7/30/2019 ergasia77 (2)

1/17

Exploring the Utility of NarrativeAnalysis an Diagnostic Deasion Making:Picture-Bound Reerence, Elaboration,and Fetal Alchol Spectrum Disorders

John C. TTruman E. CoHeather Carmichael _Susan J.University of Wash

homePurpose: To evaluate classification accuracy and clinical feasibility of a narrativeOlson,stley analysis tool for identifying children with.a fetal alcohol spectrum disorder (FASD).slyMethod: Picture-elicitecd narratives generated by 16 age-matc:hed pairs of school-

ington aged children (FASD vs. typical development [TD]) were coded for semanficelaboration and reference strategy by judges who Were unaware of age, gender,and group meimnbership of the participahts. Receiver operating characteristic (ROC)curves were used to examine the classification accuracy of the resulting set of narrativemeasures for making 2'classifications: (a) for the 1 .children diagnosed with FASD,low performance (n = 7) versus average performance (n = 9) on a standardizedexpressive language task and (b) FASD (n = 16) versus TD (n=16).Results: Combining the rates of semantic elaboration and pragmaticallyinappropriate reference perfectly matched a classification based on performanceon the standardized language task. More importantly, the rate of ambiguous nominalreference was highly accorate in classifying children with an FASD regardless oftheir performance on the standardized language task (area under the ROCcurve = .863, confidence interval =.736-.991).Conclusion: Results support further study'of the-diagnostic utility of narrative analysisusing discourse level measures ofelaboration andtchildren's strategic uie of referernce.KE Y WORDS: assessment procedures, pragmatics, discourse analysis,,language sample analysis, diagnostics'

arrative discourse permeates our social lives from an early age,'aking it a critical, area to address in the measurement oflanguage abilities. For more than 2 decades, narrative analysishas been recommended as an ecologically valid' way to assess the pro-duction ofmeaningful language in socially integrated discourse (see Owens,,1999). Underlying this recommendation is the assumption that narrativeanalysis provides a more integiated appraisal of a child's communicativeabilities than is -possible "via standardized language measures (Adams,'Iloyd, Aldred, &Baxendale, 2006; Botting & Adams, 2005; Culatta, -Page,& Ellis, 1983; Norbury & Bishop, 2003; Wagner, Nettelbladt, Sahlen, &Nilholm, 2000). Thus, narrative analysis 'should be able to identify chil-dren with meaningful communicative impairments that might be missedusing conventional standardized assessment instruments. The current studyexamined this largely untested assumption by retrospectively comparingnarratives produced by ,school-aged children!with a fetal alcohol spectrumdisorder (FASD) with those produced by , age- and gender-matched

7/30/2019 ergasia77 (2)

2/17

typically developing (TD) peefs. Comparisons were madevia the Semantic Elaboration Coding System' (Thorne,2004), which systematically measures the use of prag-matically appropriate strategies of reference and thesemantic elaboration of concepts.

FASDs and LanguageChildren with prenatal alcohol exposure exhibit a

wide range of abilities across all body systems (Astley &Clarren, 2000; Carmichael-Olson, Morse, & Huffine,1998; Streissguth, 1997). When specific growth, facial,and central nervous system impairments are presentwithin a well-specified range, a diagnosis from within thecontinuum of FASD can be rendered (see Astley, 2004;Chudley et al., 2005). Because the development and use oflanguage have been reported to be affected by high levelsof prenatal alcohol exposure (Mattson & Riley, 1998;Streissguth, Barr, Kogan, & Bookstein, 1996), measure-ment of language ability has been an important featureof interdisciplinary assessment of these individuals. Thepreponderance of evidence regarding language behav-ior in children with an FASD has been gathered usingstandardized, norm-referenced tests (Abkarian, 1992;Becker, Warr-Leeper, &Leeper, 1990; Carney & Chermak,1991; Church, Eldis, Blakley, & Bawle, 1997; Church &Kaltenbach, 1997; Gentry et al., 1998; Janzen, Nanson,& Block, 1995; Weinberg, 1997). The goal of these stud-ies has been to establish how well children with anFASD comprehend and/or produce language structuresin standardized contexts. Typically, these contexts mea-sure language using discrete responses at or below thelevel of single-sentence utterances. Although the resultshave revealed an array of performance deficits, no coredeficit profile has emerged.

Because no recognizable deficit profile has resultedfrom research using standardized language tests, re-searchers have begun to look at suprasentential dis-course in school-aged children diagnosed with an FASD.Inpreliminary research, discourse level deficits have beendocumented in children with an FASD including reducedability to provide sufficient information for listeners bothduring conversations (Hamilton, 1981) and in narratives(Coggins, Friet, & Morgan, 1998; Coggins, Olswang,Carmichael-Olsen, & Timler, 2003). In addition, caregiversreport that children with an FASD often fail to accommo-date the perspectives of others during interaction (Tinler,Olswang, & Coggins, 2005). This early research suggeststhat despite widely'variable performance on standard-ized tests, children with an FASD may have difficultyproducing integrated extended discourse that requires

'An unpublished training manual for the Semantic Elaboration CodingSystem is available from the first author via jct6@u.washington.edu-

them to balance linguistic and social-cognitive taskdemands (Coggins et al., 2003). This emerging profile,when coupled with our ability to identify an FASD in-dependent of communication ability, makes this hetero-geneous group of children an ideal population to test thediscriminative utility of a narrative analysis system.

NarrativeAnalysisAs a primary form of extended discourse, narratives

provide children with a means of verbally recapitulat-ing experiences (Bishop &Edmundson, 1987; Feagans &Appelbaum, 1986; Feagans & Short, 1984) and are animportant source ofknowledge about inference, social cog-nition, and perspective taking (Owens, 1999). The abilityto produce contextually integrated extended discourse isdifficult to measure using the discrete responses typicalof standardized tests. Analysis of narrative samples offers a viable alternative.

Unlike standardized measures, narrative analysisallows for measurement of discourse level parameters ofcommunication that result directly from the pragmaticsof a relatively communicative interaction (Owens, 1999).These parameters of behavior manifest in the history ofconcepts as they are developed across sentences in thenarrative text and should provide information regarding language ability that is unavailable in the noncom-municative context of standardized testing. Arguablythe most informative context in which to sample chil-dren's narrative ability is on e that obligates them to organize and generate narratives without an adult modelor other contextual supports (Curenton &Justice, 2004Juncos-Rabadan, Pereiro, & Rodriguez, 2005; Norbury&Bishop, 2003). This decontextualized narrative discoursestresses the language system by limiting the nonlinguistictools available during discourse. These limitations determine the type of discourse breakdowns that can be predicted in children with compromised cognitive systemsThe Semantic Elaboration Coding System (Thorne, 2004was designed to capture these predictable discourse levebehaviors in school-aged children.The Semantic ElaborationCoding System

"The Semantic Elaboration Coding System implements a framework for narrative analysis based uponcognitive linguistics (Croft & Cruse, 2004; Langacker1991; Talmy, 2000b; Tomasello, 2003). Cognitivelinguistics seeks to account for structural properties of languagein terms of its relation to more general conceptual structures and functions. It has, therefore, examined "thlinguistic structuring of basic ideational and affectivecategories attributed to cognitive agents, such as attention, perspective, volition, and intention, and expectation and effect" (Talmy, 2000b, Vol. I, p. 3).

