Journal of Autism and Developmental Disorders

, Volume 43, Issue 7, pp 1732–1740

Brief Report: Conveying Subjective Experience in Conversation: Production of Mental State Terms and Personal Narratives in Individuals with High Functioning Autism


    • School of Communication Sciences and DisordersMcGill University
  • Jesse Burns
    • School of Communication Sciences and DisordersMcGill University
  • Aparna Nadig
    • School of Communication Sciences and DisordersMcGill University
Brief Report

DOI: 10.1007/s10803-012-1716-4

Cite this article as:
Bang, J., Burns, J. & Nadig, A. J Autism Dev Disord (2013) 43: 1732. doi:10.1007/s10803-012-1716-4


Mental state terms and personal narratives are conversational devices used to communicate subjective experience in conversation. Pre-adolescents with high-functioning autism (HFA, n = 20) were compared with language-matched typically-developing peers (TYP, n = 17) on production of mental state terms (i.e., perception, physiology, desire, emotion, cognition) and personal narratives (sequenced retelling of life events) during short conversations. HFA and TYP participants did not differ in global use of mental state terms, nor did they exhibit reduced production of cognitive terms in particular. Participants with HFA produced significantly fewer personal narratives. They also produced a smaller proportion of their mental state terms during personal narratives. These findings underscore the importance of assessing and developing qualitative aspects of conversation in highly verbal individuals with autism.


High-functioning autismConversationPersonal narrativeMental state terms


“She drives me nuts with her music, and she thinks I’m supposed to get used to it.”- Child talking about his sister playing the flute.

Evidence that children have acquired an understanding of their own and others’ psychological states can be gleaned from their spoken language. Children’s use of terms such as think, pretend and know generally begins before the third year of life and has been taken to reflect a rudimentary understanding of psychological states, also known as theory of mind (see Bretherton and Beeghly 1982; Bretherton et al. 1981; Shatz et al. 1983, among others). These psychological states have been proposed to be scaled, where the understanding of cognitive mental states is scaffolded by the earlier understanding of desires. This pattern has been observed similarly across cultures (Wellman et al. 2006). Likewise, the production of mental state terms in a maturing child’s lexicon has been documented to follow a developmental pattern with references to desire and emotion states appearing before those to cognition (Bartsch and Wellman 1995; Bretherton and Beeghly 1982; Fenson et al. 1994). Cross-linguistically, cognitive terms are also found to appear latest in the expressive vocabularies of young children (Pascual et al. 2008; Poulin-Dubois et al. 2009; Tardif & Wellman 2000). Because the acquisition of mental state terms involves a developmental progression, the greatest differences may be observed earlier in development; therefore we focus on speech produced by children and adolescents in this paper.

Following the popular account of theory of mind abilities as a primary impairment in autism (Baron-Cohen et al. 1985; Baron-Cohen 2000; cf. Frith and Happe 1994; Klin et al. 1992; Mottron et al. 2006 for discussion of the limitations of theory of mind as an account of autism), a natural corollary is that the global use of mental state terms, including perception, physiology, desire, emotion, and cognition terms, may be reduced in individuals with autism. However, studies examining spontaneous speech or story-telling tasks have shown that global use of mental state terms does not differ between children with autism and language-matched developmentally delayed (DD) and/or typically-developing (TYP) samples (Capps et al. 1998; Capps et al. 2000; Norbury and Bishop 2003; Norbury 2011). Furthermore, Müller and Schuler (2006) reported an increased, rather than decreased, use of emotion terms by children with autism during interactions with family members at home.

