regular article systemizing empathy: teaching...

REGULAR ARTICLE

Systemizing empathy: Teaching adults withAsperger syndrome or high-functioningautism to recognize complex emotions usinginteractive multimedia

OFER GOLAN and SIMON BARON-COHENAutism Research Centre, University of Cambridge

AbstractThis study evaluated Mind Reading, an interactive systematic guide to emotions, for its effectiveness in teachingadults with Asperger syndrome ~AS! and high-functioning autism ~HFA! to recognize complex emotions in facesand voices. Experiment 1 tested a group of adults diagnosed with AS0HFA ~n � 19! who used the software at homefor 10–15 weeks. Participants were tested on recognition of faces and voices at three different levels ofgeneralization. A matched control group of adults with AS0HFA ~n � 22! were assessed without any intervention.In addition, a third group of general population controls ~n � 24! was tested. Experiment 2 repeated the design ofExperiment 1 with a group of adults with AS0HFA who used the software at home and met in a group with a tutoron a weekly basis. They were matched to a control group of adults with AS0HFA attending social skills training andto a general population control group ~n � 13 for all three groups!. In both experiments the intervention groupimproved significantly more than the control group on close, but not distant, generalization tasks. Verbal IQ hadsignificant effects in Experiment 2. Using Mind Reading for a relatively short period of time allows users to learn torecognize a variety of complex emotions and mental states. However, additional methods are required to enhancegeneralization.

The ability to recognize and understand emo-tions and mental states develops in the firstyear of life, with infants able to discriminate

different emotions from facial expressions andvocalizations ~Fernald, 1992; Haviland & Lel-wica, 1987!. Development of emotion recog-nition skill continues through childhood andadulthood as part of “theory of mind” ~Asting-ton, Harris, & Olson, 1988!, or what is alsoreferred to as “mindreading” ~Wellman, 1992!or “empathizing” ~Baron-Cohen, 2002, 2003!.

It is well established that emotion recogni-tion and mental state recognition are coredifficulties in people with autism spectrumconditions ~ASCs; Baron-Cohen, 1995; Hob-son, 1994!. These underlie the social difficul-ties that are diagnostic. Such difficulties have

The first author was supported by the National Alliancefor Autism Research, the Corob Charitable Trust, the Cam-bridge Overseas Trust and B’nai B’rith Leo Baeck schol-arships. The second author was supported by the ShirleyFoundation, the Medical Research Council, and the ThreeGuineas Trust. We are grateful to the following for help-ing with this study: our participants, Red Green and BlueLtd, Jessica Kingsley Ltd, Pamela Yates for arranging andrunning the social skills groups, Andrea Macleod andRob Whiskens ~Autism West Midlands!, Jill Howard ~Au-tism London!, Lynne Moxon ~European Services for Peo-ple with Autism!, and Harriet Fisher and James Graham~The Interact Centre!. Many thanks to Jacqueline Hill,Chris Ashwin, Sally Wheelwright, Matthew Belmonte,Yael Golan, Sarah Johnson, Emma Chapman, and IlariaMinio Paluello. A version of this work was presented atthe International Meeting For Autism Research ~IM-FAR!, May 2004.

Address correspondence and reprint requests to: OferGolan, Autism Research Centre, University of Cam-bridge, Department of Psychiatry, Douglas House, 18bTrumpington Road, Cambridge CB2 2AH, UK; E-mail:[email protected].

Development and Psychopathology 18 ~2006!, 591–617Copyright © 2006 Cambridge University PressPrinted in the United States of AmericaDOI: 10.10170S0954579406060305

591

been found through cognitive, behavioral, andneuroimaging studies, and across different sen-sory modalities, both visual and auditory ~Frith& Hill, 2004!.

Because the human face is central in boththe expression and communication of emo-tion, the majority of studies have focused onthe face and tested recognition of six emo-tions ~happiness, sadness, fear, anger, sur-prise, and disgust!. These “basic emotions”are expressed and recognized universally ~Ek-man, 1993; Ekman & Friesen, 1971!. Somestudies reveal emotion recognition deficitsamong individuals with ASC, compared to typ-ical or clinical control groups, using both static~Celani, Battacchi, & Arcidiacono, 1999;Deruelle, Rondan, Gepner, & Tardif, 2004;Macdonald et al., 1989! and dynamic stimuli~Hobson, 1986a, 1986b; Yirmiya, Sigman,Kasari, & Mundy, 1992!. Other studies havefound children and adults with high-functioningautism ~HFA! or Asperger syndrome ~AS! haveno difficulties in recognizing these basic emo-tions from pictures ~Adolphs, 2001; Gross-man, Klin, Carter, & Volkmar, 2000! or films~Loveland et al., 1997!, and that the deficitonly becomes apparent when testing recogni-tion of more “complex” emotions ~such asembarrassment, insincerity, intimacy, etc.! inboth adults and children with ASC ~Baron-Cohen, Wheelwright, Hill, Raste, & Plumb,2001; Baron-Cohen, Wheelwright, & Jolliffe,1997; Baron-Cohen, Wheelwright, Spong, Sca-hill, & Lawson, 2001; Golan, Baron-Cohen,& Hill, in press!. These findings suggest rec-ognition of basic emotions is relatively pre-served among high-functioning individualswith ASC, and that they show greater difficul-ties recognizing more complex emotional andmental states.

Neuroimaging studies of emotion recogni-tion from faces reveal that people with ASCshow less activation in brain regions central toface processing, such as the fusiform gyrus~Critchley et al., 2000; Pierce, Muller, Am-brose, Allen, & Courchesne, 2001; Schultzet al., 2003!. Behavioral studies show that chil-dren and adults with ASCs process faces dif-ferently compared to controls: participants withASC tend to process faces in a feature-basedapproach, whereas controls process faces con-

figurally ~Hobson, Ouston, & Lee, 1988;Schultz et al., 2003; Teunisse & De Gelder,1994; Young, 1998!. There is also evidence ofreduced activation in brain areas that play amajor role in processing of emotion, such asthe amygdala, when individuals with ASC pro-cess social–emotional information ~Ashwin,Baron-Cohen, Wheelwright, O’Riordan, &Bullmore, 2005; Baron-Cohen et al., 1999;Critchley et al., 2000!.

Emotion recognition from voices has beenstudied less frequently. Here, too, there arecontradictory findings in relation to recogni-tion of basic emotions ~Boucher, Lewis, &Collis, 2000; Loveland, Tunali Kotoski, Chen,& Brelsford, 1995; Loveland et al., 1997!.Regarding recognition of complex emotionsfrom voices, several studies report a deficit inperformance in high-functioning adults withASC compared to controls ~Golan et al., inpress; Golan, Baron-Cohen, Hill, & Ruther-ford, 2005; Kleinman, Marciano, & Ault, 2001;Rutherford, Baron-Cohen, & Wheelwright,2002!. To date, there are no reported neuro-imaging studies assessing emotion recogni-tion in voices in ASC, perhaps because thesound of the scanner itself ~the echo-planarsystem! makes auditory brain scanning diffi-cult. One study used positron emission tomog-raphy to measure brain activity of participantswith ASC and matched controls while listen-ing to theory of mind stories. Both groupsshowed similar patterns of brain activity, butactivation in the medial frontal area of thebrain was less intensive and extensive in theAS group, compared to controls ~Nieminen-von Wendt et al., 2003!.

Studies assessing the ability of individualswith ASC to identify emotions and mentalstates from context have also shown deficitsrelative to the general population or to otherclinical control groups ~Baron-Cohen, Leslie,& Frith, 1986; Fein, Lucci, Braverman, &Waterhouse, 1992!. For example, adolescentsand adults with ASC have difficulties answer-ing questions on the Strange Stories Test~Happe, 1994; Jolliffe & Baron-Cohen, 1999!.This test assesses the ability to providecontext-appropriate mental state explanationsfor nonliteral statements made by story char-acters ~e.g., ironic or sarcastic statements!.

592 O. Golan and S. Baron-Cohen

When using this task in a neuroimaging study,reduced activation of the left medial prefron-tal cortex was found in people with ASC com-pared to matched controls ~Happe et al., 1996!.

The integration of cross-modal emotionalinformation from faces, voices, and contextallows understanding and prediction of oth-ers’ emotions and mental states ~Baron-Cohen, 1995!, and the more informationavailable, the easier should be the recogni-tion of these emotions or mental states. Stud-ies assessing complex emotion and mentalstate recognition from ecologically rich so-cial situations, containing multimodal sourcesof information, show a deficit in individualswithASC, compared to controls ~Golan, Baron-Cohen, Hill, & Golan, in press; Heavey, Phil-lips, Baron-Cohen, & Rutter, 2000; Klin, Jones,Schultz, Volkmar, & Cohen, 2002!. These dif-ficulties may be related to a failure to pick upthe right emotional cues, and0or to a failureintegrating them, explained by weak centralcoherence in the cognitive level ~Frith, 1989!,and underconnectivity between brain regionsin the neurobiological level ~Belmonte, Allen,et al., 2004; Belmonte, Cook, et al., 2004;Critchley et al., 2000!. Although emotion rec-ognition deficits in ASC are life-long, somehigh-functioning individuals develop compen-satory strategies that allow them to recognizebasic emotions. However, when recognitionof more complex emotions and mental statesis required, many find them hard to interpret.

In contrast to these difficulties, individu-als with ASC show good and sometimes evensuperior skills in “systemizing” ~Baron-Cohen, 2003!. Systemizing is the drive to an-alyze or build systems, to understand andpredict the behavior of nonagentive eventsin terms of underlying rules and regularities.Individuals with ASC are hyperattentive todetail and prefer predictable, rule-based envi-ronments, features that are intrinsic to system-izing. In addition, individuals with ASC aresuperior to controls on various tasks that in-volve searching for detail, analyzing and ma-nipulating systems ~Baron-Cohen, Richler,Bisarya, Gurunathan, & Wheelwright, 2003;Baron-Cohen, Wheelwright, Spong, et al.,2001; Jolliffe & Baron-Cohen, 1997; Law-son, Baron-Cohen, & Wheelwright, 2004;

O’Riordan, Plaisted, Driver, & Baron-Cohen,2001; Shah & Frith, 1993!.

If high-functioning individuals with ASCpossess good systemizing skills, it is possiblethey could use them to compensate for someof their empathizing difficulties. This mightbe hard to implement, because the socioemo-tional world is a context-related open system~Lawson, 2003!, often unpredictable and dif-ficult to conceptualize with strict rules. How-ever, if provided with a system of emotions, itis plausible that systemizing skills could beharnessed to help individuals with ASC learnto recognize emotions.

