Kriminologiska institutionen
Personality type, prediction and
recidivist offending
An evaluation of extraversion and neuroticism in the
context of re-offending
Examensarbete 15 hp Kriminologi Kriminologi, kandidatkurs (30 hp) Höstterminen 2015 Sigrid Laiberg
Table of Contents 1. Introduction ............................................................................................................................................. 1
2. Literature review ...................................................................................................................................... 3
2.1. Theories on personality and crime ............................................................................................................... 3
2.2. Eysenck on personality and crime ................................................................................................................ 5
2.2.1. Extraversion ........................................................................................................................................... 5
2.2.2. Neuroticism ........................................................................................................................................... 7
2.3. The Eysenck Personality Inventory ............................................................................................................... 7
3. Previous research ..................................................................................................................................... 9
3.1. Eysenck’s personality test in the Cambridge study ...................................................................................... 9
3.2. Eysenck’s personality test in a study on recidivism .................................................................................... 10
3.3. Hypotheses ................................................................................................................................................. 11
4. Method .......................................................................................................................................................... 12
4.1. Population .................................................................................................................................................. 12
4.2. Significance testing .................................................................................................................................... 13
4.3. Reliability and validity ................................................................................................................................. 14
4.4. The EPI Questionnaire ................................................................................................................................ 16
4.5. Independent variable – results from the EPI .............................................................................................. 16
4.6. Dependent variable - recidivism ................................................................................................................ 17
5. Results ................................................................................................................................................... 18
5.1. Frequencies ................................................................................................................................................. 18
5.2. Dark number offending – official and self-reported recidivism .................................................................. 19
5.3. Official recidivism ....................................................................................................................................... 19
5.4. Self-reported recidivism ............................................................................................................................. 20
5.5. Correlation between the EPI and recidivism .............................................................................................. 20
5.5.1. Extraversion ......................................................................................................................................... 20
5.5.2. Neuroticism ......................................................................................................................................... 22
5.5.3. Neurotic Extraversion .......................................................................................................................... 23
6. Discussion .............................................................................................................................................. 25
6.1. Extraversion ................................................................................................................................................ 25
6.2. Neuroticism ................................................................................................................................................ 26
6.3. Neurotic Extraversion ................................................................................................................................. 27
6.4. Conduct disorder – an alternative theory ................................................................................................... 28
7. Conclusion.............................................................................................................................................. 29
8. Proposal for future research .................................................................................................................. 30
9. References ............................................................................................................................................. 31
10. Appendix .............................................................................................................................................. 32
Abstract
The purpose of this study was to determine whether the results from the Eysenck Personality
Inventory could significantly predict recidivism in the Cambridge Study in Delinquent
Development. The study is designed as an attempt at contributing to the general
criminological body of research, as well as an inquiry into one form of psychological
explanation for delinquency, namely personality theory. The use of personality testing is
based in psychologist Hans Eysenck’s theory on the relationship between personality type and
crime, and functions as an attempt at identifying and evaluating what individual risk factors
that predict criminal behaviour. From the population, two recidivist samples were selected;
officially registered as well as self-reported recidivists. The personality types tested were
extraversion, neuroticism and neurotic extraversion, as originally conceived by Eysenck. The
recidivist sample consisted of respondents in the Cambridge study who had received at least
one juvenile conviction between the ages of 10 and 16. Out of the three personality types,
extraversion could significantly predict both officially registered and self-reported recidivist
offending in the population. Neurotic extraversion predicted officially registered recidivism,
whilst no support was found for the claim that neuroticism could predict recidivism. The
study concludes that the results only partially support Eysenck’s claims and that a number of
factors influence personality type testing and predictions based therein. Thus, personality type
was not proven to be a significant risk factor for predicting crime and delinquency.
1
1. Introduction
When it comes to the causes of crime, there are a multitude of theories. Throughout the 20th
century, many of these explanation models have focused on the sociologically conditioned
origins of delinquent behaviour. Theories based in ideas of strain-, control- and labeling
aspects, to name a few, are some examples of sociologically characterized explanations of
crime that have had significant impact on criminological research in the past century (Kubrin,
Stucky & Krohn 2009: 50). Sociological approaches have also enjoyed great success in the
conduction of criminological research, and have resulted in many influential discoveries
within the discipline. However, owing to the scientific concentration on social environment,
sociologists have sometimes tended to disregard biological and psychological aspects of
behaviour. Whilst early biological notions of ‘the criminal man’1 have long since been
abandoned, researchers within modern psychology have suggested that cognitive,
developmental and personality-related facets can have explanatory value in understanding the
causes of delinquency and crime. Some behavioural psychologists have argued that criminal
conduct emerges through reinforcement and punishment, and cognitive scientists have held
that a person’s mental interpretation and method of information processing can affect
delinquent tendencies (ibid: 54-55). Moreover, some social psychologists have suggested that
an individual’s personality characteristics or traits can be used to predict criminal behaviour
in that same individual (ibid: 71-72).
Psychologist Hans Eysenck spent much of his career arguing for the link
between personality traits and crime. Specifically, he believed that people whose personalities
were characterised by high levels of extraverted and neurotic traits ran a higher risk of
antisocial behaviour2. Eysenck argued that extraversion is characterised by the individual
being more social and outgoing and prone to acting out and also to committing crimes.
Neuroticism, according to Eysenck, is a state of heightened sensitivity and emotionality,
which drives an individual to act in certain ways. He argues that extraversion inhibits
behavioural conditioning and the development of a ‘conscience’ whilst neuroticism helps
reinforce habitual behaviour, including delinquent tendencies. His personality test, the
Eysenck Personality Index, has been used to assess extraversion and neuroticism in different
populations in several studies, among them the English cohort study known as the Cambridge
Study in Delinquent Development (1961-1981). Due to its scope and detail, the Cambridge
study has been subject to vast amounts of criminological research since its initial conclusion
1 The early criminological idea that criminality is inherited; that a person is “born criminal”. 2 See section 2.2. for more information on the use of the term ‘antisocial behaviour’.
2
in 1981, and follow-up studies have also been conducted on its participants since then. In the
study, and in the criminological discipline overall, one area of interest has been the issue of
recidivism; in this paper defined as relapsing into criminal behaviour after an initial
delinquent act3. The extent of an individual’s recidivist offending is an expression of the
stretch of that individual’s criminal career - research has shown that juvenile convictions
predict persistent offending and long criminal careers (Murray & Farrington 2010: 635).
Recidivism is thus an important risk factor in predictive research, longitudinal explanations of
crime as well as in retroactively understanding criminal behaviour. The identification of
certain risk factors in predicting criminal behaviour is one way of assessing what methods can
be implemented to prevent individuals from committing criminal acts (Farrington 2005: 186).
In addition to increasing the body of criminological knowledge, predictive studies that
identify instruments for risk assessments with regards to recidivism can benefit the reduction
of crime and other social problems.
Results have been varied in determining what types of offending can be
predicted by extraverted and neurotic personality traits respectively. Some studies have found
support that both traits predict offending (Farrington 2005: 178-179). Other research has
indicated that extraversion was the more relevant trait in predicting offending (Van Dam,
Janssens & De Bruyn 2005: 8). Some argue that there is very little support for Eysenck’s
theory about personality and crime (Farrington, Biron & LeBlanc 1982: 196). Not much
research has yet been conducted on the relation between Eysenck’s personality constructs and
recidivism after an initial conviction (Van Dam et al: 8).This paper aspires to contribute to
theories involving psychological explanations of recidivism, in assessing if extraverted and
neurotic personality traits in an individual are of relevance for the risk of relapse into criminal
activity.
The study conducted and presented in this paper is a quantitative examination of
the correlation between the results from the Eysenck Personality Index and recidivist
offending in the Cambridge study. More specifically, the purpose of this paper is to determine
whether there is a statistically significant relationship between extraverted and neurotic
personality traits and the risk of recidivism in the Cambridge study: whether there is support
for the hypothesis that extraverted and neurotic personality traits in Eysenck’s test predict the
risk of relapsing into delinquent behaviour after an initial juvenile conviction. The question
formulations for this paper are:
3 See section 4.5 for further explication of this operationalisation of the concept of recidivism.
3
(1) Is there a statistically significant correlation between extraverted and neurotic personality
traits and official and self-reported recidivism in the Cambridge Study in Delinquent
Development?
(2a) Is there a difference between extraverted and neurotic traits respectively with regards to
recidivism?
(2b) Is there a difference between personality traits with regards to official and self-reported
recidivism respectively?
(3) Depending on the answer to (1), (2a) and (2b), what conclusions can be drawn with
regards to how personality type predicts recidivism?
The answer to question formulation (3) will serve as foundation for the concluding discussion.
If a correlation is found, it can be considered support for Eysenck’s hypothesis that these traits
predict offending, and furthermore, that they predict recidivism as well. If no correlation is
found, it corroborates the assumption that Eynseck’s theory is incorrect, or that the measure or
method is inadequate.
I shall give a background to trait-related explanations of crime, and personality
theory in relation to crime from Eysenck, with emphasis on extraverted and neurotic
personality traits. Following this I will present some results from personality type testing in
longitudinal studies. There will be a methodological discussion, and a review of the data
treatment, after which I will present the results from the significance testing. In conclusion, I
will bring a discussion on the results and their implications for the hypothesis, as well as make
some remarks on potential future research in the field.
2. Theory
2.1. Theories on personality and crime
As stated in the introduction, modern criminological discourse includes several theories
related to psychological explanations for criminal or otherwise antisocial behaviour. Common
to all psychological trait theories is the focus on developmental issues4, cognitive- and
intelligence related aspects as well as personality related explanations of antisocial behaviour
(Kubrin, Stucky & Krohn 2009: 54). Within the field itself, researchers often take different
approaches to psychological trait theory. Behaviour theorists have argued that antisocial
behaviour is not inherent through genetic make-up, but learned. Learning can occur through
imitation; the replication of the behaviour of others, and through conditioning processes,
4 Issues relating to the “development of human beings’ cognitive, emotional, intellectual, and social capabilities and functioning over the course of the life span” (In Encyclopædia Britannica 2015).