Offering a more precise hypothesis based on the theory of mind account, Tager-Flusberg (1992) proposed that terms referring to cognition specifically are impaired in autism, and found evidence of this in a comparison between young children with autism and those with Down syndrome. Yet, other studies with pre-adolescents found no differences between groups in the number of cognition terms elicited during story-telling tasks (Capps et al. 2000; Norbury 2011; Tager-Flusberg 1995). Although additional studies have not reported a reduction in cognitive state terms per se, they have found positive relationships between the production of cognition terms and theory of mind performance in children with autism (Capps et al. 2000, Tager-Flusberg 2003; Tager-Flusberg and Sullivan 1995; Ziatas et al. 1998), suggesting that cognition terms may provide an index of theory of mind abilities in this population during the relevant developmental window.

Many of these studies focused on young language-delayed children with autism and employed specific prompts or story-telling tasks to elicit mental state terms, rather than observing how they are used spontaneously. In this study, we examine how pre-adolescent children with autism and without language impairment, who should have acquired the full developmental range of mental state terms, use them in spontaneous conversation to speak about their own experiences. Unlike tasks designed to elicit mental state language about story characters, the nature of spontaneous conversation may provide greater insight into the tendency of individuals with autism to refer to mental states in daily life (Müller and Schuler 2006). We use “mental state terms” as a general term referring to the full range of this trajectory (i.e., perception, physiology, desire, emotion, cognition), that is, words or phrases that refer to any kind of internal state. Only one previous study examined the usage of mental state terms by pre-adolescents with high-functioning autism (HFA) during conversational interaction (Capps et al. 1998). The authors reported no differences when the HFA group was compared to a language-matched DD group with respect to “perception, emotion, desire and mental state terms”; however, details of the data were not provided and it was not clear if cognition terms were examined separately.

Beyond referencing internal states, conversational partners often spontaneously share narratives of their personal experiences (henceforth referred to as personal narratives), or sequenced retelling of autobiographical events that provide some indication of the speaker’s reflecting upon and evaluating his or her experience (Labov and Waletzky 1967). It is important to distinguish these personal narratives of one’s own experience from story narratives, which are elicited by asking participants to recount events about characters depicted in a scene or previously-encountered story (Begeer et al. 2010; Klin 2000; Klin and Jones 2006; Norbury and Bishop 2003; Tager-Flusberg 1995). We focus on the former, as there is evidence that individuals with ASD perform differently in situations of direct interaction versus describing story characters they are not directly engaged with (Begeer et al. 2010). Personal narratives are the reformulations of an event of personal interest to the speaker from the speaker’s own perspective (Labov and Waletzky 1967; Labov 1997). They serve to convey the speaker’s subjective experience of the event and are more evocative than objective descriptions. For instance, when a speaker describes a video game during conversation he could focus on facts and details, as exemplified by one of our participants: “from the beginning, you start at the beginning with Planet Alden…go up the stair, across the bridge, and you go into a room that forms to the left and to the right.” In contrast, as shown in this excerpt from another participant, the quality of conversation is quite different when the speaker describes a video game by using a personal narrative to evaluate his experience: “yeah one thing that happened, it [the character] was so angry it stopped breathing and then it turned purple.” This more subjective account of the experience via a personal narrative demonstrates the role of narrative as a vehicle of meaning and reflection, in which the child can practice, for instance, understanding another’s internal state through discourse (Bruner 1990).

The examination of personal narrative in autism, which overlaps with investigations of autobiographical memory, began with Capps et al. (1998). These authors proposed that, given more general difficulties in organizing events in narrative form (Loveland and Tunali 1991; Tager-Flusberg 1995), and decreased cultural knowledge of what people are likely to do or think in different situations (which can be gained through narrative constructions and understanding; Bruner 1990), individuals with autism would relate fewer spontaneous narratives of their personal experience. Their prediction held, and unlike the conflicting results with respect to cognition terms, results from studies of personal narrative have consistently shown that individuals with autism provide fewer and shorter personal narratives when compared with TYP and DD groups (Capps et al. 1998; Crane et al. 2010; Goldman 2008; Losh and Capps 2006). The aforementioned study by Capps et al. (1998) was also the only study to examine the spontaneous production of personal narratives during conversation. These authors noted that not only did pre-adolescents with HFA recount fewer personal narratives than the language-matched DD group, but the DD group’s personal narratives also included more “psychologically meaningful, personal events.” In the same vein of examining qualitative differences among personal narratives, Crane et al. (2010) reported that the adults with autism in their study provided fewer memories of specific, as opposed to habitual, events than a TYP comparison group. Specific autobiographical events may be more likely to relate psychologically meaningful experiences than general or habitual events.

