29, p > .1. The group x VIQ level interaction was not significant, F(1,31) =
.81, p > .1.
Because the Dewey Stories task was of uncertain validity as a measure of ToM,
correlations were conducted between the total score and the false belief variables. Raw
correlations were significant for all false belief variables except the simple false belief
task, however when age, PIQ and VIQ were partialled out, there were no longer any
significant correlations between false belief variables and the Dewey Stories total score.
It therefore appears that the validity of the Dewey Stories task as a measure of
mentalising ability is questionable, and it may be better considered as a measure of
social awareness and understanding of acceptable social behaviours. Throughout
subsequent analyses, the Dewey Stories task is considered a measure of “social
cognition”.
4.3.2.3 Tower of London (ToL)
The main performance indices of the ToL were the overall sum of adjusted extra move
scores (from here on referred to as the total adjusted extra move score) and the total
number of problems completed in the minimum number of moves. These two scores
were highly correlated, r = -.96, p < .001, hence only the total adjusted extra move score
was used in analyses. The number of rule violations per block administered was also
analysed. Because many participants committed no or very few rule violations, this
variable was highly skewed and was recoded as a dichotomous variable, with
participants making 0-1 violations per block being given a score of 0 (“low rule
violators”) and participants making any higher number of violations scored as 1 (“high
rule violators”).
Two participants had missing data on the ToL, one from the ASD group and one
from the control group, and were not included in analyses (resulting in n = 45 for the
ASD group and n = 47 for the control group). A t-test comparing the total adjusted
142
extra move scores of the ASD group (M = 26.31, SD = 7.78) and control group (M =
22.53, SD = 7.35) revealed that the ASD group made a higher number of extra moves
than the control group, t(90) = 2.40, p < .05. A chi-square analysis also showed that
significantly more participants in the ASD group (44.4%) than the control group
(23.4%) were high rule violators, χ2 (1, N = 92) = 4.56, p < .05.
Both ToL indices were significantly correlated with age (r = -.41, p < .001, for
the total adjusted extra move score; r = -.37, p < .001, for rule violations), and VIQ (r =
-.39, p < .001, for the total adjusted extra move score; r = -.26, p < .05, for rule
violations). An ANCOVA conducted on the total adjusted extra move score, with group
and VIQ level as the IVs and age as the covariate, revealed that the group difference
remained significant when age and VIQ were controlled, F(1,85) = 4.98, p < .05. The
group x VIQ level interaction was not significant, F(2,85) = .48, p > .1. Group also
remained a significant predictor of rule violation status (low/high) when age and VIQ
were assessed independently in a logistic regression, z = 4.82, p < .05.
4.3.2.4 IDED Set-shifting task
All set-shifting variables were highly skewed, with a large number of participants
making no errors or only one error to criterion in each stage. As a result, all variables
were recoded such that any error score of 0 or 1 was coded as 0 and any higher number
of errors was coded as 1. Because the reversal stages were not crucial to the current
study, only the first reversal stage (SDR) in each condition (i.e., Perseveration and
Learned Irrelevance) was included in analyses. Only the extra-dimensional shift (EDS)
stages were included in subsequent analyses (i.e., correlations etc.) as they were the
central variables of interest.
The overall N for the task (which had a restricted age range) was 72 (n = 36 in
both the ASD and control groups). Due to computer malfunction, data for the
Perseveration condition from one participant in the ASD group were invalid and not
included in analyses. The percentage of participants in each group making only 0 or 1
errors (i.e., “low error scorers”) for each stage in each task condition is displayed in
Table 5. There were no significant group differences on any variable. However, there
was a marginally significant trend for a smaller proportion of participants from the ASD
group to be low error scorers on the EDS stage of the Learned Irrelevance condition, χ2
(1, N = 72) = 3.77, p = .052, suggesting that children with ASDs may have found it
more difficult to shift their attention to a previously irrelevant stimulus dimension.
143
No variables were significantly correlated with age or PIQ, and only the IDS
stage of the Learned Irrelevance condition was significantly correlated with VIQ, r =
-.34, p < .01. The effect of group remained non-significant when a logistic regression
on this variable was performed with VIQ and group as predictors.
Table 5. IDED Set-shifting task results: Percentage of low error scorers in each group
for each stage of each task condition, and significance of group comparisons
ASD group Control group p p with age/
IQ control
Perseveration condition
SD stage 60.0 66.7 -
SDR stage 52.9 47.1 -
CD stage 45.3 54.7 -
IDS stage 44.9 55.1 -
EDS stage 60.0 72.2 -
Learned Irrelevance condition
SD stage 63.9 66.7 -
SDR stage 77.8 77.8 -
CD stage 63.9 61.1 -
IDS stage 77.8 77.8 - -
EDS stage 13.9 33.3 -
* p < .05; ** p < .01; *** p < .001; - p > .05.
4.3.2.5 Response Inhibition and Load (RIL) task
For all RIL task conditions, error variables (i.e., the percentage of errors made) were
highly skewed, with many participants making only 0-2% errors. These variables (with
the exception of the percentage of errors made in choosing the most recently displayed
shape in Condition 3) were recoded such that 0-2% errors was coded as 0 (a “low error
score”), and any higher percentage of errors was coded as 1 (a “high error score”).
However, as this precluded the use of a repeated measures ANOVA to compare
increments in performance across Conditions 1-3, the main error variables used in
analyses for these conditions were:
144
i) the inhibition error difference score - the difference between the scores
for Condition 2 (the inhibition condition) and 1 (the control condition);
ii) the load error difference score – the difference between the scores for
Condition 3 (the working memory load condition) and Condition 2; and
iii) the inhibition + load error difference score – the difference between the
scores for Conditions 3 and 1.
These difference scores were normally distributed. Four outliers were trimmed: one
control participant’s inhibition and inhibition + load error difference scores, another
control participant’s load error difference score, and one ASD participant’s inhibition +
load error difference score. The distribution of the percentage of errors made in
choosing the most recently displayed shape in Condition 3 (or the shape error score, a
measure of working memory ability under conditions requiring inhibitory control) was
also skewed, but a square root transformation was effective for this variable.
Although the median RT variables for all conditions demonstrated roughly
normal distributions, for the sake of consistency, an inhibition RT difference score, load
RT difference score, and inhibition + load RT difference score were also calculated
(representing the same comparisons between conditions as for the error data). One
outlier on the inhibition + load RT difference score from a participant in the ASD group
was trimmed.
The overall N for the task was 71 (n = 36 in the ASD group, n = 35 in the
control group). Due to computer malfunction, one participant in the ASD group had
incomplete data in Condition 3, and so error and RT data from that condition as well as
the difference scores involving Condition 3 were not included in analyses. Table 6
displays the mean and SD of each group (and the significance of group comparisons) for
the error and RT difference scores, and the shape error score. There were no significant
group differences on t-tests of the error difference scores, although there was a trend for
the ASD group to show a larger inhibition + load error difference score, t(68) = 1.72, p
= .09. A t-test comparing the shape error scores for Condition 3 revealed that the ASD
group made significantly more shape errors, t(68) = 2.03, p < .05, indicating that in an
inhibition task with a working memory load, individuals with ASDs were less able to
respond accurately on a measure of their working memory ability.
Examination of the error data for each condition separately revealed that there
was a significantly lower proportion of low error scorers in the ASD group (27.8%) than
the control group (51.4%) in Condition 2, χ2 (1, N = 71) = 4.16, p < .05. However, the
inhibition error difference score did not show a significant group difference, probably
145
because there was also a trend for fewer children in the ASD group to be low error
scorers in Condition 1, the control condition (47.2% vs. 68.6% in the control group), χ2
(1, N = 71) = 3.31, p = .07. The proportion of low error scorers in the ASD group was
also marginally lower for Condition 3 (22.9% vs. 42.9%), χ2 (1, N = 70) = 3.17, p = .08.
Thus, the overall pattern of results for the error scores in Conditions 1-3 suggested that
the ASD group tended to make more errors on all tasks, but their performance accuracy
was not proportionally worse in task conditions with inhibitory and working memory
demands (at least on the inhibitory aspect of the task – note that the ASD group
performed significantly worse on the shape error score, an index of their working
memory performance when the task contained both inhibitory and working memory
demands).
There were no significant differences between groups on any RT difference
scores, and no trends were evident. There were no significant RT differences between
the groups when Conditions 1-3 were analysed separately. In subsequent analyses, only
the error and RT difference scores and the shape error score were used, and separate
error and RT data for Conditions 1-3 were not included.
Table 6. RIL task results: Mean (and SD) of each group, and significance of group
comparisons, for error and RT difference scores and the shape error score
ASD group Control group p p with age/
IQ control
Error difference scores:
Inhibition 3.43 (6.37) 1.24 (6.20) -
Load 2.14 (7.91) 0.66 (4.66) -
Inhibition + load 5.23 (8.41) 1.95 (7.49) -
RT difference scores:
Inhibition 194.53 (166.88) 191.95 (198.21) - -
Load 116.25 (200.42) 174.53 (193.95) - -
Inhibition + load 314.48 (232.88) 366.57 (214) - -
Working memory measure:
Shape error score 25.52 (21.48) 15.43 (13.89) * **
* p < .05; ** p < .01; *** p < .001; - p > .05.
Note: The means and SDs shown for the shape error score are for the raw data, prior to
transformation.
146
It should be noted that for both the ASD and control groups, there was a significant
increase in both the number of errors made and the time taken to respond in the
inhibition condition (and load condition) compared with the control condition,
indicating that these conditions were more difficult than the control condition and
therefore that the instruction to respond to the opposite colour to the stimulus did
require inhibitory control.
Age was significantly correlated with the shape error score (r = -.32, p < .01),
the inhibition RT difference score (r = -.30, p < .05), and the inhibition + load RT
difference score (r = -.33, p < .01). VIQ was correlated with the load RT difference
score (r = -.24, p < .05). ANCOVAs on the inhibition and the inhibition + load RT
difference scores with age as the covariate did not change the non-significant effect of
group for these variables: F(1,68) = .28, p > .1, for inhibition and F(1,67) = .22, p > .1,
for inhibition + load. The group difference in the shape error score remained significant
when an ANCOVA was conducted with age as a covariate, F(1,67) = 7.84, p < .01. The
group difference on the load RT difference score also remained non-significant when a
two-way ANOVA with group and VIQ as the IVs was conducted, F(1,64) = .22, p > .1.
The interaction between group and VIQ was not significant, F(2,64) = .07, p > .1.
4.3.2.6 Opposite Worlds task
Opposite Worlds task variables used in group comparisons were the Same World error
score, Opposite World error score, Same World time score and Opposite World time
score (each score equating to the sum of two trials). There was one outlier in the ASD
group on the Opposite World time score, which was trimmed to 3 SD from the mean.
In subsequent analyses (to be reported in sections to follow), the error and time
difference scores between the Opposite and Same World conditions were the main
variables used as these were thought to be appropriate summary scores (representing the
performance decrement when inhibitory demands are introduced) for use in correlations
and other analyses. Means and SDs for all variables are displayed in Table 7.
The N for the task was 65 (n = 29 for the ASD group, n = 36 for the control
group). For the error scores, a two-way repeated measures ANOVA was conducted
with group (ASD, control) as the between-subjects factor and condition (Same World,
Opposite World) as the within-subjects factor. There was a significant main effect of
condition, F(1, 63) = 22.96, p < .001, but the main effect of group was not significant,
F(1, 63) = 1.62, p > .1. The interaction approached significance, F(1, 63) = 3.01, p =
147
.09, suggesting there was a trend for the ASD group to make comparatively more errors
in the Opposite World condition. Follow-up simple effects analyses showed that there
was no significant difference between the groups in the number of errors made in the
Same World condition, t(63) = .32, p > .1, but there was a marginally significant
difference in the Opposite World error scores, t(63) = 1.71, p = .09.
A two-way repeated measures ANOVA with group as the between-subjects
factor and condition as the within-subjects factor was also conducted on the time scores.
There was a significant main effect of condition, F(1, 63) = 107.77, p < .001, and a
significant effect of group, F(1, 63) = 5.2, p < .05. The interaction was also significant,
F(1, 63) = 7.36, p < .01, indicating that participants in the ASD group took
comparatively longer to complete the Opposite World condition (in other words, they
showed a larger performance decrement from the Same World to the Opposite World
condition compared with the control group). Follow-up analyses confirmed that there
was no significant difference between the ASD and control group on the Same World
time scores, t(63) = 1.36, p > .1, but the ASD group took significantly longer in the
Opposite World condition, t(63) = 2.66, p < .05.
Table 7. Opposite Worlds results: Mean (and SD) of each group for error/time scores
in each condition and difference scores, and significance of group comparisons
ASD group Control group p p with age/
IQ control
Error variables:
Same World error score 1.21 (1.57) 1.08 (1.56) -
Opposite World error score 2.69 (2.54) 1.78 (1.74) - -
Error difference score 1.48 (2.2) 0.69 (1.45) - -
Time variables:
Same World time score 27.27 (6.42) 25.0 (6.89) - -
Opposite World time score 38.42 (12.55) 31.53 (8.26) * *
Time difference score 11.12 (9.02) 6.53 (4.2) ** **
* p < .05; ** p < .01; *** p < .001; - p > .05.
Note: The difference scores relate to the interaction term on repeated measures
ANOVAs.
148
Neither VIQ nor PIQ correlated with any task variables, but age was
significantly correlated with the Same World time score, r = -.42, p < .001, the Opposite
World time score, r = -.43, p < .001, and the Opposite World error score, r = -.28, p <
.05. When age was introduced as a covariate in two-way repeated measures
ANCOVAs, there was no change in any of the results.
4.3.2.7 Relational Complexity
In this task, the main variable used for analyses was simply the total score (i.e., total
number correct), summed across all trials. There was one outlier in the ASD group for
this variable, which was trimmed.
A t-test comparing the total score of the ASD group (M = 9.66, SD = 3.9) and
control group (M = 9.71, SD = 4.2) was not significant, t(92) = .05, p > .1.
The total score correlated with both age, r = .58, p < .001, and VIQ, r = .28, p <
.01. An ANCOVA conducted on the total score with group and VIQ level as IVs and
age as a covariate did not influence the non-significant effect of group, F(1, 87) = .003,
p > .1. There was no significant interaction between group and VIQ level, F(2, 87) =
.18, p > .1.
4.3.2.8 Pattern Meanings
All error variables (i.e., redundant, repetitive, incorrect, and uninterpretable responses)
showed significant positive skew. For redundant responses, a square root
transformation was effective. Repetitions were recoded such that 0 or 1 repetition(s)
was coded as 0 and 2 or more repetitions were coded as 1. Due to the very small
number of incorrect and uninterpretable responses, these two variables were summed to
form a combined incorrect/uninterpretable responses variable, which was recoded such
that 0 errors remained at 0, and 1 or more errors was coded as 1. A “sum of errors”
variable was created, where the number of error responses was summed across all
categories. This variable was also skewed, and was transformed using a logarithm
equation. The other major variable for the Pattern Meanings task was the number of
correct responses, which was normally distributed.
There were no statistically significant group differences in the mean number of
correct responses produced, t(92) = 1.38, p > .1, or the sum of errors, t(92) = .14, p > .1.
Similarly, individual analyses of error variables did not reveal any significant group
149
differences. There was no significant difference in the mean number of redundant
responses, t(92) = .46, p > .1. The proportion of low error scorers in each group was not
significantly different for repetitions, χ2 (1, N = 94) = .001, p > .1, or incorrect/
uninterpretable responses, χ2 (1, N = 94) = .68, p > .1.
However, as mentioned previously, this task was the only one on which the “full
criteria” (i.e., autism) and “partial criteria” (i.e., other PDD) subgroups showed
significant differences. The partial criteria subgroup made significantly more errors
than the full criteria subgroup on the sum of errors variable, t(44) = 2.62, p < .05. When
specific error types were analysed, it was found that the partial criteria subgroup made
significantly more redundant responses, t(44) = 2.49, p < .05, and a higher proportion of
the partial criteria subgroup made a high number of repetitions, χ2 (1, N = 46) = 4.31, p
< .05. There was also a trend for the partial criteria subgroup to make more correct
responses, t(44) = 1.83, p = .07. This pattern of results therefore suggests that the
partial criteria subgroup generated more responses overall, whether correct or not. Each
of the two subgroups was then compared to the control group. It was found that the full
criteria subgroup demonstrated significantly fewer correct responses than controls, t(80)
= 2.06, p < .05, and the partial criteria subgroup produced significantly more error
responses overall than controls, t(58) = 2.44, p < .05 (in terms of specific error types,
the partial criteria subgroup produced significantly more redundant responses than
controls but there were no significant differences for other error types). Because the
two subgroups displayed different patterns of performance, Table 8 displays means,
SDs, the percentage of low scorers for dichotomous variables, and significance of group
comparisons for all variables separately for the two subgroups9.
Across the whole sample, the sum of errors was significantly correlated with
age, r = -.46, p < .001, as were all individual error variables: redundant responses, r = -
.32, p < .01, repetitions, r = -.25, p < .05, and incorrect/uninterpretable responses, r = -
.26, p < .05. VIQ was correlated with the number of correct responses, r = .20, p < .05,
and the sum of errors, r = -.20, p < .05, but of the individual error variables, only
repetitions were significantly correlated with VIQ, r = -.22, p < .05. All group
differences remained non-significant after controlling for these variables when the
whole ASD sample was analysed as one group. When the separate analyses for the full
and partial criteria subgroups were conducted with the relevant age and IQ variables
9 The two subgroups were not analysed separately in subsequent analyses involving correlations with behavioural and ToM variables, as it was of interest to see whether Pattern Meanings performance correlated with symptom severity (or ToM performance) across the whole sample.
150
controlled, the difference between the full criteria subgroup and controls on correct
responses became non-significant, F(1, 76) = 2.31, p > .1, and the difference between
the partial criteria subgroup and controls on the sum of errors became only marginally
significant, F(1, 53) = 3.02, p = .09, although the difference between these two groups
on redundant responses remained significant, F(1, 57) = 7.17, p < .05. There were no
significant interactions between group and VIQ level in any analyses.
Table 8. Pattern Meanings results: Mean (and SD) of each subgroup [or the percentage
of low error scorers for dichotomous variables], and significance of group comparisons
ASD group Control group p p with
____________________________ age/IQ
Full subgroup Partial subgroup control
Correct responses 21.32 (7.87) 26.50 (9.83) 25.1 (8.42) *1 -
Sum of errors 7.21 (9.12) 15.75 (10.81) 7.4 (7.51) *2 -
Individual error types:
- Redundant 4.06 (4.94) 8.92 (7.51) 4.23 (4.34) *2 *2
- Repetition [67.6] [33.3] [58.3] - -
- Incorrect/uninterpretable [73.5] [58.3] [77.1] - -
* p < .05; ** p < .01; *** p < .001; - p > .05.
Note: The means and SDs shown for the sum of errors and redundant responses are for
the raw data, prior to transformation. 1 Difference was between full criteria subgroup and controls 2 Difference was between partial criteria subgroup and controls
4.3.2.9 Uses of Objects
As for the Pattern Meanings task, all error variables (including the additional non-useful
responses variable) were positively skewed. For redundant responses and repetitions,
log transformations were effective. A square root transformation improved the
distribution of non-useful responses. Again, incorrect and uninterpretable responses
were summed to form a combined variable, which was recoded as dichotomous in the
same way as for the Pattern Meanings task. A “sum of errors” variable was also
created, which was normally distributed for this task. One outlier on this variable from
the control group was trimmed. The total number of correct responses, as well as the
number of correct responses for conventional and non-conventional items separately, all
151
had approximately normal distributions. Means, SDs, the percentage of low scorers for
dichotomous variables, and significance of group comparisons for all variables are
displayed in Table 9.
Table 9. Uses of Objects results: Mean (and SD) of each group [or the percentage of
low error scorers for dichotomous variables], and significance of group comparisons
ASD group Control group p p with age/
IQ control
Correct responses:
- Total 19.07 (8.99) 26.42 (9.5) *** **
- Conventional items 7.04 (3.86) 10.25 (4.35)
- Non-conventional items 12.02 (5.77) 16.17 (6.02)
Sum of errors 18.41 (12.84) 17.52 (10.09) - -
Individual error types:
- Redundant 6.02 (5.79) 5.42 (3.76) - -
- Repetition 4.28 (3.99) 5.13 (4.55) - -
- Non-useful 6.61 (6.05) 6.38 (5.53) - -
- Incorrect/uninterpretable [63.0] [64.6] - -
* p < .05; ** p < .01; *** p < .001; - p > .05.
Note: The means and SDs shown for redundant responses, repetitions, and non-useful
responses are for the raw data, prior to transformation.
To examine whether or not the ASD group produced proportionally fewer correct
responses on the conventional versus non-conventional items, a two-way repeated
measures ANOVA was performed with group as the between-subjects factor and
condition (conventional, non-conventional) as the within-subjects factor. There was a
significant main effect of group, F(1, 92) = 14.82, p < .001, and condition, F(1, 92) =
155.97, p < .001, but the interaction was not significant, F(1, 92) = 1.16, p > .1,
indicating that the ASD group produced fewer correct responses than the control group
for both conventional and non-conventional items (with the conventional items being
more difficult for both groups), but were not proportionally worse on conventional
items. Because of this, the separate totals for conventional and non-conventional items
were not used in further analyses.
There was no significant difference between groups on the sum of errors, t(92) =
.37, p > .1, and individual analyses of error variables did not reveal any significant
152
group differences. There was no significant difference in the mean number of
redundant responses, t(92) = .64, p > .1, repetitions, t(92) = 1.22, p > .1, or non-useful
responses, t(92) = .08, p > .1. The proportion of low error scorers in each group was not
significantly different for incorrect/uninterpretable responses, χ2 (1, N = 94) = .02, p >
.1.
Age was significantly correlated with the number of correct responses, r = .31, p
< .01, the sum of errors, r = -.34, p < .01, and all individual error variables (except
repetitions): redundant responses, r = -.27, p < .05, non-useful responses, r = -.26, p <
.05, and incorrect/uninterpretable responses, r = -.38, p < .001. VIQ was correlated
with the number of correct responses, r = .45, p < .001, repetitions, r = -.27, p < .05, and
incorrect/uninterpretable responses, r = -.25, p < .05. The group difference in the
number of correct responses remained significant in an ANCOVA with group and VIQ
level as the IVs and age as a covariate, F(1, 87) = 12.66, p < .01. Group remained a
non-significant effect on the sum of errors when age was introduced as a covariate in an
ANCOVA, F(1, 91) = 1.05, p > .1. Similarly, group differences in all individual error
variables remained non-significant when age and/or VIQ was partialled out using either
ANCOVA or logistic regression. There were no significant interactions between group
and VIQ level in any analyses.
4.3.2.10 Stamps task
Both the rule adherence and restriction scores demonstrated highly skewed distributions
and were recoded as dichotomous variables. For rule adherence, a score between 0 and
6 inclusive was coded as 0 and a score of 7 or 8 was coded as 1. For restriction, a score
of 0 was left as 0 and a score between 1 and 8 inclusive was coded as 1. The
complexity and originality scores showed approximately normal distributions. Means
and SDs for the latter two variables and the proportion of low scorers for the former two
variables, along with the significance of group comparisons for all scores, are presented
in Table 10.
The N for the task was 87 (n = 41 for the ASD group, n = 46 for the control
group). T-tests revealed significant group differences on the complexity score, t(85) =
2.73, p < .01, indicating that the ASD group produced less complex patterns than the
control group, and on the originality score, t(85) = 2.81, p < .01, indicating that the ASD
group produced fewer original patterns than the control group. It was found using chi-
square analysis that there was a lower percentage of low scorers in the ASD group on
153
the restriction score, χ2 (1, N = 87) = 5.76, p < .05, indicating that a larger proportion of
the ASD group tended to use fewer stamps than were available. For the rule adherence
score, there was a marginally significant trend for a smaller proportion of the ASD
group to produce patterns adhering to one rule, χ2 (1, N = 86) = 3.50, p = .06, which was
contrary to expectation.
The originality score was significantly correlated with both age, r = .34, p < .01,
and VIQ, r = .40, p < .001. The restriction score correlated with VIQ, r = -.36, p < .01.
In a two-way ANCOVA with group and VIQ level as the IVs and age as a covariate, the
group difference in the originality score remained significant, F(1, 80) = 4.55, p < .05.
The interaction between group and VIQ level was not significant, F(2, 80) = .96, p > .1.
When a logistic regression was performed on the restriction score, group was no longer
a significant predictor when it was assessed independently of VIQ, z = 1.47, p > .1.
Table 10. Stamps task results: Mean (and SD) of each group [or the percentage of low
scorers for dichotomous variables], and significance of group comparisons
ASD group Control group p p with age/
IQ control
Complexity score 18.63 (3.02) 20.39 (2.98) **
Originality score 3.17 (2.51) 4.78 (2.8) ** *
Restriction score [82.9] [97.8] * -
Rule adherence score [26.8] [11.1] -
* p < .05; ** p < .01; *** p < .001; - p > .05.
4.3.2.11 Summary and effect sizes of group comparisons
Table 11 presents a summary of the results of group comparisons on the main variables
from each cognitive task. Overall, participants in the ASD group performed
significantly more poorly than controls on tasks measuring false belief understanding,
planning, verbal inhibition, working memory (under conditions where inhibition was
required), and both verbal and non-verbal generativity (with different patterns of results
for the two subgroups of ASD participants meeting “full criteria” and “partial criteria”
on the Pattern Meanings task); but not awareness of social norms, set-shifting, non-
verbal inhibition or relational reasoning (although marginally significant differences
were obtained on certain measures of social awareness, set-shifting and non-verbal
inhibition). Age and VIQ influenced some of these results, reducing the significance of
154
group comparisons for two false belief variables, two verbal generativity variables and
one non-verbal generativity variable.
Table 11. Summary and effect sizes of significant group differences
Measure
Significant
group
difference?
Significant
difference with
age/IQ control?
Effect size:
r (and d)
ToM:
Simple false belief: Own belief .35 (.75)
Other’s belief - .28 (.58)
First-order false belief .36 (.77)
Second-order false belief - .24 (.50)
False belief aggregate .30 (.63)
False belief alternative aggregate - .26 (.54)
Social Cognition:
Dewey Stories - - .28 (.58)
Planning:
ToL: Adjusted extra move score .24 (.50)
Rule violations .21 (.43)
Set-shifting:
IDED Perseveration condition:
EDS stage errors - -
IDED Learned Irrelevance cond.:
EDS stage errors - - .23 (.47)
Inhibition:
RIL task error difference scores:
Inhibition - -
Load - -
Inhibition + load - - .23 (.47)
RIL task RT difference scores:
Inhibition - -
Load - -
Inhibition + load - -
155
Table 11 continued
Measure
Significant
group
difference?
Significant
difference with
age/IQ control?
Effect size:
r (and d)
Inhibition continued:
Opposite Worlds:
Error difference score - - .21 (.43)
Time difference score .31 (.65)
Working memory:
RIL task shape error score .27 (.56)
Relational reasoning:
Relational Complexity score - -
Generativity:
Pattern Meanings:
Correct responses 1 - .23 (.47)
Sum of errors 2 - .34 (.73)
Uses of Objects:
Correct responses .37 (.80)
Sum of errors - -
Stamps task:
Complexity score .28 (.58)
Originality score .29 (.61)
Restriction score - .26 (.54)
Rule adherence score - - .20 (.41)
�significant to at least p < .05 level; - p > .05. 1 Difference was between full criteria subgroup and controls only 2 Difference was between partial criteria subgroup and controls only
156
It should be noted that while Bonferroni corrections were not performed, the fact that
group differences followed a consistent pattern and were all in the expected direction
(such that ASD participants performed more poorly than controls) signifies that the
results are likely to be valid. Table 11 also lists the effect sizes obtained for all
significant and marginally significant group differences, as a measure of the strength of
each effect. The “effect size correlation”, or r (Rosenthal, 1991), was used as the
primary measure of effect size. The effect size correlation simply measures the size of
the correlation between the independent and dependent variable (a phi correlation was
calculated for dichotomous variables, which is equivalent to Pearson’s r and point
biserial correlations for continuous variables). However, all values of r were also
converted to d (as shown in Table 11) using an equation supplied by Rosenthal (1991),
and the size of each effect was evaluated using Cohen’s (1988) system for classifying
small, medium and large effects. The largest effect size, and the only one to classify as
a large effect, was for Uses of Objects correct responses - a measure of verbal
generativity. Most other effect sizes fell in the medium range, including the Dewey
Stories total score, on which there was only a marginally significant group difference
but for which there was a small sample size. All other variables for which only
marginally significant group differences were found displayed small effect sizes, and
the ToL rule violations also showed only a small effect size.
4.3.3 Universality of ToM and EF deficits
Ozonoff et al. (1991) assessed universality of ToM and EF deficits in their study by
calculating the proportion of individuals in their autism group who scored below the
mean of the control group. As discussed in Section 2.2.3, this is a lenient criterion for
defining a deficit. In this study, it was decided to adopt the stricter criterion of a score
more extreme (in the direction of poorer performance) than 1 SD from the mean of the
control group (i.e., in the extreme 16% of control scores for a normal distribution) as the
definition of “impairment”. The universality of a deficit on continuous variables was
therefore assessed by calculating the proportion of participants in the ASD group
scoring more poorly than 1 SD from the mean. This was done only for variables where
significant group differences were found (including variables for which the group
difference did not remain significant when age and IQ variables were partialled out, but
not including variables on which only marginally significant group differences were
found).
157
For variables coded dichotomously, the “more poorly than 1 SD from the mean”
strategy was obviously not feasible, but it was necessary for the calculation of
universality to be comparable to that for continuous variables. To address this, the
percentage of control participants gaining a score of 0 (or 1 if a higher score was poorer)
was calculated, and if it was approximately 16%, the percentage of ASD participants
gaining that score was considered a comparable measure of the universality of a deficit
on that variable (as a score at the 16th percentile corresponds to a score at 1 SD below
the mean for a normal distribution). For the false belief variables, the alternative
aggregate score (see Section 4.3.2.1) was considered the best measure to use in
assessing universality10, as 19.6% of control participants gained a score of 0.
Universality was also calculated for the first-order false belief task, on which 17.4% of
control participants scored 0. The percentages for the two other dichotomous variables
where significant group differences were found were not quite as ideal. For ToL rule
violations, 23.4% of control participants gained a high error score of 1. This variable
was therefore recoded using a more lenient criterion such that a score between 0 and 1.5
rule violations per block scored 0, which resulted in a more appropriate 17% of control
participants scoring 111. For the Stamps task restriction score, only 2.2% of control
participants gained a high restriction score of 1, so this variable was not included in the
universality calculations.
The percentages of ASD participants demonstrating a deficit on the ToM and EF
variables where significant group differences were found are displayed in Table 12. It is
evident that neither ToM nor EF deficits are universal within the ASD sample12. The
percentages of ASD participants showing deficits also appear to be fairly comparable
across ToM and EF variables, although there was some variability among the EF
variables. Within the EF tasks, deficits in verbal inhibition and verbal generativity were
the most prevalent.
10 It is worth noting that although an aggregate score was used for the false belief variables, an aggregate or composite score was not calculated across the EF tasks (as was done by Ozonoff et al., 1991) for these universality calculations or for subsequent analyses because it was not thought to be valid or meaningful, particularly in light of the fact that one of the aims of the study was to examine the specific profile of EF deficits in ASDs and the relationship of each EF component with behavioural symptomatology and with ToM. In support of this, although there were some intercorrelations between EF domains, for the most part EF task variables were not significantly correlated with each other and appeared to be measuring different constructs (these correlations are presented in Appendix B and discussed further in Section 4.4.1). The fact that group differences were found on some EF tasks but not others solidifies this view. In addition, within EF domains, verbal and non-verbal measures often did not correlate with each other (i.e., for the different tests of inhibition and generativity). EF variables were therefore considered separately throughout analyses. 11 Group differences were still significant for this recoded variable, χ2 (1, N = 93) = 4.70, p < .05. 12 Even when Ozonoff et al.’s (1991) more lenient criterion for defining a deficit was used, ToM and EF “deficits” still could not be considered universal, with proportions ranging from 60.0 to 82.6%.
158
Table 12. Universality of ToM and EF deficits in the ASD group
% of ASD group displaying a deficit
ToM:
False belief alternative aggregate score 44.2
First-order false belief 51.2
Planning:
ToL: Adjusted extra move score 28.9
Rule violations 37.0
Inhibition:
Opposite Worlds time difference score 48.3
Working memory:
RIL task shape error score 37.1
Generativity:
Pattern Meanings: Correct responses 28.3
Sum of errors 26.7
Uses of Objects correct responses 41.3
Stamps task: Complexity score 19.5
Originality score 29.3
4.3.4 Ability of ToM and EF variables to predict group membership
In order to investigate the “uniqueness” of ToM and EF impairments to autism (as
compared with matched controls), a logistic regression analysis was conducted to
examine which cognitive task variables were best able to discriminate the ASD group
from the control group. A direct logistic regression was performed with group as the
outcome variable, and VIQ and all ToM and EF variables on which there were
significant group differences as the predictors. Logistic regression was chosen as the
method of analysis rather than discriminant function analysis because logistic regression
is more suitable when there is a mixture of dichotomous and continuous predictor
variables (Tabachnik & Fidell, 1996). Direct logistic regression evaluates the
independent contribution made by each predictor over and above that of the other
predictors (i.e., each predictor is assessed as if it entered the equation last).
Because not all participants completed every task (mainly due to age limits on
certain tasks, as well as missing data), only those participants with data for all the
159
predictor variables were included in the logistic regression. There were 27 participants
in the ASD group and 32 in the control group who met these criteria, and these limited
groups were matched on age (M = 11.26, SD = 3.18 for the ASD group; M = 10.13, SD
= 2.27 for the control group), t(57) = 1.58, p > .1, and PIQ (M = 94.52, SD = 15.78 for
the ASD group; M = 99.78, SD = 18.68 for the control group), t(57) = 1.16, p > .1.
A test of the full model with all 12 predictors against a constant-only model was
statistically reliable, χ2 (12, N = 59) = 31.03, p < .01, indicating that the predictors, as a
set, reliably distinguished children with ASDs from controls. 84.4% of the control
group and 77.8% of the ASD group were classified correctly by the model. Table 13
presents regression coefficients, Wald statistics, odds ratios, and 95% confidence
intervals for odds ratios for each of the 12 predictors. According to the Wald criterion,
the only reliable predictors of group membership were the Opposite Worlds time
difference score (a verbal measure of inhibition) and the number of correct responses on
the Uses of Objects task (a measure of verbal generativity). Performance on first-order
false belief questions approached significance as a predictor (p = .08).
Two possible limitations with this initial analysis were that i) correlations
between variables derived from the same task (or set of tasks) may have affected the
ability of individual variables from those tasks to emerge as a significant predictor, and
ii) the ratio of cases to predictors was lower than it should be in the ideal regression. In
order to address these limitations, another logistic regression was conducted where only
one variable from each task was included (VIQ, first-order false belief, ToL adjusted
extra move score, RIL shape error score, Opposite Worlds time difference score, Uses
of Objects correct responses, and Stamps task originality score). The ratio of cases to
predictors was therefore substantially higher in this alternative analysis. Variables were
chosen on the basis of the effect size of group comparisons and their representativeness
of task performance. Results were almost the same as the initial regression, with the
only difference being that the level of significance of the Uses of Objects correct
responses variable dropped from p = .04 to .07. The first-order false belief task variable
remained only marginally significant as a predictor13 (p = .08). The initial logistic
regression was therefore interpreted as a valid indicator of the ability of each task
variable to predict group membership14.
13 When the false belief alternative aggregate score was included instead (as this was the variable used in the universality calculations), the results also remained the same with the exception that the false belief aggregate was a non-significant, rather than a marginally significant, predictor of group membership. 14 As it was possible that VIQ and false belief variables may have affected each other’s contribution due to their significant correlation, another logistic regression (with the initial set of task variables) was conducted without including VIQ. First-order false belief performance was found to be a significant
160
Table 13. Logistic regression analysis of group membership as a function of VIQ, ToM
and EF variables
95% C. I. for odds ratio
Wald test ___________________
Variables B (z-ratio) Odds ratio Upper Lower
VIQ -.04 2.46 .96 .91 1.01
False belief tasks:
Simple .68 .15 1.98 .06 59.92
1st - order -2.05 2.96 .13 .01 1.33
2nd - order .87 .53 2.38 .23 24.56
ToL:
Adj. extra move score .03 .18 1.03 .90 1.19
Rule violations -.07 .01 .93 .15 5.93
RIL task:
Shape error score -.02 .01 .98 .70 1.37
Opposite Worlds:
Time difference score .16 3.98* 1.17 1.0 1.38
Uses of Objects:
Correct responses -.10 4.03* .91 .82 1.0
Stamps task:
Complexity score -.19 1.36 .83 .60 1.14
Originality score .08 .25 1.08 .80 1.47
Restriction score -2.26 1.26 .10 .0 5.43
*p < .05; ** p < .01; *** p < .001.
predictor in this analysis, z = 4.12, p < .05. However, rather than suggesting that false belief performance was actually a meaningful predictor of group membership, this pattern of results (i.e., the change in significance of false belief as a predictor when VIQ was included) indicates that false belief understanding did not add significant additional variance to the regression beyond that contributed by VIQ.
161
4.3.5 Behavioural measures: Group comparisons and derivation of indices
used in correlational analyses
4.3.5.1 Repetitive Behaviours Interview (RBI)
Group comparisons. Severity summary scores were the main RBI variables used in
analyses. Distributions of the severity summary scores were frequently skewed for the
ASD group, and highly skewed for the control group. However, all transformations
were ineffective. Non-parametric statistics were used for group comparisons of the
severity of different types of repetitive behaviours. As expected, Mann-Whitney U tests
revealed that children in the ASD group exhibited significantly more severe repetitive
behaviours in all categories of the RBI (all ps < .001, except for self-injurious
behaviours, where p < .01)15. Medians and ranges of the severity summary scores
(expressed as t scores) for the ASD and control groups are presented in Table 14.
Table 14. Median (and range) of RBI severity summary scores for the ASD and control
groups
Median (range) of severity summary scores
RBI category ASD group Control group
Stereotyped manipulation of objects 54 (45-119) 45 (45-75)
Stereotyped movements 58 (46-110) 46 (46-63)
Tic-like behaviours 49 (47-130) 47 (47-57)
Self-injurious behaviours 48 (48-172) 48 (48-67)
Compulsive behaviours 60 (46- 99) 46 (constant)
Object attachments 53 (46-108) 46 (46-60)
Insistence on sameness of environment 60 (45- 98) 45 (45-69)
Rigid adherence to routines and rituals 61 (47-119) 47 (47-54)
Repetitive use of language 61 (46-109) 46 (46-62)
Circumscribed interests 64 (45- 91) 45 (45-73)
Derivation of indices used in correlational analyses. Consistent with Turner’s (1996,
1997) study, severity summary scores from the RBI were summed across categories to
form composite severity summary scores (i.e., Repetitive Movements, Sameness
15 Non-parametric group comparisons were also conducted for the “presence of behaviour” summary scores, and the outcomes were identical.
162
Behaviour, Compulsive Behaviours, Repetitive Language, and Circumscribed Interests
composites; see Section 3.5.1.2 in Chapter 3), which were used in correlational analyses
with cognitive measures16. These composite scores generally demonstrated normal
distributions in the ASD group. One outlier (in the ASD group) on the Repetitive
Movements composite score was trimmed. For variables with skewed distributions,
scatterplots were examined for evidence of curvilinearity and multivariate outliers, and
no major problems were identified.
In order to examine the factor structure of the RBI and the statistical validity of
Turner’s (1996, 1997) categories and composite scores (which were based on classes of
repetitive behaviour derived from the literature), principal components analysis with
varimax rotation was conducted on the severity summary scores from each RBI
category (including the data from both the ASD and control groups). Evaluation of
two- and three-factor solutions indicated that a two-factor model appeared to be more
meaningful. The two factors explained 57.0% of the total variance in the RBI, with
39.7% accounted for by a High-level Repetitive Behaviours factor (eigenvalue 3.97),
and 17.3% by a Low-level Repetitive Behaviours factor (eigenvalue 1.73). Factor
loadings are displayed in Table 15.
Table 15. Factor loadings of RBI severity summary scores
RBI category
Factor 1: High–level
Repetitive Behaviours
Factor 2: Low-level
Repetitive Behaviours
Stereotyped manipulation of objects .449 .707
Stereotyped movements .793
Tic-like behaviours .779
Self-injurious behaviours .540
Compulsive behaviours .752
Object Attachments .668
Insistence on sameness of environment .865
Rigid adherence to routines and rituals .835
Repetitive use of language .721
Circumscribed interests .419
Note: Factor loadings lower than .4 are not shown
16 Correlations were also conducted using the “presence of behaviour” summary scores (which were also summed to form composite scores). These showed an almost identical pattern of correlations with cognitive measures, as well as being highly correlated with the severity summary scores.
163
As the factors derived differed from the composite scores used by Turner (1996, 1997),
factor scores for each participant were calculated using a regression equation, and these
factor scores were also used in correlational analyses with cognitive measures.
4.3.5.2 Social and communicative functioning
Group comparisons. Group comparisons were conducted for each of the three measures
of social and communicative functioning separately. On the Social Behaviour
Questionnaire (SBQ), one outlier in the control group was trimmed. A t-test revealed
that, as expected, participants in the ASD group (M = 16.22, SD = 5.79) scored
significantly higher on the SBQ, indicating more abnormal social behaviours than the
control group (M = 5.06, SD = 4.75), t(88) = 10.0, p < .001. Unsurprisingly, there were
also significant group differences indicating a higher number of abnormal current
behaviours in the ASD group in the social domain of the ADI-R, (ASD group: M =
15.02, SD = 7.92; Control group: M = 2.33, SD = 3.21), t(47) = 2.74, p < .01, and in the
communication domain, (ASD group: M = 17.0, SD = 5.54; Control group: M = 4.0, SD
= 4.36), t(47) = 3.97, p < .001.
Derivation of indices used in correlational analyses. As mentioned in Section 3.5.2.2
of the previous chapter, a principal components analysis was conducted with scores
from the SBQ and scores on current behaviours only from the Social and
Communication domains of the ADI-R, which showed that all three measures loaded on
one factor (smallest factor loading = .80) which explained 75.29% of the variance in the
sample (eigenvalue 2.26). Factor scores for each participant were calculated using a
regression equation, on which higher scores indicated more abnormal
social/communicative functioning. This social/communication score was used in all
correlational analyses.
164
4.3.6 Correlations between ToM/EF and behavioural measures
The explanatory value of ToM and EF impairments was examined by correlating
cognitive task performances with behavioural indices17. As the incidence of repetitive
behaviours and abnormal social behaviours was very low in the control group,
correlations between cognitive and behavioural measures were conducted for the ASD
group only. If raw correlations were statistically significant, partial correlations
(controlling for age, PIQ and VIQ) were also conducted. Table 16 displays raw
correlations and relevant partial correlations between cognitive measures and
behavioural factors (i.e., the two RBI factor scores and the social/communication factor
score). High-level repetitive behaviours correlated only with the Uses of Objects correct
responses (in an unexpected direction, such that a higher number of correct responses
was associated with more severe high-level repetitive behaviours), but this correlation
was not significant when age and IQ variables were partialled out. Low-level repetitive
behaviours showed significant raw and partial correlations with the Opposite Worlds
time difference score, a verbal measure of inhibition (in the expected direction, such
that poorer inhibitory ability was correlated with increased severity of low-level
repetitive behaviours). The social/communication factor showed a significant raw
correlation in the expected direction with the Stamps task complexity score, which
remained significant when age and IQ variables were controlled.
These results demonstrate that the behavioural symptoms of ASDs showed
different patterns of correlation with cognitive measures. In general, however, there
were few significant correlations (with high-level repetitive behaviours in particular
being poorly explained by the available data). Of note, the false belief aggregate score
did not correlate significantly with any behavioural factors18. Dewey Stories, a higher
level measure of social cognition, also showed no significant correlations with any
behavioural factors, including social/communicative functioning. The only EF variables
to correlate significantly with behavioural factors were select measures of verbal
inhibition and non-verbal generativity.
17 As for the correlations conducted between age, PIQ, VIQ, and cognitive task variables, the same computational formula was used for correlations between continuous variables (i.e., Pearson product-moment correlation coefficients) and correlations between continuous and dichotomous variables (i.e., point biserial correlation coefficients), as recommended in Cohen and Cohen (1983). 18 Correlations were also conducted for all false belief tasks individually as well as for the alternative aggregate score, but no significant correlations emerged.
165
Table 16. Raw and partial correlations between cognitive measures and behavioural
factors within the ASD group
Factor score
Cognitive task
High-level
Rep. Behaviours
Low-level
Rep. Behaviours
Social/
Communication
ToM (n = 43):
False belief aggregate .17 .08 .24
Social Cognition (n = 17):
Dewey Stories total -.18 -.08 .09
Planning:
ToL (n = 46):
Adj.extra move score -.02 -.04 -.04
Rule violations -.06 .01 -.02
Set-shifting:
IDED Perseveration condition (n = 35):
EDS stage errors .10 -.33 -.14
IDED Learned Irrelevance condition (n = 36):
EDS stage errors .0 .20 .0
Inhibition:
RIL task error difference scores (n = 35 except inhibition score, n = 36):
Inhibition .03 .12 .22
Load .04 -.21 -.08
Inhibition + load .05 -.09 .12
RIL task RT difference scores (n = 35 except inhibition score, n = 36):
Inhibition -.03 .14 -.07
Load -.12 -.23 -.15
Inhibition + load -.11 -.10 -.19
Opposite Worlds (n = 29):
Error diff. score .06 -.04 -.03
Time diff. score -.07 .38* .47* .12
166
Table 16 continued
Factor score
Cognitive task
High-level
Rep. Behaviours
Low-level
Rep. Behaviours
Social/
Communication
Working memory:
RIL shape error score .10 .22 .20
Relational Reasoning:
Relational Complexity (n = 46):
Total score .18 -.09 -.10
Generativity:
Pattern Meanings (n = 46):
Correct responses .06 .03 -.06
Sum of errors -.07 .05 .0
Uses of Objects (n = 46):
Correct responses .33* .21 -.05 .0
Sum of errors -.03 -.12 .16
Stamps task (n = 41):
Complexity score .10 -.07 -.38* -.52**
Originality score .28 .09 .13
Restriction score .0 -.15 -.03
Rule adherence score -.03 -.16 -.29
* p < .05; ** p < .01; *** p < .001.
Note: Partial correlations controlled for age, VIQ and PIQ. All tests were two-tailed.
Ns listed for each task show the sample size for correlations with the behavioural
factors.
167
Correlations between cognitive task variables and RBI composite scores (equivalent to
those used by Turner, 1996, 1997) were also of interest, both in terms of examining
patterns of correlations with more specific types of behaviour and determining whether
these results replicate those reported by Turner. These are presented in Table 17.
Repetitive Movements demonstrated significant raw and partial correlations in the
expected direction with the Opposite Worlds time difference score, a verbal measure of
inhibitory capacity (consistent with the correlation between this variable and Low-level
Repetitive Behaviours). Sameness Behaviour, Compulsive Behaviours and Repetitive
Language did not correlate significantly with any cognitive task variables.
Circumscribed Interests demonstrated significant raw correlations with three variables
(false belief aggregate, Uses of Objects correct responses, and Stamps task restriction
score), all in the opposite direction than expected; however, none of these correlations
remained significant when age and IQ were partialled out.
Overall, each RBI composite score demonstrated a unique pattern of correlations
with cognitive task variables, although again there were few significant correlations,
with only the Repetitive Movements composite showing a significant partial correlation
with a cognitive variable. When age and IQ were controlled, ToM and social cognition
variables did not correlate with any RBI composite scores. Only one EF measure, of
verbal inhibition, correlated significantly with an RBI composite (Repetitive
Movements, as described above).
168
Table 17. Raw and partial correlations between cognitive measures and RBI composite scores within the ASD group RBI composite score Cognitive task
Repetitive Movements
Sameness Behaviour
Compulsive Behaviours
Repetitive Language
Circumscribed Interests
ToM (n = 43): False belief aggregate .14 .03 .08 .11 .32* .12 Social Cognition (n = 17): Dewey Stories total -.09 -.14 -.14 -.14 -.05 Planning: ToL (n = 46):
Adj.extra move score -.03 -.04 -.04 .0 -.03 Rule violations -.05 -.01 -.19 .24 -.04 Set-shifting: IDED Perseveration condition (n = 35): EDS stage errors -.31 -.02 .06 .07 -.11 IDED Learned Irrelevance condition (n = 36): EDS stage errors .23 .0 .08 .03 -.05 Inhibition: RIL task error difference scores (n = 35 except inhibition score, n = 36): Inhibition .09 .16 .13 -.01 -.20 Load -.22 .01 -.10 .13 -.16 Inhibition + load -.11 .10 .01 .10 -.29 RIL task RT difference scores (n = 35 except inhibition score, n = 36): Inhibition .10 .0 -.03 .16 .14 Load -.28 -.12 -.14 -.07 -.13 Inhibition + load -.18 -.07 -.14 .07 -.01
169
Table 17 continued RBI Composite Score Cognitive task
Repetitive Movements
Sameness Behaviour
Compulsive Behaviours
Repetitive Language
Circumscribed Interests
Inhibition continued: Opposite Worlds (n = 29): Error diff. score -.01 -.01 .04 .05 -.06 Time diff. score .37* .48* .08 .0 .09 .10 Working memory (n = 35): RIL shape error score -.18 .15 .12 .31 .07 Relational Reasoning: Relational Complexity (n = 46): Total score -.06
.08
.27
.02
.23
Generativity: Pattern Meanings (n = 46): Correct responses .07 .0 -.08 .07 .15 Sum of errors .01 -.06 -.09 .13 .05 Uses of Objects (n = 46): Correct responses .06 .18 .27 .08 .35* .17 Sum of errors -.10 -.05 -.11 .11 -.24 Stamps task (n = 41): Complexity score -.03 -.02 .15 -.13 .15 Originality score .12 .13 .23 .24 .31 Restriction score -.16 -.10 .10 .21 -.31* -.15 Rule adherence score -.16 -.05 -.10 -.08 -.09 *p < .05; ** p < .01; *** p < .001. Note: Partial correlations controlled for age, VIQ and PIQ. All tests were two-tailed. Ns listed for each task show the sample size for correlations with the RBI severity summary scores.
170
4.3.7 Relationship between ToM and EF
The relationship between ToM and EF in the ASD and control groups was investigated
by examining both correlations and dissociations between the two domains. The Dewey
Stories total score was omitted from these analyses, for two main reasons: firstly
because it does not appear to measure ToM and therefore any relationships with EF
would be difficult to interpret within the theoretical frameworks that exist regarding the
ToM-EF relationship; and secondly because only older participants completed the task
and so the sample size for correlations was small19.
4.3.7.1 Correlations between ToM and EF
Correlations between task variables were calculated separately for the ASD and control
groups. Again, partial correlations (controlling for the effects of age, VIQ and PIQ)
were conducted if raw correlations were significant. Table 18 presents raw and relevant
partial correlations between ToM and EF task variables within the control group.
Correlations are displayed separately for the various false belief variables rather than the
overall aggregate score because the pattern of correlations was different for the three
tasks. In this group, simple false belief task performance correlated with the ToL
adjusted extra move score, and the Pattern Meanings and Uses of Objects sum of errors
(with all correlations in the expected direction, such that poor false belief performance
correlated with poor EF task performance); however when age, VIQ and PIQ were
controlled, only the correlation with the Pattern Meanings sum of errors remained
significant. First-order false belief task performance correlated with the ToL adjusted
extra move score, Relational Complexity total score, Uses of Objects correct responses
and Stamps task originality score (all in the expected direction), but only the
correlations with ToL and Uses of Objects variables were significant when age and IQ
were partialled out. Second-order false belief task performance correlated with the ToL
adjusted extra move score and rule violations, the RIL task load error and RT difference
scores and inhibition RT difference score, the Relational Complexity total score, the
Uses of Objects correct responses, and Stamps task originality score (all in the expected
direction except the RIL task load RT difference score); but with age and IQ controlled,
19 Correlations were conducted out of interest, but did not reveal much of importance. There were only a small number of significant correlations with EF variables in the control and ASD groups, and a few of these were in the opposite than expected direction.
171
only the correlations with ToL rule violations, RIL task load error and RT difference
scores, and the Stamps task originality score remained significant.
Overall, in the control group, ToM variables demonstrated relationships with
measures of planning, non-verbal inhibition (under working memory load conditions),
relational reasoning, and both verbal and non-verbal generativity, but several of the
correlations were mediated by age and IQ effects (there were no significant partial
correlations with relational reasoning ability). All correlations were in the expected
direction - such that poorer performance on EF tasks correlated with poorer false belief
task performance - with the exception of the RIL task load RT difference score. A
possible explanation for this is that participants who performed well on false belief tasks
made fewer errors on the working memory load condition, but at the expense of speed
(i.e., they demonstrated a cautious speed/accuracy tradeoff). Another noticeable aspect
of the pattern of correlations was that there tended to be more correlations with the
second-order than the simple and first-order false belief tasks, which is likely to be
partly due to the fact that only a small proportion of participants in the control group
failed to obtain a perfect score on the lower-order tasks.
Table 19 displays raw and partial correlations between ToM and EF tasks within
the ASD group. In children with ASDs, simple false belief task performance correlated
with the ToL adjusted extra move score, Uses of Objects correct responses, and the
Stamps task restriction score (with all correlations in the expected direction); however
when age and IQ variables were controlled, only the correlation with the Stamps
restriction score remained significant. First-order false belief task performance
correlated with the ToL adjusted extra move score, Uses of Objects correct responses
and Stamps task originality score (all in the expected direction), but none of the
correlations were significant when age and IQ were partialled out. Second-order false
belief task performance correlated with the ToL adjusted extra move score and rule
violations and the Uses of Objects correct responses (all in the expected direction); none
of these correlations were significant with age and IQ controlled.
Overall, the ASD group showed noticeably fewer significant correlations
between ToM and EF variables than the control group, with only one correlation
remaining significant with age and IQ controlled (between simple false belief
performance and a non-verbal measure of generativity).