4
where antisocial behaviour is reinforced through rewarding (ibid: 54). Cognitive perspectives
on the causes of crime include theories that focus on how individual mental properties affect a
person’s construct and interpretation of themselves and the world around them. The cognitive
psychological perspective includes the notion that individual abilities relate to both
information processing as well as moral development. Some research has suggested
differences in moral development between criminals and non-criminals, as well as a lesser
ability among criminals to quickly and correctly process information. The theories then
assume that this in turn can lead to law-breaking behaviour (ibid: 54-56). Some psychological
theories on crime and deviance suggest that certain traits are inherited genetically – an
assumption that is often reviewed through studies on identical twins who share their genetic
make-up. Recent studies have found support for inherited traits, but conclude that
environmental factors also play a significant role in predicting behaviour (ibid: 62-63). Others
suggest that low arousal in a person makes them more sensation-seeking than someone with
higher arousal levels. The term ‘arousal’ refers to the activation of the functions of the
autonomic nervous system, indicating that a person is awake and attentive to outside stimuli
and perceptions (definition in Encyclopædia Britannica 2016). Low arousal levels indicate a
state where the subject requires stronger stimuli from the outside, to ‘hold their attention’, and
are less likely to feel excited about ordinary, non-sensational, experiences. On this view,
underaroused people are more likely to commit criminal acts in order to relieve feelings of
boredom. (ibid: 66-67).
The focus of this study is personality theory, where personality is thought of as
an individual’s “characteristic ways of thinking, feeling and behaving” (ibid: 72). Personality
theory draws from several of the above mentioned psychological approaches, and is denoted
by the assumption that personality consists of stable elements that influence a person’s
behaviour consistently across situations. That is to say, that personality does not naturally
change over time, and affects many types of behaviour, not just in regards to delinquency and
criminal acts (ibid: 72). In assessing personality, various types of inventories are commonly
employed, which are constructed as questionnaires where subjects are asked about
characteristics and behaviour. The assessment is based in the notion that certain subject
responses indicate high levels of certain major personality elements (ibid: 72). The Eysenck
Personality Inventory (EPI) measure reviewed in this paper is a form of assessment through
questionnaire, and is further explicated in the next section and in section 4.3. Some
personality theorists, such as Eysenck, incorporate aspects of different psychological
approaches to criminal behaviour. This is also further discussed in the following sections.
5
2.2. Eysenck on personality and crime
Eysenck spent much of his career arguing that criminal behavior can be explained as a
symptom of personality- and environmental factors in combination (Eysenck 1996: 143). He
holds that all behaviour, including individual reactions to social factors such as poverty and
inequality, is indisputably filtered through the individual’s psyche. Thus, he argues, inquiries
into psychological explanations for crime cannot be dismissed (ibid: 144). On Eysenck’s
view, personality is governed by three major and independent dimensions: psychoticism,
extraversion and neuroticism (ibid: 144). At the time the EPI was given to the Cambridge
population, psychoticism had not yet been added by Eysenck as an independent dimension.
Consequently, participants were only tested on extraversion and neuroticism (Farrington
1999: 256). For this reason, this paper only examines those two traits, as these were the two
traits assessed in the Cambridge study.
Eysenck argues that high levels of either extraversion or neuroticism or both
predict criminal behaviour (Eysenck 1996: 148). On his view, any behaviour is very much a
rational analysis, where acting in one’s interest and ‘taking what one wants’ is both rational
and perhaps natural, as both children and animals display this behaviour. ‘Socially desirable’
behaviour, such as not committing crimes, is dependent on certain social agreement on what
is right and what is wrong. People who behave in criminal ways, are equally aware of this
distinction, but break the law nonetheless (ibid: 148-9). In his theory of personality and crime,
Eysenck relates traits to antisocial behaviour. The term antisocial behaviour is largely related
to criminal behaviour in his explanations, and the two terms are often used in conjunction
(ibid: 147). Consequently, it is reasonable to assume that Eysenck views criminal acts as an
instance of antisocial behaviour. However, it is possible that some antisocial behaviour, for
instance substance misuse, which is not a specifically criminal act, is also included in
Eysenck’s theory. For the purpose of this paper, the term antisocial behaviour will be used to
reference criminal acts where they are discussed as relating to Eysenck’s theory. The reader
should take note that acts that are not necessarily criminal in a legal sense can also be
included in the concept.
2.2.1. Extraversion
On Eysenck’s conceptualisation, extraversion is the opposite of introversion (Eysenck 1969:
5-6). While Eysenck does not specifically define either term, he presents a clear idea of what
he considers ‘typical’ behavioural characteristics. In his behavioural account of the typical
extravert, Eysenck describes the subject as a person who likes parties, does not like reading or
6
studying by themselves, craves excitement, takes chances, is fond of practical jokes, is
carefree and easy-going, optimistic and tends to be aggressive and lose their temper quickly;
“his feelings are not kept under tight control, and he is not always a reliable person” (ibid: 8).
The typical introvert, on the other hand, is a person who is quiet, introspective, fond of books,
reserved, distrusts the impulse of the moment, likes a well-ordered mode of life and “keeps
his feelings under close control, seldom behaves in an aggressive manner, and does not lose
his temper easily” (ibid: 8).
Like some modern day psychologists, Eysenck argues that arousability
properties relate to antisocial behavior. His theory links low cortical5 -arousal and -
arousability in the brain to extraverted personality traits. High cortical arousal indicates a state
where the subject’s brain is awake and attentive to outside stimuli. This helps the subject
concentrate and fixate its attention on features of interest. Similarly, low cortical arousal
indicates lower levels of interest and attention in the subject. On Eysenck’s view, extraversion
is thus characterised by low cortical arousal because extraverted subjects require stronger
stimulation than introverted subjects. Consequently, the need for stronger stimuli makes
extraverted subjects sensation-seeking to a higher degree (Eysenck 1996: 148).
Eysenck also argues that criminal behavior can be explained, at least in part, in
terms of conditioning. How a person is conditioned influences what he conceptualises as
‘conscience’, which in turn governs behaviour. In his account of conditioning, Eysenck
explains conditioning in terms of punishment and reward. Different acts are either punished or
rewarded, forming the individual’s experience of their behaviour. Thus, conditioning is a
social practice, where behaviour is punished and rewarded by those around an individual;
parents, teachers and peers. Eysenck presents three possible reasons why different people
behave socially acceptable to varying degrees: The first is that the individual has not been
subjected to enough conditioning experiences. The second is that the ‘wrong’ experiences are
reinforced in the individual. The third, which is his chief concern, is that the conditioning
experience is relative to the subject’s psychological traits. He argues that low cortical
arousability in a subject, as reviewed in the previous paragraph, inhibits conditioning
processes. Consequently, subjects high on extraversion have greater problems aggregating
conditional experiences into a functioning ‘conscience’ than subjects low on extraversion.
Hence, extraverted subjects possess less internal resistance to antisocial behaviour. Regarding
the argument for extraversion and crime, results from experimental studies of conditioned
5 Referring to the outer layer of the brain.
7
responses have found evidence that antisocial subjects show lower conditioning compared
with non-antisocial subjects (Eysenck 1996: 149).
2.2.2 Neuroticism
As with extraversion and introversion, the corresponding opposite of neuroticism is called
‘stability’ in Eysenck’s terminology (Eysenck 1969: 5-6). Regarding neuroticism, Eysenck
finds the term almost self-explanatory. He takes inspiration from Carl Jung in describing it as
a state of emotionality or instability; an ‘anxiety state’ or a ‘reactive depression’,
characterised by a marked sensitivity to many situations, and a subjective experience of great
exhaustion. He also refers to neuroticism as a combination of correlated affective disorders
and a common symptom with subjects who display anxiousness, phobia, obsessive or
compulsive behaviour (Eysenck & Eysenck 1969: 226).
Eysenck argues that neuroticism is related to an inherited degree of lability of
the autonomic nervous system (Eysenck 1969: 7). High amounts of lability increase the risk
of emotionally excessive reactions to stress (Eysenck & Eysenck 1969: 50). Neuroticism
functions by reinforcing behaviour until certain action tendencies become habitual. It is
influenced by natural drive which makes subjects more persistent in their behaviour (Levine
& Jackson 2004: 136). High levels of neuroticism predict antisocial behaviour by virtue of
said drive properties, which can increase criminal action tendencies in the subject. Eysenck
also speculates that a relationship between neurotic traits and aggressiveness and
impulsiveness is an important contributor to the correlation with antisocial behaviour
(Eysenck 1996: 149-50). Interestingly, he also notes a difference between different
personality traits and reported offending, where high extraversion traits relate to offending in
younger samples, whilst high neuroticism traits relate to offending in older samples (ibid:
148).
2.3. The Eysenck Personality Inventory
The EPI was initially designed as a development of the earlier Maudsley Personality
Inventory, MPI. According to Eysenck, both the MPI and the latter EPI measure the two
major dimensions of personality that he, at the time of the Cambridge study, had identified;
extraversion and neuroticism. He considered the EPI an improved version of the MPI mainly
6 Eysenck uses the term ‘psychoasthenia’ - a psychological disorder characterized by phobias, obsessions, compulsions, or excessive anxiety (American Heritage Dictionary). The term is no longer in psychiatric diagnostic use.
8
because its items had been reworded to make them understandable to subjects of low
intelligence or education and it eliminated correlations between extraversion- and neuroticism
variables that had been present in the MPI results. Furthermore, the EPI was also proven to
have somewhat higher retest reliability than the MPI, even after significant time had passed.
This result agrees with the notion that personality is a stable element (Eysenck 1969: 5).
In his original manual of the EPI, Eysenck argues that most research has tended
to support the view that the neuroticism factor relates to lability of the autonomic nervous
system, whilst the extraversion factor relates to excitation and inhibition. Thus, he argues that
both personality traits are largely hereditary, with reference to studies on identical twins who
have been brought up separately but show very similar results when evaluated for both traits
(ibid: 7). Furthermore, he points out that narcotic substances can influence a subject’s position
on both scales, depending on its properties: Depressants affect the subject toward
extraversion, and stimulants toward introversion.
Eysenck also notes that most people fall somewhere on the spectrum between
both extremes. He does not completely disregard environmental influences, but makes a
distinction between genotypic7 and phenotypic8 aspects of personality. Behaviour, such as the
conduct listed in his description of typical characteristics, is a phenotypic aspect, and is the
expression of genotypic factors and environmental influences. Genotypic aspects refer to
constitutional traits such as conditioning, which are mostly inherited and are measured
experimentally and represented through personality qualities, as listed in figure 1. Observable
behaviour is thus a function of neurophysical and psychological aspects in interaction with the
subject’s environment (ibid: 8-10). Eysenck also conceptualises extraversion and neuroticism
as uncorrelated and independent dimensions of personality. This means that extraverts would
not be more likely to be neurotics than non-extraverts and vice versa (ibid: 12).