We propose that the spontaneous production of personal narratives is a more linguistically sophisticated way to share subjective experience than simply labeling internal states with mental state terms. The former may be a more informative measure of a speaker’s understanding of internal states and their causes in individuals with autism who have age-appropriate formal language. As seen above, the sheer frequency of mental state term production may not reveal group differences when groups are well-matched on language. However, qualitative aspects of mental state term use may differ, including the context in which they appear. For example, a mental state term within a personal narrative may provide meaningful insight into personal experience, in contrast to a mental state term outside of a personal narrative, which may label an internal state without further elaboration.

This study compares the production of mental state terms and personal narratives in the conversations of pre-adolescents with HFA and their TYP peers matched on language, gender, and nonverbal IQ. First, given the previous literature on speakers with autism (Capps et al. 1998; Capps et al. 2000; Norbury 2011; Tager-Flusberg 1995), we predicted no difference in the global use of mental state terms. Based on the view that cognition terms should be specifically impaired following a theory of mind account of autism (Tager-Flusberg 1992; Tager-Flusberg 2003), these were examined separately. As cognition terms have not been analyzed separately in pre-adolescents with HFA during conversation, we were neutral in our expectation of whether these would be reduced in participants with HFA. Second, following the literature on personal narratives (Capps et al. 1998; Goldman 2008; Losh and Capps 2006), we hypothesized that the HFA group would produce fewer personal narratives than the TYP group. We also examined the use of specific narratives that recount a distinct event in one’s personal history and predicted that the HFA group would produce fewer specific narratives than the TYP group (Crane et al. 2010). Third, in addition to the quantitative comparison of mental state terms in our first hypothesis, we also examined the qualitative use of mental state terms. We hypothesized that in conversation, the use of mental state terms during the context of personal narratives may contribute to the quality of shared meaning (Bruner 1990; Capps et al. 1998; Losh and Capps 2006), whereas mental state terms produced outside of personal narratives may not convey the same quality. Given that the conversational contributions of speakers with HFA are often reduced in shared or reciprocal quality (Capps et al. 1998; García-Pérez et al. 2007; Nadig et al. 2010), we predicted that, for speakers with HFA who produce mental state terms, a smaller proportion of these terms would be used within personal narratives than is the case for TYP speakers.



The study sample included 20 HFA and 17 TYP pre-adolescent children (see Table 1 for participant characteristics). Participants were recruited via a university health center database and through community outreach efforts in Northern California as part of a larger study approved by a university ethics board. Informed consent was obtained from all parents and participants.
Table 1

Participant characteristics


HFA (n = 20), Mean (SD)

TYP (n = 17), Mean (SD)


Chronological age

11;0 (1;11)

10;10 (1;5)


Language (CELF-IV Core)

107.1 (14)

112.7 (12)


Performance IQ (WASI)

106.5 (17)

112.5 (14)



90 % male

76 % male


VABS II socialization

61.0 (7)

101.3 (8)


VABS daily living

67.7 (8)

101.7 (9)



26.4 (6)

2.3 (3)


There were no significant differences between groups with respect to CA, language level, Performance IQ, or gender. The p value for gender was calculated with a Chi square test, and the remainder were calculated from t tests

Groups did not significantly differ on chronological age, gender ratio, language level, or Performance IQ. All participants obtained a standard score of 85 or higher on the Clinical Evaluation of Language Fundamentals—Fourth Edition (CELF-4; Semel et al. 2003), which indicates average or above-average formal language ability. The Block Design and Matrix Reasoning subtests of the Wechsler Abbreviated Scales of Intelligence (WASI; Weschler 1999) were used to assess Performance IQ. Information on participants’ adaptive skills and social functioning was obtained via parental report from the Daily Living Skills and Socialization subscales of the Vineland Adaptive Behaviour Scales-II (VABS; Sparrow et al. 2005).