7/30/2019 ergasia77 (2)

3/17

In a-detailed look at ,the,'conceptual structuring ofnarrative, Talmy (2000a) identified a series of concep-tual parameters in narrative that constitute a "set oforganizing principles that apply in common, across allmajor cognitive systems" (p. 422). The Semantic Elab-oration Coding System is organized along two of theseparameters that would be expected to vary monotoni-cally with quality in a decontextualized narrative. Thefirst involves the strategic use of inguistic reference toassure that concepts are explicit and uniquely,identifi-able in the text. The second involves the degree to whichsemantic concepts are elaborated or well specified in thetext.

The strategic use of reference in narrative. In anynarrative, it is essential that the concepts involved (bothentities and events) are kept'distinct from each other toreduce ambiguity. The linguistic strategies-used to makedistinctive reference to various concepts in a narrativemay vary in form from semantically complete phrasesand clauses to semantically ambiguous forms like pro-nouns dependent upon the presuppositions the narratorhas about the listener's current knowledge and atten-tion state regarding those concepts. Wong and Johnston(2004) dentify three basic reference functionsin narrativetasks related to presuppositions 'about the knowledgeand attention state ofthe listener: (a) the introductionofnew concepts into the discourse (presupposes no knowl-edge of the concept), (b ) the maintenance of foreground/in-focus concepts (presupposes both knowledge of andattention to the concept), and (c) the reintroductionofpreviously introduced background/out-of-focus conceptsin the discourse (presupposes knowledge of, but lim-ited attention to, the concept). The distinction betweenadequate and inadequate use of various strategies formeeting these discourse functions cannot be made with-out consideration of both the textual and extratextualcontext of the particular instance of use (see Cornish,1999, for a discussion; see also Levine & Klin, 2001;Maratsos, 1976; van Hoek, 1997; -Wong, 2001; Wong,Au, & Stokes, 2004).

In decontextualized narrative discourse, strategiesfor meeting all three basic reference functions are re-stricted to the linguistic code (see Halliday & Hasan,1976). To maintain unambiguous reference, storytellersmust use discourse strategies that do not presupposeunwarranted knowledge or attention on the part of theirlistener or require extralinguistic support to be inter-preted meaningfully.,This can be particularly challerg-ing for younger storytellers as they continuously adapttheir narratives to the ever-changing knowledge andattention states of their listeners (Coggins et al., 1998;'Cornish, 1999; Lewis, 2004; Wong & Johnston, 2004).

School-aged children are learning to effectively in-corporate a variety of reference stra.tegies into their lan-guage production (Stephens, 1988), allowing differentiation

between' children with typical and delayed language de -velopment (see Liles,'Duffy, Merritt,'& Purcell,'1995;Wong,2001, for example). Measurement of the strategic useof referehce in narrative serves as a primary componentof the Semantic Elaboration Coding System..

Semantic elaboration.Decontextualized discoursea so demands a greater density' of ideas, or semanticelaboration, from the stdryteller. An analysis of narra-tive elaboration must account for. the contribution ofparticular words or syntactic structures' to the listener'sgrowing conceptualization of a concept in a way that ac -counts' for the history of that concept in the precedingdiscourse (Croft & Cruse, 2004; Fauconnier, 2004; KMin,Weingartner, Guzman, & Levine, 2004; Talmy, 2000b).This requires that the measurement of elaboration beintegrated with the measurement of successful referenceto those concepts because a,structure cannot contributeto the elaboration of a concept if it does not unambig-uously make reference to that concept.

Investigators have heretofore used a variety of lex-ically based and syntactically based measures to captureelaboration in discourse. These measures have differen-tiated children with different overall language ability asmeasured by , standardized language tests (Condouris,Meyer, & Tag6r-Flusberg, 2003; Hammer,iYont, & Tomblipn,2005; Loban, 1976). Most approaches treat the structuresthey quantify independent of the textual history of'theconcepts they describe, much as is done with standard-ized tests of lexical knowledge or syntactic comhpetence.Consequently, they are unlikely' to provide inforniationregardinglanguage abilitybeyond that available throughstandardized testing (a largely uintested assumption, butsee Hesketh, 2004). A system that integrates measure-,ment of elaboration and reference strategy may provideinformation about integrated language abilities that isinaccessible to the more traditional approaches.

PurposeThe Semantic Elaboration Coding System is

designed to be used with decontextualized narrativesproduced,for a naive listener by school-aged children asthey look through the wordless picture book, FrogWhereAre You? (Mayer, 1969). It integrates analysis of twonarrative discourse parameters that w6uld be difficult toquantify using standardized measures: (a) ambiguity,.the use of inappropriate strategies of reference, and(b) elaboration, the,semantic elaboration of concepts as,they develop across the narrative (see Thorne &Coggins,2004; see also Thorne & Coggins, 2005).

With respect to children diagnosed with an FASD,the Semantic Elaboration Coding System must be ableto (a)identify children with an FASD who performpoorlyon a standardized test to establish a level' of concurrent

7/30/2019 ergasia77 (2)

4/17

validity with that standardized test and (b ) identify thosechildrenwith an FASD who perform like typically devel-oping (TD) children on the standardized language testto establish superior classification accuracy when com-paredwith that test. This study was designed to test thepotential fo r the Semantic Elaboration Coding Systemto make these key discriminations. Specifically, the re-search questions under study were as follows:1. Can narrative analysis using the Semantic Elabora-

tion Coding System correctly classify a group of school-aged children into separate groups based on typicaldevelopment versus an identified FASD?

2. Can narrative analysis using the Semantic Elabo-ration Coding System accurately predictwhich chil-dren with an identified FASD have either averageor low performance on a standardized languagetask?

3. Which specific measure or combinations of mea-sures from within the Semantic Elaboration CodingSystem are most accurate in performing these dis-crimination tasks, and, therefore, reasonably war-rant further development?

MethodParticipantsThirty-two school-aged children from tw o previous

studies (Carmichael-Olson & Astley, 2005; Coggins,1995) participated. They ranged in age from 8;5 yearsto 11;7 years (M = 9;11 years) and presented a rangeof socioeconomic and ethnic profiles. Sixteen of thechildren presented key clinical features consistent withan FASD while the remaining children were consid-ered TD.

Childrenwith an FASD. The 16 FASD participantshad a diagnosis of either (a)full or partial fetal alcoholsyndrome or (b ) a confirmed alcohol exposure accompa-nying static encephalopathy or neurobehavioral disor-der. Diagnosis was performed by an interdisciplinaryteam at the University of Washington Fetal AlcoholSyndrome Diagnostic and 1Prevention Network. Allchildren were originally diagnosed using the 1999version of the 4-DigitDiagnostic Code (Astley & Clarren,1999). All codes were translated into the 2004 version ofthe 4-Digit Diagnostic Code (Astley, 2004) to provide anup-to-date diagnostic standard for comparison. Partici-pants reflected "the true diversity and continuum of dis-ability associated withprenatal alcohol exposure" (Astley,2004, p. 13). Table 1 specifies each participant's 4-DigitDiagnostic Code, which provides information regardingtheir growth, facial morphology, brain development, andalcohol exposure (see Astley, 2004, for details of codeinterpretation).