172
Table 18. Raw and partial correlations between ToM and EF measures within the control group False belief task _______________________________________________ EF task Simple 1st-order 2nd-order ToL (n = 46): Adj. extra move score -.30* -.26 -.40** -.35* -.45** -.27 Rule violations -.06 -.01 -.48** -.34* IDED Set-shifting task condition (n = 34): Perseveration EDS stage errors -.27 -.25 -.28 Learned Irrelevance EDS stage errors -.13 -.23 -.13 RIL task (n = 33): Error difference scores: Inhibition .27 .07 -.02 Load -.16 -.33 -.45** -.54** Inhibition + load .12 -.15 -.30 RT difference scores: Inhibition -.05 -.05 -.44* -.35 Load .26 .24 .42* .52** Inhibition + load .19 .18 -.02 Shape error score -.20 -.19 -.31 Opposite Worlds (n = 35): Error difference score .08 -.20 .03 Time difference score .02 -.03 -.09 Relational Complexity (n = 46): Total score .26 .30* .07 .48** .14 Pattern Meanings (n = 46): Correct responses -.02 .28 .15 Sum of errors -.37* -.32* -.10 -.26 Uses of Objects (n = 46): Correct responses .11 .44** .30* .51*** .28 Sum of errors -.34* -.27 -.19 -.28 Stamps task (n = 45): Complexity score .01 .28 .29 Originality score .23 .40** .25 .56*** .40** Restriction score .04 .07 .09 Rule adherence score -.10 .02 -.06 * p < .05; ** p < .01; *** p < .001. Note: Partial correlations controlled for age, VIQ and PIQ. All tests were two-tailed. Ns listed for each EF task show the sample size for the correlations with the ToM tasks.
173
Table 19. Raw and partial correlations between ToM and EF measures within the ASD group False belief task _________________________________________________ EF task Simple 1st-order 2nd-order ToL (n = 43): Adj. extra move score -.33* -.07 -.53***-.24 -.35* -.04 Rule violations -.25 -.16 -.30* -.10 IDED Set-shifting task condition: Perseveration (n = 32): EDS stage errors -.15 .08 -.23 Learned Irrelevance (n = 33): EDS stage errors .15 -.22 -.19 RIL task (n = 32 except inhibition difference scores, n = 33): Error difference scores: Inhibition -.28 -.13 -.08 Load -.18 -.03 -.17 Inhibition + load -.32 -.09 -.17 RT difference scores: Inhibition .05 .13 .0 Load -.17 .07 .07 Inhibition + load -.16 .13 .04 Shape error score -.09 -.05 .05 Opposite Worlds (n = 29): Error difference score -.24 .01 .08 Time difference score -.21 .06 -.12 Relational Complexity (n = 43): Total score .10 .23 .27 Pattern Meanings (n = 43): Correct responses .19 .01 .26 Sum of errors .02 -.10 .09 Uses of Objects (n = 43): Correct responses .32* .11 .31* .03 .48** .31 Sum of errors .13 .02 .12 Stamps task (n = 41): Complexity score .23 .08 -.01 Originality score .11 .39* .10 .30 Restriction score -.56***- .46** -.16 -.21 Rule adherence score .15 .01 -.15 * p < .05; ** p < .01; *** p < .001. Note: Partial correlations controlled for age, VIQ and PIQ. All tests were two-tailed. Ns listed for each EF task show the sample size for the correlations with the ToM tasks.
174
Table 20 presents a summary of the significant partial correlations between ToM and
EF domains in the control and ASD groups, clearly demonstrating the different pattern
of correlations in the two groups.
Table 20. Summary of significant partial correlations between ToM and EF variables
in the control and ASD groups
ToM
EF domain Control group ASD Group
Planning
Set-shifting
Inhibition – Non-verbal *
Inhibition – Verbal
Working Memory
Relational Reasoning
Generativity – Verbal
Generativity – Non-verbal
* Correlations marked with an asterisk were in the opposite direction than expected.
Note: Each tick represents one significant correlation between a false belief and an EF
variable in that domain.
4.3.7.2 Dissociations between ToM and EF
While the correlative evidence presented in the previous section was suggestive of a
relative independence between ToM and EF in the ASD group compared with controls,
it was also of interest to examine the incidence and direction of dissociations between
ToM and EF deficits within the ASD group. This was achieved by defining a deficit on
any given task in the same way as for the universality calculations in Section 4.3.3. The
proportion of ASD participants with a ToM deficit who displayed unimpaired
performance on EF tasks was calculated, and conversely, the proportion of participants
with a given EF deficit who displayed unimpaired ToM performance was also
calculated. For ease and simplicity of interpretation, the false belief alternative
aggregate score was used as the measure of ToM performance (as for the universality
calculations) rather than analysing all the false belief variables separately. However, all
EF variables on which significant group differences were found were analysed
separately. The results of these calculations are displayed in Table 21.
175
Table 21. The incidence of ToM-EF dissociations in the ASD group
EF measure
% of ToM-impaired ASD participants
with unimpaired EF
N
ToL: Adjusted extra moves score 47.4 19
Rule violations 36.8 19
Opposite Worlds time difference score 44.4 9
RIL shape error score 54.5 11
Uses of Objects correct responses 47.4 19
Stamps task: Complexity score 83.3 18
Originality score 50.0 18
% of EF-impaired ASD participants
with unimpaired ToM
ToL: Adjusted extra moves score 23.1 13
Rule violations 40.0 20
Opposite Worlds time difference score 64.3 14
RIL shape error score 58.3 12
Uses of Objects correct responses 44.4 18
Stamps task: Complexity score 62.5 8
Originality score 25.0 12
These data clearly demonstrate that dissociations between ToM and EF occurred
relatively frequently (usually in around 50% of the participants showing impairments)
and in both directions, such that the presence of a ToM impairment did not necessarily
result in an EF impairment and vice versa. These results are consistent with the
correlative data in indicating an independence between ToM and EF deficits in the ASD
group.
4.4 Discussion
This section includes four subsections. The first three of these examine the profile,
primacy and independence of ToM and EF deficits in ASDs in this study, comparing the
current findings to those of previous studies and considering alternative interpretations
of the results. In the final section, an attempt is made to interpret the outcomes in terms
of the six alternative hypotheses regarding primacy and independence outlined in the
introduction.
176
4.4.1 Profile of ToM and EF deficits
As predicted, both ToM and EF deficits were found in this sample of individuals with
ASDs. However, a unique profile of spared and impaired abilities emerged, which
included both expected and unexpected features.
Profile of ToM deficits. In the ToM domain, a higher proportion of ASD
participants than controls displayed unstable performance on all false belief tasks and on
aggregate scores, although partialling out age and VIQ reduced the significance of
group comparisons on the simple and second-order tasks as well as on the alternative
aggregate score (which involved a more lenient scoring criterion). These results suggest
that false belief understanding was significantly impaired in the ASD group, but that on
two of the tasks the impairment was partially attributable to the poorer verbal skills of
the ASD participants20. This lack of robustness of ToM deficits on false belief tasks
was underscored by the relatively high percentage of ASD participants who
demonstrated errorless performance on the tasks, which ranged from 48.8% on the
standard first-order task to 72.1% on the simple (unexpected contents and unexpected
identity) tasks, with 39.5% displaying perfect performance across all tasks. According
to the alternative aggregate score which more reliably indicates poor performance,
55.8% of ASD participants were high scorers on false belief tasks. The highest
percentage of first-order false belief task passers found in previous studies was 90%
(Dahlgren & Trillingsgaard, 1996), with the next highest being 55% (Prior et al., 1990).
Although the high 72.1% on the simple false belief task is likely to be an overestimation
due to the fact that perfect performance on it was assumed if the first-order task was
passed (for 7-16 year-olds who began with the first-order task), it is nevertheless clear
that the sample of ASD participants in this study demonstrated better false belief task
performance than the majority of samples from previous studies. The finding that false
belief performance was significant correlated with both age and VIQ suggests that the
relatively old mean age and high level of verbal ability of the sample probably explains
this good false belief performance, consistent with previous studies demonstrating that
individuals with autism passing false belief tasks tend to be older and have higher verbal
ability (e.g., Eisenmajer & Prior, 1991; Prior et al., 1990; Sparrevohn & Howie, 1995).
20 Notably, the effect of group on performance on the simple and second-order false belief tasks remained significant when age only was included in the regression, but did not remain significant when VIQ only was included, suggesting that VIQ had a greater impact on the significance of group differences than did age.
177
The age and level of ability of the control sample meant that a high percentage
of controls also demonstrated flawless false belief performance. However, it is of note
that the use of a fairly strict scoring criterion prevented extreme ceiling effects, with
30.4% of controls demonstrating unstable false belief performance on the aggregate
score. Even using the more lenient alternative aggregate, 19.6% of controls emerged as
low scorers, the majority of whom were between 6 and 10 years of age. This suggests
that beyond the age of 5, either ToM is still developing or other cognitive factors may
be influencing false belief performance (this latter possibility is discussed further below
in Section 4.4.3).
Despite the fact that ceiling effects did not pose a significant problem on the
false belief tasks, the relatively high proportion of perfect performances in both the
ASD and control groups indicates that the assessment of ToM in this study would have
been significantly strengthened by the inclusion of a more advanced ToM task such as
Happé’s (1994a) “Strange Stories” task or Baron-Cohen et al.’s (1997, 2001a) “Eyes
Task”. The use of the Dewey Stories task represented an attempt to tap into higher-
level social cognitive skills, however its lack of correlation with false belief variables
(after partialling out age, PIQ, and VIQ) indicates that it is questionable as a measure of
mentalising ability and can probably be successfully performed by drawing on more
declarative knowledge of social norms. In light of this, it is noteworthy that ASD
participants did not show a significant impairment on the task, with marginally
significant differences reducing to non-significance when VIQ was accounted for
(although the medium effect size of the group difference suggested that the marginal
significance may have been due to the small size of the sample who completed the task).
This suggests that high-functioning individuals with ASDs often have intact knowledge
of what is considered “normal” or appropriate, but that this knowledge does not aid or
interact with either their mentalising skills or their own social skills (Dewey Stories
performance did not correlate significantly with social/ communicative functioning).
Interestingly, a similar pattern of results has been demonstrated previously with patients
with damage to the ventromedial prefrontal cortex (e.g., Saver & Damasio, 1991).
Profile of EF deficits. ASD participants also demonstrated an interesting pattern
of strength and weakness on the various EF components tested. Consistent with
previous research, individuals with ASDs displayed robust planning impairments on the
ToL, both in terms of the number of extra moves made and the frequency with which
the rules of the task were violated. However, the small to medium effect size was
somewhat lower than expected, with Pennington and Ozonoff (1996) reporting an
178
average effect size of 2.07 on the similar Tower of Hanoi (ToH) task across the studies
conducted up until then. This discrepancy probably cannot be attributed to the age or
level of functioning (i.e., IQ level) of the sample because previous studies using Tower
tasks have also used older, high-functioning participants. One difference between the
ToL administration procedure used in this study compared with other studies using the
ToL and ToH is that during the initial task instructions participants were actively
encouraged to plan the movements of the discs in advance. This may have positively
influenced performance on the task and reduced the size of the difference between the
ASD and control groups. Nevertheless, the fact that planning impairments persisted in
the ASD participants despite this extra cueing provides evidence of the severity of their
deficit in this domain. Furthermore, as the ToL and ToH have been found to hold
slightly different cognitive demands (e.g., Welsh et al., 1999), a comparison of effect
sizes across the two tasks should be viewed with caution (unfortunately, the only other
study to use the ToL rather than the ToH - Hughes et al.’s (1994) study - did not report
standard deviations and therefore the effect size from that study could not be directly
compared). Following from this, it should also be noted that Welsh et al. (1999) found
that ToL performance tapped working memory and inhibition as well as planning
ability, and therefore the poor ToL performance demonstrated by the ASD group may
not necessarily reflect a planning impairment. However, the lack of group differences
on separate and more direct measures of working memory and non-verbal inhibition (as
discussed below) make this unlikely, supporting the interpretation of the ToL result as
indicating a planning deficit in the ASD group.
The absence of significant impairments in attentional shifting abilities on the
IDED set-shifting task was an unexpected result, given fairly consistent evidence of set-
shifting difficulties in previous studies (e.g., Ciesielski & Harris, 1997; Hughes &
Russell, 1993; Hughes et al., 1994; Ozonoff et al., 1994). A difficulty with mental
flexibility holds an intuitive appeal in explaining autistic symptoms such as
perseveration and rigid adherence to routines and rituals, and Ozonoff (1997b) has
suggested that a shifting impairment may in fact be the key feature of the EF profile
which characterises autism. However, there have been at least two other studies which
have also failed to find set-shifting deficits in autism. Ozonoff et al. (2000) found no
significant difference between their high-functioning autistic participants and controls
on the original IDED set-shifting task from the CANTAB battery. They attributed this
result to the fact that their task was computerised, thereby facilitating the performance
of their autistic participants. However, the fact that their participants with Asperger
179
syndrome did show impaired performance on the task, along with Hughes et al.’s (1994)
previous finding of a deficit on the same computerised task in individuals with autism,
speak against this explanation.
Turner (1997) also found that her high-functioning participants with autism
displayed intact performance on both conditions of the modified IDED set-shifting task
used in the current study, although her low-functioning participants demonstrated
impairments on the EDS stage of the Perseveration condition. There was also evidence
of a marginally significant difference in set-shifting performance in the current sample,
but contrary to Turner’s results this occurred in the Learned Irrelevance condition.
These two negative results using the same task (i.e., Turner’s and the current study)
necessitate some decomposition of the requirements of the task. Although the design of
the modified IDED task allows more specific analysis of the component processes
involved in the task than the original version, it appears to do this at the expense of the
impact and obviousness of the shift. As Turner pointed out, in the modified IDED task,
the change in stimulus dimension that occurs in the EDS stage of both conditions (i.e.,
the introduction of the new relevant stimulus dimension of solidity in the Perseveration
condition and the new irrelevant dimension of size in the Learned Irrelevance condition)
signals very clearly that the task has changed. This means that it is fairly easy for the
participant to deduce the rules of responding for that condition without relating them to
the previous condition or becoming easily “stuck” in their previous mode of responding.
This in turn suggests that either the validity of the task as a measure of set-shifting is
questionable or that the nature of the shift required is too easy for high-functioning
individuals with ASDs. The fact that Owen et al. (1993) found different impairments on
the task in patients with frontal lesions and Parkinson’s disease indicates that the
problems on the task will emerge if the shifting deficit is severe enough. Hence, the
lack of convincing evidence of impairments on the task in high-functioning autism may
indicate that a set-shifting or cognitive flexibility deficit may not be as central to autism
as previously thought. Most of the initial studies on which this notion was based used
the WCST as their measure of cognitive flexibility, on which impaired performance
may be caused by a range of different factors. The variability in findings on more pure
set-shifting measures such as the IDED tasks calls into question the importance of the
role of set-shifting in the EF profile characteristic of autism.
Results obtained in the inhibition domain were also contrary to predictions and
added to previous research in an interesting way. Most earlier studies have not found
impairments in inhibition in individuals with autism (Brian et al., 2003; Ozonoff et al.,
180
1994; Ozonoff & Jensen, 1999; Ozonoff & Strayer, 1997), and those that have found
apparent inhibition deficits have used tasks on which performance could be influenced
by other EF capacities such as cognitive flexibility, working memory or generativity
(Hughes, 1996b; Rinehart et al., 2002; Williams et al., 2002). Notably, all of the studies
finding intact inhibition in autism have used non-verbal tasks except one study in which
the Stroop task was used (Ozonoff & Jensen, 1999). In the present study, previous
findings of unimpaired non-verbal inhibitory control in autism were replicated using the
newly developed RIL task, on which neither accuracy or RT measures revealed
inhibitory difficulties in the ASD group. However, significant and robust verbal
inhibition impairments were found on the Opposite Worlds test, particularly on RT
measures (a trend was also evident on error measures). This result stands in contrast to
that obtained by Ozonoff and Jensen (1999) with their autistic sample of similar size
and age range using the Stroop, which involves very similar verbal inhibitory
requirements. Closer inspection of Ozonoff and Jensen’s data reveals, though, that the
autism group in their study performed at a very similar level to their ADHD group
(autism group mean = 27.7 versus 27.4 for the ADHD group, on an unspecified scale),
the latter of which did differ significantly from the control group (mean = 32.0). The
lack of a significance difference from controls in the case of the autism group was likely
to have been due to their larger standard deviation (11.4 versus 7.0 for the ADHD
group). However, it is also notable that while the ADHD group was matched with
controls on all age and IQ variables, the autism group was not matched to the control
group on VIQ, PIQ, or Full-Scale IQ (FSIQ), and this was handled by covarying FSIQ
in all group comparisons. As discussed in Section 4.3.2, ANCOVA is not considered an
appropriate statistical technique for accounting for group differences in cases such as
this, as it may also remove part of the effect of group. Ozonoff and Jensen’s result may
therefore represent a false negative. It will be interesting to monitor the outcome of
further studies on verbal inhibition, particularly in regard to how inhibition performance
in autism may be distinguished from that displayed in ADHD.
The interaction between inhibition and working memory was another topic of
interest for this study, with Russell (1997b) proposing that impairments in these
domains only emerge in autism if the task at hand requires both abilities simultaneously.
Although inhibition deficits were revealed on a verbal task with minimal working
memory requirements, results from the non-verbal RIL task were largely consistent with
this proposal. While ASD participants were able to successfully perform the RIL task
condition involving only non-verbal inhibitory requirements (and as discussed further
181
below also showed intact performance on the Relational Complexity task, which
arguably requires working memory but not inhibition), on the condition involving both
inhibition and working memory requirements, the ASD participants made significantly
more errors on a measure of working memory capacity. There was also a trend for the
ASD group to make more errors on a measure of inhibitory capacity for this condition
as compared with the control condition. This suggests that in situations where both
(non-verbal) inhibition and working memory are required, individuals with ASDs are
unable to maintain an adequate level of performance in either domain, but particularly
in working memory (although it may be the case that the working memory component
of the task was more vulnerable in this case because that task requirement was added
after the inhibitory component and was therefore more novel; or, alternatively, because
it was tested less frequently).
No group differences were identified on the Relational Complexity task,
suggesting that the capacity to integrate multiple relations in parallel (Halford, 1993;
Halford et al., 1998; Waltz et al., 1999) is not impaired in children with ASDs. This
result further indicates that failure on false belief tasks in children with ASDs is unlikely
to have its basis in a working memory or relational reasoning deficit. This was
confirmed by the lack of significant correlations between false belief and Relational
Complexity performance in either the ASD or control group. However, although Waltz
et al. (1999) found that frontal lobe patients were significantly impaired on their version
of the Relational Complexity task, the validity of the task as a measure of relational
reasoning is yet to be determined. It could be argued that the task does not tap working
memory or integrative capacity as strongly as it first appears. As the stimuli and all
possible response choices are always present and visible to the participant, it is possible
that the participant can check the accuracy of each response choice against the
requirements of each relational change one by one, rather than having to hold in mind
all the relevant relational changes simultaneously. All that would then be required is for
the participant to notice all the relational changes which are occurring and accurately
check whether each response choice fits the sequence of each change correctly. These
requirements are quite obviously different to the relational integration arguably required
by false belief tasks. So, while results from the Relational Complexity task did not hold
much promise in suggesting a relational integration difficulty in ASDs, the use of
different kinds of relational complexity task (such as the transitive inference task also
used by Waltz et al., 1999) could be an interesting avenue for further research.
182
Results from the generativity tasks were more promising. On the verbal Uses of
Objects task, the group difference on the number of correct responses variable met the
criterion for a large effect. This is consistent with previous studies demonstrating
generativity impairments in autism using other tasks (Boucher, 1988; Craig & Baron-
Cohen, 1999; Lewis & Boucher, 1991), and in particular replicates Turner’s (1999)
study, which found that both low- and high-functioning children with autism generated
fewer responses than controls on the Uses of Objects task21. However, unlike Turner,
the ASD sample in this study did not produce a higher number of error responses.
Although the scoring systems used in the two studies were slightly different, even on
categories common to both studies such as redundant responses, there were discrepant
outcomes. Another difference between the studies was that Turner allowed 150s for her
participants to produce responses, whereas only 90s was allowed in the current study. It
is possible that during the extra 60s given in Turner’s study, a pressure to respond had
accumulated over a longer time and so the children with autism produced inappropriate
responses when they were unable to generate correct ones; whereas the children in this
study felt less of a demand to produce a response. Regardless, it appears that the
individuals with ASDs in both studies demonstrated difficulty spontaneously generating
correct verbal responses on this task.
In contrast, results from the Pattern Meanings task revealed no significant group
differences on any variable overall. This was somewhat surprising as it was also
thought to be a test of verbal generativity and ASD participants were found to produce
fewer responses and make more errors on the task in Turner’s (1999) study. However,
more detailed analyses involving the two subgroups of ASD participants meeting “full
criteria” and “partial criteria” on the ADI-R showed that the full criteria subgroup
generated fewer correct responses than controls (although this effect disappeared when
age and VIQ were controlled), and the partial criteria subgroup generated more error
responses than controls (although this effect became marginally significant when age
and VIQ were controlled). This discrepancy between the full and partial criteria
subgroups appeared to be due to a tendency for the partial criteria subgroup to produce
more responses overall than the fully autistic subgroup, such that the partial criteria
subgroup produced as many correct responses as the control group but were also more
likely to produce error responses. This suggests that the less severe subgroup (in terms
21 Turner (1999) actually calculated the total number of responses overall rather than the number of correct responses. Although it was not reported, the ASD group in the current study also produced significantly fewer responses overall than the control group, replicating Turner.
183
of the range and number of symptoms present) reacted to problems generating responses
by producing errors, whereas the more severe subgroup reacted by not producing
responses at all.
The failure to replicate Turner’s (1999) findings of significant differences in the
overall sample on the Pattern Meanings task (and the lack of robustness of subgroup
differences) requires further comment. The shorter time period allowed in the current
study may also account for the lack of robust or significant differences, but this does not
seem likely to be the sole cause given the strong generativity deficit displayed on the
Uses of Objects task in the same time period. It could be argued that Pattern Meanings
is not as good a task at discriminating those with poor generativity, as a larger range of
responses are acceptable than for the Uses of Objects task. Scoring was fairly lenient
for the task as it was often necessary to accept responses which the pattern possibly
could be, even if they were a little far-fetched. This could explain the lack of an overall
difference in the number of error responses made, but even given the lenient scoring,
one would expect a reduced number of correct and total responses if the ASD
participants experienced difficulty producing ideas. It may be that a combination of
these two explanations can account for the lack of significant overall group differences
in this study, in that the majority of children with ASDs were able to produce adequate
responses for a 90s period because they found the task easier than the Uses of Objects
task and the scoring was more lenient, but if the task had been continued for another
minute, they may have started producing fewer and more inappropriate responses.
Consistent with this interpretation, the rate of producing responses was similar for the
control participants across the two studies (approximately 1 every 3.6s in the current
study and 1 every 3.3s for the high-functioning controls in Turner’s study), but the ASD
participants in the current study produced responses at a faster rate (1 every 3.97s) than
the high-functioning ASD participants in Turner’s study (approximately 1 every 4.7s).
Results on the Pattern Meanings task in this study should not, therefore, be interpreted
as evidence against a verbal generativity deficit (although any such deficit on this task
was clearly more subtle than on the Uses of Objects task).
Non-verbal generativity impairments in ASD participants also emerged in this
study, with performances on the Stamps task revealing that individuals with ASDs
produced less complex and fewer original patterns and were more restricted in their use
of the stamps available. There was also a trend for children with ASDs to show less
adherence to one rule for each pattern. Results on the originality and restriction scores
were consistent with Frith (1972), however contrary to this study Frith found no
184
difference in the complexity of patterns produced by her sample of children with autism,
and she also found that her sample showed a very high degree of rule adherence. In the
current study, the lack of rule adherence was likely to have been attributable to a certain
proportion of the ASD participants who produced random patterns with no underlying
rule. These participants may also have been the cause of the lower mean complexity
score of the ASD group, as random or unidentifiable patterns were assigned the lowest
complexity score of 1. The main difference between the two studies was the level of
functioning of the samples, with 14 out of 20 of Frith’s participants having an estimated
PIQ below 60. It is possible that higher-functioning individuals with ASDs may have
opted to produce random patterns when unable to produce original rules, whereas
lower-functioning participants may have simply produced the same pattern repeatedly.
This hypothesis cannot be directly tested in the current sample because all participants
had PIQs above 60. Nevertheless, it is evident that the generativity impairment which
characterises autism extends across both verbal and non-verbal domains as well as
across all levels of functioning.
Concluding comments on the profile of impairments. In summary, then, the
ASD group in this study demonstrated a characteristic profile of strength and weakness
in the cognitive domains tested, with impairments on measures of ToM, planning,
verbal inhibition, tasks combining inhibition and working memory, and both verbal and
non-verbal generativity, but intact performance on tests of awareness of social norms,
set-shifting, non-verbal inhibition and relational reasoning. Consistent with predictions,
the largest effects were on verbal tasks22 (measures of false belief, verbal inhibition and
verbal generativity), consolidating the importance of including tasks involving both
verbal and non-verbal responses where possible. Certain aspects of the profile of
impairments found in this study were inconsistent with initial predictions based on
previous studies, such as the absence of set-shifting deficits, and the presence of
impairments in verbal inhibition. These findings suggest that the EF profile
characteristic of autism as proposed by Ozonoff and colleagues (e.g., Ozonoff, 1997;
Ozonoff & Jensen, 1999) may require modification, and its discriminant validity (i.e.,
its uniqueness to autism as compared with other clinical conditions) merits further
investigation.
22 It should be noted that there were no significant correlations between PIQ and any ToM or EF measures. This indicates that the measures of PIQ on which the control and ASD samples were matched measured different abilities to those measured by the non-verbal EF tasks, and therefore that the relative lack of group differences on non-verbal EF tasks compared with verbal tasks cannot be accounted for by the matching of the groups on PIQ.
185
It should be pointed out, however, that the neat profile described above of course
assumes reasonable construct validity of each task variable. This assumption deserves
some critical analysis, particularly given the well-documented difficulties with EF
measurement discussed in Chapter 2 (Section 2.2). It is possible that both i) certain
variables are not actually measuring what they are purported to and ii) there is overlap
between the EF domains measured and/or the tasks used in different domains. The ideal
way to address this uncertainty would be to conduct a factor analysis of all the EF
variables in the battery, however the high number of variables in relation to the number
of participants prevented a valid factor analysis from being performed on this sample.
Interpretation of each variable therefore relied mainly on previous literature as well as
informed qualitative analysis of the requirements of each task. The choice of relatively
pure EF tasks and/or tasks which included control conditions, allowing decomposition
of the processes involved in task performance, facilitated the ease and clarity with
which variables could be interpreted. Examination of raw and partial correlations
(partialling out age, VIQ and PIQ) between EF variables in the control group was also
informative (these are presented in Appendix B), in general demonstrating weak and
relatively few significant correlations between EF domains as well as several strong
intra-domain correlations, thereby validating the notion that the tasks measure mostly
independent constructs. This was the case even for variables which could conceivably
belong in a different category to other more central variables from the same task, such
as rule violations on the ToL or the error variables on the verbal generativity tasks (both
of which could reflect inhibition or working memory), with these variables usually
correlating more strongly with other variables from the same task than those from other
tasks. It appears, therefore, that there is no strong evidence to suggest that the
underlying abilities assumed to be measured by each of the EF variables are invalid.
4.4.2 Primacy of ToM and EF deficits
Having identified the ToM and EF profile which characterises this sample of individuals
with ASDs, the next question concerns the primacy of each of these deficits. In this
study, primacy was measured by calculating the universality, uniqueness (this criterion
was measured indirectly), and explanatory value of each variable on which significant
group differences were found (or all variables in the case of explanatory value). Results
showed that while ToM and EF deficits showed similar prevalence within the ASD
group, measures of ToM did not successfully discriminate between the ASD and control
186
groups or show any significant relationships with behavioural measures, yet several EF
indices emerged as significant predictors of autism group membership and two EF
variables correlated significantly with measures of symptomatology. Overall, it would
appear that EF deficits are relatively more primary23 than a ToM deficit in ASDs.
However, before making any strong conclusions, results derived from each index of
primacy require a more detailed discussion.
Universality. The first matter of note is that neither ToM nor EF deficits, as
defined by a score worse than one standard deviation from the mean of the control
group (or a close approximation in the case of dichotomous variables), were universal
among this sample of high-functioning individuals with ASDs. Within the ASD group,
44.2% displayed a ToM deficit and the prevalence of EF deficits ranged from 19.5% to
48.3% (with impairments in verbal inhibition and verbal generativity being the most
prevalent of the EF components). All deficits remained non-universal even using the
more lenient criterion of any score below the mean of the control group, contrary to the
results obtained by Ozonoff et al. (1991) which showed that deficits defined in this way
on their EF composite were almost universal (96%) amongst their autism group whereas
ToM deficits were not (52% on a first-order composite and 87% on a second-order
composite). However, the current results are consistent with most other studies which
report prevalence data on ToM and/or EF impairment in autism, the majority of which
have not found either ToM (see Happé, 1995) or EF deficits (e.g., Liss et al., 2001;
Hughes et al., 1994; Ozonoff & Jensen, 1999) to be universal. It should also be noted
that it is unlikely that EF deficits would have been universal in the Ozonoff et al. (1991)
study if a stricter definition of a deficit had been used.
In any case, unless the ToM and EF tasks used were too easy for a proportion of
the participants, these results suggest that neither a ToM nor EF deficit is the single
primary deficit in autism, but rather (as outlined in hypothesis 6 in the introduction),
that either i) different ToM and EF profiles are found in different subgroups of
individuals with autism, rather than both deficits co-occurring in all individuals; ii) ToM
and EF deficits underlie different aspects of symptomatology, and therefore may be
present in differing degrees of severity according to the individual’s position on the
multidimensional autism spectrum; or iii) an unidentified third deficit may be more
primary or at least equally primary. A fourth possibility is also conceivable, which is
23 The notion of “relative primacy” refers to the relative ability of each impairment to meet the criteria for a primary deficit. Although the term “primary” is usually used in the context of a single primary deficit, in a multiple primary deficits model it is also possible for one deficit to hold superior causal importance (e.g., explanatory value) over another, and therefore have superior relative primacy.
187
that different developmental stages of autism are characterised by different cognitive
profiles. These four possibilities will be re-visited later in this section and discussed
further in Section 4.4.4.
Uniqueness. Results on the uniqueness criterion more clearly discriminated
between the ToM and EF accounts, with verbal measures of inhibition and generativity
being the strongest predictors of autism group membership (deficits on these two
variables were also the most universal among the ASD group and had the largest effect
sizes of all the EF variables), while first-order false belief performance was only a
marginally significant predictor24. While these results do not allow any strong
inferences regarding the uniqueness of these deficits to autism as opposed to other
clinical groups, they do suggest that deficits in verbal inhibition and verbal generativity
are particularly central to ASDs. This is an interesting result given that mental
flexibility and planning deficits were previously thought to be the most significant in
autism. It also adds to the previous study by Ozonoff et al. (1991), which showed that
EF performance was the best predictor of autism group membership, but did not analyse
the key EF components involved.
Explanatory value. In terms of explanatory value, correlations between
cognitive and behavioural measures revealed that ToM variables did not correlate
significantly with any behavioural domain, whereas two EF measures showed
significant relationships with various aspects of autistic symptomatology. The lack of
explanatory value of the ToM tasks, particularly the non-significant relationship
between ToM and social/communicative functioning, was a somewhat surprising result,
although not without precedent (Prior et al., 1990; Sparrevohn & Howie, 1995; the lack
of relationship with repetitive behaviours is also consistent with Turner, 1997). If a
ToM deficit is the primary basis for social/communicative impairments in autism then
one would expect that those who performed poorly on the ToM tasks would have been
those who showed more severe social impairment. Yet this was not the case: although
the correlation was not significant, its direction actually suggested the opposite trend,
such that those with better performance on the false belief aggregate tended to show
more abnormal social/communicative functioning25. The reason for this is unclear, but
24 When the false belief alternative aggregate score was used, it was a non-significant predictor of group membership. 25 This unexpected trend still existed when the false belief variables were analysed separately and when the Social Behaviour Questionnaire and the Social and Communication domains of the ADI-R were analysed separately rather than together as one factor score. This suggests that it was not simply a spurious individual result, which was a possibility given that only one correlation between ToM and social/communicative functioning was conducted (in comparison with the wider range of EF tasks and measures of repetitive behaviour).
188
in any case it constitutes evidence against the notion that a ToM deficit underlies the
social/communicative impairments which characterise autism. Although an inability to
appreciate others’ mental states is an intuitively appealing explanation for abnormal
social behaviours, a one-to-one relationship between an emergent behaviour and
underlying cognitive deficit cannot be assumed; abnormal social behaviours are not
necessarily caused by an impairment in a social or ToM module (Bowler, 2001). The
existence of a significant correlation in the appropriate direction between
social/communicative functioning and an EF measure casts further doubt on the idea
that ToM deficits underlie the social/communicative symptoms of autism while EF
deficits underlie repetitive behaviours and restricted interests.
EF measures demonstrated better explanatory value than ToM variables, but
there were still only two EF variables showing significant correlations with behaviour: a
measure of verbal inhibition, which correlated with low-level repetitive behaviours
(when the RBI data was broken down further using Turner’s (1996, 1997) categories,
the verbal inhibition measure correlated with repetitive movements); and a non-verbal
generativity measure, which correlated with social/communicative functioning. This
latter result consolidates previous findings of a predictive relationship between EF
impairment and abnormal social behaviours in autism, although previous studies
identified different EF components, most commonly set-shifting, as holding explanatory
value (Berger et al., 2003; Gilotty et al., 2002; McEvoy et al., 1993). This may be
because the previous studies incorporated different measures of set-shifting to the one
used here and did not include any tests of non-verbal generativity. The significant
correlation between verbal inhibition and repetitive movements was consistent with
Turner’s (1997) findings, which showed that performance on a test of “recurrent
perseveration”, on which inhibitory control was required, also correlated significantly
with repetitive movements in her sample of children with autism. This relationship
between inhibitory control and repetitive movements makes intuitive sense; it is easy to
imagine how inhibitory impairment could lead to difficulties “stopping” a particular
movement sequence26.
However, Turner’s (1997) study demonstrated several other significant
correlations between EF and RBI measures which were not replicated in this study.
These included a significant relationship between set-shifting performance on the
26 Of course, the usual caveat about correlation and causation applies here. Arguments against the opposite direction of causation (i.e., that the behavioural symptoms may cause the EF deficits) are presented in Section 2.2.3 of Chapter 2.
189
modified IDED task and repetitive use of language and circumscribed interests; and
significant associations between performance on generativity measures (including the
verbal generativity tasks used in this study) and sameness behaviours and circumscribed
interests. The lack of any significant partial correlations between verbal generativity
variables and behavioural measures in this study was also surprising given the apparent
centrality of that domain in the analyses addressing the universality and uniqueness
criteria. These discrepancies with Turner’s study are somewhat difficult to explain.
While Turner did not partial out age or ability variables from her correlations (choosing
instead to divide her sample into low- and high-functioning subgroups), the results of
raw correlations in this study were not consistent with Turner’s findings either – the
Uses of Objects correct responses variable actually correlated significantly with
circumscribed interests in the opposite direction than predicted, and there were no other
significant raw correlations consistent with Turner’s results.
This failure to replicate is reflective of a general paucity of significant
correlations between cognitive and behavioural measures in the ASD group in this
study27. Neither ToM nor EF variables could account for the full range and extent of
autistic symptomatology measured, or even one complete symptom domain. Some
behavioural categories, such as high-level repetitive behaviours and several sub-
categories of the RBI falling under that heading, did not correlate with any cognitive
task variables at all; and conversely, some cognitive variables on which deficits were
significant and relatively prevalent did not show any relationships with
symptomatology. What might explain this? One possibility is that the behavioural
measures used were not sufficiently accurate, sensitive or wide-ranging, but this would
not seem to be the most likely reason given the well-documented diagnostic validity of
the ADI and the wide range and depth of domains covered by the RBI. The fairly
heterogeneous nature of the sample (i.e., the inclusion of participants meeting ADI-R
criteria in only one or two domains) is not a plausible explanation for these results
either, as variations in the range of behaviours displayed is more likely to increase,
rather than decrease, the probability of finding correlations.
One potentially influential factor is the behavioural therapy received by most
children with ASDs. It may be the case that relationships between underlying cognitive
deficits and behavioural expressions have been distorted because therapeutic
27 Pellicano, Maybery, Durkin, & Maley (submitted) also recently found a similar lack of significant correlations between ToM and EF measures and autistic symptomatology (as measured by the ADI-R) in a substantial sample of children with autism.
190
intervention shapes the nature and severity of the behaviours which would otherwise
occur if no intervention took place (without affecting cognitive functioning as strongly).
Parental discipline would have a similar effect, particularly in the case of repetitive
behaviours; indeed, during the administration of the RBI when questions were asked
regarding how long their child usually indulged in particular repetitive behaviours,
parents would often answer “Until I tell him/her to stop”.
This highlights the importance of considering the interaction between
environmental and genetically based cognitive influences on behavioural expression.
Just as there is no one-to-one mapping between genes and cognition (Karmiloff-Smith,
Scerif, & Thomas, 2002), it is also unlikely for direct or simple relationships to exist
between cognition and behaviour. It is probable that the relationship between cognitive
functions and behavioural outcomes is dynamic and changes continuously throughout
development. Hence, correlations between current cognitive status and current
behaviours may not reveal the cascade of processes which has shaped the nature and
severity of those behaviours, and they are likely to be weak and unreliable, resulting in
failures to replicate such as that which occurred with this and Turner’s (1997) study.
Correlations between cognitive and behavioural factors may also be weakened by the
use of parental report as the method of behavioural measurement, rather than direct
observation (this is discussed further in Chapter 7). Explanatory value would probably
be best measured using longitudinal designs, examining correlative and predictive
relationships between early cognitive deficits and both early and later behaviours, using
both observational and parental report methods of behavioural measurement.
Notwithstanding these concerns, the findings on explanatory value are consistent
with the results on the universality criterion in indicating the unlikelihood of a single
primary deficit model (or a model in which both deficits meet all criteria for primacy),
and could suggest that either i) different subgroups within the autism spectrum are
characterised by different cognitive and behavioural profiles, with this variability
obscuring and diluting clear relationships in the overall sample; or ii) a third cognitive
domain which was not measured is at least equally primary and can account for the
behaviours which did not correlate with any of the cognitive variables included in this
study. The “multidimensional spectrum” possibility, as described in hypothesis 6 (ii),
was not supported by these results, as this model (or at least the version described)
would predict strong correlations between particular cognitive deficits and the
behavioural domains they were purported to underlie, regardless of the variability of the
sample.
191
Concluding comments on the primacy of ToM and EF deficits. So far, then, it
appears that while EF deficits do not adequately or consistently meet all the criteria for
primacy in ASDs, they fare slightly better than the ToM hypothesis. In evaluating the
relative primacy of ToM and EF deficits, however, the comparative level of difficulty of
the ToM and EF tasks is perhaps one of the most important issues to be addressed, as it
is possible that these results simply reflect the fact that the ToM tasks were easier than
the EF tasks. The older, high-functioning nature of the overall sample was necessary to
achieve the aim of specifically assessing the full range of EF components, however the
consequence of this was that a larger than usual percentage of both ASD and control
participants displayed perfect performance on both first- and second-order false belief
tasks, thereby reducing the universality of ToM deficits. Although performance was not
quite at ceiling in the control group, the high level of performance overall may have
reduced the potential size of the group difference (this was also pointed out by Perner
and Lang (2000) in reference to the Ozonoff et al. (1991) study). This would have the
consequence of decreasing the ability of the ToM measures to discriminate the ASD
group from the control group, therefore affecting their performance on the uniqueness
criterion. The lack of significant correlations between ToM tasks and behavioural
measures could also have been a by-product of task difficulty, because it may have been
the case that ToM task passers still showed social and/or other behavioural impairments
- in other words, the behavioural measures may have been more sensitive than the ToM
measures, thereby reducing the strength of the relationship.
What may be said in defence of the validity of results derived from the ToM
measures in this study? Firstly, the inclusion of the second-order false belief task,
which has previously been failed by 10-18 year-old individuals with autism (Baron-
Cohen, 1989b) extended the range of difficulty in the ToM task domain. Secondly, it is
noteworthy that ToM and EF deficits were of roughly equal prevalence (using the ToM
scoring criterion which more clearly indicates low scorers), with a tendency for the
ToM deficit to be slightly more prevalent than most EF deficits in the current sample.
Given this, its lack of uniqueness and explanatory value cannot be readily discounted as
an artefact of the unequal difficulty of the tasks. Similarly, the significant proportion of
individuals who showed impaired performance on ToM tasks but unimpaired
performance on EF tasks (discussed below) indicates that for some individuals, the false
belief tasks were more difficult than the EF tasks (at least when evaluated with
reference to control group performance). Thirdly, performance on all of the false belief
tasks was far from the ceiling in the ASD group, allowing enough variability in the
192
sample for correlations with behavioural measures to emerge. The fact that false belief
performance showed medium level correlations with VIQ confirms that it was not at
ceiling and also suggests that it was assessed with some reliability. Finally, these
findings are consistent with previous research, with non-universality typical of all ToM
studies in autism (see Section 2.1.3 in Chapter 2), the discrepancy between the
uniqueness or discriminability of ToM and EF consistent with Ozonoff et al.’s (1991)
results, and the lack of explanatory value of ToM replicating studies by Turner (1997)
on repetitive behaviours and by Prior et al. (1990) and Sparrevohn and Howie (1995) on
social behaviours. For these reasons, the lack of universality, uniqueness and
explanatory value of the ToM deficit in this sample cannot be convincingly rejected as
an uninteresting consequence of the level of difficulty of the false belief tasks.
One additional alternative interpretation of the results indicating superior
primacy of EF deficits on the uniqueness criterion also requires acknowledgement.
When commenting on Ozonoff et al.’s (1991) results, Perner (1998; Perner & Lang,
2000) argued that the finding that an EF deficit discriminates better between ASD and
control groups than a ToM deficit does not necessarily indicate that the EF deficit is
more primary. He argues that a partial impairment in ToM (which he equates with
metarepresentational capacity) should actually result in a more severe impairment in EF,
as the SAS (Supervisory Attentional System) depends on metarepresentational capacity
and so any metarepresentational impairment will be magnified during EF task
performance. However, three findings are inconsistent with this explanation: i) the
roughly equal effect sizes of ToM and EF deficits in the ASD group, which suggest that
the deficits are equally severe; ii) the lack of explanatory value of the ToM tasks
(relative severity of impairment is irrelevant to that index of primacy, and if ToM was
primary it should show relationships with symptoms of autism); and iii) the presence of
a significant proportion of cases showing impaired ToM but intact EF (discussed
below), which this account does not allow for. Therefore, the evidence suggesting
better discriminative ability of EF deficits in this ASD sample appears to be a valid
indicator of superior primacy.
4.4.3 Independence of ToM and EF deficits
Given that EF deficits appear to be more primary than a ToM deficit in ASDs, is it
possible, then, that they can explain or subsume the ToM deficit which also
characterises these individuals? The relative absence of significant correlations and the
193
frequency of dissociations between ToM and EF performances in the ASD group
suggest that this is not in fact the case, and instead indicate fairly persuasively that the
two deficits are largely independent in individuals with ASDs. The fact that the
dissociations occurred in both directions importantly demonstrated that mastery of one
domain did not appear to be a prerequisite for the other. Instead, they suggest that the
two deficits co-occur in ASDs probably because of their proximal neuroanatomical
substrates. These results stand in contrast to the handful of studies which have found
significant correlations between ToM and EF in autism (Colvert et al., 2002; Ozonoff et
al., 1991; Zelazo et al., 2002), but consolidate previous reports of ToM-EF dissociations
in autistic individuals (Baron-Cohen & Robertson, 1995; Baron-Cohen et al., 1999b;
Ozonoff et al., 1991). As described in Section 2.3.2 of Chapter 2, the three studies
which have found an association between ToM and EF in autism have either failed to
partial out the effects of age and/or IQ or used only one type of EF task28. The
importance of partialling out the effects of age and ability was verified in this study, as
almost all of the several significant raw correlations in the ASD group between false
belief tasks and measures of planning and verbal generativity became non-significant
when these factors were accounted for. The current results are therefore likely to
represent a more reliable indication of the nature of the specific relationship between
ToM and EF in ASDs.
A number of alternative interpretations of these results are conceivable,
however. In their meta-analysis of the studies on the ToM-EF relationship conducted
up to that point, Perner and Lang (1999) found significant non-homogeneity among the
size of the correlations and proposed that the length of the testing session may be an
important confounding factor. They found a significant negative correlation between
the estimated testing duration per session and the size of the ToM-EF correlation
reported, leading them to suggest that longer testing sessions could result in fatigue
which would affect performance and decrease the strength of the correlation. It is
possible, then, that the relatively long testing sessions in this study (approximately 2.5
hours in total, including all tests in the WAFSASD battery; this was usually divided into
two sessions) may have influenced the strength of the correlations. Similarly, the fact
that the order of test administration was the same for all participants could potentially
28 In the one study (Colvert et al., 2003) which did partial out age and ability variables, only one EF task was included, the DCCS task. As this task is multifactorial and may be failed for a number of different reasons (see Perner & Lang, 2002), no equivalent task was included in the current battery. It is interesting to note that when ToM-EF correlations have been found in autism, the EF measures have been impure and/or consisted of composite scores.
194
have introduced extra fatigue-related variance to performance on the tasks completed
towards the end of each session. The questionable reliability of both ToM and EF tasks
(as discussed in Chapter 2) could also leave the correlations vulnerable to extraneous
variance. However, while these factors may have introduced a degree of extra variance
to the data, it is unlikely that they could have differentially affected the ASD and control
groups in such a way as to fully account for the striking difference in the number of
significant correlations observed in the two groups. Also, it is not the case that the tasks
at the end of the battery showed the weakest correlations, in either the ASD or control
group (e.g., the Opposite Worlds test, which was the last test to be administered,
showed strong correlations with repetitive movements in the ASD group; and the
generativity tasks, which were also administered towards the end of the battery,
demonstrated significant correlations with false belief variables in the control group).
In Section 2.3.1 of Chapter 2, it was argued that the close relationship between
ToM and EF which has been consistently demonstrated in typically developing 3 – 5
year-olds may not necessarily hold for older age groups. It is therefore also possible
that the lack of significant correlations in the ASD group may be a consequence of their
age, and that a relationship would be observed in a younger sample. However, the
presence of a range of significant partial correlations between ToM and EF measures in
the age-matched control group in this study suggests that age was not the most
important factor causing the outcome in the ASD group. Nevertheless, the pattern of
correlations demonstrated in the control group revealed a number of differences to those
typically observed in younger children, suggesting that the nature of the ToM-EF
relationship may change throughout development. Firstly, the significant correlations
occurred mostly (although not always) with the second-order false belief task, which
could reflect both the larger proportion of non-perfect scorers on this task as well as the
higher EF demands of the task (consistent with Tager-Flusberg & Sullivan, 1994b).
Second and more importantly, some EF components such as inhibition and working
memory did not show the usual relationship with false belief performance (the RIL task
Load error difference score, which reflects performance on a task combining inhibitory
and working memory requirements, did correlate with the second-order false belief task,
but other indices of that task such as the shape error score did not correlate with false
belief scores); whereas other variables which have not commonly been associated with
ToM performance, such as planning and generativity, did show significant correlations
with variables from both the first- and second-order tasks.
195
In order to further explore changes in the ToM-EF relationship with age, the
control group was divided into two age subgroups (5-8 year-olds and 9-18 year-olds)
and raw and relevant partial correlations were conducted separately for the two
subgroups (these are presented in Appendix C). Although the sample size for some of
the correlations was quite small in the younger age subgroup because some tasks were
administered only to participants over the age of 6, the results showed clearly that there
were a larger number of significant ToM-EF correlations in the younger subgroup, and
that the pattern of correlations was different to that observed in the older subgroup29.
While the larger number of significant correlations in the younger subgroup may simply
be due to the increased failure rate on false belief tasks in that subgroup (although note
that several controls over the age of 8 did not demonstrate perfect performance), the fact
that correlations with different EF variables were revealed in the older subgroup
suggests that there are also qualitative differences in the ToM-EF relationship at
different developmental stages. These findings are consistent with the few studies
which have been conducted previously on the ToM-EF relationship in children over the
age of 5, which have also demonstrated a smaller number and different pattern of
correlations compared to studies of younger children (Charman et al., 2002; Perner et
al., 2002a). The mechanisms underlying these developmental changes remain open to
speculation. One possibility is that a functional dependence between ToM and certain
aspects of EF such as inhibition exists as both of these abilities are developing (as
outlined in the emergence accounts), but once a certain level of development takes place
the ToM-EF relationship revolves more around performance-based factors (as proposed
by the expression accounts). However, the dissociability of impairments in the ASD
group is inconsistent with both emergence accounts (this is discussed further later).
Another possibility is that performance-based (or expression) factors explain the
relationship at all ages, but the EF components which influence ToM performance
change with age, as it may be that different skills are required for children of different
ages to solve ToM tasks (e.g., inhibitory control may be more important at a young age
as one’s own perspectives may be more salient)30. Although the typical development of
29 Separate correlations for the same two age subgroups were also conducted within the ASD group. Both subgroups showed no significant ToM-EF correlations after age and IQ variables were partialled out, with the exception of the correlation between the Stamps task restriction score and the simple false belief task, which was only significant in the older subgroup. This suggests that ToM and EF deficits were independent in all age ranges included in this sample of ASD participants. 30 The “common conceptual bases” account tested in this study, involving relational complexity, did not receive support in the current results – individuals with ASDs showed a ToM impairment but no relational reasoning deficit, and there were no significant correlations between the Relational Complexity task and false belief variables in either group. Nonetheless, this type of account would be unlikely to
196
ToM and EF is not the major focus of this thesis, these ideas and results certainly merit
further exploration in future studies using a wider range of ToM tasks and including
participants with a wider range of ages.
Returning to the ToM-EF relationship in ASDs, given that the lack of
correlations between ToM and EF found in the ASD group as compared with the control
group cannot be easily dismissed as a result of the length of the testing session or age of
the sample (as the two groups were matched on these factors), the question remains as
to why ToM and EF should be correlated in typically developing children but largely
uncorrelated in children with ASDs, where deficits in both co-exist. Something akin to
the ToMM-SP model proposed by Leslie and colleagues (Leslie & Thaiss, 1992; Leslie
& Roth, 1993) could potentially account for this pattern of results. In this model,
typically developing children may fail false belief tasks because of their processing
requirements (based on the ToM-EF correlations in the control group in this study, the
SP would include planning and generativity as well as inhibitory/working memory
processes), but children with autism fail because they lack a ToMM. Consistent with
the results from this study and in accordance with their predictions, Roth and Leslie
(1998) found that performances on a false belief task and a non-mentalistic control task
with similar processing requirements were significantly correlated in typically
developing children, but not autistic children. Similarly, the lack of correlations
between ToM and EF in the ASD group in this study could reflect the notion that EF
factors did not add any extra variance to their ToM performance – the false belief tasks
were failed because of ToM-specific factors and not because of poor EF. This
interpretation of the current results is consistent with the notion that ToM may be a
domain-specific capacity, although the ToMM-SP model in its original form cannot
adequately account for other results obtained in this study as it holds that a ToM deficit
is primary to autism. This interpretation would also suggest that the correlations
observed in the control group were caused by some individuals performing poorly on
the false belief tasks because of weaknesses in aspects of EF and not because of a ToM
impairment. This explanation would therefore favour an expression account of the
ToM-EF relationship in typical development.
However, while this explanation can account for the lack of ToM-EF
correlations in the ASD participants who failed ToM tasks, it does not explain the lack
of a relationship in ASD participants who showed EF impairments but intact ToM. The
explain developmental changes in the ToM-EF relationship as the common conceptual basis occurs because of common task structures, regardless of age.
197
expression account of the ToM-EF relationship would predict that those ASD
participants showing impairments in the EF components which were correlated with
ToM performance in the control group (e.g., planning, generativity) should sometimes
fail ToM tasks because of poor EF, thereby resulting in correlations between ToM and
EF performance. These correlations would only occur in the half of the ASD sample
showing impaired EF and may therefore have been too weak to emerge as significant.
Another possibility, though, is that those ASD participants who scored well on false
belief tasks did so via non-conventional routes to success, such as by using the
compensatory strategies described in Section 2.1.3 of Chapter 2. If so, then those EF
capacities which are normally required for successful ToM performance may not have
been needed, as the task-solving strategy may have been previously learned or taught
and therefore not dependent on online problem-solving skills. This speculation requires
empirical confirmation, however. One method of testing it would be to examine
correlations between EF measures and higher-level, more advanced and ecologically
valid measures of ToM, for which compensatory strategies may be less likely to have
developed.
It is not the case, however, that there was a complete absence of correlations
between ToM and EF in the ASD group. One non-verbal generativity variable (the
Stamps task restriction score) showed a significant correlation with simple false belief
task performance such that poorer generativity (higher restriction) was associated with
unstable performance on the simple false belief task. While generativity has not
previously been an EF component of particular focus in the literature on the ToM-EF
relationship, its potential role in the false belief performance of children with autism has
been highlighted previously by Peterson and Bowler (2000). Peterson and Riggs (1999)
had earlier argued that tests of false belief and subtractive reasoning (assessed by asking
a question such as “If the marble had not been moved, where would it be now?) require
similar counterfactual reasoning capabilities, in that they both involve processing a
negative counterfactual question of the form “If not-F, then Q” (where F is a known fact
and Q is a question). However, Peterson and Bowler’s (2000) results, which showed
that subtractive reasoning ability appeared to be necessary but not sufficient for accurate
false belief performance in children with autism, led them to suggest that the false belief
task required a crucial additional factor: that of generativity. They argued that in
subtractive reasoning tasks, children are given the supposition “not-F” as part of the
problem, but in false belief tasks it must be generated. A generativity impairment could
therefore explain why children with autism found false belief tasks more difficult than
198
subtractive reasoning tasks in their study. They proposed that both subtractive
reasoning and generativity were additional requirements for successful false belief
performance, beyond basic mentalistic understanding. Although no subtractive
reasoning tasks were included in this study, this kind of model fits quite well with the
current data, which suggested that ToM performance was largely independent of EF-
related factors in the ASD group, but that generativity played some role in false belief
performance. It is not clear, however, why the restriction score did not show significant
correlations with performance on the more difficult first- and second-order false belief
tasks (although these correlations were in the predicted direction). Nevertheless, this
result requires a slight modification to the two ToMM-SP-like and compensatory
strategy accounts proposed above, indicating that it may be the case that some
individuals with ASDs showed unstable performance on simple false belief tasks
because of a generativity impairment.
