A New Computerised Advanced Theory of Mind Measure for Children with Asperger Syndrome: The ATOMIC (original) (raw)

A New Computerised Advanced Theory of Mind Measure for Children with Asperger Syndrome: The ATOMIC

Renae B. Beaumont ⋅\cdot Kate Sofronoff

Published online: 13 July 2007
© Springer Science+Business Media, LLC 2007

Abstract

This study examined the ability of children with Asperger Syndrome (AS) to attribute mental states to characters in a new computerised, advanced theory of mind measure: The Animated Theory of Mind Inventory for Children (ATOMIC). Results showed that children with AS matched on IQ, verbal comprehension, age and gender performed equivalently on central coherence questions, but more poorly on the theory of mind questions compared with controls. A significant relationship was found between performance on ATOMIC and accuracy of mental state explanations provided on (Happé’s, Journal of Autism and Developmental Disorders, 24, 129-154, 1994) Strange Stories Task, supporting the validity of the new measure. Limitations of the study and suggestions for future research are discussed.

Keywords Asperger syndrome ⋅\cdot Autism spectrum disorders ⋅\cdot Theory of mind ⋅\cdot Central coherence

Introduction

Asperger Syndrome (AS) is a neuro-developmental disorder characterised by social impairments, circumscribed interests and/or a rigid adherence to routines (APA, 2000). Although qualitatively similar to autism, AS is distinguished by the absence of cognitive or language delays. This is the basis for the argument that AS and autism exist on a single spectrum, differing only in terms of the severity

[1]of autistic symptomatology and associated language and learning difficulties (Lawson, Baron-Cohen, & Wheelwright, 2004).

Theory of Mind and Autism Spectrum Disorders
Over the past 20 years, extensive research has sought to determine the core cognitive deficit that best accounts for the defining features of autism spectrum disorders (ASDs). One of the most productive lines of research has explored the hypothesis that individuals with ASDs have difficulty imputing mental states such as beliefs, intentions and desires to themselves and others (Leslie, 1987). This proposal, known as the ‘theory of mind’ deficit account of ASDs, provides a compelling explanation for the core social, communicative and imaginative deficits that characterise this spectrum of disorders (Baron-Cohen, 1995).

Substantial empirical evidence supports the theory of mind deficit account of ASDs, especially in children with severe autistic symptomatology and/or associated intellectual impairment (e.g. Baron-Cohen, Leslie, & Frith, 1985; Phillips, Baron-Cohen, & Rutter, 1998). However, theory of mind impairments in individuals with highfunctioning autism (HFA) or AS appear to be more subtle. Several investigators have shown that these individuals display intact theory of mind performance on relatively simple tasks, such as classic first- and second-order falsebelief tasks (e.g. Bowler, 1992; Dahlgren & Trilingsgaard, 1996). Thus, in recent years, researchers have developed more challenging, naturalistic measures to examine theory of mind abilities in individuals with HFA and AS. These tasks have attempted to circumvent the possibility of highfunctioning individuals with ASDs deriving solutions to test items through non-theory of mind, verbally mediated compensatory strategies.

1. R. B. Beaumont ⋅\cdot K. Sofronoff ( ⊠\boxtimes )
   School of Psychology, The University of Queensland, Brisbane, QLD, Australia
   e-mail: kate@psy.uq.edu.au ↩︎

Several studies have successfully demonstrated mindreading difficulties in individuals with HFA and AS using more advanced theory of mind measures. For example, individuals with HFA and/or AS have been shown to display difficulties providing mental state justifications for story characters’ non-literal utterances (Happé, 1994; Jolliffe & Baron-Cohen, 1999), in addition to demonstrating impairments in inferring people’s emotions from pictures of their eyes (Baron-Cohen, Jolliffe, Mortimore, & Robertson, 1997; Baron-Cohen, Wheelwright, Hill, Raste, & Plumb, 2001) or their voice tone (Rutherford, Baron-Cohen, & Wheelwright, 2002). Individuals with HFA or AS have also shown difficulties in recognising faux pas (Bar-on-Cohen, O’Riordan, Stone, Jones, & Plaisted, 1999), and in providing mental-state explanations for the movement of geometric shapes in cartoon animations (Castelli, Frith, Happé, & Frith, 2002; Klin, 2000). Although these tasks are arguably more valid measures of theory of mind than simplistic false-belief tasks, they are still characterised by limitations. These include the presentation of static stimuli (Castelli et al., 2002; Klin, 2000 are notable exceptions) in a single modality (e.g. visual and auditory), which seriously threatens the ecological validity of the measures.

In an attempt to create a measure that more closely approximates the on-line processing demands imposed in everyday social situations, Heavey, Phillips, Baron-Cohen, and Rutter (2000) created the Awkward Moments Test, in which television commercials were presented as stimuli to assess the theory of mind abilities of adults with HFA or AS. Results showed that, after differences in the comprehension skills between the autistic and control participants were controlled for, the participants with HFA or AS were more impaired relative to controls in their ability to infer the characters’ emotions, and in their ability to provide appropriate explanations for the intentions behind characters’ actions. Similar results were found by Roeyers, Buysse, Ponnet, and Pichal (2001), who demonstrated that adults with pervasive developmental disorders (PDDs) were impaired relative to age- and IQ-matched controls at inferring the mental states of strangers engaged in a videotaped conversation.

Although Heavey et al.'s (2000) and Roeyers et al.'s (2001) advanced theory of mind measures are among the most successful to date in approximating the demands of everyday social functioning, they too have weaknesses. First, no clear evidence has been found to support the validity of either task as an advanced theory of mind measure. Second, neither measure included a scale to measure the potential contribution that weak central coherence made to the poor performance of the participants with ASDs on the theory of mind questions. Finally, the majority of advanced theory of mind measures, including those developed by Heavey et al. (2000) and Roeyers et al.
(2001) have been designed for adults. To our knowledge, there are no multimodal advanced theory of mind measures currently available for children. Such assessment tools are clearly needed to enable the early identification of mild to moderate theory of mind deficits in children, and to inform the development and evaluation of interventions designed to treat these impairments.