Existing nonverbal and -verbal measures -providedan additional basis for selection and are also provided inTable 1. The Matrices subtest from the Kaufman BriefIntelligence Test (Kaufman &Kaufinan, 1990) providedan overall measure of nonverbal problem solving, withparticipants excluded based on a score 1.5 standarddeviations below the mean (range: 79-130; M = 101).The 16 participants in the FASD group were dichoto-mizedinto tw o performance groups based on their scoresfrom the Re-Creating Sentences subtest of the Test ofLanguage Competence (RS-TLC; Wiig & Secord, 1989):(a) an average-performance group (n = 9) with standardscores within one standard deviation of the mean (be-tween 7 and 10) and (b) a low-performance group (n = 7)with standard scores two or more standard deviationsbelow the mean (between 3 and 4).

The resulting sample included 9 females and 7 males.Family income for the group ranged from $15,000 to$220,000 per annum (M=$88,000, Mdn = $75,000). Thegroup included 11 children identified as Caucasian, 3 asbi- or multiracial, and 1 each as African American andNative American. Only 3 ofthe children were still livingwith their biological parent(s) at the time data were col-lected. The remaining 13 were in adoptive or legal guard-ianship placements (5 with relatives).

TD peers. Each participant with FASD was pairedwith a TD peer matched on chronological age (12months,mean difference = 3.5 months). Thirteen TD age-matchedpeers also matched the gender of their FASD counterpart(15 females, 17 males). Table 1 displays age and genderfor all participant pairs.

The TD aged-matched peers were recruited fromelementary schools representing tw o school districts inthe greater metropolitan Seattle area. Median familyincomes were similar across school districts ($61,435-$62,195). The sample included 12 children identified asCaucasian, 2 as Asian, and 1 each as African Americanand Hispanic (representative of the home county forbothdistricts).

No intelligence or standardized language measureswere available for TD participants. However, a schoolpsychologist familiar with the 16 children and with theprofile of FASD screened school records for each childwith respect to school performance, social ability, and general behavior. Based on this review of available recordseach was judged to be following a typical developmen-tal course due to their unremarkable behavior and adequate school achievement. The TD participants did noundergo the same interdisciplinary assessment as thechildren with FASD.Materials

Self-generated, decontextualized narratives were selected from tw o independent databases: on e from a study

7/30/2019 ergasia77 (2)

5/17

Table 1. Participant characteristics: diagnosis, language performance group, test scores, age, and gender.FASD Group TD Group.Diagnostic Code Re-Creating Sentences- Kaufman Brief Intelligence.&Category' Test of Language Competence" Test-Matricesc Gender Age Gender Age2443- A 3: low 79 F 8;9. F 8;8

1234 =F 3: low 90 M 9;2 M 8;113432 =B 3:low 95 F 8;11 F 9;32444 = A 3:'16w 8.0 M 10;8 M 10;114234 = E 4- low 80 F 10;6 M 9;94344 = C 3: low 96 M 11;1 M 11;41324=G 3:low. 87 F 11;2 - M 11;61124=H 7: average 92 F 8;5 F : 8;41124 = H 7: average 128 M 8;8 M 8;91224 = H 7: average 130 M 8;10 M 8;113344 =C 7: average 101 M 9;3 M, 9;11224 H 8: average 114 F 9;5 F 9;21223 H 9: average 113 F 10;6 F 9;61223 =H 9: average 122 F, 10;6 F 10;101224 H 7: average 98 F 11;2. M 11;53233 =E 10: average 105 M 11;5 M 11;7Mean Average: 8; low:. 3 101 9;11 9;10Note. Diagnostic code provides information, from left to right, regarding growth, facial morphology, brain development, 'and alcohol exposure. Scoresrange from 1 (unremarkable) o 4 (severe). FASD = fetal alcohol spectrum disorder; TD = typical development; F= female; M = male; FASD categories:A = FAS (alcohol exposed); B= FAS (alcohol exposure unknown); C = partial FAS (alcohol exposed); E-H indicate the remaining FASD categories.(with confirmed alcohol exposure). Details for interpretation of.the 4-Digit Diagnostic Code of FASD can be found in Astley (2004).OAstley (2004). bM= 10, SD= 3. CM= 100, SD= 15.

involving children diagnosed with an FASD (Carmichael-Olson &Astley, 2005) and the second from a normativestudy of TD school-aged children (Coggins, 1995).

ProceduresAll narratives from the two respective databases

were elicited using FrogWhereAre You? (Mayer, 1969).In both studies, participants were tested individuallyand received the same instructions. Each child was in-structed to look through Mayer's book to become familiarwith the story line. When the child completed preview-ing the story, the examiner exhorted the participant totell the best story possible while using the picture bookas a visual prompt. In each case, examiners were seated'across the. room from the child, with the storybook ou tof their line of sight.

Transcriptionand CodingNarratives were recorded on audiocassette and ortho-

graphically transcribed by trained graduate students. Thesecond author supervised the narrative collection andtranscription process and then stripped all transcriptsofidentifyinginformation while assigning each a random

code so that relevant information could be retrieved forlater data analysis.Transcripts were coded by the first author using theSemantic Elaboration Coding System (Thome, 2004). Thesystem assigns codes along the param6ters of(a) ambigu-ity, he consequence of inappropriate reference strategies,and (b) elaboration of concepts.Ambiguity. Operationally, measurement of ambigu-ity involved coding references to concepts as eithler un-ambiguous or ambiguous. To reduce the number ofcodingcategories, ambiguous anaphoric reference strategies formaintenance, and reintroduction of concepts an d ambig-uous introduction strategies thdt make unwarranted pre-

suppositions about listener knowledge and attention-werecollapsed into.just two ambiguity categories (nominaland pronominal) because their use has similar impacton a listener's discourse processing (Cornish, 1999; vanHoek, 1997).Elaboration. Measurement of elaboration involvedcore lexical items that unambiguously introduced con-cepts into the story,that were coded as either schematic(i.e., minimally characterized) or elaborated. Additionalwords that helped elaborate concepts were also coded.Fo r this study, each word in a transcript was as-signed 1 of 10 mutually exclusive scoring codes along

I I

7/30/2019 ergasia77 (2)

6/17

these two parameters, or a null code. The 10 scoringcodes are presented later with a brief definition of eachcategory. Further details can be found in the SemanticElaboration Coding System (Thorne, 2004).

Ambiguity codes.Two codes identified unambiguousanaphoric reference to concepts and made a distinctionbetween nominal forms and pronominal forms servingmaintenance and reintroduction functions.1. Nominal reference (NR): an unambiguous nominal

form used to maintain or reintroduce a concept pre-viously introduced into the discourse (e.g., a dogintroduced into the narrative later referred to un-ambiguously as the dog).

2. Pronounreference (PR):an unambiguous pronomi-nal form used to maintain or reintroduce a conceptpreviously introduced into the discourse (e.g., a doglater referred to unambiguously as he or it).

Two additional codes were used to identify cases in whichreference was ambiguous.3. Ambiguous nominal reference (ANR): an ambiguous

use ofa nominal form attempting to introduce, main-tain, or reintroduce a concept.