4.4.4 Towards a “multiple primary deficits” model of ToM and EF in ASDs
The next challenge is to attempt to unite this rather complex set of results on the profile,
primacy, and independence of ToM and EF deficits into a coherent theoretical
framework. In the introduction to this chapter, six hypotheses regarding the primacy
and independence of ToM and EF in ASDs and their implications for theories of autism
and models of the ToM-EF relationship were considered. The first hypothesis was that
there is only a single, primary deficit in ASDs, with no secondary impairments. As both
ToM and EF deficits were present in this sample of individuals with ASDs, this
hypothesis was not supported. Hypotheses 2 and 3 represented different scenarios in
which ToM and EF deficits were related in ASDs, either because one caused the other
or because both were caused by a third, more primary deficit. Neither of these
hypotheses were supported in this study, with the evidence suggesting that ToM and EF
deficits are largely independent in ASDs, with their co-occurrence most likely explained
by the neuroanatomical proximity. Hypotheses 4, 5, and 6 all proposed that ToM and
EF deficits were independent, but differed in terms of the primacy of those deficits.
Notwithstanding the other explanations for the results considered in previous sections of
this discussion, the non-universality and incomplete explanatory value of both ToM and
EF deficits in this study indicate that neither ToM nor EF deficits meet all the criteria
for primacy. This rules out hypothesis 4 (that either ToM or EF is the single primary
199
cognitive impairment of ASDs) and also excludes hypothesis 5 (that both ToM and EF
deficits are primary).
This leaves hypothesis 6: that ToM and EF impairments are independent in
ASDs, and neither meets all criteria for primacy. Of the six hypotheses, this found the
most support in the results from this study. Three different versions of this “multiple
deficits” model were presented in the introduction: i) different ToM and EF profiles are
found in different subgroups of individuals with autism, rather than both deficits co-
occurring in all individuals (in this model, explanatory value across all ASD individuals
may be low because the presence of different subgroups may obscure relationships in
the overall sample); ii) ToM and EF deficits underlie different aspects of
symptomatology, and therefore may be present in differing degrees of severity
according to the individual’s position on a multidimensional autism spectrum; or iii)
there may be an unidentified third deficit which is at least equally primary (and may
underlie symptoms which were unrelated to ToM and EF). A fourth version was also
proposed in Section 4.4.2 of this discussion: that iv) different stages of development of
individuals with ASDs may be associated with different primary cognitive deficits.
Each of these four possibilities will now be considered in turn.
i) Subgroups. Subgroups of individuals with ASDs can be classified or defined
in several different ways, such as by severity of symptoms, the domains in which
symptoms are present, or level of intellectual impairment (e.g., Beglinger & Smith,
2001; Prior et al., 1998). The only subgroups which were explicitly examined in this
study were the “full criteria” and “partial criteria” subgroups, defined according to the
number of domains in which a higher-than-threshold level of symptomatology was
present (as assessed by the ADI-R). Although these two subgroups showed different
patterns of performance on the Pattern Meanings task, which suggested that their verbal
generativity difficulties may be expressed in slightly different ways, there were no other
differences between the two subgroups on any other EF or ToM measures. This
indicates that the number of domains in which symptoms are present does not relate
systematically to the profile of ToM and EF deficits displayed. However, this
subgrouping method did not distinguish which symptom domains were present within
the “partial criteria” subgroup, which may have obscured more fine-grained differences.
While other subgroup divisions were not specifically analysed, the lack of
significant or strong correlations between ToM and EF variables and any measures of
symptom severity also suggests that subgroups based on overall symptom severity
(rather than presence of symptoms in particular domains) are also unlikely to be
200
associated with consistent profiles of ToM and EF deficits. It is a stronger possibility
that subgroups based on level of functioning (as measured by IQ) may be characterised
by different ToM and EF profiles, as several group differences on both ToM and EF
measures were mediated by VIQ. Previous research has also suggested that level of
functioning (which has been measured by adaptive skills as well as IQ) has shown the
most promise in discriminating subgroups and predicting outcome (see Beglinger &
Smith, 2001; Fein et al., 1999; Stevens et al., 2000). When comparisons were
conducted between “low VIQ” and “high VIQ” subgroups within the ASD group31, it
was found that the low VIQ subgroup performed significantly more poorly on ToM
measures (including both aggregate scores), and on one EF measure (the ToL).
However, there were no other EF task differences such that the high VIQ subgroup
performed more poorly than the low VIQ subgroup, which suggests that this subgroup
division did not map directly onto “ToM-impaired, EF-intact” and “EF-impaired, ToM-
intact” subgroups, instead indicating that the low VIQ group was more impaired in ToM
and equally impaired in most domains of EF relative to the high VIQ group.
However, this assumes that there are only two subgroups based on ToM-EF
performance and VIQ. This is unlikely, as there is at least a third subgroup showing
both ToM and EF deficits (as the incidence of ToM-EF dissociations was not 100%).
Furthermore, there may be more subgroup divisions which vary according to the
specific EF profile displayed. A better way of determining how many subgroups based
on ToM and EF performance there are and how they relate to other measures such as IQ
or symptomatology would be to employ cluster analysis, where the characteristics of
ToM-EF clusters could be examined to determine how the subgroups should be defined
behaviourally. This would require a large sample which varied considerably on IQ and
symptomatology. Conclusions about the validity of the subgroup notion therefore await
further investigations, although subgroups based on symptom domains or symptom
severity were not strongly supported by the current data.
ii) An autism spectrum with multiple dimensions. Rather than proposing several
discrete subgroups, this model conceives of autistic symptomatology as varying on a
more continuous spectrum. In a single primary deficit model, this spectrum could be
unidimensional, but in a multiple deficits model, there would need be more than one
dimension, with ToM and EF deficits each underlying a different dimension. In the
version of this model presented in the introduction, these dimensions corresponded to
31 A PIQ-based division did not reveal any significant subgroup differences in ToM or EF performance.
201
symptom domains, such that, for example, a ToM deficit was the basis for social
impairment and EF deficits accounted for repetitive behaviours. Thus, each ASD
individual’s profile of ToM and EF deficits would determine the nature and severity of
their symptomatology (so, a more severe ToM deficit would be associated with more
severe social impairment). In this model, the apparent presence of subgroups of “ToM-
impaired, EF-intact” and “EF-impaired, ToM-intact” individuals would be an artefact of
the arbitrary cutoff for “impairment” within a continuous distribution of scores. There
were no bimodal distributions of continuous variables in this study (although ToM
performance was highly skewed), suggesting the dimensional variation notion is
appropriate at least for EF performances. However, the particular version of the model
alluded to above was not supported in this study, with ToM performance showing no
significant correlations with behavioural measures and various EF deficits correlating
significantly with both social/communicative functioning and repetitive behaviours. As
discussed earlier, environmental and developmental factors may have contributed to
these results; however, as it stands, the data are not consistent with this model.
Alternative versions of this model are nevertheless possible. For example, rather
than the dimensions corresponding to symptom domains, there may be one dimension
for number and severity of symptoms and one for level of functioning (as proposed by
Szatmari et al., 2002). Perhaps EF deficits could then be associated with the former
dimension (as they showed greater explanatory value) and a ToM deficit could be
associated with level of functioning (as ToM performance covaried more strongly with
VIQ). The weak and incomplete explanatory value of EF deficits is inconsistent with
this possibility, but the notion of a multidimensional spectrum appears more suited to
the distributions of scores on cognitive tasks (particularly EF tasks) and warrants further
investigation.
iii) A third deficit. As neither ToM nor EF deficits were able to account for the
full range of symptoms displayed by this sample of individuals with ASDs, it is possible
that there is a third (or more) cognitive deficit(s) which could explain those symptoms.
As stated in the introduction, this possibility is compatible with both of the accounts just
described, rather than competing with them. The relative primacy and the relationship
of this third deficit to ToM and EF would be open for investigation. This study does not
allow any inferences about what this deficit might be, but based on current research the
most obvious candidate would be weak central coherence, which has been found to
characterise individuals with autism in a number of studies (Happé, 1994b, 1996, 1997;
Jolliffe & Baron-Cohen, 2000; Shah & Frith, 1983, 1993). Happé (2000) has argued
202
that weak central coherence is independent from ToM and can explain aspects of autism
which ToM cannot, although another study found that ToM and central coherence were
related (Jarrold, Butler, Cottington, & Jimenez, 2000). The universality, uniqueness,
causal precedence, and in particular the explanatory value of weak central coherence
and its relationship with ToM and EF will be interesting topics for further research.
iv) Different developmental stages. This fourth variant of a multiple primary
deficits model holds that the primacy of ToM and EF deficits in autism may not remain
consistent throughout different stages of development. This would mean, of course, that
the criterion of “causal precedence” would not necessarily be an appropriate index of
primacy. Previous research has generally not found strong EF deficits in young children
with autism, while ToM deficits (or least impairments in the proposed precursors to
ToM) and social abnormalities have been more consistently documented (see Sections
2.1.3 and 2.2.3 in Chapter 2). It could be hypothesised that an impairment in ToM
holds more explanatory value in the early stages of autism, but that deficits in EF
somehow become more primary with age. Furthermore, deficits in the various
components of EF could also change in primacy as they develop; for example,
inhibition impairments could be more important early on (as inhibitory control is
typically one of the first EF components to develop), with planning and generativity
impairments (which typically reach their capacity during adolescence) becoming more
central to autism later in development32. It may be the case that the age at which a
particular capacity usually develops is the age at which its abnormal development has
the most impact on behaviour. The relatively old mean age of the sample could
therefore explain why the correlations between ToM and behavioural measures were not
significant, and the variability in the age of the sample could account for the relatively
small number and size of significant correlations with EF components, as well as the
non-universality of both ToM and EF deficits. This is obviously speculative and relies
on the findings of previous studies given that the early stages of the development of
autism were not studied in this research. This hypothesis would be best assessed using
longitudinal studies of the development of ToM and EF and their relationship with
behaviour throughout development. Notably, its plausibility is supported by previous
findings that in children with Williams syndrome (which is also a genetically based
32 When the sample was divided into younger (5-8 year-old) and older (9-18 year-old) subgroups, the results of group comparisons were consistent with this hypothesis: group differences in verbal inhibition were significant for the younger subgroup but not the older subgroup, and planning and generativity impairments were significant for the older subgroup but only marginally significant for the younger subgroup (these analyses are presented in Appendix D).
203
developmental disorder), a change in the nature of cognitive deficits is observed at
different developmental stages (Paterson, Brown, Gsodl, Johnson, & Karmiloff-Smith,
1999).
If this developmental account is accepted, is it then possible that ToM and EF
are related processes in younger children with autism (i.e., below the age of five years)?
In Section 2.3.1 of Chapter 2, it was argued that the existence of dissociable
impairments in two abilities at a certain age cannot be used to infer the independence of
the two abilities throughout earlier development. As suggested earlier in regard to
typically developing children, it may be the case that aspects of EF depend on ToM for
their development, as argued by Perner (or vice versa as suggested by Russell, although
this is less likely because of the lack of EF deficits found in young children with autism
as well as EF’s later developmental trajectory), but that the two domains become
independent after the crucial stage of development has passed. However, if this was the
case, it would be unlikely at later ages for deficits in ToM to exist without deficits in EF
(this occurred in a significant proportion of this sample for all EF components). While
double dissociations could occur if impairment in one domain was acquired after the
initial stage of development of the other domain, “ToM-impaired, EF-intact”
dissociations could not occur if ToM was impaired from an early age, as this would
result in abnormal development of EF (and vice versa if early EF impairments caused a
ToM deficit). This suggests that EF deficits in ASDs are not a consequence of an early
ToM deficit. Moreover, the presence of double dissociations in this sample provides
evidence against both emergence accounts of the ToM-EF relationship (as well as the
“common conceptual basis” accounts). The only situation in which an emergence
account may be plausible would be if EF deficits caused ToM to develop abnormally,
but this ToM deficit was not apparent at later ages because the use of compensatory
strategies “masked” the impairment. Nevertheless, it appears that ToM and EF deficits
in ASDs are best explained as occurring independently, most likely linked by their
neurobiological substrates, but possibly varying in primacy according to the age at
which they usually have the most impact on behaviour.
In sum, then, results from this study suggest that deficits in ToM and certain aspects of
EF characterise individuals with ASDs; neither of these deficits meet criteria for a
single primary deficit, but EF deficits (in particular, deficits in verbal inhibition and
generativity) are relatively more primary; and the deficits appear to be independent.
This pattern of results suggests that a “multiple primary cognitive deficits” account best
204
explains ASDs, but it remains to be seen which version of this model is most
appropriate (or, perhaps, which combination of these models – this is discussed further
in the General Discussion in Chapter 7). Distinguishing between these models relies
fairly heavily on determining the relationship of each cognitive deficit with behavioural
symptom domains; however, the difficulties with measuring the explanatory value of
cognitive deficits in a precise and thorough manner (due to the indirectness and
complexity of cognitive-behavioural relationships, as discussed in Section 4.4.2)
prevented strong conclusions from being made on this basis. Similarly, while the non-
universality of both ToM and EF deficits indicated that neither of them was singularly
primary, the inferior primacy of ToM based on its lack of explanatory value remains
debatable (although its non-significant ability to discriminate ASD from control
individuals supported this inference). Another method of confirming the primacy of
ToM and EF deficits and testing various multiple deficits models is to examine the
prevalence of these deficits and their independent occurrence or co-occurrence in
relatives of individuals with ASDs – thereby determining their potential as independent
subclinical markers of the ASD genotype. That is the focus of Chapters 5 and 6.
205
206
CHAPTER 5
Literature Review: The Broad Autism Phenotype
5.1 Autism as a genetic disorder
5.2 The broad phenotype
5.2.1 The behavioural phenotype
5.2.2 The cognitive phenotype
5.2.2.1 General intellectual ability
5.2.2.2 Specific cognitive deficits
207
As mentioned at the end of Chapter 4, the role of ToM and EF deficits as subclinical
markers of the ASD genotype is another method of determining their primacy to ASDs.
If a cognitive deficit is primary to autism, then its prevalence in individuals who carry
the autism genotype (or at least the genotype for the relevant autistic trait), including
those with a milder or lesser variant who do not meet criteria for an ASD diagnosis,
should be higher than in the normal population (Bailey et al., 1996). An elevated
incidence of a particular cognitive weakness in first-degree relatives of individuals with
autism therefore provides evidence of the centrality of that deficit to autism. The
independent incidence of those weaknesses in certain subgroups of families, and their
relationship with behavioural traits, would also have implications for the various
multiple deficits models presented in the previous chapter (this is discussed further in
the introduction to Study Two in Chapter 6).
This chapter contains a literature review of the genetics of autism and the broad
autism phenotype, as a background for the rationale and methodology developed in
Study Two. As the arguments outlined above depend upon the assumption that autism
is a genetic disorder, the first section of the review presents evidence for that
assumption. The second section of the review describes research on the behavioural and
cognitive characteristics of the broad autism phenotype, including previous studies of
ToM and EF deficits in relatives of individuals with ASDs. Throughout the review, it
will hopefully become evident why it is important to study ToM and EF in the broad
phenotype and what needs to be addressed in further studies.
5.1 Autism as a genetic disorder
Because autism did not appear to run in families (i.e., it was rare for children with
autism to have parents with autism), for several years a genetic basis to the disorder was
rejected in favour of environmental causes such as cold, detached child-rearing practices
or “refrigerator” parenting (Bettelheim, 1967; Eisenberg & Kanner, 1956). Early
reports by Kanner and Asperger themselves of social and communicative difficulties
and obsessional characteristics in parents of children with autism and Asperger
syndrome were commonly interpreted as causing abnormal development in their
children rather than reflecting a genetically based milder phenotype. However, these
notions came under increasing doubt as it was realised that it would be rare for autistic
individuals to develop close relationships and therefore have children, and as studies
failed to find evidence of abnormal parenting styles (see Cantwell, Baker, & Rutter,
208
1979). Recognition of associations with mental retardation (Lockyer & Rutter, 1969)
and epilepsy (Rutter, 1970) provided further evidence of a biological basis. A key study
by Folstein and Rutter (1977) helped establish autism as a genetic disorder, finding a
significant difference in the concordance rates for autism in monozygotic (MZ; 36%)
versus dizygotic (DZ; 0%) same-sex twins. Furthermore, they found that the majority
of MZ twins who did not have autism showed some type of cognitive deficit, usually
involving language. These findings have since been replicated in two large-scale
studies (Bailey, Le Couteur, Gottesman, & Bolton, 1995; Steffenburg et al., 1989),
although Bailey et al. (1995) found higher MZ concordance rates of 60% for autism and
92% for a broader spectrum of cognitive or social abnormalities (the DZ concordance
for this broad spectrum was also higher at 10%). Based on their results, Bailey et al.
(1995) estimated that the heritability for autism is greater than 90%.
An elevated recurrence risk for autism has also been observed in siblings,
ranging from 2% (Boutin et al., 1997; Minton, Campbell, Green, Jennings, & Samit,
1982) to 6% (Baird & August, 1985) and averaging at around 2.2% across studies
(Szatmari, Jones, Zwaigenbaum, & MacLean, 1998), compared with a population base
rate of around 0.1% (Fombonne, 2003). An increased rate of ASDs more broadly in
twins and other relatives of individuals with autism has also been reported (Bailey et al.,
1995; Bolton et al., 1994; Le Couteur et al., 1996), indicating that the genetic liability is
not restricted to a narrowly defined disorder. Family members with ASDs do not
always covary in diagnostic subtype or symptom severity, with MacLean et al. (1999)
finding no familial aggregation of ASD subtype (i.e., autism, Asperger syndrome, or
PDDNOS), and Le Couteur et al. (1996) finding as much variation in symptom severity
and intellectual ability within MZ twin pairs as between pairs. These findings, along
with the rapid decrease in risk rates from MZ twins to DZ twins and siblings to more
distant relatives (the latter being very low; e.g., Delong & Dwyer, 1988), indicate that
the genetic mechanisms of autism are not simple or Mendelian in nature and are likely
to involve epistatic effects involving interactions among several genes (Pickles et al.,
1995; Rutter, 2000; Szatmari, 1999). Studies involving linkage analysis more directly
indicate the presence of multiple susceptibility loci for autism (e.g., Risch et al., 1999;
Yonan et al., 2003). The existence of MZ twins discordant for autism and high
phenotypic variability within twin pairs also suggests that environmental or other
factors play a role, although it remains unclear what these factors may be. Bailey,
Palferman, Heavey, and Le Couteur (1998) favour genetic instability (e.g., caused by a
209
somatic mutation), gene-environment interactions, and/or stochastic factors as
explanations for variability in phenotypic expression.
5.2 The broad phenotype
Numerous studies have now demonstrated that milder forms or lesser variants of autistic
symptomatology, which do not meet criteria for a diagnosis of autism, are frequently
exhibited in relatives of individuals with autism (Bolton et al., 1994; Landa et al., 1992;
Le Couteur et al., 1996; Pickles et al., 2000; Piven et al., 1990b, 1991, 1994; Piven,
Palmer, Jacobi, Childress, & Arndt, 1997b), giving rise to the notion of a spectrum of
autistic traits or “broad phenotype” of autism. As described earlier, studying the
characteristics of the broad phenotype in non-autistic relatives is a useful method of
identifying which traits are primary to autism. Exploration of the broad phenotype has
also been helpful in identifying possible genetic mechanisms (e.g., whether traits are
contributed by both parents) and in increasing the power to identify genes linked with
autism. If the broad phenotype is considered to be a collection of individual traits, each
of which could be related to one of the several genes which combine to cause autism,
then using the broader phenotype in linkage analysis can boost the power to find genes
involved in autism by increasing the number of “affected” individuals available for
analysis (Folstein, Bisson, Santangelo & Piven, 1998; Piven, 1999).
5.2.1 The behavioural phenotype
Most studies attempting to identify or define the broad autism phenotype have focussed
on documenting behavioural signs, generally either by conducting family history
interviews about the presence of social and communicative difficulties and repetitive
behaviours in family members, or by conducting more direct interviews and
assessments of personality characteristics and psychiatric disorders. As already noted,
family history studies (which have generally used the Family History Interview, a semi-
structured interview specifically designed to assess the broad autism phenotype) have
consistently found social abnormalities, communicative difficulties, and repetitive
stereotyped behaviours in a substantial minority of relatives of individuals with autism
(Bailey et al., 1995; Bolton et al., 1994; Le Couteur et al., 1996).
The use of personality assessment tools such as the Personality Assessment
Schedule (PAS; Tyrer, 1988) has revealed that parents of children with autism rate
210
significantly higher than controls on several personality characteristics relating to social
interaction such as aloof, untactful, shy, schizoid, oversensitive to criticism and lacking
in empathy (Murphy et al., 2000; Narayan, Moyes, & Wolff, 1990; Piven et al., 1991,
1994, 1997c; Wolff, Narayan, & Moyes, 1988). Using Baron-Cohen, Wheelwright,
Skinner, Martin, and Clubley’s (2001b) Autism Spectrum Quotient (a self-report
questionnaire designed to assess features of the broad autism phenotype), Bishop et al.
(in press-a) recently found elevated ratings on the “social skills” and “communication”
subscales in parents of children with ASDs1. Abnormal pragmatic communication
styles have also been detected in some parents using both interviews and direct
assessments of narrative discourse (Landa et al., 1992; Wolff et al., 1988), although
structural language skills are usually found to be intact (Bishop et al., in press-b;
Pilowsky, Yirmiya, Shalev, & Gross-Tsur, 2003). A history of language delay appears
to be a more equivocal finding, with most studies reporting language delay in only a
small proportion of relatives (for a review, see Bailey et al., 1998). Obsessional traits
and repetitive behaviours have been found to be relatively less common than social and
communicative impairments in relatives of autistic individuals (Bailey et al., 1995;
Bolton et al., 1994), although Piven et al. (1997c) found fairly high rates (almost 50%)
of the personality trait “rigid” in the parents of multiplex families in their study. As
pointed out by Bailey et al. (1998), the infrequency of behaviours in this category in the
broad phenotype may be a consequence of the insensitivity or inappropriateness of the
measures used rather than reflecting the secondary or unimportant nature of those
symptoms in autism. In support of the importance of repetitive behaviours, Silverman
et al. (2002) found that the severity of repetitive behaviours showed a high level of
familiality in multiplex families, whereas there was little evidence for familiality in
social or verbal communication domains.
The risk of psychiatric disorders other than autism in relatives of autistic
individuals has also been found to be elevated. In particular, increased rates of major
depression have been documented in parents (Bolton, Pickles, Murphy, & Rutter, 1998;
Piven et al., 1990b, 1991; Piven & Palmer, 1999; Smalley, McCracken, & Tanguay,
1995). In all of these studies, the majority of depressive episodes have been found to
occur prior to the birth of the child with autism, suggesting that they cannot be
explained by the burden of caring for a disabled child. The incidence of anxiety
disorders may also be increased in relatives of autistic probands, but these findings have
1 These parents were part of the WAFSASD, and therefore were the parents of the probands and siblings in the current research.
211
been less consistent. Piven et al. (1991) found an elevated rate of anxiety disorder in
parents, and increased rates of social phobia have also been reported (Piven & Palmer,
1999; Smalley et al., 1995), but other studies have not found evidence of phobic
disorders (Bolton et al., 1998; Piven et al., 1991) or anxiety disorders in general (Bolton
et al., 1998). However, two studies have found higher rates of obsessive-compulsive
disorder in relatives of autistic probands (Bolton et al., 1998; Hollander, King, Delaney,
Smith, & Silverman, 2003) with the recent study by Hollander et al. showing that the
occurrence of obsessive-compulsive traits or disorder in parents of multiplex families
was significantly more likely if the autistic children showed high levels of repetitive
behaviours. There is no consistent evidence for higher rates of other psychiatric
disorders such as schizophrenia or substance abuse (Bolton et al., 1998; Piven et al.,
1991; Smalley et al., 1995).
5.2.2 The cognitive phenotype
While studies of the behavioural features of the broad autism phenotype have been
informative, individual behavioural signs suffer from the problem of low diagnostic
specificity (Bailey et al., 1998), and are therefore of limited utility as unique indicators
of the broad phenotype. In addition, because behavioural phenotypes are multiply
determined and have indirect and complex relationships with underlying genotypes (i.e.,
the same genotype can give rise to different phenotypes, and the same phenotype can
arise from a range of genotypes; Gottesman & Gould, 2003; Karmiloff-Smith et al.,
2002), they are not an ideal basis for identifying genetic mechanisms. Researchers have
therefore concurrently searched for a more basic subclinical marker of the autism
genotype – or “endophenotype” - at the level of cognition. An endophenotype may be
described as an “intermediate phenotype” or “vulnerability marker” which is unseen by
the unaided eye (e.g., a neurophysiological, biochemical, endocrinological, or cognitive
feature – i.e., not at the level of behaviour) and is somewhere between the disorder’s
phenotype and the distal genotype (Gottesman & Gould, 2003). Endophenotypes are
believed to represent a genetic liability to the disorder in unaffected individuals, and
may only be indirectly related to classic symptoms of the disorder (Leboyer et al., 1998;
Skuse, 2001). The identification of endophenotypes for complex genetic disorders may
help address questions about aetiology and establish markers for diagnosis and
classification (Gottesman & Gould, 2003).
212
The presence of a cognitive endophenotype is suggested when unaffected
relatives of individuals with autism show a raised incidence of a cognitive deficit (or
strength) that is associated with autism, but to a milder degree than in individuals with
autism themselves (Hill & Frith, 2003). Studies of the cognitive phenotype have tended
to focus either on the IQ profiles of relatives of autistic individuals or have investigated
the presence of specific deficits in ToM, EF, and central coherence.
5.2.2.1 General intellectual ability
Because approximately 70% of individuals with autism are mentally retarded
(Fombonne, 2003) and autistic individuals in general tend to have better Performance
than Verbal IQ (e.g., Lockyer & Rutter, 1970), several studies have examined the
possibility that the broad phenotype may be similarly characterised by an increased
incidence of mental retardation and/or a Verbal-Performance IQ discrepancy. Several
small early studies involving very low-functioning autistic probands found a higher rate
of mental retardation in their relatives than in the general population (August, Stewart,
& Tsai, 1981; Baird & August, 1985; Minton et al., 1982). However, larger and more
recent studies have not replicated this result, finding that mental retardation occurs only
in association with autism and not in isolation, or at least at no greater incidence than
for the general population (Bailey et al., 1995; Folstein et al., 1999; Fombonne, Bolton,
Prior, Jordan, & Rutter, 1997; Freeman et al., 1989; Piven et al., 1990b; Smalley &
Asarnow, 1990; Szatmari et al., 1993). This suggests that the genetic liability for autism
is not usually for mental retardation alone (Bailey et al., 1998). The discrepancy
between earlier and later studies may be due to the severe retardation of the probands in
earlier studies. Consistent with this possibility, August et al. (1981), Baird and August
(1985) and Boutin et al. (1997) all reported higher rates of cognitive disabilities
(including language delay, learning disabilities, and mental retardation) in relatives of
low-functioning probands (but see Piven et al., 1990b, and Szatmari et al., 1993, both of
whom found no association between the proband’s IQ and the cognitive and academic
functioning of their relatives; Starr et al., 2001 also found comparable familial loading
for the broad phenotype in low and high IQ autism families).
Although there does not appear to be an increased incidence of mental
retardation, a number of studies have found significantly lower Verbal or Performance
IQs than controls and/or significant discrepancies between verbal and non-verbal ability
in first-degree relatives of individuals with autism. Consistent with the pattern typically
213
observed in autistic individuals, Minton et al. (1982) found that siblings of autistic
children had significantly lower VIQ than PIQ on the WISC-R and WAIS. Similarly,
Leboyer, Plumet, Goldblum, Perez-Diaz, and Marchaland (1995) found that siblings of
autistic females showed significantly lower verbal abilities than siblings of Down
syndrome controls, but there was no difference in visuospatial abilities across the two
siblings groups. The lower verbal abilities in the siblings of autistic probands appeared
to be due to a proportion of brothers who showed particularly discrepant verbal and
visuospatial abilities. However, other studies examining the IQ profile of relatives of
autistic probands have either reported no IQ differences from controls at all (Freeman et
al., 1989; Ozonoff, Rogers, Farnham, & Pennington, 1993; Szatmari et al., 1993) or
have found exactly the opposite pattern of discrepancy. Three large-scale studies have
found small but significant VIQ-PIQ discrepancies in parents of individuals with autism
whereby VIQ was significantly higher than PIQ (Folstein et al., 1999; Fombonne et al.,
1997; Piven & Palmer, 1997). Fombonne et al. (1997) found this pattern in both parents
and siblings of autistic probands irrespective of the test version used (WISC-R versus
WAIS) and after controlling for SES. Folstein et al. (1999) observed superior VIQ to
PIQ only in parents, finding no difference in siblings. While Fombonne et al. (1997)
and Piven and Palmer (1997) both used Down syndrome controls, in the former study
VIQ was significantly higher in the autism relatives and there was no difference
between the groups in PIQ, whereas in the latter study PIQ was significantly lower in
the Down syndrome relatives and there was no difference in VIQ.
There may be a number of reasons for these inconsistencies regarding the
presence and direction of VIQ-PIQ discrepancies in relatives of autistic individuals.
Firstly, siblings and parents of autistic probands do not appear to demonstrate the same
IQ profile, with most sibling studies finding superior PIQ to VIQ or no difference (with
the exception of Fombonne et al., 1997), while studies involving parents tend to find the
opposite pattern. It has been argued that parents are by definition “selected” for
parenthood in that their social and communicative functioning must be sufficient for
partnership and children, and they may therefore be less impaired than siblings in
domains such as VIQ (e.g., Piven & Palmer, 1997). Secondly, there is some evidence
that there may be at least two subgroups of parents (and possibly siblings) showing
different IQ profiles. In Folstein et al.’s (1999) study, parents with early language
delays demonstrated lower VIQ than parents without language delays and no VIQ-PIQ
discrepancy, leading the authors to suggest that there may be two or more patterns of IQ
performance in parents of autistic probands. Consistent with this, Freeman et al. (1989)
214
reported that approximately equal numbers of relatives showed VIQ-PIQ discrepancies
in both directions (although this could equally reflect random differences). While
subgroups of parents (and siblings) showing better VIQ than PIQ appear to contradict
the pattern typically found in individuals with autism, several studies have now shown
that children with high-functioning autism and Asperger syndrome often show higher
VIQ than PIQ (Goodman, 1989; Klin, Volkmar, Sparrow, Cicchetti, & Rourke, 1995;
Szatmari et al., 1990). The possibility that parents in general may be less impaired than
siblings is therefore consistent with the finding that parents more often show IQ
discrepancies in favour of VIQ (mirroring the pattern observed in higher-functioning
probands). There is also the possibility that that high-functioning autism is genetically
different to low-functioning autism (MacLean et al., 1999; Szatmari et al., 2002), and is
associated with different IQ profiles in relatives; direct correlations between proband
and relative IQ have generally not been reported, however. Finally, the role of other
methodological differences between studies such as the IQ subtests used, sampling
methods, unit of analysis (aggregation of familial data versus inclusion of individual
sibling scores), age and gender of the probands and/or relatives, and range of ASD
diagnoses included are yet to be clarified.
5.2.2.2 Specific cognitive deficits
Given the variability in studies of IQ profiles, researchers have increasingly turned their
attention to the investigation of specific cognitive deficits as potential endophenotypes
for autism. Studies of the specific cognitive phenotype have been driven by concurrent
research on primary cognitive deficits in autism, focussing on the three main current
cognitive theories: ToM, EF, and weak central coherence. This not only allows more
precise delineation of the broad cognitive phenotype, but also represents a method of
testing the primacy of deficits in those domains to autism.
To date, only three published studies have examined the mentalising abilities of
relatives of individuals with autism, with contrasting results. Ozonoff et al. (1993)
found no significant differences between siblings of autistic individuals and learning-
disabled controls on three ToM tasks. However, they acknowledged that according to
their power analyses, the ToM measures used were not sensitive enough to detect any
deficits in non-autistic siblings, and they suggested the use of higher-level tasks. Baron-
Cohen and Hammer (1997) employed Baron-Cohen et al.’s (1997) Eyes Task with
parents of children with Asperger syndrome. They found that both mothers and fathers
215
in the Asperger group showed subtle but significant impairment on the task compared
with control mothers and fathers. Using the same task, a recent study by Dorris, Espie,
Knott, & Salt (2004) replicated these findings in siblings of children with Asperger
syndrome, who displayed poorer performance on the task compared with control
siblings.
EF performance in relatives of autistic probands has been addressed in several
studies, most of which have focussed on measures of planning and set-shifting.
Significantly poorer performance by parents of individuals with autism compared with
control parents (including parents of children with learning disabilities and Down
syndrome) on Tower tasks (i.e., the Towers of Hanoi and London and the Stockings of
Cambridge test from the CANTAB battery) was found by Hughes, Leboyer and
Bouvard (1997) and Piven and Palmer (1997), and the same result in siblings was
obtained by Ozonoff et al. (1993) and Hughes, Plumet, and Leboyer (1999). In Hughes
et al.’s (1997) study, the difference in planning ability was restricted to fathers only, and
in both of the studies by Hughes et al. (1997, 1999) a planning deficit was restricted to
a subset of the relatives of autistic probands, with group differences only emerging
clearly when the proportions of participants showing a deficit were compared.
Findings of no group differences on the WCST in parents (Szatmari et al., 1993)
or siblings (Ozonoff et al., 1993) were initially suggestive of intact cognitive flexibility
in relatives of individuals with autism. However, two subsequent studies using the
IDED set-shifting task found that a subset of both parents and siblings of autistic
probands demonstrated difficulties with the extra-dimensional shift stage of the task
(Hughes et al., 1997, 1999). Hughes et al. (1999) postulated that this discrepancy
between the results observed using the WCST and the IDED task may be due to the fact
that the IDED task involves a total change of stimuli at each shift and so perseverative
responses are limited to high-level dimensional shifting difficulties rather than specific
exemplars. However, an argument that the WCST is lower-level than the IDED task is
inconsistent with findings on probands themselves, who generally show difficulties on
the WCST more often than the IDED task. It may be the case that subsets of the
samples tested with the WCST may have shown a deficit as in the Hughes et al. studies,
but this was not directly examined.
The two studies by Hughes et al. (1997, 1999) also incorporated working
memory measures, with different patterns of results for parents and siblings. Both
studies included a spatial working memory task involving a high demand for strategy
use and therefore purportedly the “central executive” (Baddeley, 1986), and a simple
216
spatial span task with low executive or strategic requirements which served as a control
task. Parents of autistic probands showed intact spatial spans, but fathers made a
significantly higher number of errors than normal control fathers on the more strategic
working memory task (Hughes et al., 1997). However, there was no difference between
the fathers of autistic probands and the fathers of learning disabled controls, indicating
that the deficit was not unique to autism families. By contrast, siblings of autistic
probands showed superior spatial spans to siblings of both developmentally delayed and
normal controls (as well as better verbal short-term memory for recently presented
items), but there were no group differences on the more strategic working memory task
(Hughes et al., 1999). Together, these results suggest that a working memory deficit is
not as reliable or unique a characteristic of the broad phenotype as problems with
planning and set-shifting.
Other components of EF such as inhibition and generativity have not been as
well studied in relatives of autistic probands. Hughes et al. (1999) included a verbal
generativity task (word fluency) in their battery with siblings, and found both a
significant group difference overall in the number of words generated and a higher
proportion of “low fluency” participants in the autism sibling group. This promising
result requires replication and extension to parent samples using both verbal and non-
verbal generativity tasks. Similarly, tests of both verbal and non-verbal inhibition are
yet to be employed with either siblings or parents.
The possibility that weak central coherence may characterise the broad autism
phenotype has also received some attention. No evidence of a relative strength in the
Block Design subtest from the Wechsler scales, which purportedly indicates weak
central coherence (Shah & Frith, 1993; Happé, 1994c), was found by Szatmari et al.
(1993) or Fombonne et al. (1997) in parents or siblings, even when the analysis was
restricted to relatives with the broad phenotype. Using the Embedded Figures Test,
arguably a more direct test of central coherence, Baron-Cohen and Hammer (1997)
found that both mothers and fathers of children with Asperger syndrome were faster to
identify hidden shapes (indicating a tendency for piecemeal, detail-focussed
processing). Happé, Briskman, and Frith (2001) included a larger range of both verbal
and visuospatial measures of central coherence with both parents and siblings, and
found that parents of children with autism – particularly fathers – showed a significant
bias towards piecemeal processing across the four tasks used compared with parents of
children with dyslexia and with no disorder. There were no significant differences
among the sibling groups, however.
217
Research on specific cognitive deficits has therefore revealed that deficits in
ToM, EF, and central coherence may all be characteristic of the broader phenotype, but
that results across studies are often inconsistent. In all three domains, significant
differences have been found in parents of autistic probands but often not in their
siblings, contrary to the notion that parents should be less impaired than siblings
because of the selection for parenthood described earlier. Even on measures of planning
and set-shifting where significant differences among sibling groups were found, Hughes
et al. (1999) noted that the deficits were not as strong in siblings as in parents. While it
is not clear that the selection for parenthood should extend beyond social and
communicative capabilities to cognitive characteristics, these parent-sibling
discrepancies still require explanation. Hughes et al. (1999) proposed that the use of
computerised tasks may favour young participants over parents, but this does not
account for studies using non-computerised tasks. Happé et al. (2001) suggested that
the tasks used may not be sufficiently sensitive in younger subjects. However, many
studies, including theirs, have found differences in parents (who one would expect to be
at a higher level than their children) using the same tasks as for siblings2. These authors
also suggested that genetically determined cognitive weaknesses may only emerge at a
certain age or become more pronounced with age. This would be an unusual finding in
domains such as ToM and certain aspects of EF, though, which typically develop
relatively early in life. These parent-sibling discrepancies therefore remain difficult to
explain.
A number of issues appear worthy of further investigation in future broad
phenotype studies. Firstly, identification of the profile of EF performance in relatives of
individuals with autism on tasks measuring the full range of EF components is yet to be
achieved and will augment research on EF deficits in probands. Secondly, comparison
of inter-task correlations in relatives of probands with autism versus control relatives
could be informative, with Hughes et al. (1997, 1999) finding unusual associations
between task performances in both parents and siblings of autistic individuals, which
they interpreted as suggesting the use of different strategies in performing the tasks.
Thirdly, given that cognitive deficits are often only found in a subset of relatives, it
remains to be seen whether this subset display both ToM and EF deficits and therefore
represent a general “cognitive impairment” subgroup, or whether there are different
subgroups with different types of cognitive deficit (most studies have only examined
2 Low sensitivity may be a result of floor effects in children as well as ceiling effects, but there is no evidence of floor level performance in the sibling studies reported above.
218
one cognitive domain). Fourthly, the relationship between performance on cognitive
tasks and the presence of certain behavioural traits is another important issue. Hughes
et al. (1997) found a modest but significant correlation between a composite EF score
and interviewers’ pre-test impressions of social abnormalities in parents of autistic
probands, and Briskman, Happé, and Frith (2001) found that parents of autistic
individuals who reported more preference for nonsocial activities in everyday life
tended to show weaker central coherence on testing. These findings show that the
cognitive weaknesses observed in the broad autism phenotype may hold relevance in
accounting for subtle behavioural traits also displayed by parents and siblings, but the
nature and specificity of these cognitive-behavioural relationships remains unclear, and
could be important for assessing the validity of the “multidimensional spectrum”
version of the multiple primary deficits model of ASDs (see Section 4.4.4 in Chapter 4).
Finally, while several studies have examined relationships between the IQ of the
proband and the behavioural or cognitive traits of family members, no published studies
have directly correlated performances of probands and relatives on the same ToM or EF
tasks. This could be a useful method of identifying which aspects of cognitive
functioning in autism are the most highly familial (and therefore which may be most
strongly coded in the autism genotype). In sum, greater specification of the cognitive
and behavioural characteristics of the broad autism phenotype will hopefully aid
progress in identifying relationships between genotype, endophenotype, and behavioural
phenotype both in autism and its milder variants. These issues are examined in Study
Two.
219
220
CHAPTER 6
Study Two: Theory of Mind and Executive Function in Siblings of Individuals with Autism Spectrum Disorders
6.1 Introduction 6.1.1 Aims 6.1.2 Hypotheses
6.2 Method
6.2.1 Participants 6.2.2 Procedure
6.3 Results
6.3.1 Sibling group comparisons on ToM and EF tasks 6.3.1.1 False belief tasks 6.3.1.2 Tower of London 6.3.1.3 IDED Set-shifting task 6.3.1.4 Response Inhibition and Load task 6.3.1.5 Opposite Worlds task 6.3.1.6 Pattern Meanings 6.3.1.7 Uses of Objects 6.3.1.8 Stamps task 6.3.1.9 Summary of sibling group comparisons
6.3.2 Comparisons between ASD siblings and ASD probands 6.3.3 Ability of cognitive variables to predict sibling group membership 6.3.4 Proband-sibling relationships within the ASD families
6.3.4.1 Correlations between proband IQ and siblings’ cognitive performances
6.3.4.2 Correlations between probands’ and siblings’ cognitive performances
6.3.5 Prevalence of deficits in ASD siblings 6.3.6 Correlations between ToM and EF 6.3.7 Dissociations between ToM and EF 6.3.8 Results from behavioural measures
6.4 Discussion
6.4.1 Endophenotype status of ToM and EF impairments 6.4.2 Differentiating the multiple deficits models
221
6.1 Introduction
6.1.1 Aims
As reviewed in Chapter 5, weaknesses in both ToM and EF have been identified in
parents and/or siblings of autistic probands in previous studies, but findings have been
inconsistent and there are several empirical issues yet to be examined. Study Two is an
investigation of ToM and EF deficits in siblings of individuals with ASDs. The main
aims of this study were i) to identify whether ToM or EF performance meets criteria for
an endophenotype or vulnerability marker for the autism genotype, and thereby seek
confirmation of the results of Study One regarding the relative primacy of ToM and EF
in ASDs; and ii) to collect further information relevant to distinguishing the various
multiple deficits models presented in Chapter 4 (Section 4.4.4). These aims were
addressed in several ways.
i) Aim 1: Determining endophenotype status. In this study, those ToM and EF
tasks on which probands with ASDs showed significantly poorer performance than
control probands in Study One were administered to siblings of individuals with ASDs
(“ASD siblings”) and control siblings. These tasks therefore included measures of false
belief understanding, planning, and both verbal and non-verbal inhibition and
generativity (even though non-verbal inhibition was found to be intact in probands using
the RIL task, that task was administered because probands showed difficulties in the
condition combining working memory and inhibitory requirements). Although only
marginal differences were found between proband groups on the IDED set-shifting task,
this task was also administered with siblings to enable comparison with Hughes et al.’s
(1999) previous findings on siblings using the original IDED task. The inclusion of
both verbal and non-verbal inhibition and generativity tasks represents an extension to
previous research.
Gottesman and Gould (2003, p. 639) outline five criteria for the identification of
an endophenotype. The following numbered points list these criteria and describe how
they were tested in the current research.
1. “The endophenotype is associated with illness in the population”. This was
demonstrated in Study One, which showed that both ToM and EF deficits were
associated with having an ASD diagnosis.
222
2. “The endophenotype is primarily state-independent (manifests in an individual
whether or not illness is active)”. This criterion is somewhat irrelevant in the case
of ASDs, as the disorder is present throughout the lifetime of the affected individual
(unlike, for example, depression or schizophrenia). Of course, one would still
expect the endophenotype to manifest in affected individuals.
3. “Within families, endophenotype and illness co-segregate”. This was indirectly
assessed by examining whether there was an increased incidence of ASDs in
siblings of ASD probands as compared with the normal population (the control
siblings were not used as a comparison group in this situation because having any
child with a clinical diagnosis of an ASD was an exclusion criterion for control
families). However, comparisons between affected siblings and controls on ToM
and EF variables (to assess whether the siblings with ASDs showed a similar
profile of deficits as the ASD probands) were not conducted because the size of the
affected group was too small for meaningful analyses (see Section 6.2.1),
particularly on some tasks which were administered to participants within a
restricted age range.
4. “The endophenotype found in affected family members is found in nonaffected
family members at a higher rate than in the general population”. This was tested by
examining whether there were any group differences in ToM or EF performance
between ASD and control siblings which remained significant after siblings with
ASD diagnoses were excluded. The ability of any deficits to discriminate ASD
siblings from control siblings was also calculated, as any useful endophenotype
should be unique to the disorder in question (Skuse, 2001).
5. “The endophenotype is heritable”. This was assessed by calculating correlations
between the ToM and EF performances of the ASD siblings in Study Two, and i)
the level of intellectual ability and ii) the ToM and EF performances of the ASD
probands in Study One, with the assumption that significant correlations would be
suggestive of a degree of familiality to the trait.
A sixth feature would also be expected, which is:
6. “The severity of the endophenotype in nonaffected family members is milder than
in the affected family members” (Slaats-Willemse, Swaab-Barneveld, de
Sonneville, van der Meulen, & Buitelaar, 2003). Relative severity of any ToM and
EF deficits in unaffected siblings as compared with affected probands was assessed
by comparing the effect sizes of any significant differences between sibling groups
with effect sizes for the proband groups.
223
Hence, this study addressed criteria 4, 5, and 6. The ability of ToM and/or EF deficits to
sufficiently meet these three criteria would suggest that they could be endophenotypes
for ASDs and therefore that they have some degree of primacy to ASDs. If one deficit
is better able to meet these criteria than the other, this would suggest superior relative
primacy.
ii) Aim 2: Testing multiple deficits models. If ToM and/or EF deficits were
identified in ASD siblings, the pattern of results would have implications for the various
versions of the multiple deficits model presented in Section 4.4.4 of Chapter 4.
Although the “third primary deficit” and “different developmental stages” versions of
the multiple primary deficit model were not tested in this study, it was possible to
examine (somewhat indirectly) the plausibility of the other two versions (the
“subgroups” and “multidimensional spectrum” models). If there were different
subgroups of ASDs displaying different ToM and EF profiles, then assuming these
subgroups corresponded with different ASD genotypes, it would be predicted that
similar subgroupings would be evident in the broad autism phenotype. This would be
demonstrated by results indicating the presence of “ToM-impaired, EF-intact” and “EF-
impaired, ToM-intact” siblings as was the case for probands (although impairments
would be more subtle), and furthermore it would be expected that siblings
demonstrating a particular ToM-EF profile would be the siblings of probands
demonstrating that same profile. These possibilities were examined in several ways.
Firstly, the prevalence of any deficits identified in group comparisons was calculated, to
examine whether they appeared to occur only in a subset of ASD siblings. Secondly,
correlations between ToM and EF in both ASD and control sibling groups were also
conducted to investigate whether ASD siblings showed a similar independence between
ToM and EF as was the case in probands; or if not, whether they may show unusual
patterns of association between the two domains. Thirdly, the incidence of ToM-EF
dissociations was examined. Finally, the correlations between proband and sibling
scores on ToM and EF tasks would be indicative of possible familial aggregation of
ToM-EF profiles, as described earlier.
The version of the multidimensional spectrum model examined in this study was
the one in which ToM and EF were purported to underlie different domains of
symptomatology, although it is acknowledged that other versions (which may be more
plausible based on the results of Study One) are possible. Abnormal social behaviours
and repetitive behaviours in ASD siblings were both measured in this study. Although
224
it would be expected that unaffected ASD siblings would not show symptoms of autism
even if they showed a cognitive endophenotype, under this version of the spectrum
model it would still be predicted that ToM or EF weaknesses would be associated with
increased levels of symptomatology in the relevant domain, even if that
symptomatology was subclinical. Therefore, this model was examined by first
analysing sibling group differences on behavioural measures (to confirm that some
subclinical symptomatology was present in ASD siblings), and then conducting
correlations between ToM and EF performances and these behavioural measures within
the ASD sibling group. However, it was recognised that this methodology may be
subject to the same problems as cognitive-behavioural correlations conducted in Study
One.
6.1.2 Hypotheses
Predictions for endophenotype status. Given that neither ToM nor EF deficits met all of
the criteria for primacy in Study One, no strong predictions were made with regard to
whether or not either domain would adequately meet all criteria for an endophenotype
of ASDs, although previous research has suggested that both ToM and EF have
potential endophenotype status. More confident predictions could be made in terms of
relative primacy, as EF deficits were found to be relatively more primary in Study One.
On this basis, it would be expected that i) significant weaknesses in ASD siblings on EF
tasks (which are less severe than the deficits displayed by probands) would be more
likely than on ToM tasks, and would be better able to predict group membership; and ii)
correlations between the EF performance of ASD siblings and probands would be
stronger than correlations between ToM performances of ASD siblings and probands.
Furthermore, it would be expected that the EF variables which demonstrated the
strongest evidence of primacy in ASD probands (i.e., verbal inhibition and verbal
generativity) would be the most likely variables on which weaknesses in ASD siblings
would emerge. However, given the concerns with interpretation of some of the findings
relevant to primacy in Study One, the possibility remained open that a ToM deficit may
also meet criteria as an endophenotype to an equal degree as EF deficits.
Predictions for multiple deficits models. Based on the results of Study One and on
previous research, it was predicted that if ToM and EF weaknesses were found, they
would only be evident in a subset of ASD siblings. Beyond that, however, given that
225
the results of the analyses relevant to the different multiple deficits models were very
much dependent upon the results of analyses relevant to determining endophenotype
status, no specific predictions were made prior to conducting the study. This aspect of
the study may therefore be considered exploratory.
6.2 Method
6.2.1 Participants
Siblings of ASD Group (“ASD siblings”)1. There were 108 siblings in this group,
ranging in age from 4 to 29 years. These siblings came from 68 families, thus in some
cases there was more than one sibling per family. Six siblings had received clinical
diagnoses of ASDs: three with diagnoses of autism, one with Asperger syndrome, and
two with PDDNOS. Three additional siblings had received diagnoses indicating
language impairment. As for the control group in Study One, autistic symptomatology
in siblings was assessed using the ASQ, and the ADI-R was administered for anyone
scoring above 10. All six siblings with clinical diagnoses of ASDs and two of the three
with language impairment met either full or partial criteria for autism on the ADI-R. In
addition, two other siblings without clinical diagnoses met partial criteria on the ADI-R.
Hence, there were 10 siblings altogether who met criteria for an ASD, which is 9.3% of
the ASD sibling group – a 31-fold increase compared to the population prevalence for
ASDs, which is around 0.3% (including ASDs besides autism; Fombonne, 2003).
Exclusion criteria were the same as for Study One (genetic abnormalities or
neurological dysfunction, except for epilepsy). No siblings were excluded for these
reasons. There were 10 ASD siblings with other clinical diagnoses (6 with ADHD, 2
with epilepsy, 1 with dyspraxia, and 1 with dyslexia).
Siblings of Control Group (“Control siblings”). Sixty-seven control siblings ranging in
age from 4 to 24 years participated in the study. These siblings came from 49 families.
No siblings in this group had clinical diagnoses of ASDs or exceeded the cutoff
1 This group included siblings of some probands with ASDs who were not included in Study One of this thesis because they were too low-functioning (but who were recruited as participants in the WAFSASD). Given that the ASD siblings were themselves matched with the siblings of the control group on age and PIQ, the inclusion of the siblings of low-functioning children with ASDs was considered to be valid. The relationship between siblings’ performance on cognitive tasks and the level of functioning of the proband was also examined, as reported in Section 6.3.4.1.
226
criterion on the ASQ. Three control siblings had other clinical diagnoses (2 with
ADHD and 1 with epilepsy).
Demographic characteristics of each group are presented in Table 22. The ASD and
control siblings were matched on chronological age, t(173) = .90, p > .1, and PIQ,
t(173) = .08, p > .1. All participants had a PIQ and VIQ of 60 or above. ASD siblings
had significantly lower VIQs than control siblings, t(173) = 2.28, p < .05. However,
when ASD siblings who met full or partial ADI-R criteria were excluded, the difference
in VIQs became only marginally significant, t(163) = 1.74, p = .08. There was a higher
proportion of girls in the control sibling group (68.7% vs. 42.6% in the ASD sibling
group), χ2 (1, N = 175) = 11.27, p < .01. This is likely to be due to the fact that in
attempting to select proband samples matched on gender, often the male child in the
family was selected as a control proband, resulting in a higher proportion of female
siblings in the control sibling group. To ensure sibling group comparisons were not
influenced by gender, it was included as an independent variable in all analyses. This
also enabled evaluation of any group by gender interactions (e.g., it may be that brothers
of ASD probands show greater heritability of autistic-like cognitive traits than sisters).
Table 22. Demographic characteristics of the sibling samples
ASD siblings
ASD siblings,
ADI-R subgroup
excluded
Control siblings
N 108 98 67
Age: Mean (SD, range) 11.33 (5.38, 4-29) 11.62 (5.42, 4-29) 10.61 (4.71, 4-24)
Male: Female 62: 46 53: 45 21: 46
PIQ: Mean (SD, range) 107.63
(17.08, 70-149)
107.9
(17.7, 70-149)
107.42
(16.18, 58-146)
VIQ: Mean (SD, range) 102.41
(15.02, 66-141)
103.67
(14.37, 66-141)
107.61
(14.14, 72-138)
With an n of 108 in the ASD sibling sample and 67 in the control sibling sample, the
power of the study to detect medium sized effects (i.e., d = .5) at an alpha level of .05
was excellent at .94.
227
6.2.2 Procedure
The same tests, questionnaires and interviews were used in this study as in Study One2,
all of which are described in Chapter 3. The procedure for this study was also identical
to that used in Study One (see Section 4.2.2 of Chapter 4).
6.3 Results
This section includes analyses addressing i) group comparisons between ASD and
control siblings on ToM and EF tasks, both before and after exclusion of siblings who
met ADI-R criteria for an ASD; ii) the relative severity of any weaknesses in ASD
siblings compared with ASD probands; iii) the ability of task variables to predict
ASD/control sibling group membership; iv) relationships between probands’ and
siblings’ scores; v) the prevalence of deficits in the ASD sibling group; vi) correlations
between ToM and EF variables; vii) dissociations between ToM and EF performances;
and viii) results from behavioural measures. Hence, analyses i) to iv) assess the
endophenotype status of ToM and EF, and v) to viii) are aimed primarily at assessing
the subgroup and multidimensional spectrum models. As for Study One, SPSS Version
10.0.5 was used for all analyses. Data screening was handled in the same way as for
Study One (see Section 4.3.1 in Chapter 4).