Past attempts to teach emotion recognitionto adults and children with ASC have eitherfocused on the basic emotions ~Hadwin,Baron-Cohen, Howlin, & Hill, 1996; Howlin,Baron-Cohen, & Hadwin, 1999!, or have beenpart of social skills training courses, usuallyrun in groups ~Barry et al., 2003; Howlin &Yates, 1999; Rydin, Drake, & Bratt, 1999!.These training programs typically do not fo-cus specifically on systematically teachingemotion recognition, but instead address otherissues, such as conversation, reducing so-cially inappropriate behavior, personal hy-giene, and so forth. In such groups it isdifficult to target the individual’s specific paceof learning. Finally, such groups are sociallydemanding, and might therefore deter moresocially anxious participants.

Other attempts to teach individuals withASC social skills have used computer-basedtraining ~Bernard-Opitz, Sriram, & Nakhoda-Sapuan, 2001; Bolte et al., 2002; Hetzroni &Tannous, 2004; Rajendran & Mitchell, 2000;Silver & Oakes, 2001; Swettenham, 1996!.The use of computer software for individualswith ASC has several advantages: first, indi-viduals with ASC favor the computerized en-vironment because it is predictable, consistent,and free from social demands, which they mayfind stressful. Second, users can work at theirown pace and level of understanding. Third,lessons can be repeated over and over again,until mastery is achieved. Fourth, interest andmotivation can be maintained through differ-ent and individually selected computerized re-wards ~Bishop, 2003; Moore, McGrath, &Thorpe, 2000; Parsons & Mitchell, 2002!. Pre-

Teaching adults with AS0HFA to recognize emotions 593

vious studies have found that the use of com-puters can help individuals with autism passfalse belief tasks ~Swettenham, 1996!, recog-nize basic emotions from cartoons and stillphotographs ~Bolte et al., 2002; Silver &Oakes, 2001!, and solve problems in illus-trated social situations ~Bernard-Opitz et al.,2001!. However, participants find it hard togeneralize their knowledge from learnt mate-rial to related tasks.

The computer-based interventions aboveused drawings or photographs for training, ratherthan more lifelike stimuli.This might have madegeneralization harder than if more ecologicallyvalid stimuli were used. In addition, the pro-grams teaching emotion-recognition focused onbasic emotions, and only on facial expressions.No reported program to date has systemati-cally trained complex emotion recognition inboth visual and auditory channels, with lifelikefaces and voices. The study reported here eval-uates such a program in two formats: individ-ually based, and via a group intervention. Thequestion tested in both was “Can the goodsystemizing skills that individuals with ASCpossess be used to teach them to improve theirrecognition of complex emotions?” To test thisquestion, emotions and mental states had to bepresented as a system for learners to analyzeand study.

Mind Reading: A Systematic Guide toEmotions

Mind Reading ~Baron-Cohen, Golan, Wheel-wright, & Hill, 2004! is an interactive guide toemotions and mental states. It is based on ataxonomic system of 412 emotions and men-tal states, grouped into 24 emotion groups,and six developmental levels ~from age 4 toadulthood!. The emotions and mental statesare organized systematically, according to theemotion groups and developmental levels. Eachemotion group is introduced and demon-strated by a short video clip giving some cluesfor later analysis of the emotions in this group.Each emotion is defined and demonstrated insix silent films of faces, six voice recordings,and six written examples of situations thatevoke this emotion. The resulting library of

emotional “assets” ~video clips, audio clips,or brief stories! comprise 412 � 18 � 7,416units of emotion information to learn to rec-ognize or understand. This is therefore a richand systematically organized set of educa-tional material. The face videos and voice re-cordings comprise actors of both genders,various ages, and ethnicities, to facilitate gen-eralization. Faces and voices are presented sep-arately for each emotion ~i.e., silent face filmsand faceless voice recordings! to encourageanalysis of the emotion in each modality. Allface video clips and voice recordings werevalidated by a panel of 10 independent judges,and were included in Mind Reading if at least8 judges agreed the emotional label given de-scribed the face0voice ~p, .05, binomial test!.

This emotion database is accessed usingthree applications: ~a! The emotion library al-lows users to browse freely through the differ-ent emotions and emotion groups, play thefaces, voices and scenarios giving examplesof the emotions, read stories, add their ownnotes, and compare different emotional expres-sions in the face and the voice using a scrap-book. ~b! The learning center uses lessons,quizzes, and several reward collections to teachabout emotions in a more structured and di-rective way. In addition to teaching about the24 emotion groups, it also includes lessonsand quizzes about the 20 and 100 most com-monly used emotions, as well as a “build yourown lesson0quiz” option. The various rewardcollections were chosen for their potential ap-peal to users with ASC and were arrangedsystematically ~e.g., pictures and informationabout space elements, clips of birds arrangedby families, different types of trains to collect,etc.!. A reward is given when a quiz questionis answered correctly on the first attempt. ~c!The game zone comprises five educationalgames, allowing users to enjoy a game whilestudying about emotions. The software wascreated for the use of children and adults ofvarious levels of functioning. Vocal and ani-mated helpers give instructions on everyscreen. Figure 1 shows screen shots from thesoftware ~www.jkp.com0mindreading!.

The study reported here includes two ex-periments. Experiment 1 tested for any im-provement in adults with AS0HFA in emotion


Figure 1. Screenshots from Mind Reading: The Interactive Guide to Emotions, by S. Baron-Cohen, O. Golan, S. Wheelwright, &J. J. Hill, 2004, London: Jessica Kingsley Limited. Copyright 2003 by the University of Cambridge. Reprinted with permission.@A color version of this figure can be viewed online at www.journals.cambridge.org#

595

recognition skills following independent useof the software, and the extent to which theseusers can generalize their acquired knowl-edge. Experiment 2 tested the same in a groupof participants who, in addition to using thesoftware individually, attended a weekly groupsession led by a tutor. Using group discus-sion, role play, worksheets, and analysis ofemotions in newspapers and television pro-grams, this group aimed to consolidate partici-pants’ computer-acquired knowledge toimprove generalization. Both interventionstook place over a period of 10–15 weeks, toassure a meaningful period for training, rec-ognizing that a longer duration might lead toindividuals dropping out. This allowed us totest whether the educational software ~used inboth experiments! has an effect independentof any human tutor or social group effects, orwhether addition of a human tutor and socialgroup enhances any effects. In both experi-ments, participants were tested before and af-ter the intervention. A no-computer interventioncontrol group of adults with AS0HFA wasmatched to the intervention group in both ex-periments. This AS0HFA control group wasalso tested before and after a similar period oftime. In Experiment 1 the AS0HFA controlgroup had no intervention, thereby simply con-trolling for the passage of time. In Experiment2 the AS0HFA control group undertook a 10–15 week social skills course, but without usingMind Reading, thereby controlling for the spe-cific use of this educational software. Finally,a typical control group from the general pop-ulation was matched to the intervention groupsin both experiments. This group was only testedonce, to obtain baseline measures.

For both experiments we predicted thefollowing:

1. the performance of participants with AS0HFA would be lower than typical controlson all emotion recognition tasks at Time 1;

2. AS0HFA groups using the software wouldperform better on all emotion recognitiontasks at Time 2, compared to Time 1, acrossgeneralization levels; and

3. participants using the software would im-prove more than AS0HFA controls on alltasks.

In addition, the association of softwareusage time with improvement on the differenttasks was examined for the intervention groups.

Experiment 1

This experiment compared the effect of usingMind Reading at home alone to that of takingthe assessment twice with no intervention, atthree levels of generalization. The need for ano-intervention AS0HFA group was to assesswhether any improvement was related to theintervention or merely to taking the tasks twice.

Method

Design. A controlled trial design with multi-ple repeated measures was used for this exper-iment. Two groups, the home intervention andthe AS0HFA control, were assessed twice; atTime 1 and, 10–15 weeks after, at Time 2. Ineach assessment, generalization was assessedusing stimuli in two modalities ~visual andauditory! and three generalization levels:

1. Close generalization: this was tested usingfaces and voices that were included in thecomputer intervention but were presentedusing different software, and with morechallenging multiple choice answers.

2. Feature-based distant generalization: usingfaces and voices that were not included inthe training, we tested the ability to trans-fer emotion recognition skills separately infaces and voices.

3. Holistic distant generalization: using scenesfrom feature films, this level comprised ho-listic socioemotional stimuli, includingfaces, voices, body language, and context.Participants were then asked to identify theemotion of one of the characters in thescene, thus testing the ability to transferacquired knowledge to complex holisticstimuli. This level was only assessed atTime 2.

In addition, the two groups were comparedto the typical control group at Time 1.

Participants. Three groups took part in thisexperiment; one AS0HFA intervention group,


one AS0HFA control group, and one typicalcontrol group. Participants in the clinicalgroups had all been diagnosed with AS0HFAin specialist centers using established criteria~American Psychiatric Association, 1994;World Health Organization, 1994!. Partici-pants filled in the Autism Spectrum Quotient~AQ; Baron-Cohen, Wheelwright, Skinner,Martin, & Clubley, 2001!, to assess their self-reported level of autistic traits. Eighty percentof the participants scored above a cutoff pointof 32, which matches the percentage origi-nally reported by Baron-Cohen et al. ~2001!.Eighty-eight percent of the participants scoredabove 26, which has recently been suggestedin two separate studies as a more sensitivecutoff point for the AQ ~Kurita, Koyama, &Osada, 2005; Woodbury-Smith, Robinson,Wheelwright, & Baron-Cohen, 2005!. Partici-pants were recruited from several sources, in-cluding a volunteer database, a local clinic foradults with AS0HFA, and an advert in theNational Autistic Society magazine Commu-nication. Participants were accepted to thestudy only if they had not participated in anyrelated intervention during the last 3 monthsand had no plans for engaging in another in-tervention while the study was ongoing. Par-ticipants were randomly allocated into twogroups:

Software home users. Nineteen partici-pants ~14 males, 5 females! were asked to usethe software ~provided free of charge! at homeby themselves for 2 hr0week over a period of10 weeks, a total of 20 hr. Participants wereincluded in the study if they completed a min-imum of 10 hr of work with the software. Ifthey did not complete this minimum, partici-pants were given an extension of up to 4 weeksto do more work with the software. Out of 24participants originally recruited to this group,3 withdrew during the 10-week period and 2others were excluded at the end, as they failedto reach the 10-hr minimum. No specific pat-tern was found for these participants: they var-ied in their age range ~21– 43!, education ~3had carried on studying beyond compulsoryeducation, 2 had not!, and employment status~2 were unemployed, 3 were employed!. TheirIQ, AQ, and Time 1 assessment task scores

ranged within 1 SD of their group means. Allof them related dropping out0not completingtheir work to being too busy and not getting todo the required amount of work. This findingwill be discussed further.

AS0HFA control group. Twenty-two partici-pants ~17 males, 5 females! attended the as-sessment meetings with a 10–15-week periodbetween them, during which they did not takepart in any intervention related to emotionrecognition.