4. Ambiguous pronounreference (APR): an ambiguoususe of a pronominal form attempting to introduce,maintain, or reintroduce a concept.Semantic elaboration codes. Six codes were used

to quantify semantic elaboration. Four of these codesdichotomized core lexical items as either schematic orelaborated. Two codes identified core lexical units asschematic.5. Schematic verb (SV): a word that introduced basic

information regarding an event into the discourse(i.e., the fact that somethinghappened: went,got, is,going).

6. Schematic nominal (SN): a word that introducedbasic information regarding an entitjr into the dis-course (e.g., boy, dog, frog, ar,animal, hing).

Two additional codes identified those core lexical itemsthat were relatively elaborated.7. Elaboratedverb (EV): a word that introduced ela-

borated information regarding an event into the dis-course (i.e., the manner in which somethinghappened:ran, ell, chased,yelled).

8. Elaboratednominal (EN): a word that introducedelaborated information regarding an entity unam-biguously into the discourse (e.g., 2Tmmy, elk, bullfrog,wife, bumblebee).

Tw o final codes identified word forms associated withthese core lexical items that provided additional seman-tic elaboration. These two codes are the most frequentin the Semantic Elaboration Coding System, making

up a significant portion of the information the systemgathers.9. Verb satellite (VS): a word providing elaborating in-

formation about averb (e.g., went away, ran quickly,in the morningwhen he got up).

10 . Nominal modifier (NM): a word providing elabo-rating information regarding a nominal (e.g., bigmad owl, frog thatranaway).

A null code was used to indicate that a word did not fitany of the 10 scoring categories.Null code (null): a word not meeting operationaldefinitions for any ambiguity or elaboration codecategory in the system.

AnalysisIntercoderagreement.A graduate student in speech

and hearing sciences was recruited and trained to func-tion as a secondary coder. Coder competence was estab-lished when intercoder agreement between the primary(the first author) and secondary coder reached a kappaof .7 or better for each code in the system on a set of fivetraining narratives taken from the CIULDES databank(MacWhinney, 2000).

The primary coder then scored all 32 of the studynarratives while the secondary coder independentlyscored 25% of the narratives (n = 8) randomly selectedusing SPSS for Windows (SPSS, 1998). Random selectionand all coding were completed before either coder knewthe diagnostic status, age, or gender of the storytellers.The resulting sample contained 5 narratives from the TDgroup and three from the FASD group.

For both training and study narratives, kappa wascalculated as a measure of agreement between codersfor each of the 10 Semantic Elaboration Coding Sys-tem codes and for the null code. Calculation of kappa wasconducted separately foreach code, with agreement basedon a binary decision-every word in the narratives wasidentified as carrying the designated code or not (follow-ing Kraemer, n.d.; see also Bakeman & Gottman, 1997).All words not coded by both judges with the designatedcode were treated as disagreements in the calculationof kappa for that code. Because it reveals performancefor individual codes and avoids overly optimistic anddifficult-to-interpret estimates of overall agreement thatcan occur in multicode calculations of kappa, this is aconservative method of estimating intercoder agreement(Acklin, McDowell, Verschell, & Chan, 2000; see alsoKraemer, Periyakoil, &Noda, 2004). Results are presentedin Table 2.

Eleven kappa statistics were computed. The preci-sion of 10 kappa scores ranged from substantial (i.e., .6-.8to almost perfect (i.e., >.8; Kraemer, Periyakoil, &Noda,

7/30/2019 ergasia77 (2)

7/17

Table 2. Interrater agreement (kappa) for all Semantic ElaborationCoding System codes.

Code XVS .836SV .893EV .935NM .653NR .876SN .793EN .792PR .767ANR .764APR .540Null (no code) .836Note. Calculation of the kappa was conducted separately for each code,with agreement based on a binary decisi6n: Every word in the narrativeswas identified as carrying the designated code or not. All words notcoded by both judges with the designated code were treated as disagree-ments in the calculation of the kappa for that code. VS = verb satellite;SV = schematic verb; EV = elaborated verb; NM = nominal modifier;NR =nominal reference; SN = schematic nominal; EN = elaboratednominal; PR = pronoun reference; ANR = ambiguous nominal reference;APR = ambiguous pronoun reference.

2004; Landis & Koch, 1977). The kappa statistic forthe coding of ambiguous pronoun referdnce was in themoderate range of precision (i.e., .4-.6). Because it in-creases chances of Type H error, the moderate level ofintercoder agreement on this single code might be con-sidered insufficient for clinical application (cf.Bakeman& Gottman, 1997; Cicchetti & Sparrow, 1981), high-lighting the need for attention to coder training and mea-surement stability in subsequent development of thesystem (see the Discussion section for more on pronounsand intercoder agreement).

Data preparation.' Several preparatory steps weretaken to ready the data for analysis. First, the,totalnumber of words (TW) was calculated fo r each narrative(excluding mazes, range: 154-796; M = 308) using SALT(Miller, 2004). Second, the raw frequencies of each of the11 scoring codes in the 32 narratives were calculatedalso using SALT. Next, the raw frequencies of codes forAPR and ANR were combined to create 'a summaryambiguity score, while the raw frequencies of codes forEV forms, EN forms, VS forms, and NM forms werecombined to create a summary elaboration score.

Next, code and score rates were computed by divid-ing each measure by,the TW in the narrative. Becausethe Semantic Elaboration Coding System examines nar-ratives on a word-by-word basis, the TW in a story wasconsidered to best represent the length of the story andthereby became the denominator used in the calculationofcode and score rates. Using a common denominator for

the calculation of all rates also facilitated comparisonbetween rates.2 This process resulted in 26 narrativemeasures for analysis: 11'code frequencies, 2 summaryscores, and 13 associated rates as a function ofnarrativelength. Table 3 presents all 26 measures.

Classification accuracy analysis. To explore theaccuracy of classification for each of the 26 measures,empirical classification rates including sensitivity (truepositive rate), specificity (true negative rate), and ef-ficiency (overall accuracy rate) were examined. Methodsfrom signal detection theory based on these three mea-surement parameters were implemented to judge therelative potential of each ofthe 26 measures to match theclassification of participants provided by the appropri-ate reference standard (Kraemer, 1988, 1992; Kraemer,Noda, & O'Hara, 2004; McFall &Treat, 1999). The twoclassifications of particular interest were as follows:Classification1-the accuracy of each measure in classi-fying participants as members of the FASD or TD group;Classification2-the accuracy of each measure in classi-fying participants diagnosed with an FASD as a memberof the group with average performance or low. perfor-mance on the RS-TLC.

More specifically, for both classifications, each ofthe26 measures was analyzed using an empirical receiveroperating characteristics (ROC) curvehbased on valuesobtained from the 32 narratives (for recent applicationsof this method, see Dwolatzky et aL,, 2003; see also'Heilmann, Weismer, Evans, & Hollar, 2005). This anal-ysis compared the classification-of each measure againstthe classification by the appropriate reference standard,and provided anROC curve plottingthe sensitivity against1 minus the specificity of each measure for all obtainedvalues. The RO C curve, the area under the ROC curve(AUC), and the asymptotic significance, standard error,and 95% AUC confidence int6rvals were all calculatedusing SPSS.