6.3.1 Sibling group comparisons on ToM and EF tasks
The consistent approach to group comparisons that was used in Study One (as described
in Section 4.3.2) was also followed in this study. The ASD and control sibling groups
remained matched on age and PIQ for all tests which were administered to participants
within a restricted age range. For some tasks (all false belief tasks, the Opposite Worlds
task, the RIL task, the IDED set-shifting task, and the Stamps task), the two groups of
siblings who completed those tasks, including participants meeting full or partial criteria
on the ADI-R, were also matched on VIQ.
As mentioned in Section 6.2.1, gender was included as an independent variable
in all sibling group comparisons on continuous variables. In the case of dichotomous
2 Although there were no group differences on the Dewey stories and Relational Complexity tasks in the proband study, these tasks were actually also administered to the siblings in this study. Hence, the order and length of task administration was identical across the two studies. Results from these tasks were analysed out of interest and there were no significant differences between the sibling groups.
228
variables, gender effects were assessed by conducting separate chi-square analyses for
brothers (i.e., ASD brothers versus control brothers) and sisters (ASD sisters versus
control sisters). These separate analyses are only reported if there were significant
group differences for one gender but not another (or if group differences were in
opposite directions for the two genders); otherwise, only the results of overall chi-
square analyses including both genders are reported. For all variables, separate means
and standard deviations (or proportions of high/low scorers) for brothers and sisters are
only reported if there were significant group by gender interactions on the task, or if
displaying separate data for brothers and sisters was meaningful for other reasons.
Similarly, in analyses where age, PIQ and/or VIQ were covaried or controlled, gender
was only included as an independent variable if significant interactions involving
gender had been found in initial group comparisons.
In all sibling group comparisons where there were significant group differences,
analyses were repeated after excluding all ASD siblings who met full or partial criteria
on the ADI-R. Results of these repeat analyses are reported in each of the relevant
sections.
As in Study One, the influence of participants with non-ASD diagnoses (of
which there were 10 in the ASD sibling group and 3 in the control sibling group) was
checked by repeating all group comparisons after excluding these participants from the
sample. This did not affect any of the results (i.e., both non-significant and significant
differences remained so, both before and after exclusion of participants meeting full or
partial ADI-R criteria), with the exception of the Stamps task complexity score. The
change in the result for this variable is reported in Section 6.3.1.8.
6.3.1.1 False belief tasks
As in Study One, a large proportion of participants gained perfect scores for both belief
and control questions on all false belief tasks. All variables were recoded as
dichotomous such that a perfect score was coded as 1 and any other score as 0.
Four participants (two ASD siblings and two control siblings) were not
administered the First-order and Second-order false belief tasks due to equipment
malfunction. These participants had all passed the Simple false belief task and were
therefore assigned the mean value of other participants in their group who had passed
the Simple false belief task. The overall sample size for all false belief tasks (which
were administered to participants within a restricted age range) was 148 (87 ASD
229
siblings and 61 control siblings). As for Study One, the ns for the memory and reality
questions, as well as the “own belief” questions in the Simple false belief task, were
limited to those who actually did the task (as these questions were not assumed to be
passed or failed according to performance on other false belief tasks, as was the case for
the belief questions). Percentages of participants gaining perfect scores on belief
questions (i.e., “perfect scorers”) in each group for each false belief task are presented
in Table 23.
i) Simple false belief task. Chi-square analyses revealed that there was no
statistically significant difference between the ASD and control siblings on reality
questions, χ2 (1, N = 51) = 1.34, p > .1, belief questions referring to the participant’s
own previous belief, χ2 (1, N = 51) = .84, p > .1, or belief questions referring to other’s
beliefs, χ2 (1, N = 148) = .95, p > .1. There were no significant differences when
brothers and sisters’ results were analysed separately.
Performance on the questions relating to the participant’s own previous belief
was significantly correlated with VIQ, r = .34, p < .05. Performance on others’ belief
questions was significantly correlated with both age, r = .33, p < .001, and VIQ, r = .35,
p < .001. Group remained a non-significant predictor of performance on both own
belief questions, z = .1, p > .1, and others’ belief questions, z = .48, p > .1, when VIQ
(and age in the case of the latter variable) was controlled using logistic regression.
ii) First-order false belief task. The performance of ASD and control
siblings did not differ significantly on reality questions, χ2 (1, N = 113) = .31, p > .1,
memory questions, χ2 (1, N = 113) = .01, p > .1, or belief questions, χ2 (1, N = 148) =
1.20, p > .1. Results were the same for brothers and sisters.
Performance on belief questions was significantly correlated with both age, r =
.51, p < .001, and VIQ, r = .20, p < .05. In a logistic regression with age, VIQ and
group as predictors of performance on belief questions, the independent contribution of
group remained non-significant, z = 1.97, p > .1.
iii) Second-order false belief task. Again, there was no significant difference
between the ASD and control siblings on reality questions, χ2 (1, N = 127) = .06, p > .1,
memory questions, χ2 (1, N = 127) = .09, p > .1, or belief questions, χ2 (1, N = 148) =
.11, p > .1, and no difference in the results when brothers and sisters were examined
separately.
Scores on belief questions were significantly correlated with age, r = .47, p <
.001, and VIQ, r = .34, p < .001. Group was not a significant predictor of performance
230
on belief questions in a logistic regression with age, VIQ and group as the predictors, z
= .04, p > .1.
iv) Overall false belief performance indices. As for the probands, an
aggregate score and a more lenient alternative aggregate score were calculated for
siblings. There was no significant group difference in the proportion of perfect scorers
on the aggregate score, χ2 (1, N = 148) = .01, p > .1, or in the proportion of high scorers
on the alternative aggregate, χ2 (1, N = 148) = .57, p > .1. Results were the same when
brothers and sisters were analysed separately.
Both aggregate scores were correlated with age (r = .52, p < .001, for the
aggregate score and r = .48, p < .001, for the alternative aggregate) and VIQ (r = .32, p
< .001, for the aggregate score and r = .34, p < .001, for the alternative aggregate).
When logistic regressions were performed with age, VIQ, and group as the predictors,
group was not a significant predictor of either the aggregate score, z = .42, p > .1, or the
alternative aggregate, z = .14, p > .1.
Table 23. False belief task results: Percentage of siblings in each group with perfect
scores [or high scores for the alternative aggregate] on belief questions, and significance
of group comparisons
ASD siblings Control siblings p p with age/
IQ control
Simple false belief:
Own belief 74.2 85.0 - -
Others’ belief 90.8 95.1 - -
First-order false belief 82.8 75.4 - -
Second-order false belief 74.7 77.0 - -
Aggregate score 71.3 70.5 - -
Alternative aggregate [80.5] [85.2] - -
* p < .05; ** p < .01; *** p < .001; - p > .05.
6.3.1.2 Tower of London (ToL)
As in Study One, the number of rule violations per block administered was highly
skewed and was recoded as a dichotomous variable, with participants making 0-1
violations per block being given a score of 0 (“low rule violators”) and participants
making any higher number of violations scored as 1 (“high rule violators”).
231
Two ASD siblings had missing data on the ToL, and were not included in
analyses. A two-way ANOVA comparing the total adjusted extra move scores of the
ASD siblings and control siblings revealed no significant effect of group or gender, and
no significant interaction; largest F(1, 169) = .06, p > .1 (ASD siblings: M = 21.65, SD
= 8.72; Control siblings: M = 21.99, SD = 9.60). A chi-square analysis also showed that
the proportion of high rule violators in the ASD sibling group (25.5%) did not differ
significantly from the proportion in the control sibling group (29.9%), χ2 (1, N = 173) =
.40, p > .1. This difference was non-significant for both brothers and sisters.
The total adjusted extra moves score correlated significantly with age, r = -.73, p
< .001, and rule violations were significantly correlated with both age, r = -.58, p <
.001, and VIQ, r = -.17, p < .05. An ANCOVA conducted on the total adjusted extra
move score revealed that the group difference remained non-significant when age was
covaried, F(1,170) = .50, p > .1. Group also remained a non-significant predictor of
rule violation status (low/high) when age and VIQ were assessed independently in a
logistic regression, z = .19, p > .1.
6.3.1.3 IDED Set-shifting task
All set-shifting variables were again highly skewed, and all variables were recoded such
that any participant making 0 or 1 errors was assigned a score of 0 (“low error scorers”)
and any participant making a higher number of errors was given a score of 1 (“high
error scorers”).
The overall N for the task (which had a restricted age range) was 129 (81 ASD
siblings and 48 control siblings). Due to computer malfunction, data for the
Perseveration condition from one ASD sibling were invalid and not included in
analyses. The percentage of low error scorers for each stage in each task condition is
displayed in Table 24. There were no significant group differences overall on any
variable. When brothers and sisters were analysed separately, a significant difference
was observed in the SD stage of the Learned Irrelevance condition such that there was a
higher proportion of high error scorers among sisters of ASD probands than among
control sisters, χ2 (1, N = 70) = 5.08, p < .05. There was no significant difference
between the brother groups on this variable, and no discrepancies in the results from
brothers and sisters on other variables.
Errors made on the SD stage of the Perseveration condition correlated
significantly with both age, r = -.24, p < .01, and PIQ, r = -.17, p < .05. Age was also
232
significantly correlated with errors made in the Learned Irrelevance condition on the SD
stage, r = -.21, p < .05, and the IS stage, r = -.19, p < .05. Group remained a non-
significant predictor of performance on these variables when logistic regressions were
performed with age and group (and PIQ where relevant) as predictors; the largest z =
1.80, p > .1. When brothers and sisters were analysed separately for the Learned
Irrelevance SD stage variable, group was again a significant predictor of performance
on that variable for sisters when age was accounted for, z = 4.14, p < .05. The group
difference also remained significant when sisters meeting full or partial ADI-R criteria
were excluded, χ2 (1, N = 69) = 5.29, p < .05.
Table 24. IDED Set-shifting task results: Percentage of low error scorers in each sibling
group for each stage of each task condition, and significance of group comparisons
ASD siblings Control siblings p p with age/
IQ control
Perseveration condition:
SD stage 76.3 75.0 - -
SDR stage 57.5 70.8 -
CD stage 71.3 75.0 -
IDS stage 76.3 83.3 -
EDS stage 68.8 75.0 -
Learned Irrelevance condition
SD stage – brothers only 74.3 85.7 - -
SD stage – sisters only 80.0 97.1 * *
SDR stage 77.8 81.3 -
CD stage 70.4 83.3 -
IDS stage 77.8 87.5 - -
EDS stage 27.2 25.0 -
* p < .05; ** p < .01; *** p < .001; - p > .05.
6.3.1.4 Response Inhibition and Load (RIL) task
For all RIL conditions, error variables (representing the percentage of errors made) were
again highly skewed, with many participants making a low percentage of errors. These
variables (with the exception of the shape error score) were recoded such that 0-2%
233
errors was coded as 0 (a “low error score”), and any higher percentage of errors was
coded as 1 (a “high error score”). However, as for Study One, the main error variables
used in analyses were the inhibition error difference score, load error difference score,
and inhibition + load error difference score. These difference scores were normally
distributed. Five outliers were trimmed: one ASD sibling’s inhibition and load error
difference scores, and the inhibition error difference score for two other ASD siblings
and one control sibling. The distribution of the shape error score was slightly positively
skewed but transformation was not considered necessary.
Median RT variables for all conditions demonstrated roughly normal
distributions, but again, an inhibition RT difference score, load RT difference score, and
inhibition + load RT difference score were also calculated. Five outliers were trimmed:
one control sibling’s inhibition RT difference score, one ASD sibling’s load and
inhibition + load RT difference scores, and the inhibition + load RT difference scores of
one ASD and one control sibling.
The overall N for the task was 126 (79 ASD siblings and 47 control siblings).
Table 25 displays the mean and SD of each group (and the significance of group
comparisons) for error and RT difference scores, and the shape error score. On the error
difference scores, there were no significant main effects of group or gender and no
significant group x gender interactions. There were no significant differences in any of
the individual conditions when error data were examined separately for each condition
(either overall or for brothers or sisters). The shape error score did not differ
significantly between ASD and control siblings, F(1, 122) = 1.90, p > .1, and there was
no significant effect of gender, F(1, 122) = .40, p > .1, and no significant interaction,
F(1, 122) = .01, p > .1. On both the RT difference scores and the separate RT data for
each condition, there were no significant main effects of group or gender and no
significant interactions. In subsequent analyses, only the error and RT difference scores
were used, and separate error and RT data for Conditions 1-3 were not included (nor
was gender included as a factor).
There were a number of significant correlations between age and IQ variables
and both error and RT difference scores from the RIL task. The inhibition + load error
difference score correlated significantly with VIQ, r = -.24, p < .01. The shape error
score correlated significantly with age, r = -.39, p < .001, PIQ, r = -.30, p < .01, and
VIQ, r = -.18, p < .05. The inhibition RT difference score was significantly correlated
with age, r = -.24, p < .01, and VIQ, r = -.24, p < .01. Finally, the inhibition + load RT
difference score correlated significantly with age, r = -.20, p < .05. Group differences
234
remained non-significant (or group was a non-significant predictor) for all of the above
variables, with the exception of the shape error score, when age and/or IQ variables
were partialled out using ANCOVA or logistic regression. For the shape error score,
group differences became significant when age, PIQ, and VIQ were introduced as
covariates in an ANCOVA (VIQ was not “blocked” and used as an additional IV
because the groups were matched on VIQ for the RIL task), F(1, 121) = 4.40, p < .05.
This indicates that when extraneous variance caused by age and IQ factors was
removed, ASD siblings were found to make a significantly higher number of errors than
control siblings on a measure of working memory on a task where inhibition was
required. Importantly, this group difference in the shape error score remained
significant when siblings who met full or partial ADI-R were excluded from the sample,
F(1, 116) = 4.08, p < .05.
Table 25. RIL task results: Mean (and SD) of each sibling group, and significance of
group comparisons, for error and RT difference scores and the shape error score
ASD siblings Control siblings p p with age/
IQ control
Error difference scores:
Inhibition 0.76 (2.80) 1.10 (3.39) -
Load 0.73 (3.67) 0.35 (3.47) -
Inhibition + load 1.43 (3.88) 1.49 (3.50) - -
RT difference scores:
Inhibition 159.35 (147.76) 140.92 (141.82) - -
Load 145.44 (132.66) 188.73 (137.83) -
Inhibition + load 304.52 (203.07) 329.05 (216.90) - -
Working memory measure:
Shape error score 13.42 (13.57) 9.50 (12.02) - *
* p < .05; ** p < .01; *** p < .001; - p > .05.
6.3.1.5 Opposite Worlds task
No transformations were required on Opposite Worlds task variables. Two ASD
siblings and two control siblings demonstrated outlying scores (two on the Same World
error score, one on the Same World time score, and one on both the Same and Opposite
235
World time scores), which were trimmed. Means and SDs for all variables are
displayed in Table 26.
The N for the task was 100 (56 ASD siblings and 44 control siblings). For the
error scores, a three-way repeated measures ANOVA was conducted with group and
gender as between-subjects factors and condition (Same World, Opposite World) as the
within-subjects factor. There was a significant main effect of condition, F(1, 96) =
19.77, p < .001, but there was no significant main effect of group, F(1, 96) = .24, p > .1,
or gender, F(1, 96) = 1.80, p > .1. There was a significant interaction between group
and condition, F(1, 98) = 4.63, p < .05, but no other significant interactions. Follow-up
simple effects analyses showed that there was no significant difference between the
groups in the Same World error score, t(98) = 1.31, p > .1, or the Opposite World error
score, t(98) = 1.02, p > .1, however the control siblings demonstrated a significantly
larger error difference score than the ASD siblings (as reflected in the interaction).
Examination of the pattern of results suggested that this was due to a combination of
both the ASD siblings tending to make slightly more Same World errors than the
control siblings, and the control siblings making slightly more Opposite World errors
than ASD siblings.
Time scores were also analysed using a three-way repeated measures ANOVA
with group and gender as the between-subjects factors and condition as the within-
subjects factor. There was a significant main effect of condition, F(1, 96) = 182.56, p <
.001, but no significant effect of group, F(1, 96) = 1.55, p > .1, or gender, F(1, 96) =
.02, p > .1. The interaction between group and gender was not significant, F(1, 96) =
.02, p > .1, however there was a marginally significant interaction between group and
condition, F(1, 96) = 3.93, p = .05, and a significant interaction between gender and
condition, F(1, 96) = 9.51, p < .01. The group x gender x condition interaction was not
significant, F(1, 96) = .27, p > .1. With regard to the group x condition interaction,
follow-up analyses showed that there was a marginally significant difference between
the groups in the Same World time score such that ASD siblings took slightly longer
than control siblings, t(98) = 1.81, p = .07, but no significant difference in the Opposite
World time score, t(98) = .77, p > .1. In terms of the gender x condition interaction,
follow-up analyses indicated that there was no significant gender difference in either the
Same World time score, t(98) = .64, p > .1, or the Opposite World time score, t(98) =
.98, p > .1, but brothers demonstrated a significantly larger time difference score than
sisters (as reflected in the interaction). The pattern of results suggested that this was due
236
to a tendency both for sisters to take slightly longer in the Same World condition and for
brothers to take slightly longer in the Opposite World condition.
Table 26. Opposite Worlds results: Mean (and SD) and significance of group
comparisons for each sibling group for error/time scores in each condition and
difference scores, and for each gender for time scores
ASD siblings Control siblings p p with age/
IQ control
Error variables:
Same World error score 1.22 (1.41) 0.87 (1.27) - -
Opposite World error score 1.68 (1.86) 2.05 (1.67) - -
Error difference score 0.42 (1.70) 1.12 (1.56) * *
Time variables:
Same World time score 25.72 (7.04) 23.41 (5.24) - -
Opposite World time score 31.75 (8.95) 30.47 (7.39) - -
Time difference score 6.19 (5.60) 6.94 (4.16) - *
Brothers Sisters p p with age/
IQ control
Same World time score 24.27 (6.01) 25.09 (6.73) - -
Opposite World time score 32.05 (9.03) 30.42 (7.57) - -
Time difference score 7.97 (5.27) 5.24 (4.42) ** **
* p < .05; ** p < .01; *** p < .001; - p > .05.
Note: The difference scores relate to the interaction term on repeated measures
ANOVAs.
Age was significantly correlated with all task variables: the Same World error score, r =
-.26, p < .01, Opposite World error score, r = -.27, p < .01, Same World time score, r = -
.50, p < .001, and the Opposite World time score, r = -.56, p < .001. VIQ correlated
significantly with the Same World error score, r = -.20, p < .05, and the Opposite World
time score, r = -.27, p < .01. Age and VIQ were introduced as covariates (VIQ was
covaried because the groups were matched on VIQ for this task) in a two-way group x
condition repeated measures ANCOVA on the error scores and three-way ANCOVA on
the time scores (including gender as a between-subjects factor, as there were
interactions involving gender for the time scores). There was no change in any of the
237
results with the exception that the group x condition interaction on the time scores
increased in significance, F(1, 94) = 6.56, p < .05. This interaction remained significant
when siblings meeting full or partial criteria on the ADI-R were excluded, F(1, 90) =
6.94, p < .05. The interaction between group and condition for the error scores also
remained significant when siblings meeting ADI-R criteria were excluded, F(1, 94) =
6.69, p < .05 (age and VIQ were not covaried in this analysis as there was no difference
in the original result when these variables were accounted for).
6.3.1.6 Pattern Meanings
Individual error types were not analysed in this study (this level of detail was not
considered essential, particularly given that the ASD and control groups in Study One
did not show significant differences in error variables). As for Study One, the sum of
errors variable was skewed and transformed using a logarithm equation. The number of
correct responses variable was normally distributed.
There was no significant difference in the number of correct responses produced
by ASD siblings (M = 26.71, SD = 9.55) and control siblings (M = 26.52, SD = 8.13),
no significant effect of gender on this variable, and no significant group by gender
interaction; largest F(1, 171) = 1.90, p > .1. The sum of errors was not significantly
different between ASD siblings (Median = 4, Range = 0-42, prior to transformation) and
control siblings (Median = 5, Range = 0-56), F(1, 171) = 2.58, p > .1. There was a
trend for brothers to make more error responses than sisters, F(1, 171) = 3.65, p = .06,
but the interaction between group and gender was not significant for the sum of errors
variable, F(1, 171) = .16, p > .1.
The sum of errors was correlated with age, r = -.51, p < .001, and VIQ, r = -.24,
p < .01. Group comparison of the sum of errors remained non-significant in an
ANCOVA with group and VIQ level as the IVs and age as a covariate, F(1, 168) =
2.23, p > .1, and the interaction between group and VIQ level was not significant, F(2,
168) = .06, p > .1.
6.3.1.7 Uses of Objects
As for the Pattern Meanings task, individual error types were not analysed. No
transformation was necessary for the sum of errors variable, but four outliers were
trimmed (for 2 ASD siblings and 2 control siblings). The total number of correct
238
responses, as well as the number of correct responses for conventional and non-
conventional items separately, were all normally distributed. Table 27 displays means
and SDs for these variables for each sibling group, and the significance of group
comparisons.
One ASD sibling had missing data and was not included in analyses. In a three-
way repeated measures ANOVA on the number of correct responses with group and
gender as the between-subjects factors and condition (conventional, non-conventional)
as the within-subjects factor, there was a significant main effect of condition such that
more correct responses were generated in the non-conventional condition than in the
conventional condition, F(1, 170) = 305.84, p < .001, but no significant main effect of
group, F(1, 170) = .67, p > .1, or gender, F(1, 170) = 2.50, p > .1. There were no
significant interactions between any variables. Separate totals for conventional and
non-conventional items were not used in further analyses. In a two-way group x gender
ANOVA on the sum of errors, there was no main effect of group F(1, 170) = 1.12, p >
.1. Brothers (M = 19.38, SD = 13.56) produced significantly more error responses than
sisters (M = 15.09, SD = 12.60), F(1, 170) = 5.21, p < .05, but there was no significant
interaction between group and gender, F(1, 170) = .01, p > .1.
The number of correct responses was significantly correlated with both age, r =
.51, p < .001 and VIQ, r = .24, p < .01. The sum of errors also correlated significantly
with both age, r = -.36, p < .001, and VIQ, r = -.22, p < .01. Group differences in both
correct responses and the sum of errors remained non-significant in ANCOVAs where
group and VIQ level were the IVs and age was covaried, and there were no significant
interactions between group and VIQ level in either analysis.
Table 27. Uses of Objects results: Mean (and SD) of each sibling group, and
significance of group comparisons
ASD siblings Control siblings p p with age/
IQ control
Correct responses:
- Total 25.57 (10.32) 27.55 (11.46) - -
- Conventional items 19.57 (7.80) 21.0 (8.99)
- Non-conventional items 22.76 (8.22) 24.34 (9.75)
Sum of errors 16.73 (12.69) 17.72 (14.04) - -
* p < .05; ** p < .01; *** p < .001; - p > .05.
239
6.3.1.8 Stamps task
Both the rule adherence and restriction scores demonstrated highly skewed distributions
and were recoded as dichotomous variables, in the same way as for Study One. For rule
adherence, a score between 0 and 6 inclusive was coded as 0 and a score of 7 or 8 was
coded as 1. For restriction, a score of 0 was left as 0 and a score between 1 and 8
inclusive was coded as 1. The complexity and originality scores were approximately
normally distributed. Means and SDs for the latter two variables and the proportion of
low scorers for the former two variables, along with the significance of group
comparisons for all scores, are presented in Table 28.
Table 28. Stamps task results: Mean (and SD) of each sibling group [or the percentage
of low scorers for dichotomous variables], and significance of group comparisons
ASD siblings Control siblings p p with age/
IQ control
Complexity score 18.90 (3.28) 20.18 (3.86) * *
Originality score 4.21 (3.04) 4.34 (2.64) - -
Restriction score [94.6] [88.7] -
Rule adherence score [19.8] [21.3] - -
* p < .05; ** p < .01; *** p < .001; - p > .05.
The N for the task was 154 (92 ASD siblings and 42 control siblings). There was a
significant group difference on the complexity score, F(1, 150) = 4.68, p < .05,
indicating that the ASD siblings produced less complex patterns than control siblings.
The effect of gender was not significant for this variable, F(1, 150) = .10, p > .1, and the
interaction between group and gender was not significant, F(1, 150) = .02, p > .1. In a
two-way ANOVA on the originality scores, there was no significant effect of group or
gender, and no significant interaction, largest F(1, 150) = .67, p > .1. Chi-square
analyses revealed that there was no significant group difference in the percentage of low
scorers on the restriction score, χ2 (1, N = 154) = 1.77, p > .1, or the rule adherence
score, χ2 (1, N = 154) = .05, p > .1. Results were the same when brothers and sisters
were analysed separately.
Originality scores were significantly correlated with both age, r = .46, p < .001,
and VIQ, r = .23, p < .01. Age also correlated significantly with complexity scores, r =
.44, p < .001, and rule adherence scores, r = .35, p < .001. In a two-way ANCOVA
240
with group and VIQ level as the IVs and age as a covariate, the group difference in the
originality score remained non-significant, F(1, 147) = .03, p > .1. The interaction
between group and VIQ level was not significant, F(2, 147) = .20, p > .1. When a
logistic regression was performed on the rule adherence score, group remained a non-
significant predictor when it was assessed independently of age, z = .01, p > .1. In an
ANCOVA with age as a covariate, the group difference in the complexity score
remained significant, F(1, 151) = 5.78, p < .05. The difference also remained
significant when participants meeting full or partial criteria on the ADI-R were excluded
from the sample, F(1, 142) = 4.0, p < .05. However, when participants with non-ASD
diagnoses were additionally excluded, the group difference in the complexity score
became only marginally significant, F(1, 131) = 2.96, p = .09.
6.3.1.9 Summary of sibling group comparisons
In summary, ASD siblings performed significantly more poorly than control siblings on
tasks measuring working memory (under conditions where inhibition was required) and
non-verbal generativity. When participants with non-ASD diagnoses were excluded,
the group difference on the non-verbal generativity measure became only marginally
significant. Sisters of ASD probands also made more errors than sisters of control
probands on the simple first stage of the IDED set-shifting task Learned Irrelevance
condition. Control siblings showed larger error and time difference scores on the
Opposite Worlds test, which appeared to be a fairly spurious result which was equally
attributable to ASD siblings performing slightly (but not significantly) more poorly on
the Same World condition and control siblings performing slightly (but not
significantly) more poorly on the Opposite World condition. There were no significant
group differences on measures of ToM, planning, set-shifting (i.e., in the EDS stages),
non-verbal inhibition, or verbal generativity.
All of the significant group differences remained significant when ASD siblings
meeting full or partial criteria on the ADI-R were excluded, which indicates that
criterion 4 for endophenotype status was met (see Section 6.1.1). However, as the
poorer performance on the simple first stage of the IDED set-shifting task in sisters of
ASD probands did not correspond with a deficit displayed by the ASD probands
themselves, this suggests that the weakness displayed by ASD sisters on this task was
not indicative of an endophenotype for ASDs (as it violates criterion 1). Therefore, this
variable is not included in subsequent analyses examining other criteria for
241
endophenotype status. In addition, because the group differences on the Opposite
Worlds error and time difference scores appeared to be spurious outcomes deriving
from two non-significant differences in opposite directions and therefore did not
represent meaningful strengths or weaknesses in the ASD sibling group (i.e., were not
candidates for an endophenotype), these variables were not included in subsequent
analyses either. Hence, the RIL task shape error score and the Stamps task complexity
score were the only two candidate endophenotype variables remaining.
6.3.2 Comparisons between ASD siblings and ASD probands
If these two variables are possible endophenotypes for ASDs, then it would be expected
that the performance displayed by ASD siblings would be poorer than that of control
siblings, but not as poor as that of ASD probands. To compare the severity of deficits
across these three groups, it was decided not to directly compare the scores, as the
groups were not matched on PIQ or VIQ (and therefore any differences could be
attributable to those variables). Instead, the effect sizes of the differences found
between ASD and control siblings were calculated and compared with the effect sizes of
the differences between the two proband groups in Study One. These two sets of effect
sizes are presented in Table 29. It is evident that the effect sizes for the sibling group
differences (both small effects) are smaller than those for the proband group differences
(both medium effects), consistent with predictions.
Table 29. Effect sizes, r (and d), of significant group differences between sibling
groups and between proband groups
Measure ASD versus control
siblings
ASD versus control
probands
Working memory:
RIL task shape error score .15 (.31) .27 (.56)
Generativity:
Stamps task complexity score .18 (.36) .28 (.58)
242
6.3.3 Ability of cognitive variables to predict group membership
In order to examine if either of two candidate endophenotype variables were able to
discriminate ASD siblings from control siblings, a direct logistic regression was
performed with group as the outcome variable, and the RIL task shape error score and
Stamps task complexity score as the predictors. There were 65 ASD siblings and 42
control siblings with data for both predictor variables, and these limited groups were
matched on age (M = 11.27, SD = 2.62 for ASD siblings; M = 11.59, SD = 2.74 for
control siblings), t(105) = .61, p > .1, PIQ (M = 107.69, SD = 17.74 for ASD siblings; M
= 103.83, SD = 15.81 for control siblings), t(105) = 1.15, p > .1, and VIQ (M = 105.65,
SD = 14.60 for ASD siblings; M = 107.43, SD = 13.33 for control siblings), t(105) =
.64, p > .1. Thus, none of these matching variables were included in the regression.
A test of the full model with both predictors against a constant-only model was
statistically reliable, χ2 (2, N = 107) = 8.69, p < .05, indicating that the two predictors
together reliably distinguished ASD from control siblings. 90.8% of ASD siblings and
35.7% of control siblings were classified correctly by the model, with an overall success
rate of 69.2%. This pattern of results suggests that the model was sensitive to ASD
sibling group membership, but not specific. Table 30 presents regression coefficients,
Wald statistics, odds ratios, and their 95% confidence intervals for each predictor.
According to the Wald criterion, only the Stamps task complexity score was a
significant individual predictor of group membership. Of note, when age, PIQ, and VIQ
were included in the regression, the RIL task shape error score also became marginally
significant as a predictor, z = 2.74, p = .098 (note that the group difference in the shape
error score also became significant only after the age and IQ variables were controlled).
Table 30. Results of logistic regression analysis of sibling group membership
95% C. I. for odds ratio
Wald test ___________________
Variables B (z-ratio) Odds ratio Upper Lower
RIL task:
Shape error score -.02 1.85 .98 .95 1.01
Stamps task:
Complexity score .20 5.16* 1.22 1.03 1.44
*p < .05; ** p < .01; *** p < .001.
243
6.3.4 Proband-sibling relationships within the ASD families
To examine correlations between ASD probands’ and ASD siblings’ scores on cognitive
measures, data were used from one sibling in each family who was closest in age to the
proband. Correlations were conducted for all ToM and EF variables regardless of
whether or not they were tasks on which ASD siblings demonstrated weaknesses, as it
was possible that sibling performances could covary with proband performances even if
the sibling performances were in the normal range. Because of concerns that age would
strongly mediate relationships between probands’ and siblings’ scores, age-scaled
scores were calculated for use in all correlations. For each variable, the regression
equation: predicted score = slope x age + intercept was calculated using the combined
control proband and sibling samples. If the relationship between the variable and age
was curvilinear, the log of age was used in this equation instead, if it resulted in a more
linear relationship (this was the case for ToL rule violations, IDED set-shifting Learned
Irrelevance condition EDS stage errors, the sum of errors on the Pattern Meanings and
Uses of Objects tasks, and the Stamps task complexity score). ASD participants (both
probands and siblings) then had their scores converted to age-scaled z-scores by
subtracting the predicted score from the obtained score and dividing by the standard
error of the estimate. Because linear regression equations could not be calculated for
dichotomous variables, some of these were not used in the proband-sibling correlations
(all IDED set-shifting variables except the EDS stage in the Learned Irrelevance
condition, and the Stamps task rule adherence and restriction scores). For variables
which were not excessively skewed before dichotomisation, the original continuous
form of the variable was used in correlations (the false belief aggregate score, ToL rule
violations, and the IDED Learned Irrelevance EDS stage errors).
6.3.4.1 Correlations between proband IQ and siblings’ cognitive
performances
The relationship between ASD probands’ level of functioning and their siblings’
performances on cognitive tasks was assessed by calculating correlations between
probands’ IQ scores and siblings’ scores on cognitive measures. This was particularly
important as the probands of the sibling groups were not matched on either VIQ or PIQ
(as probands who were part of the WAFSASD but were too low-functioning to
participate in Study One had siblings who were included in Study Two). The IQ data
244
from the low-functioning probands who were not included in Study One (but whose
siblings were included in Study Two) were included in these correlations. Age-scaled
scores were used for the siblings’ scores, but no z-scores were calculated for probands’
IQs (as these were already age-scaled). If raw correlations were significant, partial
correlations were also conducted where sibling VIQ and PIQ were controlled, to ensure
that the correlations were not mediated by relationships between proband and sibling IQ
(the correlations between proband and sibling PIQ and VIQ were both significant; PIQ:
r = .25, p < .01; VIQ: r = .30, p < .01).
Table 31. Raw and partial correlations between proband PIQ and VIQ and siblings’ scores on ToM and EF measures Proband IQ score Sibling score on cognitive task PIQ VIQ False belief aggregate score .45** .25* .32* .08 ToL: Adjusted extra moves score .24* .08 .26* .07 Rule violations -.11 -.10 IDED set-shifting Learned Irrelevance condition EDS stage errors
.01 .16
RIL task: Inhibition error difference score -.13 -.29* -.05 Load error difference score -.13 .0 Inhibition + load error diff. score -.20 -.25 Inhibition RT difference score -.15 -.09 Load RT difference score .24 .24 Inhibition + load RT diff. score .06 .10 Shape error score -.05 -.01 Opposite Worlds: Error difference score .05 -.01 Time difference score -.19 .03 Pattern Meanings: Correct responses -.07 -.12 Sum of errors -.22 -.02 Uses of Objects: Correct responses .03 -.06 Sum of errors -.20 -.16 Stamps task: Complexity score -.11 .05 Originality score -.08 -.07 *p < .05; ** p < .01; *** p < .001. Results of these correlations are displayed in Table 31. Siblings’ false belief aggregate
score and ToL adjusted extra moves score both correlated significantly with both the
PIQ and VIQ of probands, and the RIL task inhibition error difference score correlated
significantly with proband VIQ. However, when sibling PIQ and VIQ were controlled,
245
only the correlation between proband PIQ and siblings’ false belief aggregate score
remained significant (this correlation also remained significant when ASD siblings
meeting full or partial ADI-R criteria were excluded). None of the variables on which
significant differences between sibling groups were observed correlated significantly
with either proband PIQ or VIQ, suggesting that the non-matching of the autistic
probands of Study Two’s participants did not affect the outcome of group comparisons
in this study. Overall, these results indicate that the ToM performance of siblings of
autistic individuals was related to the level of functioning of probands, but EF variables
were not.
6.3.4.2 Correlations between probands’ and siblings’ cognitive
performances
Correlations between probands’ and siblings’ cognitive task performances were limited
to the sample of siblings whose autistic brother or sister participated in Study One.
Only correlations between identical task variables for each family member were
examined (rather than all correlations between different tasks). Surprisingly, there were
no significant raw correlations between probands’ and siblings’ scores on any variable
(therefore, no partial correlations were conducted). This suggests that ToM and EF
performances were not strongly familial.
6.3.5 Prevalence of deficits in ASD siblings
The prevalence of deficits in ASD siblings on the two potential endophenotype
variables was calculated in the same way as the universality of deficits in Study One.
The proportion of ASD siblings scoring below the 16th percentile of control siblings (or
above the 74th percentile in the case of the error variable) was 24.1% on the RIL task
shape error score, and 19.6% on the Stamps task complexity score. Therefore,
impairments on these variables clearly only occurred in a subset of ASD siblings.
6.3.6 Correlations between ToM and EF
Although no ToM deficit was identified in the ASD sibling group as compared with
control siblings, suggesting that the notion of a “ToM-impaired, EF-intact” subgroup
was somewhat invalid, correlations between ToM and EF were still of interest to
246
determine whether ASD siblings showed an unusual pattern of association between the
two domains (which may suggest that they used different strategies to solve the tasks,
even if no performance decrement was observed). Correlations between ToM and EF
variables were calculated separately for the ASD and control sibling groups. As in
Study One, partial correlations (controlling for the effects of age, VIQ and PIQ) were
also conducted if significant raw correlations were observed.
Table 32 presents the raw and relevant partial correlations between ToM and EF
task variables within control siblings. As for Study One, correlations are displayed
separately for the various false belief variables rather than the overall aggregate score
because the pattern of correlations was different for the three tasks. In the control
sibling group, simple false belief task performance correlated with measures of planning
and verbal generativity (with all correlations in the expected direction, such that poor
false belief performance correlated with poor EF task performance); however when age,
VIQ and PIQ were controlled for, none of these correlations remained significant. First-
order false belief task performance correlated with measures of planning, non-verbal
inhibition (with working memory load), working memory (with inhibition
requirements), verbal inhibition, and both verbal and non-verbal generativity (all in the
expected direction with the exception of the RIL task load error difference score).
However, when age and IQ variables were partialled out, correlations remained
significant only with measures of planning, verbal inhibition, and non-verbal
generativity. Second-order false belief task performance correlated with measures of
planning and verbal and non-verbal generativity (all in the expected direction); with
age, PIQ and VIQ controlled, there were no significant correlations with planning
measures.
Overall, in the control group, ToM variables demonstrated significant
relationships with all EF domains measured except for set-shifting, but several of the
correlations were mediated by age and IQ (there were no significant partial correlations
with non-verbal inhibition measures). All correlations were in the expected direction,
such that poorer performance on EF tasks correlated with poorer false belief task
performance. Ceiling effects on the simple false belief task resulted in a paucity of
significant correlations with EF variables for that task.
247
Table 32. Raw and partial correlations between ToM and EF variables within control siblings False belief task EF task Simple First-order Second-order ToL (n = 61): Adj.extra move score -.26* -.01 -.61*** -.32* -.55*** -.21 Rule violations -.01 -.49*** -.22 -.37** -.05 IDED Set-shifting Perseveration condition (n = 42): EDS stage errors a -.17 .01 IDED Set-shifting Learned Irrelevance condition (n = 42): EDS stage errors a -.01 -.01 RIL task (n = 41): Error difference scores: Inhibition a -.13 -.09 Load a .32* .31 -.07 Inhibition + load a .18 -.15 RT difference scores: Inhibition a .09 .15 Load a .22 -.01 Inhibition + load a .20 .09 Shape error score a -.35* -.21 -.13 Opposite Worlds (n = 43): Error diff. score a -.34* -.42** -.13 Time diff. score a -.38* -.32* -.12 Pattern Meanings (n = 61): Correct responses .14 .20 .25* .28* Sum of errors -.15 -.34** -.03 -.30* .02 Uses of Objects (n = 61): Correct responses .31* .17 .36** .13 .52*** .34** Sum of errors -.02 -.23 -.19 Stamps task (n = 61): Complexity score .23 .45*** .33* .41** .26* Originality score .06 .28* .07 .24 Restriction score -.16 -.03 -.05 Rule adherence score -.25 -.16 -.19 * p < .05; ** p < .01; *** p < .001. Note: Partial correlations controlled for age, VIQ and PIQ. All tests were two-tailed. Ns listed for each task show the sample size for correlations with the ToM tasks. a = No correlation could be calculated as all participants had perfect scores on the false belief task.
248
For ASD siblings, correlations were conducted both before and after excluding
individuals meeting ADI-R criteria for an ASD. For the sake of brevity, only
correlations after exclusion of this subgroup are reported, as priority was given to
determining the pattern characteristic of the broad phenotype without any siblings with
ASDs in the sample3. Table 33 displays these raw and partial correlations. In this
group, simple false belief task performance correlated with two measures of non-verbal
generativity (with both correlations in the expected direction), both of which remained
significant when age and IQ variables were partialled out. First-order false belief task
performance correlated with variables from all EF domains tested except for set-shifting
and verbal generativity (all in the expected direction), but only the correlations with one
planning measure, non-verbal inhibition (with a working memory load), and one non-
verbal generativity measure were significant when age and IQ were controlled. Second-
order false belief task performance correlated with measures of planning, non-verbal
inhibition, and verbal and non-verbal generativity (all in the expected direction), but
only one correlation with a non-verbal inhibition measure remained significant when
age and ability variables were partialled out.
Overall, the ASD siblings showed a fairly similar pattern of raw correlations as
the control siblings. However, partial correlations showed a different pattern from
controls, with the ASD siblings showing no significant partial correlations between
ToM measures and verbal inhibition or verbal generativity variables, but demonstrating
significant partial correlations of ToM variables with measures of non-verbal inhibition.
Like control siblings, ASD siblings also showed significant partial correlations between
ToM variables and measures of planning and non-verbal generativity.
Table 34 presents a summary of the significant partial correlations between ToM
and EF domains in the control and ASD sibling groups as well as the control and ASD
proband groups from Study One. The most striking aspect of this table is the clear
relative absence of significant ToM-EF correlations in ASD probands compared with all
other groups. It also shows that while the pattern of correlations displayed by ASD
siblings did not mirror that demonstrated by the ASD probands, it was also qualitatively
different to the pattern displayed by control siblings. Nevertheless, it is additionally
evident that the control groups from both studies did not show identical patterns of
correlation (this is discussed further in Section 6.4.2).
3 When siblings with ASDs were included, the pattern of correlations was similar.
249
Table 33. Raw and partial correlations between ToM and EF variables within ASD siblings False belief task EF task Simple First-order Second-order ToL (n = 78): Adj.extra move score -.15 -.29** -.04 -.34** -.17 Rule violations -.20 -.24* -.23* -.13 IDED Set-shifting Perseveration condition (n = 58): EDS stage errors a .16 -.04 IDED Set-shifting Learned Irrelevance condition (n = 59): EDS stage errors a .01 .15 RIL task (n = 57): Error difference scores: Inhibition a .0 -.34** -.31* Load a -.41** -.42** .11 Inhibition + load a -.42** -.40** -.20 RT difference scores: Inhibition a -.07 -.02 Load a .20 .04 Inhibition + load a .09 .02 Shape error score a -.32* -.07 -.19 Opposite Worlds (n = 45): Error diff. score a -.09 -.19 Time diff. score a -.29* -.09 -.25 Pattern Meanings (n = 79): Correct responses .14 -.02 -.01 Sum of errors -.09 -.21 -.15 Uses of Objects (n = 79): Correct responses .15 .21 .30** .08 Sum of errors -.10 -.19 -.08 Stamps task (n = 76): Complexity score .33** .24* .38** .30* .24* .09 Originality score .22 .30** .15 .32** .16 Restriction score .05 -.10 -.16 Rule adherence score -.35** -.28* -.28* -.22 -.08 * p < .05; ** p < .01; *** p < .001. Note: Partial correlations controlled for age, VIQ and PIQ. All tests were two-tailed. Ns listed for each task show the sample size for correlations with the ToM tasks. a = No correlation could be calculated as all participants had perfect scores on the false belief task.
250
Table 34. Summary of partial correlations between ToM and EF variables in the
control and ASD probands and siblings
ToM
EF domain Control
siblings
ASD
siblings
Control
probands
ASD
probands
Planning
Set-shifting
Inhibition – Non-verbal *
Inhibition – Verbal
Working Memory
Generativity – Verbal
Generativity – Non-verbal
* Correlations marked with an asterisk were in the opposite direction than expected.
Note: Each tick represents one significant correlation between a false belief and an EF
variable in that domain.
6.3.7 Dissociations between ToM and EF
The presence of ToM-EF dissociations in the ASD sibling group was assessed in the
same way as for Study One. It should again be noted that because ToM was not
impaired in ASD siblings relative to the control siblings, the presence of any “ToM-
impaired, EF-intact” dissociations is somewhat misleading in that ToM was not
impaired in the group as a whole. However, these calculations were still of interest as it
may have been the case that those siblings showing EF deficits were also more likely to
be low scorers on ToM tasks. As in Study One, the false belief alternative aggregate
score was used as the measure of ToM performance (14.8% of control siblings were low
scorers on this variable, indicating that defining ASD siblings with low scores as
“impaired” was comparable with the definition of impairment for continuous variables).
The two candidate endophenotype EF variables were analysed separately. The results
of these calculations are displayed in Table 35, and demonstrate that ToM and EF
impairments did not always co-occur in the same ASD siblings. Rather, the EF-
impaired siblings were equally or more likely to show intact ToM than impaired ToM.
However, it is also notable that both the ToM-impaired and the EF-impaired siblings
were more likely than the sibling group as a whole to demonstrate impairments in the
other domain (e.g., 55.6% of ASD siblings showing impaired non-verbal generativity
251
also scored poorly on the false belief aggregate, compared with 19.5% of the ASD
sibling group as a whole).
Table 35. The incidence of ToM-EF dissociations in the ASD siblings
EF measure
% of ToM-impaired ASD siblings
with unimpaired EF
N
RIL task shape error score 50.0 4
Stamps task complexity score 52.9 9
% of EF-impaired ASD siblings
with unimpaired ToM
RIL task shape error score 88.9 18
Stamps task complexity score 44.4 18
6.3.8 Results from behavioural measures
Both the SBQ (which measures current social behaviours) and the RBQ (which
measures the lifetime presence of repetitive behaviours) were completed by parents of
siblings (both are described in Chapter 3, Section 3.5). The ASQ was not used as a
measure of subclinical behavioural traits as it was not considered valid for use in this
way, being designed to discriminate individuals with autism from typically developing
individuals rather than measure the severity of any autistic-like symptomatology in
typically developing individuals. As the RBQ was originally intended only as a
screening measure (not enough siblings demonstrated an adequate number of repetitive
behaviours for analyses using the RBI to be meaningful), only the overall sum was used
rather than composite scores for each behavioural category.
Comparisons between ASD and control siblings on the SBQ and RBQ were
conducted to assess whether there was an increased incidence of behavioural
symptomatology in ASD siblings. These comparisons were conducted both with the
overall group and with siblings meeting full or partial ADI-R criteria excluded. The
summary variables from both measures were highly skewed and were not amenable to
transformation, so non-parametric tests (Mann-Whitney U) were used for group
comparisons. Two ASD siblings and three control siblings had missing data on all three
questionnaires and were not included in analyses.
252
On the SBQ overall sum, there was no significant difference between ASD and
control siblings, U = 3176.5, N1 = 106, N2 = 64, p > .1, and the difference remained
non-significant when siblings meeting full or partial ADI-R criteria were excluded, U =
2710.0, N1 = 97, N2 = 64, p > .1. On the RBQ, however, there was a trend for parents
to report more repetitive behaviours in control siblings than in ASD siblings, U =
2818.0, N1 = 106, N2 = 64, p = .06, and this difference became significant when
siblings meeting ADI-R criteria for an ASD were excluded, U = 2320.5, N1 = 97, N2 =
64, p < .01. One explanation for these unexpected findings4 could be that parents of a
child with autism were more likely to under-report autistic-like symptomatology in their
non-autistic children, as their benchmark for comparison (e.g., what might be
considered to be repetitive use of language) was set much higher. This seems more
likely than ASD siblings actually displaying less behavioural symptomatology than
control siblings, given previous research on the broad behavioural phenotype (see
Chapter 5, Section 5.2.1). Of note, ASD siblings meeting full or partial ADI-R criteria
scored significantly higher than remaining ASD siblings on both the SBQ and RBQ, as
would be expected, which suggests that these measures successfully discriminated
individuals with ASD diagnoses from those without ASD diagnoses, but were not
accurate measures of symptom severity in individuals without ASD diagnoses.
While it was initially intended to use data from these two questionnaire
measures to examine correlations between cognitive and behavioural measures within
the ASD sibling group, the outcomes of these group comparisons suggest that the
behavioural data are not likely to be a valid indicator of behavioural severity, making
correlations difficult to interpret. When these correlations were conducted with the
overall ASD sibling sample, there were a number of significant correlations between
both ToM and EF variables and behavioural measures (most of which remained
significant when age, PIQ, and VIQ were partialled out), but when siblings meeting full
or partial ADI-R criteria were excluded, many of these correlations became non-
significant. Therefore, because of concerns about their interpretation, these correlations
are not reported.
4 When results from the ASQ were analysed, a similar pattern emerged: when siblings with ASD diagnoses were excluded, it was found that control siblings scored more highly than ASD siblings in both the communication and interests domains as well as overall.
253
6.3 Discussion
6.4.1 Endophenotype status of ToM and EF impairments
In the introduction to this chapter, six features of endophenotypes were described, with
criteria 4, 5, and 6 being tested in the current study. Did any ToM or EF variables meet
these three criteria?
i) Criterion 4. The first criterion tested in this study was that the
endophenotype should be found in siblings of probands with ASDs at a higher rate than
in the general population (or in this case, a higher rate than siblings of control
probands). Group comparisons between siblings of individuals with ASDs and
siblings of controls revealed few significant differences. There were no significant
group differences on measures of ToM, planning, set-shifting, non-verbal inhibition, or
verbal generativity. However, weaknesses in working memory (within an inhibition
task) and non-verbal generativity emerged as the two main candidate endophenotypes.
Importantly, group differences on these variables remained significant when siblings
with ASDs were excluded. Although the group difference in non-verbal generativity
became only marginally significant when individuals with non-ASD diagnoses were
excluded, this does not necessarily mean that a non-verbal generativity deficit in the
broad phenotype is an unimportant artefact of pathology unrelated to ASDs, but may
reflect the possibility that the broad phenotype is itself characterised by higher rates of
non-ASD diagnoses and these individuals also display a more abnormal cognitive
profile5.
Sisters of ASD probands also performed significantly more poorly than control
sisters on the simple first stage of the IDED set-shifting task (in the Learned Irrelevance
condition). As there were no significant differences observed in latter stages of the task
which involve shifting set, this difference probably reflects either attentional or
motivational differences rather than a deficit in a component of EF. As previously
stated, the fact that ASD probands did not display a deficit on this simple first stage
variable further indicates that it is not likely to represent a useful endophenotype for
ASDs. It is unclear why the difference occurred only in sisters and only in one of the
task conditions, but it may have been that ASD sisters were more prone to fatigue (the
Learned Irrelevance condition was administered after the Perseveration condition).
5 Of the ASD sibling group, 9.3% had a non-ASD diagnosis compared with 4.5% of control siblings.
254
There were no other significant interactions between group and gender, suggesting that
brothers and sisters of ASD probands were equally susceptible to any inherited
cognitive weaknesses.
Unexpectedly, control siblings showed significantly larger error and time
difference scores on the Opposite Worlds task, which is superficially suggestive of a
strength in verbal inhibition in ASD siblings. However, this difference resulted from a
non-significant weakness in ASD siblings in the control condition combined with a non-
significant weakness in control siblings in the inhibition condition. As the control
siblings were not actually significantly poorer in the inhibition condition, it would be
difficult to argue that the result reflects a strength in inhibition in the ASD siblings, and
it appears more likely that it was a rather spurious outcome of two non-meaningful but
additive differences.
Overall, then, the results relevant to criterion 4 were consistent with the
prediction, based on the results of Study One, that EF deficits would be more likely to
demonstrate superior relative primacy over a ToM deficit as measured by their presence
in siblings of individuals with ASDs. Non-autistic siblings of ASD probands exhibited
a broad cognitive phenotype characterised by weaknesses in the non-verbal generation
of novel ideas and working memory performance in situations combining working
memory and inhibitory requirements, but no impairment in ToM. However, there are a
number of caveats and additions to this conclusion. Firstly, the sensitivity of the ToM
tasks used in this study may not have been sufficient to detect subtle weaknesses in
mentalising abilities (although while ceiling effects were observed on the simple false
belief task, there was a significant proportion of both ASD and control siblings who
showed unstable performance on the other tasks). This limitation is particularly
pertinent given the wide age range of participants in this study, which was larger than in
Study 1. The use of more advanced and naturalistic ToM tasks would strengthen future
broad phenotype studies (previous findings using higher-level ToM tasks are discussed
further below). Nevertheless, it is apparent that the broad autism phenotype is not
characterised by a significant impairment in basic ToM abilities. Secondly, the
prediction that the EF variables which showed the strongest evidence of primacy in
Study One (i.e., verbal inhibition and verbal generativity) would be the most likely to
emerge as endophenotypes in siblings was not borne out. ASD siblings did not show
impairments in either verbal inhibition or verbal generativity (or in planning, which was
also found to be impaired in probands). These negative findings call into question the
primacy of those EF domains to ASDs – although alternatively, it is possible either that
255
i) the tasks used in these domains were also lacking in sensitivity, or ii) impairments in
these domains as well as in ToM always result in an ASD phenotype, and therefore are
not seen in a milder form in unaffected relatives. Thirdly, only non-verbal generativity
performance significantly predicted membership in the ASD sibling group (although the
RIL task shape error score also became a marginally significant predictor when age and
IQ variables were included in the regression); and while the two variables together
successfully predicted membership of the ASD sibling group in 90.8% of cases, they
also misclassified 64.3% of control siblings. Hence, their utility as endophenotypes is
limited by their poor uniqueness or specificity as markers of genetic vulnerability.
How do the results of sibling group comparisons compare with previous studies?
As reviewed in Section 5.2.2.2 of Chapter 5, studies on cognitive deficits in siblings of
autistic probands have generally found fewer and smaller differences than studies with
parents, and in that sense this study is consistent with other sibling studies6. Only one
previous study has employed similar false belief tasks with siblings of children with
autism (Ozonoff et al., 1993), which also found no evidence of mentalising deficits.
However, the ToM tasks used in both Ozonoff et al.’s and the current study may not
have been difficult or high-level enough to detect subtle weaknesses, especially in older
siblings. Dorris et al.’s (2004) finding of impaired performance on the higher-level
Eyes Task in siblings of children with Asperger syndrome suggests that ToM
difficulties may indeed be revealed if more advanced ToM tasks are used, although the
“purity” and validity of the Eyes task as a measure of ToM is questionable (Dorris et al.,
2004).
EF in siblings of probands with ASDs has been investigated in two previous
studies (Hughes et al., 1999; Ozonoff et al., 1993). Neither the interaction between
working memory and inhibition or non-verbal generativity were tested in these two
studies, so the positive results in these domains in the current study are new findings.