Weak Central Coherence and Autism Spectrum Disorders

Despite its success in explaining the triad of social, communicative and imaginative impairments that characterise ASDs, the theory of mind deficit account fails to explain the non-triadic features of the disorder, including spiky IQ profiles, circumscribed interests and pre-occupations with parts of objects (Hoy, Hatton, & Hare, 2004). Frith (1989) and Frith and Happé (1994) proposed that these aspects of ASDs are best understood as manifestations of a qualitatively distinct processing style in individuals with ASDweak central coherence. Frith (1989) conceptualised central coherence as the spontaneous tendency of typically developing individuals to integrate stimuli into coherent, meaningful wholes. This feature of information processing is believed to be disturbed in ASDs, resulting in a failure to process information in context (Happé, 1999).

The Weak Central Coherence account of ASDs is supported by studies in both the visual and linguistic domains. For example, Shah and Frith (1983) and Jolliffe and BaronCohen (1997) demonstrated that participants with ASDs were faster than matched controls on the Embedded Figures Task, in which simple shapes have to be located within the context of more complex figures. Individuals with ASDs have also been shown to perform faster than controls on the block design task in which a pattern has to be constructed from painted blocks (Hoy et al., 2004; Shah & Frith, 1983). When the pattern is pre-segmented, controls perform as well as individuals with ASDs, suggesting that individuals with ASDs ‘pre-segment’ the design in the standard task, focusing on the constituent parts rather than the whole. With respect to linguistic tasks, a well-replicated finding is that people with ASDs fail to use the context provided by a sentence to determine the correct pronunciation of a homograph (e.g. ‘In her eyes was a large tear’ versus, ‘In her dress was a large tear’.) (Frith & Snowling, 1983; Hoy et al., 2004; Jolliffe & Baron-Cohen, 1999).

However, a number of studies have failed to support the weak central coherence account of ASDs, particularly with respect to the claim that individuals with ASDs are impaired at processing stimuli at the global or ‘gestalt’ level. For example, several investigations using the Navon hierarchical figures paradigm have shown that when instructed

to do so, participants with ASDs are as proficient as controls at identifying targets appearing at the global level of contextually embedded letter stimuli (e.g. Mottron, Burack, Stauder, & Robaey, 1999; Plaisted, Swettenham, & Rees, 1999). Recent investigations employing more sophisticated methodologies than past research have also failed to replicate the initial findings of weak central coherence in ASD populations. For example, Brian and Bryson (1996) and Ozonoff, Pennington, and Rogers (1991) found no differences in the performance of control participants and participants with a diagnosis of PDD or AS on the Embedded Figures Test and Block Design Task.

Putative explanations for the lack of consensus in research findings are subtle differences in the target populations sampled (e.g. children versus adults; participants with autism versus AS versus PDDs), and the methodologies used. Furthermore, much of the research examining central coherence deficits in ASD populations has limitations, including an inadequate definition of the ‘central coherence’ construct, a failure to guard against ceiling effects and a focus on central coherence at the perceptual rather than conceptual level (i.e. processing stimuli for meaning).

Jolliffe and Baron-Cohen (2001) are among the few investigators who have examined coherence abilities in high-functioning individuals with ASDs at a conceptual level, with their research yielding mixed results. Although the participants with ASDs were impaired relative to controls at integrating drawings of objects into coherent scenes, identifying scenes depicted in line drawings, and spontaneously identifying objects that were incongruent with the scenes, the majority of participants with ASDs provided coherent descriptions of the scenes, did not pay preferential attention to local details in their descriptions, and were equivalent to controls in their speed of locating a named object that was incongruent with the surrounding context.

In an attempt to account for the equivocal findings regarding central coherence abilities in individuals with ASDs, revisions to the original Weak Central Coherence theory have been proposed. First, it has been suggested that persons with ASDs may simply exhibit a non-conscious processing preference for local information, with no concomitant global processing deficit (the ‘enhanced perceptual processing hypothesis’; Mottron & Burack, 2001). Although some experimental findings support this proposal (e.g. Mottron, Peretz, & Menard, 2000; Plaisted et al., 1999) others plead against it (e.g. Jolliffe & Baron-Cohen, 2001). To account for this discrepancy in findings, Jolliffe and Baron-Cohen (1999) proposed that the capacity of individuals with ASD to appreciate the ‘gestalt’ is only revealed when they are instructed to process information in a coherent manner, or task demands clearly require them to do so. Evidence for this notion comes from Snowling and

Frith (1986), who found that individuals with autism who previously failed to pronounce homographs in context were capable of doing so after being given explicit instructions. Jolliffe and Baron-Cohen (2001) also suggested that the number of items of information that need to be processed simultaneously (i.e. working memory load) may influence whether individuals with ASDs process information coherently. This proposal is consistent with the ‘executive functioning deficit’ account of ASDs, which purports that higher-order cognitive abilities such as working memory, planning and response-shifting are impaired in individuals on the autistic spectrum.

The Present Study

The aim of the present study was to develop a computerised, naturalistic, multimodal measure to examine the theory of mind abilities of children with AS. The Animated Theory of Mind Inventory for Children (ATOMIC) was designed to improve on a limitation of past tests by including a scale to assess the contribution that central coherence deficits make to children’s theory of mind task performance. Furthermore, the test comprised questions about characters’ cognitions and complex emotions, and included items displaying a variety of child, adolescent and adult themes. No existing theory of mind measures appear to have examined whether individuals with AS are differentially impaired at inferring cognitions relative to complex emotions, and/or whether they are better at attributing mental states to others in personally relevant situations. A subsidiary aim of the study was to empirically examine whether children with AS displayed global processing deficits on a configural task, and whether their propensity to process information in context varied as a function of the number of items of information that needed to be integrated simultaneously. As computers constitute a special interest for many children with AS (Attwood, 1998), utilising a computerised presentation format was intended to enhance participants’ enjoyment of, and performance on the task, in addition to enabling automated scoring and recording of response times.

