See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/7298531

The Cambridge Mindreading (CAM) Face-Voice Battery: Testing Complex

Emotion Recognition in Adults with and without Asperger Syndrome

Article  in  Journal of Autism and Developmental Disorders · March 2006

DOI: 10.1007/s10803-005-0057-y · Source: PubMed

CITATIONS READS
205 1,937

3 authors:

Ofer Golan Simon Baron-Cohen

Bar Ilan University University of Cambridge
67 PUBLICATIONS   2,146 CITATIONS    810 PUBLICATIONS   76,225 CITATIONS   

SEE PROFILE SEE PROFILE

Jacqueline J Hill
University of Exeter
28 PUBLICATIONS   4,027 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Mixed methods analysis for health services research View project

ASC Inclusion View project

All content following this page was uploaded by Simon Baron-Cohen on 22 May 2014.

The user has requested enhancement of the downloaded file.

Journal of Autism and Developmental Disorders, Vol. 36, No. 2, February 2006 (Ó 2006)
DOI 10.1007/s10803-005-0057-y
Published Online: February 14, 2006

The Cambridge Mindreading (CAM) Face-Voice Battery:

Testing Complex Emotion Recognition in Adults with and
without Asperger Syndrome

Ofer Golan,1,2,* Simon Baron-Cohen,1 and Jacqueline Hill1

Adults with Asperger Syndrome (AS) can recognise simple emotions and pass basic theory of
mind tasks, but have difficulties recognising more complex emotions and mental states. This
study describes a new battery of tasks, testing recognition of 20 complex emotions and mental
states from faces and voices. The battery was given to males and females with AS and matched
controls. Results showed the AS group performed worse than controls overall, on emotion
recognition from faces and voices and on 12/20 specific emotions. Females recognised faces
better than males regardless of diagnosis, and males with AS had more difficulties recognising
emotions from faces than from voices. The implications of these results are discussed in
relation to social functioning in AS.

KEY WORDS: Emotion recognition; complex emotions; face perception; voice perception; theory of
mind; Asperger Syndrome; autism spectrum; psychometrics.

The ability to understand people’s emotional and spectrum, and be clinically useful in diagnosis and
other mental states, also known as ‘‘theory of mind’’ treatment. The present study reports a new test
or ‘‘mindreading’’ (Wellman, 1992) underlies funda- battery to assess recognition of complex mental states.
mental social skills. This ability is also sometimes The standard test of emotion recognition (Ekman
referred to as empathising (Baron-Cohen, 2002; & Friesen, 1971) focuses on the ability to recognise six
Baron-Cohen, 2003). Individuals with autism spec- basic emotions that are recognised cross-culturally:
trum conditions have difficulties understanding happiness, sadness, anger, fear, surprise, and disgust.
others’ mental (including their emotional) states, Some studies report that people with autism spectrum
which is thought to play a major role in their social conditions have difficulties recognising such emotions
and communicational difficulties (Baron-Cohen, from photographs (Bormann-Kischkel, Vilsmeier, &
1995). The measurement and analysis of such diffi- Baude, 1995), films (Yirmiya, Sigman, Kasari, &
culties may shed light on the phenotype of the autistic Mundy, 1992) and voice recordings (Loveland, Tunali
Kotoski, Chen, & Brelsford, 1995) as well as with
1
matching faces and voices of these emotions (Bu-
Department of Psychiatry, Autism Research Centre, Cambridge
University, Douglas House, 18b Trumpington Road, CB2 2AH,
itelaar, Van der Wees, Swabb Barneveld, & Van der
Cambridge, UK. Gaag, 1999; Hobson, 1986a, b). However, other
2
Correspondence should be addressed to: Ofer Golan, Department studies (Grossman, Klin, Carter, & Volkmar, 2000;
of Psychiatry, Autism Research Centre, Cambridge University, Loveland et al., 1997) have found that children and
Douglas House, 18b Trumpington Road, CB2 2AH, Cambridge, adolescents with high functioning autism or Asperger
UK; Tel: 44 (0)1223 746062; Fax: 44 (0)1223 746033; e-mail:
[email protected]
Syndrome (AS) have no difficulties in basic emotion

169
0162-3257/06/0200-0169/0 Ó 2006 Springer ScienceþBusiness Media, Inc.
170 Golan, Baron-Cohen, and Hill

recognition. Adults with autism spectrum condi- strategy, which may mask their deficits under
tions may also not differ from typically developed certain circumstances.
controls in the ability to recognise the 6 basic For these reasons, in order to examine mental
emotions (Adolphs, Sears, & Piven, 2001). These state recognition abilities in people with autism
groups also passed first-and second-order theory of spectrum conditions of normal intelligence, there is
mind tasks, which could be taken to indicate no a need for more a fine grained test battery, examining
theory of mind difficulty in autism. more subtle emotions.
This lack of difficulty in basic emotion recogni- To address this need, the Strange Stories Test was
tion can be explained in different ways. For example, developed (Happe, 1994). This assesses the ability to
one can distinguish situation-based emotions (such as provide context-appropriate mental state explana-
sadness and happiness) from belief-based emotions tions for non-literal statements made by story char-
(such as surprise) (Baron-Cohen, Spitz, & Cross, acters (e.g. ironic or sarcastic statements). Happe’s
1993). According to the theory of mind hypothesis, study with adolescents found specific deficits on this
individuals with autism have specific deficits in task, a result that was later replicated with adults with
recognising belief-based emotions, compared to situ- AS or HFA (Jolliffe & Baron-Cohen, 1999).
ation-based emotions. The latter study found that, A different approach involves the ‘‘Reading the
compared to learning disabled and typically develop- Mind in the Eyes’’ test, which measures the ability to
ing controls, matched on mental age, children with identify complex mental states from still pictures of the
autism had no difficulty in recognising sadness or eye region (Baron-Cohen, Wheelwright, & Jolliffe,
happiness, but they found it harder to recognise 1997; Baron-Cohen, Wheelwright, Hill, Raste, &
surprise. Similarly, children with high functioning Plumb, 2001). This approach has been extended in
autism (HFA) found it harder to describe situations two studies, using tasks in which the participant is
in which they felt embarrassed and to recognise required to identify mental states from recordings of
embarrassment (a belief-based emotion) from photo- brief spoken phrases (Kleinman, Marciano, & Ault,
graphs, relative to a matched task involving sadness 2001; Rutherford, Baron-Cohen, & Wheelwright,
or happiness (Capps, Yirmiya, & Sigman, 1992). 2002). All these tasks showed that individuals with
These findings support the distinction between ‘‘sit- AS or HFA, despite having at least normal intelli-
uational’’ and ‘‘cognitive’’ emotions (Harris, 1989). gence, have specific difficulties in identifying subtle
The lack of difficulty in passing first- and mental states.
second-order theory of mind tasks by people with However, there are a few methodological and
HFA and AS can be explained by the fact this conceptual issues that these tasks do not address:
competence is usually achieved by typically develop-
1. The ‘‘Reading the Mind in the Eyes’’ (the
ing children as young as 6 years old. The fact that
Eyes task) and the ‘‘Reading the Mind in the
people with AS or HFA pass these tasks may simply
Voice’’ (the Voice task) tasks test different
indicate that their understanding of mental states is at
complex emotions and mental states, but they
least equal to that of an average 6 year-old.
do not systematically examine the recognition
The success with basic emotion recognition by
of particular emotions. Rather, these tests
people with AS may also reflect that such
include a variety of stimuli, which generate
individuals (who have normal intelligence) may be
one overall score of emotion recognition.
using compensation strategies to bypass their earlier
2. The Eyes task, like most visual tasks of
difficulties. For example, Grossman et al. (2000)
emotion recognition, uses still pictures rather
showed children with AS pictures of 5 ‘‘basic’’
than motion. The result is a relatively narrow
emotions, with matching or mismatching labels.
range of emotions which can be studied, as
The children with AS had no problem recognising
the distinction between many emotions re-
these emotions or identifying the emotions when
quires motion (e.g. relief). In addition, this
labelled with matching words, but had difficulties in
makes the test less naturalistic, and therefore
recognising the emotions in the mismatching labels
may not assess an individual’s actual ability to
condition. This result suggests that, instead of
identify emotions from moving faces.
recognising the emotions in the face, children with
3. No tasks have studied understanding of
AS were using the written label to answer the
complex mental states via both visual and
question. The authors concluded that individuals
auditory channels. This has been tested with
with AS use verbal mediation as a compensatory
The Cambridge Mindreading (CAM) Face-Voice Battery 171