Typical control group. Twenty-eight par-ticipants were recruited for this group from alocal employment agency. Participants re-ported no psychiatric history and no occur-rence of ASC in their families. After screeningfor autistic spectrum conditions using the AQ~Baron-Cohen, Wheelwright, Skinner, et al.,2001!, four participants were excluded for scor-ing above the cutoff of 26. The remaining 24participants ~19 males, 5 females! attendedone assessment meeting.

All participants were given the WechslerAbbreviated Scale of Intelligence ~WASI!,comprising the vocabulary, similarities, blockdesign, and matrix reasoning tests. The WASIproduces verbal, performance, and full-scaleIQ scores, with correlations of .88, .84, and.92 with the full Wechsler scales ~Wechsler,1999!. All participants scored above 70 onboth verbal and performance scales.

All participants completed the AQ ~Baron-Cohen,Wheelwright, Skinner, et al., 2001!. One-way analysis of variance ~ANOVA!ofAQ scoreswas significant, F ~2, 62! � 81.01, p , .001.Tukey post hoc comparisons revealed that thetwo clinical groups scored significantly higheron theAQ compared to the typical control group.Participants were asked to report any psychi-atric comorbid diagnoses. Five participants ineach AS0HFA group had another psychiatricdiagnosis, such as depression or attentiondeficit0hyperactivity disorder. No differencewas found between the AS0HFA groups in theproportion of participants with psychiatric co-morbidity, x2 ~1!� .07, ns ~see Table 1!.

The three groups were matched on age,verbal and performance IQ, handedness, andgender. They spanned an equivalent range of


employment and educational levels. As shownin Table 1, no significant differences werefound between the groups for age, F ~2, 62!�2.14, ns, verbal IQ, F ~2, 62!� 2.31, ns, per-formance IQ, F ~2, 62! � 0.53, ns, gender,x2 ~2!� 0.18, ns, handedness, x2 ~2!� 1.91,ns, education, x2 ~2!� 0.47, ns, and employ-ment, x2 ~2!� 0.20, ns. In addition, no differ-ence was found between the two AS0HFAgroups in the time between the two assess-ment meetings, t ~27.7!� 1.57, ns.

Instruments.

The Cambridge Mindreading (CAM) Face–Voice Battery. This battery ~Golan et al., inpress! tests complex emotion and mental staterecognition from short silent clips of facesand from voice recordings. Fifty faces and 50voices, taken from Mind Reading, test recog-nition of 20 different complex emotions andmental states ~e.g., intimate, insincere!. Theface task comprises silent clips of adult ac-tors, both male and female, of different eth-nicities, expressing the emotions in the face.An example showing one frame from one ofthe clips is shown in Figure 2. The voice task

comprises recordings of short sentences ex-pressing various emotional intonations. In bothtasks four adjectives were presented after eachstimulus is played and participants were askedwhich adjective best describes how the personfeels. Both tasks were run on a computer, usingthe experimental software DMDX ~Forster &Forster, 2003!. A handout of definitions of allthe adjectives used in the battery was avail-able for the participants at the beginning andthrough the assessment. There was no timelimit for answering. The battery provides anoverall facial and an overall vocal emotionrecognition score ~max � 50 for each of them!,as well as individual scores for each of the 20emotions assessed ~pass0fail, i.e., recognizedabove chance or not! and an overall number ofthe emotions correctly recognized ~max � 20!.Individuals with AS0HFA have been found toscore significantly lower than controls on allthree scores of the battery ~Golan et al., inpress!. Test–retest correlations, calculated forthe AS0HFA control group in the current studywere found to be r � .94 for the face task andr � .81 for the voice task, p , .001 for both.This battery was used for the assessment ofclose generalization in both faces and voices.

Table 1. Means (standard deviations) and ranges of age, IQ, and AQ scores for theintervention and control groups of Experiment 1

Software Home Users~N � 19!

AS0HFA Controls~N � 22!

Typical Controls~N � 24! F ~2, 62!

Age 30.5 ~10.3! 30.9 ~11.2! 25.3 ~9.1! 2.1417.5– 48 17.5–52 17.5–51

Verbal IQ 108.3 ~13.3! 109.7 ~10.0! 115.8 ~13.7! 2.3180–127 93–129 86–138

Performance IQ 112.0 ~12.6! 115.3 ~12.3! 112.5 ~8.9! 0.5394–134 97–140 92–129

AQ 37.2 ~8.4! 38.2 ~7.5! 14.0 ~5.9! 81.02*20– 47 16– 49 6–26

x2~2!

Females ~%! 26.3 22.7 20.8 0.18Left handed ~%! 21.1 9.1 8.3 1.91Employed ~%! 47.4 40.9 45.8 0.20A levels or above ~%!a 68.4 63.6 58.3 0.47Comorbid diagnoses ~%! 26.3 22.7 NA 0.07

aA levels are the first component of noncompulsory education in Britain.*p , .001. All other test results are not significant ~ p . .05!.


Figure 2. Examples of visual tasks from the three generalization levels: ~top! close generalizationadapted from the CAM Face task from Golan et al. ~in press!, ~middle! feature-based distant general-ization adapted from Reading the Mind in the Eyes Test from Baron-Cohen et al. ~2001!, and ~bottom!holistic distant generalization adapted from Reading the Mind in Films task ~Golan & Baron-Cohen2006!. Screenshot from Lost for Words, by D. Longden, 1999, London: Yorkshire Television. Copyright1999 ITN. Reprinted with permission. @A color version of this figure can be viewed online at www.jour-nals.cambridge.org#

599

Reading the Mind in the Eyes task (re-vised, adult version). The task ~Baron-Cohen,Wheelwright, Hill, et al., 2001! has 36 items,in which participants are presented with aphotograph of the eyes region of the face andmust choose one of four adjectives or phrasesto describe the mental state of the person pic-tured ~see Figure 2 for an example!. A defini-tion handout is provided at the beginning ofthe task. In the present study, the pictures andadjectives were presented on the computerscreen ~using DMDX software!, to avoid pos-sible difficulties due to communication with ahuman examiner ~Ozonoff, 1995a!. Items werepresented in a random order. There was notime limit for answering. Test–retest correla-tion, calculated for the AS0HFA control groupin the current study was found to be r � .86~p , .001!. This task was used as a facialfeature based generalization task.

Reading the Mind in the Voice task (re-vised). We used a revised version ~Golan et al.,2005! of the original task ~Rutherford et al.,2002!. In the original task, 40 segments ofspeech, taken from BBC drama series, wereplayed on audio tape to the participants, whowere asked to choose one out of two possibleanswers, describing the speaker’s mental statein each item. The task was revised as follows:each of the test items was allocated two morefoils, taken from the same level: one levelabove or one level below the correct answer~based on the emotion taxonomy in Mind Read-ing!. Foils were selected to match the contentof the verbalizations but not the intonation,thus making the task harder to answer. Thisavoided ceiling effects to which the originalversion of the test was prone. Seven itemswere removed because the authors found thecorrect answer inappropriate to the verbaliza-tion. Eight more items were excluded aftervalidation by a sample of 15 typically devel-oping adults. The final task included 25 itemswith four possible answers for each of them.The test items were “cleaned” as far as possi-ble and played on a computer ~using DMDXsoftware! in random order, preceded by aninstruction slide and two practice items. Par-ticipants were given a definition handout be-fore the beginning of the task. There was notime limit for answering. Test–retest correla-

tion, calculated for the AS0HFA control groupin the current study was r � .80 ~ p , .001!.This task was used as a vocal feature basedgeneralization task.

Reading the Mind in Film task. This task~Golan, Baron-Cohen, Hill, et al., in press! com-prises 22 short social scenes taken from featurefilms. Each scene includes visual, vocal, andsome contextual information. Scenes are pre-sented on the computer screen ~using DMDXsoftware!. Participants are presented with fouradjectives and are asked to choose the one thatbest describes the way a target character feelsat the end of the scene. Foils were selected tomatch some aspects of the scene ~e.g., contentof speech! but not all of them ~e.g., facial ex-pression, intonation, etc.!, thus making the taskharder to answer. A handout of definitions ofall the adjectives used in the task was availablefor the participants at the beginning and throughthe assessment. There was no time limit for an-swering. Task score is the number of correctlyrecognized emotions ~max � 22!. Participantswith AS0HFA were found to perform signifi-cantly worse on this task, comparing to matchedcontrols. Figure 2 presents a screenshot fromone of the task items. This task was used as aholistic distant generalization task and was onlytaken at Time 2.

Procedure

Participants were individually tested at the Au-tism Research Center in Cambridge. The firstauthor and three trained assistants helped theparticipants through the assessments. Theassistants were blind as to which group theparticipants belonged. Participants in the in-tervention group were asked to help in theevaluation of a new piece of software. It wasexplained they would need to commit to usingMind Reading for 2 hr0week over a period of10 weeks, and to be assessed before and afterthis training period. Participants of both con-trol groups were asked to take part in an emo-tion recognition study, helping to validate newtasks. For this reason, participants in the AS0HFA control group were asked to come fortwo assessments, separated by a 10- to 15-week period. Participants’ written consent was


obtained. All participants were told they werefree to withdraw from the study at any time.

In the first assessment background infor-mation was collected, followed by the emo-tion recognition tasks. Participants were seatedin front of IBM compatible computers with15-in. monitors and were given headphonesfor the voice tasks. The emotion recognitiontasks were presented to them in random order~with the exception of the Film task!. The CAMbattery included two breaks in each task dueto its length. Participants were also allowed abreak to freshen up between tasks. In betweenthe tasks, two subtests of the WASI were ad-ministered. After the assessment was com-pleted, Mind Reading was introduced to theparticipants of the intervention group in de-tail, including a presentation of the emotiontaxonomy, the different areas, and a demon-stration of a systematic analysis of an emo-tion, comparing different faces and voices toidentify the unique facial0intonation featuresof this emotion. Participants were encouragedto analyze the facial and vocal stimuli system-atically. They were asked to use the emotionslibrary and learning center as they please, butnot to use the game zone for more than a thirdof the usage time. They were also told MindReading logs their work, and they were askedto bring the log file to the second assessmentmeeting, for usage time verification. The wholeassessment meeting took about 3 hr. Duringthe time between the two assessments partici-pants of the intervention group were ap-proached by telephone at least once, to checkthey were still committed to the study and toworking with the software.

In the second assessment participants tookthe same emotion recognition tasks, includingthe Film task and the other two WASI sub-tests. Task administration order was changed.The log files of participants in the interven-tion group were checked to verify they hadused the software for the required amount oftime. Average usage time was 17.5 hr ~SD �6.7, range � 10–36!. These participants werethen asked for their feedback about the pro-gram, their experience with it, and commentsabout its usefulness. Participants of the AS0HFA control group were asked about any pos-sible effects of the first assessment meeting

on their interest in emotions during the periodbetween the assessments. All participants werethen debriefed about the aims and design ofthe study and were rewarded with a compli-mentary copy of Mind Reading ~or were al-lowed to keep the copy they used, with nocharge!. This assessment meeting took about3 hr.