Based on previous research (e.g. Heavey et al., 2000; Roeyers et al., 2001), it was hypothesised that the children with AS would perform more poorly than matched controls on the theory of mind questions of the ATOMIC. It was also predicted that the children with AS would take longer to answer these questions relative to controls, as they were expected to try to ‘hack-out’ answers to the questions via indirect cognitive routes rather than through the intuitive attribution of mental states (see Bowler, 1992). As research suggests that individuals with ASDs experience difficulties attributing both complex emotions and cognitions to others (e.g. Happé, 1994; Heavey et al., 2000), participants with

AS were expected to be equally impaired on the cognition and emotion theory of mind questions relative to the control participants. However, the participants with AS were predicted to be more competent on the theory of mind items that depicted child themes than those portraying adolescent or adult themes, as the participants would presumably have more personal experience with the former situations.

On the ATOMIC central coherence questions, the children with AS were predicted to perform equivalently to the matched controls, as evidence suggests that individuals with ASDs have a preserved capacity to process information coherently at a global level when task demands clearly require them to do so (e.g. Snowling & Frith, 1986). The structured, forced-choice response format of the central coherence questions was believed to be conducive to participants consciously attempting to integrate the information contained within the vignettes to correctly answer the questions. However, as global processing is purported to be a non-preferential processing style in individuals with ASDs, it was expected that the children with AS would take longer to respond to the central coherence questions than controls. To investigate whether the performance of children with AS varied as a function of the number of items of information that needed to be processed in parallel, some of the central coherence questions required the serial integration of only two pieces of information, whereas others required the simultaneous integration of multiple stimuli. Consistent with Jolliffe and Baron-Cohen’s (2001) proposal that working memory load may determine the central coherence abilities of individuals with ASDs, it was predicted that the children with AS would perform more poorly on the central coherence questions that required the simultaneous integration of multiple items of information relative to questions that required the integration of only two items of information.

Finally, it was hypothesised that there would be no difference between the AS and control groups in their accuracy of responses and response times for the ATOMIC memory control questions, as research suggests that individuals with AS do not suffer from general attention or memory deficits (e.g. Ozonoff & Strayer, 2001; Russell, Jarrold, & Henry, 1996).

Methods

Participants

Thirty-nine children with AS aged between 8 and 10 years and an equivalent number of typically developing children were recruited for the study. Children with AS were recruited via an article in a local Brisbane newspaper, an advertisement placed in a support group newsletter, and
letters forwarded to eligible clients of a private practitioner specialising in ASDs and the Developmental Disabilities Unit of a children’s hospital. Children with AS were only permitted to participate if their diagnosis of AS had been confirmed by a paediatrician. Parents were asked to complete the childhood Asperger syndrome test (CAST) and a brief questionnaire to check that participants met DSM-IVTR criteria for Asperger’s disorder (APA, 2000). Thirtynine control participants were recruited from local schools via letters forwarded to students’ parents. To ensure that control participants did not have ASD symptomatology, parents of control participants were also required to complete the CAST and a brief screening questionnaire. Inclusion criteria for all participants included a WISC-III Prorated Full-Scale IQ of 85 or greater, and no current reading difficulties as reported by the child’s primary caregivers.

Preliminary analyses indicated that the control participants had significantly higher IQs than the participants with AS. Although statistical analyses were initially attempted with IQ entered as a covariate, heterogeneity of regression slopes threatened the validity of the results from these analyses. Thus, final data analyses were conducted on a sub-sample of 25 children from each group to ensure that the groups were matched on IQ. The 25 children who attained the highest WISC-III Full-Scale IQ scores in the AS sample were selected, in addition to a computer-generated random sample of 25 children from the control group. Each group comprised 23 males and 2 females. Mean age, WISC-III pro-rated Verbal Comprehension Index scores and pro-rated Full-Scale IQ scores for the participants with AS and typically developing controls are shown in Table 1. One-way ANOVAs revealed no significant differences between the groups in terms of age, F(1,48)<1F(1,48)<1, pro-rated WISC-III Full-Scale IQ, F(1,48)=2.21,p>.14F(1,48)=2.21, p>.14, or prorated WISC-III Verbal Comprehension Index scores, F(1,48)=1.66,p>.20F(1,48)=1.66, p>.20.

Materials

Childhood Asperger Syndrome Test (CAST)

The CAST was used to check participants’ diagnostic status because it is quick to administer, has high sensitivity ( 100%100 \% ) and specificity ( 97%97 \%; Williams et al., 2005), and has been shown to discriminate well between children with AS and typically developing children (Scott, Baron-Cohen, Bolton, & Brayne, 2002).

Animated Theory of Mind Inventory for Children (ATOMIC)

The ATOMIC consisted of 18 cartoons ( 1 practice item, 17 test stimuli) depicting a range of child, adolescent and adult

Table 1 Pro-rated full-scale IQ, verbal comprehension scores and age data for participants with AS and typically developing controls

themes, each followed by two multiple-choice questions. Twelve of the cartoons served as the test stimuli for the theory of mind items, with the remaining five cartoons constituting the stimuli for the central coherence items. Following each cartoon was a theory of mind or central coherence question, and finally a memory control question. Each question had four response options. Example theory of mind and central coherence vignettes and accompanying questions are shown in Appendix A. Cartoons were presented in a computer-generated, pseudo-random order, whereby each fourth cartoon was a central coherence vignette. Accuracy and response time scores were automatically computed by the ATOMIC programme. The programme took ∼25\sim 25 min to complete.

The ATOMIC theory of mind questions asked the respondent to infer characters’ mental states, and inherently required the integration of information contained within the cartoons. Seven of the theory of mind questions involved inferring characters’ emotions (e.g. ‘How was Lucy most likely to be feeling when Mark told her that he couldn’t go away?’). These questions required the inference of complex emotions (e.g. guilt, embarrassment and jealousy) that could not be deduced from simple visual clues alone (e.g. facial expression recognition). The remaining five theory of mind questions were about characters’ cognitions (e.g. 'Why did Phillip say, “I know, I can’t wait to get on it!”?).

The central coherence questions examined participants’ ability to integrate information unrelated to mental state phenomena. Three of these questions required the simultaneous integration of multiple items of information to enable the recognition of a scene (e.g. a boy’s bedroom), and the identification of an incongruent object (e.g. a female doll). The remaining two required participants to make inferences by integrating two items of information (e.g. ‘What is the most likely reason that the phone is not working?’).