Whether Eysenck’s assumptions are correct is largely dependent on whether
they are supported in empirical studies. If no correlation is found between extraversion or
neuroticism and offending, that result suggests that Eysenck’s theory is flawed. He holds an
interactionist view, claiming that criminal behaviour is the result of personality and
environmental factors in conjunction. This position is not always reflected in his arguments,
which tend to be focused on the importance of psychological traits, disregarding factors such
as class, gender and ethnicity. The importance of environmental aspects has also been
supported in numerous studies, even among those that have suggested that some traits are
7 Genotype refers to something that is part of the genetic makeup which determines a specific characteristic. 8 From Greek: to show – a composite of an organism’s observable characteristics or traits.
9
inherited (Kubrin, Stucky & Krohn: 62-63). However, the EPI does not measure
environmental aspects at all. Aside from Eysenck’s accounts of how ‘typical’ extraverts and
introverts interact with others, not very much is said about the influence of an individual’s
social environment. Eysenck seems to hold that personality affects social behaviour rather
than vice versa and that crime is the result of how individuals with high scores of certain traits
have a tendency to breaks social norms. This means that the results of the EPI, as they are
tested in this paper, cannot really be thought of as evidence for or against Eysenck’s theory on
personality, environmental influence and crime, only for his claims that high scores of either
trait predict certain behaviour. Furthemore, the test itself could suffer from issues relating to
validity, the question of whether the EPI questionnaire really measures personality traits.
Within the modern psychological discourse, many have argued that personality is something
very complex and disputed the claim that personality is stable across situations. Contemporary
researchers often suggest that humans are very intricate in their behaviour and that personality
therefore is very difficult to measure (ibid: 73-74). The next section explores such and
argument regarding the validity of the EPI.
3. Previous research
3.1. Eysenck’s personality test in the Cambridge study
David Farrington, head researcher for the Cambridge study, holds that the results from the
EPI in the study show that those respondents high on both extraversion and neuroticism
tended to be adult official offenders (Farrington 2005: 178-9). However, in their extensive
review of the predictive capacity of the EPI in the Cambridge study and in a similar study
from Montreal, Farrington, Biron and LeBlanc found very little evidence that the EPI in itself
was an adequate measure for predicting official or self-reported offending, in juveniles or
adults. Their study showed that the individual items of the EPI lacked independency, and
significance was lost when the measure was held for other variables. In the Cambridge study,
no individual item on the scale was significantly related to either self-reported or official
offending. Results suggested instead that it was the trait of impulsiveness that predicted
offending (Farrington et al 1982: 183, 196). Consequently, Farrington et al concluded that the
EPI seemed to be an inadequate method of measuring personality dimensions, and for
predicting offending, mainly because of the non-independency of the scales (ibid: 196).
Farrington’s review can be considered a sort of criticism of Eysenck’s theory as
well as his method. The results suggest that correlations between the traits measured in the
EPI and offending is because certain measure impulsiveness, instead of personality. Even if
10
impulsiveness is a characteristic of extraversion, other facets of personality traits do not
explain much if significance is lost when held for impulsiveness. Impulsiveness is also often
listed as an important individual risk factor by many, separate from personality trait constructs
(Farrington 2010: 635). For the purpose of this study, it is worth noting that Farrington et al’s
review focused on juvenile and adult official, as well as self-reported offending as dependent
variables, but excluded recidivism.
3.2. Eysenck’s personality test in a study on recidivism
In their study of the personality-recidivism relationship, van Dam, Janssens and De Bruyn
used Eysenck’s newer model, PEN, which includes the Psycopathy scale, as well as the Big 5
personality test, which also measures extraversion and neuroticism. Both studies’ respective
measures of extraversion and neuroticism have shown high resemblance (Van Dam et al
2003: 12). Their purpose was to determine which model was better at distinguishing between
students and offenders, as well as recidivists and non-recidivists. Neither model had at that
time been prominently used in studies on recidivism in particular. Van Dam et al argue that
previous results on the predictive qualities of Eysenck’s model have been varied, but show
that some studies have found significant correlation between different traits and offending.
Regarding recidivism, one study conducted by Eysenck and Eysenck, indicated that
neuroticism showed little relevance, but that extraversion appeared to be the better predictor
(ibid: 8).
Van Dam et al conducted their study on two samples. The offender sample
consisted of 96 male adolescent offenders aged 13-25, with a mean age of 18.69. 36% of the
subjects in the sample were incarcerated during the assessment, and 64% were assessed after
their release. The college student sample consisted of 204 male adolescents, aged 15-24 with
a mean age of 17.23. The samples were selected to be similar in regards to education level
(ibid: 10-11). Both populations were assessed with the Eysenck Personality Questionnaire
(EPQ) measuring the PEN, as well as on the Big 5. Recidivism was measured through a self-
report list asking respondents if they had committed one or more of 20 criminal acts listed
since they were released, as well as official criminal records with a follow-up period of 7-52
months (ibid: 11-12). For the purpose of this paper, the results from the Big 5 will not be
exhaustively elaborated, since this is a review of Eysenck’s theory. Some notes on the results
from the Big 5 will however be made due to the previously mentioned similarities between
extraversion and neuroticism in both tests.
The results of the testing show that both models were able to differentiate
11
between students and offenders. Interestingly, as opposed to Eysenck’s theory, students
scored higher on PEN extraversion than offenders, but not on the Big 5 test (ibid: 13). The
authors argue that this result can in part be explained by incarceration-effects; where
offenders who were in custody at the time of the personality assessment had difficulty
properly answering questions about social activity, since their normal social behaviour was
restrained whilst incarcerated. Van Dam et al compared results from incarcerated and non-
incarcerated offenders and found significant differences in their answers regarding
extraversion items (ibid: 16). No significant differences between students and offenders were
found with regards to neuroticism (ibid: 13).
A univariate analysis showed that recidivists scored significantly higher on PEN
extraversion than non-recidivists, in line with Eysenck’s view. Thus, Eysenck’s extraversion
trait predicted recidivism in official criminal records. Similarly to Farrington, van Dam et al
argue that this result can be due to extraverts being more impulsive, sociable and adventurous.
This would also cause them to run a higher risk of attracting the attention of the police or
other authorities, increasing the risk of getting caught (ibid: 16).
Neither extraversion nor neuroticism on the PEN model showed significance
with self-reported recidivism in a univariate analysis. Self-reported recidivists did however
score higher on neuroticism in the Big 5 test (ibid: 13-14, table 3). Van Dam et al conclude
their discussion by remarking on the importance of inquiry in to the reasons why there is such
a difference between personality traits and self-reported and official recidivism respectively.
They also point out that their results indicate that different personality traits are significant for
predicting the two different kinds of recidivism, and that only the PEN extraversion trait could
be considered having a relevant correlation with officially recorded recidivism (ibid: 17).
3.3. Hypotheses
The evidence in support for Eysenck’s theory of personality and crime seems to be
inconclusive at best. Farrington et al hold that the EPI as a measure is inadequate and that any
correlation is due to the test picking up on other risk factors. Because Farrington et al’s review
of the EPI suggests that it is not a good predictor for offending it would be unreasonable to
assume that it is a good predictor for recidivism: In order to distinguish between offenders and
re-offenders, the measure should be able to distinguish between offenders and non-offenders
first, since offending is a prerequisite for recidivism in this study. However, van Dam et al’s
results suggest that the extraversion trait is relevant for officially registered recidivism in
particular, independently of offending. This is owing to the fact that students (non-offenders)
12
scored higher on extraversion than offenders, whilst officially registered recidivists scored
higher than non-recidivists. Consequently, their results suggest that whilst the extraversion
trait could not predict officially registered offending, it successfully predicted officially
registered recidivism. As mentioned in section 3.2., a few different explanations can be given
for this result; such as incarceration effects and an increased risk of detection for extraverts.
In accordance with Eysenck’s theory, this study will also test respondents high
on both extraversion and neuroticism. Following the Cambridge study terminology this
classification is hereinafter referred to as neurotic extraversion (Farrington 1999: 261).
Assuming that van Dam’s results with regards to recidivism can be reproduced, the results
expected in this study are:
- Extraversion significantly predicts officially registered recidivism.
- No significant correlation between extraversion and self-reported recidivism.
- No significant correlation between neuroticism and officially registered recidivism.
- No significant correlation between neuroticism and self-reported recidivism.
- Neurotic extraversion significantly predicts officially registered recidivism.
- No significant correlation between neurotic extraversion and self-reported recidivism.
4. Method
4.1. Population
Through the use of significance testing in SPSS, this study aims to establish whether there is a
correlation between personality type and recidivism in the Cambridge study. Thus, it concerns
an age cohort of a total population of 411working-class boys, born in 1953, living in London
at the launch of the study in 1961. Because one of the chief purposes of the Cambridge study
was to research delinquency, a working class population was selected, as researchers wanted
to find high enough frequencies of delinquency for comparative purposes (ibid: section I).
Therefore, the population does not constitute a simple random sample. The data was collected
starting when the members of the population were 8, and then again at 10 and 14, in their
schools. At 16, 18, 21 and 24 they were interviewed in a research office or asked to supply
answers via post. Some non-response occurred during the duration of the study, do to death,
relocation and refusal to partake (ibid: section I). Missing values due to non-response will be
disregarded and excluded in the significance testing in this study (Bjurfeldt & Barmark 2009
27-28). Some remarks on the potential effects of this non-response in the context of this study
will be made in the concluding discussion. No new data collection will be made, so it is
13
entirely reliant on secondary data. The population has previously been reviewed by many
researchers, including Farrington et al in relation to the EPI, as discussed in section 3.1.