Results

The performance of the three groups on theemotion recognition tasks at Time 1 was ex-plored first. Five one-way analyses of vari-ance were conducted, testing group differenceson the emotion recognition tasks used atTime 1. Using Holm’s sequential rejective Bon-ferroni procedure ~Holm, 1979; see also Zhang,Quan, Ng, & Stepanavage, 1997!, significantdifferences were found between the groups onthe CAM face task, F ~2, 62! � 13.82, p ,.001, voice task, F ~2, 60!� 11.53, p , .001,the number of emotional concepts recognized,F ~2, 60! � 12.77, p , .001, the Reading theMind in the Eyes task, F ~2, 62!� 6.10, p ,.01, and the Reading the Mind in the Voice-Rtask, F ~2, 62! � 4.92, p , .02. Preplannedcomparisons with Bonferroni corrections re-vealed no significant differences between thetwo clinical groups on any of the task scores,and significantly higher scores of the typicalcontrol group on all tasks, comparing to thetwo AS0HFA groups. These findings supporthypothesis 1. Table 2 shows the means andstandard deviations of the groups’ emotion rec-ognition scores at Time 1.

Next, five multivariate analyses of covari-ance ~MANCOVA! with repeated measureswere conducted, to examine the differences be-tween the intervention and AS0HFA controlgroup on the various tasks at Time 1 and Time 2.Age, verbal, and performance IQ were usedas covariates. Using Holm’s sequential rejec-tive Bonferroni procedure, significant Time �Group interactions were found on CAM faces,FWilks ~1, 35!�11.82, p, .002, CAM voices,FWilks ~1, 34! � 7.51, p , .01, CAM numberof concepts recognized, FWilks ~1, 33!� 8.38,p , .01, but not for Reading the Mind in theEyes, FWilks ~1, 36!� 1.46, ns, or Reading theMind in the Voice-R, FWilks ~1, 36!� 0.47, ns.


No main effects were found significant, butthe effect of the covariate verbal IQ was sig-nificant in relation to CAM voice scores,F ~1, 34!� 5.11, p , .05.

Simple main effects analyses for the threeCAM scores with Bonferroni corrections re-vealed the intervention group improved sig-nificantly from Time 1 to Time 2 on all threescores: for faces, t ~17!� 5.37, p , .001; forvoices, t ~16! � 5.24, p , .001; for conceptsrecognized: t ~15!� 3.96, p , .005. The AS0HFA control group scores did not change sig-nificantly from Time 1 to Time 2 on the CAMvoices, t ~21! � 1.43, ns, and the number ofconcepts recognized, t ~21!�1.25, ns, but didso on the CAM faces task, t ~21!� 3.51, p ,.005. However, when a t test was conductedon Time 2–Time 1 score differences, the im-provement of the intervention group was sig-nificantly greater than that of the AS0HFAcontrol group, t ~38!� 3.38, p, .005 . Table 2and Figure 3 show the mean scores of all tasksat Time 1 and Time 2.

Then, we computed score differencesfor the 20 CAM concepts ~Time 2 score mi-nus Time 1 score!. A multivariate ANOVA~MANOVA! was conducted for the 20 differ-ence scores, with group as the independentvariable. The MANOVA did not yield a sig-nificant group difference beyond emotional

concept, FWilks ~20, 19!� 1.36, ns, but yieldedsignificant individual between group effectsfor the concepts: grave, F ~1, 38!� 5.81, p ,.05, uneasy, F ~1, 38! � 5.2, p , .05, lured,F ~1, 38!� 6.98, p, .05, intimate, F ~1, 38!�5.70, p , .05, insincere, F ~1, 38! � 6.79,p , .05, and nostalgic, F ~1, 38! � 18.48,p , .001. These findings suggest that the in-tervention group’s recognition of these emo-tional concepts improved more than it did inthe AS0HFA control group. However, careshould be taken when interpreting these find-ings, due to the lack of an overall effect forthe MANOVA, and the multiple tests con-ducted under it.

Next, an analysis of the Film task scoreswas conducted, to test for holistic distant gen-eralization. This task was only used at Time 2,so differences could only be measured be-tween groups. Therefore, a one way ANOVAwas conducted on the three groups’ Film taskscores and was significant, F ~2, 62! � 7.68,p, .01. Preplanned contrasts with Bonferronicorrection revealed no difference between theintervention group ~M � 11.8, SD � 3.8! andthe AS0HFA control group ~M � 12.8, SD �3.4! for this task, t ~62!� 1.07, ns. However,the typical control group ~M � 15.5, SD �2.4! scored significantly higher than bothgroups, t ~62!� 3.81, p , .001. These results

Table 2. Experiment 1 means (standard deviations) of the three groups on all tasksat Time 1 and Time 2

SoftwareHome Usersa

AS0HFAControls

Time 1 Time 2 Time 1 Time 2TypicalControls

CAM face task ~max score � 50! 31.3 37.5 32.5 34.8 42.0~8.8! ~7.8! ~8.4! ~8.2! ~5.2!

CAM voice task ~max score � 50! 33.8 38.9 35.2 36.6 42.1~6.6! ~6.2! ~7.4! ~7.9! ~4.2!

CAM no. concepts recognized 9.8 13.6 10.5 11.3 16.1~max score � 20! ~5.2! ~4.8! ~5.2! ~5.4! ~3.0!

Reading the Mind in the Eyes 23.1 23.8 23.9 23.0 28.5~max score � 36! ~6.7! ~4.7! ~6.7! ~7.3! ~3.1!

Reading the Mind in the Voice 16.1 16.7 16.1 17.4 18.6~max score � 25! ~2.9! ~3.9! ~3.9! ~3.5! ~2.4!

Reading the Mind in Films 11.8 12.8 15.5~max score � 22! ~3.8! ~3.4! ~2.4!

aTwo participants of this group did not complete the CAM voice task at Time 1 and one participant did not completethe CAM faces at Time 2. These participants were excluded from the analysis of the three CAM scores. Other than that,groups’ sizes are identical to Table 1.


suggest the intervention group did not per-form better than the AS0HFA control groupon this level of generalization.

Last, a correlation analysis was conductedbetween the time participants had used thesoftware and the improvement scores of eachtask. Nonparametric analysis was used, due tothe small group size. Significant Spearmanrho correlations were found for software usagetime with improvement on the CAM voicetask ~rSpearman � .59, p, .01!, and with Read-ing the Mind in Film task scores ~rSpearman �.45, p , .05!. No other correlations weresignificant.

Discussion

In this experiment, the use of Mind Reading athome was compared to no intervention. Re-sults showed that following 10–20 hr of usingthe software over a period of 10–15 weeks,users significantly improved in their ability torecognize complex emotions and mental statesfrom both faces and voices, compared to theirperformance before the intervention, and rel-ative to the control group. This finding is in-teresting, considering the short usage time andthe large number of emotions included in thesoftware, and because participants were notasked to study these particular emotions.

Time 1 results confirmed the marked diffi-culties individuals with ASC have in recogniz-ing emotions from faces, voices, and eyes, andfrom social situations including the above withcontext and body language. These findings rep-licate results of previous studies ~Baron-Cohenet al., 1997; Baron-Cohen, Wheelwright, Hill,et al., 2001; Golan, Baron-Cohen, & Hill, inpress; Golan, Baron-Cohen, Hill, et al., in press;Heavey et al., 2000; Kleinman et al., 2001; Ruth-erford et al., 2002!. Because no differences werefound between the clinical groups at Time 1,any difference at Time 2 can be attributed tothe intervention.

The intervention group improved sig-nificantly on close generalization measures,including faces, voices, and the emotionsindividuals with AS0HFA had particular diffi-culties with ~Golan et al., in press!. Improve-ment in the ability of individuals with ASC torecognize mental states such as intimate, in-

sincere, or grave might have a positive effecton their confidence, willingness, and function-ing in interpersonal situations. This, togetherwith participants’ reports of greater attentionto faces and emotions, and improved eye con-tact, suggests that the analysis of emotionsusing Mind Reading allows people with ASCto improve emotion recognition skills fromboth faces and voices.

Improvement following the intervention waslimited to close generalization tasks, that is, tofaces and voices taken from Mind Reading.No difference was found between the inter-vention and AS0HFA control group on eitherfeature based or holistic tasks of distant gen-eralization. Similar findings of poor general-ization have been found in studies teachingtheory of mind, emotion recognition, and so-cial skills to individuals with ASC. These find-ings will be discussed in more detail later.However, software usage time was positivelycorrelated with film task scores, suggestingthat the more participants used the software,the higher they scored on the distant general-ization task. It is possible that a longer periodof usage would have led to improved general-ization among software users.

Interestingly, the control group signifi-cantly improved on the CAM face task be-tween Time 1 and Time 2, despite having noformal intervention. This could be the resultof taking the same task for a second time aftera relatively brief interval. In addition, wheninterviewed at the end of the second assess-ment, many participants in this group re-ported greater interest in emotions followingthe assessment at Time 1. Therefore, it is pos-sible that the assessment itself served as alimited short-term intervention, arousing par-ticipants’ awareness of the importance of facesand emotions. This new awareness was notsufficient to cause an improvement on voices,but did allow for improvement on faces, whichmight suggest this domain is more easilychanged through intervention. However, theimprovement in this group’s CAM face taskscores was significantly smaller than that ofthe software home users’ group.

Unlike our previous findings ~Golan et al.,in press!, verbal IQ was found to have a sig-nificant effect on CAM voice task scores. The


Figure 3. Mean scores ~with standard error bars! of the three groups on two levels of generalization for Experiment 1. @A color version of thisfigure can be viewed online at www.journals.cambridge.org#

605

inclusion of verbal content in the voice tasksegments of speech ~rather than just intona-tion! might account for this effect and for thelack of it in the face task. However, no sucheffect was found for Reading the Mind in theVoice ~Revised!, which also includes verbalcontent in its items. Experiment 2 was in partintended to help determine how central therole is that verbal IQ plays in the ability torecognize emotions.

Experiment 2

This experiment compared the effect of usingMind Reading with a weekly support of a tu-tor in group meetings to that of taking part ina 10-week social skills training course with-out using the software. Once again, improve-ment of emotion recognition skills in the threelevels of generalization was the target out-come measure. This experiment allowed us toexamine any extra value of using the softwareto that of attending group training for socialskills that had less if any systematic methodfor teaching emotion recognition.

Method

Design and instruments. The design and theinstruments used in this experiment were iden-tical to those of Experiment 1.