In an initial screening, individuals were excluded if they had a metabolic disorder known to cause autistic features, major medical issues, or a physical disability. Participants in the HFA group met DSM-IV diagnostic criteria for Autistic Disorder. Diagnoses were confirmed with the Autism Diagnostic Observation Schedule-Module 3 (ADOS-3; Lord et al. 1999) and the Social Communication Questionnaire-Lifetime (SCQ; Rutter et al. 2003). Eight of the 20 participants had a history of early language delay defined as parental report of first word onset after 2 years of age or phrase speech after 3 years of age. Children with complex medical histories, psychiatric conditions, developmental delay, first- or second-degree relatives with autism, or autistic symptoms by parent report (SCQ) were excluded from the TYP group.


To elicit conversational speech each participant was asked to discuss a special interest or favorite hobby, as well as a generic topic (i.e., siblings, pets, or friends) with an adult research assistant who was blind to group status. The special interest or favorite hobbies initiated by participants from both groups included common interests such as sports, reading books, video or computer games, and animals (e.g., lizards or bugs). Less often participants in both the HFA and TYP groups initiated idiosyncratic topics (e.g., canning fruit or 18th century military history). Conversations, approximately 2–6 min in duration, were videotaped and transcribed. These conversations took place at the end of the participant’s 2nd or 3rd visit to the lab, with each visit lasting approximately 2 h. The same experimenter interacted with the participants across the visits, making the experimental conversation partner somewhat familiar to the participant. Variables reflecting the verbal exchange were calculated and collapsed over both topics (analysis of other elements of these conversations was presented in Nadig et al. 2010). HFA and TYP speakers did not differ in terms of the overall number of questions or comments addressed to participants by adult partners, the total number of participant utterances, the ratio of participant to adult utterances, mean length of utterance, or lexical diversity.

Conversations were transcribed and then coded for participants’ use of mental state terms and personal narratives. Mental state terms were defined and coded into the categories of perception,physiology, desire,emotion, or cognition (see Table 2 for a comprehensive list of terms observed in our sample and the category they were assigned to), according to previous taxonomies (Bretherton and Beeghly 1982; Shatz et al. 1983; Tager-Flusberg 1992). Part of speech (i.e., verb, noun, adverb, or adjective) was also identified for each term. Conversational context was used to judge if terms or phrases referred to internal states. Phrases coded as referring to mental states included those that could be clearly paraphrased into a mental state term (e.g., drive me crazy and gets on my nerves could be paraphrased as annoy, an emotion term). Following conservative conventions employed in the aforementioned literature, idiomatic or conversational uses of mental state terms (e.g., “Know what?”) were not included in our analyses. Terms or phrases that were used by the experimenter in the previous utterance were also not included, as their use may have resulted from repetition of the experimenter’s words. Finally, words or phrases were not included if they referred to personality traits (e.g., shy) or if they were used to express moral or value-laden judgments of observable behavior (e.g., stupid was not coded in “…my sister likes torun around the house acting stupid.”).
Table 2