Unlike this study, planning difficulties in ASD relatives were reported in both of the
two previous studies, and Hughes et al. (1999) also found evidence of weaknesses in
set-shifting and verbal generativity in their ASD sibling sample. The sample size was
much larger for the current study than for either of these two previous studies (108
siblings of probands with ASDs compared with 18 in Ozonoff et al. and 31 in Hughes et
al.), ruling out power as an explanation for these discrepancies. As was the case in this
study, Ozonoff et al. also included siblings of probands with other ASDs besides
6 Moreover, the parents of probands with ASDs who were tested as part of the WAFSASD did show more EF deficits than the siblings (Wong, Maybery, Bishop, Maley, & Hallmayer, in preparation).
256
autism, however Hughes et al. included only siblings of probands with a full diagnosis
of autism, and the mean IQ of the probands in that study was also lower than in this
study (even though siblings of the lower-functioning probands in WAFSASD were
included, the mean PIQ and VIQ of the probands was still higher in this study than in
Hughes et al.’s study). It is possible that the broad cognitive phenotype is expressed
more strongly when the proband has a full autism diagnosis and is lower-functioning,
therefore explaining the increased incidence of planning, set-shifting and verbal
generativity deficits in the siblings in that study. However, contrary to this explanation,
there were no significant partial correlations between proband IQ and sibling
performance on planning, set-shifting or verbal generativity tasks in this study.
Another possible explanation for the discrepancies in the EF results obtained
across sibling studies is that the tasks employed differed in such a way as to favour the
siblings in this study. As discussed in Section 4.4.1 of Chapter 4, the administration of
the ToL in this study differed slightly from most other studies in that forward planning
was actively encouraged, which may have bolstered any weaknesses. While Hughes et
al. (1999) used the original version of the IDED set-shifting task which had also been
used to demonstrate set-shifting deficits in probands (Hughes et al., 1994), the modified
version used in this study did not reveal deficits in probands either in Study One of this
research or in high-functioning probands in previous research (Turner, 1997), making it
unsurprising that siblings demonstrated intact performance on it in this study. It is also
interesting, however, that Ozonoff et al. (1993) did not find any evidence of a deficit in
cognitive flexibility in siblings using the WCST. The lack of any significant differences
on the verbal generativity tasks used in this study was a little more surprising, as Turner
(1999) found that autistic probands actually showed more striking deficits on ideational
fluency tasks (used in this study) than on the word fluency task used by Hughes et al.
(1999). However, while only 90s were allowed to generate responses in this study,
Hughes et al. allowed 120s, which may have been the extra time needed to reveal
generativity difficulties in siblings of autistic probands.
Of additional note, Hughes et al. (1999) did not find overall group differences in
their sibling groups on either of their continuous measures of planning or set-shifting,
but found differences only when particular variables were dichotomised and the
proportions of siblings classified as passers or failers were compared. In this study, set-
shifting variables were already dichotomised (although using a different criterion from
257
Hughes7), but no differences in the proportion of poor performers were found. When
the current ToL data were re-analysed using a dichotomous pass/fail performance
criterion (such that a “failer” was anyone who completed less than 50% of the problems
in the minimum number of moves), again no group differences were revealed. It
therefore does not appear that existing group differences were merely “hidden” by the
methods of analysis used in this study; furthermore, the fact that Hughes et al. were only
able to find significant group differences on certain variables after dichotomising their
data suggests that the deficits observed lacked robustness and were not highly prevalent.
ii) Criterion 5. The familiality of ToM and EF abilities was assessed by
calculating correlations between the cognitive functioning of ASD siblings and the IQ
and cognitive functioning of ASD probands. The only significant relationship to
emerge (after the mediating effect of sibling IQ was controlled) was between proband
PIQ and siblings’ false belief performance, indicating that siblings were more likely to
perform poorly on false belief tasks if the proband with autism was low-functioning
(non-verbally). Interestingly, this implies a small degree of familiality of ToM
performance in ASD siblings, even though they did not display evidence of a ToM
impairment and the correlation between proband and sibling false belief performance
was not itself significant. There were no significant relationships between siblings’ EF
performances and proband IQ or EF performances. This suggests that siblings’ EF
abilities were not strongly familial, even on the tasks on which they displayed
significant weaknesses. Hence, the results did not support the prediction that EF
performances would be more likely to show evidence of familiality than ToM
performance. Impairments in non-verbal generativity and working memory (in an
inhibitory context), which were identified as potential endophenotypes on the basis of
sibling group comparisons, did not meet the criterion of heritability.
Do these results indicate that autism is not a genetic disorder, that there is no
broad autism phenotype, or that the cognitive impairments displayed by ASD siblings
were random and unrelated to genetic vulnerability? There are several alternative
explanations. It is possible that i) different genetic factors underpin the variation found
in siblings than in probands - for example, many genes are known to cause mental
retardation, but these genes do not influence the normal variation of IQ; ii) the measures
used lacked sensitivity to the milder deficits displayed by siblings, thereby weakening
7 When Hughes’ criterion was used (i.e., achieving six consecutive correct responses on the EDS stage), there were still no significant group differences, although on this task version there were ceiling effects using this method (i.e., the large majority of siblings achieved criterion in both conditions).
258
proband-sibling correlations; or iii) non-shared environmental factors contributed a
significant amount of variance, therefore reducing the size of correlations. It is also of
note that even under a perfect model of a monogenic disorder, the correlation between
siblings would only be 0.5. The general lack of significant proband-sibling
relationships in this study is consistent with previous studies by Piven et al. (1990) and
Szatmari et al. (1993), both of which found no association between proband IQ and the
cognitive functioning of first-degree relatives. No previous studies have reported direct
correlations between probands’ and siblings’ performances on specific cognitive tasks.
iii) Criterion 6. The notion that any endophenotype displayed in nonaffected
family members would be less severe than in affected probands was not so much a
criterion, but rather an expected feature of endophenotypes. The comparison of effect
sizes of significant differences in this study with the proband differences in Study One
confirmed this expectation, with smaller effect sizes displayed for both candidate
endophenotype variables in this study. However, it should be noted that the smaller
effect sizes could have been caused by a smaller proportion of siblings than probands
showing a deficit, rather than the severity of the deficit being milder across the sibling
sample. It was not possible to directly compare the performances of the probands and
siblings, as they were not matched on PIQ or VIQ8. The proportion of siblings showing
a deficit is discussed further below, in Section 6.4.2.
Concluding comments on endophenotype status. In summary, weaknesses in
working memory (in a context where inhibition was required) and non-verbal
generativity were identified in ASD siblings when compared with control siblings,
making these two variables candidates for endophenotypes of ASDs (ASD sisters also
showed poorer performance on a simple discrimination stage of the IDED set-shifting
task, but this was ruled out as a potential endophenotype because probands did not show
a deficit on that variable). However, while these two variables also met criterion 6 (i.e.,
deficits in those domains were less severe in siblings than in probands), they failed to
meet criterion 5 (heritability). They also lacked specificity as predictors of ASD sibling
group membership (misclassifying a high proportion of control siblings), and they did
not show the strongest evidence of primacy in Study One. Therefore, the evidence for
their validity and utility as endophenotypes for ASDs was not strong or consistent.
8 It was possible to compare the performances of only those probands and siblings who showed a deficit (as defined by a score worse than 1 SD from the mean) on the RIL task shape error score, as the proband and sibling samples were matched on age, PIQ, and VIQ in this case. Consistent with expectation, the ASD siblings showed significantly better performance than ASD probands on that variable when z-scores were compared, t(34) = 2.76, p < .01.
259
Overall, while EF deficits demonstrated superior relative primacy than a ToM deficit,
neither ToM nor EF showed convincing evidence of their primacy in this study (these
results are fairly consistent with Study One, but were even less compelling). This
outcome occurs in the context of frequent inconsistencies across sibling studies in the
autism field and between studies of probands, siblings, and parents, and probably
reflects the likelihood that genotype-phenotype relationships in the autism spectrum are
complex and indirect, even when the phenotype is at the level of cognition (this is
discussed further in Chapter 7). Nevertheless, further studies incorporating higher-
level, more sensitive tasks may still prove useful in identifying more subtle weaknesses
in family members.
6.4.2 Differentiating the multiple deficits models
The “subgroups” and “multidimensional spectrum” versions of the multiple primary
deficits model of ASDs were both indirectly examined in this study. The lack of any
ToM impairment in ASD siblings was in itself inconsistent with both of these models,
indicating that the notion of a ToM-impaired sibling subgroup or the idea that a ToM
impairment was the basis for certain types of subclinical symptomatology both lacked
support. The prediction that any cognitive deficits would only occur in a subset of ASD
siblings was confirmed, however, with impairments in working memory (in an
inhibition context) and non-verbal generativity being demonstrated by 24.1% and 19.6%
of siblings respectively. This is consistent with previous studies of cognitive abilities in
relatives of individuals with ASDs, and suggests that endophenotypes or markers of
genetic vulnerability are only expressed in a certain subgroup of relatives (although, it is
possible that impairments in other domains not measured here may turn out to be more
prevalent among relatives).
The results of analyses examining the presence of ToM-EF dissociations further
suggested that there may be more than one of these subgroups, each with a different
cognitive profile. While impairment in ToM did not occur with a significantly greater
frequency in ASD siblings than in control siblings, those ASD siblings who did display
an unstable ToM were also more likely to show EF deficits than the ASD sibling group
as a whole (and vice versa, those with a non-verbal generativity deficit were also more
likely to show unstable ToM performance). However, there was also a high frequency
of ToM-EF dissociations in both directions. This pattern of results suggests that within
the subgroup of relatives expressing the endophenotype, there were two further
260
subgroups: an “EF-impaired, ToM-intact” group and a “both ToM and EF impaired”
group. However, it is also possible that these do not represent valid subgroups, but
instead that ToM and EF performance varied on a more continuous spectrum (with the
two spectrums covarying to some degree), and only some siblings fell on the “impaired”
side of the arbitrary cutoff for impairment. It would be interesting to test the validity of
the various sibling “subgroups” by investigating whether there are systematic
differences between their genotypes (or, perhaps, gene-environment interactions; see
Bauminger & Yirmiya, 2001).
Correlations between ToM and EF in siblings of individuals with ASDs were
also investigated in this study, although given the lack of ToM impairment this set of
analyses did not really address the “subgroups”-driven idea that ToM and EF
impairments may be independent in ASD siblings (this would not be expected, given
that the ASD siblings did not display a ToM deficit and the account outlined in Study 1
proposes that the relative independence between ToM and EF in probands occurs partly
because their ToM deficit is caused by ToM-specific factors). Instead, the results of
ToM-EF correlations were more relevant to the ancillary issue of whether ASD siblings
showed unusual patterns of association between the two domains. Results demonstrated
that, compared to control siblings, ASD siblings did show a different pattern of
associations between ToM and EF performances. Unlike control siblings, ASD siblings
showed no significant partial correlations between measures of ToM and verbal
inhibition or verbal generativity, but showed three significant partial correlations
between ToM and non-verbal inhibition measures. Hughes et al. (1999) found similar
evidence of unusual associations between tasks for ASD siblings (although the
correlations were between EF tasks in that study). This is consistent with the notion
that ASD siblings may use different strategies to solve cognitive tasks than siblings of
children without autism, even though their overall level of performance may not be
impaired. Nevertheless, it must be noted that control siblings and control probands also
showed a different pattern of correlations between ToM and EF. One explanation for
this may be that the siblings in this study were older, on average, than the probands in
Study One and the age range was larger. If the hypothesis that the ToM-EF relationship
changes with development is correct, these differences in correlations across the two
control samples would be expected. As the ASD and control sibling groups were
matched on age and similar in age range, the difference in the pattern of correlations
between these two groups more meaningfully suggests that ASD siblings are
characterised by unusual associations between ToM and EF. However, it is also
261
possible that the different patterns of ToM-EF correlations displayed in different control
samples could simply reflect the fact that ToM-EF relationships are weak or even
spurious in some cases, resulting in variable outcomes in different samples. The
hypothesis that siblings (and probands) use unconventional strategies to solve ToM (or
EF) tasks would be better addressed by systematically varying the problem-solving
demands of ToM tasks and observing the differential effects of these manipulations in
ASD and control samples.
It was initially intended to examine the “multidimensional spectrum” idea in this
study by calculating correlations between cognitive and behavioural measures within
the ASD sibling group. Unfortunately, the behavioural measures of social impairment
and repetitive behaviours did not appear to be valid indicators of behavioural severity in
siblings without ASD diagnoses, with the higher levels of symptoms reported in control
siblings probably reflecting a tendency for parents of children with ASDs to under-
report subtle autistic-like symptomatology in non-autistic siblings (e.g., the parent of a
child with autism may answer the question “Does your child pace or move around
repetitively?” negatively with regard to their non-autistic child as compared with their
child with autism, whereas a control parent may answer positively because many
children display occasional restlessness). This meant that correlations between
cognitive and behavioural measures could not be calculated, as the behavioural
measures lacked validity. It was interesting, however, that this under-reporting was not
evident on the measure of social behaviour, which may be an indication of an increased
incidence of social impairment in ASD siblings (although this is highly speculative).
Future investigations of cognitive-behavioural relationships in relatives of individuals
with ASDs may benefit from the employment of observational measures of behaviour or
other more direct measures which do not rely on parental report (this is discussed
further in Chapter 7).
In sum, then, the results from this study were not able to contribute as much as
was hoped to the question of which multiple deficits model may be the most appropriate
for ASDs. Indeed, the results relevant to endophenotype status did not identify multiple
deficits in siblings (i.e., deficits in both ToM and EF domains). It was nevertheless
evident that the two EF deficits showing the most promise as endophenotypes only
characterised a subgroup of ASD siblings, and that the presence of an additional ToM
deficit may represent a more subsidiary subgroup. However, it was not possible to
determine whether these represented valid subgroups (as opposed to ends of a spectrum)
or whether they were also associated with increased levels of subclinical behavioural
262
symptomatology. The role of ToM and EF in ASDs therefore remains a question with a
somewhat nebulous answer.
263
264
CHAPTER 7
General Discussion: Constructing an Explanatory Model for ASDs
7.1 Summary of the findings
7.2 Methodological strengths and limitations
7.3 Conclusions on constructing an explanatory model for ASDs
7.4 Future directions
265
7.1 Summary of the findings
The major findings of this research may be summarised as follows:
1. Individuals with ASDs demonstrated a profile of spared and impaired cognitive
abilities which differed in important ways from previous research. They showed
impairments in ToM, planning, verbal inhibition, working memory (when inhibitory
control was also required), and both verbal and non-verbal generativity, but intact
performance on tests of awareness of social norms, set-shifting, non-verbal
inhibition and relational reasoning. Deficits on verbal tasks were more common
than on non-verbal tasks, and several task performances were mediated by VIQ.
The deficits in verbal inhibition and in working memory in an inhibitory context
were new findings which suggested that the previously proposed “typical EF
profile” of individuals with ASDs, in which inhibition is spared (e.g., Ozonoff &
Jensen, 1999), may need revision.
2. Results did not support a single primary cognitive deficit model of ASDs. Neither
ToM nor EF deficits met the criteria of universality or explanatory value. This
confirms and supports the findings of several previous studies which have
demonstrated similar outcomes (as described in Chapter 2).
3. However, EF deficits showed superior relative primacy compared with a ToM
deficit, as judged by their superior ability to discriminate individuals with ASDs
from controls, and the higher number of significant correlations with aspects of
behavioural symptomatology. In particular, deficits in verbal inhibition and verbal
generativity appeared to be the most primary.
4. ToM and EF were found to be largely independent deficits in ASDs, as measured by
the paucity of significant correlations between the two domains and the
dissociability of the impairments in both directions. This was the first study to
demonstrate this independence of the two deficits in ASDs. It indicated that
although EF deficits were relatively more primary, they could not explain or
subsume ToM as a secondary deficit. These findings were also inconsistent with
both the “common conceptual bases” and “emergence” accounts of the ToM-EF
relationship in typical development, thereby providing the most support for either
the “common neuroanatomical bases” and/or “expression” accounts.
5. No ToM or EF variables demonstrated strong or consistent potential as
endophenotypes for ASDs, although EF deficits showed better potential than a ToM
266
deficit. Weaknesses in working memory (in a context where inhibition was
required) and in non-verbal generativity were identified in ASD siblings when
compared with control siblings; however, performance in these domains was not
strongly familial and the variables lacked specificity as predictors of ASD sibling
versus control group membership.
Hence, in sum, ToM and EF were found to be independently impaired in ASDs, but
neither impairment was universal, showed strong relationships with symptoms, or was a
useful candidate for an endophenotype for ASDs. EF deficits consistently showed
superior primacy in comparison with ToM. The results of both studies indicated that a
multiple primary deficit model is more suitable for ASDs than a single primary deficit
model, but it was not possible to determine which type of multiple deficits model was
the most appropriate. There appeared to be different subgroups of both ASD probands
and ASD siblings demonstrating different cognitive profiles, but results were also
compatible with the notion of a more continuous spectrum (with the apparent subgroups
an artefact of the arbitrary cutoff for the definition of “impairment”). The way in which
these subgroups or spectrums should be defined behaviourally was also unclear,
although results were more consistent with a classification system based on level of
functioning as opposed to symptom domains or symptom severity. The possibility that
the primacy of deficits changes with development also remains open, and there may be
other equally primary deficits which were not measured in this research.
7.2 Methodological strengths and limitations
As described in Chapters 4 and 6, the current research incorporated several
methodological improvements upon previous studies. One of the major strengths was
the use of relatively process-pure tests of a range of EF components, several of which
included in-built control conditions allowing isolation of the relevant ability. Tests
requiring both verbal and non-verbal responses were used, and most tasks had several
levels of difficulty in order to be suitable for individuals of a wide range of ages. Large
sample size was also a significant strength of this research; for example, the number of
siblings who participated in Study Two was more than three times higher than the
number for the largest previous study on ToM or EF in siblings. Statistical approaches
were thorough and the effects of potentially confounding variables such as age and IQ
were carefully examined and accounted for throughout all analyses.
267
One of the methodological weaknesses of this research was the limited range of
ToM measures employed. While the Dewey Stories task was included as a higher-level
social cognition measure, it was of questionable validity as a measure of ToM. The
otherwise exclusive choice of false belief tasks was the result of i) the fact that theories
of the ToM-EF relationship have been based largely around false belief, ii) the need to
constrain the length of the test battery, and iii) the fact that the few high-level, advanced
ToM tasks that exist suffer from the problem of a lack of process purity (this is
discussed further in Section 7.2). Based on previous research (e.g., Baron-Cohen,
1989b), it was also expected that the failure rate of ASD probands on the false belief
tasks used (particularly the second-order tasks) would be higher than was found in this
research – with the current result indicating that a ToM deficit is not as severe or
prevalent in ASDs than some authors have argued (e.g., Baron-Cohen, 1995; Leslie &
Roth, 1993). The high success rates of control probands and both ASD and control
siblings on the false belief tasks caused difficulties for the interpretation of task results
as indicative of a lack of ToM impairment in ASD siblings or of a mild severity of
impairment in ASD probands. However, as discussed in Section 4.4.2 of Chapter 4, the
lack of discriminative ability (or “uniqueness”) and explanatory value of ToM in Study
One was not easily dismissable as a consequence of the level of difficulty of false belief
tasks, as i) ToM and EF deficits were of roughly equal prevalence in the ASD group, ii)
a significant proportion of individuals showed impaired performance on ToM tasks but
unimpaired performance on EF tasks, and iii) performance on all of the false belief tasks
was far from the ceiling in the ASD group – as confirmed by the significant medium-
level correlations between false belief performance and VIQ (which also suggests that
the measures have some reliability). In addition, although ToM performance in the
control group and in the two sibling groups was high, it was not at ceiling. The use of
an additional higher-level, more sensitive ToM task would nevertheless have
strengthened this research, particularly for the detection of any subtle weaknesses in
mentalising ability in ASD siblings, especially those in middle childhood and older.
Another possible limitation was the use of parental questionnaires and
interviews as indices of the presence and severity of behavioural symptomatology.
Parental report is subjective and dependent upon the individual parent’s framework for
judging abnormality. More direct and objective methods such as systematic behavioural
observation techniques may have provided more valid measures of behavioural
variation in individuals without ASDs and possibly resulted in stronger relationships
with underlying cognitive deficits. Bishop and Norbury (2002) found that diagnostic
268
measures of autism based on parental interview (the ADI-R) and direct observation (the
ADOS-G: Autism Diagnostic Observation Schedule – Generic; Lord et al., 2000)
resulted in widely discrepant outcomes for several children, and they noted that it is
usually recommended that information from parental report and observational
techniques be combined. However, observational methods have their own limitations,
such as time-intensiveness and the possibility of poor ecological validity (as behaviour
can only be observed for a limited time and in a restricted range of situations, and the
presence of an observer may alter the nature of the behaviour displayed). There are also
essentially no observational scales available for social/communicative functioning and
repetitive behaviours which are appropriate for recording both normal and abnormal
variation in these behavioural domains, rather than being directed at diagnosing
pathology.
While the characteristics of the ASD proband sample (e.g., age, level of
functioning, range of symptom severity) were not considered to be a major limitation as
the sample was generally appropriate for the research aims and the tasks used, it should
be recognised that the sample characteristics limit the scope of the conclusions.
Szatmari and colleagues have suggested that low-functioning autism may arise from
different genetic mechanisms from high-functioning autism (e.g., Szatmari, 1999;
Szatmari et al., 2002), therefore the high-functioning nature of the ASD probands in this
research may limit the generalisability of the findings. However, the inclusion of low-
functioning probands would have required substantial alterations to the design of the
study, as many of the tasks were inappropriate for individuals with mental retardation.
Similarly, the relatively old age of the sample limited the conclusions that could be
drawn particularly with regard to the possibility of changes in the primacy of and
relationship between ToM and EF impairments with development, but again the
inclusion of participants below the age of five years would have caused difficulties for
task selection. The non-matching of the ASD and control probands on VIQ was not
ideal, but this is a somewhat inevitable aspect of autism research given the typical PIQ-
VIQ discrepancy displayed by individuals with ASDs (making it difficult to match
controls on both PIQ and VIQ), and VIQ was taken into account in all analyses. The
analysis of ASD probands with different ASD diagnoses (e.g., autism, Asperger
syndrome) together as one sample may also be subject to criticism, but its validity was
attested by the finding that significant differences between probands meeting full ADI-
R criteria for autism and those meeting partial ADI-R criteria occurred only on one task.
Nevertheless, the variability of the sample is likely to have increased standard
269
deviations which may have reduced the likelihood of finding significant or large
differences from controls in group comparisons (marginal differences were found on
several tasks, and several significant differences were attenuated when age and/or IQ
variables were controlled).
Finally, although there were a small number of control probands with mild
mental retardation, the control proband group consisted largely of typically developing
individuals, matched to ASD probands on age and PIQ. Without having a control group
of individuals with other disabilities (e.g., Down’s syndrome), it is not possible to test
whether simply having any developmental disability may have resulted in some of the
deficits observed. This distinction is necessary for any deficit to meet the uniqueness
criterion for primacy, and it is also relevant for the interpretation of deficits in ASD
siblings, who may show adverse effects of living with a sibling with a disability
(although it is difficult to see why this would affect certain EF components and not
others; for further discussion, see Bauminger & Yirmiya, 2001).
7.3 Conclusions on constructing an explanatory model for ASDs
Taking into account these constraints, what broad conclusions about explanatory models
of ASDs can be made on the basis of this research? As already stated, we can fairly
confidently reject a conceptualisation of autism as a unitary syndrome with a single
primary cognitive deficit. Notwithstanding the possibility that there is another cognitive
deficit which was not measured in this research and which could explain both ToM and
EF deficits as secondary to it (which is unlikely as ToM and EF were found to be
unrelated impairments), the current findings clearly and consistently demonstrate that
ASDs can not be explained by a single primary deficit.
These findings consolidate recent research on cognitive impairments in ASDs,
which has increasingly moved away from the notion of a single primary deficit.
Psychologists studying cognitive deficits in ASDs have to some extent lagged behind
other researchers focussing on genetic and neurobiological aetiologies, who have been
arguing for some time that “any attempt to demonstrate a single cause for all cases of
autism appears to be futile” (Gillberg & Coleman, 1992, p. 283). This lag was driven
by the hope that the identification of a single cognitive deficit would provide a
diagnostic marker for autism and a unified explanation for the range of unusual
behaviours displayed by individuals with ASDs. However, it could be argued that the
failure to find such a cognitive marker was somewhat predictable given the
270
heterogeneity evident at the genetic, neurobiological, and behavioural levels of
explanation. This highlights the importance of an integrated approach to ASD research,
where findings from all levels of explanation constrain and inform each other (see
Bailey et al., 1996; Tager-Flusberg, 1999a).
Nevertheless, recognition of this need for integration is only the first step
towards the discovery of which kind of multiple deficits model may best explain ASDs.
Several key questions remain. Are ASDs best conceptualised as a group of distinct
subtypes or a multidimensional spectrum? How should these subgroups or spectrums
be defined and operationalised? Are they associated with different genotypes or
neuropathologies? Are there other cognitive impairments besides ToM and EF which
may be equally primary in ASDs, and if so what are they? Do the various cognitive
impairments change in primacy or causal status with development? It may be the case
that a combination of the various multiple deficits models will end up forming the best
explanatory paradigm; for example, there may be a multidimensional autism spectrum
within which certain clusters often occur (this kind of model has been proposed
previously by Beglinger & Smith, 2001), where more than two cognitive deficits are
present and these deficits change in primacy throughout development. The problem
with this sort of model is that its complexity makes it very difficult to test empirically.
The methodological and conceptual difficulties with determining which
integrated causal model can explain ASDs are characteristic of research on complex
genetic disorders in general, especially when dealing with disorders of development.
The task of beginning with behaviour and tracing the causal chain back through
cognition to the level of biology is full of hazards. The mapping of genotype to
phenotype is neither direct nor specific (Karmiloff-Smith et al., 2002), and a
neurological abnormality which occurs early in development can trigger a complex
chain of both structural and functional changes. This means that diverse pathogenic
processes may lead to similar behavioural phenotypes, and conversely, similar
pathogenic processes may lead to divergent behavioural symptoms (Courchesne,
Townsend, & Chase, 1995). In the words of Gottesman and Gould (2003):
In diseases with classic or Mendelian genetics as their distal causes, genotypes are usually
indicative of phenotypes. However, this degree of genetic certainty does not exist for diseases
with complex genetics. Genetic probabilism aptly describes the process by which a particular
genotype gives rise to phenotype. Epigenetic factors may also be of critical importance for
modifying the development of phenotypes, and such modifications may be influenced by
genotype or environment or be entirely stochastic in origin. Thus, models of complex genetic
271
disorders predict a ballet choreographed interactively over time among genotype, environment,
and epigenetic factors, which gives rise to a particular phenotype (p. 636 – references not
included).
The recognition of this multilayered complexity will be necessary to make progress in
the construction of an explanatory model for ASDs. While admirable attempts at
integrated models have been made (e.g., Courchesne et al., 1995; Dawson et al., 2002b;
Waterhouse et al., 1996), we are still far from understanding how to tie together the
diverse array of often inconsistent findings across all levels of explanation. This not
least in part because our understanding of the interactions between genes, neurobiology,
cognition and behaviour in typical development is crude and fragmentary at best.
7.4 Future directions
An integrative approach to autism research ideally requires both clarity within each
level of explanation and consistency and integration between the various levels of
explanation. Therefore, further research needs to address remaining questions at the
cognitive level of explanation, as well as the integration of cognitive findings with
research on behavioural outcomes, neurobiological substrates, and genetic mechanisms.
So which issues at the cognitive level of explanation deserve further attention?
Although neither ToM or EF impairments appear to be a core marker for autism, the
study of their nature, primacy and relationship can still inform research on causal
models of ASDs as well as the study of ToM, EF, and their relationship in typical
development. Firstly, there is a clear need to develop more high-level, ecologically
valid measures of ToM. It is evident that ToM develops beyond the ability to
understand false belief, and yet there are few tasks available for investigating more
advanced ToM development. Those that are available (e.g., the Eyes task, Strange
Stories) suffer from a lack of process purity, relying heavily on other abilities such as
face perception, emotion recognition, and verbal comprehension. Ecological validity is
an important task property as attempts to rigorously control the conditions of the task
can end up altering the essence of the phenomenon under study (Volkmar et al., 2004),
but the search for ecological validity often comes at the expense of task purity. The
challenge is therefore to develop ecologically valid tasks which have in-built control
conditions that allow isolation of the target ability. Such tasks would allow more
precise examination of the extent of higher-level ToM development in individuals with
ASDs and their first-degree relatives. They would also aid the investigation of the
272
nature of the ToM-EF relationship in children and adults over the age of five, which will
be important both for the extension of theory and empirical findings on the ToM-EF
relationship at later ages and for the interpretation of findings on the ToM-EF
relationship in clinical samples in this older age range.
The development of ToM tasks which incorporate the component process
approach may also help uncover any compensatory strategies which may be used to aid
ToM performance. The use of alternative strategies often becomes a “default”
explanation for intact performance on ToM tasks, yet this hypothesis has not been
directly tested, relying only on indirect evidence such as neuroimaging data. More
direct tests could involve systematic manipulation of the problem-solving requirements
of high-level ToM tasks to examine how performance is affected with certain strategies
cannot be used. Such multiple-condition tasks may also represent a method of
investigating the validity of one of the proposed explanations for the intriguing lack of
correlations between ToM and EF in individuals with ASDs (i.e., the possibility that the
lack of significant correlations between ToM and EF in individuals with ASDs who
show EF impairment is due to the use of alternative strategies for ToM performance; see
Section 4.4.3 in Chapter 4).
A number of aspects of EF in ASDs also merit further investigation. The
findings of this research suggest that deficits in generativity play a key role in ASDs.
The impairment in verbal generativity displayed by ASD probands demonstrated the
largest effect size, was one of the most prevalent deficits, and was a significant
discriminator between the ASD and control groups, and an impairment in non-verbal
generativity was one of the few significant deficits to emerge in ASD siblings. Further
investigation of generativity with individuals with ASDs of a larger age range and using
a wider range of tasks therefore appears worthwhile, although it awaits the development
of generativity tasks appropriate for young children. It is interesting to note that
generativity tasks are generally the most unstructured of EF tasks, as they require the
participant to produce novel responses, as opposed to reacting to stimuli presented to
them as part of a structured task. It would therefore also be interesting to see whether
the apparent severity of impairment displayed on generativity tasks is due to a specific
problem with generativity, or whether EF impairments in general are better detected and
therefore more severe on unstructured tasks which are more representative of many real-
life situations (i.e., have higher ecological validity), regardless of which EF component
is involved. This could be addressed by designing more unstructured tests of other EF
components.
273
Impairments in verbal inhibition and on tasks requiring a combination of
inhibitory and working memory requirements were also new findings in this research
which await replication. If further studies confirm the existence of an inhibitory
impairment in ASDs, this raises additional questions about the discriminant validity of
the EF profile in ASDs as compared with other disorders such as ADHD. Even if EF
impairments are not singularly primary in autism and therefore do not strictly need to
meet the uniqueness criterion, the question remains as to why similar EF impairments
result in such different behaviours in different disorders. Is it the case that EF
impairments must co-occur with certain other cognitive impairments, or emerge at a
particular point in development - or both - in order to produce the unique behaviours
displayed by individuals with ASDs?
The integration of cognitive and behavioural levels of explanation is the next
challenge in constructing an integrated explanatory model of ASDs. To begin with, in
order to examine the relationships between cognition and behaviour in a more precise
manner, we first require more accurate measures of behaviour. As mentioned
previously, parental report and observational techniques each have their own set of
problems. A combination of approaches may be necessary to gain a complete picture of
the nature and severity of behavioural symptomatology (Bishop & Norbury, 2002).
However, it will first be necessary to develop observational scales which are appropriate
for capturing both the normal range of variation in behaviour and the extremes of
abnormality. Without this, it remains unclear whether ToM and EF impairments do
actually underlie the behaviours that they are commonly purported to – or indeed
whether they hold any explanatory value at all. If ToM and/or EF do not show strong
relationships with behaviour, it remains possible that they are simply pleiotropic effects
– that is, they may be related to the genetic mechanisms which cause autism but
unrelated to its behavioural phenotype. This does not seem plausible given the results
of previous research and our knowledge about the behavioural effects of cognitive
deficits such as EF impairment in other disorders (e.g., individuals with frontal lobe
damage), but it is a possibility which needs to be ruled out using valid behavioural
measures.
The need for longitudinal studies which track the development of cognition and
behaviour in children with ASDs from an early age is a theme which has recurred
throughout this thesis as well as autism research in general. These studies will be
crucial for i) determining whether ToM or EF impairments have causal precedence, ii)
examining relationships between ToM and EF impairments and their proposed
274
precursors such as joint attention (e.g., Leekam & Moore, 2001; Mundy, 2003) and
imitation (Rogers, 1999; Rogers & Pennington, 1991), and iii) investigating how early
cognitive deficits affect the nature and severity of both early and later behavioural
symptomatology. Until recently, cognitive theories of ASDs have largely ignored the
process of development and instead proposed essentially static impairments which
supposedly persist throughout the affected individual’s lifetime. However, the
importance of considering developmental factors when conducting research on
developmental psychopathologies is being increasingly emphasised (e.g., Bishop, 1997;
Karmiloff-Smith, 1992; Tager-Flusberg, 1999a; Thomas & Karmiloff-Smith, 2002).
For example, Karmiloff-Smith (1997) recommended six changes of approach for
research in developmental cognitive neuroscience:
1. The recognition that plasticity is the rule, not simply a specialised response to injury.
2. The identification of constraints on plasticity.
3. A focus on the dynamics of development at multiple levels.
4. The recognition that specialisation within some brain regions is the product of development,
not its starting point.
5. A focus not only on the end state but also how the child progressively develops to the end
state.
6. The in-depth analysis of the different processes by which seemingly normal surface
behaviour can be produced by a brain that has developed differently from the outset. (p.
514).
Increased recognition of the role of interactive developmental processes in cognitive
performance and behavioural outcomes in the field of autism has paralleled this more
general shift (e.g., Bowler, 2001; Burack, Charman, Yirmiya, & Zelazo, 2001;
Courchesne et al., 1995; Happé, 2001; Steele, Joseph, & Tager-Flusberg, 2003; Tager-
Flusberg, 2001). However, while the need for longitudinal studies is often discussed, it
is rarely enacted. This is partly because autism is still difficult to diagnose early,
although recent progress in early diagnosis (see Charman & Baird, 2002) may facilitate
the ease with which longitudinal studies can be conducted. Targeting newborn siblings
of individuals with ASDs, who have an increased likelihood of developing an ASD, is
another method of identifying possible participants early for longitudinal monitoring.
The ability to conduct studies of cognitive impairment from a very young age has also
been hampered by the lack of appropriate tasks for young children, although again such
tasks are becoming increasingly available.
275
How might findings at the cognitive level of explanation inform research on the
neurobiological substrates of ASDs? One possible avenue results from the finding that
ToM and EF deficits were independent in ASDs, which indicates that their co-
occurrence is most likely to be explained by their neuroanatomical proximity. This
suggests that the functioning of the prefrontal cortex, both in its ventromedial and
dorsolateral aspects, is disrupted in individuals with ASDs. However, there is no clear
evidence of structural frontal abnormality in ASDs. Instead, it is possible that cortical
networks involving frontal regions may have been disrupted during development, or that
neurotransmitters which are particularly active in these regions are deficient1 (e.g., a
dopaminergic deficit may underlie the range of cognitive deficits displayed by
individuals with autism, as suggested by Pennington et al., 1997). It therefore appears
that investigations of the development of cortical networks involving the prefrontal
cortex and neurotransmitter systems which heavily populate frontal areas would be
worthwhile targets for neurobiological studies of ASDs.
The relationship between cognitive (and behavioural and neurobiological)
findings with underlying genetic mechanisms will arguably be one of the most
important and fruitful links in the search for the aetiology of autism. This research
suggested that there may be different subgroups of individuals with ASDs, possibly
defined by their level of functioning, which have different ToM and EF profiles.
Similarly, relatives of individuals with ASDs who demonstrated endophenotypes or
cognitive vulnerability markers also appeared to show a variety of cognitive profiles.
However, in both probands and siblings, it was also possible (and perhaps even more
likely) that cognitive performances varied on a more continuous spectrum, with the
apparent subgroups a result of classifying individuals scoring below a certain point as
“impaired”. One method of distinguishing between these two possibilities would be to
conduct a cluster analysis (based on cognitive performances) on a large sample of
individuals with ASDs and their relatives, and test firstly whether any meaningful
clusters emerged which showed unique behavioural characteristics, and secondly
whether any such clusters showed distinct genotypic markers and/or neuropathologies.
This kind of research has been conducted previously with promising results (Dawson,
Klinger, Panagiotides, Lewy, & Vastelloe, 1995; this study used the subgroup
classification system proposed by Wing and Gould (1979) based on differences in social
1 However, note that if the severity of ToM and EF deficits depended directly on the extent of neurotransmitter deficiency, then significant correlations between ToM and EF might be expected in the ASD population.
276
behaviour rather than using cluster analysis). The notion of different subgroups of
individuals with ASDs characterised by different genetic mechanisms has been
previously proposed to explain the heterogeneity evident at all levels of explanation
(e.g., Szatmari, 1999; Tager-Flusberg & Joseph, 2003). However, it remains to be seen
how these subgroups should be defined, or indeed, whether the notion of subgroups
holds any validity at all. The kind of multi-level approach proposed above seems the
most appropriate way of approaching this problem, although cluster analysis is limited
by the absence of objective rules for defining the boundaries of each subgroup (Lorr,
1994).
The current research has consolidated and extended previous work by
demonstrating that i) neither ToM or EF impairments meet criteria for a single primary
deficit in ASDs, ii) ToM and EF impairments are independent and do not explain each
other, and iii) multiple deficits models involving subgroups or spectrums which are
probably not based on symptom domains or severity, and where deficits are not
considered static and unchanging, are the best place to focus future research efforts.
Studies of cognitive mechanisms in ASDs and their relationship with behaviour and
biological substrates should move away from attempting to find a specific core
cognitive deficit which could “explain autism” and instead focus upon mapping the
profile of deficits and examining how these deficits change over time and interact with
the other levels of explanation. The challenge will be to develop creative methods and
strategies for implementing an integrated, developmental approach which recognises the
complexities and dynamics of the genotype-phenotype interactions that underlie autism.
277
278
REFERENCES
Abu-Akel, A. (2003). A neurobiological mapping of theory of mind. Brain Research
Reviews, 43, 29-40.
Akshoomoff, N., Pierce, K., & Courchesne, E. (2002). The neurobiological basis of
autism from a developmental perspective. Development & Psychopathology,
14(3), 613-634.
Alexander, M. P. (2002). Disorders of language after frontal lobe injury: Evidence for
the neural mechanisms of assembling language. In D. T. Stuss & R. T. Knight
(Eds.), Principles of frontal lobe function (pp. 159-167). London: Oxford
University Press.
American Psychiatric Association. (1994). Diagnostic and statistical manual of mental
disorders. (4th ed.). Washington, DC: Author.
Ames, D., Cummings, J. L., Wirshing, W. C., Quinn, B., & Mahler, M. (1994).
Repetitive and compulsive behavior in frontal lobe degenerations. Journal of
Neuropsychiatry & Clinical Neurosciences, 6(2), 100-113.
Anderson, C. V., Bigler, E. D., & Blatter, D. D. (1995). Frontal lobe lesions, diffuse
damage, and neuropsychological functioning in traumatic brain-injured patients.
Journal of Clinical & Experimental Neuropsychology, 17(6), 900-908.
Anderson, P. (2002). Assessment and development of executive function (EF) during
childhood. Child Neuropsychology, 8(2), 71-82.
Anderson, P., Anderson, V., & Lajoie, G. (1996). The Tower of London Test:
Validation and standardization for pediatric populations. Clinical
Neuropsychologist, 10(1), 54-65.
Anderson, S. W., Bechara, A., Damasio, H., Tranel, D., & Damasio, A. R. (1999).
Impairment of social and moral behavior related to early damage in human
prefrontal cortex. Nature Neuroscience, 2(11), 1032-1037.
Anderson, S. W., Damasio, H., Jones, R., & Tranel, D. (1991). Wisconsin Card Sorting
Test performance as a measure of frontal lobe damage. Journal of Clinical &
Experimental Neuropsychology, 13(6), 909-922.
Anderson, V. (1998). Assessing executive functions in children: Biological,
psychological, and developmental considerations. Neuropsychological
Rehabilitation, 8(3), 319-349.
279
Anderson, V. A., Anderson, P., Northam, E., Jacobs, R., & Catroppa, C. (2001).
Development of executive functions through late childhood and adolescence in
an Australian sample. Developmental Neuropsychology, 20(1), 385-406.
Anderson, V., Levin, H. S., & Jacobs, R. (2002). Executive functions after frontal lobe
injury: A developmental perspective. In D. T. Stuss & R. T. Knight (Eds.),
Principles of frontal lobe function (pp. 504-527). London: Oxford University
Press.
Astington, J. W., Harris, P. L., & Olson, D. R. (Eds.). (1988). Developing theories of
mind. Cambridge: Cambridge University Press.
Atkinson, R., & Shiffrin, R. (1968). Human memory: A proposed system and its control
processes. In K. W. Spence & J. T. Spence (Eds.), The psychology of learning
and motivation . New York: Academic Press.
August, G. J., Stewart, M. A., & Tsai, L. (1981). The incidence of cognitive disabilities
in the siblings of autistic children. British Journal of Psychiatry, 138, 416-422.
Bach, L., Davies, S., Colvin, C., Wijeratne, C., Happé, F., & Howard, R. (1998). A
neuropsychological investigation of theory of mind in an elderly lady with
frontal leucotomy. Cognitive Neuropsychiatry, 3(2), 139-159.
Bach, L. J., Happé, F., Fleminger, S., & Powell, J. (2000). Theory of mind:
Independence of executive function and the role of the frontal cortex in acquired
brain injury. Cognitive Neuropsychiatry, 5(3), 175-192.
Bachevalier, J. (1994). Medial temporal lobe structures and autism: A review of clinical
and experimental findings. Neuropsychologia, 32(6), 627-648.
Bachevalier, J., & Loveland, K. A. (2003). Early orbitofrontal-limbic dysfunction and
autism. In D. Cicchetti & E. Walker (Eds.), Neurodevelopmental mechanisms in
psychopathology (pp. 215-236). New York: Cambridge University Press.
Baddeley, A. D. (1986). Working memory. Oxford: Oxford University Press.
Baddeley, A. (1996). Exploring the central executive. Quarterly Journal of
Experimental Psychology A, 49A(1), 5-28.
Baddeley, A. (2002). Fractionating the central executive. In D. T. Stuss & R. T. Knight
(Eds.), Principles of frontal lobe function (pp. 246-260). London: Oxford
University Press.
Baddeley, A. D., & Hitch, G. (1974). Working memory. In G. H. Bower (Ed.), The
psychology of learning and motivation (Vol. 8). New York: Academic Press.
280
Bailey, A., Le Couteur, A., Gottesman, I., & Bolton, P. (1995). Autism as a strongly
genetic disorder: Evidence from a British twin study. Psychological Medicine,
25(1), 63-77.
Bailey, A., Palferman, S., Heavey, L., & Le Couteur, A. (1998). Autism: The phenotype
in relatives. Journal of Autism & Developmental Disorders, 28(5), 369-392.
Bailey, A., Phillips, W., & Rutter, M. (1996). Autism: Towards an integration of
clinical, genetic, neuropsychological, and neurobiological perspectives. Journal
of Child Psychology and Psychiatry, 37(1), 89-126.
Baird, G., Charman, T., Baron-Cohen, S., Cox, A., Swettenham, J., Wheelwright, S., &
Drew, A. (2000). A screening instrument for autism at 18 months of age: A 6-
year follow-up study. Journal of the American Academy of Child and Adolescent
Psychiatry, 39(6), 694-702.
Baird, T. D., & August, G. J. (1985). Familial heterogeneity in infantile autism. Journal
of Autism & Developmental Disorders, 15(3), 315-321.
Baron-Cohen, S. (1988). Social and pragmatic deficits in autism: Cognitive or affective?
Journal of Autism & Developmental Disorders, 18(3), 379-402.
Baron-Cohen, S. (1989a). Are autistic children "behaviorists"? An examination of their
mental-physical and appearance-reality distinctions. Journal of Autism &
Developmental Disorders, 19(4), 579-600.
Baron-Cohen, S. (1989b). The autistic child's theory of mind: A case of specific
developmental delay. Journal of Child Psychology & Psychiatry & Allied
Disciplines, 30(2), 285-297.
Baron-Cohen, S. (1989c). Do autistic children have obsessions and compulsions?
British Journal of Clinical Psychology, 28(3), 193-200.
Baron-Cohen, S. (1989d). Perceptual role taking and protodeclarative pointing in
autism. British Journal of Developmental Psychology, 7(2), 113-127.
Baron-Cohen, S. (1991a). Do people with autism understand what causes emotion?
Child Development, 62, 385-395.
Baron-Cohen, S. (1991b). Precursors to a theory of mind: Understanding attention in
others. In A. Whiten (Ed.), Natural theories of mind: Evolution, development
and simulation of everyday mindreading (pp. 233-251). Oxford: Blackwell.
Baron-Cohen, S. (1991c). The theory of mind deficit in autism: How specific is it?
British Journal of Developmental Psychology, 9(2), 301-314.
281
Baron-Cohen, S. (1992). Debate and argument: On modularity and development in
autism: A reply to Burack. Journal of Child Psychology & Psychiatry & Allied
Disciplines, 33(3), 623-629.
Baron-Cohen, S. (1994). How to build a baby that can read minds: Cognitive
mechanisms in mindreading. Cahiers de Psychologie Cognitive, 13(5), 513-552.
Baron-Cohen, S. (1995). Mindblindness: An essay on autism and theory of mind.
Cambridge, MA: MIT Press.
Baron-Cohen, S. (1998). Does the study of autism justify minimalist innate modularity?
Learning & Individual Differences, 10(3), 179-191.
Baron-Cohen, S. (2000). Theory of mind and autism: A fifteen year review. In S.
Baron-Cohen, H. Tager-Flusberg, & D. J. Cohen (Eds.), Understanding other
minds: Perspectives from developmental cognitive neuroscience (2nd ed., pp. 3-
20). London: Oxford University Press.
Baron-Cohen, S., Allen, J., & Gillberg, C. (1992). Can autism be detected at 18 months?
The needle, the haystack, and the CHAT. British Journal of Psychiatry, 161,
839-843.
Baron-Cohen, S., Campbell, R., Karmiloff-Smith, A., Grant, J., & Walker, J. (1995).
Are children with autism blind to the mentalistic significance of the eyes?
British Journal of Developmental Psychology, 13, 379-398.
Baron-Cohen, S., & Goodhart, F. (1994). The "seeing-leads-to-knowing" deficit in
autism: The Pratt and Bryant probe. British Journal of Developmental
Psychology, 12(3), 397-401.
Baron-Cohen, S., & Hammer, J. (1997). Parents of children with Asperger syndrome:
What is the cognitive phenotype? Journal of Cognitive Neuroscience, 9(4), 548-
554.
Baron-Cohen, S., Jolliffe, T., Mortimore, C., & Robertson, M. (1997). Another
advanced test of theory of mind: Evidence from very high functioning adults
with autism or Asperger Syndrome. Journal of Child Psychology & Psychiatry
& Allied Disciplines, 38(7), 813-822.
Baron-Cohen, S., Leslie, A. M., & Frith, U. (1985). Does the autistic child have a
"theory of mind"? Cognition, 21(1), 37-46.
Baron-Cohen, S., Leslie, A. M., & Frith, U. (1986). Mechanical, behavioural and
Intentional understanding of picture stories in autistic children. British Journal
of Developmental Psychology, 4(2), 113-125.
282
Baron-Cohen, S., & Ring, H. (1994). A model of the mindreading system:
Neuropsychological and neurobiological perspectives. In C. Lewis & P. Mitchell
(Eds.), Children's early understanding of mind: Origins and development (pp.
183-207). Hove, UK: Lawrence Erlbaum Associates.
Baron-Cohen, S., Ring, H., Moriarty, J., Schmitz, B., Costa, D., & Ell, P. (1994).
Recognition of mental state terms: Clinical findings in children with autism and
a functional neuroimaging study of normal adults. British Journal of Psychiatry,
165(5), 640-649.
Baron-Cohen, S., Ring, H. A., Wheelwright, S., Bullmore, E. T., Brammer, M. J.,
Simmons, A., & Williams, S. C. R. (1999a). Social intelligence in the normal
and autistic brain: An fMRI study. European Journal of Neuroscience, 11, 1891-
1898.
Baron-Cohen, S., & Robertson, M. M. (1995). Children with either autism, Gilles de la
Tourette syndrome or both: Mapping cognition to specific syndromes.
Neurocase: Case Studies in Neuropsychology, Neuropsychiatry, & Behavioural
Neurology, 1(2), 101-104.
Baron-Cohen, S., & Swettenham, J. (1997). Theory of mind in autism: Its relationship
to executive function and central coherence. In D. J. Cohen & F. R. Volkmar
(Eds.), Handbook of autism and pervasive developmental disorders (2nd ed., pp.
880-893). New York: John Wiley & Sons.
Baron-Cohen, S., Wheelwright, S., Hill, J., Raste, Y., & Plumb, I. (2001a). The
"Reading the mind in the eyes" Test revised version: A study with normal adults,
and adults with Asperger syndrome or high-functioning autism. Journal of Child
Psychology & Psychiatry & Allied Disciplines, 42(2), 241-251.
Baron-Cohen, S., Wheelwright, S., Skinner, R., Martin, J., & Clubley, E. (2001b). The
Autism-Spectrum Quotient (AQ): Evidence from Asperger syndrome/high-
functioning autism, males and females, scientists and mathematicians. Journal of
Autism & Developmental Disorders, 31(1), 5-17.
Baron-Cohen, S., Wheelwright, S., Stone, V., & Rutherford, M. (1999b). A
mathematician, a physicist and a computer scientist with Asperger syndrome:
Performance on folk psychology and folk physics tests. Neurocase, 5(6), 475-
483.
Bartsch, K., & Wellman, H. (1989). Young children's attribution of action to beliefs and
desires. Child Development, 60(4), 946-964.
283
Bartsch, K., & Wellman, H. M. (1995). Children talk about the mind. New York:
Oxford University Press.
Bauman, M. L. (1999). Autism: Clinical features and neurobiological observations. In
H. Tager-Flusberg (Ed.), Neurodevelopmental disorders: Developmental
cognitive neuroscience (pp. 383-399). Cambridge, MA: The MIT Press.
Bauman, M. L., & Kemper, T. L. (1994). Neuroanatomic observations of the brain in
autism. In M. L. Bauman & T. L. Kemper (Eds.), The neurobiology of autism
(pp. 119-145). Baltimore, MA: John Hopkins.
Bauminger, N., & Kasari, C. (1999). Brief report: Theory of mind in high-functioning
children with autism. Journal of Autism & Developmental Disorders, 29(1), 81-
86.
Bauminger, N., & Yirmiya, N. (2001). The functioning and well-being of siblings of
children with autism: Behavioral-genetic and familial contributions. In J. A.
Burack, T. Charman, N. Yirmiya, & P. R. Zelazo (Eds.), The development of
autism: Perspectives from theory and research (pp. 61-80). Mahwah, NJ:
Lawrence Erlbaum Associates.
Beglinger, L. J., & Smith, T. H. (2001). A review of subtyping in autism and proposed
dimensional classification model. Journal of Autism & Developmental
Disorders, 31(4), 411-422.
Bell, M. A., & Fox, N. A. (1992). The relations between frontal brain electrical activity
and cognitive development during infancy. Child Development, 63(5), 1142-
1163.
Bennetto, L., Pennington, B. F., & Rogers, S. J. (1996). Intact and impaired memory
functions in autism. Child Development, 67, 1816-1835.
Benson, G., Abbeduto, L., Short, K., Bibler-Nuccio, J., & Maas, F. (1993).
Development of a theory of mind in individuals with mental retardation.
American Journal on Mental Retardation, 98(3), 427-433.
Berger, H. J., Aerts, F. H., van Spaendonck, K. P., Cools, A. R., & Teunisse, J.-P.
(2003). Central coherence and cognitive shifting in relation to social
improvement in high-functioning young adults with autism. Journal of Clinical
& Experimental Neuropsychology, 25(4), 502-511.
Berger, H. J., Van Spaendonck, K. P., Horstink, M. W., Buytenhuijs, E. L., Lammers, P.
W. J. M., & Cools, A. R. (1993). Cognitive shifting as a predictor of progress in
social understanding in high-functioning adolescents with autism: A prospective
study. Journal of Autism & Developmental Disorders, 23(2), 341-359.
284
Bertrand, J., Mars, A., Boyle, C., Bove, F., Yeargin-Allsopp, M., & Decoufle, P. (2001).
Prevalence of autism in a United States population: The Brick Township, New
Jersey, investigation. Pediatrics, 108(5), 1155-61.
Berument, S. K., Rutter, M., Lord, C., Pickles, A., & Bailey, A. (1999). Autism
screening questionnaire: Diagnostic validity. British Journal of Psychiatry, 175,
444-451.
Bettelheim, B. (1967). The empty fortress: Infantile autism and the birth of the self.
New York: Free Press.
Beveridge, M., Jarrold, C., & Pettit, E. (2002). An experimental approach to executive
fingerprinting in young children. Infant & Child Development, 11(2), 107-123.
Biro, S., & Russell, J. (2001). The execution of arbitrary procedures by children with
autism. Development & Psychopathology, 13(1), 97-110.
Bishop, D. V. (1993). Annotation: Autism, executive functions and theory of mind: A
neuropsychological perspective. Journal of Child Psychology & Psychiatry &
Allied Disciplines, 34(3), 279-293.
Bishop, D. V. M. (1997). Cognitive neuropsychology and developmental disorders:
Uncomfortable bedfellows. Quarterly Journal of Experimental Psychology A,
50A(4), 899-923.
Bishop, D. V. M. (2000). What's so special about Asperger syndrome? The need for
further exploration of the borderlands of autism. In A. Klin & F. R. Volkmar
(Eds.), Asperger syndrome (pp. 254-277). New York: Guilford Press.