To correctly answer the memory control questions, participants need to attend to non-essential visual and verbal information presented in the cartoons (e.g. a score on a test). These questions were designed to be of a similar difficulty level to the theory of mind and central coherence items.

The ATOMIC vignettes and test questions were written by the chief investigator, and animated by University stu-
dents who completed the programme to fulfil requirements for a multimedia course. Child and adult volunteers were recruited from local acting agencies to record voice overs for the characters featured in the programme. Prior to this study, extensive pilot testing was conducted with typically developing children and children with AS to ensure that there was consensus on the correct responses to items, and that the questions were of an appropriate difficulty level. Through this process, the number of cartoons included in the programme was reduced from 24 to 18 . All of the theory of mind and central coherence questions retained in the final version of the ATOMIC were correctly answered by a minimum of 11 of 15 typically developing children aged 8−118-11 years in pilot testing (see Table 2 for item analysis). This accuracy level is comparable to that obtained by Rutherford et al. (2002) when piloting the ‘Reading the Mind in the Voice’ task.

Short-form Version of the WISC-III

Donders’ (1997) short-form of the WISC-III was used to provide an estimate of participants’ Full-Scale IQ and Verbal Comprehension Index scores. This short form is comprised of the Vocabulary, Similarities, Picture Completion, Block Design, Arithmetic and Coding sub-tests of the WISC-III. The Full-Scale IQ and Verbal-Comprehension Index estimates derived from the Donders’ (1997) short-form have been shown to yield high reliability (.91-.94) and validity (.90–93) coefficients when compared to full-scale scores. Participants’ Block Design scores were also used to validate the ATOMIC Central Coherence scale.

Happe"s (1994) Strange Stories Task

A selection of ten stories was chosen from Happé’s (1994) Strange Stories Task to validate the theory of mind questions of the ATOMIC.

The Children’s Embedded Figures Test (CEFT)

The CEFT was used to validate the central coherence questions of the ATOMIC. This task involves finding hidden target shapes (e.g. a tent) within more complex pictures (e.g. a pram) as quickly as possible. The speed and

Table 2 Number of child participants endorsing each of the response options for the theory of mind (ToM) and central coherence (CC) items

accuracy of an individual’s responses are interpreted as a reflection of the tendency to process information in an integrated versus piecemeal manner. This measure is one of the most frequently used tests of central coherence with ASD populations (e.g. Brian & Bryson, 1996; Jolliffe & Baron-Cohen, 1997).

Procedure

Participants were tested individually by the chief investigator in either a small room at the Behaviour Research and Therapy Centre, University of Queensland (children with AS), or their local school (control participants). Participants initially completed the ATOMIC computer programme, followed by the WISC-III short-from, as these tasks were the most cognitively taxing. After a short break, participants were administered the Strange Stories Task and CEFT, with the order of test administration counterbalanced across participants in each group.

Strange Stories were read aloud to participants by the chief investigator, and responses to test questions were recorded on a response sheet. Detailed scoring criteria for the stories were developed based on guidelines provided by Francesca Happé at the author’s request, and information contained in published studies using the measure (e.g. Happé, 1994; Jolliffe & Baron-Cohen, 1999). Specifically, responses to the justification questions of the task were scored according to whether they referred to physical or mental state phenomena, and whether they were inaccurate (score of 0 ), partially accurate (score of 1) or completely correct (score of 2). The responses of seven participants in
each group were re-scored by a second rater, blind to participants’ diagnostic status. The percentage agreement between raters was 94%94 \% for accuracy of response, and 96%96 \% for the classification of responses as physical or mental state ( K=.88K=.88 for both measures).

For the CEFT, participants were allowed a maximum of 30 s to find the hidden shapes in the complex figures, as recommended by Nebelkopf and Dreyer (1970). Participants were scored on the total number of shapes that they located on their first attempt and within the 30 -s time limit, in addition to their mean response time for all correct items.

Results

Preliminary checks indicated that data was not normally distributed for many of the focal variables. However, data were still analysed using mixed-model ANOVAs, as parametric tests are considered to be quite robust to nonnormal data distributions if sample sizes are reasonably large (i.e. >30>30, Stevens, 1996). Due to the small number of items comprising the ATOMIC sub-scales, all simple effects analyses were rerun using non-parametric procedures. As the substantive results from these analyses remained the same, only data from parametric tests is reported.

Validity of Theory of Mind and Central Coherence Scales

A significant correlation was found between participants’ performance on the ATOMIC theory of mind questions and

the accuracy of their mental state justifications on the abbreviated Strange Stories Task, after shared variance due to verbal comprehension ability was statistically removed, partial r=.49,p<.001r=.49, p<.001. This finding supported the validity of the ATOMIC theory of mind scale. However, no significant relationships were found between participants’ ATOMIC central coherence scores and the number of shapes that they located on the CEFT, r=.09,p>.53r=.09, p>.53, their mean response time for correct items on the CEFT, r=−.01,p>.93r=-.01, p>.93, or their WISC-III Block Design task scores, r=−.15,p>.28r=-.15, p>.28. These results questioned the validity of the ATOMIC central coherence scale.

Number of Correct Answers

Due to the unequal number of theory of mind, central coherence and control questions included in the ATOMIC, participants’ total scores on each of these sub-scales were expressed as a proportion of the total number of items comprising each scale. The proportion of correct responses provided by participants on the theory of mind, central coherence and memory questions of the ATOMIC was entered into a 2×32 \times 3 mixed-model ANOVA with a between subjects factor of Group (AS versus Controls) and a within subjects factor of Question Type (theory of mind versus central coherence versus memory). Descriptive data for this analysis is presented in Fig. 1. Results showed a significant main effect of group, F(1,48)=8.76,p<.01,η2=.15F(1,48)=8.76, p<.01, \eta^{2}=.15, a significant main effect for Question Type, FF (2, 47)=13.51,p<.001,η2=.3747)=13.51, p<.001, \eta^{2}=.37 and a significant interaction between these two factors, F(2,47)=3.47,p<.05F(2,47)=3.47, p<.05, η2=.13\eta^{2}=.13. Planned simple effects analyses with a Bonfer-roni-adjusted alpha value ( α=.05/3=.016\alpha=.05 / 3=.016 ) showed that there was no significant difference between the groups on the central coherence questions, F<1F<1, although there was a trend for the AS group to perform more poorly than the

Fig. 1 Mean proportion of theory of mind (ToM), central coherence (CC)(C C) and memory questions correctly answered by control participants and participants with AS
control group on the memory questions, F(1,48)=3.45F(1,48)=3.45, p<.07,η2=.07p<.07, \eta^{2}=.07. Furthermore, the control group significantly outperformed the AS group on the theory of mind questions, F(1,48)=15.13,p<.001,η2=.24F(1,48)=15.13, p<.001, \eta^{2}=.24.