Participants. Three groups took part in this ex-periment: two AS0HFA intervention groups~who were independent from those in Experi-ment 1! and one typical ~general population!control group. Participants in the clinical groupshad all been diagnosed with AS0HFA in spe-cialist centers using established criteria ~Amer-ican Psychiatric Association, 1994; WorldHealth Organisation, 1994!. They were re-cruited via two support organizations and twocolleges for individuals withASC, where groupmeetings were also held. Because they were re-cruited via organizations that had volunteeredto help with the study, participants were not ran-domly allocated to the groups, but instead wereassigned by their recruiting organization.

Participants filled in the AQ ~Baron-Cohen,Wheelwright, Skinner, et al., 2001! to assesstheir self-reported level of autistic traits. Forty

percent of them scored above the cutoff of 32,a result that was unexpectedly low given ear-lier studies using the AQ ~Woodbury-Smithet al., 2005!, so a parent version of the AQ wassent to the parents or tutors of those who scoredbelow the cutoff to check if this reflected un-derreporting of symptoms by this group or ifdiagnosis was in question. All but threeparents0tutors returned the questionnaires.Using these reports, 70% of the participantsscored above the cutoff of 32 ~Baron-Cohen,Wheelwright, Skinner, et al., 2001! and allparticipants scored above the more sensitivecutoff of 26 ~Kurita et al., 2005; Woodbury-Smith et al., 2005!, an indicator that diagnosiswas reliable. Participants were accepted to thestudy if they had not participated in any re-lated intervention during the last 3 monthsand had no plans for engaging in another in-tervention while the study was ongoing. Therewere three groups in this experiment.

Software and tutor group. Thirteen partici-pants ~12 males, 1 female! were asked to useMind Reading alone for 2 hr0week, over a10-week period. In addition, these partici-pants attended 10 weekly sessions in smallgroups of up to six members. A tutor workedwith each group following a protocol that in-cluded analysis of features in different facialand vocal expressions of emotions, examplesof situations from participants’ everyday life,and the emotions they evoke, analysis ofemotions in pictures from newspapers andscenes from feature films and TV programs.Tutors were free to choose the materials theywanted to analyze in the group, and were askedto relate lessons to emotion groups in MindReading, to help associate the software withgroup activities and with everyday life. Eachof the tutors was given a copy of Mind Read-ing and was asked to become familiar withit before the course. Two of the tutors weresupport workers and one was a teacher, allexperienced in working with adults with ASC.Three such groups were run: two in supportcenters for individuals with ASC and one in acollege for adolescents and young adults withAS0HFA.

As in Experiment 1, participants were in-cluded if they completed a minimum of 10 hr


using Mind Reading. Out of 18 participants whooriginally started the course, 3 withdrew dur-ing the course, and 2 were excluded after fail-ing to complete the 10 hr of minimumuse. Two groups were given extra time, so thatparticipants could complete the minimum usagetime.As in Experiment 1, no pattern was foundfor the participants who withdrew or did notcomplete the program: their age range variedbetween 19 and 46; 2 of them had continuedstudying beyond compulsory education,whereas 3 did not; 1 was unemployed, 3 werestudents, and 1 was employed. Their IQ levelsandTime 1 assessment task scores ranged within1 SD from their group average. The reasons fortheir not having completed the program in-cluded not being able to commit to working withthe software for 2 hr0week due to their studiesor work. One participant had left the group af-ter falling in love with another participant andbeing rebuked.

Social skills course group. Thirteen partici-pants ~10 males, 3 females! took part in 10sessions of social skills training. Two courses,with 9 participants in each, were facilitated bya clinical psychologist who specializes in so-cial skill training for individuals with ASC,together with a staff member of the institutionwhere the training took place. The facilitatorsalso recruited the participants. The two groupsfollowed the same curriculum, which in-cluded themes such as conversation rules, emo-tional expressions and body language, jobinterviews, and friendship. The facilitators useda variety of techniques, such as stand-up teach-ing, group discussion, role play, and pictureanalysis. The facilitators were blind to the cur-riculum of the other research group and to thesoftware. In addition, the experimenters wereblind to the group’s curriculum until after theTime 2 assessment. Out of the 18 participantswho originally attended the course, 2 did notcomplete the course and 3 others were ex-cluded because their IQ scores fell below 70.

Typical (general population) control group.Thirteen participants ~10 males, 3 females! ofthe typical control group described in Experi-ment 1 were matched to the two AS0HFAgroups in this experiment.

A one-way ANOVA conducted on groups’AQ scores was significant for both self-report,F ~2, 36!� 16.98, p , .001, and parent0tutorreport, F ~2, 36! � 69.34, p , .001. Tukeypost hoc comparisons revealed that the twoclinical groups scored significantly higher onthe AQ compared to the typical control group,and did not differ from each other. Partici-pants were also asked to report any comorbiddiagnosis. Five participants in the softwareand tutor group and four participants in thesocial skills group had comorbid diagnosessuch as depression, learning difficulties orOCD. The two AS0HFA groups did not differon the proportion of participants with comor-bid diagnoses, x2 ~1!� 0.17, ns ~see Table 3!.

The three groups were matched on age, ver-bal and performance IQ, handedness, andgender. They spanned an equivalent range ofsocioeconomic classes and educational levels.No significant differences were found betweenthe groups for age, F ~2, 36!� 0.7, ns, verbalIQ, F ~2, 36! � 2.36, ns, performance IQ,F ~2, 36!� 2.70, ns, gender, x2 ~2!�1.39, ns,handedness, x2 ~2! � 2.44, ns, education,x2 ~2!� 0.22, ns, and employment, x2 ~2!�0.21, ns. In addition, no difference was foundbetween the two AS0HFA groups on thetime between the two assessment meetings,t ~17.3!�1.61, ns. Table 3 presents the groups’age and IQ data.

Procedure

Participants were tested at the local supportcenters and colleges for individuals with ASC.Participants were tested in groups of three bythe first author and one of three trained assis-tants. The assistants were blind to which groupparticipants belonged.

Participants of the software and tutor groupwere asked to help in the evaluation of a newintervention program. They were asked to com-mit to using Mind Reading for 2 hr0week overa period of 10 weeks and to attend all 10group sessions. Mind Reading was briefly in-troduced to the participants of this group atthe first assessment. The group tutors thenintroduced it in more detail at the first groupmeeting.


Participants of the social skills group weretold the study was to evaluate how social skillsgroups teach people to recognize emotions.They were asked to take part in the assess-ments at the beginning and the end of thecourse ~which was free of charge!. All partici-pants were told they were free to withdrawfrom the study at any time. In the first assess-ment some background information was ob-tained, followed by administration of theemotion recognition tasks. Participants wereseated in front of IBM compatible laptop com-puters with 15-in. monitors, and were givenheadphones for the voice tasks. The testingprocedure was similar to that of Experiment 1.The assessment meetings took about 3 hr. Dur-ing the intervention time between the two as-sessments, participants’ use of the softwarewas monitored by the tutors. The tutors werealso in charge of collecting the log files madeby the software to verify usage time. Averageusage time of the software in the tutor andsoftware group was 14.9 hr ~SD � 3.1, range �10–23!. In the second assessment, partici-pants of both intervention groups were asked

for their detailed feedback about the program.They were then debriefed about the aims anddesign of the study and were rewarded with acomplimentary copy of Mind Reading ~or wereallowed to keep the copy they used with nocharge!. This assessment meeting also tookabout 3 hr.

Results

For differences in group performance atTime 1, five one-way ANOVAs were con-ducted on the emotion recognition taskscores, using Holm’s sequential rejectiveBonferroni procedure. Significant differenceswere found between the groups for the CAMface task, F ~2, 36! � 9.76, p , .001, voicetask, F ~2, 36!� 5.64, p , .01, and the num-ber of emotional concepts recognized,F ~2, 36! � 7.77, p , .005; as well as forReading the Mind in the Eyes, F ~2, 36! �6.75, p , .005, and Reading the Mind in theVoice-R, F ~2, 36! � 4.99, p , .02. Pre-planned comparisons with Bonferroni correc-tions revealed no significant differences

Table 3. Means (standard deviations) and ranges of age, IQ, and AQ scores forthe intervention and control groups of Experiment 2

Software andTutor Group~N � 13!

Social SkillsGroup~N � 13!

TypicalControls~N � 13! F ~2, 36!

Age 25.5 ~9.3! 24.4 ~6.4! 25.5 ~9.6! 0.7017–50 17– 42 17–51

Verbal IQ 105.7 ~16.1! 96.5 ~15.5! 109.2 ~14.4! 2.3776–123 78–126 86–128

Performance IQ 103.9 ~19.8! 95.5 ~6.0! 106.3 ~6.0! 2.7271–129 87–105 92–117

AQ ~self-report! 25.1 ~7.1! 29.7 ~8.1! 13.9 ~5.7! 16.98*16–37 17– 45 8–26

AQ ~with parent0tutor reporta! 33.7 ~4.4! 34.2 ~4.9! 13.9 ~5.7! 69.34*27– 42 29– 45 8–26

x2 ~2!

Females ~%! 7.7 23.1 23.1 1.39Left handed ~%! 15.4 30.8 7.7 2.44Employed ~%! 38.5 38.5 46.2 0.21A levels or above ~%!b 38.5 30.8 38.5 0.22Comorbid diagnoses ~%! 38.5 30.8 NA 0.17

aParent0tutor filled AQ replaced self-report if the score was lower than the cutoff.bA levels are the first component of noncompulsory education in Britain.*p , .001. All other test results are not significant ~ p . .05!.


between the two clinical groups on any of thetask scores, and significantly higher scoresby the typical control group on all tasks, com-pared to the two AS0HFA groups. These find-ings support hypothesis 1. Means and standarddeviations of the groups’ emotion recognitionscores at Time 1 appear in Table 4.

Five MANCOVAs with repeated measureswere conducted to test for group differenceson the various tasks at Time 1 and Time 2.Verbal IQ was entered as a covariate. UsingHolm’s sequential rejective Bonferroni proce-dure, Time � Group interactions were foundfor CAM voices, FWilks ~1, 23! � 6.5, p ,.012, CAM number of concepts recognized,FWilks ~1, 23! � 6.04, p , .016, and Readingthe Mind in the Eyes, FWilks ~1, 23!� 8.4, p,.01, but not for CAM faces, FWilks ~1, 23! �0.23, ns, or Reading the Mind in the Voice-R,FWilks ~1, 23! � 0.11, ns. Verbal IQ had asignificant effect as a covariate on all tasks,beyond time, F ~1, 23!�17.89 for CAM faces,19.4 for CAM voices, 17.05 for CAM numberof concepts, 11.2 for Reading the Mind inthe Eyes, 10.0 for Reading the Mind in theVoice-R, p , .01 for all.