ResultsCriteria or a ReasonableMeasureThe AU G is a widely accepted measure of overall

accuracy (McFall &Treat, 1999). An index of effect size,the AU G can distinguish between tests that are' ran-dom (AUG = 0.5), poorly accurate (0.5-0.7), moderately21t is common to calculate proportional rates for referential terms based ontotal number of referential opportunities rather than total words. In thecase of the Semantic Elaboration Coding System; this would mean dividingeach ambiguity code frequency by the total number ofNRs + PRs,+ ANRs +APRs + SNs + ENs: To assure that results using total word rates were notsubstantially skewed when compared with those using total opportunityrates, correlations and ROC curves were'run for ambiguity code ratescalculated using both methods. Rates were substantially correlated (r> .9),and AUC results were not significantly different (p > A4).AUC data arereported using rates calculated with TW to facilitate comparison betweenelaboration rates and ambiguity rates.'

Thorne et aL.: Narrative Analysis in DiagnosticDecisionMaking 465

7/30/2019 ergasia77 (2)

8/17

Table 3. Semantic Elaboration Coding System measures evaluated for classification accuracy.Raw Code Frequencies Summary Scores and Formula Rates as a Function of TWAmbiguity measure

NR NR/TWPR PR/TWANR ANR/TWAP R APR/TWAS =ANR +APR AR =AS/TW

Elaboration measureEV EV/lTWEN EN/TWVS VS/TWNM NM/TWES = EV + EN + VS + NM ER = ES/TW

Schematic codeSV SV/TWSN SN/TWNull Null/TW'

Note. TW = total words; AS = ambiguity score;AR = ambiguity rate; ES = elaboration score; ER = elaboraterate; Null = no code.

accurate (0.7-0.9), and highly accurate (0.9-1.0; seeSwets, 1988). To be considered reasonable in the currentanalysis, a measure required an AUC with an asymp-totic significance better than .02 and an AUC 95% con-fidence interval with a lower bound above 0.7. Thesecriteria assured not only that the AUC was significantlydifferent from a random test (i.e., AUC = 0.5) but alsothat it had at least a moderate chance of accurately clas-sifying cases.

Sensitivity, specificity, and efficiency values werecalculated at the best possible cut-points for measuresobtaining an AUC with an asymptotic significance bet-ter than .02. A "best cut-point" was defined as the ob-tained value along the ROC curve with the highestefficiency (overall accuracy rate). If multiple cut-pointshad equivalent efficiency, the on e with the highest sen-sitivity was chosen.DataPresentation

Figure 1 presents AULTC data for both of the testedclassifications. The top portion of Figure 1 displays re-sults for Classification 1-FASD versus TD group mem-bership. The lower portion of Figure 1 displays resultsfo r Classification 2-performance grouping for the FASDparticipants: low performance versus average perfor-mance on the RS-TLC. Figure 1 includes asymptoticsignificance levels, AUC values, and 95% confidence in-tervals fo r each measure (i.e., test) that achieved anasymptotic significance better than .02 for each classifi-cation. The shaded regions of Figure 1 indicate the rangeused to determine that a particular test's confidence

interval indicated reasonable accuracy. The four measuresthat reached the criteria for a reasonable measure areindicated by a double asterisk. Those measures not in-cluded on Figure I were not statistically different froma random test.

Table 4 displays information on the sensitivity,spec-ificity, and efficiency at the best cut point for the mostpromising measures. These measures were chosen basedon visual inspection of ROC curve shape (see Kraemer,1992). The top of Table 4 includes data for Classifica-tion 1-TTD versus FASD, while the lower portion dis-plays data for Classification 2--RS-TLC performancegroup.

Accuracy for Classification I-FASDVersus TD GroupMembershipAs can be seen in the top portion of Figure 1, the rate

of ambiguous nominal reference (ANR/TW) obtained anAUC of.863, asymptotic significance = .0001. This AUCindicates classification accuracy solidly within themoderate-to-high accuracy range with a 95% confidenceinterval from .736 to .991.

As revealed in Table 4, the rate 6f ambiguous nom-inal reference (ANR/TW) achieves strong sensitivityspecificity, and efficiency for Classification 1. At the bestpossible cut point, >.0165 (i.e., greater than 1.65% oftotal words being ambiguous nominal references equalsa positive test for FASD status), ANNR/TW achieved asensitivity of87.5%, a specificity of 75%, and overall efficiency of 81.25%. At this particular cut point, it correctly

7/30/2019 ergasia77 (2)

9/17

Figure 1. Area under the receiver operating characteristic (AUC) curve with 95% onfidence intervals for measuresasymptotically different than a random test. Analysis compares dassification by Semantic Elaboration CodingSystem measures indicated and that by reference standard. For Classification 1, eference standard isdiagnosis ofan FASD by interdisciplinary team. For Classification 2, reference standard isbased on average performance (within1 SDof the mean) or low performanc6 (92% of 185 ambiguous nominalreferences). In a concordance search ofall 32 stories, only14 (

7/30/2019 ergasia77 (2)

10/17

Table 4. Sensitivity, specificity, and efficiency of most promising measuresat their best cut-point for both Classification 1 and Classification 2.Measure Cut-Point Sensitivity(%) Specificity(%) Efficiency(%)

Classification 1: FASD vs. TD group membershipANR/TW >.01 65 87.50 75 81.25PR 5.0 81.25 81.25 81.25

Classification 2: RS-TLC performance group (FASD only)AR >.0588 85.7 100 93.75ER 50.4816 71.4 100 87.50VS 589.0 100 66.7 81.25Note. Best cut-point chosen as most efficient test weighing sensitivity overspecificity for equally efficient tests. High sensitivity and a negative testhelp rule out diagnosis, which isuseful for screening. High specificity anda positive test help to confirm a diagnosis. Higher efficiency indicatesbetter overall diagnostic performance. RS-TLC = Re-Creating Sentencessubtest of the Test of Language Competence.

achieved for the AR. The obtained value of .921, asymp-totic significance = .005, places this result in the highlyaccurate range, with a 95% confidence interval rangingfrom .763 to 1.00. The lower portion of Table 4 showsthat at the best cut-point, >.0588 (i.e., more than 5.88%ofwords being ambiguous indicates low RS-TLC perfor-mance group), this measure obtained a sensitivity of85.7%, a specificity of 100%, and an efficiency of 93.75%.At this cut-point, the measure accurately classified 6 of7 children in the low RS-TLC performance group and all9 of those in the average RS-TLC performance group.

The AUC of the ER, as shown in the lower portionof Figure 1, is .905, asymptotic significance = .007. ThisAUC falls in the highly accurate range, with a 95% con-fidence interval ranging from .753 to 1.00. As seen inlower portion of Table 4, at the best cut-point,

7/30/2019 ergasia77 (2)

11/17

an orrule correctly classified all 16 childrenmwithrespectto their performance on the RS-TLC.

These results support the notion that narrative arial-ysis should play an important role in diagnostic decisionmaking. Classification using information from a narra-tive analysis perfectly matched that based on performanceon a standardized measure of expressive language. Testingof the Semantic Elaboration Coding System against' awider range of standardized language measures to es-tablish a better understanding ofits concurrent Validityappears warranted.ClassificationAccuracyfor FASD Diagnosis

A second goal of this feasibility research was to de-termine whether the Semantic Elaboration Coding Sys-tem would provide the information needed for a moreaccurate classification than the standardized measureto support the idea that narrative analysis provides amore ecologically valid description of a child's ability toproduce meaningful language in socially integrated dis-course. This is clearly shown with the results of ourbetween-groups classification.