the basic emotions (Hobson, 1986a, b; Love- (Baron-Cohen, Hill, Golan, & Wheelwright, 2002).
land et al., 1995). Every film and audio recording of every emotion was
validated, using a panel of 10 volunteers from the
In the construction of The Cambridge Mind-
general population. If at least 8 out of 10 judges agreed
reading (CAM) Face-Voice Battery, we address these
that the label given to a face or voice could be a
issues. The test was designed to assess the wide
reasonable description of the particular emotion, the
emotional repertoire of adults and to examine each
film or voice recording was included in the database.
emotion thoroughly through both visual and audi-
This database is used in a piece of educational
tory modalities, using motion in the visual task.
software, for teaching emotion recognition (for
The test is based on a new taxonomy of emotion
details, see www.jkp.com/mindreading).
(Baron-Cohen, Golan, Wheelwright, & Hill, 2004).
The CAM battery evaluates a selection of 20
This model comprises 412 unique emotion concepts,
emotion concepts, taken from the above mentioned
including all the emotion terms in the English
taxonomy, representing 18 of the 24 emotion groups.
language, as well as epistemic mental states with an
This covers a good variety of emotions and mental
emotional dimension (e.g. doubting). Mental states
states, while remaining relatively brief for adminis-
that could be a purely bodily state (e.g. hungry) and
tration. Coverage of all 412 emotion concepts would
epistemic states with no emotional dimension (e.g.
have made the test itself unfeasible in a single session.
reasoning) are not included.
The battery includes two tasks: emotion recognition
These 412 emotions are grouped into 24 mutu-
in the face and emotion recognition in the voice. Each
ally exclusive emotion groups (such as the happy
of these tasks has fifty questions, in which the
group, the thinking group, the sneaky group, etc). A
participant is either watching 3–5 second silent clips
list of the 24 emotion groups is shown in Fig. 1.
of actors portraying an emotion (facial task), or
Developmental testing resulted in the 412 emotion
listening to short sentences, spoken in a particular
concepts being further subdivided into six different
emotional intonation (vocal task). After watching the
levels, on the basis of word frequency in the English
clip/listening to the voice recording, the participant is
language and verbal comprehension. The six levels
presented with four adjectives and is asked to
represent an age range from preschool through to
‘‘choose the word that best describes how the person
adulthood. The full list of emotions, according to
is feeling’’.
emotion groups and developmental levels can be
In order to make sure that the chosen concepts
found elsewhere (Baron-Cohen, Golan, Hill &
are taken from the adult emotional repertoire, they
Wheelwright, submitted; Baron-Cohen et al., 2004).
were all selected from the higher levels of the
Using this taxonomy, and together with a multi-
taxonomy; 6 concepts from level 4 (concepts under-
media production company, we created a multimedia
stood by typical 15–16 year olds), 13 concepts from
database on DVD-ROM, in which actors of both
level 5 (understood by typical 17–18 year olds) and
sexes, different age groups and ethnicities enacted
one concept from level 6 (words understood by less
each of the 412 emotions. These were captured using
than 75% of typical 17–18 year olds). A detailed
video of their facial expression, and using audio of their
description of the level validation process is available
vocalisation. Each of the 412 emotions therefore
elsewhere (Baron-Cohen et al., submitted).
has 6 films and 6 voice recordings expressing it
We included mental states that are ‘‘positive’’ in
valence, such as empathic and intimate, as well as
Afraid Excited Liked Surprised concepts that are ‘‘negative’’, such as guarded and
insincere. An attempt was also made to include
Angry Fond Romantic Thinking emotions of varying intensity, i.e.—subtle emotions
on the one hand (e.g.—uneasy, subdued) and intense
Bored Happy Sad Touched
ones on the other (e.g.—appalled, mortified). Con-
Bothered Hurt Sneaky Unfriendly cepts were coded for intensity and valence when
creating the taxonomy by 3 independent judges
Disbelieving Interested Sorry Unsure (Baron-Cohen et al., submitted). In total, the CAM
battery has 5 ‘‘positive’’ concepts, 12 ‘‘negative’’, and
Disgusted Kind Sure Wanting
3 ‘‘neutral’’. 3 concepts were coded as having high
Fig. 1. The 24 groups included in the emotion taxonomy (adapted intensity. The twenty concepts with their level,
from Baron-Cohen et al., 2004). valence and intensity coding are listed in Table I.
172 Golan, Baron-Cohen, and Hill

Table I. The 20 Emotional Concepts Included in the CAM, their Emotion Group, Developmental Level, Valence and Intensitya

Concept Emotion Group Level Valence Intensity

Appalled Surprised 4 Negative Strong
Appealing (asking for) Wanting 4 Neutral
Confronted Hurt 5 Negative
Distaste Disgusted 5 Negative Strong
Empathic Kind 5 Positive
Exonerated Happy 6 Positive
Grave Sad 5 Negative
Guarded Disbelieving 4 Negative
Insincere Sneaky 5 Negative
Intimate Romantic 4 Positive
Lured Interested 5 Neutral
Mortified Sorry 5 Negative Strong
Nostalgic Touched 5 Neutral
Reassured Liked 4 Positive
Resentful Unfriendly 5 Negative
Stern Unfriendly 5 Negative
Subdued Sad 5 Negative
Subservient Unsure 5 Negative
Uneasy Afraid 4 Negative
Vibrant Excited 5 Positive

Note: aconcepts with high emotional intensity are marked as ‘‘strong’’.