To determine whether the superior performance of the control group on the ATOMIC theory of mind questions was due to them attending to, and remembering the cartoons better than the AS group, a one-way ANCOVA was conducted with memory question performance entered as a covariate. Results showed that even after group difference in memory question performance was statistically controlled, the control group significantly outperformed the AS group on the theory of mind questions, F(1,47)=13.61F(1,47)=13.61, p<.001,η2=.23p<.001, \eta^{2}=.23.

The ATOMIC theory of mind questions could be divided into those that required the attribution of emotions (n=5)(n=5) versus those that required cognitions to be inferred (n=7)(n=7). To ascertain whether children with AS were differentially impaired relative to control participants at inferring people’s emotions relative to their cognitions, a 2×22 \times 2 mixed model ANOVA was conducted, with a between subjects factor of Group (AS versus Controls), and a within subjects factor of Theory of Mind Question Type (Emotion versus Cognition). Due to the unequal number of emotion and cognition theory of mind items, proportional scores on each of these sub-scales were entered into the ANOVA. Results showed a significant main effect of Group, F(1,48)=12.76,p<.01,η2=.21F(1,48)=12.76, p<.01, \eta^{2}=.21, with the control group performing significantly better on the theory of mind questions than the AS group (see Fig. 1 for descriptive data), but no significant main effect of Theory of Mind Question Type nor a significant interaction between these two factors.

To determine whether participants performed more competently on the theory of mind items with age-matched themes, a 2×22 \times 2 mixed model ANOVA was performed on proportional sub-scale scores. Group was the between subjects factor, and Theme (child versus adolescent/adult) was the within subjects factor. Of the 12 theory of mind cartoons included in the ATOMIC, four focused on adolescent/adult themes (e.g. a wedding) and the remainder portrayed scenes that children were more likely to have had personal contact or experience with (e.g. submitting a school project). The ANOVA revealed a significant main effect of Group, F(1,48)=14.87,p<.001,η2=.24F(1,48)=14.87, p<.001, \eta^{2}=.24, with the control group outperforming the AS group on the theory of mind questions, and a significant main effect of Theory of Mind Question Type, F(1,48)=35.89F(1,48)=35.89 p<.001,η2=.43p<.001, \eta^{2}=.43, with participants performing better on the adult-themed theory of mind questions ( M=.82M=.82, SD .21) than the child-themed theory of mind items ( M=.65M=.65, SD .18). However, the interaction between these two factors was not significant.

A 2 (Group) ×2\times 2 (Central Coherence Question Type) mixed-model ANOVA was also conducted to examine whether participants were differentially impaired on central coherence items that required the simultaneous integration of multiple items of information (complex), relative to those that required the integration of only two items of information (simple). As there were fewer simple central coherence questions than complex items, statistical analyses were conducted on proportional sub-scale scores. Results showed no significant Group main effect, nor a significant Group ×\times Central Coherence Question Type interaction. However, there was a significant Central Coherence Question Type main effect, F(1,48)=37.99F(1,48)=37.99, p<.001,η2=.44p<.001, \eta^{2}=.44, with participants correctly answering a higher proportion of simple central coherence items ( M=.88M=.88, SD .22) than complex items ( M=.57M=.57, SD .30).

Reaction Time

Reaction time data displayed floor effects and was positively skewed. Thus, all analyses were re-run with logarithmic transformations made to focal variables. As the substantive findings of the analyses did not differ, results are reported for parametric analyses only.

To test hypotheses pertaining to participants’ response times for the ATOMIC questions, a 2×32 \times 3 mixed-model ANOVA was conducted. This consisted of a between subjects factor of Group (AS versus Control), and a within subjects factor of Question Type (theory of mind, central coherence or memory). To control for differences in the number of items constituting each of the ATOMIC subscales, analyses were performed on average response time data. The ANOVA revealed a significant main effect of Question Type, F(2,47)=15.56,p<.001,η2=.39F(2,47)=15.56, p<.001, \eta^{2}=.39 but no significant main effect for Group or for Group ×\times Question Type interaction. Follow-up analyses of the Question Type Main Effect indicated that participants took significantly longer to respond to the ATOMIC theory of mind ( M=3.31 s,SD1.85M=3.31 \mathrm{~s}, \mathrm{SD} 1.85 ) and central coherence questions ( M=3.63 s,SD2.52M=3.63 \mathrm{~s}, \mathrm{SD} 2.52 ), than to the memory questions ( M=2.58M=2.58, SD 1.30).

Mixed-model ANOVAs also indicated that there was no significant difference in the mean response times of the AS and control groups on the cognition and emotion theory of mind questions, nor on the theory of mind questions depicting child themes, versus those displaying adolescent or adult interactions.

To examine whether participants took longer to answer the complex central coherence questions than the simple items, a 2 (Group: AS versus Control) ×2\times 2 (Central Coherence Question Type: two item-integration versus multiple-item integration) mixed-model ANOVA was
conducted on the mean response times for each central coherence sub-scale. Results showed no significant Group main effect, although the main effect of Question Type was significant, F(1,48)=8.25,p<.01,η2=.15F(1,48)=8.25, p<.01, \eta^{2}=.15, with participants taking longer to answer the complex central coherence questions ( M=4.23 sM=4.23 \mathrm{~s}, SD 3.51 ) than the simple items ( M=2.73 s,SD2.35M=2.73 \mathrm{~s}, \mathrm{SD} 2.35 ). No interaction between Group and Central Coherence Question Type was evident.