One ambiguity measure satisfied strict accuracy cri-teria in matching the reference standard classification ofstudy participants into FASD and TD groups. The rateof ANR (calculated as a function of narrative length)accurately identified participants previously identifiedby an interdisciplinary team as having an FASD regard-less of the participant's performance on a standardizedexpressive language task. At its'.best value, the rate ofANR matched the interdisciplinary team classificationfor 14 of 16 (87.5%) children with an FASD, and iden-tified 12 of 16 (75%) children with typical development.Moreover, this measure achieved a substantially reliablekappa statistic (O= .764).

This finding is arguably the most important from thecurrent study.The data demonstrate a procedure for reli-ably quantifying meaningful performance differences inthe use of reference strategies that has the potential to bediagnostically informative for apopulation of children whohave resisted easy classification using standardized lan-guage tests. In spite of the FASD group's wide range ofcognitive and linguistic abilities, a single narrativemea-sure keyed to the rate at which a child used inappropri-ate nominal reference strategies was able to match aninterdisciplinary team diagnosis of an FASD for all but2 children.Ambiguous NominalReferenceby Children in the FASD Group

A closer look at the posthoc analysis of ambiguousnominal reference sheds light on the behaviors children

in the FASD group were using in' their narratives 'thatwas captured bythe ANR code. For a reference to be ratedambiguous in the Semantic Elaboration Coding System,an equivocal word choice must occur. In other words, thestoryteller has selected a word that fails in its mission toeither introduce or unambiguouslk reference a concept.There are three basic conditions under which a nominalreference may be considered,,ambiguous in the codingsystem.

In Condition 1, an existing concept (e.g., THEBOY)is treated as ifit were new-by using an indefinite nominalform (e.g., a boy) to maintain or reintroduce it. Condition 1involves pragmatically inappropriate use of an indefi-nite nominal. Post hoc analysis found this type, of errorto be rare (

7/30/2019 ergasia77 (2)

12/17

into decontextualized discourse (indefinite forms) thatare distinct from those for accessing (i.e., maintaining orreintroducing) file cards already present in the discourse(definite nominals and pronouns; see Croft, 2001; Klinet al., 2004; Langacker, 1991; Maratsos, 1976; van Hoek,1997; Wong & Johnston, 2004).

Applying this analogy, if jar is used to introduce theentity concept GLASS CONTAINER, it will be the nom-inal form listed first on the listener's file card for thatconcept. It is not the only lexical form available on thefile card, but it will be the on e that most effectively refersto the concept. When storytellers who have introducedthe concept into the story with the nominal complexa jar decide to maintain or reintroduce that same con-cept using a different nominal form, let's say the bottle,they are acting ostensively as if their listener will recog-nize the overlapping nature ofJAR and BOTTLE (bothare glass containers) and will assign reference appropri-ately. However, context plays a significant role in whenand how the quite distinct categories JAR and BOTTLEare equivalent and when they are not.

In a shared visual context in which only on e item fit-ting in the category GLASS CONTAINER is seen, visualinformation aids the listener in disambiguating a refer-ence to that glass container whether the term jar or bottleis used. In a decontextualized discourse, however, thatvisual information is unavailable to the listener (despitebeing available to the storyteller). With only their devel-oping conception of the entities involved in the story tosupport inferences about the referent, listeners may no tbe able to quickly and easily determine if the bottle beingreferred to is the same GLASS CONTAINER as the jarintroduced into the story earlier.

When this occurs, a storyteller who has visual sup--porthas not recognized the potential increase in pro-cessing demands that a switch in reference forms causesfor the listener who does not have access to that visualsupport (Wong & Johnston, 2004). This increased pro-cessing effort may or may not lead to an equivalent un-derstanding of the concepts in the story as listenersattempt to find the most efficient way to resolve the am-biguity. Because this switching of reference forms wouldnot create an equivalent difficulty for a listener who sharedthe visual context, we refer to this reference strategy aspicture-bound eference (following Shapiro & Hudson, 1991).

A picture-bound reference strategy is even more ap-parent in Condition 3, in which new concepts are intro-duced as if they already existed in the discourse. Whenthere is a shared visual context, picture-bound referenceis a reasonable and pragmatically appropriate strategyfor introducing concepts because visual information willsupport listeners as they quickly and easily disambig-uate the reference. There is not an obligation to verballyintroduce concepts into the discourse if they can be in-troduced visually.

In decontextualized narrative discourse, the visualinformation is not available to the listener. So, storytellerswho use picture-bound reference strategies to introduceconcepts are not recognizing the increased processing de-mands they are placing on their listeners and are riskingthat their listeners will not develop an equivalent under-standing of the concepts in the story. By identifying thispicture-bound reference strategy, the Semantic Elabora-tion Coding System was able to substantially match inter-disciplinary team classification of children into the groupwith an FASD independent of the child's performance ona standardized expressive language task.Pronounsand Picture-BoundReferencing

As a definite form, pronouns have the potential to bemarkers of picture-bound reference. Our results, how-ever, do not show ambiguous pronoun reference to be areasonable measure for accurately classifying children.It may be that pronouns represent a more complexreference form providing for more fine-grained manip-ulation of listener attention than full nominal phrases(Gundel, Hedberg, & Zacharski, 2001). As relativelycomplex forms, it is more likely that pronominal formswill be used in error by children in this age group(cf. Schelletter & Leinonen, 2003; van Der Lely, 1997;Wigglesworth, 1997; Wong &Johnston, 2004). This maybe seen in the relatively poor specificity of PR in Table 4.Results indicate that despite the fact that the children inthe FASD group tended to use relatively few unambig-uous pronoun references, the same can be said for manyof theirTTD peers, leadingto arelativelyhighrate of false-positive classifications. The poor classification accuracyofAPR may also reflect the fact that, unlike PR (c = .767),APR was relatively difficult for judges to agree upon(c = .540). Lower precision in a measure results in ahigher chance of a Type IHerror and potentially masksthe utility of the underlying construct. Given the uncer-tainties, this study highlights the need to consider nomi-nal and pronominal lexical forms separately in narrativeresearch but does not diminish the need for continuedstudy of children's development and use of pronominalforms of reference in discourse.Conclusion

Whether they exhibited average or low performanceon a standardized language measure, children in thisstudy who had an existing interdisciplinary team diag-nosis of a disorder on the fetal alcohol spectrum (FASD)were more likely than were their typically developing peersto use a picture-bound reference strategy during story-telling. This strategy could be identified with reasonableaccuracy using the rate of ANR calculated as a functionof narrative length as defined in the Semantic Elabora-tion Coding System. If this result can be replicated in a

7/30/2019 ergasia77 (2)

13/17

7/30/2019 ergasia77 (2)

14/17

Carmichael-Olson, H., & Astley, S. J. (2005). [Interveningwith children/adolescents with FAS/ARND]. Unpublishedraw data.Carmichael-Olson, H., Morse, B. A. , & Huffine, C. (1998).Development and psychopathology: Fetal alcohol syndromeand related conditions. Seminars in ClinicalNeuropsychia-try, 3, 262-284.Carney, L. J., & Chermak, G. D. (1991). Performance of