The fact that the population is not a random sample can be considered both
positive and negative for this study. Because most boys came from working class
backgrounds any results obtained here can be considered less suitable for generalising across
larger populations. The same is true for gender as only boys were studied, making it very
difficult to say anything about what the results would have been in an all-female or mixed
population. However, the theory this study evaluates assumes that personality is mostly
independent from factors relating to gender and social class. Eysenck’s theory of personality
thus claims to be generalisable to a certain extent. This means that results obtained through
this study of a quite homogenous group, could be compared to a different group in order to
evaluate the evidence for or against Eysenck’s position that personality is genetic and
hereditary, and not chiefly a social construct. For example, a similar study could be conducted
on a sample of 400 girls, and results could then be compared. The issue with such a
comparison, however, would be that the frequency of delinquency would likely be much
lower in an all-female sample. Because the Cambridge study population was chosen with the
hope of finding significant levels of delinquency, a study such as this on recidivism is also
possible. Had there not been any delinquency in the population, there would of course not
have been any recidivism (as it is defined in this study).
4.2. Significance testing
Significance will be measured using crosstables and Chi Square-testing, comparing observed
frequencies with expected frequencies. The confidence interval is set to 95% for this study,
meaning that a significant result is a result where there is a 95% chance that the correlation
reflects a relation between personality and recidivism, rather than a sampling error. A null
hypothesis is the position that there is no relationship between the independent and dependent
variables measured; that is to say, that there is no difference between scores of traits and
recidivism in this study. A confidence interval of 95% means that the risk of rejecting a true
null-hypothesis is 5%. Even if a significance is calculated, and the hypothesis is accepted, the
risk that the null hypothesis is true (no correlation) is still 5%, meaning that no result is 100%
certain. For any Chi Square-test in this study, p-values of .05 and lower will indicate
significance at the set confidence level of 95% (Djurfeldt, Larsson & Stjärnhagen 2010:116-
117, 188).
The study will also use odds ratios for significant results. The odds ratio
14
expresses the association between two factors in a population. The factors in the population
assessed in this study are (1) a score of a certain personality trait and (2) committing an act
defined in this study as a recidivist act. Odds ratios describe the relationship between different
probabilities. They are calculated by taking the probability of something occurring and
dividing it by the probability of the same thing not occurring (Ribe 1999: 15). In this study,
that the odds ratio reflects the relationship between the probability of committing an act of
recidivism and not committing such an act, for the personality traits tested. Odds ratio values
larger than 1 indicate an increased risk of something occurring. The higher the odds ratio
value the greater the risk of that occurrence (Ribe 1999: 14-15). In this study an odds ratio
higher than 1 describes the over-risk of recidivism (as defined) entailed by a certain score of
the personality traits measured. The hypotheses stated in section 3.3 assume that the odds
ratio value will be higher for extraversion and neurotic extraversion. If a significant
correlation is found as described in the previous paragraph, the hypotheses assume that the
odds ratio will be higher than 1 for those traits, meaning that the risk of committing recidivist
acts is higher for respondents with high scores.
4.3. Reliability and validity
Reliability refers to the extent to which the result found in this study can be reproduced. The
success of such a reproduction is likely largely dependent on the method used. If another
researcher would use the same population, data, variables, significance- and odds ratio
measures, the results would be identical to those found here (Bryman 2011: 49). This is
because such treatment of data is computed by algorithms that are preprogrammed in the
analysis system - in this study, SPSS was used. The method as such does not involve personal
or subjective interpretation in significance testing indicating that the reliability of this study is
quite high in that sense. However, in the interpretation of what the results actually reflect
about reality, there could be certain differences between researchers in how they view the
strength of the support for the hypotheses. In order to show transparency in the interpretation,
I shall try to view the implications of each result not just in the context of the hypotheses, but
also reflect on what each variable measured actually expresses. With quantitative research,
there is a certain risk of jumping to conclusions when a covariance is established. Unlike in an
interview setting, secondary data does not allow for asking open questions or following up on
answers, which means that the question of what variables actually measure is especially
important. In the concluding discussion, I shall try to review the results critically, as well as
compare them to the results found by van Dam et al, who studied a different population but
15
used a quantitative method and a somewhat similar approach.
Validity is an expression of whether this study actually measures what it sets out
to measure. As previously mentioned, the validity of Eysenck’s personality test has been
questioned. The first main criticisms against the validity of the EPI is that the personality
construct is incorrect and does not adequately measure personality type, even if there could be
a correlation between some personality construct and certain behaviour (Kubrin, Stucky &
Krohn: 62-63). The second is that the measure picks up on other relevant factors, such as
impulsiveness, rendering a sort of false positive correlation between personality and criminal
behaviour (Farrington et al 1982: 196). Whilst the purpose of this study is not to evaluate the
validity of either of those positions, they will be kept in mind in the discussion of the results
found. Furthermore, as this study in large part is an evaluation of whether the results support
Eysenck’s theory, the problem of personality can be avoided to a certain extent. The results
will either support or not support Eysenck’s theory – this does not mean that they say much
about the actual empirical correlation between personality and crime.
Of course, there is also always the chance that personality is of no importance
when it comes to recidivism at all. This notion could be supported if no correlations are
found. However, such an alternative theory of the causes of crime is not in itself an argument
against the validity of this study; the validity is dependent on if the EPI measures personality
and the recidivism variables measure recidivism. Regarding recidivism, the variable
measuring officially registered reoffending is likely quite valid as it is derived from official
records of convictions. Official convictions do not, of course, reflect all of the actual
offending, and the difference between the two is often referred to as the dark number. For
illustrative purposes and an approximation of dark numbers, a short summary of the
difference between officially registered and self-reported juvenile delinquency is presented in
section 5.2. Regarding self-reported recidivism, there is always the risk that respondents have
answered the questionnaire falsely, which makes that variable slightly more problematic.
Furthermore those who are excluded from testing due to non-response can constitute a sample
with traits or behaviours that would have had bearing on the results, for example respondents
who died young or who were incarcerated. However, the samples remain large enough that at
least some conclusions can be drawn from it (Bryman 2011:51). In addition to this, the
Cambridge study also specifies that respondents who were incarcerated at the different times
of the self-report index data collection were sent surveys via post (Farrington 1999: section I).
Van Dam et al also bring a discussion on incarceration effects on personality type in regards
to the validity of their study. This argument is reflected upon throughout this study as well
16
(van Dam et al 2005: 16).
4.4. The EPI questionnaire
In order to give the reader an overview of how the EPI assessment was actually conducted,
the questionnaire given to the Cambridge study population is presented in the appendix9. The
letter before the number indicates what trait the question assessed: ‘E’ – extraversion and ‘N’
- neuroticism. The Lie scale questions originally included are not presented, as lying
tendencies were not of interest in this study. A ‘-‘ sign before a letter means that that question
measures a low level of the trait, ‘-E’ expresses negative extraversion, i.e. introversion. Each
question was answered either with either ‘yes’ or ‘no’.
4.5. Independent variable – results from the EPI
The sample were assessed according to the EPI at age 16. The results of each scale;
extraversion and neuroticism respectively, are presented both as raw data as well as divided in
to quartiles in the Cambridge study. In this paper, the Cambridge quartile division has been
used to dichotomise both personality trait variables, so that each individual in the sample is
classified either as displaying ‘low extraversion’ or ‘high extraversion’ as well as ‘low
neuroticism’ or ‘high neuroticism’. The two quartiles representing the lower range of
extraversion scores in the Cambridge division constitute the ‘low extraversion’ classification
in this study, whilst the two higher scoring groups have been merged into the ‘high
extraversion’ classification. The same method vas used for dichotomising responses on the
neuroticism scale. An identical approach is also implemented by the Cambridge study
researchers for EPI variables (Farrington 1999: 261). Consequently, the use of this
dichotomisation of the independent variables in this study appears reasonable. Furthermore, in
Eysenck’s terminology ‘low extraversion’ is synonymous and interchangeable with
introversion and low neuroticism is interchangeable with stability (Eysenck 1969: 6). For this
reason, it seems favourable to categorise those with scores on the higher half of the scale as
more extraverted and those with lower scores as less extraverted, bearing in mind that it is a
scale and not a dichotomy. The terms ‘introversion’ and ‘stability’ are also used for lower
scores of extraversion and neuroticism in the Cambridge study. For the purpose of not
confusing the reader, this study will exclude the introversion- and stability terms from the
significance testing, on behalf of ‘low extraversion’ and ‘low neuroticism’.
9 See Table 1 in Appendix.
17
4.6. Dependent variable – recidivism
For the purpose of operationalisation within the constraints of this study, and with regard to
the vast scope of information in the Cambridge data, a few restrictions were made. At the time
of the study, the English system convicted 16-year olds as juveniles, and 17-year olds as
adults (Farrington 1999: 255). In this study, an initial juvenile conviction is as a prerequisite
for recidivism, which also corresponds well with how this variable was operationalised in the
study by van Dam et al (van Dam et al 2005: 8). Their variable expresses officially registered
recidivism and self-reported recidivism respectively after a period of incarceration. In this
study, those with juvenile convictions are considered potential recidivists, directly excluding
those who never received a conviction as juveniles, even with high scores of self-reported
delinquency. This entails that the dependent variable in this study is based on a legal
definition of recidivism. The legal definition has been chosen with the aim to distinguish
between those convicted as juveniles who reoffend, officially and/or self-reportedly, and those
who merely admit one act as juveniles plus one act as adults through self-reports. In fact, only
20.4% of respondents had a juvenile conviction, but 97.3% reported delinquent acts as
juveniles. Nearly 74% of the total population self-reported both juvenile delinquency and
criminal acts as adults. Consequently, a recidivism variable constructed from only self-report
would comprise a majority of the total population, including any respondents who have
admitted to any kind of delinquency between ages 10-16 and any form of violation between
ages 17-24. The objective of the juvenile conviction prerequisite is to avoid the risk of
including too large a sample of respondents who have merely acted mischievously or
prankishly for the 15 year duration. As this study, to some extent, is an inquiry into
psychological circumstances surrounding re-offending, such a reservation seemed favourable.
Two measures will be taken for recidivism. Officially registered recidivism
refers to respondents registered for an offense, at least one as juveniles and at least one as
adults. Considering the English penal system at the time, this means those respondents who
had been officially registered for an offense; once before their 17th birthday and once after
their 17th birthday but before their 25th birthday, giving a follow-up period of eight years
(Farrington 1999: section XVII, var. 9-10). Self-reported recidivism refers to respondents
registered for at least one offense as juveniles who have self-reported at least one criminal act
as adults. The measure was constructed by adding population responses regarding a range of
criminal acts in the data, to create an index that was later dichotomised. A similar method is
used in the Cambridge study to represent delinquency in the 18-19-year old age group, this is
not dichotomised but classified to three ‘delinquency levels’ although all boys that were
18
classified as ‘low’ in the Cambridge study denied any act of delinquent behaviour (ibid: 356).