Discussion

Theory of Mind and ASDs
Consistent with the first hypothesis, participants with AS performed more poorly than matched controls on the ATOMIC Theory of Mind Questions. This did not appear to be due to general difficulties in integrating the information contained within the cartoons, as participants with AS performed equivalently to the control group on the central coherence questions. The trend for participants with AS to perform more poorly than matched controls on the memory questions suggested that attention and/or memory deficits may have contributed to their poor theory of mind task performance. However, statistical analyses showed that the impaired performance of the group with AS on the theory of mind questions prevailed after differences on the memory questions were statistically controlled. These results support previous research demonstrating mental state attribution difficulties in high-functioning individuals with ASDs on advanced theory of mind measures consisting of static and/or unimodal stimuli (e.g. Baron-Cohen et al., 2001; Happé, 1994). They also supplement findings from studies using multimodal, dynamic mind reading measures by showing that the poor mentalising abilities displayed by high-functioning individuals with ASD do not appear to be due to generalised deficits in integrating contextual information, or in making inferences.

The unexpected finding that participants with AS and controls responded to the ATOMIC theory of mind questions at an equally rapid pace suggests that the cognitive strategies used by the participants with AS to respond to these questions were highly developed and largely automatic. Future research may use neuroimaging studies to determine whether both neurotypical individuals and those with HFA or AS use similar mental strategies to attribute mental states to people engaged in social interactions.

As predicted, the participants with AS were impaired on both the cognition and emotion theory of mind questions relative to the control participants. This finding is consistent with past research demonstrating that high-functioning individuals with ASDs are impaired at attributing a wide range of mental states to others, including cognitions and

complex emotions (e.g. Happé, 1994; Heavey et al., 2000). However, contrary to expectations, participants with AS did not appear to perform more competently on theory of mind items depicting child themes compared with adolescent or adult themes. Alternatively, both the control participants and participants with AS performed better on the questions accompanying the adolescent/adult themed cartoons than the child-themed cartoons. This finding is difficult to interpret due to the small number of adolescent/ adult themed cartoons included in the ATOMIC (4). It is possible that even though the child-themed theory of mind items portrayed scenarios that participants were familiar with (e.g. classroom and playground), they may not have had personal experience with the specific situations depicted in the cartoons (e.g. feeling jealous of a competitor winning a race). Furthermore, participants are likely to have had exposure to television programmes, commercials and movies depicting adult interactions similar to those displayed in the adolescent/adult themed cartoons. Thus, one cannot assume that participants had more experience with, and exposure to, the child-themed scenarios than the adult-based interactions. This restricts the conclusions that can be made from this study regarding the facilitatory effect that personal experience has on a child’s ability to infer the thoughts and feelings of others.

Central Coherence and ASDs

As hypothesised, children with AS performed equivalently to matched controls on the ATOMIC central coherence questions. This finding supports and extends previous research showing that individuals with ASDs are able to integrate details into a global gestalt on perceptual processing tasks (e.g. Mottron et al., 1999; Plaisted et al., 1999), by demonstrating that participants with AS were also able to process multimedia information in context at a conceptual or ‘meaning-based’ level.

Contrary to expectations, the control participants and participants with AS took an equivalent amount of time to respond to the ATOMIC central coherence items. This finding is inconsistent with the notion that processing information in a coherent manner is a non-preferred processing style for individuals with ASDs, requiring more cognitive effort. The results suggest that when individuals with ASD are aware that they are required to integrate information, they are as competent at performing this task as are typically developing persons.

The hypothesis that the participants with AS would be differentially impaired relative to controls on the central coherence questions that required the simultaneous integration of multiple items of information, relative to those that only required the integration of two items of information was not supported. Although participants per-
formed more poorly on the central coherence questions with high-simultaneous processing demands, and took longer to respond to these items than those with lowsimultaneous processing demands, this was evident for both the control and AS groups. This result is inconsistent with Jolliffe and Baron-Cohen’s (2001) finding that participants with ASDs were impaired relative to matched controls in their ability to spontaneously identify objects that were incongruent with scenes. As Jolliffe and BaronCohen’s (2001) central coherence task was very similar to that used in the ATOMIC programme, differences in task demands are unlikely to account for this discrepancy in findings. Alternatively, differences in sample characteristics (i.e. participants in the present study were younger and had higher IQs than those in Jolliffe and Baron-Cohen’s investigation) may account for the dissimilar results. Another explanation is that there was an insufficient number of items comprising the ATOMIC central coherence subscale to detect subtle between group differences in the ability to simultaneously integrate multiple items of information.

Attention and Memory and ASDs

Contrary to expectations, there was a trend for participants with AS to perform more poorly than matched controls on the memory control questions of the ATOMIC. This finding is inconsistent with previous research (e.g. Ozonoff & Strayer, 2001) demonstrating that high-functioning individuals with ASDs are as capable as matched controls at attending to, and retaining, information. However, it is possible that the performance of the participants with AS on the ATOMIC memory control questions may have been impaired by the competing tasks of attempting to integrate important information contained within the cartoons, and making inferences about physical or mental state phenomena. These tasks may have taken more cognitive effort for the participants with AS than the controls, leaving fewer processing resources available to attend to the details that were the focus of the memory control questions.

Limitations and Future Directions

Although the ATOMIC represents an innovative methodology for examining theory of mind abilities in highfunctioning children, a note of caution is necessary when interpreting these findings. First, although the clinical diagnoses of the participants with AS were checked using care-giver questionnaire responses and CAST ratings, it would have been preferable for more rigorous screening methods to have been used. Resource constraints, together

with the anticipated unwillingness of participants and their caregivers to undergo lengthy diagnostic testing prior to participating in the study precluded standardised diagnostic instruments from being used. Another limitation of the study was the failure to use random sampling procedures when selecting the AS participant sample. However, selecting a sub-sample of children with AS with very highIQs provided a more rigorous test of the sensitivity of the ATOMIC in detecting subtle mentalising difficulties and of the applicability of the theory of mind and central coherence deficit accounts of ASD to very high-functioning individuals on the autism spectrum. A third limitation of the ATOMIC was the small number of items included in the central coherence scale. Items were not added to this scale, however, due to concerns about increased task duration resulting in fatigue effects on participants’ performance.