American Indian children with fetal alcohol syndrome onthe test of language development. Journalof Communica-tion Disorders,24, 123-134.Chudley, A. E., Conry, J., Cook, J. L., Loock, C., Rosales,T., & LeBlanc, N. (2005). Fetal alcohol spectrum disorder:Canadian guidelines for diagnosis. CanadianMedical

Association Journal,175(Suppl. 5), $1-S21. *Church, M. W., Eldis, F., Blakley, B. W., & Bawle, E. V.(1997). Hearing, language, speech, vestibular, and dentofa-cial disorders in fetal alcohol syndrome.Alcoholism: Clinicaland ExperimentalResearch,21, 227-237.Church, ML W., &Kaltenbach, J. A. (1997). Hearing, speech,language, and vestibular disorders in the fetal alcohol

syndrome: A literature review. Alcoholism: ClinicalandExperimentalResearch,21, 495-512.Cicchetti, D. V., & Sparrow, S. S. (1981). Developing criteriafor establishing interrater reliabiity of specific items:

Applications to assessment of adaptive behavior. AmericanJournalof Mental Deficiency, 86, 127-137.Coggins, T. E. (1995). [Narratives produced by typically

developing school-aged children]. Unpublished raw data.Coggins, T. E., Friet, T., & Morgan, T. (1998). Analysingnarrative productions in older school-age children andadolescents with fetal alcohol syndrome: An experimentaltool for clinical applications. ClinicalLinguistics&Phonetics,12 , 221-236.Coggins, T. E., Olswang, L. B., Carmichael-Olsen, I., &Timler, G. (2003). On becoming socially competent com-

municators: The challenge for children with fetal alcoholexposure. InternationalReview of Research in MentalRetardation,27, 121-150.

Condouris, K., Meyer, E., & Tager-Flusberg, H. (2003).The relationship between standardized measures of lan-guage and measures of spontaneous speech in children withautism. American Journalof Speech-LanguagePathology,12 , 349-358.

Cornish, F. (1999).,Anaphora, discourse,and understanding:Evidence from English andFrench.New York. OxfordUniversity Press.Croft, W. (2001). Radicalconstructiongrammar:Syntactictheory in typologicalperspective. Oxford, England: OxfordUniversity Press.Croft, W., & Cruse, D. A. (2004). Cognitive linguistics.Cambridge, England: Cambridge University Press..Culatta, B., Page, J. L., & Ellis, J. (1983). Story retelling asa communicative performance screening tool. Language,

Speech, and HearingServices in Schools, 14, 67-74.Curenton, S. M., & Justice, L M. (2004). African Americanand Caucasian preschoolers' use of decontextualized lan-guage: Literate language features in oral narratives.

Language,Speech, and HearingServices in Schools, 35,240-253.

Dwolatzky, T., Whitehead, V., Doniger, G., Simon, E.,Schweiger, A. , Jaffe, D., et al. (2003). Validity of a novelcomputerized cognitive battery for mild cognitive impair-ment. BMC Geriatrics, (1), 4.Eisenberg, S., & Gillam, R. B. (2005, June). Complex nounphraseelaboration n children'sstories. Poster sessionpresented at the Symposium on Research in ChildLanguageDisorders, Madison, WI.Fauconnier, G. (2004). Mental spaces, language modalities,conceptual integration. In S. Davis &B. S. Gillon (Eds.),

Semantics:A reader.Oxford, England: Oxford UniversityPress.Feagans, L., & Appelbaum, M. (1986). Validation of lan-guage subtypes in learning disabled children. Journalof

EducationalPsychology, 78, 358-364.Feagans, L, & Short, E. J. (1984). Developmental differ-ences in the comprehension and production of narratives byreading-disabled and normally achieving children. ChildDevelopment, 55, 1727-1736.Gentry, B., Griffith, L., Dancer, J., Davis, P., Eaton, B., &Schulz,E. (1998). Prenatal alcohol exposure and communi-cation, behavior, and nonverbal intelligence of 3 school-age

children. PerceptualandMotorSkills, 86(3,Pt. 1), 1089-1090.Gundel, J. K., Hedberg, N., & Zacharski, R. (2001).Definite descriptions and cognitive status in English:

Why accommodation is unnecessary. EnglishLanguageandLinguistics,5(02), 273-295.Hialliday, Li.A. K., & Hasan, R. (1976). Cohesionin English.London: Longman.Hamilton, M. (1981). Linguisticabilitiesof children with fetalalcoholsyndrome.Unpublished doctoral dissertation, Uni-versity of Washington.Hammer, C. S., Yont, K. M., & Tomblin, J. B. (2005).Language sampling for kindergarten children with andwithout SLI: Mean length of utterance, IPSYN, and NDW.Journalof CommunicationDisorders,38, 197-213.Heilmann, J., Weismer, S. E., Evans, J., & Hollar, C.(2005). Utility of the MacArthur-Bates CommunicativeDevelopment Inventory in identifying language abilities oflate-talking and typically developing toddlers.American

Journalof Speech-LanguagePathology,14, 40-51.Hesketh, A. (2004). Grammatical performance of childrenwith language disorder on structured elicitation andnarrative tasks. ClinicalLinguistics& Phonetics,18, 161-182.Janzen, L.A., Nanson, J. L., &Block, G. W. (1995). Neuro-psychological evaluation of preschoolers with fetal alcohol

syndrome. Neurotoxicologyand Teratology, 17, 273-279.Juncos-Rabadan, 0., Pereiro, A. X., & Rodriguez, M. S.(2005). Narrative speech in aging: Quantity, informationcontent, and cohesion. Brain andLanguage, 95, 423-434.Kaufman, A. S., & Kaufman, N. L. (1990). The KaufmanBrief ntelligenceTest. Circle Pines, MN: AGS.Klin, C.A., Weingartner, K. IV., Guzman, A. E., & Levine,W. H. (2004). Readers' sensitivity to linguistic cues innarratives: How salience influences anaphor resolution.

Memory & Cognition,32, 511-522.Kraemer, H. C. (1988). Assessment of 2X 2 associations:Generalization of signal-detection methodology. TheAmer-

ican Statistician,42(1), 37-49.

7/30/2019 ergasia77 (2)

15/17

Kraemer,H. C. (1992). Evaluatingmedical tests: Objectiveand quantitativeguidelines.London: Sage Publications.Kraemer,H. C. (n.d.). Quality of measurement. RetrievedDecember 20 , 2004, from http://mirecc.stanford.edu/video-streaming.htm#MethodsKraemer,HL C., Noda, A., & O'Hara, I. (2004). Categoricalversus dimensional approaches to diagnosis: Methodologicalchallenges. JournalofPsychiatricResearch, 38, 17-25.Kraemer, H. C., Periyakoil, V. S., &Noda, A. (2004).Tutorial in biostatistics: Kappa coefficients in medicaliesearch. In R. B. D'Agostino (Ed.), Tutorials n biostatistics:Statisticalmethods in clinical studies (Vol. 1, pp. 85-105).Hoboken, NJ: VTiley.Landis, J. R., & Koch, G. G. (1977). The measurement ofobserver agreement for categorical data. Biometrics, 33,159-174.Langacker; I. W. (1991). Concept, image and symbol: The

cognitive basisofgrammar.Berlin: Mouton de Gruyter.Levine, W. H., & Klin, C. M. (2001). Tracking of spatial infor-mation in narratives. Memory & Cognition,29, 327-335.Lewis, D. (2004). Scorekeeping in a language game.