Following van Dam et al, those who report a minimum of one criminal act in the follow-up
period are considered recidivists (van Dam et al 2005: 11). An index was made for age groups
18-19, 21-22 and 24-25. At 20-21 and at 24-25 the respondents were asked to report acts
going back two years (Farrington 1999: 373, 392). The index was made up out of a total of
eight variables on a self-report card: (1) stealing from work, (2) driving a vehicle without
permission, (3) obtaining money from the government by telling lies, (4) breaking and
entering to steal money or things, (5) stealing from parked vehicles, (6) damaging property
without stealing anything, (7) stealing from slot machines and (8) shoplifting (ibid: 373-6,
392-5). The same types of criminal acts were measured at 21-22 and at 24-25. At 18-19 they
were asked to report three years back (ibid: 355). To eliminate those who reported juvenile
acts from the adult index, responses at 18-19 were controlled for responses at 14-17. 74.5% (n
306) reported committing at least one criminal act between the ages 18-2510. 20.7% (n 85)
denied having committed any criminal act and 4.9% (n 20) never answered the questionnaire.
5. Results
5.1. Frequencies
The initial testing showed the following frequencies of respondents where percentages are in
relation to the total population:
Table 1. Frequency of personality traits and recidivism
N %
Recidivism
Official 55 13.9
Self-reported 76 18.5
Extraversion
Low 185 46.5
High 213 53.5
Neuroticism
Low 191 48.0
High 207 52.0
Neurotic extraversion
Yes 118 29.6
No 280 70.4
10 Table 4 in Appendix.
19
5.2. Dark number offending – official and self-reported juvenile delinquency
Van Dam et al argue that one of the main problems with studies of this kind is the
methodological issue surrounding dark number offending (van Dam et al: 9). Especially
where recidivism is concerned, as those who are convicted at one point are not necessarily
convicted again even if they engage in substantial recidivist behaviour. In order to determine
the extent of dark number offending, it can be useful to compare self-reported delinquency
with officially registered delinquency. In the Cambridge study, a total of 20.4% (n 84) had at
least one juvenile conviction between 10-1611. 18% (n 74) had at least one juvenile conviction
between ages 14-16. 26 were convicted of at least one offense the year of their 17th birthday,
making the total percentage of official offending 23.1% (n 95) for ages 14-17. The official
records were then compared to self-reported delinquency between ages 14-17. Frequencies
are shown in table 2 below.
Table 2. Approximation of dark number offending 14-17
N %
Official offending
14-17 95 23.1
Self-reported offending (not convicted)
Low 379 92.2
High 32 7.8
The results show that 7.8% (n 32) of respondents who were not convicted between the ages
14-17 reported high scores of delinquency, 21 or more acts admitted (ibid: 255-6)12.
Compared to the 23.1% with convictions, the dark number can be estimated to equal slightly
more than one third of the official offending numbers in respondents between ages 14-17,
going by self-reported acts.
5.3. Official recidivism
In order to classify the official recidivists group, those with juvenile convictions were cross-
compared to those with adult convictions The table below shows the frequency of respondents
11 Table 2 in Appendix 12 Table 3 in Appendix.
20
who had been officially registered for an offense, once before their 17th birthday and once
after their 17th birthday but before their 25th birthday.
Table 3. Officially registered recidivism
Convicted as juvenile 10-16
No conviction 1 or more convictions Total
Convicted as adult 17-24
No conviction 265 23 288
1 or more convictions 52 55 107
Total 317 78 395
The 55 individuals in the official recidivist group make up 13.4% of the total population, and
had received both juvenile and adult convictions. This is the sample called official recidivists
in this study.
5.4. Self-reported recidivism
The same method of cross-comparison was used for the self-report variable. The table below
shows the frequency of self-reported recidivists; those who had a juvenile conviction and who
self-reported at least one criminal act after their 17th birthday.
Table 4. Self-reported recidivism
Convicted as juvenile 10-16
No conviction 1 or more convictions Total
Adult criminal act
1 or more admitted 230 76 306
None admitted 79 6 85
Total 309 82 391
The results show that 18.5% (n 76) of the total population were both convicted as juveniles
and reported criminal acts in adulthood. Those 76 individuals are the self-reported recidivists
in this study.
5.5. Correlation between the EPI and recidivism
5.5.1. Extraversion
The results from the test of extraversion and officially registered recidivism showed that the
trait could significantly predict official recidivism13. Among the 54 recidivists assessed, 37
had had high scores of extraversion when evaluated at age 16, compared to only 17 with low
13 Chi-Square and odds ratio tests – table 5 in Appendix.
21
scores. The corresponding numbers for non-recidivists were 172 with high scores of
extraversion and 159 with low scores. This makes those high on extraversion more than twice
as frequent among officially registered recidivists, than those low on the trait. This result is
also reflected by the odds ratio of 2.012, stating that the risk of being a recidivist is
approximately twice as high for extraverts compared to non-extraverts. Among non-
recidivists, the difference between high and low scores was not as significant.
Table 5. Correlation between extraversion and official recidivism
EPI Extraversion
Official recidivism Low E High E Total
Non-recidivist 159 172 331
90,3% 82,3% 86,0%
Recidivist 17 37 54
9,7% 17,7% 14,0%
Total 176 209 385
100,0% 100,0% 100,0%
p < .05 Exp(B) 2.012
This result also supports Eysenck’s theory and the hypothesis on extraversion and official
recidivism in section 3.4. It also reflects the findings by van Dam et al, where high
extraversion in the PEN model could predict official recidivism in their sample (van Dam et
al: 16). The results do not say anything about whether the significance is due to the
extraversion measure picking up on some other relevant trait or aspect, but confirms
Farrington’s observation that those high on the trait tended to be official offenders (Farrington
2005: 178-9).
Regarding self-reported recidivism, the hypothesis states that no statistically
significant correlation will be found. However, among the 75 self-reported recidivists who
were assessed in this particular test, 48 were high on extraversion and only 27 were low.
Because of this ratio, significance should be expected at the confidence level set here.
Table 6. Correlation between extraversion and self-reported recidivism
EPI Extraversion
Self-reported recidivism Low E High E Total
Non-recidivist 153 160 313
85,0% 76,9% 80,7%
Recidivist 27 48 75
15,0% 23,1% 19,3%
Total 180 208 388
22
100,0% 100,0% 100,0%
p < .05 Exp(B) 1.700
A significant correlation was found and the hypothesis was not confirmed14. The results
support Eysenck’s theory that those high on extraversion are more likely to offend. However,
self-reports of criminal acts are not necessarily the same as offending. In contrast to the
results found in van Dam et al’s study, in which those high on the trait were not more likely to
self-report recidivism, this study found a correlation with the trait (ibid: 13).
5.5.2. Neuroticism
Concerning neuroticism, differences between those high and low on the trait, were very small
with regards to official recidivism. Although a higher percentage of recidivists reported high
scores of neurotic traits, no correlation was found at the significance level, confirming the
hypothesis formulated in section 3.4.15
Table 7. Correlation between neuroticism and official recidivism
EPI Neuroticism
Official recidivism Low N High N Total
Non-recidivist 160 171 331
87,4% 84,7% 86,0%
Recidivist 23 31 54
12,6% 15,3% 14,0%
Total 183 202 385
100,0% 100,0% 100,0%
p > .05
The results contest Eysenck’s assumptions regarding the effects of neurotic traits on criminal
behaviour, suggesting that neurotics are not more habitual than non-neurotics in regards to
offending and detection by authorities. Again, the study mirrors van Dam et al’s, who did not
find significance for high scores of neurotic traits on the PEN, although they found a
correlation with neuroticism in the Big 5 test (ibid: 13).
When neuroticism was tested for self-reported recidivism, the results were
similar to those found with official data. Of the 75 self-reported recidivists tested, 43 were
high on neuroticism and 32 were low on the trait. Among non-recidivists, the results were
almost 50/50 with regards to neuroticism: 158 had high scores of the trait and 155 had low
14 Chi-Square and odds ratio tests – table 6 in Appendix. 15 Chi-Square tests - table 7 in Appendix.
23
scores. Consequently, the hypothesis that there is no association between neuroticism and
self-reported offending seems likely. The results are shown below.
Table 8. Correlation between neuroticism and self-reported recidivism
EPI Neuroticism
Self-reported recidivism Low N High N Total
Non-recidivist 155 158 313
82,9% 78,6% 80,7%
Recidivist 32 43 75
17,1% 21,4% 19,3%
Total 187 201 388
100,0% 100,0% 100,0%
p > .05
The hypothesis was also confirmed through the significance test, which showed no
correlation16. Again, the results provide no support for Eysenck’s theory, but correspond with
those found by van Dam (ibid: 13).
5.5.3. Neurotic extraversion
Respondents who had scores in the higher half of both the extraversion scale and the
neuroticism scale were classified as neurotic extraverts. In a survey of the percentages, those
high on both traits make up about 30% of the total of respondents, 114 of 385 individuals
tested. 23 of the 54 who were officially registered as recidivists had high scores on both traits,
whilst 31 did not. This makes neurotic extraverts in excess of 40% of the official recidivist
sample, whilst they make up less than 30% of the non-recidivist sample. Consequently, a
statistically significant result seems plausible, which would confirm the hypothesis in section
3.4. When tested, significance was found with a p-value of less than .05 and an odds ratio of
1.95717.
Table 9. Correlation between neurotic extraversion and official recidivism
EPI Neurotic Extraversion
Official recidivism Low on E+N High on E+N Total
Non-recidivist 240 91 331
88,6% 79,8% 86,0%
Recidivist 31 23 54
11,4% 20,2% 14,0%
16 Chi-square tests - Table 8 in Appendix. 17 Chi-square test and odds ratio – table 9 in Appendix.
24
Total 271 114 385
100,0% 100,0% 100,0%
p < .05 Exp(B) 1.957
The results support Eysenck’s notion that those high on both traits run a higher risk of
becoming delinquent than those high on only one (Eysenck 1996: 148). Interestingly, the odds
ratio is about the same as for extraversion, indicating an over-risk of about 2 for those high on
both traits. However, the odds ratio is slightly lower for neurotic extraverts than for all
extraverts, implying that high levels of neuroticism decrease the risk of official recidivism for
extraverts. This combination of traits was not tested by van Dam et al, but it is conceivable
that they had gotten a similar outcome in their sample, since the studies have corresponded
well in regards to the first four hypotheses.