Each of the 20 emotion concepts tested was greater than 20 is above chance at the p < .01
expressed by 5 examples (or items). Our criterion for level (Binomial test).
passing a concept was correct recognition of at least 4 4. Concepts correctly recognised: The concepts
out of 5 items. Achieving 4 or more out of 5 would can be studied in two ways: The sum of con-
represent above chance recognition of the concept cepts correctly recognised, ranging from 0 to
(Binomial test, p < .05). Of the 20 concepts, 10 20 (any score greater than 2 is above chance at
concepts were measured by the following combina- the p < .01 level, according to Binomial test),
tion of 5 items: 3 faces and two voices. The other 10 or the particular concepts correctly answered,
concepts were measured by a variation on this: 3 analysed individually/by subgroups
voices and two faces. This design allowed us to keep (e.g.—positive/negative).
the task brief overall.
In this study, we predicted that participants with
There are four different scores that can be
AS would show deficits, relative to matched controls,
derived from the CAM:
on all 4 scores above.
1. An overall emotion recognition score: defined Another comparison we carried out was between
as the sum of all the correctly answered the ability to recognise emotions from the face and
questions, ranging from 0 to 100, describing from the voice. Such a comparison is possible since
overall emotion and mental state recognition. the CAM includes the same emotions in the facial
Any score greater than 35 is above chance at and vocal scale. We also tested for a group difference
the p < .01 level (Binomial test). on the face-voice comparison (i.e. a group by
2. Facial emotion recognition score: defined as modality interaction).
the sum of all items answered correctly from Lastly, we predicted that CAM scores would
the facial items, ranging from 0 to 50. Any positively correlate with the Revised ‘‘Reading the
score greater than 20 is above chance at the Mind in the Eyes’’ test (Baron-Cohen et al., 2001)
p < .01 level (Binomial test). and with a revised version of the ‘‘Reading the Mind
3. Vocal emotion recognition score: defined as the in the Voice’’ test (Golan, Baron-Cohen, Rutherford,
sum of all items answered correctly in the & Hill, submitted), which includes 4 foils for each
vocal items, ranging from 0 to 50. Any score question instead of two in Rutherford et al.’s original
The Cambridge Mindreading (CAM) Face-Voice Battery 173

task (Rutherford et al., 2002). The CAM scores were significant differences between the groups at the
also predicted to correlate negatively with scores of p < .05 level (see Table II).
the Autism Spectrum Quotient (AQ) (Baron-Cohen,
Wheelwright, Skinner, Martin, & Clubley, 2001). Instruments
The CAM: Test Development
METHOD
Twenty-five concepts were selected from the
emotion taxonomy (Baron-Cohen et al., submitted).
Participants
Six of the twenty-five were taken from level 4, 15
Group 1 comprised twenty-one adults (15 males from level 5 and four from level 6, which are all adult
and 6 females), aged 18–50 (Mean age = 30.2, levels. A pilot study carried out with adult partici-
SD = 10.5). Participants had all been diagnosed pants with AS using the basic emotions resulted in
with Asperger Syndrome in specialist centres using ceiling effects. This confirmed that choice of the adult
established criteria (APA, 1994; World-Health-Orga- level emotion concepts was appropriate.
nisation, 1994). They were recruited from several Selection of the 25 concepts followed 3 princi-
sources, including a local clinic for adults with AS, ples: concepts should be (a) selected from all 24
support organisations, and colleges for individuals emotion groups, (b) mainly subtle, and (c) important
with autism spectrum conditions around the UK. All for everyday social functioning. Although most of the
participants were given the Wechsler Abbreviated 24 emotion groups were represented in the CAM by
Scale of Intelligence (WASI), comprising the vocab- one emotion concept, the larger emotion groups
ulary, similarities, block design and matrix reasoning (unfriendly and sad) were represented by 2 emotion
tests. The WASI produces verbal, performance and concepts each. For each concept, six test questions
full scale IQ scores, with correlations of .88, .84 and were created, using a computer programme which
.92, respectively, with the full Wechsler scales randomly selected three video films and three audio
(Wechsler, 1999). All participants scored above 70 voice recordings out of the six existing on the
on both verbal and performance scales. Mean verbal multimedia database, and matched them with foil
IQ score was 114.4 (SD = 9.02) and mean perfor- words from levels 4 and 5 of the emotion taxonomy,
mance IQ score was 107.71 (SD = 15.95). ensuring that foils were not from the same emotion
Group 2 comprised 21 adults recruited from a group as the target answer. Although choosing foils
local employment agency. After screening for autistic from other emotion groups possibly makes the tasks
spectrum conditions using the AQ (Baron-Cohen easier, it was believed that foils taken from the same
et al., 2001), four participants were excluded for categories as targets might be too similar and increase
scoring above cut-off of 32. The remaining seventeen, the dependency of performance on verbal ability (i.e.
12 males and 5 females, matched the clinical group in the ability to distinguish one emotion from another
age (range = 18–51, mean age = 27.1, SD = 10.3) within the same group). Some of the groups might
verbal IQ (mean = 118.47, SD = 7.42) and perfor- still provide quite difficult foils to other emotion
mance IQ (mean = 110.82, SD = 8.57). They groups with a similar valence and theme. For
spanned an equivalent range of socio-economic example—emotions from the unfriendly group were
classes and educational levels as that seen in the used with targets from the angry group, emotions
clinical group. Chi-square test for sex and t-tests for from the sad group were used with targets from the
age, verbal and performance IQ revealed no hurt group, etc.

Table II. Means, Standard Deviations and Ranges of Chronological Age and WASI Scores for the AS and Control Groups

AS group (n = 21) Control group (n = 17)

Mean SD Range Mean SD Range t

Age 30.23 10.50 17.9–49.9 27.08 10.25 17.6–51.2 .93
VIQ 114.38 9.02 100–127 118.47 7.42 104–130 1.50
PIQ 107.71 15.95 72–134 110.82 8.57 92–125 .72
FSIQ 112.48 12.26 84–131 116.29 6.71 106–128 1.22

Note: p > .1 for all t-tests.