Exhaustive list of mental state terms or phrases coded from participant conversations






hear; heard; see; saw; seeing

messed up; sick

choose; concentration; decide; decides; distracts; expect; figure out; forget; forgets; forgot; ideas; know; knew; knowledges; mean; meant; memory; pretend; pretending; remember; think; thinks; thought; unsuspecting

afraid; angry; annoy; annoys; bugging; drives me nuts; enjoy; favorite; frightened; fun; funny; gets on my nerves; getting tired; happy; hate; hates; like; likes; liked; love; loves; mad; obsessed; pester; pesters; scare; scared; surprises

hopefully; hoping; try; tries; trying; tried; want; wanna; wants; wish

A personal narrative was defined as discourse describing a temporally-ordered event or sequence of events (Capps et al. 1998; Labov and Waletzky 1967). Temporal order may be implied by the provision of an outcome or effect for the described event. Temporal order was often identified by the use of terms such as “and then after”, “well first we…”, or “before.” Descriptions of how to play a game did not qualify as an event on their own. Moreover, for our coding, the participant had to be personally involved in the event (e.g., it was evident that the participant was present at the time of the event, but it was not necessary that the event happened to the participant; “this one time I saw my friend…”), and the personal narrative had to contain a minimum of two utterances. Personal narratives were further identified as portraying habitual versus specific events (Crane et al. 2010). Habitual events referred to a repeated or routine event (e.g., “…I ride my bike and he tries chasing me on his bike.”), whereas specific events made reference to a particular point in time (e.g., “…last year we won one game and then everybody got to go the playoffs…”). Please refer to Appendix for examples of personal narratives.

All conversations were coded individually by the first two authors for use of mental state terms and personal narratives. Agreement between the two authors was 93.56 %. Discrepancies were discussed and resolved by consensus. In addition, a rater blind to group membership and to the study hypotheses coded 27 % of the conversations. Agreement between the blind rater and the authors’ consensus codes was 88.89 % for mental state terms. For personal narratives the first two authors established a protocol to code for the presence of personal narratives as well as differentiation between specific and habitual ones; consensus codes were obtained for each conversation. A different rater who was blind to group membership and to the study hypotheses coded 27 % of the conversations for personal narratives. The inter-coder agreement rate for identifying personal narratives was 100 % (i.e., of 10 conversations coded, the same 9 personal narratives were identified by both the blind rater and the authors’ consensus). There was 88.89 % agreement on the type of personal narrative (i.e., specific or habitual; of the 9 personal narratives coded, 8 were classified consistently as specific or habitual). Few habitual narratives were produced by participants in any group, therefore analyses on habitual narratives were not conducted.


Nonparametric tests were used given the non-normal distribution of the data. However, means and ranges are also provided to facilitate interpretation. The Mann–Whitney U test was used for group comparisons on continuous variables, and categorical variables were analyzed with Chi square tests. Because multiple comparisons were carried out on the same sample, the Bonferroni-adjusted alpha level was set at .005 (.05/10 comparisons). Effect sizes are reported with Pearson’s correlation coefficient, r, for Mann–Whitney U tests and with Cramer’s V for Chi square tests. For both Cramer’s V (degree of freedom = 1) and r, values of .10 represent small effects, .30 medium effects, and .50 large effects (Cohen 1988; Cohen 1992). Although gender proportion was not significantly different between groups, a larger proportion of the HFA group was male (90 % HFA vs. 79 % TYP). To examine if gender influenced the group differences, all analyses were re-conducted using the male participants only. The pattern of results was identical to those with the full group; therefore, the analyses reported below reflect the full sample.

Mental State Terms

Our first hypothesis was that there would be no difference between groups in a summary score of total mental state (MS) terms, and our hypothesis with regard to cognition terms specifically was neutral. Results for all variables coded are provided in Table 3. As predicted, there was no significant difference between groups with respect to the total number of MS terms produced. We divided this figure by the number of utterances spoken by each participant to obtain the proportion of their utterances containing a MS term, which also did not differ between groups. Examination of the parts of speech used by speakers to express mental states showed that the groups were similar in using verbs approximately 80 % of the time (U = 195.50; p = .44, r = .16), and adjectives approximately 10 % of the time (U = 178.50; p = .78, r = .00), whereas nouns (U = 146.50; p = .29, r = − .17) and adverbs (U = 180.00; p = .28, r = .18) were used less than 5 % of the time. For each individual category of perception, physiology, desire, and emotion, we found no significant difference between groups and relatively small effect sizes. Furthermore, the production of cognition terms did not differ significantly between the HFA and TYP groups.
Table 3