When self-reported recidivism was controlled, only 28 of the 110 neurotic
extraverts with juvenile convictions tested had self-reported criminal acts.
Table 10. Correlation between neurotic extraversion and self-reported recidivism
EPI Neurotic Extraversion
Self-reported recidivism Low on E+N High on E+N Total
Non-recidivist 228 85 313
82,9% 75,2% 80,7%
Recidivist 47 28 75
17,1% 24,8% 19,3%
Total 275 113 388
100,0% 100,0% 100,0%
p > .05
This number is a lot smaller than the corresponding number for those who self-reported in
both the extraverted and the neurotic sample. In both of those groups more than 40 individuals
self-reported recidivism. Consequently, no significant correlation was found with self-
reported recidivism in this instance either18.
In conclusion, officially registered recidivism could be predicted by high levels
of both extraverted traits and neurotic extraverted traits, but not by neurotic traits. A
significant correlation was found with self-reported recidivism for extraversion, but not for
the other two traits tested. The odds ratio showed that high scores of extraversion turned out
to involve a slightly higher over-risk for official recidivism than extraverted neurotic traits,
although the respective risks were almost equal and at about twice as high as for those without
high scores on the traits. For self-reported recidivism the over-risk was not as high for
18 Chi-Square tests – Table 10 in Appendix.
25
extraverts, although prevalent. The results confirmed five of six hypotheses in section 3.3 and
largely correspond with the study by van Dam et al. Some results support Eysenck’s theory
whilst others contest it. However, none of the correlations found conclusively prove a relation
between personality type and recidivism in this study. Eysenck’s theory is mainly supported
by the significance found for extraverted traits. With regards to neuroticism, none of the
findings indicate a correlation between the trait and recidivism, despite the fact that neurotic
extraversion predicted officially registered recidivism.
6. Discussion
6.1. Extraversion
In regards to the correlation found between extraversion and recidivism, a number of
reasonable assumptions can be made. One possible reason for the correlation is that there is an
actual relation between certain personality types and certain forms of recidivist offending, and
that the EPI adequately measures this relation. It could also be because the extraversion
measure includes other psychological or personality-related traits, which in turn predict some
types of offending. This study does not evaluate that possibility, and if Farrington and his
colleagues are right in their assumption, impulsiveness could be a key factor in the findings of
this study and the one conducted by van Dam et al (Farrington et al 1982: 196). Furthermore,
a range of methodological issues and issues relating to subject responses and data collection
procedures influence the results found here. As previously discussed, it is also conceivable
that the relevance of extraversion for officially registered recidivism is influenced by
extraverts’ higher risk of detection by authorities, due to characteristics related to extraversion
drawing more attention in general (van Dam et al: 16). This is also reflected by the fact that
extraverts had a higher over-risk of being convicted than they had of self-reporting recidivism
in this study. If true, extraverts are not necessarily more antisocial, only more likely to have
their delinquent acts spotted. In order to determine whether this is the case, a ‘detection risk’
measure could be used in combination with the personality index. If support is found for a
detection effect, this would be a problem for Eysenck’s theory, which postulates that
personality relates to antisocial behaviour and not the risk for detection.
Interestingly, extraversion was found to be significant in both studies, despite
the fact that large parts of van Dam et al’s sample were incarcerated at the time of personality
assessment, and thus would perhaps respond differently to extraversion measures due to
incarceration effects. It is not known how many of the Cambridge study respondents were
incarcerated at the time of the EPI but it is well recognised that most of them were not. Unlike
26
van Dam’s sample, the population was not selected because they had a previous conviction, or
had spent time incarcerated. This would indicate that the Cambridge population would have
more responses indicating high extraversion, due to the population not dealing with
incarceration effects, unlike in van Dam’s study. However, when the Cambridge population
self-reported criminal acts as adults, those incarcerated at the time were asked to supply their
answers via post. This could have affected the non-response. If those who were incarcerated
were more likely to not respond, this would affect the non-response rates of the sample
assessed in this study. More specifically, such non-response would leave out some of those
that would have been classified as recidivists. Despite this, the study conducted here found
that extraversion predicts self-reported offending, whilst this result was not apparent in van
Dam’s study, perhaps due to incarceration effects on responses about social attitude (ibid: 16).
Because of the similarities between results found in this study and van Dam’s, it
can be argued that similar results could have been found in this study if many of the
Cambridge study respondents had been incarcerated as well. However, the fact that van
Dam’s study assesses the newer PEN model could also be important for determining
comparability between results. Furthermore, it is conceivable that the PEN measure could
have predicted recidivism in the Cambridge study as well, due to its similarity with the EPI
tested here. The significance of extraversion can thus be seen as an indication of similarity
between the EPI and the later PEN model. This is however not particularly surprising as the
models were evaluated to show high resemblance (ibid: 12).
6.2. Neuroticism
Because no significant correlation was found with neurotic traits at all, the results correspond
with those found by van Dam et al in that they show that neurotic personality traits appear to
have little importance with regards to recidivism (ibid: 13). This result can be viewed as an
argument against Eysenck’s notion that neurotic traits function by way of reinforcing certain
behaviours, even if those have been previously punished (Levine & Jackson 2004: 136).
Furthemore, neurotic extraverts were in fact a little less likely than those with only high
scores of extraversion to be official recidivists, indicating negative effects of neuroticism on
recidivism.
It is however worth noting that not all antisocial behaviour is expressed through
the recidivism variables used in this study. As reviewed in section 5.3, the self-report index
consists mainly of theft- and damage-related violations. Following Eysenck’s symptom
description from Jung, neuroticism as a term appears to be referencing states of depression
27
and stress (Eysenck & Eysenck 1969: 22). For that reason, it is reasonable to question whether
subjects high on neurotic traits, as they are described, are perhaps more inclined towards other
forms of antisocial behaviour, such as substance abuse or self-harm. The implication that
neuroticism is closely related to anxiety disorders is also apparent when viewing the N-
indicating questions in the EPI questionnaire19, which reference sadness, nightmares and
feelings of insecurity. By today’s standards, many of the symptom descriptions for neurotic
personality traits are perhaps related to an increased risk of other forms of unhealthy
behaviour, rather than those included in the self-report indexes in the Cambridge study. If
true, this line of reasoning could indicate that Eysenck’s theory about a link between
neuroticism and antisocial behaviour is correct, even though that particular form of antisocial
behaviour is not properly assessed in this study.
6.3. Neurotic extraversion
Interestingly, the EPI was significant in one other aspect, in that neurotic extraversion
adequately predicted official recidivism in this study. This result can be considered further
support for Eysenck’s theory of personality and crime, although it does not have to be for the
same reasons as listed in the sections above. Eysenck holds that high scores of both traits
dispositions the subject to an even higher risk of delinquent behaviour than high scores on
only one trait. In Eysenck’s theory, neurotic extraverts are less resistant to delinquency, due to
inadequate aggregation of conditioning experiences, by virtue of their extraversion.
Additionally, high levels of neuroticism reinforce the delinquent behaviour, causing it to
become habitual. Recidivism, by definition, can be thought of as acting out of habit. It could
also be considered an instance of inadequate conditioning, as described by Eysenck, since the
initial conviction has not deterred the subject from partaking in further delinquency. Whilst
this is not likely an exhaustive explanation of recidivism, the results found here relating to
neurotic extraversion support that notion, at least in regards to officially registered recidivism.
Contrary to Eysenck’s theory and as previously mentioned, neuroticism could
not predict recidivism. In support of the theory, more than 40% of officially registered
recidivist possessed high levels of both traits (n 23/n 54), whilst they only constitute less than
30% (n 114) of the total population. Based on this result, it can be argued that a high score of
neuroticism in combination with extraversion is relevant for recidivist behaviour. Van Dam et
al argue that extraverted personality traits imply a greater risk for detection, and thus for being
19 Table 1 in Appendix.
28
officially registered for crime (van Dam et al: 13). It is conceivable that this is also true for
Eysenck’s neurotic extraverts who, as previously mentioned, are thought of as unstable in
their extraversion; impulsive, aggressive and excitable. Commonsensically, it is possible that
subjects with traits such as those described as unstable extraversion traits by Eysenck run a
higher risk of detection by authorities. However, the results found through this study support
the opposite notion, as neuroticism actually has a negative effect on the extravert’s risk of
official recidivism, even though the negative effect was very negligible.
With regards to self-reported recidivism, no correlation was found with neurotic
extraversion. One explanation for this is the low rates of neurotic extraverts in the self-
reported group. In fact, neurotic extraverts were much more likely to be official recidivists,
with an odds ratio indicating an over-risk of close to 2 in that group. Official recidivist
constituted about 47% of neurotic extraverts (54 out of 114), whilst self-reported recidivists
were only about 37% (28 out of 113) of that group.
6.4. Conduct disorder – an alternative theory
Farrington has held that when EPI results are shown to have significance, this is likely
because the measure picks up on some individual characteristic which in itself is predictive of
delinquency. Along with Murray, he has argued for the importance of impulsiveness in
relation to offending. Farrington and Murray also suggest that impulsiveness is one of the
chief risk factors for what they call ‘conduct disorder’, referring to behaviour which violates
societal norms and the rights of others (Murray & Farrington 2010: 633-5). Other critics have
also highlighted the problem with overlapping concepts in regards to personality theory
(Kubrin, Stucky & Krohn 2009: 73). This form of criticism can gain further support from
results found in the Cambridge study. When held for the variable ‘conduct disorder’, assessed
by school teachers when respondents were aged 8, some of the significance of extraverted
traits was lost. In fact, extraversion could significantly predict officially recorded recidivism
only for boys who were considered ‘well-behaved’ by their teachers, but not for boys whom
teacher’s deemed ‘moderately badly’ or ‘very badly’ behaved20. When held for conduct
disorder, the correlation between extraversion and official recidivism gave a p-value of .04 for
‘well-behaved’ boys. That is to say, the correlation was significant at the confidence level.
However, for ‘moderately badly’ behaved boys and ‘very badly’ behaved boys, the p-values
were .059 and .488 respectively. Whilst this result suggests a partial overlap between conduct
20 Chi-Square tests – table 11 in Appendix.
29
disorder and extraversion, it also supports the idea that they are separate properties, as
extraversion still increased the risk for recidivism in otherwise well-behaved boys.