174 Golan, Baron-Cohen, and Hill

Two tasks (face recognition and voice recogni- available for the participants at the beginning of the
tion) were created and run on an IBM compatible assessment. Participants were encouraged to go
computer, using the experimental software DMDX through it and make sure they were familiar with
(Forster & Forster, 2003). Each task started with an all the words, and to use the handout in case of any
instruction slide, followed by two practice items, to doubt during the task. Hence, response time was
ensure understanding of the task. In both tasks the unrestricted and not measured.
test items were presented in a random order. The face Item validation was conducted before carrying
task comprised silent clips of adult actors, both male out any group analysis. The data from the 21 adults
and female, and of different ethnicities, expressing the in the control group was first analysed as follows: An
emotions in the face (though with shoulders visible). item was considered valid if at least 11 out of 21
Example questions showing one frame from two of (>50%) of these participants selected the target word
the clips are shown in Fig. 2. The voice task and no more than 6 (<33%) selected any one of the
comprised recordings of short sentences expressing foils. Using these criteria, 5 of the concepts (3 from
various emotional intonations. An attempt was made level 6 and 2 from level 5) were excluded from the
to use as neutral content as possible for the voice battery. Eight other concepts had one invalid item
recordings (e.g. ‘‘Where did you get that?’’), so that each, and these items were removed. In order to keep
recognition had to rely primarily on intonation rather the same number of items for all the concepts and to
than semantic content. keep the battery brief to administer, one item was
In both tasks four adjectives, numbered from 1 then randomly removed from each of the remaining
to 4, were presented after playing each stimulus. 12 concepts, so that the final battery comprised 20
Participants were asked to press 1, 2, 3 or 4 on a concepts with 5 items for each concept. The number
keyboard to select their preferred answer. After of concepts having 3 face and 2 voice items or 3 voice
choosing an answer the next item was presented. and 2 face items were counterbalanced.
No feedback was given during the task. A handout of Autism Spectrum Quotient (AQ, Baron-Cohen
definitions of all the adjectives used in the task was et al., 2001): The AQ is a self-report questionnaire,
which measures the degree to which any adult of
normal IQ possesses traits related to the autistic
spectrum. Scores range from 0 to 50, and the higher
the score, the more autistic traits a person possesses.
Reading the Mind in the Eyes Task [Revised,
adult version] (Baron-Cohen et al., 2001): The task
has 36 items, in which participants are presented with
a photograph of the eyes region of the face and must
choose one of four adjectives or phrases to describe
the mental state of the person pictured. A definition
handout is provided at the beginning of the task and
a practice item precedes the first trial. In the present
study, the pictures and adjectives were presented on
the computer screen (using DMDX software), to
avoid possible difficulties due to communication with
a human examiner (Ozonoff, 1995). Items were
presented in a random order. There was no time
limit for answering.
Reading the Mind in the Voice Task (Revised).
We used a revised version of the original task
(Rutherford et al., 2002). In the original task, 40
segments of speech, taken from BBC drama series,
were played on audio tape to the participants, who
were asked to choose one out of two possible
answers, describing the speaker’s mental state in each
Fig. 2. Examples of questions from the emotion recognition in the item. The task was revised as follows: Each of the test
face task (showing only one frame out of the full clip). items was allocated two more foils, taken from the
The Cambridge Mindreading (CAM) Face-Voice Battery 175

same level, one level above or one level below the the independent variables. Verbal IQ, performance
correct answer (based on the emotion taxonomy). IQ, and age were entered as covariates.
Foils were selected to match the content of the A main effect of group was found in all four
verbalisations but not the intonation, thus making analyses. The scores of the AS group were signifi-
the task harder to answer. This avoided ceiling effects cantly lower than those of the control group on the
that the original version of the test is prone to. Seven facial scale (F [1,31] = 15.61, p < .001), the vocal
items were removed since the authors found the scale (F [1,31] = 21.26, p < .001), the overall score
correct answer inappropriate to the verbalisation. (F [1,31] = 25.32, p < .001) and the number of
Eight more items were excluded after validation by a concepts passed (F [1,31] = 15.67, p < .001).
sample of 15 typically developing adults. The final In addition, a main effect of sex was found for
task included 25 items with 4 possible answers for the facial scale (F [1,31] = 5.02, p < .05). Females
each of them (For a full description of the revision were significantly better than males at recognising
and validation, see Golan et al., submitted). The test emotions from faces in both groups, regardless of
items were ‘‘cleaned’’ from background noise as far diagnosis. No other effect or covariate contribution
as possible and played on a computer (using DMDX was significant. The means and standard deviations
software) in random order, preceded by an instruc- of the 4 scores for males and females in the AS and
tion slide and two practice items. Participants were control group are shown in Table III and in Fig. 3.
again given a definition handout before the beginning
of the task. There was no time limit for answering. Table III. Facial, Vocal and Overall Mean Scores and Number of
Concepts Passed for Males and Females in the AS & Control
Groups
Procedure
Participants were tested either at the Autism ASc Controlc Total
Research Centre in Cambridge, or at local support a. Facial scale (Max = 50)
centres and colleges for individuals with autism Females Mean 35.33 46.00 40.18b
spectrum conditions. Participants were seated in SD 6.95 2.55 7.60
front of IBM compatible computers with 15 inch Males Mean 31.13 42.50 36.19b
SD 8.24 4.17 8.78
monitors and were given headphones for the voice Total Mean 32.33a 43.53a 37.34
tasks. The CAM took about 45 minutes to complete. SD 7.96 4.03 8.55
b. Vocal scale (Max = 50)
Females Mean 33.17 44.80 38.45
SD 9.62 2.59 9.27
RESULTS Males Mean 36.73 41.92 39.04
SD 4.22 3.96 4.81
A calculation of facial, vocal and overall scores Total Mean 35.71a 42.76a 38.87
SD 6.19 3.78 6.29
was made by counting the number of correct answers
c. Overall score (Max = 100)
in each of the scales and across the whole battery. Females Mean 68.50 90.80 78.64
Furthermore, a tally of concepts correctly recognised SD 15.76 3.63 16.28
was made. A minimum of 4 correct answers out of 5 Males Mean 67.87 84.42 75.22
items was considered successful recognition of a SD 10.32 5.85 11.92
Total Mean 68.05a 86.29a 76.21
concept.
SD 11.69 5.99 13.19
All participants scored above chance (p < .01, d. Number of concepts passed (Max = 20)
Binomial test) on the vocal scale and on the overall Females Mean 11.50 18.40 14.64
score. All except for two participants from the AS SD 5.68 .89 5.43
group scored above chance on the facial scale Males Mean 10.47 16.08 12.96
SD 4.32 2.50 4.56
(p < .01). These two participants scored just below
Total Mean 10.76a 16.76a 13.45
the threshold (20 correct answers out of 50), one of SD 4.62 2.39 4.81
whom scored above chance at the p < .05 level. They
were therefore included in the analysis. Notes: aTotal difference between AS and control groups is signifi-
Univariate analyses of variance (ANOVA) were cant for all 4 measures (p < .001).
b
Total difference between males and females is significant for the
performed on the facial scale score, the vocal scale facial scale (p < .05).
score, the CAM overall score and the number of c
Group sizes: 6 females and 15 males in the AS group (total=21). 5
concepts passed. Group (AS/control) and Sex were females and 12 males in the control group (total=17).
176 Golan, Baron-Cohen, and Hill