Bishop, D. V. M., Maybery, M., Maley, A., Wong, D., Hill, W., & Hallmayer, J. (in
press-a). Using self-report to identify the broad phenotype in parents of children
with autistic spectrum disorders: A study using the Autism-Spectrum Quotient.
Journal of Child Psychology & Psychiatry.
Bishop, D. V. M., Maybery, M., Wong, D., Maley, A., Hill, W., & Hallmayer, J. (in
press-b). Are phonological processing deficits part of the broad autism
phenotype? American Journal of Medical Genetics (Neuropsychiatric Genetics).
Bishop, D. V., & Norbury, C. F. (2002). Exploring the borderlands of autistic disorder
and specific language impairment: A study using standardised diagnostic
instruments. Journal of Child Psychology & Psychiatry & Allied Disciplines,
43(7), 917-929.
Bjorklund, D. F., & Harnishfeger, K. K. (1990). The resources construct in cognitive
development: Diverse sources of evidence and a theory of inefficient inhibition.
Developmental Review, 10(1), 48-71.
285
Blair, J., Sellars, C., Strickland, I., Clark, F., Williams, A., Smith, M., & Jones, L.
(1996). Theory of mind in the psychopath. Journal of Forensic Psychiatry, 7(1),
15-25.
Bolton, P., Macdonald, H., Pickles, A., Rios, P., Goode, S., Crowson, M., Bailey, A., &
Rutter, M. (1994). A case-control family history study of autism. Journal of
Child Psychology & Psychiatry & Allied Disciplines, 35(5), 877-900.
Bolton, P., Pickles, A., Murphy, M., & Rutter, M. (1998). Autism, affective and other
psychiatric disorders: Patterns of familial aggregation. Psychological Medicine,
28(2), 385-395.
Boone, K. B., Ponton, M. O., Gorsuch, R. L., Gonzalez, J. J., & Miller, B. L. (1998).
Factor analysis of four measures of prefrontal lobe functioning. Archives of
Clinical Neuropsychology, 13(7), 585-595.
Boucher, J. (1988). Word fluency in high-functioning autistic children. Journal of
Autism & Developmental Disorders, 18(4), 637-645.
Boucher, J. (1996). What could possibly explain autism? In P. Carruthers & P. K. Smith
(Eds.), Theories of theories of mind (pp. 223-241). Cambridge: Cambridge
University Press.
Boutin, P., Maziade, M., Merette, C., Mondor, M., Bedard, C., & Thivierge, J. (1997).
Family history of cognitive disabilities in first-degree relatives of autistic and
mentally retarded children. Journal of Autism & Developmental Disorders,
27(2), 165-176.
Bowler, D. M. (1992). "Theory of mind" in Asperger's syndrome. Journal of Child
Psychology & Psychiatry & Allied Disciplines, 33(5), 877-893.
Bowler, D. M. (2001). Autism: Specific cognitive deficit or emergent end point of
multiple interacting systems? In J. A. Burack, T. Charman, N. Yirmiya, & P. R.
Zelazo (Eds.), The development of autism: Perspectives from theory and
research (pp. 219-235). Mahwah, NJ: Lawrence Erlbaum Associates.
Bowler, D. M., & Briskman, J. A. (2000). Photographic cues do not always facilitate
performance on false belief tasks in children with autism. Journal of Autism &
Developmental Disorders, 30(4), 305-316.
Brian, J. A., Tipper, S., Weaver, B., & Bryson, S. (2003). Inhibitory mechanisms in
autism spectrum disorders: Typical selective inhibition of location versus
facilitated perceptual processing. Journal of Child Psychology & Psychiatry &
Allied Disciplines, 44(4), 552-560.
286
Briskman, J., Happé, F., & Frith, U. (2001). Exploring the cognitive phenotype of
autism: Weak "central coherence" in parents and siblings of children in autism:
II. Real-life skills and preferences. Journal of Child Psychology & Psychiatry &
Allied Disciplines, 42(3), 309-316.
Brothers, L. (1996). Brain mechanisms of social cognition. Journal of
Psychopharmacology, 10(1), 2-8.
Brown, R., Hobson, R., Lee, A., & Stevenson, J. (1997). Are there "autistic-like"
features in congenitally blind children? Journal of Child Psychology &
Psychiatry & Allied Disciplines, 38(6), 693-703.
Bruner, J., & Feldman, C. (1993). Theories of mind and the problem of autism. In S.
Baron-Cohen, H. Tager-Flusberg, & D. J. Cohen (Eds.), Understanding other
minds: Perspectives from autism. Oxford: Oxford University Press.
Brunet, E., Sarfati, Y., Hardy-Baylé, M. C., & Decety, J. (2000). A PET investigation of
the attribution of intentions with a non-verbal task. NeuroImage, 11, 157-166.
Bryson, S. E., Landry, R., & Wainwright, J. A. (1997). A componential view of
executive dysfunction in autism: Review of recent evidence. In J. A. Burack & J.
T. Enns (Eds.), Attention, development, and psychopathology (pp. 232-255).
New York: The Guilford Press.
Buitelaar, J. K., Swaab, H., van der Wees, M., Wildschut, M., & van der Gaag, R. J.
(1996). Neuropsychological impairments and deficits in theory of mind and
emotion recognition in a non-autistic boy. European Child & Adolescent
Psychiatry, 5(1), 44-51.
Burack, J. A. (1994). Selective attention deficits in persons with autism: Preliminary
evidence of an inefficient attentional lens. Journal of Abnormal Psychology,
103(3), 535-543.
Burack, J. A., Charman, T., Yirmiya, N., & Zelazo, P. R. (Eds.). (2001). The
development of autism: Perspectives from theory and research. Mahwah, NJ:
Lawrence Erlbaum Associates.
Burgess, P. W. (1997). Theory and methodology in executive function research. In P.
Rabbitt (Ed.), Methodology of frontal and executive function (pp. 81-116). Hove,
UK: Psychology Press.
Burgess, P. W. (2000). Strategy application disorder: The role of the frontal lobes in
human multitasking. Psychological Research, 63(3-4), 279-288.
287
Burgess, P. W., Alderman, N., Evans, J., Emslie, H., & Wilson, B. A. (1998). The
ecological validity of tests of executive function. Journal of the International
Neuropsychological Society, 4(6), 547-558.
Cabeza, R., & Nyberg, L. (2000). Imaging cognition II: An empirical review of 275
PET and fMRI studies. Journal of Cognitive Neuroscience, 12(1), 1-47.
Cantwell, D. P., Baker, L., & Rutter, M. (1979). Families of autistic and dysphasic
children: I. Family life and interaction patterns. Archives of General Psychiatry,
36(6), 682-687.
Capps, L., Kehres, J., & Sigman, M. (1998). Conversational abilities among children
with autism and children with developmental delays. Autism, 2(4), 325-344.
Carlin, D., Bonerba, J., Phipps, M., Alexander, G., Shapiro, M., & Grafman, J. (2000).
Planning impairments in frontal lobe dementia and frontal lobe lesion patients.
Neuropsychologia, 38(5), 655-665.
Carlson, S. M., & Moses, L. J. (2001). Individual differences in inhibitory control and
children's theory of mind. Child Development, 72(4), 1032-1053.
Carlson, S. M., Moses, L. J., & Breton, C. (2002). How specific is the relation between
executive function and theory of mind? Contributions of inhibitory control and
working memory. Infant & Child Development, 11(2), 73-92.
Carlson, S. M., Moses, L. J., & Hix, H. R. (1998). The role of inhibitory processes in
young children's difficulties with deception and false belief. Child Development,
69(3), 672-691.
Carpenter, M., Pennington, B. F., & Rogers, S. J. (2001). Understanding of others'
intentions in children with autism. Journal of Autism & Developmental
Disorders, 31(6), 589-599.
Carruthers, P. (1996). Autism as mind-blindness: An elaboration and partial defence. In
P. Carruthers & P. K. Smith (Eds.), Theories of theories of mind (pp. 257-273).
Cambridge: Cambridge University Press.
Carruthers, P., & Smith, P. K. (Eds.). (1996). Theories of theories of mind. Cambridge:
Cambridge University Press.
Casanova, M. F., Buxhoeveden, D. P., Switala, A. E., & Roy, E. (2002). Minocolumnar
pathology in autism. Neurology, 58(3), 428-432.
Case, R. (1985). Intellectual development: From birth to adulthood. New York:
Academic Press.
288
Castelli, F., Frith, C., Happé, F., & Frith, U. (2002). Autism, Asperger syndrome and
brain mechanisms for the attribution of mental states to animated shapes. Brain,
125, 1839-1849.
Chakrabarti, S., & Fombonne, E. (2001). Pervasive developmental disorders in
preschool children. Jama: Journal of the American Medical Association,
285(24), 3093-3099.
Chandler, M. J., Fritz, A. S., & Hala, S. M. (1989). Small scale deceit: Deception as a
marker of 2-, 3- and 4-year-olds' early theories of mind. Child Development, 60,
1263-1277.
Chandler, M., & Hala, S. (1994). The role of personal involvement in the assessment of
early false belief skills. In C. Lewis & P. Mitchell (Eds.), Children's early
understanding of mind: Origins and development (pp. 403-425). Hillsdale, NJ:
Lawrence Erlbaum.
Channon, S., & Crawford, S. (2000). The effects of anterior lesions on performance on a
story comprehension test: Left anterior impairment on a theory of mind-type
task. Neuropsychologia, 38(7), 1006-1017.
Channon, S., Flynn, D., & Robertson, M. M. (1992). Attentional deficits in Gilles de la
Tourette syndrome. Neuropsychiatry, Neuropsychology, & Behavioral
Neurology, 5(3), 170-177.
Charman, T. (2000). Theory of mind and the early diagnosis of autism. In S. Baron-
Cohen, H. Tager-Flusberg, & D. J. Cohen (Eds.), Understanding other minds:
Perspectives from developmental cognitive neuroscience (2nd ed., pp. 422-441).
London: Oxford University Press.
Charman, T., & Baird, G. (2002). Practitioner review: Diagnosis of autism spectrum
disorder in 2- and 3-year-old children. Journal of Child Psychology &
Psychiatry & Allied Disciplines, 43(3), 289-305.
Charman, T., & Baron-Cohen, S. (1992). Understanding drawings and beliefs: A further
test of the metarepresentation theory of autism: A research note. Journal of
Child Psychology & Psychiatry & Allied Disciplines, 33(6), 1105-1112.
Charman, T., & Baron-Cohen, S. (1995). Understanding photos, models, and beliefs: A
test of the modularity thesis of theory of mind. Cognitive Development, 10(2),
287-298.
Charman, T., & Campbell, A. (1997). Reliability of theory of mind task performance by
individuals with a learning disability: A research note. Journal of Child
Psychology & Psychiatry & Allied Disciplines, 38(6), 725-730.
289
Charman, T., Carroll, F., & Sturge, C. (2001). Theory of mind, executive function and
social competence in boys with ADHD. Emotional & Behavioural Difficulties,
6(1), 31-49.
Charman, T., & Lynggaard, H. (1998). Does a photographic cue facilitate false belief
performance in subjects with autism? Journal of Autism & Developmental
Disorders, 28(1), 33-42.
Chelune, G. J., & Baer, R. A. (1986). Developmental norms for the Wisconsin Card
Sorting Test. Journal of Clinical & Experimental Neuropsychology, 8(3), 219-
228.
Christensen, K. J., Kim, S. W., Dysken, M. W., & Hoover, K. M. (1992).
Neuropsychological performance in obsessive-compulsive disorder. Biological
Psychiatry, 31(1), 4-18.
Cicerone, K. D., & Tanenbaum, L. N. (1997). Disturbance of social cognition after
traumatic orbitofrontal brain injury. Archives of Clinical Neuropsychology,
12(2), 173-188.
Ciesielski, K. T., & Harris, R. J. (1997). Factors related to performance failure on
executive tasks in autism. Child Neuropsychology, 3(1), 1-12.
Clark, P., & Rutter, M. (1981). Autistic children's responses to structure and to
interpersonal demands. Journal of Autism & Developmental Disorders, 11(2),
201-217.
Cohen, J. (1988). Statistical power analysis for the behavioural sciences. (2nd ed.).
Hillsdale, NJ: Lawrence Erlbaum.
Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation analysis for the
behavioral sciences. (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum.
Collette, F., & van der Linden, M. (2002). Brain imaging of the central executive
component of working memory. Neuroscience & Biobehavioral Reviews, 26(2),
105-125.
Collette, F., van der Linden, M., & Salmon, E. (1999). Executive dysfunction in
Alzheimer's disease. Cortex, 35(1), 57-72.
Colvert, E., Custance, D., & Swettenham, J. (2002). Rule-based reasoning and theory of
mind in autism: A commentary on the work of Zelazo, Jacques, Burack and
Frye. Infant & Child Development, 11(2), 197-200.
Corcoran, R. (2000). Theory of mind in other clinical conditions: Is a selective 'theory
of mind' deficit exclusive to autism? In S. Baron-Cohen, H. Tager-Flusberg, &
D. J. Cohen (Eds.), Understanding other minds: Perspectives from
290
developmental cognitive neuroscience (2nd ed., pp. 391-421). London: Oxford
University Press.
Corcoran, R., Mercer, G., & Frith, C. D. (1995). Schizophrenia, symptomatology and
social influence: Investigating "theory of mind" in people with schizophrenia.
Schizophrenia Research, 17(1), 5-13.
Courchesne, E. (1997). Brainstem, cerebellar and limbic neuroanatomical abnormalities
in autism. Current Opinion in Neurobiology, 7(2), 269-278.
Courchesne, E., Townsend, J., Akshoomoff, N. A., Saitoh, O., Yeung-Courchesne, R.,
Lincoln, A. J., James, H. E., Haas, R. H., Schreibman, L., & Lau, L. (1994).
Impairment in shifting attention in autistic and cerebellar patients. Behavioral
Neuroscience, 108(5), 848-865.
Courchesne, E., Townsend, J., & Chase, C. (1995). Neurodevelopmental principles
guide research on developmental psychopathologies. In D. Cicchetti & D. J.
Cohen (Eds.), Developmental psychopathology (Vol. 1: Theory and methods, pp.
195-226). Oxford: John Wiley & Sons.
Cox, C. S., Fedio, P., & Rapoport, J. L. (1989). Neuropsychological testing of
obsessive-compulsive adolescents. In J. L. Rapoport (Ed.), Obsessive-
compulsive disorder in children and adolescents (pp. 73-85). Washington, DC:
American Psychiatric Press.
Craig, J., & Baron-Cohen, S. (1999). Creativity and imagination in autism and Asperger
syndrome. Journal of Autism & Developmental Disorders, 29(4), 319-326.
Cripe, L. I. (1996). The ecological validity of executive function testing. In R. J.
Sbordone & C. J. Long (Eds.), Ecological validity of neuropsychological testing
(pp. 171-202). Delray Beach, FL: GR Press/St Lucie Press.
Culbertson, W. C., & Zillmer, E. A. (1998a). The construct validity of the Tower of
LondonDX as a measure of the executive functioning of ADHD children.
Assessment, 5(3), 215-226.
Culbertson, W. C., & Zillmer, E. A. (1998b). The Tower of LondonDX: A standardized
approach to assessing executive functioning in children. Archives of Clinical
Neuropsychology, 13(3), 285-301.
Dadds, M. R., Schwartz, S., Adams, T., & Rose, S. (1988). The effects of social context
and verbal skill on the stereotypic and task-involved behaviour of autistic
children. Journal of Child Psychology & Psychiatry & Allied Disciplines, 29(5),
669-676.
291
Dagher, A., Owen, A. M., Boecker, H., & Brooks, D. J. (1999). Mapping the network
for planning: A correlational PET activation study with the Tower of London
ask. Brain, 122(10), 1973-1987.
Dahlgren, S., Dahlgren Sandberg, A., & Hjelmquist, E. (2003). The non-specificity of
theory of mind deficits: Evidence from children with communicative disabilities.
European Journal of Cognitive Psychology, 15(1), 129-155.
Dahlgren, S. O., & Trillingsgaard, A. (1996). Theory of mind in non-retarded children
with autism and Asperger's syndrome: A research note. Journal of Child
Psychology & Psychiatry & Allied Disciplines, 37(6), 759-763.
Damasio, A. R., & Maurer, R. G. (1978). A neurological model for childhood autism.
Archives of Neurology, 35, 777-786.
Davis, H. L., & Pratt, C. (1995). The development of children's theory of mind: The
working memory explanation. Australian Journal of Psychology, 47(1), 25-31.
Dawson, G. (1991). A psychobiological perspective on the early socio-emotional
development of children with autism. In D. Cicchetti & S. L. Toth (Eds.),
Rochester symposium on developmental psychopathology (Vol. 3: Models and
integrations, pp. 207-234). Rochester, NY: University of Rochester Press
Dawson, G., & Adams, A. (1984). Imitation and social responsiveness in autistic
children. Journal of Abnormal Child Psychology, 12(2), 209-225.
Dawson, G., & Fernald, M. (1987). Perspective-taking ability and its relationship to the
social behavior of autistic children. Journal of Autism and Developmental
Disorders, 17, 487-498.
Dawson, G., Klinger, L. G., Panagiotides, H., Lewy, A., & Vastelloe, P. (1995).
Subgroups of autistic children based on social behavior display distinct patterns
of brain activity. Journal of Abnormal Child Psychology, 23(5), 569-583.
Dawson, G., & Lewy, A. (1989). Arousal, attention, and the socioemotional
impairments of individuals with autism. In G. Dawson (Ed.), Autism: Nature,
diagnosis, and treatment (pp. 49-74). New York: The Guildford Press.
Dawson, G., Meltzoff, A. N., Osterling, J., & Rinaldi, J. (1998). Neuropsychological
correlates of early symptoms of autism. Child Development, 69(5), 1276-1285.
Dawson, G., Munson, J., Estes, A., Osterling, J., McPartland, J., Toth, K., Carver, L., &
Abbott, R. (2002a). Neurocognitive function and joint attention ability in young
children with autism spectrum disorder versus developmental delay. Child
Development, 73(2), 345-358.
292
Dawson, G., Osterling, J., Rinaldi, J., Carver, L., & McPartland, J. (2001). Brief report:
Recognition memory and stimulus-reward associations: Indirect support for the
role of ventromedial prefrontal dysfunction in autism. Journal of Autism &
Developmental Disorders, 31(3), 337-341.
Dawson, G., Webb, S., Schellenberg, G. D., Dager, S., Friedman, S., Aylward, E., &
Richards, T. (2002b). Defining the broader phenotype of autism: Genetic, brain,
and behavioral perspectives. Development & Psychopathology, 14(3), 581-611.
de Villiers, J. (2000). Language and theory of mind: What are the developmental
relationships? In S. Baron-Cohen, H. Tager-Flusberg, & D. J. Cohen (Eds.),
Understanding other minds: Perspectives from developmental cognitive
neuroscience (2nd ed., pp. 83-123). London: Oxford University Press.
de Villiers, J. G., & de Villiers, P. A. (2000). Linguistic determinism and the
understanding of false beliefs. In P. Mitchell & K. J. Riggs (Eds.), Children's
reasoning and the mind (pp. 191-228). Hove, UK: Psychology Press.
Deb, S., & Thompson, B. (1998). Neuroimaging in autism. British Journal of
Psychiatry, 173, 299-302.
Delis, D. C., Squire, L. R., Bihrle, A., & Massman, P. J. (1992). Componential analysis
of problem-solving ability: Performance of patients with frontal lobe damage
and amnesic patients on a new sorting test. Neuropsychologia, 30(8), 683-697.
DeLong, G., & Dwyer, J. T. (1988). Correlation of family history with specific autistic
subgroups: Asperger's syndrome and bipolar affective disease. Journal of Autism
& Developmental Disorders, 18(4), 593-600.
Dempster, F. N. (1992). The rise and fall of the inhibitory mechanism: Toward a unified
theory of cognitive development and aging. Developmental Review, 12(1), 45-
75.
Dempster, F. (1993). Resistance to interference: Developmental changes in a basic
processing mechanism. In M. L. Howe & R. Pasnak (Eds.), Emerging themes in
cognitive development (Vol. 1: Foundations, pp. 3-27). New York: Springer-
Verlag.
Dennett, D. C. (1978). Beliefs about beliefs. The Behavioral and Brain Sciences, 4, 568-
570.
Dewey, M. (1991). Living with Asperger's syndrome. In U. Frith (Ed.), Autism and
Asperger syndrome. (pp. 184-206). Cambridge: Cambridge University Press.
Diamond, A. (1985). Development of the ability to use recall to guide action, as
indicated by infants' performance on AB. Child Development, 56(4), 868-883.
293
Diamond, A. (2002). Normal development of prefrontal cortex from birth to young
adulthood: Cognitive functions, anatomy, and biochemistry. In D. T. Stuss & R.
T. Knight (Eds.), Principles of frontal lobe function (pp. 466-503). London:
Oxford University Press.
Diamond, A., & Goldman-Rakic, P. S. (1989). Comparison of human infants and rhesus
monkeys on Piaget's AB task: Evidence for dependence on dorsolateral
prefrontal cortex. Experimental Brain Research, 74, 24-40.
Diamond, A., Prevor, M. B., Callender, G., & Druin, D. P. (1997). Prefrontal cortex
cognitive deficits in children treated early and continuously for PKU.
Monographs of the Society for Research in Child Development, 62(4), 1-205.
Diamond, A., & Taylor, C. (1996). Development of an aspect of executive control:
Development of the abilities to remember what I said and to "Do as I say, not as
I do". Developmental Psychobiology, 29(4), 315-334.
Donnellan, A. M., Anderson, J. L., & Mesaros, R. A. (1984). An observational study of
stereotypic behavior and proximity related to the occurrence of autistic child-
family member interactions. Journal of Autism & Developmental Disorders,
14(2), 205-210.
Dorris, L., Espie, C. A. E., Knott, F., & Salt, J. (2004). Mind-reading difficulties in the
siblings of people with Asperger's syndrome: Evidence for a genetic influence in
the abnormal development of a specific cognitive domain. Journal of Child
Psychology & Psychiatry, 45(2), 412-418.
Downes, J. J., Roberts, A. C., Sahakian, B. J., Evenden, J. L., Morris, R. G., & Robbins,
T. W. (1989). Impaired extra-dimensional shift performance in medicated and
unmedicated Parkinson’s disease: Evidence for a specific attentional
dysfunction. Neuropsychologia, 27, 1329-1343.
Drewe, E. (1974). The effect of type and area of brain lesion on Wisconsin Card Sorting
Test performance. Cortex, 10(2), 159-170.
Drewe, E. A. (1975). An experimental investigation of Luria's theory on the effects of
frontal lobe lesions in man. Neuropsychologia, 13, 421-429.
Duncan, J., Burgess, P., & Emslie, H. (1995). Fluid intelligence after frontal lobe
lesions. Neuropsychologia, 33(3), 261-268.
Duncan, J., Emslie, H., Williams, P., Johnson, R., & Freer, C. (1996). Intelligence and
the frontal lobe: The organization of goal-directed behavior. Cognitive
Psychology, 30(3), 257-303.
294
Eisenberg, L., & Kanner, L. (1956). Early infantile autism, 1943-55. American Journal
of Orthopsychiatry, 26, 556-566.
Eisenmajer, R., & Prior, M. (1991). Cognitive linguistic correlates of "theory of mind"
ability in autistic children. British Journal of Developmental Psychology, 9(2),
351-364.
Elliott, R., McKenna, P., Robbins, T., & Sahakian, B. (1995). Neuropsychological
evidence for frontostriatal dysfunction in schizophrenia. Psychological
Medicine, 25(3), 619-630.
Eslinger, P. J. (1996). Conceptualizing, describing, and measuring components of
executive function: A summary. In G. R. Lyon & N. A. Krasnegor (Eds.),
Attention, memory, and executive function (pp. 367-395). Baltimore, MD: Paul
H. Brookes.
Eslinger, P. J. (1998). Neurological and neuropsychological bases of empathy.
European Neurology, 39(4), 193-199.
Eslinger, P. J., Biddle, K. R., & Grattan, L. M. (1997). Cognitive and social
development in children with prefrontal cortex lesions. In N. A. Krasnegor, G.
R. Lyon, & P. S. Goldman-Rakic (Eds.), Development of the prefrontal cortex:
Evolution, neurobiology, and behavior (pp. 295-335). Baltimore, MD: Paul H.
Brookes.
Eslinger, P. J., & Damasio, A. R. (1985). Severe disturbance of higher cognition after
bilateral frontal lobe ablation: Patient EVR. Neurology, 35(12), 1731-1741.
Eslinger, P. J., Grattan, L. M., Damasio, H., & Damasio, A. R. (1992). Developmental
consequences of childhood frontal lobe damage. Archives of Neurology, 49, 764-
769.
Espy, K. A., Kaufmann, P. M., Glisky, M. L., & McDiarmid, M. (2001). New
procedures to assess executive functions in preschool children. Clinical
Neuropsychologist, 15(1), 46-58.
Espy, K. A., Kaufmann, P. M., McDiarmid, M. D., & Glisky, M. L. (1999). Executive
functioning in preschool children: Performance on A-not-B and other delayed
response format tasks. Brain & Cognition, 41(2), 178-199.
Fein, D., Stevens, M., Dunn, M., Waterhouse, L., Allen, D., Rapin, I., & Feinstein, C.
(1999). Subtypes of pervasive developmental disorder: Clinical characteristics.
Child Neuropsychology, 5(1), 1-23.
295
Fine, C., Lumsden, J., & Blair, R. (2001). Dissociation between "theory of mind" and
executive functions in a patient with early left amygdala damage. Brain, 124(2),
287-298.
Flavell, J. H., Flavell, E. R., & Green, F. L. (1983). Development of the appearance-
reality distinction. Cognitive Psychology, 15(1), 95-120.
Fletcher, P. C., Happé, F., Frith, U., Baker, S. C., Dolan, R. J., Frackowiak, R. S. J., &
Frith, C. D. (1995). Other minds in the brain: A functional imaging study of
"theory of mind" in story comprehension. Cognition, 57, 109-128.
Fodor, J. A. (1983). The modularity of mind. Cambridge, MA: MIT Press.
Folstein, S. E., Bisson, E., Santangelo, S. L., & Piven, J. (1998). Finding specific genes
that cause autism: A combination of approaches will be needed to maximize
power. Journal of Autism & Developmental Disorders, 28(5), 439-445.
Folstein, S., & Rutter, M. (1977). Infantile autism: A genetic study of 21 twin pairs.
Journal of Child Psychology & Psychiatry & Allied Disciplines, 18(4), 297-321.
Folstein, S. E., Santangelo, S. L., Gilman, S. E., Piven, J., Landa, R., Lainhart, J., Hein,
J., & Wzorek, M. (1999). Predictors of cognitive test patterns in autism families.
Journal of Child Psychology & Psychiatry & Allied Disciplines, 40(7), 1117-
1128.
Fombonne, E. (2003). Epidemiological surveys of autism and other pervasive
developmental disorders: An update. Journal of Autism & Developmental
Disorders, 33(4), 365-382.
Fombonne, E., Bolton, P., Prior, J., Jordan, H., & Rutter, M. (1997). A family study of
autism: Cognitive patterns and levels in parents and siblings. Journal of Child
Psychology & Psychiatry & Allied Disciplines, 38(6), 667-683.
Fonagy, P., Steele, M., Steele, H., Leigh, T., Kennedy, R., Mattoon, G., & Target, M.
(1995). Attachment, the reflective self, and borderline states: The predictive
specificity of the Adult Attachment Interview and pathological emotional
development. In S. Goldberg, R. Muir, & J. Kerr (Eds.), Attachment theory:
Social, developmental, and clinical perspectives (pp. 233-278). New York:
Analytic Press.
Freeman, B., Ritvo, E. R., Mason-Brothers, A., Pingree, C., Yokota, A., Jenson, W.,
McMahon, W., Peterson, B., Mo, A., & Schroth, P. (1989). Psychometric
assessment of first-degree relatives of 62 autistic probands in Utah. American
Journal of Psychiatry, 146(3), 361-364.
296
Freeman, N. H., & Lacohée, H. (1995). Making explicit 3-year-olds' implicit
competence with their own false beliefs. Cognition, 56(1), 31-60.
Freeman, N. H., Lewis, C., & Doherty, M. (1991). Preschoolers' grasp of a desire for
knowledge in false-belief prediction: Practical intelligence and verbal report.
British Journal of Developmental Psychology, 9(1), 139-157.
Frith, C. D. (1992). The cognitive neuropsychology of schizophrenia. Hillsdale, NJ:
Lawrence Erlbaum.
Frith, C., & Frith, U. (2000). The physiological basis of theory of mind: Functional
neuroimaging studies. In S. Baron-Cohen, H. Tager-Flusberg, & D. J. Cohen
(Eds.), Understanding other minds: Perspectives from developmental cognitive
neuroscience (2nd ed., pp. 334-356). London: Oxford University Press.
Frith, U. (1972). Cognitive mechanisms in autism: Experiments with color and tone
sequence production. Journal of Autism and Childhood Schizophrenia, 2(2),
160-173.
Frith, U., & Happé, F. (1994). Autism: Beyond "theory of mind". Cognition, 50(1-3),
115-132.
Frith, U., Happé, F., & Siddons, F. (1994). Autism and theory of mind in everyday life.
Social Development, 3(2), 108-124.
Frith, U., Morton, J., & Leslie, A. M. (1991). The cognitive basis of a biological
disorder: Autism. Trends in Neurosciences, 14, 433-438.
Frye, D. (1999). Development of intention: The relation of executive function to theory
of mind. In P. D. Zelazo, J. W. Astington, & D. Olson (Eds.), Developing
theories of intention: Social understanding and self-control (pp. 119-132).
Mahwah, NJ: Lawrence Erlbaum Associates.
Frye, D. (2000). Theory of mind, domain specificity, and reasoning. In P. Mitchell & K.
J. Riggs (Eds.), Children's reasoning and the mind (pp. 149-167). Hove, UK:
Psychology Press.
Frye, D., & Zelazo, P. D. (1998). Complexity: From formal analysis to final action.
Behavioural and Brain Sciences, 21(6), 836-837.
Frye, D., Zelazo, P. D., Brooks, P. J., & Samuels, M. C. (1996). Inference and action in
early causal reasoning. Developmental Psychology, 32(1), 120-131.
Frye, D., Zelazo, P. D., & Burack, J. A. (1998). Cognitive complexity and control: I.
Theory of mind in typical and atypical development. Current Directions in
Psychological Science, 7(4), 116-121.
297
Frye, D., Zelazo, P. D., & Palfai, T. (1995). Theory of mind and rule-based reasoning.
Cognitive Development, 10(4), 483-527.
Fuster, J. M. (2000). Prefrontal neurons in networks of executive memory. Brain
Research Bulletin, 52(5), 331-336.
Gallagher, H. L., & Frith, C. D. (2003). Functional imaging of 'theory of mind'. Trends
in Cognitive Sciences, 7(2), 77-83.
Gallagher, H., Happé, F., Brunswick, N., Fletcher, P., Frith, U., & Frith, C. (2000).
Reading the mind in cartoons and stories: an fMRI study of 'theory of the mind'
in verbal and nonverbal tasks. Neuropsychologia, 38(1), 11-21.
Garcia-Villamisar, D., & Della Sala, S. (2002). Dual-task performance in adults with
autism. Cognitive Neuropsychiatry, 7(1), 63-74.
Garner, C., Callias, M., & Turk, J. (1999). Executive function and theory of mind
performance of boys with fragile-X syndrome. Journal of Intellectual Disability
Research, 43(6), 466-474.
Garretson, H. B., Fein, D., & Waterhouse, L. (1990). Sustained attention in children
with autism. Journal of Autism & Developmental Disorders, 20(1), 101-114.
George, M. S., Costa, D. C., Kouris, K., Ring, H. A., & Ell, P. J. (1992). Cerebral blood
flow abnormalities in adults with infantile autism. Journal of Nervous & Mental
Disease, 180(7), 413-417.
German, T. P., & Leslie, A. M. (2000). Attending to and learning about mental states. In
P. Mitchell & K. J. Riggs (Eds.), Children's reasoning and the mind (pp. 229-
252). Hove, UK: Psychology Press.
Gerstadt, C. L., Hong, Y. J., & Diamond, A. (1994). The relationship between cognition
and action: Performance of children 31/2-7 years old on a Stroop-like day-night
test. Cognition, 53(2), 129-153.
Gillberg, C., & Coleman, M. (1992). The biology of the autistic syndromes. (2nd ed.).
London: Mac Keith Press.
Gilotty, L., Kenworthy, L., Sirian, L., Black, D. O., & Wagner, A. E. (2002). Adaptive
skills and executive function in autism spectrum disorders. Child
Neuropsychology, 8(4), 241-248.
Godefroy, O., Cabaret, M., Petit-Chenal, V., Pruvo, J.-P., & Rousseaux, M. (1999).
Control functions of the frontal lobes: Modularity of the central-supervisory
system? Cortex, 35(1), 1-20.
Goel, V., Grafman, J., Sadato, N., & Hallett, M. (1995). Modeling other minds.
Neuroreport, 6(13), 1741-1746.
298
Goldberg, M., Lasker, A., Zee, D., Garth, E., Tien, A., & Landa, R. (2002). Deficits in
the initiation of eye movements in the absence of a visual target in adolescents
with high functioning autism. Neuropsychologia, 40(12), 2039-2049.
Golden, C. J. (1981). The Luria-Nebraska Children's Battery: Theory and formulation.
In G. W. Hynd & J. E. Obrzut (Eds.), Neuropsychological assessment of the
school-aged child (pp. 277-302). New York: Grune & Stratton.
Goldman-Rakic, P. S. (1995). Architecture of the prefrontal cortex and the central
executive. In J. Grafman, K. J. Holyoak, & F. Boller (Eds.), Structure and
functions of the human prefrontal cortex (pp. 71-83). New York: New York
Academy of Sciences.
Goldman-Rakic, P. S., & Leung, H.-C. (2002). Functional architecture of the
dorsolateral prefrontal cortex in monkeys and humans. In D. T. Stuss & R. T.
Knight (Eds.), Principles of frontal lobe function (pp. 85-95). London: Oxford
University Press.
Goldstein, G., Johnson, C. R., & Minshew, N. J. (2001). Attentional processes in
autism. Journal of Autism & Developmental Disorders, 31(4), 433-440.
Goodman, R. (1989). Infantile autism: A syndrome of multiple primary deficits?
Journal of Autism & Developmental Disorders, 19(3), 409-424.
Gopnik, A. (1993). How we know our minds: The illusion of first-person knowledge of
intentionality. Behavioral & Brain Sciences, 16(1), 1-14, 29-113.
Gopnik, A., & Astington, J. W. (1988). Children's understanding of representational
change and its relation to the understanding of false belief and the appearance-
reality distinction. Child Development, 59(1), 26-37.
Gopnik, A., & Meltzoff, A. N. (1997). Words, thoughts, and theories. Cambridge, MA:
MIT Press.
Gopnik, A., & Wellman, H. M. (1994). The theory theory. In L. A. Hirschfeld & S. A.
Gelman (Eds.), Mapping the mind: Domain specificity in cognition and culture
(pp. 257-293). Cambridge: Cambridge University Press.
Gordon, A. C. L., & Olson, D. R. (1998). The relation between acquisition of a theory
of mind and the capacity to hold in mind. Journal of Experimental Child
Psychology, 68(1), 70-83.
Gottesman, I. I., & Gould, T. D. (2003). The endophenotype concept in psychiatry:
Etymology and strategic intentions. American Journal of Psychiatry, 160(4),
636-645.
299
Grafman, J. (1994). Alternative frameworks for the conceptualization of prefrontal lobe
functions. In F. Boller & J. Grafman (Eds.), Handbook of Neuropsychology
(Vol. 9, pp. 187-201). Amsterdam: Elsevier Science.
Grant, C. M., Grayson, A., & Boucher, J. (2001). Using tests of false belief with
children with autism: How valid and reliable are they? Autism, 5(2), 135-145.
Grant, D. A., & Berg, E. (1948). A behavioral analysis of degree of reinforcement and
ease of shifting to new responses in a Weigl-type card-sorting problem. Journal
of Experimental Psychology, 38, 404-411.
Grattan, L. M., Bloomer, R. H., Archambault, F. X., & Eslinger, P. J. (1994). Cognitive
flexibility and empathy after frontal lobe lesion. Neuropsychiatry,
Neuropsychology, and Behavioral Neurology, 7(4), 251-259.
Gregory, C., Lough, S., Stone, V., Erzinclioglu, S., Martin, L., Baron-Cohen, S., &
Hodges, J. R. (2002). Theory of mind in patients with frontal variant
frontotemporal dementia and Alzheimer's disease: Theoretical and practical
implications. Brain, 125(4), 752-764.
Griffith, E. M., Pennington, B. F., Wehner, E. A., & Rogers, S. J. (1999). Executive
functions in young children with autism. Child Development, 70(4), 817-832.
Grodzinsky, G. M., & Diamond, R. (1992). Frontal lobe functioning in boys with
attention-deficit hyperactivity disorder. Developmental Neuropsychology, 8(4),
427-445.
Grossman, M. (2002). Frontotemporal dementia: A review. Journal of the International
Neuropsychological Society, 8(4), 566-583.
Hala, S., Hug, S., & Henderson, A. (2003). Executive function and false-belief
understanding in preschool children: Two tasks are harder than one. Journal of
Cognition & Development, 4(3), 275-298.
Hala, S., & Russell, J. (2001). Executive control within strategic deception: A window
on early cognitive development? Journal of Experimental Child Psychology,
80(2), 112-141.
Halford, G. S. (1993). Children's understanding: The development of mental models.
Hillsdale, NJ: Lawrence Erlbaum.
Halford, G. S., Wilson, W. H., & Phillips, S. (1998). Processing capacity defined by
relational complexity: Implications for comparative, developmental, and
cognitive psychology. Behavioral & Brain Sciences, 21(6), 803-864.
Happé, F. G. E. (1994a). An advanced test of theory of mind: Understanding of story
characters' thoughts and feelings by able autistic, mentally handicapped, and
300
normal children and adults. Journal of Autism & Developmental Disorders,
24(2), 129-154.
Happé, F. G. E. (1994b). Annotation: Current psychological theories of autism: The
"Theory of Mind" account and rival theories. Journal of Child Psychology &
Psychiatry & Allied Disciplines, 35(2), 215-229.
Happé, F. G. E. (1994c). Wechsler IQ profile and theory of mind in autism: A research
note. Journal of Child Psychology & Psychiatry & Allied Disciplines, 35(8),
1461-1471.
Happé, F. G. (1995). The role of age and verbal ability in the theory of mind task
performance of subjects with autism. Child Development, 66(3), 843-855.
Happé, F. G. E. (1996). Studying weak central coherence at low levels: Children with
autism do not succumb to visual illusions. Journal of Child Psychology and
Psychiatry, 37(7), 873-877.
Happé, F. G. E. (1997). Central coherence and theory of mind in autism: Reading
homographs in context. British Journal of Developmental Psychology, 15, 1-12.
Happé, F. (1999). Understanding assets and deficits in autism: Why success is more
interesting than failure. Psychologist, 12(11), 540-546.
Happé, F. (2000). Parts and wholes, meaning and minds: Central coherence and its
relation to theory of mind. In S. Baron-Cohen, H. Tager-Flusberg, & D. J.
Cohen (Eds.), Understanding other minds: Perspectives from developmental
cognitive neuroscience (2nd ed., pp. 203-221). London: Oxford University
Press.
Happé, F. (2001). Social and nonsocial development in autism: Where are the links? In
J. A. Burack, T. Charman, N. Yirmiya, & P. R. Zelazo (Eds.), The development
of autism: Perspectives from theory and research (pp. 237-253). Mahwah, NJ:
Lawrence Erlbaum Associates.
Happé, F., Briskman, J., & Frith, U. (2001). Exploring the cognitive phenotype of
autism: Weak "central coherence" in parents and siblings of children with
autism: I. Experimental tests. Journal of Child Psychology & Psychiatry &
Allied Disciplines, 42(3), 299-307.
Happé, F., Ehlers, S., Fletcher, P., Frith, U., Johansson, M., Gillberg, C., Dolan, R.,
Frackowiak, R., & Frith, C. (1996). 'Theory of mind' in the brain. Evidence from
a PET scan study of Asperger syndrome. Neuroreport, 8(1), 197-201.
301
Happé, F., & Frith, U. (1995). Theory of mind in autism. In E. Schopler & G. B.
Mesibov (Eds.), Learning and Cognition in Autism (pp. 177-197). New York:
Plenum Press.
Happé, F., & Frith, U. (1996). The neuropsychology of autism. Brain, 119, 1377-1400.
Happé, F., Malhi, G. S., & Checkley, S. (2001). Acquired mind-blindness following
frontal lobe surgery? A single case study of impaired 'theory of mind' in a
patient treated with stereotactic anterior capsulotomy. Neuropsychologia, 39(1),
83-90.
Harnishfeger, K. K., & Bjorklund, D. F. (1993). The ontogeny of inhibition
mechanisms: A renewed approach to cognitive development. In M. L. Howe &
R. Pasnak (Eds.), Emerging themes in cognitive development (Vol. 1:
Foundations, pp. 28-49). New York: Springer-Verlag.
Harris, P. (1993). Pretending and planning. In S. Baron-Cohen, H. Tager-Flusberg, & D.
J. Cohen (Eds.), Understanding other minds: Perspectives from autism (pp. 228-
246). Oxford: Oxford University Press.
Harris, P. L., & Leevers, H. J. (2000). Pretending, imagery, and self-awareness in
autism. In S. Baron-Cohen, H. Tager-Flusberg, & D. J. Cohen (Eds.),
Understanding other minds: Perspectives from developmental cognitive
neuroscience (2nd ed., pp. 182-202). London: Oxford University Press.
Head, D., Bolton, D., & Hymas, N. (1989). Deficit in cognitive shifting ability in
patients with obsessive-compulsive disorder. Biological Psychiatry, 25(7), 929-
937.
Heaton, R. K., Grant, I., & Matthews, C. G. (1991). Comprehensive norms for an
expanded Halstead-Reitan battery: Demographic corrections, research findings,
and clinical applications. Odessa, FL: Psychological Assessment Resources.
Heilman, K., Watson, R., & Valenstein, E. (1993). Neglect and related disorders. In K.
Heilman & E. Valenstein (Eds.), Clinical Neuropsychology (3rd ed., pp. 279-
336). New York: Oxford University Press.
Hermelin, B., & O'Connor, N. (1970). Psychological experiments with autistic children.
New York: Pergamon.
Hill, E. L. (2004). Executive dysfunction in autism. Trends in Cognitive Sciences, 8(1),
26-32.
Hill, E. L., & Frith, U. (2003). Understanding autism: Insights from mind and brain.
Philosophical Transactions of the Royal Society of London B, 358(1430), 281-
289.
302
Hill, E. L., & Russell, J. (2002). Action memory and self-monitoring in children with
autism: Self versus other. Infant & Child Development, 11(2), 159-170.
Hobson, R. P. (1989). Beyond cognition: A theory of autism. In G. Dawson (Ed.),
Autism: Nature, diagnosis, and treatment (pp. 22-48). New York: The Guilford
Press.
Hobson, R. P. (1993). Understanding persons: The role of affect. In S. Baron-Cohen, H.
Tager-Flusberg, & D. J. Cohen (Eds.), Understanding other minds: Perspectives
from autism (pp. 204-227). Oxford: Oxford University Press.
Hoff, A. L., & Kremen, W. S. (2003). Neuropsychology in schizophrenia: An update.
Current Opinion in Psychiatry, 16(2), 149-155.
Hollander, E., King, A., Delaney, K., Smith, C. J., & Silverman, J. M. (2003).
Obsessive-compulsive behaviors in parents of multiplex autism families.
Psychiatry Research, 117(1), 11-16.
Holroyd, S., & Baron-Cohen, S. (1993). Brief report: How far can people with autism
go in developing a theory of mind? Journal of Autism & Developmental
Disorders, 23(2), 379-385.
Hughes, C. (1996a). Brief report: Planning problems in autism at the level of motor
control. Journal of Autism & Developmental Disorders, 26(1), 99-107.
Hughes, C. (1996b). Control of action and thought: Normal development and
dysfunction in autism: A research note. Journal of Child Psychology &
Psychiatry & Allied Disciplines, 37(2), 229-236.
Hughes, C. (1998a). Executive function in preschoolers: Links with theory of mind and
verbal ability. British Journal of Developmental Psychology, 16, 233-253.
Hughes, C. (1998b). Finding your marbles: Does preschoolers' strategic behavior
predict later understanding of mind? Developmental Psychology, 34(6), 1326-
1339.
Hughes, C. (2001). Executive dysfunction in autism: Its nature and implications for the
everyday problems experienced by individuals with autism. In J. A. Burack, T.
Charman, N. Yirmiya, & P. R. Zelazo (Eds.), The development of autism:
Perspectives from theory and research (pp. 255-275). Mahwah, NJ: Lawrence
Erlbaum Associates.
Hughes, C., Adlam, A., Happé, F., Jackson, J., Taylor, A., & Caspi, A. (2000). Good
test-retest reliability for standard and advanced false-belief tasks across a wide
range of abilities. Journal of Child Psychology & Psychiatry & Allied
Disciplines, 41(4), 483-490.
303
Hughes, C., Dunn, J., & White, A. (1998). Trick or treat? Uneven understanding of
mind and emotion and executive dysfunction in "hard-to-manage" preschoolers.
Journal of Child Psychology and Psychiatry, 39(7), 981-994.
Hughes, C., & Graham, A. (2002). Measuring executive functions in childhood:
Problems and solutions? Child & Adolescent Mental Health, 7(3), 131-142.
Hughes, C., Leboyer, M., & Bouvard, M. (1997). Executive function in parents of
children with autism. Psychological Medicine, 27, 209-220.
Hughes, C., Plumet, M.-H., & Leboyer, M. (1999). Towards a cognitive phenotype for
autism: Increased prevalence of executive dysfunction and superior spatial span
amongst siblings of children with autism. Journal of Child Psychology &
Psychiatry & Allied Disciplines, 40(5), 705-718.
Hughes, C., & Russell, J. (1993). Autistic children's difficulty with mental
disengagement from an object: Its implications for theories of autism.
Developmental Psychology, 29(3), 498-510.
Hughes, C., Russell, J., & Robbins, T. W. (1994). Evidence for executive dysfunction in
autism. Neuropsychologia, 32(4), 477-492.
Hughes, C., Soares-Boucaud, I., Hochmann, J., & Frith, U. (1997). Social behaviour in
pervasive developmental disorders: Effects of informant, group and "theory of
mind". European Child & Adolescent Psychiatry, 6(4), 191-198.
Hutt, S., & Hutt, C. (1968). Stereotypy, arousal and autism. Human Development,
11(4), 277-286.
Huttenlocher, P. R., & Dabholkar, A. S. (1997). Developmental anatomy of prefrontal
cortex. In N. A. Krasnegor, G. R. Lyon, & P. S. Goldman-Rakic (Eds.),
Development of the prefrontal cortex: Evolution, neurobiology and behavior
(pp. 69-83). Baltimore, MD: Paul H. Brookes.
Jacques, S., Zelazo, P. D., Kirkham, N. Z., & Semcesen, T. K. (1999). Rule selection
versus rule execution in preschoolers: An error-detection approach.
Developmental Psychology, 35(3), 770-780.
Jarrold, C., Boucher, J., & Smith, P. K. (1994a). Executive function deficits and the
pretend play of children with autism: A research note. Journal of Child
Psychology & Psychiatry & Allied Disciplines, 35(8), 1473-1482.
Jarrold, C., Boucher, J., & Smith, P. K. (1996). Generativity defects in pretend play in
autism. British Journal of Developmental Psychology, 14(3), 275-300.
304
Jarrold, C., Butler, D. W., Cottington, E. M., & Jimenez, F. (2000). Linking theory of
mind and central coherence bias in autism and in the general population.
Developmental Psychology, 36(1), 126-138.
Jarrold, C., Smith, P., Boucher, J., & Harris, P. (1994b). Comprehension of pretense in
children with autism. Journal of Autism & Developmental Disorders, 24(4),
433-455.
Jenkins, J. M., & Astington, J. W. (1996). Cognitive factors and family structure
associated with theory of mind development in young children. Developmental
Psychology, 32(1), 70-78.
Johnson, M. H., Siddons, F., Frith, U., & Morton, J. (1992). Can autism be predicted on
the basis of infant screening tests? Developmental Medicine & Child Neurology,
34(4), 316-320.
Jolliffe, T., & Baron-Cohen, S. (2000). Linguistic processing in high-functioning adults
with autism or Asperger's syndrome. Is global coherence impaired?
Psychological Medicine, 30(5), 1169-1187.
Kain, W., & Perner, J. (2003). Do children with ADHD not need their frontal lobes for
theory of mind? A review of brain imaging and neuropsychological studies. In
M. Brüne, H. Ribbert, & W. Schiefenhövel (Eds.), The social brain: Evolution
and pathology (pp. 197-230). Chichester, UK: John Wiley & Sons.
Kane, M. J., & Engle, R. W. (2002). The role of prefrontal cortex in working-memory
capacity, executive attention, and general fluid intelligence: An individual-
differences perspective. Psychonomic Bulletin & Review, 9(4), 637-671.
Karmiloff-Smith, A. (1992). Beyond modularity: A developmental perspective on
cognitive science. Cambridge, MA: The MIT Press.
Karmiloff-Smith, A. (1997). Crucial differences between developmental cognitive
neuroscience and adult neuropsychology. Developmental Neuropsychology,
13(4), 513-524.
Karmiloff-Smith, A., Scerif, G., & Ansari, D. (2003). Double dissociations in
developmental disorders? Theoretically misconceived, empirically dubious.
Cortex, 39(1), 161-163.
Karmiloff-Smith, A., Scerif, G., & Thomas, M. (2002). Different approaches to relating
genotype to phenotype in developmental disorders. Developmental
Psychobiology, 40(3), 311-322.
Keenan, T. (1998). Memory span as a predictor of false belief understanding. New
Zealand Journal of Psychology, 27(2), 36-43.
305
Keenan, T., Olson, D. R., & Marini, Z. (1998). Working memory and children's
developing understanding of mind. Australian Journal of Psychology, 50(2), 76-
82.
Kerr, N., Dunbar, R. I., & Bentall, R. P. (2003). Theory of mind deficits in bipolar
affective disorder. Journal of Affective Disorders, 73(3), 253-259.
Kimberg, D. Y., & Farah, M. J. (1993). A unified account of cognitive impairments
following frontal lobe damage: The role of working memory in complex,
organized behaviour. Journal of Experimental Psychology: General, 122, 411-
428.
Kleinman, J., Marciano, P. L., & Ault, R. L. (2001). Advanced theory of mind in high-
functioning adults with autism. Journal of Autism and Developmental Disorders,
31(1), 29-36.
Klin, A., & Volkmar, F. (1993). The development of individuals with autism:
Implications for the theory of mind hypothesis. In S. Baron-Cohen, H. Tager-
Flusberg, & D. J. Cohen (Eds.), Understanding other minds: Perspectives from
autism (pp. 317-331). Oxford: Oxford University Press.
Klin, A., Volkmar, F. R., & Sparrow, S. S. (1992). Autistic social dysfunction: Some
limitations of the theory of mind hypothesis. Journal of Child Psychology &
Psychiatry & Allied Disciplines, 33(5), 861-876.
Klin, A., Volkmar, F., Sparrow, S., Cicchetti, D., & Rourke, B. (1995). Validity and
neuropsychological characterization of Asperger syndrome: Convergence with
nonverbal learning disabilities syndrome. Journal of Child Psychology &
Psychiatry & Allied Disciplines, 36(7), 1127-1140.
Kochanska, G., Murray, K., & Coy, K. C. (1997). Inhibitory control as a contributor to
conscience in childhood: From toddler to early school age. Child Development,
68(2), 263-277.
Koenig, K., Tsatsanis, K. D., & Volkmar, F. R. (2001). Neurobiology and genetics of
autism: A developmental perspective. In J. A. Burack, T. Charman, N. Yirmiya,
& P. R. Zelazo (Eds.), The development of autism: Perspectives from theory and
research (pp. 81-101). Mahwah, NJ: Lawrence Erlbaum Associates.
Krikorian, R., Bartok, J., & Gay, N. (1994). Tower of London procedure: A standard
method and developmental data. Journal of Clinical & Experimental
Neuropsychology, 16(6), 840-850.
306
Landa, R., Piven, J., Wzorek, M. M., Gayle, J. O., Chase, G. A., & Folstein, S. E.
(1992). Social language use in parents of autistic individuals. Psychological
Medicine, 22(1), 245-54.
Lang, B., & Perner, J. (2002). Understanding of intention and false belief and the
development of self-control. British Journal of Developmental Psychology,
20(1), 67-76.
Le Couteur, A., Bailey, A., Goode, S., Pickles, A., Robertson, S., Gottesman, I., &
Rutter, M. (1996). A broader phenotype of autism: The clinical spectrum in
twins. Journal of Child Psychology & Psychiatry & Allied Disciplines, 37(7),
785-801.
Le Couteur, A., Rutter, M., Lord, C., Rios, P., Robertson, S., Holdgrafer, M., &
McLennan, J. D. (1989). Autism Diagnostic Interview: A semi-structured
interview for parents and caregivers of autistic persons. Journal of Autism &
Developmental Disorders, 19(3), 363-387.
Leboyer, M., Bellivier, F., Nosten-Bertrand, M., Jouvent, R., Pauls, D., & Mallet, J.
(1998). Psychiatric genetics: Search for phenotypes. Trends in Neurosciences,
21(3), 102-105.
Leboyer, M., Plumet, M.-H., Goldblum, M.-C., Perez-Diaz, F., & Marchaland, C.
(1995). Verbal versus visuospatial abilities in relatives of autistic females.
Developmental Neuropsychology, 11(1), 139-155.
Leekam, S. R., & Moore, C. (2001). The development of attention and joint attention in
children with autism. In J. A. Burack, T. Charman, N. Yirmiya, & P. R. Zelazo
(Eds.), The development of autism: Perspectives from theory and research (pp.
105-129). Mahwah, NJ: Lawrence Erlbaum Associates.
Leekam, S. R., & Perner, J. (1991). Does the autistic child have a metarepresentational
deficit? Cognition, 40(3), 203-218.
Leekam, S. R., & Prior, M. (1994). Can autistic children distinguish lies from jokes? A
second look at second-order belief attribution. Journal of Child Psychology &
Psychiatry & Allied Disciplines, 35(5), 901-915.