Frequency of mental state and personal narrative production per participant for HFA and TYP groups


HFA (n = 20)


TYP (n = 17)


Mean (Range)


Mean (Range)






.20 (0–1)


.12 (0–2)






.10 (0–1)








1.40 (0–5)


1.18 (0–3)






3.40 (0–12)


3.76 (0–13)






1.85 (0–5)


2.71 (0–7)





Total mental state (MS) terms

6.95 (0–18)


7.76 (3–18)





Proportion of MS per utterance

.18 (0–.46)


.23 (.07–.47)





Personal narratives (PN)

.33 (0–1)


1.48 (0–4)





Specific PN1

.17 (0–1)


.88 (0–3)





Proportion of MS terms during PN2

.03 (0–.27)


.14 (0–.38)





1Only calculated for participants who produced personal narratives

2Only calculated for participants who produced mental state terms

* Significant at adjusted alpha level of .005

Personal Narratives

Secondly, we predicted that there would be fewer personal narratives produced by the HFA group than the TYP group, and also that they would produce fewer specific narratives (Crane et al. 2010). As can be seen in Table 3, significantly fewer personal narratives were produced by the HFA group than the TYP group, with a large effect size. Viewed with respect to the number of individual participants, significantly fewer participants in the HFA group (n = 6) produced a personal narrative than participants in the TYP group (n = 15), with a medium effect size, χ2(1) = 12.697, p < .005, Cramer’s V = .46. Within the context of lower overall personal narrative production, significantly fewer specific narratives were produced by the HFA group than the TYP group as well, with a medium effect size. Again, with respect to the number of individual participants, significantly fewer members of the HFA group (n = 3) produced a specific narrative in comparison to members of the TYP group (n = 10), with a large effect size, χ2(1) = 7.74, p < .005, Cramer’s V = .59.

Mental State Term Production During Personal Narratives

Our third hypothesis was that a smaller proportion of MS terms would be used during personal narratives by HFA participants than by TYP participants. For this analysis, only participants who produced MS terms were included (ASD, n = 18, TYP, n = 17). Per participant, the proportion of MS terms used during personal narratives relative to the total number of MS terms produced was calculated. As shown in Table 3, HFA participants who produced MS terms used a smaller proportion of them during personal narratives than did TYP participants, with a medium effect size, though this comparison failed to reach significance with the Bonferroni adjustment.


Our examination of the spontaneous production of MS terms and personal narratives provides a comprehensive view of how verbally-fluent individuals with autism convey subjective experience through conversation. No group differences were found in the global production of MS terms, nor was a reduction found in the production of cognition terms specifically. With respect to personal narratives, our HFA participants produced significantly fewer personal narratives than their TYP peers, and among those who did produce one or more personal narratives, fewer of these described a specific or distinct personal event. Finally, although it did not reach significance, our qualitative analysis of MS term use revealed that HFA participants produced a smaller proportion of MS terms during personal narratives than TYP participants, with a medium effect size.