In the instances where significance was lost, it is of interest to consider the risk
that there might be an overlap between constructs. One possibility, once again drawing on the
argument from van Dam et al, is that certain traits, such as being outgoing and impulsive,
renders a person more likely to draw the attention of authorities. This assumption is evidently
also very true for conduct disorder. In fact, this is precisely what is measured by the conduct
disorder variable: If teachers are thought of as authorities, and ‘bad behaviour’ in school is a
form of delinquency, being considered ‘badly behaved’ by a school teacher at age 8-9 is of
course such an instance of authority detection of delinquent behaviour.
In the Cambridge study, there is a fairly high chance of having one’s delinquent
behaviour go undetected. For adults, 74.5% reported delinquent acts, but only 27.1% were
convicted. This means that nearly two thirds of all adult delinquent behaviour was never
penalised. However, of those who had juvenile convictions, only 8% received no further
convictions as adults. Furthermore, it is reasonable to assume that some poor behaviour in
school has the capacity to handicap an individual in other aspects of life, in regards to factors
such as school grades, relationship with peers and parents. These factors can in turn
disposition the individual towards delinquency.
If the conduct disorder variable and the extraversion variable overlap, this is
likely because they measure similar traits. Murray and Farrington describe how low IQ, which
is a main risk factor for conduct disorder on their view, inhibits abstract thinking,
handicapping the individual from foreseeing the consequences of their action (ibid: 636).
Eysenck’s theory on extraversion includes a similar notion about how arousability in the brain
relates to poor conditioning, disposing the individual to ‘act without thinking’. Similarly, both
lines of reasoning can be compared to the elusive ‘impulsiveness’ trait as stressed by
Farrington, and included as a measure by Eysenck. Eysenck also makes a distinction between
genotypical and phenotypical aspects of personality. Whilst sociability and aggressiveness are
phenotypic aspects, it is possible that they relate to genotypic aspects such as IQ or ability to
treat abstract concepts.
7. Conclusion
Whilst most of the hypotheses this study intended to review seem to be correct, the results
presented in this paper do not provide much support for Eysenck’s theory on personality and
crime. In fact they seem to suggest that there are better models for predicting delinquency and
30
antisocial behaviour. Furthermore, the reservations with regard to operationalisation and the
discussion of the results should indicate that there are some methodological problems, but
perhaps more importantly, many issues regarding how personality theorists deal with complex
constructs and even more complex subjects. Personality theoretical explanations can by no
means be dismissed by virtue of the results found here. In fact, most of Eysenck’s notions
could be correct, only this study failed to find much support. That said, I find it quite
fascinating that this study more or less replicated the results from van Dam et al. Whilst this
paper should not function as strong evidence for or against Eysenck’s theory or any other
theory on personality and crime, I believe it illustrates some major points of how different
constructs relate and some of the issues surrounding attempts at measuring personality.
I think that Eysenck is right in that there is no escaping the psychological
aspects of social life; the filter through which everything is sent and received. Because of this,
I believe it is too early to abandon ship on all psychological explanations of crime and I am
looking forward to seeing what progress can be made within the discipline.
8. Proposal for future research
Whilst this study found little support for Eysenck’s theory of personality and crime, some
interesting observations were made, especially with regards to the functions related to
extraversion. I have spent some time discussing incarceration effects. In this study, no known
respondents were incarcerated at the time of personality type testing, and high scores of
extraversion could significantly distinguish between recidivists (re-offenders) and non-
recidivists, whereas the study by van Dam et al on incarcerated respondents found support for
incarceration effects. With regards to personality type measurements and convicted offenders,
this seems to be a particularly interesting field of study. Some personality theorists hold that
personality is consistent over time, but this does not appear to be true in the presence of
incarceration effects. Alternatively, personality measures are not developed for respondents
who cannot interact under their “normal” social circumstances during incapacitation. Beside
the more commonly discussed consequences of prisonisation on the individual’s
psychological health, it is possible that incarceration and incapacitation can have unexpected
effects on both behaviour and self-image, and perhaps even on personality. Such discoveries
could be of importance to findings on the humanitarian aspects of imprisonment as
punishment. Furthemore, psychological and psychosocial research on the effects of
incarceration can give some new insights into how deterrence is viewed in criminological
discourse.
31
9. References
Van Dam, C., Janssens, J., De Bruyn, E., (2003) “PEN, Big Five, Juvenile Delinquency and
Criminal Recidivism”, Personality and Individual Differences, 39 (2005), p. 7-19.
Djurfeldt, G. (2009). “Förberedelser – att lära känna sina data” In Göran Djurfeldt & Mimmi
Barmark (ed), Statistisk verktygslåda 2 – multivariat analys, p. 25-52. Lund: Författarna och
Studentlitteratur.
Djurfeldt, Larsson & Stjärnhagen (2003) Statistisk verktygslåda. Studentlitteratur AB.
Eysenck, H. (1996) “Personality and Crime: Where Do We Stand”, Psychology, Crime and
Law, 2008, 3/2, p. 143-152.
Eysenck, H. & Eysenk, S. (1969) Manual of the Eysenck Personality Inventory. London:
University of London Press.
Eysenck, H. & Eysenck, S. (1969) Personality Structure and Measurement. London:
Routledge & Kegan Paul.
Farrington, D. (1999) Cambridge Study in Delinquent Development [Great Britain], 1961-
1981. Institute of Criminology: Cambridge University.
Farrington, D. (2005) “Childhood Origins of Antisocial Behavior”, Clinical Psychology and
Psychotherapy, 2005, 12, p. 177-190.
Farrington, D., Biron, L., LeBlanc, M. (1982) “Personality and Delinquency in London and
Montreal” In John Gunn & David P. Farrington (ed), Abnormal Offenders, Delinquency, and
the Criminal Justice System, p.153-201. John Wiley & Sons.
Kubrin, C., Stucky, T., & Krohn, M. (2009) Researching Theories of Crime and Deviance.
New York: Oxford University Press.
Levine, S. & Jackson, C., (2004) “Eysenck’s Theory of Crime Revisited: Factors or Primary
Scales?”, Legal and Criminal Psychology, 2004, 1/9, p. 135-152.
Murray, J. & Farrington, D. (2010) “Risk Factors for Conduct Disorder and Delinquency: Key
Findings from Longitudinal Studies”, The Canadian Journal of Psychiatry, 2010, 10/55, p.
633-642.
32
10. Appendix
Table 1
E1. Do you often long for excitement?
N2. Do you often need understanding friends to cheer you
up?
E3. Are you usually carefree?
N4. Do you find it very hard to take no for an answer?
-E5. Do you stop and think things over before doing
anything?
L6. If you say you will do something do you always keep
your promise, no matter how inconvenient it might be to
do so?
N7. Does your mood often go up and down?
E8. Do you generally do and say things quickly without
stopping to think?
N9. Do you ever fell ‘just miserable’ for no good reason?
E10. Would you do almost anything for a dare?
N11. Do you suddenly feel shy when you want to talk to an
attractive stranger?
-L12. Once in a while do you lose your temper and get
angry?
E13. Do you often do things on the spur of the moment?
N14. Do you often worry about things you should not have
done or said?
-E15. Generally, do you prefer reading to meeting people?
N16. Are your feelings rather easily hurt?
E17. Do you like going out a lot?
-L18. Do you occasionally have thoughts and ideas that
you would not like other people to know about?
N19. Are you sometimes bubbling over with energy and
sometimes very sluggish?
-E20. Do you prefer to have few but special friends?
N21. Do you daydream a lot?
E22. When people shout at you, do you shout back?
N23. Are you often troubled about feelings of guilt?
L24. Are all of your habits good and desireable ones?
E25. Can you usually let yourself go and enjoy yourself at a
gay party?
N26. Would you call yourself tense or ‘highly-strung’?
E27. Do other people think of you as being very lively?
N28. After you have done something important, do you
often come away feeling you could have done better?
-E29 Are you mostly quiet when you are with other
people?
-L30. Do you sometimes gossip?
N31. Do ideas run through your head so that you cannot
sleep?
-E32. If there is something you want to know about, would
you rather look it up in a book than talk to someone about
it?
N33. Do you get palpitations or thumping in your heart?
-E34. Do you like the kind of work that you need to pay
close attention to?
N35. Do you get attacks of shaking or trembling?
L36 Would you always declare everything at the customs,
even if you knew that you could never be found out?
-E37. Do you hate being with a crowd who play jokes on
one another?
N38. Are you an irritable person?
E39. DO you like doing things in which you have to act
quickly?
N40. Do you worry about awful things that might happen?
-E41. Are you slow and unhurried in the way you move?
-L42. Have you ever been late for an appointment or
work?
N43. Do you have many nightmares?
E44. Do you like talking to people so much that you never
miss a chance of talking to a stranger?
N45. Are you troubled by aches and pains?
E46. Would you be very unhappy if you could not see lots
of people most of the time?
N47. Would you call yourself a nervous person?
-L48. Of all the people you know, are there some whom
you definitely do not like?
E49. Would you say that you were fairly self-confident?
N50. Are you easily hurt when people find fault with you
or your work?
-E51. Do you find it hard to really enjoy yourself at a lively
party?
N52. Are you troubled with feelings of inferiority?
E53. Can you easily get some life into a rather dull party?
-L54 Do you sometimes talk about things you know
nothing about?
N55. Do you worry about your health?
E56. Do you like playing pranks on others?
N 57. Do you suffer from sleeplessness?