Leslie, A. M. (1987). Pretense and representation: The origins of "theory of mind".
Psychological Review, 94(4), 412-426.
Leslie, A. M. (1991). The theory of mind impairment in autism: Evidence for a modular
mechanism of development? In A. Whiten (Ed.), Natural theories of mind:
Evolution, development and simulation of everyday mindreading (pp. 63-78).
Oxford: Blackwell.
307
Leslie, A. M. (1994a). Pretending and believing: Issues in the theory of ToMM.
Cognition, 50(1-3), 211-238.
Leslie, A. M. (1994b). ToMM, ToBy, and Agency: Core architecture and domain
specificity. In L. A. Hirschfeld & S. A. Gelman (Eds.), Mapping the mind:
Domain specificity in cognition and culture (pp. 119-148). Cambridge:
Cambridge University Press.
Leslie, A. M., & Frith, U. (1988). Autistic children's understanding of seeing, knowing
and believing. British Journal of Developmental Psychology, 6, 315-324.
Leslie, A. M., & Happé, F. (1989). Autism and ostensive communication: The relevance
of metarepresentation. Development & Psychopathology, 1(3), 205-212.
Leslie, A. M., & Polizzi, P. (1998). Inhibitory processing in the false belief task: Two
conjectures. Developmental Science, 1(2), 247-253.
Leslie, A., & Roth, D. (1993). What autism teaches us about metarepresentation. In S.
Baron-Cohen, H. Tager-Flusberg, & D. J. Cohen (Eds.), Understanding other
minds: Perspectives from autism (pp. 83-111). Oxford: Oxford University Press.
Leslie, A. M., & Thaiss, L. (1992). Domain specificity in conceptual development:
Neuropsychological evidence from autism. Cognition, 43(3), 225-251.
Levin, H. S., Culhane, K. A., Hartmann, J., Evankovich, K., Mattson, A. J., Harward,
H., Ringholz, G., Ewing-Cobbs, L., & Fletcher, J. M. (1991). Developmental
changes in performance on tests of purported frontal lobe functioning.
Developmental Neuropsychology, 7(3), 377-395.
Levin, H. S., Fletcher, J. M., Kufera, J. A., Lilly, M. A., Mendelsohn, D., Bruce, D., &
Eisenberg, H. M. (1996). Dimensions of cognition measured by the Tower of
London and other cognitive tasks in head-injured children and adolescents.
Developmental Neuropsychology, 12(1), 17-34.
Levin, H. S., Mendelsohn, D. B., Lilly, M. A., Fletcher, J. M., Culhane, K. A.,
Chapman, S. B., Harward, H., Kusnerik, L., Bruce, D., & Eisenberg, H. M.
(1994). Tower of London performance in relation to Magnetic Resonance
Imaging following closed head injury in children. Neuropsychology, 8(2), 171-
179.
Levine, B., Stuss, D. T., Milberg, W. P., Alexander, M. P., Schwartz, M., &
MacDonald, R. (1998). The effects of focal and diffuse brain damage on strategy
application: Evidence from focal lesions, traumatic brain injury and normal
aging. Journal of the International Neuropsychological Society, 4(3), 247-264.
308
Lewis, C., & Mitchell, P. (Eds.). (1994). Children's early understanding of mind:
Origins and development. Hove, UK: Lawrence Erlbaum.
Lewis, V., & Boucher, J. (1988). Spontaneous, instructed and elicited play in relatively
able autistic children. British Journal of Developmental Psychology, 6(4), 325-
339.
Lewis, V., & Boucher, J. (1991). Skill, content and generative strategies in autistic
children's drawings. British Journal of Developmental Psychology, 9(3), 393-
416.
Lewis, V., & Boucher, J. (1995). Generativity in the play of young people with autism.
Journal of Autism & Developmental Disorders, 25(2), 105-121.
Lezak, M. D. (1993). Newer contributions to the neuropsychological assessment of
executive functions. Journal of Head Trauma Rehabilitation, 8(1), 24-31.
Lezak, M. D. (1995). Neuropsychological assessment. (3rd ed.). New York: Oxford
University Press.
Liss, M., Fein, D., Allen, D., Dunn, M., Feinstein, C., Morris, R., Waterhouse, L., &
Rapin, I. (2001). Executive functioning in high-functioning children with
autism. Journal of Child Psychology & Psychiatry & Allied Disciplines, 42(2),
261-270.
Lockyer, L., & Rutter, M. (1969). A five- to fifteen-year follow-up study of infantile
psychosis: III. Psychological aspects. British Journal of Psychiatry, 115(525),
865-882.
Lockyer, L., & Rutter, M. (1970). A five- to fifteen-year follow-up study of infantile
psychosis: IV. Patterns of cognitive ability. British Journal of Social & Clinical
Psychology, 9(2), 152-163.
Lord, C., Risi, S., Lambrecht, L., Cook, E. H., Leventhal, B. L., DiLavore, P. C.,
Pickles, A., & Rutter, M. (2000). The Autism Diagnostic Observation Schedule-
Generic: A standard measure of social and communication deficits associated
with the spectrum of autism. Journal of Autism & Developmental Disorders,
30(3), 205-223.
Lord, C., Rutter, M., & Le Couteur, A. (1994). Autism Diagnostic Interview--Revised:
A revised version of a diagnostic interview for caregivers of individuals with
possible pervasive developmental disorders. Journal of Autism & Developmental
Disorders, 24(5), 659-685.
309
Lorr, M. (1994). Cluster analysis: Aims, methods, and problems. In S. Strack & M. Lorr
(Eds.), Differentiating normal and abnormal personality (pp. 179-195). New
York: Springer Publishing Co.
Lough, S., Gregory, C., & Hodges, J. R. (2001). Dissociation of social cognition and
executive function in frontal variant frontotemporal dementia. Neurocase,
7(2,Pt2), 123-130.
Lough, S., & Hodges, J. R. (2002). Measuring and modifying abnormal social cognition
in frontal variant frontotemporal dementia. Journal of Psychosomatic Research,
53(2), 639-646.
Lowe, C., & Rabbitt, P. (1998). Test/re-test reliability of the CANTAB and ISPOCD
neuropsychological batteries: Theoretical and practical issues.
Neuropsychologia, 36(9), 915-923.
Luciana, M., & Nelson, C. A. (1998). The functional emergence of prefrontally-guided
working memory systems in four- to eight-year-old children. Neuropsychologia,
36(3), 273-293.
Luna, B., Minshew, N., Garver, K., Lazar, N., Thulborn, K., Eddy, W., & Sweeney, J.
(2002). Neocortical system abnormalities in autism: An fMRI study of spatial
working memory. Neurology, 59(6), 834-840.
Luria, A. R. (1966). Higher cortical functions in man. New York: Basic Books.
Macintosh, K. E., & Dissanayake, C. (2004). Annotation: The similarities and
differences between autistic disorder and Asperger's disorder: A review of the
empirical evidence. Journal of Child Psychology and Psychiatry, 45(3), 421-
434.
MacLean, J. E., Szatmari, P., Jones, M. B., Bryson, S. E., Mahoney, W. J., Bartolucci,
G., & Tuff, L. (1999). Familial factors influence level of functioning in
pervasive developmental disorder. Journal of the American Academy of Child &
Adolescent Psychiatry, 38(6), 746-753.
Manly, T., Anderson, V., Nimmo-Smith, I., Turner, A., Watson, P., & Robertson, I. H.
(2001). The differential assessment of children's attention: The Test of Everyday
Attention for Children (TEA-Ch), normative sample and ADHD performance.
Journal of Child Psychology & Psychiatry & Allied Disciplines, 42(8), 1065-
1081.
Manly, T., Robertson, I. H., Anderson, V., & Nimmo-Smith, I. (1998). The Test of
Everyday Attention for Children (TEA-Ch). Thames Valley Test Company.
310
Marlowe, W. B. (1992). The impact of a right prefrontal lesion on the developing brain.
Brain & Cognition, 20(1), 205-213.
Maxwell, S. E., & Delaney, H. D. (1990). Designing experiments and analysing data: A
model comparison perspective. Belmont, CA: Wadsworth.
Mayes, L. C., Klin, A., Tercyak, K. P., Cicchetti, D. V., & Cohen, D. J. (1996). Test-
retest reliability for false-belief tasks. Journal of Child Psychology & Psychiatry
& Allied Disciplines, 37(3), 313-319.
Mazza, M., De Risio, A., Surian, L., Roncone, R., & Casacchia, M. (2001). Selective
impairments of theory of mind in people with schizophrenia. Schizophrenia
Research, 47(2-3), 299-308.
McEvoy, R. E., Rogers, S. J., & Pennington, B. F. (1993). Executive function and social
communication deficits in young autistic children. Journal of Child Psychology
& Psychiatry & Allied Disciplines, 34(4), 563-578.
Mega, M. S., & Cummings, J. L. (1994). Frontal-subcortical circuits and
neuropsychiatric disorders. Journal of Neuropsychiatry & Clinical
Neurosciences, 6(4), 358-370.
Mengelberg, A., & Siegert, R. J. (2003). Is theory-of-mind impaired in Parkinson's
disease? Cognitive Neuropsychiatry, 8(3), 191-209.
Miller, G. A., & Chapman, J. P. (2001). Misunderstanding analysis of covariance.
Journal of Abnormal Psychology, 110, 40-48.
Miller, J. N., & Ozonoff, S. (2000). The external validity of Asperger disorder: Lack of
evidence from the domain of neuropsychology. Journal of Abnormal
Psychology, 109(2), 227-238.
Milner, B. (1963). Effects of different brain lesions on card sorting. Archives of
Neurology, 9, 90-100.
Minshew, N. J., Goldstein, G., Muenz, L. R., & Payton, J. B. (1992).
Neuropsychological functioning nonmentally retarded autistic individuals.
Journal of Clinical & Experimental Neuropsychology, 14(5), 749-761.
Minshew, N. J., Johnson, C., & Luna, B. (2001). The cognitive and neural basis of
autism: A disorder of complex information processing and dysfunction of
neocortical systems. In L. M. Glidden (Ed.), International review of research in
mental retardation: Autism (Vol. 23, pp. 111-138). San Diego, CA: Academic
Press.
311
Minshew, N. J., Luna, B., & Sweeney, J. A. (1999). Oculomotor evidence for
neocortical systems but not cerebellar dysfunction in autism. Neurology, 52(5),
917-922.
Minter, M., Hobson, R., & Bishop, M. (1998). Congenital visual impairment and 'theory
of mind'. British Journal of Developmental Psychology, 16(2), 183-196.
Minton, J., Campbell, M., Green, W. H., Jennings, S., & Samit, C. (1982). Cognitive
assessment of siblings of autistic children. Journal of the American Academy of
Child Psychiatry, 21(3), 256-261.
Mitchell, P., & Lacohée, H. (1991). Children's early understanding of false belief.
Cognition, 39, 107-127.
Mitchell, P., & Riggs, K. J. (Eds.). (2000). Children's reasoning and the mind. Hove,
UK: Psychology Press.
Miyake, A., Friedman, N. P., Emerson, M. J., Witzki, A. H., & Howerter, A. (2000).
The unity and diversity of executive functions and their contributions to
complex "frontal lobe" tasks: A latent variable analysis. Cognitive Psychology,
41(1), 49-100.
Moore, C., Jarrold, C., Russell, J., Lumb, A., Sapp, F., & MacCallum, F. (1995).
Conflicting desire and the child's theory of mind. Cognitive Development, 10(4),
467-482.
Morton, J., & Frith, U. (1995). Causal modeling: A structural approach to
developmental psychopathology. In D. Cicchetti & D. J. Cohen (Eds.),
Developmental psychopathology (Vol. 1: Theory and methods, pp. 357-390).
Oxford: John Wiley & Sons.
Morton, J., & Frith, U. (2001). Why we need cognition: Cause and developmental
disorder. In E. Dupoux (Ed.), Language, brain, and cognitive development:
Essays in honor of Jacques Mehler (pp. 263-278). Cambridge, MA: MIT Press.
Moses, L. J., & Flavell, J. H. (1990). Inferring false beliefs from actions and reactions.
Child Development, 61(4), 929-945.
Mundy, P. (2003). The neural basis of social impairments in autism: The role of the
dorsal medial-frontal cortex and anterior cingulate system. Journal of Child
Psychology & Psychiatry & Allied Disciplines, 44(6), 793-809.
Mundy, P., & Neal, A. (2001). Neural plasticity, joint attention, and a transactional
social-orienting model of autism. In L. M. Glidden (Ed.), International review of
research in mental retardation: Autism (Vol. 23, pp. 139-168). San Diego, CA:
Academic Press.
312
Mundy, P., & Sigman, M. (1989). Specifying the nature of the social impairment in
autism. In G. Dawson (Ed.), Autism: Nature, diagnosis, and treatment (pp. 3-
21). New York: The Guilford Press.
Mundy, P., Sigman, M., & Kasari, C. (1993). The theory of mind and joint-attention
deficits in autism. In S. Baron-Cohen, H. Tager-Flusberg, & D. J. Cohen (Eds.),
Understanding other minds: Perspectives from autism (pp. 181-203). Oxford:
Oxford University Press.
Murphy, M., Bolton, P. F., Pickles, A., Fombonne, E., Piven, J., & Rutter, M. (2000).
Personality traits of the relatives of autistic probands. Psychological Medicine,
30(6), 1411-1424.
Narayan, S., Moyes, B., & Wolff, S. (1990). Family characteristics of autistic children:
A further report. Journal of Autism & Developmental Disorders, 20(4), 523-535.
Norman, D., & Shallice, T. (1980). Attention to action: Willed and automatic control of
behaviour. Center for Human Information Processing (Technical Report No.
99).
Norman, D. A., & Shallice, T. (1986). Attention to action: Willed and automatic control
of behaviour. In R. J. Davidson, G. E. Schwartz, & D. Shapiro (Eds.),
Consciousness and self-regulation (Vol. 4). New York: Plenum Press.
Ohnishi, T., Matsuda, H., Hashimoto, T., Kunihiro, T., Nishikawa, M., Uema, T., &
Sasaki, M. (2000). Abnormal regional cerebral blood flow in childhood autism.
Brain, 123(9), 1838-1844.
Olson, D. R. (1989). Making up your mind. Canadian Psychology, 30, 617-627.
Oosterlaan, J., Logan, G. D., & Sergeant, J. A. (1998). Response inhibition in AD/HD,
CD, comorbid AD/HD + CD, anxious, and control children: A meta-analysis of
studies with the stop task. Journal of Child Psychology & Psychiatry & Allied
Disciplines, 39(3), 411-425.
Ornitz, E. M. (1969). Disorders of perception common to early infantile autism and
schizophrenia. Comprehensive Psychiatry, 10(4), 259-274.
Ornitz, E. M. (1988). Autism: A disorder of directed attention. Brain Dysfunction, 1(5-
6), 309-322.
Oswald, D. P., & Ollendick, T. H. (1989). Role taking and social competence in autism
and mental retardation. Journal of Autism & Developmental Disorders, 19(1),
119-127.
313
Owen, A. M., Downes, J. J., Sahakian, B. J., Polkey, C. E., & Robbins, T. W. (1990).
Planning and spatial working memory following frontal lobe lesions in man.
Neuropsychologia, 28(10), 1021-1034.
Owen, A. M., Roberts, A. C., Hodges, J. R., Summers, B. A., Polkey, C. E., & Robbins,
T. W. (1993). Contrasting mechanisms of impaired attentional set-shifting in
patients with frontal lobe damage or Parkinson's disease. Brain, 116, 1159-1175.
Owen, A. M., Roberts, A. C., Polkey, C. E., Sahakian, B. J., & Robbins, T. W. (1991).
Extra-dimensional versus intra-dimensional set shifting performance following
frontal lobe excisions, temporal lobe excisions or amygdalo-hippocampectomy
in man. Neuropsychologia, 29(10), 993-1006.
Ozonoff, S. (1995a). Executive functions in autism. In E. Schopler & G. B. Mesibov
(Eds.), Learning and cognition in autism (pp. 199-219). New York: Plenum
Press.
Ozonoff, S. (1995b). Reliability and validity of the Wisconsin Card Sorting Test in
studies of autism. Neuropsychology, 9(4), 491-500.
Ozonoff, S. (1997a). Causal mechanisms of autism: Unifying perspectives from an
information-processing framework. In D. J. Cohen & F. R. Volkmar (Eds.),
Handbook of autism and pervasive developmental disorders (2nd ed., pp. 868-
879). New York: John Wiley & Sons.
Ozonoff, S. (1997b). Components of executive function in autism and other disorders.
In J. Russell (Ed.), Autism as an executive disorder (pp. 179-211). Oxford:
Oxford University Press.
Ozonoff, S. (2001). Advances in the cognitive neuroscience of autism. In C. A. Nelson
& M. Luciana (Eds.), Handbook of developmental cognitive neuroscience (pp.
537-548). Cambridge, MA: MIT Press.
Ozonoff, S., & Jensen, J. (1999). Brief report: Specific executive function profiles in
three neurodevelopmental disorders. Journal of Autism and Developmental
Disorders, 29(2), 171-177.
Ozonoff, S., & McEvoy, R. E. (1994). A longitudinal study of executive function and
theory of mind development in autism. Development & Psychopathology, 6(3),
415-431.
Ozonoff, S., Pennington, B. F., & Rogers, S. J. (1991). Executive function deficits in
high-functioning autistic individuals: Relationship to theory of mind. Journal of
Child Psychology & Psychiatry & Allied Disciplines, 32(7), 1081-1105.
314
Ozonoff, S., Rogers, S. J., Farnham, J. M., & Pennington, B. F. (1993). Can standard
measures identify subclinical markers of autism? Journal of Autism &
Developmental Disorders, 23(3), 429-441.
Ozonoff, S., South, M., & Miller, J. N. (2000). DSM-IV-defined Asperger syndrome:
Cognitive, behavioral and early history differentiation from high-functioning
autism. Autism, 4(1), 29-46.
Ozonoff, S., & Strayer, D. L. (1997). Inhibitory function in nonretarded children with
autism. Journal of Autism and Developmental Disorders, 27(1), 59-77.
Ozonoff, S., & Strayer, D. L. (2001). Further evidence of intact working memory in
autism. Journal of Autism and Developmental Disorders, 31(3), 257-263.
Ozonoff, S., Strayer, D. L., McMahon, W. M., & Filloux, F. (1994). Executive function
abilities in autism and Tourette syndrome: An information processing approach.
Journal of Child Psychology & Psychiatry & Allied Disciplines, 35(6), 1015-
1032.
Pantelis, C., Barnes, T. R., Nelson, H. E., Tanner, S., Weatherley, L., Owen, A. M., &
Robbins, T. W. (1997). Frontal-striatal cognitive deficits in patients with chronic
schizophrenia. Brain, 120(10), 1823-1843.
Passler, M. A., Isaac, W., & Hynd, G. W. (1985). Neuropsychological development of
behavior attributed to frontal lobe functioning in children. Developmental
Neuropsychology, 1(4), 349-370.
Paterson, S., Brown, J., Gsoedl, M., Johnson, M., & Karmiloff-Smith, A. (1999).
Cognitive modularity and genetic disorders. Science, 286(5448), 2355-2358.
Pellicano, E., Maybery, M., Durkin, K., & Maley, A. (2004). Weak central coherence in
children with autism: Its relationship to mindreading and executive functioning.
Manuscript submitted for publication.
Pennington, B. F. (1997). Dimensions of executive functions in normal and abnormal
development. In N. A. Krasnegor, G. R. Lyon, & P. S. Goldman-Rakic (Eds.),
Development of the prefrontal cortex: Evolution, neurobiology, and behavior
(pp. 265-281). Baltimore, MD: Paul H. Brookes.
Pennington, B. F., Groisser, D., & Welsh, M. C. (1993). Contrasting cognitive deficits
in attention deficit hyperactivity disorder versus reading disability.
Developmental Psychology, 29(3), 511-523.
Pennington, B. F., & Ozonoff, S. (1991). A neuroscientific perspective on continuity
and discontinuity in developmental psychopathology. In D. Cicchetti & S. L.
Toth (Eds.), Rochester symposium on developmental psychopathology (Vol. 3:
315
Models and integrations, pp. 117-159). Rochester, NY: University of Rochester
Press.
Pennington, B. F., & Ozonoff, S. (1996). Executive functions and developmental
psychopathology. Journal of Child Psychology and Psychiatry, 37(1), 51-87.
Pennington, B. F., Rogers, S. J., Bennetto, L., McMahon Griffith, E., Reed, D. T., &
Shyu, V. (1997). Validity tests of the executive dysfunction hypothesis of
autism. In J. Russell (Ed.), Autism as an executive disorder (pp. 143-178).
Oxford: Oxford University Press.
Pennington, B. F., & Welsh, M. (1995). Neuropsychology and developmental
psychopathology. In D. Cicchetti & D. J. Cohen (Eds.), Developmental
psychopathology (Vol. 1: Theory and methods, pp. 254-290). Oxford: John
Wiley & Sons.
Perner, J. (1991). Understanding the representational mind. Cambridge, MA: MIT
Press.
Perner, J. (1993). The theory of mind deficit in autism: Rethinking the
metarepresentation theory. In S. Baron-Cohen, H. Tager-Flusberg, & D. J.
Cohen (Eds.), Understanding other minds: Perspectives from autism (pp. 112-
137). Oxford: Oxford University Press.
Perner, J. (1995). The many faces of belief: Reflections on Fodor's and the child's theory
of mind. Cognition, 57(3), 241-269.
Perner, J. (1998). The meta-intentional nature of executive functions and theory of
mind. In P. Carruthers & J. Boucher (Eds.), Language and thought (pp. 270-
283). Cambridge: Cambridge University Press.
Perner, J. (2000). About + belief + counterfactual. In P. Mitchell & K. J. Riggs (Eds.),
Children's reasoning and the mind (pp. 367-401). Hove, UK: Psychology Press.
Perner, J., Baker, S., & Hutton, D. (1994). Prelief: The conceptual origins of belief and
pretence. In C. Lewis & P. Mitchell (Eds.), Children's early understanding of
mind: Origins and development. Hove, UK: Lawrence Erlbaum.
Perner, J., Frith, U., Leslie, A. M., & Leekam, S. R. (1989). Exploration of the autistic
child's theory of mind: Knowledge, belief, and communication. Child
Development, 60(3), 689-700.
Perner, J., Kain, W., & Barchfeld, P. (2002a). Executive control and higher-order theory
of mind in children at risk of ADHD. Infant & Child Development, 11(2), 141-
158.
316
Perner, J., & Lang, B. (1999). Development of theory of mind and executive control.
Trends in Cognitive Sciences, 3(9), 337-344.
Perner, J., & Lang, B. (2000). Theory of mind and executive function: Is there a
developmental relationship? In S. Baron-Cohen, H. Tager-Flusberg, & D. J.
Cohen (Eds.), Understanding other minds: Perspectives from developmental
cognitive neuroscience (2nd ed., pp. 150-181). London: Oxford University
Press.
Perner, J., & Lang, B. (2002). What causes 3-year-olds' difficulty on the dimensional
change card sorting task? Infant & Child Development, 11(2), 93-105.
Perner, J., Lang, B., & Kloo, D. (2002b). Theory of mind and self-control: More than a
common problem of inhibition. Child Development, 73(3), 752-767.
Perner, J., Leekam, S. R., & Wimmer, H. (1987). Three-year-olds' difficulty with false
belief: The case for a conceptual deficit. British Journal of Developmental
Psychology, 5(2), 125-137.
Perner, J., Ruffman, T., & Leekam, S. R. (1994). Theory of mind is contagious: You
catch it from your sibs. Child Development, 65(4), 1228-1238.
Perner, J., Stummer, S., & Lang, B. (1999). Executive functions and theory of mind:
Cognitive complexity or functional dependence? In P. D. Zelazo, J. W.
Astington, & D. R. Olson (Eds.), Developing theories of intention: Social
understanding and self-control (pp. 133-152). Mahwah, NJ: Lawrence Erlbaum.
Perner, J., & Wimmer, H. (1985). "John thinks that Mary thinks that . . .": Attribution of
second-order beliefs by 5- to 10-year-old children. Journal of Experimental
Child Psychology, 39(3), 437-471.
Peterson, C. C. (2002). Drawing insight from pictures: The development of concepts of
false drawing and false belief in children with deafness, normal hearing, and
autism. Child Development, 73(5), 1442-1459.
Peterson, C. C., & Siegal, M. (1995). Deafness, conversation and theory of mind.
Journal of Child Psychology & Psychiatry & Allied Disciplines, 36(3), 459-474.
Peterson, D. M., & Bowler, D. M. (2000). Counterfactual reasoning and false belief
understanding in children with autism, children with severe learning difficulties
and children with typical development. Autism, 4(4), 391-405.
Peterson, D. M., & Riggs, K. J. (1999). Adaptive modelling and mindreading. Mind and
Language, 14(1), 80-117.
317
Phillips, L. H. (1997). Do "frontal tests" measure executive function? Issues of
assessment and evidence from fluency tests. In P. Rabbitt (Ed.), Methodology of
frontal and executive function (pp. 191-213). Hove, UK: Psychology Press.
Phillips, W., Baron-Cohen, S., & Rutter, M. (1998). Understanding intention in normal
development and in autism. British Journal of Developmental Psychology,
16(3), 337-348.
Pickles, A., Bolton, P., Macdonald, H., Bailey, A., Le Couteur, A., Sim, C. H., &
Rutter, M. (1995). Latent-class analysis of recurrence risks for complex
phenotypes with selection and measurement error: A twin and family history
study of autism. American Journal of Human Genetics, 57(3), 717-26.
Pickles, A., Starr, E., Kazak, S., Bolton, P., Papanikolaou, K., Bailey, A., Goodman, R.,
& Rutter, M. (2000). Variable expression of the autism broader phenotype:
Findings from extended pedigrees. Journal of Child Psychology & Psychiatry &
Allied Disciplines, 41(4), 491-502.
Pilowsky, T., Yirmiya, N., Arbelle, S., & Mozes, T. (2000). Theory of mind abilities of
children with schizophrenia, children with autism, and normally developing
children. Schizophrenia Research, 42(2), 145-155.
Pilowsky, T., Yirmiya, N., Shalev, R. S., & Gross-Tsur, V. (2003). Language abilities
of siblings of children with autism. Journal of Child Psychology & Psychiatry &
Allied Disciplines, 44(6), 914-925.
Piven, J. (1999). Genetic liability for autism: The behavioural expression in relatives.
International Review of Psychiatry, 11(4), 299-308.
Piven, J., Arndt, S., Bailey, J., Havercamp, S., Andreasen, N. C., & Palmer, P. (1995).
An MRI study of brain size in autism. American Journal of Psychiatry, 152(8),
1145-1149.
Piven, J., Bailey, J., Ranson, B. J., & Arndt, S. (1997a). An MRI study of the corpus
callosum in autism. American Journal of Psychiatry, 154, 1051-1056.
Piven, J., Berthier, M. L., Starkstein, S. E., Nehme, E., Pearlson, G., & Folstein, S.
(1990a). Magnetic resonance imaging evidence for a defect of cerebral cortical
development in autism. American Journal of Psychiatry, 147(6), 734-739.
Piven, J., Chase, G. A., Landa, R., Wzorek, M., Gayle, J., Cloud, D., & Folstein, S.
(1991). Psychiatric disorders in the parents of autistic individuals. Journal of the
American Academy of Child & Adolescent Psychiatry, 30(3), 471-478.
Piven, J., Gayle, J., Chase, G. A., Fink, B., Landa, R., Wzorek, M. M., & Folstein, S. E.
(1990b). A family history study of neuropsychiatric disorders in the adult
318
siblings of autistic individuals. Journal of the American Academy of Child &
Adolescent Psychiatry, 29(2), 177-183.
Piven, J., & Palmer, P. (1997). Cognitive deficits in parents from multiple-incidence
autism families. Journal of Child Psychology & Psychiatry & Allied Disciplines,
38(8), 1011-1021.
Piven, J., & Palmer, P. (1999). Psychiatric disorder and the broad autism phenotype:
Evidence from a family study of multiple-incidence autism families. American
Journal of Psychiatry, 156(4), 557-563.
Piven, J., Palmer, P., Jacobi, D., Childress, D., & Arndt, S. (1997b). Broader autism
phenotype: Evidence from a family history study of multiple-incidence autism
families. American Journal of Psychiatry, 154(2), 185-190.
Piven, J., Palmer, P., Landa, R., Santangelo, S., Jacobi, D., & Childress, D. (1997c).
Personality and language characteristics in parents from multiple-incidence
autism families. American Journal of Medical Genetics (Neuropsychiatric
Genetics), 74(4), 398-411.
Piven, J., Wzorek, M., Landa, R., Lainhart, J., Bolton, P., Chase, G. A., & Folstein, S.
(1994). Personality characteristics of the parents of autistic individuals.
Psychological Medicine, 24(3), 783-795.
Plaisted, K. C. (2000). Aspects of autism that theory of mind cannot explain. In S.
Baron-Cohen, H. Tager-Flusberg, & D. J. Cohen (Eds.), Understanding other
minds: Perspectives from developmental cognitive neuroscience (2nd ed., pp.
222-250). London: Oxford University Press.
Plaisted, K. C. (2001). Reduced generalization in autism: An alternative to weak central
coherence. In J. A. Burack, T. Charman, N. Yirmiya, & P. R. Zelazo (Eds.), The
development of autism: Perspectives from theory and research (pp. 149-169).
Mahwah, NJ: Lawrence Erlbaum Associates.
Premack, D., & Woodruff, G. (1978). Does the chimpanzee have a theory of mind? The
Behavioral and Brain Sciences, 4, 515-526.
Price, B. H., Daffner, K. R., Stowe, R. M., & Mesulam, M. M. (1990). The
comportmental learning disabilities of early frontal lobe damage. Brain, 113,
1383-1393.
Prior, M., Dahlstrom, B., & Squires, T.-L. (1990). Autistic children's knowledge of
thinking and feeling states in other people. Journal of Child Psychology &
Psychiatry & Allied Disciplines, 31(4), 587-601.
319
Prior, M., Eisenmajer, R., Leekam, S., Wing, L., Gould, J., Ong, B., & Dowe, D.
(1998). Are there subgroups within the autistic spectrum? A cluster analysis of a
group of children with autistic spectrum disorders. Journal of Child Psychology
& Psychiatry & Allied Disciplines, 39(6), 893-902.
Prior, M., & Hoffmann, W. (1990). Brief report: Neuropsychological testing of autistic
children through an exploration with frontal lobe tests. Journal of Autism &
Developmental Disorders, 20(4), 581-590.
Pylyshyn, Z. W. (1978). When is attribution of beliefs justified? The Behavioral and
Brain Sciences, 1, 592-593.
Rabbitt, P. (Ed.). (1997). Methodology of frontal and executive function. Hove, UK:
Psychology Press.
Rapin, I. (1997). Classification and causal issues in autism. In D. J. Cohen & F. R.
Volkmar (Eds.), Handbook of autism and pervasive developmental disorders
(2nd ed., pp. 847-867). New York: John Wiley & Sons.
Razani, J., Boone, K., Miller, B. L., Lee, A., & Sherman, D. (2001).
Neuropsychological performance of right- and left-frontotemporal dementia
compared to Alzheimer's disease. Journal of the International
Neuropsychological Society, 7(4), 468-480.
Reed, T., & Peterson, C. (1990). A comparative study of autistic subjects' performance
at two levels of visual and cognitive perspective taking. Journal of Autism and
Developmental Disorders, 20, 555-568.
Reitan, R. M., & Wolfson, D. (1994). A selective and critical review of
neuropsychological deficits and the frontal lobes. Neuropsychology Review,
4(3), 161-198.
Remmel, E. R. (2003). Theory of mind development in signing deaf children.
Unpublished PhD thesis, Stanford University.
Rinehart, N. J., Bradshaw, J. L., Moss, S. A., Brereton, A. V., & Tonge, B. J. (2001). A
deficit in shifting attention present in high-functioning autism but not Asperger's
disorder. Autism, 5(1), 67-80.
Rinehart, N. J., Bradshaw, J. L., Tonge, B. J., Brereton, A. V., & Bellgrove, M. A.
(2002). A neurobehavioral examination of individuals with high-functioning
autism and Asperger disorder using a fronto-striatal model of dysfunction.
Behavioral & Cognitive Neuroscience Reviews, 1(2), 164-177.
320
Risch, N., Spiker, D., Lotspeich, L., Nouri, N., Hinds, D., Hallmayer, J., et al. (1999). A
genomic screen of autism: evidence for a multilocus etiology. American Journal
of Human Genetics, 65(2), 493-507.
Roberts, R. J., Hager, L. D., & Heron, C. (1994). Prefrontal cognitive processes:
Working memory and inhibition in the antisaccade task. Journal of
Experimental Psychology: General, 123(4), 374-393.
Roberts, R. J., & Pennington, B. F. (1996). An interactive framework for examining
prefrontal cognitive processes. Developmental Neuropsychology, 12(1), 105-
126.
Robinson, E. J., & Beck, S. (2000). What is difficult about counterfactual reasoning? In
P. Mitchell & K. J. Riggs (Eds.), Children's reasoning and the mind (pp. 101-
119). Hove, UK: Psychology Press.
Robinson, E., & Mitchell, P. (1995). Masking of children's early understanding of the
representational mind: Backwards explanation versus prediction. Child
Development, 66(4), 1022-1039.
Robinson, E., Riggs, K., & Samuels, J. (1996). Children's memory for drawings based
on a false belief. Developmental Psychology, 32(6), 1056-1064.
Rogers, S. J. (1999). An examination of the imitation deficit in autism. In J. Nadel & G.
Butterworth (Eds.), Imitation in infancy: Cambridge studies in cognitive
perceptual development (pp. 254-283). New York: Cambridge University Press.
Rogers, S. J., & Pennington, B. F. (1991). A theoretical approach to the deficits in
infantile autism. Development & Psychopathology, 3(2), 137-162.
Rosenthal, R. (1991). Meta-analytic procedures for social research. Newbury Park,
CA: Sage.
Roth, D., & Leslie, A. M. (1998). Solving belief problems: Toward a task analysis.
Cognition, 66(1), 1-31.
Rowe, A. D., Bullock, P. R., Polkey, C. E., & Morris, R. G. (2001). 'Theory of mind'
impairments and their relationship to executive functioning following frontal
lobe excisions. Brain, 124(3), 600-616.
Royall, D. R., Lauterbach, E. C., Cummings, J. L., Reeve, A., Rummans, T. A., Kaufer,
D. I., LaFrance, W., & Coffey, C. (2002). Executive control function: A review
of its promise and challenges for clinical research: A report from the committee
on research of the American Neuropsychiatric Association. Journal of
Neuropsychiatry & Clinical Neurosciences, 14(4), 377-405.
321
Ruffman, T., Perner, J., Naito, M., Parkin, L., & Clements, W. A. (1998). Older (but not
younger) siblings facilitate false belief understanding. Developmental
Psychology, 34(1), 161-174.
Rumsey, J. M. (1985). Conceptual problem-solving in highly verbal, nonretarded
autistic men. Journal of Autism & Developmental Disorders, 15(1), 23-36.
Rumsey, J. M., & Hamburger, S. D. (1988). Neuropsychological findings in high-
functioning men with infantile autism, residual state. Journal of Clinical &
Experimental Neuropsychology, 10(2), 201-221.
Russell, J. (1996). Agency: Its role in mental development. Hove, UK: Lawrence
Erlbaum.
Russell, J. (Ed.). (1997a). Autism as an executive disorder. Oxford: Oxford University
Press.
Russell, J. (1997b). How executive disorders can bring about an inadequate 'theory of
mind'. In J. Russell (Ed.), Autism as an executive disorder (pp. 215-255).
Oxford: Oxford University Press.
Russell, J., Hala, S., & Hill, E. (2003). The automated windows task: The performance
of preschool children, children with autism, and children with moderate learning
difficulties. Cognitive Development, 18(1), 111-137.
Russell, J., & Hill, E. L. (2001). Action-monitoring and intention reporting in children
with autism. Journal of Child Psychology & Psychiatry & Allied Disciplines,
42(3), 317-328.
Russell, J., Hill, E. L., & Franco, F. (2001). The role of belief veracity in understanding
intentions-in-action: Preschool children's performance on the transparent
intentions task. Cognitive Development, 16(3), 775-792.
Russell, J., & Jarrold, C. (1998). Error-correction problems in autism: Evidence for a
monitoring impairment? Journal of Autism & Developmental Disorders, 28(3),
177-188.
Russell, J., & Jarrold, C. (1999). Memory for actions in children with autism: Self
versus other. Cognitive Neuropsychiatry, 4(4), 303-331.
Russell, J., Jarrold, C., & Henry, L. (1996). Working memory in children with autism
and with moderate learning difficulties. Journal of Child Psychology and
Psychiatry, 37(6), 673-686.
Russell, J., Jarrold, C., & Hood, B. (1999). Two intact executive capacities in children
with autism: Implications for the core executive dysfunctions in the disorder.
Journal of Autism and Developmental Disorders, 29(2), 103-112.
322
Russell, J., Jarrold, C., & Potel, D. (1994). What makes strategic deception difficult for
children - the deception or the strategy? British Journal of Developmental
Psychology, 12(3), 301-314.
Russell, J., Mauthner, N., Sharpe, S., & Tidswell, T. (1991). The "windows task" as a
measure of strategic deception in preschoolers and autistic subjects. British
Journal of Developmental Psychology, 9(2), 331-349.
Russell, J., Saltmarsh, R., & Hill, E. (1999). What do executive factors contribute to the
failure on false belief tasks by children with autism? Journal of Child
Psychology & Psychiatry & Allied Disciplines, 40(6), 859-868.
Rutherford, M., & Rogers, S. J. (2003). Cognitive underpinnings of pretend play in
autism. Journal of Autism & Developmental Disorders, 33(3), 289-302.
Rutter, M. (1968). Concepts of autism: A review of research. Journal of Child
Psychology & Psychiatry & Allied Disciplines, 9(1), 1-25.
Rutter, M. (1970). Autistic children: Infancy to adulthood. Seminars in Psychiatry, 2,
435-450.
Rutter, M. (1983). Cognitive deficits in the pathogenesis of autism. Journal of Child
Psychology & Psychiatry & Allied Disciplines, 24(4), 513-531.
Rutter, M. (2000). Genetic studies of autism: From the 1970s into the millennium.
Journal of Abnormal Child Psychology, 28(1), 3-14.
Saltzman, J., Strauss, E., Hunter, M., & Archibald, S. (2000). Theory of mind and
executive functions in normal human aging and Parkinson's disease. Journal of
the International Neuropsychological Society, 6(7), 781-788.
Saver, J. L., & Damasio, A. R. (1991). Preserved access and processing of social
knowledge in a patient with acquired sociopathy due to ventromedial frontal
damage. Neuropsychologia, 29(12), 1241-1249.
Scheerer, M., Rothmann, E., & Goldstein, K. (1945). A case of "idiot savant": An
experimental study of personality organization. Psychological Monographs, 58,
1-63.
Schneider, W., & Shiffrin, R. M. (1977). Controlled and automatic human information
processing: I. Detection, search, and attention. Psychological Review, 84(1), 1-
66.
Scholl, B. J., & Leslie, A. M. (1999). Modularity, development and 'theory of mind'.
Mind & Language, 14(1), 131-153.
323
Scholl, B. J., & Leslie, A. M. (2001). Minds, modules, and meta-analysis. Commentary
on "Meta-analysis of theory-of-mind development: The truth about false belief.".
Child Development, 72(3), 696-701.
Schwartz, M. L. (1997). Organization and development of callosal connectivity in
prefrontal cortex. In N. A. Krasnegor, G. R. Lyon, & P. S. Goldman-Rakic
(Eds.), Development of the prefrontal cortex: Evolution, neurobiology and
behavior (pp. 49-67). Baltimore, MD: Paul H. Brookes.
Sergeant, J. A., Geurts, H., & Oosterlaan, J. (2002). How specific is a deficit of
executive functioning for attention-deficit/hyperactivity disorder? Behavioural
Brain Research, 130(1-2), 3-28.
Shah, A., & Frith, U. (1983). An islet of ability in autistic children: A research note.
Journal of Child Psychology and Psychiatry, 24(4), 613-620.
Shah, A., & Frith, U. (1993). Why do autistic individuals show superior performance on
the block design task? Journal of Child Psychology & Psychiatry & Allied
Disciplines, 34(8), 1351-1364.
Shallice, T. (1982). Specific impairments in planning. Philosophical Transactions of the
Royal Society of London B, 298, 199-209.
Shallice, T. (1984). More functionally isolable subsystems but fewer "modules"?
Cognition, 17(3), 243-252.
Shallice, T. (1988). From neuropsychology to mental structure. New York: Cambridge
University Press.
Shallice, T. (2002). Fractionation of the supervisory system. In D. T. Stuss & R. T.
Knight (Eds.), Principles of frontal lobe function (pp. 261-277). London: Oxford
University Press.
Shallice, T., & Burgess, P. (1991). Deficits in strategy application after frontal lobe
damage in man. Brain, 114, 727-741.
Shallice, T., & Burgess, P. W. (1996). Domains of supervisory control and the temporal
organisation of behaviour. Philosophical Transactions of the Royal Society of
London B, 351, 1405-1412.
Shallice, T., Marzocchi, G. M., Coser, S., Del Savio, M., Meuter, R. F., & Rumiati, R. I.
(2002). Executive function profile of children with attention deficit hyperactivity
disorder. Developmental Neuropsychology, 21(1), 43-71.
Sherman, M., Nass, R., & Shapiro, T. (1984). Brief report: Regional cerebral blood flow
in autism. Journal of Autism and Developmental Disorders, 14(4), 439-446.
324
Siegal, M., & Beattie, K. (1991). Where to look first for children's understanding of
false beliefs. Cognition, 38(1), 1-12.
Sigman, M., & Ruskin, E. (1999). Social competence in children with autism, Down
syndrome and developmental delays: A longitudinal study. Monographs of the
Society for Research in Child Development, 64(Serial No. 256).
Silverman, J. M., Smith, C. J., Schmeidler, J., Hollander, E., Lawlor, B. A., Fitzgerald,
M., Buxbaum, J. D., Delaney, K., & Galvin, P. (2002). Symptom domains in
autism and related conditions: Evidence for familiality. American Journal of
Medical Genetics (Neuropsychiatric Genetics), 114(1), 64-73.
Skuse, D. (2001). Endophenotypes and child psychiatry. British Journal of Psychiatry,
178, 395-396.
Skuse, D., James, R., Bishop, D., Coppin, B., Dalton, P., Aamodt-Leeper, G., Bacarese-
Hamilton, M., Creswell, C., McGurk, R., & Jacobs, P. A. (1997). Evidence from
Turner's syndrome of an imprinted X-linked locus affecting cognitive function.
Nature, 387(6634), 705-708.
Slaats-Willemse, D., Swaab-Barneveld, H., de Sonneville, L., van der Meulen, E., &
Buitelaar, J. (2003). Deficient response inhibition as a cognitive endophenotype
of ADHD. Journal of the American Academy of Child & Adolescent Psychiatry,
42(10), 1242-1248.
Smalley, S. L., & Asarnow, R. F. (1990). Brief report: Cognitive subclinical markers in
autism. Journal of Autism & Developmental Disorders, 20(2), 271-278.
Smalley, S. L., McCracken, J., & Tanguay, P. (1995). Autism, affective disorders, and
social phobia. American Journal of Medical Genetics (Neuropsychiatric
Genetics), 60, 19-26.
Smith, I. M., & Bryson, S. E. (1994). Imitation and action in autism: A critical review.
Psychological Bulletin, 116(2), 259-273.
Smith, M. L., Klim, P., & Hanley, W. B. (2000). Executive function in school-aged
children with phenylketonuria. Journal of Developmental & Physical
Disabilities, 12(4), 317-332.
Sodian, B., & Frith, U. (1992). Deception and sabotage in autistic, retarded and normal
children. Journal of Child Psychology & Psychiatry & Allied Disciplines, 33(3),
591-605.
Sparrevohn, R., & Howie, P. M. (1995). Theory of mind in children with autistic
disorder: Evidence of developmental progression and the role of verbal ability.
Journal of Child Psychology & Psychiatry & Allied Disciplines, 36(2), 249-263.
325
Stahl, L., & Pry, R. (2002). Joint attention and set-shifting in young children with
autism. Autism, 6(4), 383-396.
Starr, E., Berument, S. K., Pickles, A., Tomlins, M., Bailey, A., Papanikolaou, K., &
Rutter, M. (2001). A family genetic study of autism associated with profound
mental retardation. Journal of Autism & Developmental Disorders, 31(1), 89-96.
Steel, J., Gorman, R., & Flexman, J. E. (1984). Neuropsychiatric testing in an autistic
mathematical idiot-savant: Evidence for nonverbal abstract capacity. Journal of
the American Academy of Child Psychiatry, 23(6), 704-707.
Steele, S., Joseph, R. M., & Tager-Flusberg, H. (2003). Brief report: Developmental
change in theory of mind abilities in children with autism. Journal of Autism and
Developmental Disorders, 33(4), 461-467.
Steffenburg, S., Gillberg, C., Hellgren, L., Andersson, L., Gillberg, I., Jakobsson, G., &
Bohman, M. (1989). A twin study of autism in Denmark, Finland, Iceland,
Norway and Sweden. Journal of Child Psychology & Psychiatry & Allied
Disciplines, 30(3), 405-416.
Stevens, M. C., Fein, D. A., Dunn, M., Allen, D., Waterhouse, L. H., Feinstein, C., &
Rapin, I. (2000). Subgroups of children with autism by cluster analysis: A
longitudinal examination. Journal of the American Academy of Child &
Adolescent Psychiatry, 39(3), 346-352.
Stone, V. (2000). The role of the frontal lobes and the amygdala in theory of mind. In S.
Baron-Cohen, H. Tager-Flusberg, & D. J. Cohen (Eds.), Understanding other
minds: Perspectives from developmental cognitive neuroscience (2nd ed., pp.
253-273). London: Oxford University Press.
Stone, V. E., Baron-Cohen, S., & Knight, R. T. (1998). Frontal lobe contributions to
theory of mind. Journal of Cognitive Neuroscience, 10(5), 640-656.
Stuss, D. T., & Alexander, M. P. (2000). Executive functions and the frontal lobes: A
conceptual view. Psychological Research, 63(3-4), 289-298.
Stuss, D. T., & Benson, D. F. (1984). Neuropsychological studies of the frontal lobes.
Psychological Bulletin, 95(1), 3-28.
Stuss, D. T., & Benson, D. F. (1986). The frontal lobes. New York: Raven Press.
Stuss, D. T., Gallup, G. G., Jr., & Alexander, M. P. (2001). The frontal lobes are
necessary for "theory of mind". Brain, 124(2), 279-286.
Stuss, D. T., & Knight, R. T. (Eds.). (2002). Principles of frontal lobe function. London:
Oxford University Press.
326
Surian, L., & Leslie, A. M. (1999). Competence and performance in false belief
understanding: A comparison of autistic and normal 3-yr-old children. British
Journal of Developmental Psychology, 17(Pt 1), 141-155.
Swettenham, J., Baron-Cohen, S., Charman, T., Cox, A., Baird, G., Drew, A., Rees, L.,
& Wheelwright, S. (1998). The frequency and distribution of spontaneous
attention shifts between social and nonsocial stimuli in autistic, typically
developing, and nonautistic developmentally delayed infants. Journal of Child
Psychology & Psychiatry & Allied Disciplines, 39(5), 747-753.
Szatmari, P. (1999). Heterogeneity and the genetics of autism. Journal of Psychiatry &
Neuroscience, 24(2), 159-165.
Szatmari, P., Jones, M. B., Tuff, L., Bartolucci, G., Bartolucci, G., Fisman, S., &
Mahoney, W. (1993). Lack of cognitive impairment in first-degree relatives of
children with pervasive developmental disorders. Journal of the American
Academy of Child & Adolescent Psychiatry, 32(6), 1264-1273.
Szatmari, P., Jones, M. B., Zwaigenbaum, L., & MacLean, J. E. (1998). Genetics of
autism: Overview and new directions. Journal of Autism & Developmental
Disorders, 28(5), 351-368.
Szatmari, P., Merette, C., Bryson, S. E., Thivierge, J., Roy, M.-A., Cayer, M., &
Maziade, M. (2002). Quantifying dimensions in autism: A factor-analytic study.
Journal of the American Academy of Child & Adolescent Psychiatry, 41(4), 467-
474.
Szatmari, P., Tuff, L., Finlayson, A. J., & Bartolucci, G. (1990). Asperger's Syndrome
and autism: Neurocognitive aspects. Journal of the American Academy of Child
& Adolescent Psychiatry, 29(1), 130-136.
Tabachnik, B. G., & Fidell, L. S. (1996). Using multivariate statistics. (3rd ed.). New
York: HarperCollins College Publishers.
Tager-Flusberg, H. (1992). Autistic children's talk about psychological states: Deficits
in the early acquisition of a theory of mind. Child Development, 63, 161-172.
Tager-Flusberg, H. (Ed.). (1999a). Neurodevelopmental disorders: Developmental
cognitive neuroscience.. Cambridge, MA: The MIT Press.
Tager-Flusberg, H. (1999b). A psychological approach to understanding the social and
language impairments in autism. International Review of Psychiatry, 11(4), 325-
334.
Tager-Flusberg, H. (2000). Language and understanding minds: Connections in autism.
In S. Baron-Cohen, H. Tager-Flusberg, & D. J. Cohen (Eds.), Understanding
327
other minds: Perspectives from developmental cognitive neuroscience (2nd ed.,
pp. 124-149). London: Oxford University Press.
Tager-Flusberg, H. (2001). A reexamination of the theory of mind hypothesis of autism.
In J. A. Burack, T. Charman, N. Yirmiya, & P. R. Zelazo (Eds.), The
development of autism: Perspectives from theory and research (pp. 173-193).
Mahwah, NJ: Lawrence Erlbaum Associates.
Tager-Flusberg, H., & Joseph, R. M. (2003). Identifying neurocognitive phenotypes in
autism. Philosophical Transactions of the Royal Society of London B,
358(1430), 303-314.
Tager-Flusberg, H., & Sullivan, K. (1994a). Predicting and explaining behavior: A
comparison of autistic, mentally retarded and normal children. Journal of Child
Psychology & Psychiatry & Allied Disciplines, 35(6), 1059-1075.
Tager-Flusberg, H., & Sullivan, K. (1994b). A second look at second-order belief
attribution in autism. Journal of Autism & Developmental Disorders, 24(5), 577-
586.
Tager-Flusberg, H., & Sullivan, K. (1995). Attributing mental states to story characters:
A comparison of narratives produced by autistic and mentally retarded
individuals. Applied Psycholinguistics, 16(3), 241-256.
Tager-Flusberg, H., Sullivan, K., & Boshart, J. (1997). Executive functions and
performance on false belief tasks. Developmental Neuropsychology, 13(4), 487-
493.
Teunisse, J.-P., Cools, A. R., van Spaendonck, K. P. M., Aerts, F. H. T. M., & Berger,
H. J. C. (2001). Cognitive styles in high-functioning adolescents with autistic
disorder. Journal of Autism & Developmental Disorders, 31(1), 55-66.
Thatcher, R. W. (1997). Human frontal lobe development: A theory of cyclical cortical
reorganization. In N. A. Krasnegor, G. R. Lyon, & P. S. Goldman-Rakic (Eds.),
Development of the prefrontal cortex: Evolution, neurobiology and behavior
(pp. 85-113). Baltimore, MD: Paul H. Brookes.
Thomas, M., & Karmiloff-Smith, A. (2002). Are developmental disorders like cases of
adult brain damage? Implications from connectionist modelling. Behavioral &
Brain Sciences, 25(6), 727-787.
Tranel, D. (2002). Emotion, decision making, and the ventromedial prefrontal cortex. In
D. T. Stuss & R. T. Knight (Eds.), Principles of frontal lobe function (pp. 338-
352). London: Oxford University Press.
328
Tranel, D., Anderson, S. W., & Benton, A. (1994). Development of the concept of
'executive function' and its relationship to the frontal lobes. In F. Boller & J.
Grafman (Eds.), Handbook of Neuropsychology (Vol. 9, pp. 125-148).
Amsterdam: Elsevier Science.
Turner, M. A. (1996). Repetitive behaviour and cognitive functioning in autism.
Unpublished PhD thesis, University of Cambridge.
Turner, M. (1997). Towards an executive dysfunction account of repetitive behaviour in
autism. In J. Russell (Ed.), Autism as an executive disorder (pp. 57-100).
Oxford: Oxford University Press.
Turner, M. A. (1999). Generating novel ideas: Fluency performance in high-functioning
and learning disabled individuals with autism. Journal of Child Psychology and
Psychiatry, 40(2), 189-201.
Tyrer, P. (Ed.). (1988). Personality assessment schedule: In personality disorders:
Diagnosis, management, and course. London: Butterworth.
Veale, D. M., Sahakian, B. J., Owen, A. M., & Marks, I. M. (1996). Specific cognitive
deficits in tests sensitive to frontal lobe dysfunction in obsessive-compulsive
disorder. Psychological Medicine, 26, 1261-1269.
Vecchi, T. (1998). Visuo-spatial imagery in congenitally totally blind people. Memory,
6(1), 91-102.
Volkmar, F. R., Lord, C., Bailey, A., Schultz, R. T., & Klin, A. (2004). Autism and
pervasive developmental disorders. Journal of Child Psychology and Psychiatry,
45(1), 135-170.
Volkmar, F. R., Sparrow, S. S., Goudreau, D., Cicchetti, D. V., Paul, R., & Cohen, D. J.
(1987). Social deficits in autism: An operational approach using the Vineland
Adaptive Behavior Scales. Journal of the American Academy of Child &
Adolescent Psychiatry, 26(2), 156-161.
Wallach, M. A., & Kogan, N. (1965). Modes of thinking in young children. New York:
Holt, Rinehart, & Winston.
Walsh, K. W. (1978). Neuropsychology: A clinical approach. New York: Churchill
Livingston.
Waltz, J. A., Knowlton, B. J., Holyoak, K. J., Boone, K. B., Mishkin, F. S., de Menezes
Santos, M., Thomas, C. R., & Miller, B. L. (1999). A system for relational
reasoning in human prefrontal cortex. Psychological Science, 10(2), 119-125.
Waterhouse, L., Fein, D., & Modahl, C. (1996). Neurofunctional mechanisms in autism.