(a) found (Fwilks[1,31] = 5.06, p < .05), suggesting that

Facial and Vocal scales (Max=50)
50 females found it easier to recognise emotions from
45 faces than from voices, whereas males showed the
40 43.53 42.76
opposite preference. However, the analysis of simple
35
Mean score

30 35.71 main effects was not significant.

32.33
25 A three-way interaction of modality by group by
20 sex was also found to be significant (Fwilks [1,31] =
15 4.45, p < .05). Analysis of simple main effects
10
5 revealed that while there were no differences between
0 recognition of emotions from faces and voices in the
Facial scale Vocal Scale control group and among females with AS, there was
indeed a significant difference among males with AS
(b) CAM overall score (Max=100) (t[14] = 2.69, p < .05), who found the recognition
100
90 of emotions from faces (M = 31.1, SD = 8.24)
80 86.29 significantly harder than recognition from voices
70 (M = 36.7, SD = 4.22).
Mean score

60 68.05 Next, in order to compare the recognition of

50 individual concepts in the two groups, a 18 by 2
40
multivariate analysis of variance was performed for
30
20
the proportions of participants who correctly recog-
10 nised each concept in the two groups. Due to the
0 large number of dependent variables and since groups
Overall score were matched on sex, age and IQ, these variables
were not included. Two concepts—uneasy and
(c) Number of concepts passed (Max=20) appalled, which all of the participants of the control
18
group passed, were excluded from the analysis. The
Average no. of concepts

16 16.76
14 analysis yielded a significant overall effect of group
12 (Fwilks[18,19] = 2.60, p < .05). Individual concept
10 analyses revealed that the AS group scored signifi-
10.76
8 cantly lower than the control group in the recognition
6 of 11 of the 18 concepts. These concepts were:
4 intimate and reassured from level 4 of the taxonomy;
2 distaste, insincere, lured, mortified, nostalgic, resentful,
0 subservient and grave of level 5; and exonerated of
No. of concepts passed
level 6. The two concepts, which were excluded from
AS group Control group the analysis, were then analysed separately with a
goodness of fit test. The proportion of participants
Fig. 3. Mean scores of AS and control groups on the four mea- with AS who correctly recognised uneasy was signif-
sures of the CAM. icantly lower than that of the control group
(v2[1] = 8.2, p < .01). There was no difference
To test for differences between the vocal and between the groups in the recognition of appalled
facial scale scores among the groups (modality effect) (v2[1] = 0.831, N.S.). A Wilcoxon signed ranks test
and between them (a modality by group interaction), was held to check the difference between proportion
a multivariate ANOVA for repeated measures was of concepts passed from levels 4 and 5 of the
performed with the CAM facial and vocal scale taxonomy (level 6 was excluded since it was repre-
scores as the dependent variables, group (AS/control) sented by one concept only). No significant difference
and sex as independent variables and verbal IQ, was found for either group (Z = .47 for the AS
performance IQ and age as covariates. No effect of group and Z = .12 for the control group. p > .5 for
modality was found (Fwilks[1,31] = 2.63, n.s.), nor an both). The proportion of participants of the two
interaction of group by modality (Fwilks[1,31] = 0.39, groups who passed each of the 20 concepts is shown
n.s.). A significant interaction of modality by sex was on Table IV.
The Cambridge Mindreading (CAM) Face-Voice Battery 177

Table IV. Proportion of Participants who Correctly Recognised (a = .001) the respective power scores were: .972
the CAM 20 Concepts (facial), .761 (vocal), .933 (number of concepts
recognised), and .993 (overall).
AS group Control group
Concept (n = 21) % (n = 17) %
An examination of the correlation matrix, shown
a
in Table V reveals that as predicted, the CAM is
Appalled 95.2 100.0 strongly and positively correlated with the ‘‘Reading
Appealing (asking for) 52.4 76.5
Confronted 71.4 82.4
the Mind in the Eyes’’ task (Baron-Cohen et al., 2001)
Distaste 57.1 94.1 ** and the revised version of the ‘‘Reading the Mind in
Empathic 76.2 76.5 the Voice’’ Task (Golan et al., submitted; Rutherford
Exonerated 33.3 94.1 ** et al., 2002). All of the CAM’s scores correlated
Grave 42.9 82.4 * positively with these external criteria. Unsurprisingly,
Guarded 42.9 52.9
Insincere 28.6 88.2 **
the facial scale had a stronger correlation with the
Intimate 42.9 94.1 ** Eyes task (r = .74, p < .001) compared to the Voice
Lured 42.9 82.4 * task (r = .49, p < .01). Similarly, the vocal scale of
Mortified 66.7 94.1 * the CAM correlated more strongly with the Voice task
Nostalgic 66.7 94.1 * (r = .62, p < .001) than with the Eyes task (r = .32,
Reassured 42.9 82.4 *
Resentful 61.9 94.1 *
p < .05). The two scales of the CAM maintained a
Stern 47.6 52.9 strong correlation between themselves (r = .57,
Subdued 66.7 70.6 p < .001), as they did with the overall score and the
Subservient 28.6 70.6 ** number of correctly recognised concepts.
Uneasy 61.9 100.0 **a All of the CAM’s measures were negatively
Vibrant 81.0 94.1
correlated with the AQ score, which means that the
Notes: **p < .01. more autism spectrum characteristics one possesses,
*p < .05. the lower one’s CAM scores. All correlations of the
a
The analysis of these concepts was done using goodness of fit test CAM scores with IQ or age of the participants were
due to a ceiling effect in the control group.
found to be non-significant, which suggests that the
CAM measures are independent of both verbal and
performance IQ, as well as chronological age.