Simple main effects analyses with Bonfer-roni correction revealed the software and tu-tor group improved significantly from Time 1to Time 2 on CAM voice task, t ~12! � 4.65,p, .01, and CAM number of concepts recog-nized, t ~12! � 5.2, p , .001, whereas the

social skills group did not on either CAMvoices, t ~12! � 0.56, ns, or CAM number ofconcepts, t ~12!�1.72, ns. Simple main effectanalysis of the Time � Group interaction forReading the Mind in the Eyes with Bonferronicorrection revealed no significant time effectsfor either the software and tutor group, t ~12!�2.01, ns, or the social skills group, t ~12! ��2.5, ns. Because the strong effect of verbalIQ might have overshadowed any main effector interaction in the CAM faces scores, simplemain effect analyses were conducted for thetwo groups despite the lack of significant in-teraction. Paired t tests for CAM face scoresat Time 1 and Time 2 with Bonferroni correc-tion showed a significant improvement in thesoftware and tutor group, t ~12! � 4.2, p ,.005, but not in the social skills group, t ~12!�2.0, ns. Mean scores of all tasks at Time 1 andTime 2 can be found in Table 4 and Figure 4.

The MANOVA conducted for the 20 CAMconcept difference scores failed to reach sig-nificance, FWilks ~20, 5! � 1.16, ns. This mayhave been due to the small number of partici-pants in each group and the small range ofdifference scores for each emotional concept.Significant effects were found for two con-cepts only: vibrant, F ~1, 24!� 4.88, p , .05,and mortified, F ~1, 24!� 10.04, p , .01.

The ANOVA conducted for the film taskscores, testing for holistic distant generaliza-tion at Time 2 only, was significant, F ~2, 36!�

Table 4. Experiment 2 means (standard deviations) of the three groups on all tasksat Time 1 and Time 2

Software and Tutor Social Skills

Time 1 Time 2 Time 1 Time 2TypicalControls

CAM face task ~max score � 50! 32.3 36.2 26.8 29.3 40.8~8.1! ~8.9! ~9.7! ~9.5! ~6.0!

CAM voice task ~max score � 50! 33.2 38.9 31.1 31.8 41.1~9.1! ~7.6! ~9.1! ~10.9! ~5.2!

CAM no. concepts recognized ~max score � 20! 10.2 13.5 7.7 8.5 15.1~4.9! ~5.2! ~5.8! ~6.3! ~3.7!

Reading the Mind in the Eyes ~max score � 36! 21.6 23.8 21.5 19.2 28.2~6.3! ~4.2! ~5.6! ~6.8! ~3.6!

Reading the Mind in the Voice ~max score � 25! 15.1 16.2 13.9 14.7 18.1~2.8! ~3.5! ~4.5! ~4.6! ~2.8!

Reading the Mind in Films ~max score � 22! 11.9 10.5 14.8~3.7! ~3.2! ~2.6!

Note: N � 13 for every group.


Figure 4. Mean scores ~with standard error bars! of the three groups on two levels of generalization for Experiment 2. @A color version of thisfigure can be viewed online at www.journals.cambridge.org#

611

6.15, p , .01. However, preplanned contrastswith Bonferroni correction revealed no differ-ence between the software and tutor group~M � 11.9, SD � 3.7! and the social skillsgroup ~M � 10.5, SD � 3.2! for this task,t ~36! � 1.1, ns, although the typical controlgroup ~M � 14.8, SD � 2.6! scored signifi-cantly higher than both of the other groups,t ~36! � 3.33, p , .01. These results suggestthat as seen in Experiment 1, the AS0HFAgroup using Mind Reading did not performbetter than the AS0HFA group who did notuse it, at this level of generalization.

Nonparametric correlation analysis con-ducted for software usage time with task im-provement scores in the software and tutorgroup revealed significant correlations of soft-ware usage time with improvement on theCAM voice task ~rSpearman � .60, p, .05!, thenumber of CAM concepts correctly recog-nized ~rSpearman � .53, p , .05! and with filmtask scores ~rSpearman � .50, p, .05!. No othercorrelations were found significant.

Discussion

This experiment compared the use of MindReading at home with weekly support of atutor in group sessions to participating in asocial skills course without the use of the soft-ware. As in Experiment 1, the group using thesoftware improved significantly more than thecontrol group on two measures of close gen-eralization: recognition of voices and numberof emotional concepts recognized, both takenfrom Mind Reading. The close generalizationface task showed that the software and tutorgroup improved following training, whereasno improvement was found for the social skillsgroup. These findings confirm the value ofhome use of Mind Reading compared to theeffect of group meetings.

As in Experiment 1, software users failedto improve more than controls on feature-based distant generalization tasks and did notperform better than controls on the holisticdistant generalization level: the film task. Soft-ware usage time in the software and tutor groupwas correlated with greater improvement onCAM voices task and CAM’s number of emo-tions correctly recognized, as well as with

higher scores on the holistic film task. Thissuggests that longer use of the software leadsto improved generalization, although cautionshould be used when interpreting these re-sults, due to the small group size.

The most striking effect in this experimentwas that of Verbal IQ. Unlike Experiment 1,its effect was significant on all tasks beyondtime. Yet, despite this, the software and tutorgroup improved significantly more than thesocial skills group on the recognition of twoemotional concepts. The difference betweenthe two experiments in the potency of verbalIQ could be related to the IQ difference be-tween the groups in the two experiments.Although the groups in Experiment 1 had sig-nificantly above average intelligence, the twogroups in Experiment 2 had average ~and manylower than average! IQ, were recruited asgroups via colleges and support centers, andhad tutors take responsibility for their work.As a result of these differences, the groups ofExperiment 2 might have found the assess-ments more challenging and the words in thetasks more difficult. Although definition hand-outs were offered, it is possible that those whohave higher verbal IQ coped better with thetasks, as they were less distracted and stressedby the need to use the handouts. Future stud-ies could include nonverbal tasks in the eval-uation, to improve the validity of assessmentin lower functioning groups.

Another limitation of this experiment liesin the different number of hours the two groupshave received. The social skills group had noinput at home, to match the homework of theSoftware and tutor group. This difference inthe number of intervention hours, as well ashaving other subjects ~except for emotion rec-ognition! discussed in the social skills group,might have accounted for some of the improve-ment difference. Future studies should try tomatch the control intervention more carefullyin both time and curriculum.

Last, group size limits the power of find-ings in this experiment. This is related to prac-tical reasons of keeping the social skills groupand tutor and software groups small, to opti-mize learning conditions, as well as to dropout rate. Future studies should evaluate theuse of Mind Reading in group settings with


more participants to further validate findingsof this experiment.

General Discussion

This study evaluated a new intervention pro-gram for systematically teaching emotion rec-ognition in faces and voices to adults withASC. We investigated the effectiveness of in-dividual independent use of Mind Reading,and the effectiveness of using it in conjunc-tion with a tutor and group support. Results ofboth experiments revealed that the use of MindReading led adults with ASC to improve sig-nificantly in their emotion recognition skills.This improvement was achieved over a rela-tively short period of time, on a variety ofcomplex and socially important emotions andmental states, in both faces and voices. Theimprovement was also independent of ~or ad-ditional to! tutor and group support.

However, improvement following the useof the software in both experiments was lim-ited to different presentation and variations ontaught stimuli. Participants found it hard togeneralize their knowledge to other tasks ofemotion recognition from voices and eyes, anddid not perform better than controls on a taskinvolving integration of facial, vocal, and con-textual cues. Similar generalization problemshave been reported in previous studies attempt-ing to teach theory of mind, emotion recogni-tion, and social skills ~Bolte et al., 2002;Hadwin et al., 1996; Swettenham, 1996, 2000!.

Generalization difficulties have been re-ported to be characteristic of ASC ~Rimland,1965!. A focus on small details at the expenseof being able to see the larger picture ~Frith,1989!, abstraction difficulties, and insistenceon sameness may make generalization a chal-lenge for individuals with ASC. Various mod-els try to explain the generalization difficulties:adherence to rule-based categorization whilefailing to use prototype-based categorizations~Klinger & Dawson, 1995!; an inability torecognize the similarities between stimuli~Plaisted, 2001!; or the inability to deal withopen systems of the social world, instead fo-cusing on closed system, rule-based atomicphysical phenomena ~Klin, Jones, Schultz, &Volkmar, 2003; Lawson, 2003!.

Our own view is that reduced generaliza-tion is not so much a reflection of a deficit as areflection of the strong drive to systemize inpeople with ASC ~Baron-Cohen, 2002!. Goodsystemizing requires that one pays attention tothe small details between variables in case theseare important to understanding the workingsof the system. As such, a good systemizer re-sists grouping variables together until there isreliable evidence that these variables have nofunctional differences. To group them togetherrisks losing key information. With this view,generalization by individuals without ASC is asign of reduced systemizing, and reduced gen-eralization by individuals withASC is a sign oftheir talent at systemizing, as revealed on an-alytical tests ~Baron-Cohen, Wheelwright,Spong, et al., 2001, Folk Physics Test; Jolliffe& Baron-Cohen, 1997, Embedded FiguresTest!.Because the socioemotional world is difficultif not impossible to systemize, a systemizingstrategy for learning about emotions may leadto limited generalization.

Mind Reading was used as the interventionfor adults with ASC to exploit their goodsystemizing skills. As a systematic guide toemotions, Mind Reading trained recognitionin faces and voices separately. Our results showthat this was successful. Paradoxically, thisway of teaching encouraged an atomized learn-ing style leading to improvement, and made itharder to generalize to holistic material. Hence,we recommend that Mind Reading should beviewed as a first step in a training program.The next steps would need to deal with thesystematic introduction of context and integra-tion of different socioemotional cues into one~flexible! picture. Each step will need to beexplicitly connected to the previous ones, withthe main features pointed out, to ease gener-alization ~Ozonoff, 1995b!.

One side effect of the study was the rela-tively high dropout rate of participants in thesoftware using groups ~21% in Experiment 1,28% in Experiment 2!. The reasons individu-als gave for withdrawing were usually relatedto their inability to find the time to use thesoftware at home for 2 hr0week. Assumingthat this is not an extensive amount of time toask for, and bearing in mind that these partici-pants had joined the study willingly, while


knowing the work requirements, this findingcould relate to difficulties individuals withAS0HFA have prioritizing, planning aheadand adhering to goals. Such difficulties couldbe explained by the executive dysfunctiontheory of autism ~Ozonoff, 1995b; Russell,1997!, and suggest that even high-functioningadults with ASC may benefit from supportof this kind. Future studies will need to eval-uate the effect of possible executive dysfunc-tion on the ability of adults with ASC toindependently benefit from computer-basedinterventions.

In addition, as some of the participantspointed out, joining the study gave them anopportunity for desirable social contact with theexperimenters and their group. Whereas thisreason may have been secondary to their wishto acquire socioemotional skills, for some it mayhave been the main reason to join the study. Itis therefore possible that participants who with-drew from the study felt the social gain was notstrong enough to justify spending so much timeusing the software on their own. This againstrengthens the need to use Mind Reading inconjunction with group activities, to boost bothgeneralization and motivation.