Finally, the lack of empirical support for the ATOMIC central coherence scale is problematic. The failure to find a significant relationship between participants’ performance on the ATOMIC central coherence questions and their competency on the Embedded Figures Test and Block Design Task suggests that there may have been a discontinuity between individuals’ local versus global processing abilities. Thus, superior ability to focus on the details of a figure may not be indicative of a concurrent impairment at integrating information for meaning at the global level, and vice versa. To overcome this problem, it would have been preferable to choose a central coherence measure that, like the ATOMIC Central Coherence questions, required the conceptual integration of information (e.g. Jolliffe and Baron-Cohen’s, 2001 task). This task was not chosen for validation purposes, however, due to concerns about practice effects resulting from its similarity to the ATOMIC central coherence scale.

An alternative explanation for the lack of a relationship between participants’ performance on the central coherence measures is that the ATOMIC central coherence questions may have been measuring competency in a domain other than central coherence, such as general knowledge. Admittedly, to correctly answer the central coherence questions of the ATOMIC, participants needed to possess a basic level of general knowledge (e.g. knowing that failure to recharge a laptop battery will result in the computer switching off). However, the central coherence questions did not assess general knowledge alone, as they could not be correctly answered without integrating general knowledge with content presented in the cartoons. It is arguable that the external validity of the ATOMIC central coherence scale was enhanced by requiring the application of general knowledge to make inferences about cartoon content. However, this raises the issue of what constitutes a ‘valid’ central coherence scale:
a question that has received scant attention in the literature. Before further progress can be made in refining the central coherence deficit account of ASDs, the construct of central coherence, and the domains that comprise it, need to be more clearly defined.

Following further refinement and testing, the ATOMIC has the potential to be a valuable tool in the comprehensive clinical assessment of high-functioning children referred with suspected theory of mind difficulties. Specifically, a child who attains a very low score on the ATOMIC Theory of Mind Scale may be expected to experience difficulties attributing mental states to others in real life contexts. To further enhance the sensitivity of the instrument, it may be useful to replace the cartoons with real-life footage of social interactions as test stimuli.

In summary, the present study demonstrated that children with AS had difficulty inferring characters’ mental states relative to matched controls on a computerised theory of mind measure comprised of multimodal cartoon stimuli. Furthermore, the mental state attribution difficulties experienced by the children with AS did not appear to be due to general difficulties in attending to, remembering, or integrating the information contained within the cartoons. Although the present findings have implications for the refinement of the current test, the development of the ATOMIC is an important step forward in improving the external validity of advanced theory of mind measures for children.

Acknowledgements We would like to thank Professor Tony Attwood and Dr. Tony Leslie for their help in recruiting children with Asperger syndrome for this project. We are especially grateful to the children with Asperger syndrome and students from Christ the King School, Jindalee State School and Fig Tree Pocket State School who participated in the study.

Appendix

Example Cartoons and Questions from the ATOMIC (correct answers to questions are shown in bold).

Example Theory of Mind Cartoon

Scenario: Primary school students are running in a race. One student, Jane, narrowly wins the race over a second student, Amanda. At the end of the race, Jane congratulates Amanda on having run a good race. Amanda says to Jane in a jealous tone of voice, ‘I don’t care that you won. I didn’t really try that hard, anyway’.

Question 1: How was Amanda MOST likely to be feeling when she said, ‘I don’t care that you won. I didn’t really try that hard, anyway?’
(theory of mind question)

(a) Jealous.
(b) Angry.
© Sad.
(d) Unconcerned.

Question 2: Why won the race?
(control question)
(a) Amanda.
(b) Jane.
© Amanda and Jane tied.
(d) Another competitor.

Example Central Coherence Cartoon

Scenario: A beach scene is shown with waves crashing on the shore, and several people dressed in swimwear playing and lying on the beach. One man is pictured wearing a jumper, long pants, a woollen hat and a scarf.

Question 1: Was there anything in the scene just shown to you that looked out of place?
(central coherence question)
(a) Yes: the clothing that a man was wearing.
(b) Yes: the sunglasses that a man was wearing.
© Yes: some people were wearing hats and others were not.
(d) No: everything in the scene looked normal.

Question 2: What was the weather like in the scene? (control question)
(a) Cloudy and raining.
(b) Sunny and windy.
© Sunny with clouds in the sky.
(d) Sunny with no clouds in the sky.

References

American Psychiatric Association. (2000). Diagnostic and statistical manual of mental disorders (4th ed., text revision). Washington, DC: American Psychiatric Association.
Attwood, T. (1998). Asperger’s syndrome: A guide for parents and professionals. London: Jessica Kingsley Publishers.
Baron-Cohen, S. (1995). Mindblindness: An essay on autism and theory of mind. Boston: MIT.
Baron-Cohen, S., Jolliffe, T., Mortimore, C., & Robertson, M. (1997). Another advanced test of theory of mind: Evidence from very high functioning adults with autism or Asperger syndrome. Journal of Child Psychology and Psychiatry and Allied Disciplines, 38, 813-822.
Baron-Cohen, S., Leslie, A. M., & Frith, U. (1985). Does the autistic child have a “theory of mind”? Cognition, 21, 37-46.