Our finding that language-matched groups of HFA and TYP pre-adolescents do not differ in global MS term production during conversation is consistent with Capps et al. (1998). This is the first study to demonstrate that, in conversation, this subpopulation does not differ from a matched TYP comparison group in the production of cognition terms. This is consistent with previous literature examining the production of MS terms, and terms of cognition, in pre-adolescents during story-telling tasks (Capps et al. 2000; Norbury 2011; Tager-Flusberg 1995). This finding contrasts with Tager-Flusberg’s (1992) report of decreased use of cognition terms in autism, which is likely due to the age of our participants and their high verbal abilities. The 6 children with autism in Tager-Flusberg’s study were young at the start of data collection (3;4–7;7) and had delayed language development. As discussed above, cognition terms are the last to appear along the developmental sequence of MS terms (Bartsch and Wellman 1995; see also Tardif and Wellman 2000 for cross-linguistic studies); therefore, given more mature linguistic abilities, differences in cognition term production may no longer be evident. This interpretation is corroborated by Norbury (2011), who demonstrated that during story-telling tasks highly verbal pre-adolescents with HFA produced cognition terms at similar rates to language-matched TYP peers, whereas a language-impaired group with autism produced fewer of them. It should be noted that our conversations were only a few minutes in duration and that results pertain only to highly verbal individuals with autism; future research should investigate the production of MS terms over longer stretches of speech and how interactions may differ if the conversational partner was someone the participants knew well.

In contrast to similarities observed between groups on the quantitative measure of frequency of MS term production, the most noteworthy results from this study lie in more complex features of how speakers express subjectivity in conversation. Our results are consistent with the prior literature documenting that individuals with autism produce fewer spontaneous personal narratives (Capps et al. 1998; Goldman 2008; Losh and Capps 2006), as well as replicating a previous finding of fewer narratives describing distinct, specific life events (Crane et al. 2010). Personal narratives represent a more complex and sophisticated strategy to convey one’s subjective experience in conversation, and it is this level of complexity that appears to be reduced in high-functioning individuals with autism, despite their age-appropriate language ability. To complement studies focused on personal experiences, other measures such as the social attribution task, which depicts a “social” scene using geometric figures (Klin 2000; Klin and Jones 2006), can help us to gain a broader understanding of how individuals with autism use narrative to interpret social situations and convey meaning. We also found a non-significant trend (at the Bonferroni-adjusted alpha level of .005) for speakers with HFA to use a smaller proportion of their MS terms during personal narratives. The use of MS terms primarily outside of personal narratives may decrease the quality of shared meaning conveyed by MS terms, since they may simply be labeling an internal state whereas MS terms produced during personal narratives often play a role in elaborating or reflecting on a subjective experience.

In conclusion, our results indicate that pre-adolescents with HFA who have unimpaired formal language produce the same number of MS terms, and cognition terms specifically, in conversation as their TYP peers. However, they are less likely to use the more complex conversational device of personal narratives. This may make their conversational contributions appear less reciprocal to a listener. This pattern of findings mirrors that of other studies examining the quality of conversation in autism (García-Pérez et al. 2007; Nadig et al. 2010), where differences in conversation between highly verbal individuals with autism and language-matched groups are absent or less pronounced in quantitative measures (e.g., MS production), yet significant and consequential differences in qualitative aspects of conversation are observed. Conversation plays an important role in forming reciprocal relationships, which remains an area of difficulty for individuals with autism into adulthood (Howlin et al. 2004). The current findings add to our understanding of how subjective experience is conveyed in conversation, and which aspects require support in highly verbal individuals with autism. Future work should investigate how the production of MS terms and personal narratives, as well as other modes of expressing subjective experience, affect listener perceptions, and how they can in turn be used to foster reciprocal relationships.


This work was supported by a NRSA postdoctoral fellowship to Nadig, NIDCD F32‐DC007297. The mental state term and personal narrative data presented here builds on Nadig et al. 2010, which presented general descriptions of the verbal content of the same conversations, along with eye-tracking data obtained during these conversations. We would like to thank Sofia Sergyeyenko, Anjali Mulligan, and Sarah Justine Leduc-Villeneuve for assistance with reliability coding, Iris Lee for work on the original transcriptions, and Giacomo Vivanti, Pamela Richardson, Xiomara Watkins, Alaina White, and Nancy Vu for assistance with data collection. Finally, we extend our gratitude to the children and families who participated in this research.

Conflict of interest

The authors declare that they have no conflict of interest.

Copyright information

© Springer Science+Business Media New York 2012