The use of computer-based tasks in the eval-uation of learning and generalization in thisstudy has its limitations. Although such tasksallow for controlled and structured assess-ment of emotion recognition skills, testing dif-ferent modalities separately, they are quitedifferent to real life experience. Hence, therelevance of improvement among the soft-ware users in this study to real life functioningshould be considered with care. Indeed, someparticipants commented they found recogniz-ing emotions and mental states on the com-puter easier than doing this in real socialsituations, which requires cross-modal infor-mation processing and an immediate reactionin real time. This brings up the question of theappropriate evaluation of emotion recognitionabilities. It also raises the question of the rel-evance of emotion recognition training to ac-tual social functioning in adults with ASC.Most computerized emotion recognition andsocial skills programs were created for chil-dren with ASC. Whereas one was made foradolescents ~Parsons, Mitchell, & Leonard,

2005!, we are not aware of any other programmade to address the needs of adults with ASC.More research into such interventions and theirassociation with real life social functioning isrequired.

In both experiments, significant correla-tions were found between usage time and im-provement scores of some tasks. Of particularinterest was the positive correlation with theholistic film task, suggesting the use of thesoftware might be associated with distant gen-eralization measures. Research studies sug-gest there is both a developmental delay andlong lasting difficulties in mental state recog-nition by individuals with ASC ~Baron-Cohen,1995; Frith & Hill, 2004!. The present resultssuggest that learning aspects of empathizingskills ~such as the emotion recognition com-ponent! is possible by people with ASC eveninto adulthood and that improvement in suchareas is achieved when the intervention har-nesses their systemizing strengths. It may bethe case that even greater improvement wouldbe achieved if intervention was started at adevelopmentally earlier time point. We are cur-rently investigating the effectiveness of thismethod with children with ASC ~Golan &Baron-Cohen, 2006!.

The results of this study, as well as theparticipants’ reports of looking more at facesand engaging in more eye contact followingthe use of Mind Reading, call for neuroimag-ing studies to examine possible changes in thefunctioning of brain areas ~e.g., in the amyg-dala, fusiform gyrus, or prefrontal cortex!, andgaze tracking studies to examine subtle behav-ioral changes following the systematic studyof emotions using Mind Reading. Such stud-ies would throw light on whether the observedcognitive changes reported here are arisingfrom changes in those neural regions that aretypically recruited by the nonautistic brain, orif they are due to compensatory strategies byother neural regions. We conclude that thepresent experiments indicate that complexemotion–recognition can be improved over arelatively brief training period ~10–15 weeks!,when systematic methods such as Mind Read-ing are employed. Additional systematic meth-ods ~possibly over a longer time period! arerequired to improve generalization.


References

Adolphs, R. ~2001!. The neurobiology of social cogni-tion. Current Opinion in Neurobiology, 11, 231–239.

American Psychiatric Association. ~1994!. Diagnostic andstatistical manual of mental disorders ~4th ed.!. Wash-ington, DC: Author.

Ashwin, C., Baron-Cohen, S., Wheelwright, S., O’Riordan,M., & Bullmore, E. T. ~2005!. Differential activationof the social brain during emotion processing in adultswith autism. Manuscript submitted for publication.

Astington, J. W., Harris, P. L., & Olson, D. R. ~1988!.Developing theories of mind. Cambridge: CambridgeUniversity Press.

Baron-Cohen, S. ~1995!. Mindblindness: An essay on au-tism and theory of mind. Boston: MIT Press0BradfordBooks.

Baron-Cohen, S. ~2002!. The extreme male brain theoryof autism. Trends in Cognitive Sciences, 6, 248–254.

Baron-Cohen, S. ~2003!. The essential difference: Men,women and the extreme male brain. London: Penguin.

Baron-Cohen, S., Golan, O., Wheelwright, S., & Hill,J. J. ~2004!. Mind reading: The interactive guideto emotions. London: Jessica Kingsley Limited~www.jkp.com!.

Baron-Cohen, S., Leslie, A. M., & Frith, U. ~1986!. Me-chanical, behavioural and Intentional understandingof picture stories in autistic children. British Journalof Developmental Psychology, 4, 113–125.

Baron-Cohen, S., Richler, J., Bisarya, D., Gurunathan, N.,& Wheelwright, S. ~2003!. The Systemising Quotient~SQ!: An investigation of adults with Asperger Syn-drome or High Functioning Autism and normal sexdifferences. Philosophical Transactions of the RoyalSociety, Series B, 358, 361–374.

Baron-Cohen, S., Ring, H. A., Wheelwright, S., Bull-more, E. T., Brammer, M. J., Simmons, A., et al. ~1999!.Social intelligence in the normal and autistic brain:An fMRI study. European Journal of Neuroscience,11, 1891–1898.

Baron-Cohen, S., Wheelwright, S., Hill, J. 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-functioningautism. Journal of Child Psychology and Psychiatry,42, 241–251.

Baron-Cohen, S., Wheelwright, S., & Jolliffe, T. ~1997!.Is there a “language of the eyes”? Evidence fromnormal adults, and adults with autism or Aspergersyndrome. Visual Cognition, 4, 311–331.

Baron-Cohen, S., Wheelwright, S., Skinner, R., Martin,J., & Clubley, E. ~2001!. The Autism-Spectrum Quo-tient ~AQ!: Evidence from Asperger syndrome0high-functioning autism, males and females, scientists andmathematicians. Journal of Autism and Developmen-tal Disorders, 31, 5–17.

Baron-Cohen, S., Wheelwright, S., Spong, A., Scahill, V.,& Lawson, J. ~2001!. Are intuitive physics and intu-itive psychology independent? A test with childrenwith Asperger Syndrome. Journal of Developmentaland Learning Disorders, 5, 47–78.

Barry, T. D., Klinger, L. G., Lee, J. M., Palardy, N.,Gilmore, T., & Bodin, S. D. ~2003!. Examining theeffectiveness of an outpatient clinic-based social skillsgroup for high-functioning children with autism. Jour-nal of Autism and Developmental Disorders, 33,685–701.

Belmonte, M. K., Allen, G., Beckel-Mitchener, A., Bou-langer, L. M., Carper, R. A., & Webb, S. J. ~2004!.Autism and abnormal development of brain connec-tivity. Journal of Neuroscience, 24, 9228–9231.

Belmonte, M. K., Cook, E. H., Anderson, G. M., Ruben-stein, J. L., Greenough, W. T., Beckel-Mitchener, A.,et al. ~2004!. Autism as a disorder of neural informa-tion processing: Directions for research and targetsfor therapy. Molecular Psychiatry, 9, 646– 663. @Un-abridged edition at http:00www.cureautismnow.org0media03915.pdf#

Bernard-Opitz, V., Sriram, N., & Nakhoda-Sapuan, S.~2001!. Enhancing social problem solving in childrenwith autism and normal children through computer-assisted instruction. Journal of Autism and Develop-mental Disorders, 31, 377–398.

Bishop, J. ~2003!. The Internet for educating individualswith social impairments. Journal of Computer As-sisted Learning, 19, 546–556.

Bolte, S., Feineis-Matthews, S., Leber, S., Dierks, T.,Hubl, D., & Poustka, F. ~2002!. The development andevaluation of a computer-based program to test and toteach the recognition of facial affect. InternationalJournal of Circumpolar Health, 61~Suppl. 2!, 61– 68.

Boucher, J., Lewis, V., & Collis, G. M. ~2000!. Voiceprocessing abilities in children with autism, childrenwith specific language impairments, and young typ-ically developing children. Journal of Child Psychol-ogy and Psychiatry and Allied Disciplines, 41,847–857.

Celani, G., Battacchi, M. W., & Arcidiacono, L. ~1999!.The understanding of the emotional meaning of facialexpressions in people with autism. Journal of Autismand Developmental Disorders, 29, 57– 66.

Critchley, H. D., Daly, E. M., Bullmore, E. T., Williams,S. C., Van Amelsvoort, T., Robertson, D. M., et al.~2000!. The functional neuroanatomy of social behav-iour: Changes in cerebral blood flow when peoplewith autistic disorder process facial expressions. Brain,123, 2203–2212.

Deruelle, C., Rondan, C., Gepner, B., & Tardif, C. ~2004!.Spatial frequency and face processing in children withautism and Asperger syndrome. Journal of Autismand Developmental Disorders, 34, 199–210.

Ekman, P. ~1993!. Facial expression and emotion. Amer-ican Psychologist, 48, 384–392.

Ekman, P., & Friesen, W. ~1971!. Constants across cul-tures in the face and emotion. Journal of Personalityand Social Psychology, 17, 124–129.

Fein, D., Lucci, D., Braverman, M., & Waterhouse, L.~1992!. Comprehension of affect in context in chil-dren with pervasive developmental disorders. Journalof Child Psychology and Psychiatry, 33, 1157–1167.

Fernald, A. ~1992!. Human maternal vocalizations to in-fants as biologically relevant signals: An evolutionaryperspective. In J. H. Barkow, L. Cosmides, & J. Tooby~Eds.!, The adapted mind: Evolutionary psychologyand the generation of culture ~pp. 391– 428!. London:Oxford University Press.

Forster, K. L., & Forster, J. C. ~2003!. DMDX: A win-dows display program with millisecond accuracy. Be-havior Research Methods, Instruments, and Computers,35, 116–124.

Frith, U. ~1989!. Autism: Explaining the enigma. Oxford:Blackwell.


Frith, U., & Hill, E. ~2004!. Autism: Mind and brain.Oxford: Oxford University Press.

Golan, O., & Baron-Cohen, S. ~2006!. Teaching childrenwith Asperger syndrome and high functioning autismto recognize emotions using interactive multimedia.Unpublished manuscript.

Golan, O., Baron-Cohen, S., & Hill, J. ~in press!. TheCambridge Mindreading ~CAM! Face–Voice Battery:Testing complex emotion recognition in adults withand without Asperger Syndrome. Journal of Autismand Developmental Disorders.

Golan, O., Baron-Cohen, S., Hill, J. J., & Golan, Y. ~inpress!. “Reading the Mind in Films”—Testing recog-nition of complex emotions and mental states in adultswith and without autism spectrum conditions. SocialNeuroscience.

Golan, O., Baron-Cohen, S., Hill, J. J., & Rutherford,M. D. ~2005!. The “Reading the Mind in the VoiceTest—Revised”: A study of complex emotion recogni-tion in adults with and without autism spectrum con-ditions. Manuscript submitted for publication.

Grossman, J. B., Klin, A., Carter, A. S., & Volkmar, F. R.~2000!. Verbal bias in recognition of facial emotionsin children with Asperger syndrome. Journal of ChildPsychology and Psychiatry and Allied Disciplines,41, 369–379.