33
Table 2
Convicted as juvenile 10-16
Frequency Percent Valid Percent Cumulative Percent
Valid
No convictions 327 79,6 79,6 79,6
1 or more convictions 84 20,4 20,4 100,0
Total 411 100,0 100,0 Table 3
Self-reported non-official delinquency 14-17 * Convicted at 17 Crosstabulation
Convicted at 17 Total
No conviction
1 or more convictions
Not at risk
Self-reported non-official delinquency 14-17
Not known Count 0 0 2 2
Official (10-16) or no reported delinquency (14-17)
Count 327 40 3 370
Self-reported non-official delinquency
Count 32 7 0 39
Total Count 359 47 5 411
Table 4
Adult criminal acts
Frequency Percent Valid Percent Cumulative Percent
Valid
Not known whole period 20 4,9 4,9 4,9
1 or more criminal acts 306 74,5 74,5 79,3
No criminal acts 85 20,7 20,7 100,0
Total 411 100,0 100,0
34
Table 5
Juvenile-adult official recidivism * EPI Extraversion Crosstabulation
EPI Extraversion Total
Low Extraversion
High Extraversion
Juvenile-adult official recidivism
Non-recidivist
Count 159 172 331
% within EPI Extraversion
90,3% 82,3% 86,0%
Recidivist
Count 17 37 54
% within EPI Extraversion
9,7% 17,7% 14,0%
Total
Count 176 209 385
% within EPI Extraversion
100,0% 100,0% 100,0%
Chi-Square Tests
Value df Asymp. Sig. (2-sided)
Exact Sig. (2-sided)
Exact Sig. (1-sided)
Pearson Chi-Square 5,127a 1 ,024 Continuity Correctionb 4,482 1 ,034 Likelihood Ratio 5,265 1 ,022 Fisher's Exact Test ,027 ,016
Linear-by-Linear Association
5,114 1 ,024
N of Valid Cases 385
Variables in the Equation
B S.E. Wald df Sig. Exp(B)
Step 1a ExtravertD ,699 ,313 4,990 1 ,025 2,012
Constant -2,935 ,542 29,367 1 ,000 ,053
a. Variable(s) entered on step 1: ExtravertD.
35
Table 6
Self-reported recidivism * EPI Extraversion Crosstabulation
EPI Extraversion Total
Low Extraversion
High Extraversion
Self-reported recidivism
Non-recidivist
Count 153 160 313
% within EPI Extraversion
85,0% 76,9% 80,7%
Recidivist
Count 27 48 75
% within EPI Extraversion
15,0% 23,1% 19,3%
Total
Count 180 208 388
% within EPI Extraversion
100,0% 100,0% 100,0%
Chi-Square Tests
Value df Asymp. Sig. (2-
sided)
Exact Sig. (2-
sided)
Exact Sig. (1-
sided)
Pearson Chi-Square 4,037a 1 ,045
Continuity Correctionb 3,536 1 ,060
Likelihood Ratio 4,094 1 ,043
Fisher's Exact Test ,053 ,029
Linear-by-Linear Association 4,027 1 ,045
N of Valid Cases 388
a. 0 cells (0,0%) have expected count less than 5. The minimum expected count is 34,79.
b. Computed only for a 2x2 table
Variables in the Equation
B S.E. Wald df Sig. Exp(B)
Step 1a ExtravertD ,531 ,266 3,985 1 ,046 1,700
Constant -2,265 ,449 25,481 1 ,000 ,104
a. Variable(s) entered on step 1: ExtravertD.
36
Table 7
Juvenile-adult official recidivism * EPI Neuroticism Crosstabulation
EPI Neuroticism Total
Low Neuroticism
High Neuroticism
Juvenile-adult official recidivism
Non-recidivist
Count 160 171 331
% within EPI Neuroticism
87,4% 84,7% 86,0%
Recidivist
Count 23 31 54
% within EPI Neuroticism
12,6% 15,3% 14,0%
Total
Count 183 202 385
% within EPI Neuroticism
100,0% 100,0% 100,0%
Chi-Square Tests
Value df Asymp. Sig. (2-sided)
Exact Sig. (2-sided)
Exact Sig. (1-sided)
Pearson Chi-Square ,615a 1 ,433 Continuity Correctionb ,406 1 ,524 Likelihood Ratio ,617 1 ,432 Fisher's Exact Test ,465 ,263
Linear-by-Linear Association ,613 1 ,434 N of Valid Cases 385 a. 0 cells (0,0%) have expected count less than 5. The minimum expected count is 25,67. b. Computed only for a 2x2 table
Table 8
Self-reported recidivism * EPI Neuroticism Crosstabulation
EPI Neuroticism Total
Low Neuroticism
High Neuroticism
Self-reported recidivism
Non-recidivist Count 155 158 313
% within EPI Neuroticism 82,9% 78,6% 80,7%
Recidivist Count 32 43 75
% within EPI Neuroticism 17,1% 21,4% 19,3%
Total Count 187 201 388
% within EPI Neuroticism 100,0% 100,0% 100,0%
Chi-Square Tests
Value df Asymp. Sig. (2-sided)
Exact Sig. (2-sided)
Exact Sig. (1-sided)
Pearson Chi-Square 1,138a 1 ,286 Continuity Correctionb ,880 1 ,348 Likelihood Ratio 1,143 1 ,285 Fisher's Exact Test ,305 ,174
Linear-by-Linear Association 1,135 1 ,287 N of Valid Cases 388 a. 0 cells (0,0%) have expected count less than 5. The minimum expected count is 36,15. b. Computed only for a 2x2 table
37
Table 9
Juvenile-adult official recidivism * EPINeurotExtrav Crosstabulation
EPINeurotExtrav Total
Not high on N+E
High on N+E
Juvenile-adult official recidivism
Non-recidivist
Count 240 91 331
% within EPINeurotExtrav
88,6% 79,8% 86,0%
Recidivist
Count 31 23 54
% within EPINeurotExtrav
11,4% 20,2% 14,0%
Total
Count 271 114 385
% within EPINeurotExtrav
100,0% 100,0% 100,0%
Chi-Square Tests
Value df Asymp. Sig. (2-sided)
Exact Sig. (2-sided)
Exact Sig. (1-sided)
Pearson Chi-Square 5,079a 1 ,024 Continuity Correctionb 4,380 1 ,036 Likelihood Ratio 4,806 1 ,028 Fisher's Exact Test ,036 ,020
Linear-by-Linear Association 5,066 1 ,024 N of Valid Cases 385 a. 0 cells (0,0%) have expected count less than 5. The minimum expected count is 15,99. b. Computed only for a 2x2 table
Variables in the Equation
B S.E. Wald df Sig. Exp(B)
Step 1a EPINeurotExtrav ,671 ,301 4,958 1 ,026 1,957
Constant -2,718 ,447 36,905 1 ,000 ,066
a. Variable(s) entered on step 1: EPINeurotExtrav.
Table 10
Self-reported recidivism * EPINeurotExtrav Crosstabulation
EPINeurotExtrav Total
Not high on N+E High on N+E
Self-reported recidivism
Non-recidivist Count 228 85 313
% within EPINeurotExtrav 82,9% 75,2% 80,7%
Recidivist Count 47 28 75
% within EPINeurotExtrav 17,1% 24,8% 19,3%
Total Count 275 113 388
% within EPINeurotExtrav 100,0% 100,0% 100,0%
38
Chi-Square Tests
Value df Asymp. Sig. (2-sided)
Exact Sig. (2-sided)
Exact Sig. (1-sided)
Pearson Chi-Square 3,036a 1 ,081 Continuity Correctionb 2,563 1 ,109 Likelihood Ratio 2,930 1 ,087 Fisher's Exact Test ,090 ,056
Linear-by-Linear Association 3,028 1 ,082 N of Valid Cases 388 a. 0 cells (0,0%) have expected count less than 5. The minimum expected count is 21,84. b. Computed only for a 2x2 table
Table 11
Juvenile-adult official recidivism * EPI Extraversion * CONDUCT DISORDER OF BOY Crosstabulation
CONDUCT DISORDER OF BOY EPI Extraversion Total
Low Extraversion
High Extraversion
WELL BEHAVED
Juvenile-adult official recidivism
Non-recidivist
Count 100 112 212
% within EPI Extraversion
95,2% 87,5% 91,0%
Recidivist
Count 5 16 21
% within EPI Extraversion
4,8% 12,5% 9,0%
Total
Count 105 128 233
% within EPI Extraversion
100,0% 100,0% 100,0%
MODERATELY BADLY
Juvenile-adult official recidivism
Non-recidivist
Count 24 19 43 % within EPI Extraversion
88,9% 67,9% 78,2%
Recidivist Count 3 9 12 % within EPI Extraversion
11,1% 32,1% 21,8%
Total Count 27 28 55 % within EPI Extraversion
100,0% 100,0% 100,0%
VERY BADLY BEHAV
Juvenile-adult official recidivism
Non-recidivist
Count 19 22 41 % within EPI Extraversion
70,4% 78,6% 74,5%
Recidivist Count 8 6 14 % within EPI Extraversion
29,6% 21,4% 25,5%
Total Count 27 28 55 % within EPI Extraversion
100,0% 100,0% 100,0%
Total
Juvenile-adult official recidivism
Non-recidivist
Count 143 153 296
% within EPI Extraversion
89,9% 83,2% 86,3%
Recidivist
Count 16 31 47
% within EPI Extraversion
10,1% 16,8% 13,7%
Total
Count 159 184 343
% within EPI Extraversion
100,0% 100,0% 100,0%
39
Chi-Square Tests
CONDUCT DISORDER OF BOY Value df Asymp. Sig. (2-sided)
Exact Sig. (2-sided)
Exact Sig. (1-sided)
WELL BEHAVED
Pearson Chi-Square 4,212c 1 ,040
Continuity Correctionb 3,321 1 ,068
Likelihood Ratio 4,465 1 ,035
Fisher's Exact Test ,064 ,032
Linear-by-Linear Association
4,194 1 ,041
N of Valid Cases 233
MODERATELY BADLY
Pearson Chi-Square 3,564d 1 ,059 Continuity Correctionb 2,438 1 ,118 Likelihood Ratio 3,704 1 ,054 Fisher's Exact Test ,101 ,058
Linear-by-Linear Association
3,500 1 ,061
N of Valid Cases 55
VERY BADLY BEHAV
Pearson Chi-Square ,487e 1 ,485 Continuity Correctionb ,151 1 ,698 Likelihood Ratio ,488 1 ,485 Fisher's Exact Test ,547 ,349
Linear-by-Linear Association
,478 1 ,489
N of Valid Cases 55
Total
Pearson Chi-Square 3,321a 1 ,068
Continuity Correctionb 2,772 1 ,096
Likelihood Ratio 3,386 1 ,066
Fisher's Exact Test ,083 ,047
Linear-by-Linear Association
3,311 1 ,069
N of Valid Cases 343 a. 0 cells (0,0%) have expected count less than 5. The minimum expected count is 21,79. b. Computed only for a 2x2 table c. 0 cells (0,0%) have expected count less than 5. The minimum expected count is 9,46. d. 0 cells (0,0%) have expected count less than 5. The minimum expected count is 5,89. e. 0 cells (0,0%) have expected count less than 5. The minimum expected count is 6,87.