Baron-Cohen, S., O’Riordan, M., Stone, V., Jones, R., & Plaisted, K. (1999). Recognition of faux pas by normally developing children and children with Asperger syndrome or high-functioning autism. Journal of Autism and Developmental Disorders, 29, 407−418407-418.
Baron-Cohen, S., Wheelwright, S., Hill, J., Raste, Y., & Plumb, I. (2001). The ‘reading the mind in the eyes’ test revised version: A study with normal adults, and adults with Asperger syndrome or high-functioning autism. Journal of Child Psychology and Psychiatry and Allied Disciplines, 42, 241-251.
Bowler, D. M. (1992). Theory of mind in Asperger’s syndrome. Journal of Child Psychology and Psychiatry and Allied Disciplines, 33, 877-893.
Brian, J. A., & Bryson, S. E. (1996). Disembedding performance and recognition memory in autism/PDD. Journal of Child Psychology and Psychiatry and Allied Disciplines, 37, 865-872.
Castelli, F., Frith, C., Happé, F., & Frith, U. (2002). Autism, Asperger syndrome and brain mechanisms for the attribution of mental states to animated shapes. Brain, 125, 1839-1849.
Dahlgren, S. O., & Trilingsgaard, A. (1996). Theory of mind in nonretarded children with autism and Asperger’s syndrome: A research note. Journal of Child Psychology and Psychiatry and Allied Disciplines, 37, 759-763.
Donders, J. (1997). A short form of the WISC-III for clinical use. Psychological Assessment, 9, 15-20.
Frith, U. (1989). Autism: Explaining the enigma. Oxford: Blackwell.
Frith, U., & Happé, F. (1994). Autism: Beyong “theory of mind”. Cognition, 50, 115-132.
Frith, U., & Snowling, M. (1983). Reading for meaning and reading for sound in autistic and dyslexic children. British Journal of Developmental Psychology, 1, 329-342.
Happé, F. (1999). Understanding assets and deficits in autism: Why success is more interesting than failure. The Psychologist, 12, 540−546540-546.
Happé, F. G. E. (1994). An advanced test of theory of mind: Understanding of story characters’ thoughts and feelings by able autistic, mentally handicapped and normal children and adults. Journal of Autism and Developmental Disorders, 24, 129-154.
Heavey, L., Phillips, W., Baron-Cohen, S., & Rutter, M. (2000). The Awkward moments test: A naturalistic measure of social understanding in autism. Journal of Autism and Developmental Disorders, 30, 225-236.
Hoy, J. A., Hatton, C., & Hare, D. (2004). Weak central coherence: A cross-domain phenomenon specific to autism? Autism, 8, 267281.

Jolliffe, T., & Baron-Cohen, S. (1997). Are people with autism and Asperger syndrome faster than normal on the embedded figures test? Journal of Child Psychology and Psychiatry and Allied Disciplines, 38, 527-534.
Jolliffe, T., & Baron-Cohen, S. (1999). The strange stories test: A replication with high-functioning adults with autism or Asperger syndrome. Journal of Autism and Developmental Disorders, 29, 395-406.
Jolliffe, T., & Baron-Cohen, S. (1999). A test of central coherence theory: Linguistic processing in high-functioning adults with autism or Asperger syndrome: Is local coherence impaired? Cognition, 71, 149-185.
Jolliffe, T., & Baron-Cohen, S. (2001). A test of central coherence theory: Can adults with high-functioning autism or Asperger syndrome integrate objects in context? Visual Cognition, 8, 67101.

Klin, A. (2000). Attributing social meaning to ambiguous visual stimuli in higher-functioning autism and Asperger syndrome: The social attribution task. Journal of Child Psychology and Psychiatry and Allied Disciplines, 41, 831-846.

Lawson, J., Baron-Cohen, S., & Wheelwright, S. (2004). Empathising and systemising in adults with and without Asperger syndrome. Journal of Autism and Developmental Disorders, 34, 301−310301-310.
Leslie, A. M. (1987). Pretense and representation: The origins of “theory of mind.”. Psychological Review, 94, 412-426.
Mottron , L., & Burack, J. A. (2001). Enhanced perceptual functioning in the development of autism. In J. A. Burack, T. Chapman, N. Yirmiya & R. R. Zelazo (Eds.), The development of autism: Perspectives from theory and research (pp. 131-148). Mahwah, NJ: Lawrence Erlbaum Associates Publishers.
Mottron, L., Burack, J. A., Stauder, J. E. A., & Robaey, P. (1999). Perceptual processing among high-functioning persons with autism. Journal of Child Psychology and Psychiatry and Allied Disciplines, 40, 203-211.
Mottron, L., Peretz, I., & Menard, E. (2000). Local and global processing of music in high-functioning persons with autism: Beyond central coherence? Journal of Child Psychology and Psychiatry and Allied Disciplines, 41, 1057-1065.
Nebelkopf, E. B., & Dreyer, A. S. (1970). Perceptual structuring: Cognitive style differences in the perception of ambiguous stimuli. Perceptual and Motor Skills, 30, 635-639.
Ozonoff, S., Pennington, B. F., & Rogers, S. J. (1991). Executive function deficits in high-functioning autistic individuals: Relationships to theory of mind. Journal of Child Psychology and Psychiatry and Allied Disciplines, 32, 1081-1105.
Ozonoff, S., & Strayer, D. L. (2001). Further evidence of intact working memory in autism. Journal of Autism and Developmental Disorders, 31, 257-263.

Phillips, W., Baron-Cohen, S., & Rutter, M. (1998). Understanding intention in normal development and in autism. British Journal of Developmental Psychology, 16, 337-348.
Plaisted, K. C., Swettenham, J., & Rees, L. (1999). Children with autism show local precedence in a divided attention task and global precedence in a selective attention task. Journal of Child Psychology and Psychiatry and Allied Disciplines, 40, 733-742.
Roeyers, H., Buysse, A., Ponnet, K., & Pichal, B. (2001). Advancing advanced mind-reading tests: Empathic accuracy in adults with a pervasive developmental disorder. Journal of Child Psychology and Psychiatry and Allied Disciplines, 42, 271-278.
Rutherford, M. D., Baron-Cohen, S., & Wheelwright, S. (2002). Reading the mind in the voice: A study with normal adults and adults with Asperger syndrome and high functioning autism. Journal of Autism and Developmental Disorders, 32, 189-194.
Scott, F. J., Baron-Cohen, S., Bolton, P., & Brayne, C. (2002). The CAST (childhood Asperger syndrome test): Preliminary development of a UK screen for mainstream primary-school-age children. Autism, 6, 9-31.
Shah, A., & Frith, U. (1983). An islet of ability in autistic children: A research note. Journal of Child Psychology and Psychiatry and Allied Disciplines, 24, 613-620.
Snowling, M., & Frith, U. (1986). Comprehension in ‘hyperlexic’ readers. Journal of Experimental Child Psychology, 42, 392-415.
Stevens, J. (1996). Applied multivariate statistics for the social sciences (3rd ed.). Mahway, NJ: Lawrence Erlbaum.
Williams, J., Scott, F., Stott, C., Allison, C., Bolton, P., Baron-Cohen, S., et al. (2005). The CAST (childhood Asperger syndrome test): Test accuracy. Autism, 9, 45-68.