Hadwin, J., Baron-Cohen, S., Howlin, P., & Hill, K. ~1996!.Can we teach children with autism to understand emo-tions, belief, or pretence? Development and Psycho-pathology, 8, 345–365.

Happe, F. ~1994!. An advanced test of theory of mind:Understanding of story characters’ thoughts and feel-ings by able autistic, mentally handicapped, and nor-mal children and adults. Journal of Autism andDevelopmental Disorders, 24, 129–154.

Happe, F., Ehlers, S., Fletcher, P., Frith, U., Johansson,M., Gillberg, C., et al. ~1996!. “Theory of mind” inthe brain. Evidence from a PET scan study of As-perger syndrome. Neuroreport, 8, 197–201.

Haviland, J. M., & Lelwica, M. ~1987!. The induced af-fect response: 10-week-old infants’ responses to threeemotion expressions. Developmental Psychology, 23,97–104.

Heavey, L., Phillips, W., Baron-Cohen, S., & Rutter, M.~2000!. The Awkward Moments Test: A naturalisticmeasure of social understanding in autism. Journal ofAutism and Developmental Disorders, 30, 225–236.

Hetzroni, O. E., & Tannous, J. ~2004!. Effects of acomputer-based intervention program on the commu-nicative functions of children with autism. Journal ofAutism and Developmental Disorders, 34, 95–113.

Hobson, P. ~1994!. Understanding persons: The role ofaffect. In S. Baron-Cohen, H. Tager Flusberg, & D.Cohen ~Eds.!, Understanding other minds ~pp. 204–227!. London: Oxford University Press.

Hobson, R. P. ~1986a!. The autistic child’s appraisal ofexpressions of emotion. Journal of Child Psychologyand Psychiatry, 27, 321–342.

Hobson, R. P. ~1986b!. The autistic child’s appraisal ofexpressions of emotion: A further study. Journal ofChild Psychology and Psychiatry, 27, 671– 680.

Hobson, R. P., Ouston, J., & Lee, A. ~1988!. What’s in aface? The case of autism. British Journal of Psychol-ogy, 79, 441– 453.

Holm, S. ~1979!. A simple sequentially rejective multipletest procedure. Scandinavian Journal of Statistics, 6,65–70.

Howlin, P., Baron-Cohen, S., & Hadwin, J. ~1999!. Teach-

ing children with autism to mind-read: A practicalguide for teachers and parents. Chichester: Wiley.

Howlin, P., & Yates, P. ~1999!. The potential effective-ness of social skills groups for adults with autism.Autism, 3, 299–307.

Jolliffe, T., & Baron-Cohen, S. ~1997!. Are people withautism and Asperger syndrome faster than normal onthe Embedded Figures Test? Journal of Child Psychol-ogy and Psychiatry and Allied Disciplines, 38, 527–534.

Jolliffe, T., & Baron-Cohen, S. ~1999!. The Strange Sto-ries Test: A replication with high-functioning adultswith autism or Asperger Syndrome. Journal of Autismand Developmental Disorders, 29, 395– 406.

Kleinman, J., Marciano, P. L., & Ault, R. L. ~2001!. Ad-vanced theory of mind in high-functioning adults withautism. Journal of Autism and Developmental Disor-ders, 31, 29–36.

Klin, A., Jones, W., Schultz, R., & Volkmar, F. ~2003!.The enactive mind, or from actions to cognition: Les-sons from autism. Philosophical Transactions of theRoyal Society, 358, 345–360.

Klin, A., Jones, W., Schultz, R., Volkmar, F., & Cohen, D.~2002!. Visual fixation patterns during viewing of nat-uralistic social situations as predictors of social com-petence in individuals with autism. Archives of GeneralPsychiatry, 59, 809–816.

Klinger, L. G., & Dawson, G. ~1995!. A fresh look atcategorization abilities in persons with autism. In E.Schopler & G. Mesibov ~Eds.!, Learning and cogni-tion in autism ~pp. 119–136!. New York: Plenum Press.

Kurita, H., Koyama, T., & Osada, H. ~2005!. Autism-spectrum quotient—Japanese version and its shortforms for screening normally intelligent persons withpervasive developmental disorders. Psychiatry andClinical Neuroscience, 59, 490– 496.

Lawson, J. ~2003!. Depth accessibility difficulties: AnAlternative conceptualisation of autism spectrum con-ditions. Journal for the Theory of Social Behaviour,33, 189–202.

Lawson, J., Baron-Cohen, S., & Wheelwright, S. ~2004!.Empathising and systemising in adults with and with-out Asperger Syndrome. Journal of Autism and De-velopmental Disorders, 34, 301–310.

Loveland, K. A., Tunali Kotoski, B., Chen, R., & Brels-ford, K. A. ~1995!. Intermodal perception of affect inpersons with autism or Down syndrome. Develop-ment and Psychopathology, 7, 409– 418.

Loveland, K. A., Tunali Kotoski, B., Chen, Y. R., Orte-gon, J., Pearson, D. A., Brelsford, K. A., et al. ~1997!.Emotion recognition in autism: Verbal and nonverbalinformation. Development and Psychopathology, 9,579–593.

Macdonald, H., Rutter, M., Howlin, P., Rios, P., Le Con-teur, A., Evered, C., et al. ~1989!. Recognition andexpression of emotional cues by autistic and normaladults. Journal of Child Psychology and Psychiatry,30, 865–877.

Moore, D., McGrath, P., & Thorpe, J. ~2000!. Computer-aided learning for people with autism—A frameworkfor research and development. Innovations in Educa-tion and Training International, 37, 218–228.

Nieminen-von Wendt, T., Metsahonkala, L., Kulomaki, T.,Alto, S.,Autti, T., Vanhala, R., et al. ~2003!. Changes incerebral blood flow inAsperger syndrome during theoryof mind tasks presented by the auditory route. Euro-pean Child and Adolescent Psychiatry, 12, 178–189.

O’Riordan, M. A., Plaisted, K. C., Driver, J., & Baron-Cohen, S. ~2001!. Superior visual search in autism.


Journal of Experimental Psychology: Human Percep-tion and Performance, 27, 719–730.

Ozonoff, S. ~1995a!. Reliability and validity of the Wis-consin Card Sorting Test in studies of autism. Neuro-psychology, 9, 491–500.

Ozonoff, S. ~1995b!. Executive functions in autism. In E.Schopler & B. Mesibov ~Eds.!, Learning and cogni-tion in autism ~pp. 199–219!. New York: Plenum Press.

Parsons, S., & Mitchell, P. ~2002!. The potential of virtualreality in social skills training for people with autisticspectrum disorders. Journal of Intellectual Disabili-ties Research, 46, 430– 443.

Parsons, S., Mitchell, P., & Leonard, A. ~2005!. Do ado-lescents with autistic spectrum disorders adhere tosocial conventions in virtual environments? Autism,9, 95–117.

Pierce, K., Muller, R. A., Ambrose, J., Allen, G., &Courchesne, E. ~2001!. Face processing occurs out-side the fusiform “face area” in autism: Evidence fromfunctional MRI. Brain, 124, 2059–2073.

Plaisted, K. C. ~2001!. Reduced generalization in autism:An alternative to weak central coherence. In J. A.Burack ~Ed.!, The development of autism: Perspec-tives from theory and research ~pp. 149–169!. Mah-wah, NJ: Erlbaum.

Rajendran, G., & Mitchell, P. ~2000!. Computer mediatedinteraction in Asperger’s syndrome: The Bubble Dia-logue program. Computers and Education, 35, 189–207.

Rimland, B. ~1965!. Infantile autism: The syndrome andits implications for a neural theory of behavior. Lon-don: Methuen.

Russell, J. ~1997!. Autism as an executive disorder. Ox-ford: Oxford University Press.

Rutherford, M. D., Baron-Cohen, S., & Wheelwright, S.~2002!. Reading the mind in the voice: A study withnormal adults and adults with Asperger syndrome andhigh functioning autism. Journal Autism and Devel-opmental Disorders, 32, 189–194.

Rydin, O. T., Drake, J., & Bratt, A. ~1999!. The effects oftraining on emotion recognition skills for adults withan intellectual disability. Journal of Applied Researchin Intellectual Disabilities, 12, 253–262.

Schultz, R. T., Grelotti, D. J., Klin, A., Kleinman, J., Vander Gaag, C., Marois, R., et al. ~2003!. The role of thefusiform face area in social cognition: Implications

for the pathobiology of autism. Philosophical Trans-actions of the Royal Society London B: BiologicalSciences, 358, 415– 427.

Shah, A., & Frith, U. ~1993!. Why do autistic individualsshow superior performance on the block design task?Journal of Child Psychology and Psychiatry and Al-lied Disciplines, 34, 1351–1364.

Silver, M., & Oakes, P. ~2001!. Evaluation of a new com-puter intervention to teach people with autism or As-perger syndrome to recognize and predict emotions inothers. Autism, 5, 299–316.

Swettenham, J. ~1996!. Can children with autism be taughtto understand false belief using computers? Journalof Child Psychology and Psychiatry, 37, 157–165.

Swettenham, J. ~2000!. Teaching theory of mind to indi-viduals with autism. In S. Baron-Cohen, H. Tager-Flusberg, & D. J. Cohen ~Eds.!, Understanding otherminds: Perspectives from developmental cognitiveneuroscience ~2nd ed., pp. 442– 456!. Oxford: OxfordUniversity Press.

Teunisse, J. P., & De Gelder, B. ~1994!. Do autistics havea generalized face processing deficit? InternationalJournal of Neuroscience, 77, 1–10.

Wechsler, D. ~1999!. Wechsler Abbreviated Scale of Intel-ligence (WASI). San Antonio, TX: The PsychologicalCorporation.

Wellman, H. M. ~1992!. The child’s theory of mind. Cam-bridge, MA: MIT Press.

Woodbury-Smith, M. R., Robinson, J., Wheelwright, S.,& Baron-Cohen, S. ~2005!. Screening adults for As-perger syndrome using the AQ: A preliminary studyof its diagnostic validity in clinical practice. Journalof Autism and Developmental Disorders, 35, 331–335.

World Health Organisation. ~1994!. International Classi-fication of Diseases ~10th ed.!. Geneva: Author.

Yirmiya, N., Sigman, M. D., Kasari, C., & Mundy, P.~1992!. Empathy and cognition in high-functioningchildren with autism. Child Development, 63, 150–160.

Young, A. W. ~1998!. Face and mind. Oxford: OxfordUniversity Press.

Zhang, J., Quan, H., Ng, J., & Stepanavage, M. E. ~1997!.Some statistical methods for multiple endpoints inclinical trials. Controlled Clinical Trials, 18, 204–221.


regular article systemizing empathy: teaching...

Documents