"The Frog Ate the Bug and Made his Mouth Sad": Narrative Competence in Children with Autism [+].

Lisa Capps [1]

Molly Losh [1,3]

Christopher Thurber [2]

This study compares the narrative abilities of 13 children with autism, 13 children with developmental delays, and 13 typically developing children matched on language ability. Although groups did not differ in their use of causal language or internal state terms, children with autism and children with developmental delays were less likely than typical children to identify the causes of characters' internal states. Rather, they tended simply to label emotions and explain actions. Children with autism and children with developmental delays also relied on a more restricted range of evaluative devices, which both convey point of view and maintain listener involvement. In addition, the narrative abilities of children with autism were linked to performance on measures of theory of mind and an index of conversational competence, whereas this was not the case among children with developmental delays. Findings are discussed in relation to the social, cognitive, and emotional underpinnings and consequences of narrativ e activity.

KEY WORDS: Autism; narrative; of mind; conversational competence.

Jerome Bruner described narrative as "among the earliest powers of mind to appear in the young child and among the most widely used forms of organizing human experience" (1990, p. 9). Through narrative, individuals imbue experience with meaning and impose order on otherwise disconnected events by sequencing them in time and rendering them from a particular point of view (e.g., Bruner, 1983, 1986, 1990; Berman & Slobin, 1994; Labov & Waletzky, 1968; Ochs & Capps, 1996; Polkinghome, 1988; Propp, 1968; Ricoeur, 1988; White, 1980).

Given its potential to illuminate not only linguistic skills but also cognitive and social knowledge, researchers have recently begun to analyze narrative practices as a means of investigating atypical development. This approach is particularly promising in relation to autism because of autistic individuals' social-communicative deficits in general (e.g., Fein, Pennington, Markowitz, Braverman, & Waterhouse, 1986), and pragmatic deficits in particular (e.g., Baron-Cohen, 1989; Frith, 1989; Tager-Flusberg, 1993). Analysis of narrative practices affords unique insights into autistic impairment in that narrative involves both pragmatic language skills and an appreciation of the role of mental states in predicting and explaining behavior. The use of narrative may prove useful both in diagnosis of high-functioning individuals and in interventions aimed at improving their social understanding and pragmatic skills.

There have been surprisingly few detailed investigations of narrative abilities in individuals with autism. Moreover, different investigators have examined narrative in different contexts. Loveland and colleagues (Loveland, McEvoy, Tunali, & Kelley, 1990), for example, asked children with autism to narrate a puppet show or video sketch they had previously viewed. In comparison to children with Down Syndrome matched on chronological and verbal mental age, children with autism tended to produce impoverished narratives, more frequently made grammatical errors, included bizarre or otherwise inappropriate information, neglected to mention central themes, and misinterpreted story events. Similarly, in an exploratory study in which four nonretarded adolescents with autism were read and then asked to retell stories involving deception, they tended to mention the deception but failed to organize their retelling around it (Bruner & Feldman, 1993). In both cases, the investigators interpreted their findings as evidence of pervasive difficulties in both taking listeners' needs into account and appreciating the function of narrative activity (see also Loveland & Tunali, 1993; Sigman & Capps, 1997).

In addition to these global narrative deficits, individuals with autism have also been shown to exhibit specific difficulty narrating stories that revolve around characters' mental states. Baron-Cohen, Leslie, & Frith (1985), for example, presented subjects with three sets of pictures, each of which could be arranged to tell a story. One set depicted a chain of events attributable to a physical cause ("mechanical"), another depicted a sequence of actions that could be described through reference to behaviors ("behavioral-descriptive"), and the third required reference to a character's intentions ("psychological-intentional"). In comparison to typically developing children and children with Down Syndrome, children with autism had considerable trouble ordering the psychological-intentional sequence, and did not use mental state terms, such as thinking and knowing, to narrate the story. In contrast, children with autism were as apt as comparison children in narrating the mechanical and behavioral-descriptive pi cture sets. These findings were interpreted as evidence of the difficulty using psychological concepts to construct causal frameworks in children with autism (see also Bruner & Feldman 1993). The results of this study are consistent with additional evidence that children with autism less often referred to mental states during the context of conversations with caregivers than did children with Down Syndrome (Tager-Flusberg, 1992).

Subsequent investigations (Capps, Kehres, & Sigman, 1998; Tager-Flusberg, 1995; Tager-Flusberg & Sullivan, 1995), however, have challenged the notion that children with autism are less inclined to reference mental states. In a study of an informal conversation, Capps et al. (1998) found that children with autism were as likely to use mental state terms as were nonautistic, comparison children with mental retardation matched on language ability, but that individuals with autism were less likely to offer narratives.

In a detailed study of the narratives generated using the wordless picture book, Frog, Where Are You? (Mayer, 1969), Tager-Flusberg (1995), identified no differences in the proportion of mental state terms used by children with autism, nonautistic children with mental retardation, and typically developing children matched on verbal mental age. With regard to linguistic structure, the narratives of children with autism were less grammatically complex than were those of comparison children. Surprisingly, in light of pragmatic language deficits, few group differences emerged in relation to the use of evaluative devices, which convey the narrators' point of view and maintain listener involvement. Although they were less likely to use sound effects, children with autism used character speech, emphatic stress, and other "audience hookers" as often as did children in comparison groups. Group differences did emerge with respect to story length and structure, in that children with autism told significantly shorter st ories than did comparison children, were less likely to construct stories with a resolution, and were more apt to introduce new characters in an ambiguous way than were typically developing children. In addition, there were no instances of causal attribution in any of the stories from the autistic group, while causal language occurred with equal frequency in the narratives of children in both comparison groups. This result was consistent with the previous finding that autistic children had significant difficulty explaining human action in a psychological framework within the context of a structured task designed to elicit such explanations (Tager-Flusberg & Sullivan, 1994).

Tager-Flusberg and Sullivan (1995) further explored causal attributions, mental state language, and theory of mind in a follow-up study by substituting a picture book deemed more likely to elicit mental state language, Frog on His Own (Mayer, 1973), for the one previously used. In addition, they asked subjects a series of probe questions about characters' emotions immediately after they told the story. Finally, they investigated the association between the narrative variables and performance on false belief tasks. The sample was larger and more linguistically competent than was true for the sample in the previous narrative study by Tager-Flusberg (1995). In this study, there were fewer group differences than had been true in previous studies. Groups did not differ in their use of mental state terms or causal statements. However, when probed, they were less accurate in labeling characters' emotions and …