In S. Davis & B. S. Gillon (Eds.), Semantics:A reader(pp. 803-816). Oxford, England: Oxford University Press.Liles, B. Z., Duffy, R. J, Merritt, D. D., & Purcell, S. L.(1995). Measurement of narrative discourse ability in chil-dren with language disorders. Journal ofSpeech andHearingResearch, 38, 415-425.Loban, W. (1976). Languagedevelopment: Kindergartenthroughgrade twelve (No. 18). Urbana, MI: National Councilof Teachers of English.MacWhinney, B. (2000).,The CHILDESproject:Tools,for analyzing talk (3rd ed.). Mahwah, NJ: Erlbaum.Maratsos, AL P. (1976). The use of definite and indefinitereference n young children:Anexperimentalstudy ofsemantic

acquisition.New York. Cambridge University Press.Mattson, S.N., & Riley, E. P. (1998). A review of theneurobehavioral deficits in children with fetal alcoholsyndrome or prenatal exposure to alcohol. Alcoholism:Clinicaland ExperimentalResearch, 22, 279-294.Mayer, .VL1969). Frogwhere areyou? New York-Dial Press.McFall, RL RL, &Treat, T. A. (1999). Quantifying the infor-mation value ofclinical assessments with signal detectiontheory.Annual Review ofPsychology, 50, 215-241.Miller, J. F. (2004). Systematic Analysis of Language Tran-scripts (Version 8.02, Research edition) [Computer software].Madison, WI: University of Wisconsin-Madison.Norbury, C. F., & Bishop, D. V. (2003). Narrative skills ofchildren with communication impairments. InternationalJournalof CommunicationDisorders,38, 287-313.Owens, RL (1999). Language disorders:A unctionalapproachto assessmentand intervention (3rd ed.). Boston: Allyn &Bacon.Schelletter, C., & Leinonen, E. (2003). Normal andlanguage-impaired children's use of reference: Syntacticversus pragmatic processing. Clinical Linguistics &Phonetics,17(4-5), 335-343.Shapiro, L. IL , &Hudson, J. A. (1991). Tell me a make-believe story: Coherence and cohesion in young children's

picture-elicited narratives. DevelopmentalPsychology,27, 960-974.SPSS. (1998): SPSS for Windows (Version 9.0) [Computer

software]. Chicago: Author.Stephens, M. I. (1988). Pragmatics. In M. A. Nippold (Ed.),Later languagedevelopment:Ages nine through nineteen(pp. 247-261). Boston: College-Hill.Streissguth, A. P. (1997). Fetalalcoholsyndrome:A guide or

families and communities.' Baltimore: Brookes:Streissguth, A. P., Barr, H. M., Kogan, J., &Bookstein,F. L. (1996). Understanding he occurrenceof secondarydisabilitiesin clients with fetal alcohol syndrome (FAS) andfetal alcoholeffects (FAE):Finalreport.Seattle: Universityof Washington.

Swets, J. A. (1988, June 3). Measuring the accuracy ofdiagnostic systems. Science, 240, 1285-1293.Talmy, L. (2000a). Toward a cognitive semantics: 7ypologyandprocess in conceptstructuring Vol. II, pp. 417-482).Cambridge, MA . MIT Press.Talmy, L. (2000b). Toward a cognitive semantics:Volumes Iand II. Cambridge, MA : MIT Press.Thorne, J. C. (2004). The semantic elaborationcodingsystem.Unpublished training manual, ,Universityof Washington.Thorne, J. C., & Coggins, T. E. (2004, November). Semanticelaboration n narrativesof childrenwith FASD: The Se-mantic ElaborationCoding System. Poster presented atthe annual convention of the American Speech-Language-Hearing Association, Philadelphia, PA.Thorne, J. C., & Coggins, T. E. (2005, June). Exploring theutility of narrativeanalysis to aid in the diagnosis of etal

alcoholspectrumdisorders FASD): The SemanticElaborationCodingSystem (SECS).Poster presented at the Symposiumon Research in Child Language Disorders, Madison, WI.Timler, G.R., Olswang, L. B., & Coggins, T. E. (2005). "Do I

know what I need to do?"A social communication interven-tion for children with complex clinical profiles. Language,Speech, and HearingServices in Schools, 36, 73-85.

Tomasello, M. (2003). Constructinga language:A usage-basedtheory of languageacquisition.Cambridge,MA : Harvard University Press.van Der Lely, H. K. J. (1997). Narrative discourse in gram-matical specific language impaired children: A modularlanguage deficit? Journalof Child Language,24, 221-256.van Hoek, K. (1997). Anaphora and conceptualstructure.Chicago: The University of Chicago Press.Wagner, C. R., Nettelbladt, U., Sahlen, B., & Nilholm, C.(2000). Conversation versus narration in pre-school children

with language impairment. International Journal ofLanguageand CommunicationDisorders,35, 83-93.Weinberg, N. Z. (1997). Cognitive and behavioral deficitsassociated with parental alcohol use. Journal of theAmericanAcademy of Child and Adolescent Psychiatry,

36, 1177-1186.Wigglesworth, G. (1997). Children's individual approachesto the organization of narrative. JournalofChild Language,24 , 279-309.Wiig, E. H., & Secord, W. A. (1989). Test ofLanguageCompetence-Expandededitiom.San Antonio, TX- Psycho-logical Corporation.

Thome et al.: NarrativeAnalysis in 'DiagnosticDecision Making 473

7/30/2019 ergasia77 (2)

16/17

Wong, A. M.-Y. (2001). Who are you talkingabout?The de-velopment of reference in Cantonese-speakingchildren withandwithout specific language impairment.Unpublisheddoctoral dissertation, University of British Columbia.

Wong, A. M.-Y., Au, C. W. S. , & Stokes, S. F. (2004). Threemeasures of language production for Cantonese-speakingschool-age children in a story-retelling task- JournalofSpeech, Language,andHearingResearch,47, 1164-1178.

Wong, A. M.-Y., & Johnston, J. EL (2004). The developmentof discourse referencing in Cantonese-speaking children.Journalof ChildLanguage, 31 , 633-660.

Received July 27 , 2005Revision received December 23 , 2005Accepted July 13, 2006DOI: 10.1044/1092-4388(2007/032)Contact author: John C. Thorne, Department of Speech

and Hearing Sciences, University of Washington,1417 NRE. 42nd Street, Seattle, WA 98105-6246.E-mail: jct6@u.washington.edu.

7/30/2019 ergasia77 (2)

17/17

COPYRIGHT INFORMATION

TITLE: Exploring the Utility of Narrative Analysis in Diagnostic

Decision Maki

SOURCE: Journal of Speech, Language, and Hearing Research 50

no2 Ap 2007

PAGE(S): 459-74

(C) The American-Speech-Language-Hearing Association is the

publisher of this article and holder of the copyright. Further

reproduction of this article in violation of copyright is prohibited

without the consent of the publisher. To contact the publisher:

http://www.asha.org/