Power calculations for the different scales (with

a = 0.01) show that the scales are quite powerful in DISCUSSION
differentiating the group with AS from the control
group. Power scores were 1-b=.996 for the facial This study introduces the CAM—a new bat-
scale, .939 for the vocal scale, .989 for the tally of tery for testing recognition of complex emotions
correctly recognised concepts and .999 for the CAM and mental states in the face and the voice. It
overall score. With a more conservative alpha level expands previous work that found difficulties

Table V. Correlation of the CAM Scores with Each Other, with External Criteria and with Background Parameters

No. of concepts
CAM—overall Facial scale Vocal scale passed
Facial scale .920**
Vocal scale .847** .571**
No. of concepts passed .975** .904** .816**
Reading the mind in the eyes .632** .737** .324* .647**
Reading the mind in the voice-R .611** .489** .616** .609**
AQ ).574** ).472** ).563** ).512**
AGE ).039 ).122 .084 ).037
Verbal IQ .235 .257 .143 .271
Performance IQ .186 .184 .140 .176

Notes: **Correlation is significant at the 0.01 level (2-tailed).

*Correlation is significant at the 0.05 level (2-tailed).
178 Golan, Baron-Cohen, and Hill

in this domain among adults of normal intelli- almost everything, responses do of course require
gence, diagnosed with autism spectrum conditions some minimal inhibition (the ability to go through all
(Baron-Cohen et al., 1997; Baron-Cohen, Jolliffe, the answers and choose the best one) and working
Mortimore, & Robertson, 1997; Baron-Cohen memory.
et al., 2001; Rutherford et al., 2002). However, A review of the mental states with which the AS
unlike previous studies, the CAM battery allowed group had significant difficulties reveals no clear
a test of the recognition of specific emotions and pattern: Of the concepts included in the CAM, the
mental states as well as overall performance, and groups did not find positive emotions easier to
recognition in the two perceptual channels recognise than negative ones; concepts known to
separately. It also tested recognition of complex 15–16 year olds were not recognised more easily than
emotions and mental states using films of faces concepts known to adults. There was even no
rather than still pictures. preference for recognition of more intense emotions
Results showed that individuals with Asperger in comparison to more subtle ones: a group difference
Syndrome (AS), when compared to general popula- was found for distaste and mortified which were both
tion controls, had more difficulties in recognising marked as ‘‘strong’’, whereas other, more subtle,
mental states from both faces and voices. In addition, mental states such as subdued or empathic did not
participants with AS recognised fewer mental state produce a group difference. However, emotion
concepts then controls. In twelve out of the twenty valence and subtlety were not systematically studied
emotions and mental states tested in the CAM, a in the CAM, and these could be studied in their own
significantly lower number of participants with AS right in future studies.
successfully recognised the concept, compared to age- The lack of a developmental level effect may be
and IQ-matched controls. attributed to the AS group members having a verbal
The fact that controls were matched on chrono- IQ within the normal range. In other words, individ-
logical, verbal and nonverbal mental age, and the uals who are older than 18 (as the participants of this
lack of correlations between the CAM scores and study were) would be expected to be familiar with
these factors, suggests the independence of complex these concepts. An alternative explanation is the
emotions and mental state recognition from verbal relative similarity in complexity between level 4 and
and non-verbal ability. The strong negative correla- level 5 concepts (which the CAM comprises). In order
tion of the CAM scores with the participants’ Autism to test for an effect of level, i.e. differential recogni-
Spectrum Quotient (AQ) score (Baron-Cohen et al., tion of complex and simpler mental states, the task
2001) supports the relevance of emotion and mental should compare concepts from the higher levels (4–6)
state recognition difficulties in high-functioning with lower levels (1–3). A comparison of this kind,
adults with autism spectrum conditions. These using still pictures of faces and pictures of the
are important components of empathising (Baron- eyes and mouth region (Baron-Cohen et al., 1997)
Cohen, 2003). Despite their ability to recognise basic revealed that whilst adults with autism and AS were
emotions, such adults still find it hard to ‘‘mindread’’ able to detect basic mental states in the whole face,
complex mental states from faces and voices. The they were impaired at recognising complex mental
relatively high correlation of the CAM facial and states, and especially at recognising such mental
vocal scales with the ‘‘Reading the Mind in the Eyes’’ states from the eyes alone.
(Baron-Cohen et al., 2001) and ‘‘Reading the Mind in Despite the small number of female participants
the Voice-R’’ (Golan et al., submitted) tasks, respec- in this study, a sex difference was found on the facial
tively, provides the task with important measures of scale in this study, and this was independent of
external validity. The CAM goes beyond these two diagnosis: females recognised emotions in faces
tasks by using motion in the facial scale items and by significantly better than males. This has been found
allowing the opportunity to analyse individual in typically developed controls using the ‘‘Reading
concepts. Its power levels show it is sensitive to the Mind in the Eyes’’ (Baron-Cohen et al., 2001) or
group differences across all scales and scores. the ‘‘Profile for Non verbal Sensitivity’’ (Baron-
The CAM tests recognising emotions indepen- Cohen, 2003; Hall, 1984). The lack of studies
dent of weak central coherence (Frith, 1989) or investigating females with autism spectrum
executive function (Ozonoff, Pennington, & Rogers, conditions calls for a thorough investigation of their
1991) because there is minimal context or planning, profile. The absence of a sex difference on the vocal
which burden working memory. However, like scale mirrors the lack of such a difference on the
The Cambridge Mindreading (CAM) Face-Voice Battery 179