Psychological Review, 103(3), 457-489.
329
Weinberger, D. (2002). Schizophrenia, the prefrontal cortex, and a mechanism of
genetic susceptibility. European Psychiatry, 17(Suppl4), 355-362.
Wellman, H. M. (1990). The child's theory of mind. Cambridge, MA: MIT Press.
Wellman, H. M., Cross, D., & Watson, J. (2001). Meta-analysis of theory-of-mind
development: The truth about false belief. Child Development, 72(3), 655-684.
Wellman, H. M., & Gelman, S. A. (1998). Knowledge acquisition in foundational
domains. In D. Kuhn & R. Siegler (Eds.), Handbook of child psychology:
Cognition, perception and language (5th ed., pp. 523-573). New York: Wiley.
Wellman, H. M., & Lagatutta, K. H. (2000). Developing understandings of mind. In S.
Baron-Cohen, H. Tager-Flusberg, & D. J. Cohen (Eds.), Understanding other
minds: Perspectives from developmental cognitive neuroscience (2nd ed., pp.
21-49). London: Oxford University Press.
Wellman, H. M., & Woolley, J. D. (1990). From simple desires to ordinary beliefs: The
early development of everyday psychology. Cognition, 35(3), 245-275.
Welsh, M. C., & Pennington, B. F. (1988). Assessing frontal lobe functioning in
children: Views from developmental psychology. Developmental
Neuropsychology, 4(3), 199-230.
Welsh, M. C., Pennington, B. F., & Groisser, D. B. (1991). A normative-developmental
study of executive function: A window on prefrontal function in children.
Developmental Neuropsychology, 7(2), 131-149.
Welsh, M. C., Pennington, B. F., Ozonoff, S., Rouse, B., & McCabe, E. (1990).
Neuropsychology of early-treated phenylketonuria: Specific executive function
deficits. Child Development, 61(6), 1697-1713.
Welsh, M. C., Satterlee-Cartmell, T., & Stine, M. (1999). Towers of Hanoi and London:
Contribution of working memory and inhibition to performance. Brain &
Cognition, 41(2), 231-242.
Whiten, A. (Ed.). (1991). Natural theories of mind: Evolution, development and
simulation of everyday mindreading. Oxford: Blackwell.
Williams, M. A., Moss, S. A., Bradshaw, J. L., & Rinehart, N. J. (2002). Random
number generation in autism. Journal of Autism & Developmental Disorders,
32(1), 43-47.
Wilson, B. A., Evans, J. J., Emslie, H., Alderman, N., & Burgess, P. (1998). The
development of an ecologically valid test for assessing patients with
dysexecutive syndrome. Neuropsychological Rehabilitation, 8(3), 213-228.
330
Wimmer, H., & Mayringer, H. (1998). False belief understanding in young children:
Explanations do not develop before predictions. International Journal of
Behavioral Development, 22(2), 403-422.
Wimmer, H., & Perner, J. (1983). Beliefs about beliefs: Representation and constraining
function of wrong beliefs in young children's understanding of deception.
Cognition, 13(1), 103-128.
Wing, L., & Gould, J. (1979). Severe impairments of social interaction and associated
abnormalities in children: Epidemiology and classification. Journal of Autism
and Developmental Disorders, 9, 11-29.
Wolff, S., Narayan, S., & Moyes, B. (1988). Personality characteristics of parents of
autistic children: A controlled study. Journal of Child Psychology & Psychiatry
& Allied Disciplines, 29(2), 143-53.
Wong, D., Maybery, M., Bishop, D. V. M., Maley, A., & Hallmayer, J. (2004). Profiles
of executive function performance in parents and siblings of individuals with
autism spectrum disorders. Manuscript in preparation.
World Health Organization. (1992). The ICD-10 classification of mental and behavioral
disorders: Clinical descriptions and diagnostic guidelines. Geneva, Switzerland:
Author.
Yirmiya, N., Erel, O., Shaked, M., & Solomonica-Levi, D. (1998). Meta-analyses
comparing theory of mind abilities of individuals with autism, individuals with
mental retardation, and normally developing individuals. Psychological Bulletin,
124(3), 283-307.
Yirmiya, N., & Shulman, C. (1996). Seriation, conservation, and theory of mind
abilities in individuals with autism, individuals with mental retardation, and
normally developing children. Child Development, 67(5), 2045-2059.
Yirmiya, N., Solomonica-Levi, D., Shulman, C., & Pilowsky, T. (1996). Theory of
mind abilities in individuals with autism, Down syndrome, and mental
retardation of unknown etiology: The role of age and intelligence. Journal of
Child Psychology & Psychiatry & Allied Disciplines, 37(8), 1003-1014.
Yonan, A. L., Alarcon, M., Cheng, R., Magnusson, P. K., Spence, S. J., Palmer, A. A.,
Grunn, A., Juo, S. H., Terwilliger, J. D., Liu, J., Cantor, R. M., Geschwind, D.
H., & Gilliam, T. C. (2003). A genomewide screen of 345 families for autism-
susceptibility loci. American Journal of Human Genetics, 73(4), 886-97.
331
Zelazo, P. D. (2000). Self-reflection and the development of consciously controlled
processing. In P. Mitchell & K. J. Riggs (Eds.), Children's reasoning and the
mind (pp. 169-189). Hove, UK: Psychology Press.
Zelazo, P. D., Burack, J. A., Benedetto, E., & Frye, D. (1996a). Theory of Mind and
rule use in individuals with Down's Syndrome: A test of the uniqueness and
specificity claims. Journal of Child Psychology & Psychiatry & Allied
Disciplines, 37(4), 479-484.
Zelazo, P. D., Burack, J. A., Boseovski, J. J., Jacques, S., & Frye, D. (2001). A
cognitive complexity and control framework for the study of autism. In J. A.
Burack, T. Charman, N. Yirmiya, & P. R. Zelazo (Eds.), The development of
autism: Perspectives from theory and research (pp. 195-217). Mahwah, NJ:
Lawrence Erlbaum Associates.
Zelazo, P. D., Carter, A., Reznick, J. S., & Frye, D. (1997). Early development of
executive function: A problem-solving framework. Review of General
Psychology, 1(2), 198-226.
Zelazo, P. D., & Frye, D. (1998). Cognitive complexity and control: II. The
development of executive function in childhood. Current Directions in
Psychological Science, 7(4), 121-126.
Zelazo, P. D., Frye, D., & Rapus, T. (1996b). An age-related dissociation between
knowing rules and using them. Cognitive Development, 11(1), 37-63.
Zelazo, P. D., Jacques, S., Burack, J. A., & Frye, D. (2002). The relation between theory
of mind and rule use: Evidence from persons with autism-spectrum disorders.
Infant & Child Development, 11(2), 171-195.
Zelazo, P. D., & Müller, U. (2002). Executive function in typical and atypical
development. In U. Goswami (Ed.), Blackwell handbook of childhood cognitive
development (pp. 445-469). Malden, MA: Blackwell Publishers.
Zelazo, P. D., & Reznick, J. (1991). Age-related asynchrony of knowledge and action.
Child Development, 62(4), 719-735.
Ziatas, K., Durkin, K., & Pratt, C. (1998). Belief term development in children with
autism, Asperger syndrome, specific language impairment, and normal
development: Links to theory of mind development. Journal of Child
Psychology & Psychiatry & Allied Disciplines, 39(5), 755-763.
Zilbovicius, M., Garreau, B., Samson, Y., Remy, P., Barthelemy, C., Syrota, A., &
Lelord, G. (1995). Delayed maturation of the frontal cortex in childhood autism.
American Journal of Psychiatry, 152(2), 248-252.
332
APPENDIX A Repetitive Behaviours Interview – Current Version
Instructions: In this interview I will ask you for details about some of the behaviours covered in the repetitive behaviours questionnaire which you would have completed in regard to each of your children. I’ll just be asking you about questions which you answered ‘yes’ to in that questionnaire. I’ll start by asking if [name] currently displays a particular behaviour and by this I mean a behaviour s/he has displayed once a week or more over the last 3 months. If s/he has, I’d like you to try to describe the behaviour, and I’ll also ask how often he/she shows this behaviour. [I’ll then ask you whether he/she has ever shown this behaviour at least once a week for a period of three months or more. If he/she has shown this behaviour in the past, I’d like you to try to describe the behaviour me and if possible to tell me at what age the behaviour was most frequent.]* Ask me questions at any time if things don’t seem clear. I am interested in all the repetitive behaviours shown by [name], so please tell me anything that you think may be of any interest. All the information you give me will be confidential. Any queries before we start? * [These instructions are only for parents of participants who are over the age of 12.]
333
STEREOTYPED MANIPULATION OF OBJECTS 1. Does [name] currently manipulate objects repetitively in any way? For example, does he/she spin, twiddle, bang, tap, twist, flick or wave objects or other materials repetitively? Has s/he ever? a) HOW OFTEN DOES HE/SHE DO THIS? b) HOW LONG DOES IT LAST? (O) never (1) less 60 secs (1) 1-2x’s per week (2) 1-3 mins (2) 3-6x’s per week (3) 4-9 mins (3) 1-4x’s per day (4) 10-29 mins (4) 5-14x’s per day (5) 30 mins + (5) 15-29x’s per day (9) no information (6) 30+x’s per day (99) not applicable (7) almost constantly (9) no information Describe objects and actions- 2. Does [name] currently operate light switches, taps, the toilet flush etc, repeatedly? Has s/he ever? a) HOW OFTEN DOES HE/SHE DO THIS? b) HOW LONG DOES IT LAST? (O) never (1) less 60 secs (1) 1-2x’s per week (2) 1-3 mins (2) 3-6x’s per week (3) 4-9 mins (3) 1-4x’s per day (4) 10-29 mins (4) 5-14x’s per day (5) 30 mins + (5) 15-29x’s per day (9) no information (6) 30+x’s per day (99) not applicable (7) almost constantly (9) no information Describe actions- 3. Does [name] currently arrange objects in rows or other patterns? Has s/he ever? a) HOW OFTEN DOES HE/SHE DO THIS? b) HOW LONG DOES IT LAST? (O) never (1) less 60 secs (1) 1-2x’s per week (2) 1-3 mins (2) 3-6x’s per week (3) 4-9 mins (3) 1-4x’s per day (4) 10-29 mins (4) 5-14x’s per day (5) 30 mins + (5) 15-29x’s per day (9) no information (6) 30+x’s per day (99) not applicable (7) almost consta tly n(9) no information c) DOES (NAME] ALWAYS LINE UP THE SAME OBJECTS IN THE SAME ORDER? (1) different objects and different order (2) same objects and different order (3) same objects and same (9) no information (99) not applicable d) DOES [NAME] SEEM TO NOTICE INSTANTLY IF AN OBJECT IS MISSING OR MOVED? (1) no (2) frequently (3) always (9) no information (99) not applicable
e) DOES [NAME] OBJECT IF THESE ROWS OR PATTERNS ARE MOVED OR PACKED AWAY? (1) no (2) frequently (3) always (9) no information (99) not applicable Describe objects and arrangements-
334
4. Does [name] currently mouth or suck objects or parts of him/herself repeatedly? For example, does he/she mouth or suck his/her fingers, a favourite object, his/her shirt collar or the like? Has s/he ever? a) HOW OFTEN DOES HE/SHE DO THIS? b) HOW LONG DOES IT LAST? (O) never (1) less 60 secs (1) 1-2x’s per week (2) 1-3 mins (2) 3-6x’s per week (3) 4-9 mins (3) 1-4x’s per day (4) 10-29 mins (4) 5-14x’s per day (5) 30 mins + (5) 15-29x’s per day (9) no information (6) 30+x’s per day (99) not applicable (7) almost constantly (9) no information Describe objects or body parts- 5. Does [name] currently stare closely at objects or his/her body parts? For example, does he/she stare at lights, spinning objects, a certain toy, his/her fingers etc.? Has s/he ever? a) HOW OFTEN DOES HE/SHE DO THIS? b) HOW LONG DOES IT LAST? (O) never (1) less 60 secs (1) 1-2x’s per week (2) 1-3 mins (2) 3-6x’s per week (3) 4-9 mins (3) 1-4x’s per day (4) 10-29 mins (4) 5-14x’s per day (5) 30 mins + (5) 15-29x’s per day (9) no information (6) 30+x’s per day (99) not applicable (7) almost constantly (9) no information Describe objects or body parts- 6. Does [name] currently obsessively collect or hoard items of any sort? Has s/he ever? (0) no obsessive, or unusually keen, collecting or hoarding (1) very keen collector of usual items (eg. stamps, football cards etc.) (2) hoards unusual or odd items (eg. leaflets, jar lids, sticks etc.), irregularly or on occasion and is reticent to throw
anything that has been collected away. (3) hoards unusual or odd items on a very regular basis, which, because of the volume of items hoarded, leads to
regular difficulties and conflicts (9) no information Details-
STEREOTYPED MOVEMENTS
7. Does [name] currently pace or move around repetitively? For example, does he/she walk to and fro across a room or around the house or garden repetitively? Has s/he ever? a) HOW OFTEN DOES HE/SHE DO THIS? b) HOW LONG DOES IT LAST? (O) never (1) less 60 secs (1) 1-2x’s per week (2) 1-3 mins (2) 3-6x’s per week (3) 4-9 mins (3) 1-4x’s per day (4) 10-29 mins (4) 5-14x’s per day (5) 30 mins + (5) 15-29x’s per day (9) no information (6) 30+x’s per day (99) not applicable (7) almost constantly (9) no information Describe movement, route and location-
335
8. Does [name] currently often spin him/herself around and around? Has s/he ever? a) HOW OFTEN DOES HE/SHE DO THIS? b) HOW LONG DOES IT LAST? (O) never (1) less 60 secs (1) 1-2x’s per week (2) 1-3 mins (2) 3-6x’s per week (3) 4-9 mins (3) 1-4x’s per day (4) 10-29 mins (4) 5-14x’s per day (5) 30 mins + (5) 15-29x’s per day (9) no information (6) 30+x’s per day (99) not applicable (7) almost constantly (9) no information Describe movement- 9. Does [name] currently rock rhythmically backwards and forwards, or side to side, either when sitting or when standing? Has s/he ever? a) HOW OFTEN DOES HE/SHE DO THIS? b) HOW LONG DOES IT LAST? (O) never (1) less 60 secs (1) 1-2x’s per week (2) 1-3 mins (2) 3-6x’s per week (3) 4-9 mins (3) 1-4x’s per day (4) 10-29 mins (4) 5-14x’s per day (5) 30 mins + (5) 15-29x’s per day (9) no information (6) 30+x’s per day (99) not applicable (7) almost constantly (9) no information Describe whether sitting or standing- 10. Does [name] currently touch parts of his/her body or clothing repeatedly? For example, does he/she repeatedly rub his/her legs, pull at the buttons on his/her clothing, or touch his/her ear or elbow etc.? Has s/he ever? a) HOW OFTEN DOES HE/SHE DO THIS? b) HOW LONG DOES IT LAST? (O) never (1) less 60 secs (1) 1-2x’s per week (2) 1-3 mins (2) 3-6x’s per week (3) 4-9 mins (3) 1-4x’s per day (4) 10-29 mins (4) 5-14x’s per day (5) 30 mins + (5) 15-29x’s per day (9) no information (6) 30+x’s per day (99) not applicable (7) almost constantly (9) no information Describe action and body part or clothing- 11. Does [name] currently make repetitive arm, hand and/or finger movements? For example, does he/she repetitively wave, flick, flap or twiddle his/her hands or fingers repetitively? Does he/she repetitively clap or clasp his/her hands? Has s/he ever? a) HOW OFTEN DOES HE/SHE DO THIS? b) HOW LONG DOES IT LAST? (O) never (1) less 60 secs (1) 1-2x’s per week (2) 1-3 mins (2) 3-6x’s per week (3) 4-9 mins (3) 1-4x’s per day (4) 10-29 mins (4) 5-14x’s per day (5) 30 mins + (5) 15-29x’s per day (9) no information (6) 30+x’s per day (99) not applicable (7) almost constantly (9) no information Describe movements and whether this occurs near his/her eyes-
336
12. Does [name] currently make any repetitive movements with his/her feet or legs? For example, does he/she repetitively tap his/her feet, swing his/her legs or jump etc.? Has s/he ever? a) HOW OFTEN DOES HE/SHE DO THIS? b) HOW LONG DOES IT LAST? (O) never (1) less 60 secs (1) 1-2x’s per week (2) 1-3 mins (2) 3-6x’s per week (3) 4-9 mins (3) 1-4x’s per day (4) 10-29 mins (4) 5-14x’s per day (5) 30 mins + (5) 15-29x’s per day (9) no information (6) 30+x’s per day (99) not applicable (7) almost constantly (9) no information Describe movements-
TIC-LIKE BEHAVIOURS 13. Does [name] currently make any particular words, noises etc. that he/she uses repeatedly? For example, does he/she repeat single words or nonsense words? Or other sounds such as hums, growls, clicking of the tongue, or clearing the throat? Has s/he ever? a) HOW OFTEN DOES HE/SHE DO THIS? b) HOW LONG DOES IT LAST? (O) never (1) less 60 secs (1) 1-2x’s per week (2) 1-3 mins (2) 3-6x’s per week (3) 4-9 mins (3) 1-4x’s per day (4) 10-29 mins (4) 5-14x’s per day (5) 30 mins + (5) 15-29x’s per day (9) no information (6) 30+x’s per day (99) not applicable (7) almost constantly (9) no information Describe words, noises etc- 14. Does [name] currently make any repetitive head or neck movements? For example, does he/she nod or shake his/her head repetitively, or show any jerky tic-like movements? Or does he/she show other repetitive movements of the face muscles such as raising eyebrows or moving the muscles around the lips? Has s/he ever? a) HOW OFTEN DOES HE/SHE DO THIS? b) HOW LONG DOES IT LAST? (O) never (1) less 60 secs (1) 1-2x’s per week (2) 1-3 mins (2) 3-6x’s per week (3) 4-9 mins (3) 1-4x’s per day (4) 10-29 mins (4) 5-14x’s per day (5) 30 mins + (5) 15-29x’s per day (9) no information (6) 30+x’s per day (99) not applicable (7) almost constantly (9) no information Describe movements-
337
15. Does [name] currently make any repetitive eye movements? For example, does he/she blink, roll or move his/her eyes repeatedly? Has s/he ever? a) HOW OFTEN DOES HE/SHE DO THIS? b) HOW LONG DOES IT LAST? (O) never (1) less 60 secs (1) 1-2x’s per week (2) 1-3 mins (2) 3-6x’s per week (3) 4-9 mins (3) 1-4x’s per day (4) 10-29 mins (4) 5-14x’s per day (5) 30 mins + (5) 15-29x’s per day (9) no information (6) 30+x’s per day (99) not applicable (7) almost constantly (9) no information Describe movements- 16. Does [name] currently make any repetitive mouth and/or tongue movements? For example, does he/she grind his/her teeth, smack his/her lips, or make sucking movements repetitively? Has s/he in the past? a) HOW OFTEN DOES HE/SHE DO THIS? b) HOW LONG DOES IT LAST? (O) never (1) less 60 secs (1) 1-2x’s per week (2) 1-3 mins (2) 3-6x’s per week (3) 4-9 mins (3) 1-4x’s per day (4) 10-29 mins (4) 5-14x’s per day (5) 30 mins + (5) 15-29x’s per day (9) no information (6) 30+x’s per day (99) not applicable (7) almost constantly (9) no information Describe movements-
SELF-INJURIOUS BEHAVIOUR 17. Does [name] currently bang his/her head? Does he/she do this
repeatedly? Has s/he in the past? a) HOW OFTEN DOES HE/SHE DO THIS? b) HOW LONG DOES IT LAST? (O) never (1) less 60 secs (1) 1-2x’s per week (2) 1-3 mins (2) 3-6x’s per week (3) 4-9 mins (3) 1-4x’s per day (4) 10-29 mins (4) 5-14x’s per day (5) 30 mins + (5) 15-29x’s per day (9) no information (6) 30+x’s per day (99) not applicable (7) almost constantly (9) no information Describe what head is banged against- 18. Does [name] currently ever injure himself/herself? For example does he/she bite, scratch, knock or pick at himself/herself? Does he/she do this repeatedly? Has s/he in the past? a) HOW OFTEN DOES HE/SHE DO THIS? b) HOW LONG DOES IT LAST? (O) never (1) less 60 secs (1) 1-2x’s per week (2) 1-3 mins (2) 3-6x’s per week (3) 4-9 mins (3) 1-4x’s per day (4) 10-29 mins (4) 5-14x’s per day (5) 30 mins + (5) 15-29x’s per day (9) no information (6) 30+x’s per day (99) not applicable (7) almost consta tly n(9) no information
338
GENERAL 19. Has [name] always shown one or more of these behaviours, or have there been periods when he/she hasn't shown any repetitive behaviours for 3 months or more? (1) at times has shown no repetitive behaviours for 3 months or more (2) has always shown one or more behaviours (3) has always shown at least one repetitive activity (9) no information (99) not applicable- items 1-18 all received a (0) rating Details of time periods-
COMPULSIVE BEHAVIOURS 20. Cleaning/Washing Compulsions: Does [name] currently wash his/her hands, shower, bathe or groom himself/herself, more than is necessary? Is he/she overly concerned about dirt and contamination, or take measures to prevent contact with contaminants? Does s/he clean household items or other objects excessively? Has s/he in the past? (0) no obsessive-or compulsive behaviour of this type- washes hands at appropriate times (e.g. at meal times, after
using the toilet), but does not consistently wash at inappropriate times. Is not unusually concerned about dirt or contamination.
(1) suspicious or mild obsessive or compulsive behaviour- washes hands 10 -14 times a day (2) clear obsessive or compulsive behaviour- washes hands 15+ times a day, or is preoccupied with worry about dirt
and contamination (9) no information b) IS THIS WASHING BEHAVIOUR CARRIED OUT IN A RITUALISED FASHION? (i.e. is it always carried out in the same order or in the same way) (1) no (2) frequently (3) always (9) no information (99) not applicable Describe cleaning/washing behaviour - 21. Checking Compulsions: Does [name] currently often check repeatedly that things are switched off, locked up or put away etc? Does s/he check other things like that nothing bad has happened, or that s/he did not make a mistake? Does he/she check these things more often than is necessary? Has s/he in the past? (0) no obsessive-or compulsive behaviour of this type- may check that an item has been switched off ·etc. once, but is
not preoccupied with whether or not items have been checked (1) suspicious or mild obsessive or compulsive behaviour- checks that one or more items have been turned off etc. on
two separate occasions on a daily basis (2) clear obsessive or compulsive behaviour- checks that one or more items has been switched off etc. on at least
three separate occasions on a daily basis, or is preoccupied with items being safely handled in order to avert disaster
(9) no information b) IS THIS CHECKING BEHAVIOUR CARRIED OUT IN A RITUALISED FASHION? (i.e. is it always carried out in the same order or in the same way) (1) no (2) frequently (3) always (9) no information (99) not applicable Describe checking and items checked-
339
22. Repeating Rituals: Does [name] currently perform any rituals where s/he has to keep repeating a certain action? For example, does s/he reread or rewrite excessively, or repeat routine activities such as going in and out of a door or getting up and down from a chair? (0) no obsessive-or compulsive behaviour of this type- (1) suspicious or mild obsessive or compulsive behaviour- performs repeating routine 3-10 times a day (2) clear obsessive or compulsive behaviour- performs routine 10+ times a day (9) no information b) IS THIS REPEATING BEHAVIOUR CARRIED OUT IN A RITUALISED FASHION? (i.e. is it always carried out in the same order or in the same way) (1) no (2) frequently (3) always (9) no information (99) not applicable Describe repeating ritual- 23. Counting Compulsions: Does [name] currently count objects repeatedly? Does s/he perform any rituals, which involve counting? Has s/he in the past? (0) no obsessive-or compulsive behaviour of this type- may count money or other objects but not excessively or
inappropriately (1) suspicious or mild obsessive or compulsive behaviour- counts objects inappropriately less than 5 times a day (2) clear obsessive or compulsive behaviour- counts objects more than 5 times per day (9) no information b) IS THIS COUNTING BEHAVIOUR CARRIED OUT IN A RITUALISED FASHION? (i.e. is it always carried out in the same order or in the same way) (1) no (2) frequently (3) always (9) no information (99) not applicable Describe counting behaviour -
24. Does [name] currently engage in any other compulsive behaviours? For example does s/he write lists excessively? Does s/he repeatedly touch, tap or rub certain things? Any other superstitious behaviours? Has s/he in the past? (0) no obsessive-or compulsive behaviour of this type- (1) suspicious or mild obsessive or compulsive behaviour- (2) clear obsessive or compulsive behaviour (9) no information b) IS THIS COMPULSIVE BEHAVIOUR CARRIED OUT IN A RITUALISED FASHION? (i.e. is it always carried out in the same order or in the same way) (1) no (2) frequently (3) always (9) no information (99) not applicable Describe compulsive behaviour - 25. How much time do you think s/he spends on these compulsive behaviours per day? (1) 0-1 hrs/day (2) 1-3 hrs/day (3) 3-8 hrs/day (4) >8 hrs/day (9) no information (99) not applicable
340
OBJECT ATTACHMENTS 26. Is [name] currently attached to any particular objects? For example, does he/she carry a teddy, a blanket or a stick etc. around with him/her? Does he/she want to sleep with this item? Does he/she become distressed if it is lost or forgotten? Has s/he in the past? [In order to be considered an object attachment the individual must insist on sleeping with the item, or must carry it with him/her at specific times or in specific situations (e.g. whenever out of the house). The individual should also be concerned or distressed if the item is mislaid. (0) no attachments to objects (1) attachments to objects which are commonly used as comforters (e.g. teddies, blankets etc.) (2) attachments to unusual objects or junk materials (eg. sticks, tins etc.). Rate here even if unusual attachments cc-
exist with more usual object attachments
(9) no information b) [If score on the previous item is (1) or (2), then also complete the following item] (1) insists that the object must be in bed every night, but only when the individual is at home (2) insists that the object must be in bed every night whether the individual is at home or away (3) insists that the object must be with the individual at times other than when tired or sleeping (9) no information Describe objects-
G. INSISTENCE ON SAMENESS OF ENVIRONMENT
27. Does [name] currently insist on things about the house staying the same? For example, does he/she insist on furniture staying in the same place, or curtains being open or closed etc.? (0) No fixed insistence on furniture, ornaments etc. remaining in the same places simply because he/she doesn't like
things to be moved (1) any relatively inflexible example which does not impact on other family members daily, as it primarily concerns
items that belong to, or are used by, the individual only, or if this is not the case, he/she is able to tolerate alterations when others are present
(2) any pervasive example which is very rigid and impacts on the other members of the family on a daily basis (e.g. having to have lounge furniture organised in a particular way, or insisting that everybody's bedroom door must be closed etc, at all times)
(9) no information Describe items and location- 28. Does [name] currently insist on other items being put out, kept or stored in the same way? For example, does he/she like ornaments, toys or cassette tapes kept in the same places or positions? Has s/he in the past? (0) no fixed insistence that items must be stored in the same places or the same way (1) any example which does not interfere with other family members on a daily basis, as it primarily concerns the
individuals own personal possessions although it may be very inflexible a (e.g. the arrangement of personal toiletries- he/she will not tolerate others moving them, even when cleaning the bathroom).
(2) any pervasive example which is very rigid and impacts on the other members of the family daily (e.g. insisting that a family video collection must always be stored in precisely the same way.)
(9) no information Describe items and location-
341
29. Is there anything else that [name] currently likes to remain just so? Has s/he in the past? (0) no (1) yes, any relatively inflexible example which is consistently observed by the individual, but has only a limited impact
on the family (i.e. does not impact on the remainder of the family on a daily basis) (2) yes, any pervasive example which is highly rigid and impacts on the other family members on a daily basis (9) no information Describe- 30. Does [name] currently play the same music, game or video, or read the same book repeatedly? Has s/he in the past? (0) does not have any music, games, videos or books that s/he uses more than normal (1) plays the same, music, game or video or reads the same book (excepting continuing on with a novel) at least
once a day (2) plays the same, music, game or video or reads the same book (excepting continuing on with a novel) at least
three times a day and prevention or interruption of this activity causes a marked negative reaction (9) no information Describe the book, game or music- 31. Does [name] currently insist on using the same objects or items in any other situation? For example, does he/she insist on using the same chair, plate, bed linen or door? Has s/he in the past? [Do note rate insistence on using the same mug or cup] (0) no fixed insistence on always using precisely the same items (excepting a mug or cup) in any situation- will
generally use any item that he/she is given or the first item that is available (1) any example which is unusually restricted or fixed, but can generally be modified if it is important to do so (e.g. if
the item is in the dishwasher, if someone else is using it) (2) any pervasive example which is very rigid and leads to regular confrontations with others, or requires extra effort
on the part of the individual or others (e.g. insisting on using a certain plate etc. even if it is dirty or someone else is using it), on a regular or daily basis
(9) no information Describe item and situation- 32. Does [name] currently insist on wearing the same clothes or refuse to wear new clothes? Has s/he in the past? (0) no insistence on wearing the same items of clothes- wears a range of different items and is keen to have new
clothes (1) insists on wearing the same item of clothing (e.g. jumper, trousers), in most situations, including frequently when it
is inappropriate. Or refuses, or shows marked reticence, to wear new clothes. Will wear alternative clothing for at least certain, or special, occasions if prompted.
(2) insists on wearing the same (or substantially the same), outfit most or all of the time so that it is difficult for this outfit to be washed and any deviation from this usual outfit causes an extreme negative reaction.
(9) no information Describe clothing- 33. Does [name] currently insist that certain items of clothing must always be worn, or worn in the same situation or in the same way? For example, does he/she insist on always wearing a vest, or wearing a hat to the shops, or always buttoning a shirt to the collar? Has s/he in the past? (0) no unusually fixed ways of wearing clothes- will modify clothing and the way in which it is warn etc. as appropriate
(e.g. will take off coat if hot, or if wet or dirty etc.) (1) consistently dresses in the same fixed manner, or wears the same clothes in the same situations, in a manner that
is odd or unusual, but can modify this behaviour if it is necessary or important to do so (e.g. generally wears tops done up and with the hood up, but will undo this if it is hot etc.)
(2) has very fixed ways of wearing clothes, or always wears the same clothes in the same situations, and this is adhered to strictly even when it is very odd and impractical (e.g. always wears hat to the shops, always wears a coat outside irrespective of the weather)
(9) no information Describe clothing and situation-
342
34. Does [name] currently insist on eating the same foods, or a very small range of foods, at every meal? Has s/he in the past? (0) eats a range of foods, although there may be a limited number of foods that he/she doesn't like to eat (1) eats a limited range of foods and it is regularly the case that the he/she will eat a different meal to the rest of the
family- will not try new foods (2) eats fewer than five separate food types (9) no information Describe foods- 35. How does [name] respond if you introduce him/her to a new activity or place? Would he/she have any objection to trying something new and different? Would he/she be anxious? [Rate usual, or most common, reaction] (0) participates/will visit without hesitation (1) will be persuaded, but shows some reticence because the activity/place is new or different (2) refuses to take part in anything new or different (3) shows a high degree of stereotyped behaviour when trying something new (9) no information Describe reaction-
RIGID ADHERENCE TO ROUTINES AND RITUALS 36. Are there any aspects of routine that [name] currently insists must remain the same? For example, does he/she insist on always bathing before breakfast, on going to the shops every afternoon, or on watching a video after every meal? Has s/he in the past? (0) has no rigid routine - preferred routines can be modified if it is necessary or appropriate to do so (1) has a set routine which is inflexible and consistently impacts on other family members because he/she is unable to
take "shortcuts" in his/her routine (e.g. the individual is unable to finish early in the bathroom if someone needs it, or take their walk on another day if a family outing is planned etc.)
(2) has a very fixed or inflexible routine which involves not just the self but also other family members and so has a substantial impact on the family (e.g. expects everyone to go swimming on a Saturday morning and is upset if this routine is violated)
(9) no information Describe routine- 37. Does [name] currently make rituals out of everyday activities such as eating, dressing, getting in the car, walking up stairs etc.? Are these activities always carried out in exactly the same way? Has s/he in the past? (0) has no regular rituals or set ways of doing things- preferred ways of doing things can be modified if it is appropriate
to do so (e.g. may always put socks on first, but if no clean socks are available will put on other items of clothing first)
(1) has set rituals, which are inflexible and impact on other family members to some degree because the individual is unable to modify these rituals when it is important to do so. These rituals concern the individual only and are not excessively time-consuming. They do not incorporate unnecessary or redundant steps and actions.
(2) has very elaborate and inflexible rituals which may, or may not, involve others, but take considerable time (i.e. take significantly more time than the same activity would take in non-ritualised fashion) and cannot be abbreviated. These rituals affect all family members because of the large amounts of time taken up with these rituals on a daily basis (e.g. having to check that every bodies seat belt is fastened and that the glove box contains certain items before setting out on any car journey, no matter how short.)
(9) no information Describe activity and precise ritual-
343
38. Does [name] currently have any rituals that are linked to particular occasions or places? For example, does he/she have specific rituals for the supermarket, the Doctor's surgery or a relative's house? Has s/he in the past? (0) has no fixed rituals for particular places or occasions- preferred ways of doing things can be modified if it is
appropriate to do so (e.g. if in a hurry, if the weather is not appropriate etc.) (1) has certain fixed activities or rituals that he/she insists on at particular occasions or particular places. These rituals
concern the individual only and have minimal impact on the remainder of the family (e.g. always rides on the swings in the same fixed order, or always orders the same food in a cafe)
(2) has one or more very fixed and inflexible rituals which have a severe impact on the family as it is highly intrusive or involves other family members (e.g. must always enter certain shops in certain order when shopping)
(9) no information Describe ritual and occasion or place- 39. Does [name] currently insist on moving or travelling by the same route? For example, does he/she insist on taking the same route when moving about the house, going for a walk, or travelling in the car? Has s/he in the past? (0) has no set route for moving or travelling- preferred ways of doing things can be modified if it is appropriate to do so (1) has a set route that he/she will always take to one or more specific locations if on his/her own or if given the
choice. Finds it very difficult to accept deviations from this, but will accept an alternative if there is a good reason for doing so.
(2) will take only one route to at least one specific destination and will not tolerate any deviation from this, no matter what the need or justification for the change is.
(9) no information Describe mode of travelling and journey- 40. Is there anything else that[name] currently likes to be done in a certain way, or at a certain time? Has s/he in the past? (0) no (1) yes, any relatively inflexible example which is consistently observed by the individual, but has only a limited impact
on the family (i.e. does not impact on the remainder of the family on a daily basis) (2) yes, any pervasive example which is highly rigid and impacts on the other family members on a daily basis no
information (9) no information Describe- 41. Does [name] currently incorporate any unnecessary, or unusual, behaviours as part of any rituals or routines? For example, does he/she tap the plate after every mouthful when eating, or touch specific objects when walking through a room? Has s/he in the past? (0) no unnecessary, idiosyncratic behaviours incorporated in routines (1) yes, any relatively inflexible example which is consistently observed by the individual, but has a limited impact on
the family- he/she can refrain from the behaviour when asked to do so for at least 10 minutes
(2) yes, any pervasive and unusual example, which is very rigid and is observed by the individual at all times. He/she is unable (or unwilling) to suppress this behaviour.
(9) no information Describe ritual or routine and unnecessary activity-
344
REPETITIVE USE OF LANGUAGE 42. Does [name] currently mimic others or repeat speech? Has s/he in the past? a) HOW OFTEN DOES HE/SHE DO THIS? b) HOW LONG DOES IT LAST? (O) never (1) less 60 secs (1) 1-2x’s per week (2) 1-3 mins (2) 3-6x’s per week (3) 4-9 mins (3) 1-4x’s per day (4) 10-29 mins (4) 5-14x’s per day (5) 30 mins + (5) 15-29x’s per day (9) no information (6) 30+x’s per day (99) not applicable (7) almost constantly (9) no information C) DOES [NAME] (A) REPEAT WHAT IS SAID IMMEDIATELY AFTER IT IS SAID OR, (B) REPEAT WHAT HAS BEEN SAID SOME TIME AFTER IT HAS BEEN SAID? (1) A (2) B (3) combination of A and B (9) no information (99) not applicable Describe the type of speech repeated- Items 43-45 inclusive specifically address spontaneous language and exclude echolalia, or language that is copied from other sources. If [name] does not have at least good phrase speech, skip items 43-45 (and score (99), not applicable). 43. Does [name] currently say the same things, or sing the same songs, repeatedly? For example, does [name] recite the same thing over and over, or have stock phrases that he/she often uses? Has s/he in the past? a) HOW OFTEN DOES HE/SHE DO THIS? b) HOW LONG DOES IT LAST? (O) never (1) less 60 secs (1) 1-2x’s per week (2) 1-3 mins (2) 3-6x’s per week (3) 4-9 mins (3) 1-4x’s per day (4) 10-29 mins (4) 5-14x’s per day (5) 30 mins + (5) 15-29x’s per day (9) no information (6) 30+x’s per day (99) not applicable (7) almost constantly (9) no information c) DOES [NAME] (A) SAY THE SAME THING OVER AND OVER AGAIN AT ONE POINT IN TIME OR, (B) SAY THE SAME THING AT DIFFERENT TIMES? (1)A (2) B (3) combination of A and B (9) no information· (99) not applicable Describe sentences or songs- 44. Does [name] currently ask the same questions repeatedly? Has s/he in the past? a) HOW OFTEN DOES HE/SHE DO THIS? b) HOW LONG DOES IT LAST? (O) never (1) less 60 secs (1) 1-2x’s per week (2) 1-3 mins (2) 3-6x’s per week (3) 4-9 mins (3) 1-4x’s per day (4) 10-29 mins (4) 5-14x’s per day (5) 30 mins + (5) 15-29x’s per day (9) no information (6) 30+x’s per day (99) not applicable (7) almost constantly (9) no information c) DOES [NAME] (A) SAY THE SAME THING OVER AND OVER AGAIN AT ONE POINT IN TIME OR, (B) SAY THE SAME THING AT DIFFERENT TIMES? (1) A (2) B (3) combination of A and B (9) no information (99) not applicable d) DOES HE/SHE DEMAND THAT OTHERS ALWAYS GIVE THE SAME ANSWERS? (1) no (2) frequently (3) always (9) no information (99) not applicable Describe questions -
345
45. Does [name] currently talk about the same topic over and over again? Has s/he in the past? [Rate only repeated attempts to raise the same topic in conversation. These attempts may incorporate some echoed speech, but must also include spontaneous speech and attempts to talk around the topic.] a) HOW OFTEN DOES HE/SHE DO THIS? b) HOW LONG DOES IT LAST? (O) never (1) less 60 secs (1) 1-2x’s per week (2) 1-3 mins (2) 3-6x’s per week (3) 4-9 mins (3) 1-4x’s per day (4) 10-29 mins (4) 5-14x’s per day (5) 30 mins + (5) 15-29x’s per day (9) no information (6) 30+x’s per day (99) not applicable (7) almost constantly (9) no information Describe topic and whether it’s based on fantasy or reality-
CIRCUMSCRIBED INTERESTS
46. Does [name] have any unusual preoccupations? Does he/she regularly talk about and seek out a particular type of object? Has s/he ever? (0) no preoccupations or preoccupation with objects that are common in their age group and not to the exclusion of
other interests or activities (1) preoccupation with items common in their age but to such a degree that it significantly limits involvement in other
interests or activities (2) preoccupation with unusual items (9) no information How long has [name] been preoccupied with [interest]? Please describe the preoccupation- 47a. Does [name] have any particular interests? Is there anything unusual about this interest? Would you describe this interest as particularly keen or obsessional? Does he/she pursue this interest to the exclusion of other interests and hobbies? What other interests and hobbies does [name] have? Has s/he ever had any unusual or obsessional interests? (0) usual topic of hobby or interest (e.g. computers or football teams)- casual to keen interest (1) usual topic of hobby or interest (e.g. computers or football teams)- abnormally keen or obsessional interest OR
mildly unusual topic of hobby or interest (e.g. road maps or record covers)- casual to keen interest (2) mildly unusual topic of hobby or interest (e.g. road maps or record covers) – abnormally keen or obsessional
interest (3) abnormally keen or obsessional interest in highly unusual topic of hobby or interest (e.g. DIY tools or street lamps)
- abnormally keen or obsessional interest (9) no information How long has [name] had this particular interest(s)? Please describe the interest(s)- 47b. Summary Rating (0) has a varied pattern of interests, which are pursued meaningfully. (1) one or more abnormally keen or highly circumscribed interests, but also more usual interests which are pursued
meaningfully. (2) has only obsessional interests which are either pursued to an abnormally keen extent, or are highly circumscribed
in nature (3) has no particular interests or hobbies that he/she will pursue spontaneously (DO NOT RATE WATCHING
TELEVISION) (9) no information 47c. How is this interest or hobby manifested (0) usual manifestation of interest- collecting, sorting, reading, playing/using relevant materials (1) mildly unusual or idiosyncratic manifestation of interest- odd or unusual activity (2) highly unusual or idiosyncratic manifestation of interest- highly stereotyped or ritualised activity (9) not applicable- item received a (O) rating Please describe how it is manifest-
346
GENERAL ITEMS Skip items 48-52 inclusive (and score (99), not applicable), if all interview items have received a (O) rating. 48a. Does [name] ever make any attempt to cover up, hide or change any of the behaviours you have described? For example, does he/she leave the room to engage in repetitive activities, or does he/she suppress them if he/she knows that other people are watching? Has s/he in the past? (0) never (1) occasionally- but not at specific or predictable times (2) most often- but not at specific or predictable times (3) at all times (4) only, or mainly, when calm and relaxed (5) only, or mainly, at school (6) only, or mainly, with new people or in social situations (excluding solely school) (7) only, or mainly, when likely to be reprimanded (8) at other times (9) no information (99) not applicable- all interview items received a (O) rating Describe the way in which the behaviour has been covered up- 48b. Which behaviours? [Rate the category that the behaviour that s/he attempts to cover belongs to.] (1) repetitive movements – (a) stereotypies (b) repetitive use of objects (c) tic like movements (d) self injurious behaviours (2) object attachments (3) insistence on sameness of environment (4) insistence on sameness of activity or item (5) adherence to routine and rituals (6) repetitive use of language (7) circumscribed interests (8) compulsive behaviours (9) no information (99) not applicable- all interview items received a (O) rating Briefly describe the behaviour- 49. Have you, or anyone else, ever made any attempt to reduce any of the behaviours shown by [name] that we have talked about? (1) no (2) yes, at different times (3) yes, continually and consistently (9) no information (99) not applicable - all interview items received a (0) rating 50. What was the earliest repetitive activity that you remember [name] showing? How old was he/she when this began? [Rate the category that this activity belongs to and the age at which it began.] (1) repetitive movements – (a) stereotypies (b) repetitive use of objects (c) tic like movements (d) self injurious behaviours (2) object attachments (3) insistence on sameness of environment (4) insistence on sameness of activity or item (5) adherence to routine and rituals (6) repetitive use of language (7) circumscribed interests (8) compulsive behaviours (9) no information (99) not applicable- all interview items received a (O) rating
347
[The following two items apply only to repetitive activities which have been evident during the last three months.] 51a. Of the repetitive behaviours and rituals and special interests that we have discussed, which one would you say is the most marked or the most noticeable? [Rate the category that this activity belongs to.] (1) repetitive movements – (a) stereotypies (b) repetitive use of objects (c) tic like movements (d) self injurious behaviours (2) object attachments (3) insistence on sameness of environment (4) insistence on sameness of activity or item (5) adherence to routine and rituals (6) repetitive use of language (7) circumscribed interests (8) compulsive behaviours (9) no information (99) not applicable- all interview items received a (O) rating b. Which would come second? c. Which would you think comes third? 52a. Of all of the repetitive behaviours and rituals and special interests etc. that we have talked about, which one would you say causes the greatest problem in day-to-day life? (1) repetitive movements – (a) stereotypies (b) repetitive use of objects (c) tic like movements (d) self injurious behaviours (2) object attachments (3) insistence on sameness of environment (4) insistence on sameness of activity or item (5) adherence to routine and rituals (6) repetitive use of language (9) circumscribed interests (10) compulsive behaviours (9) no information (99) not applicable- all interview items received a (O) rating b. Which would you think comes second? c. Which would you think comes third?
348
APPENDIX B Correlations between EF task variables in the control group (Study One)
Table B1. Raw correlations between EF variables in the control group (N.B.: Intra-domain correlations are depicted in bold) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 181 2 .30*
.
3 05 .22
4 .11 .17 .04 5 -.07 -.20 -.42* -.266 .11 -.06 .18 .10 -.03 7 .02 -.19
-.22
-.15
.79** .59** 8 .01 .05 .06 .04 -.06 -.07 -.099 -.04 .10 .24 .16 .15 -.11 .06 -.07 10 .14 .14 .28 -.02
-.08
-.22 -.19 .05 .45** 11 -.40**
-.16 -.25 .19 .04 -.18 -.09 -.33 -.14 -.34*
12 -.11 -.09 -.17 -.16 .16 .12 .20 -.49** -.14 -.19 .29*13 .35* .15 -.02 -.28 -.02 .09 .04 .04 -.05 .11 -.51** .18 14 -.44** -.32* .20 -.01 -.24 .12 -.12 -.31 -.24 -.14 .47** .45** -.25 15 .29* .17 -.24 -.32 .26 -.04 .19 -.11 .02 .27 -.36* .33* .48** -.45** 16 -.25 -.14 -.42* .12 .08 .11 .13 .07 -.03 -.13 .08 -.02 -.13 .18 -.12 17 -.23 -.40** -.40* -.08 .14 -.01 .10 -.23 -.05 -.22 .30* .24 -.21 .50** -.28 .51**
18 .05 .27 .27 .13 -.73** -.34 -.78** .15 .03 .34* -.02 -.13 -.01 .04 -.07 -.12 -.10 19 .11 .15 .13 -.34 .42* -.38* .02 -.11 .13 .07 .02 -.11 -.09 -.13 -.10 -.53** -.28 a 1 = ToL adjusted extra moves score; 2 = ToL rule violations; 3 = IDED set-shifting task Perseveration Condition EDS stage errors; 4 = IDED set-shifting task Learned Irrelevance Condition EDS stage errors; 5 = RIL task inhibition error difference score; 6 = RIL task load error difference score; 7 = RIL task inhibition + load error difference score; 8 = RIL task shape error score; 9 = Opposite Worlds error difference score; 10 = Opposite Worlds time difference score; 11 = Relational Complexity total score; 12 = Pattern Meanings correct responses; 13 = Pattern Meanings sum of errors; 14 = Uses of Objects correct responses; 15 = Uses of Objects sum of errors; 16 = Stamps task complexity score; 17 = Stamps task originality score; 18 = Stamps task restriction score; 19 = Stamps task rule adherence score. *p < .05; ** p < .01. All tests were two-tailed. a = Correlation could not be computed because one of the variables was constant. Note: The RIL task RT difference scores are not included in this table for the sake of brevity.
349
Table B2. Partial correlations between EF variables in the control group (N.B.: Intra-domain correlations are depicted in bold) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 181 2 .14 3 .03 .18
4 .15 .21 .07 5 -.11 -.23 -.50** -.226 .11 -.07 .19 .09 .0 7 -.02 -.22 -.26
-.10
.76**
.65** 8 -.28 -.15 .02 .03 .03 -.10 -.049 .01 .10 .22 .17 .21 -.10 .10 -.09 10 .04 .03 .24 .06 -.27 -.22 -.35 -.01 .50**
11 -.01 .21 -.26 .24 .08 -.24 -.09 -.04 -.23 -.2312 -.02 .01 -.14 -.17 .16 .12 .20 -.44* -.12 -.14 .21 13 .05 -.07 -.08 -.31 -.07 .09 .0 -.24 -.03 -.04 -.18 .36* 14 -.20 -.11 .33 -.03 -.37* .18 -.17 -.07 -.31 .0 .06 .41** .10 15 .07 .0 -.32 -.34 .32 -.06 .21 -.36* .03 .20 -.07 .49** .31* -.26 16 -.19 -.10 -.43* .10 .18 .11 .21 .12 -.06 -.05 -.05 -.04 -.03 .12 -.05 17 .0 .0 -.38* -.11 .16 -.01 .12 -.03 -.04 -.11 -.09 .14 .05 .31* -.11 .53**
18 .02 .02 .26 .11 -.74** -.36 -.79** .07 .0 .41* .11 -.08 -.07 .16 -.13 -.15 -.01 19 .09 .09 .11 -.31 .37* -.39* -.06 -.12 .15 -.01 .11 -.11 -.16 -.13 -.15 -.51** -.30 a 1 = ToL adjusted extra moves score; 2 = ToL rule violations; 3 = IDED set-shifting task Perseveration Condition EDS stage errors; 4 = IDED set-shifting task Learned Irrelevance Condition EDS stage errors; 5 = RIL task inhibition error difference score; 6 = RIL task load error difference score; 7 = RIL task inhibition + load error difference score; 8 = RIL task shape error score; 9 = Opposite Worlds error difference score; 10 = Opposite Worlds time difference score; 11 = Relational Complexity total score; 12 = Pattern Meanings correct responses; 13 = Pattern Meanings sum of errors; 14 = Uses of Objects correct responses; 15 = Uses of Objects sum of errors; 16 = Stamps task complexity score; 17 = Stamps task originality score; 18 = Stamps task restriction score; 19 = Stamps task rule adherence score. *p < .05; ** p < .01. All tests were two-tailed. a = Correlation could not be computed because one of the variables was constant. Note: The RIL task RT difference scores are not included in this table for the sake of brevity.
350
APPENDIX C Separate ToM-EF correlations for young and old age subgroups within the
control sample (Study One)
Table C1. Raw and partial correlations between ToM and EF variables within “young” control participants (aged 5-8 years) False belief task EF task Simple 1st-order 2nd-order ToL (n = 25): Adj.extra move score -.34 -.46* -.48* -.39 Rule violations -.11 -.02 -.60** -.55** IDED Set-shifting task condition (n = 13): Perseveration EDS stage errors a -.68* -.90*** -.64* -.84** Learned Irrelevance EDS stage errors a -.19 -.08 RIL task (n = 12): Error difference scores: Inhibition a .06 .02 Load a -.64* -.80* -.51 Inhibition + load a -.69* -.66 -.57 RT difference scores: Inhibition a .18 -.45 Load a .21 .47 Inhibition + load a .46 .10 Shape error score a -.31 -.12 Opposite Worlds (n = 14): Error diff. score a -.24 -.11 Time diff. score a -.11 -.14 Relational Complexity (n = 25): Total score .14 .38 .40* .13 Pattern Meanings (n = 25): Correct responses -.19 .39 .17 Sum of errors -.56** -.58** -.13 -.14 Uses of Objects (n = 25): Correct responses .08 .46* .24 .46* .25 Sum of errors -.48* -.39 -.13 -.21 Stamps task (n = 25): Complexity score -.04 .36 .35 Originality score .05 .41* .25 .48* .36 Restriction score a a a Rule adherence score .13 -.01 .17 * p < .05; ** p < .01; *** p < .001. Note: Partial correlations controlled for age, VIQ and PIQ. All tests were two-tailed. a = No correlation could be calculated as one of the variables was constant
351
Table C2. Raw and partial correlations between ToM and EF variables within “old” control participants (aged 9-18 years) False belief task EF task Simple 1st-order 2nd-order ToL (n = 21): Adj.extra move score -.20 -.15 -.20 Rule violations .09 .13 .09 IDED Set-shifting task condition (n = 21): Perseveration EDS stage errors -.28 -.08 -.28 Learned Irrelevance EDS stage errors -.18 -.26 -.18 RIL task (n = 21): Error difference scores: Inhibition .29 .07 .29 Load -.32 -.05 -.32 Inhibition + load .10 .03 .10 RT difference scores: Inhibition -.16 -.26 -.16 Load .44* .54* .30 .44* .54* Inhibition + load .21 .01 .21 Shape error score -.31 -.20 -.31 Opposite Worlds (n = 21): Error diff. score .18 -.17 .18 Time diff. score .02 .02 .02 Relational Complexity (n = 21): Total score .43 .08 .43 Pattern Meanings (n = 21): Correct responses .20 .14 .20 Sum of errors -.08 .12 -.08 Uses of Objects (n = 21): Correct responses .10 .27 .10 Sum of errors -.02 -.17 -.02 Stamps task (n = 20): Complexity score .07 .10 .07 Originality score .59** .43 .26 .59** .43 Restriction score .05 .08 .05 Rule adherence score .06 .08 .06 * p < .05; ** p < .01; *** p < .001. Note: Partial correlations controlled for age, VIQ and PIQ. All tests were two-tailed.
352
APPENDIX D Separate group comparisons for young and old age subgroups on EF tasks (Study One)
Table D1. Group comparisons for “young” (5-8 years) and “old” (9-18 years) participants on inhibition, planning, and generativity tasks N Mean (SD)
Age subgroup ASD Control ASD Control t p Young Inhibition: participants Opposite Worlds: Error difference score 10 14 2.60 (2.59) 0.43 (1.87) 2.39 .03* Time difference score
10 14 15.67 (11.99)
7.01 (4.55)
2.48 .02*
Planning: ToL: Adjusted extra moves score
20 25 29.80 (8.17)
25.36 (7.19)
1.94 .06
Generativity: Uses of Objects: Correct responses 20 25 16.75 (8.28) 22.00 (9.51) 1.95 .06 Stamps task: Complexity score 20 25 18.25 (3.48) 20.12 (3.15) 1.89 .07 Originality score 20 25 2.50 (2.16)
3.96 (2.99)
1.83 .07
Old Inhibition: participants Opposite Worlds: Error difference score 19 22 0.89 (1.76) 0.86 (1.13) .07 .95 Time difference score
19 22 8.73 (6.09)
6.22 (4.03)
1.57 .12
Planning: ToL: Adjusted extra moves score
25 22 23.52 (6.33)
19.32 (6.24)
2.29 .03*
Generativity: Uses of Objects: Correct responses 26 23 20.85 (9.26) 31.22 (6.93) 4.39 .00*** Stamps task: Complexity score 21 21 19.00 (2.55) 20.71 (2.80) 2.08 .04* Originality score 21 21 3.81 (2.69) 5.76 (2.23) 2.56 .01* Note: Only continuous variables on which significant overall group differences were found are included. This table is intended to demonstrate that the EF components which are impaired in individuals with ASDs (in comparison with age-matched controls) change with development.
353