‘‘Reading the Mind in the Voice’’ task (Rutherford difficulty in matching incongruent faces and emotion
et al., 2002). labels among children with AS (Grossman et al.,
The reason for the difference between facial and 2000). It is possible that on the CAM items for
vocal emotion recognition when testing sex differ- insincere, the participants with AS were distracted by
ences might actually derive from the linguistic nature the content of what was said in the vocal items, rather
of the vocal task. Our results show that males with than judging the intonation, which caused them to
AS had higher vocal scores than their facial ones. It make errors. Similarly, they might have used more
might be that they used the content to guess the obvious cues in the face (e.g. the fact that the person
emotions, to compensate for their difficulties in in the film was smiling) to answer the facial items
emotion recognition (Grossman et al., 2000). while missing other, more subtle facial cues (e.g. gaze
Although the dependent variables in this study that was incongruent with the smile). An example of
measured number of correctly recognised emotions, it such an error is the preference of the distracter
is nevertheless interesting to examine the errors made ‘‘spellbound’’ when asked to recognise insincerity in
by the groups. Due to the wide variety of emotions one of the face items. The participants might have
used as distracters, it was not possible to locate interpreted the smile and avoiding gaze as a spell-
specific error patterns. However, there were some bound rather than an insincere emotional state. Since
interesting anecdotal examples of errors made only deception is tested in the false belief tasks (which
by participants in the AS group. These errors were adults with AS passed with no difficulty), the CAM
mainly about missing subtleties in face items and might provide a more subtle alternative to these
answering according to the content while ignoring the tasks.
intonation in voice items. Though in most cases Another emotional concept the AS group had
chosen distracters were of the same valence as targets, particular difficulty recognising was subservient. For
in some items participants in the AS group preferred example, two thirds of the AS group preferred the
a cross valence distracter upon the target answer (i.e. label ‘‘miffed’’ for a subservient face item (comparing
preferred a negative distracter when target was to 11% in the control group). This could reflect their
positive and vice versa). All examples of the errors confusion between dominant and submissive charac-
quoted in the discussion section below were chosen ters in a social situation. Since dominance hierarchies
by at least 30% of the participants with AS. are widely recognised in social primates (De Waal,
When looking at the group differences on 1998) it is surprising that people with AS should find
individual concept recognition, the deficit among this emotion difficult. However, it may be that non-
the AS group in recognising insincerity is most human primates rely on other cues to judge domi-
striking. Less than a third of the AS group members nance and subservience (e.g. physical size or success
recognised this mental state successfully, which sup- in conflict). It is likely that people with AS would
ports evidence from other studies, showing specific have no difficulty in understanding social hierarchies
difficulty in understanding deception (Baron-Cohen, from non-emotional cues (such as rank). It may
1992; Sodian & Frith, 1992). One study found that therefore be that their deficit arises only when the
high-functioning adults with autism had difficulties cues are from emotional expression. Such misunder-
assessing the trustworthiness and approachability of standing of hierarchical human systems and social
people from pictures of their faces (Adolphs et al., relations might, for example, lead to the use of an
2001). Their participants tended to evaluate people in inappropriate attitude towards authority.
the photos as being more trustworthy and approach- A similar problem might arise for the misunder-
able than controls’ evaluations. Understanding of standing of intimacy—another clearly interpersonal
hidden intent, masked behind an incongruent facial emotion. The AS group had difficulties spotting the
expression, is a major area of difficulty for individuals interpersonal aspect in this mental state. More than
with autistic spectrum conditions. In her book 40% of participants with AS mislabelled intimate face
Thinking in Pictures, Temple Grandin described her items as ‘‘determined’’ and ‘‘carefree’’. Similarly,
experience as a high functioning adult with autism: 30% of them mislabelled an intimate voice item as
‘‘It is easy for me to understand the concept of ‘‘subservient’’, possibly relying on its content (‘‘ask
deception when it involves playing tricks...but under- me anytime’’) while not picking up the intonation. It
standing the social cues that indicate an insincere is easy to imagine how such a cognitive deficit could
person is much more difficult’’ (Grandin, 1995, lead to difficulties in relationships and to difficulties
p. 137). Other studies have also reported a specific distinguishing genuine closeness and interest on one
180 Golan, Baron-Cohen, and Hill

hand, from politeness or even boredom in an of subtle fear. Our study is consistent with these
encounter on the other. The AS group’s difficulties findings. Future studies should investigate amygdala
in recognising resentful, grave or mortified also may activity in individuals with AS not only when
reflect longstanding difficulties in reading emotions watching but also when listening to fear items of
and mental states. The lack of group difference for different intensity (e.g. uneasy, afraid, terrified).
mental states like guarded or stern may reflect these Another significant difference in recognition
items being hard for both groups, since both groups between the AS and control groups was with the
scored relatively low on them. Further examination concept distaste from the disgusted group in the
of these mental states and the groups they represent taxonomy. This high-intensity concept, was recogni-
will be required. This applies especially to guarded sed by nearly all of the controls (94.1%) but only by a
and its group disbelieving due to its relevance to little more than half of the participants with AS
theory of mind and the understanding of deception. (57.1%). More than a third of the participants in the
Another mental state which the AS group had AS group mislabelled a face item of this emotion as
significant difficulties with was exonerated. A third of ,offended’ and two thirds of them mislabelled a voice
the participants in the AS groups mistook a face item item as ‘‘battered’’ (‘‘you’ve done it again’’). Surpris-
of exonerated for ‘‘remote’’, which is a cross valence ingly, not many studies have studied recognition of
distracter. Similarly, they confused the positive exon- disgust. Those which have, found difficulties in
erated for a negative label, ‘‘resigned’’ in a voice item, disgust recognition among severely autistic adoles-
again using the content only (‘‘now I can get down to cents with mental retardation (Hobson, Ouston, &
work’’). This concept was the only remaining repre- Lee, 1988) and no difficulty in its recognition among
sentative of level 6 in the taxonomy (emotion words high-functioning adults with autistic spectrum con-
which were known to less than 75% of adult ditions (Adolphs et al., 2001). The group difference
participants in the validation study (Baron-Cohen we found in the recognition of faces and voices
et al., submitted)). Nevertheless, analysis revealed no expressing distaste (which is slightly more subtle than
significant effect of verbal IQ in the ability to disgust) suggests that even high-functioning individ-
recognise this emotion. This mental state (which uals with AS might have problems recognising this in
concerns relief from blame, like reassured which also faces and voices. So far, no brain mapping studies
had a significant group effect), requires the use of a have assessed the recognition of disgust by individ-
theory of mind for its interpretation, which might uals with autism spectrum conditions. However,
have made its recognition especially hard for the AS studies of patients with insula and basal ganglia
group. lesions report specific difficulties with the recognition
One of the emotional concepts which was and experience of disgust, suggesting brain specificity
recognised by all members of the control group but for this emotion (Calder, Keane, Manes, Antoun, &
only 62% of the AS group is uneasy. This emotion, Young, 2000; Calder, Lawrence, & Young, 2001).
expressing very subtle degrees of fear, is a good Interestingly, there was no difference between
example of the difficulty individuals with AS might the AS and control groups in the recognition of
have with picking up subtle emotional cues. A new empathy (the concept empathic), and this was not
fMRI study of the amygdala, a key brain area predicted. The reason for this might have been the
underlying the detection if fear in others, involved easy foils which appeared with this concept. It will be
showing pictures of people expressing fear at different interesting to explore other emotions in the kind
intensities to individuals with AS and controls group, to test this concept more thoroughly. Equally
(Ashwin, Baron-Cohen, Wheelwright, O’Riordan, & surprising was the lack of difference in recognising
Bullmore, submitted). Amygdala activation among mental states such as appalled, vibrant, confronted and
the participants with AS was significantly reduced subdued. These results suggest that in such adults with
relative to controls. Furthermore, the AS group AS, their ‘‘mindblindness’’ is by no means total.
showed no difference in amygdala activation between The CAM presents both visual and vocal stimuli,
intense, moderate and mild degrees of fear. This including motion in the face task and requiring word
study supports the amygdala theory of autism labels for the emotion. As these different kinds of
(Adolphs et al., 2001; Baron-Cohen et al., 2000; stimuli activate different brain areas, one might
Howard et al., 2000) and shows that the brain in wonder which of these underlies the difficulties of
high-functioning adults with autistic spectrum con- the AS group. Such studies were mainly done with
ditions does not respond differentially to expressions typically developed and with brain damaged
The Cambridge Mindreading (CAM) Face-Voice Battery 181

participants. In a recent study assessing a patient with dynamic emotion recognition using the videos from
bilateral amygdala lesions, Adolphs and colleagues the CAM in a functional MRI experiment involving
found that while the patient could not recognise the 6 adults with and without autism.
‘‘basic’’ emotions from still faces, he could easily The CAM includes only 20 out of the full set of
recognise them when they were expressed by a live 412 emotions and mental states in the emotion
model. They argued that still and motion emotional taxonomy. Our team is currently working on a
stimuli activate different areas of the brain: while the version for children (Golan & Baron-Cohen, in
fusiform gyrus, as well as the posterior and superior preparation). Future studies using the taxonomy
temporal cortex and the amygdala and insula are and the multimedia database could explore each
involved in recognising emotions from still faces, particular emotion group in depth, and compare
emotion recognition from faces in motion is related certain emotions in different intensities. Temple
to middle temporal and middle superior temporal Grandin mentioned the visual collection of social
areas of the brain, together with parietal and frontal ‘‘clips’’ she holds in her mind for reference, which
sectors (Adolphs, Tranel, & Damasio, 2003). help her cope with interpersonal situations (Grandin,
Kilts and colleagues conducted a PET imaging 1995). Matching facial and vocal expressions of
study assessing intensity of anger and happiness from mental states could be used to teach individuals with
still and video stimuli. They found that judgment of autism spectrum conditions to improve their emotion
anger in dynamic expressions was associated with recognition. We are currently engaged in such an
increased right-lateralised activity in the medial, intervention study.
superior, middle, and inferior frontal cortex and We conclude by suggesting that this new test of
cerebellum, while judgments of happiness were asso- subtle emotion recognition in the face and the voice
ciated with relative activation of the cuneus, temporal reveals that adults with HFA or AS have residual
cortex, and the middle, medial, and superior frontal difficulties in recognising complex emotions and
cortex. In contrast, the perception of anger or mental states, independent of IQ, language, central
happiness in static facial expressions activated a coherence, or executive function.
motor, prefrontal, and parietal cortical network
(Kilts, Egan, Gideon, Ely, & Hoffman, 2003).
As with visual stimuli, there are different find- ACKNOWLEDGMENTS
ings with regards to emotion recognition from the
voice. In a study involving over a hundred partici- OG was supported by the Corob Charitable
pants with focal brain damage, Adolphs and col- Trust, the Cambridge Overseas Trust and the
leagues found that lesions in right cortical regions National Alliance for Autism Research (NAAR).
impair recognition of emotion in prosody. Phillips SBC and JH were supported by the Shirley Founda-
and colleagues found in a functional MRI study that tion, Medical Research Council (MRC), and the
the amygdala, as well as the superior temporal gyrus, Three Guineas Trust. We are grateful to our col-
were involved with recognition of fear and disgust in leagues at Red Green and Blue Ltd, Autism West
both visual and vocal stimuli (Phillips et al., 1998). Midlands, Autism London, European Services for
However, Adolphs and Tarnel found no difficulty in People with Autism (ESPA), and to Chris Ashwin,
recognising emotions from prosody among amygdala Sally Wheelwright, Yael Golan and Sarah Johnson.
lesioned patients (Adolphs & Tranel, 1999). Evi-
dently, this calls for further inquiry of brain struc-
tures involved in emotion recognition from voices. REFERENCES
The above mentioned studies assessed emotion
recognition using the 6 ‘‘basic’’ emotions. Further Adolphs, R., Sears, L., & Piven, J. (2001). Abnormal processing of
social information from faces in autism. Journal of Cognitive
investigation into activation of brain regions in Neuroscience, 13, 232–240.
recognition of complex emotions from still, dynamic Adolphs, R., & Tranel, D. (1999). Intact recognition of emotional
and vocal stimuli, such as those that appear in the prosody following amygdala damage. Neuropsychologia, 37,
1285–1292.
CAM is still required. Since this study was purely Adolphs, R., Tranel, D., & Damasio, A. R. (2003). Dissociable
behavioural, no conclusions could be drawn of neural systems for recognizing emotions. Brain and Cognition,
relevant brain regions involved in the recognition of 52, 61–69.
American Psychiatric Association. (1994). DSM-IV Diagnostic and
the CAM’s different emotions. Work is under way in statistical manual of mental disorders (4th ed.). Washington
our lab to investigate the neural basis underlying DC: American Psychiatric Association.
182 Golan, Baron-Cohen, and Hill

Ashwin, C., Baron-Cohen, S., Wheelwright, S., O’Riordan, M., Forster, K. L., & Forster, J. C. (2003). DMDX: A windows display
& Bullmore, E. T. (submitted). Differential activation of the program with millisecond accuracy. Behavior Research Meth-
social brain during emotion processing in adults with ods Instruments & Computers, 35, 116–124.
autism. Frith, U. (1989). Autism: Explaining the enigma. Oxford: Blackwell.
Baron-Cohen, S. (1992). Out of sight or out of mind? Another look Golan, O., Baron-Cohen, S., Rutherford, M. D., & Hill, J. J.
at deception in autism. Journal of Child Psychology and Psy- (submitted). The Reading the Mind in the Voice test–Revised:
chiatry, 33, 1141–1155. a study of complex emotion recognition in adults with and
Baron-Cohen, S. (2002). The extreme male brain theory of autism. without Autism Spectrum Conditions.
Trends in Cognitive Sciences, 6, 248–254. Grandin, T. (1995). Thinking in pictures. New York: Doubleday.
Baron-Cohen, S., Wheelwright, S., & Jolliffe, T. (1997). Is there a Grossman, J. B., Klin, A., Carter, A. S., & Volkmar, F. R. (2000).
‘‘language of the eyes’’? Evidence from normal adults, and Verbal bias in recognition of facial emotions in children with
adults with autism or Asperger syndrome. Visual Cognition, 4, Asperger syndrome. Journal of Child Psychology and Psychi-
311–331. atry and Allied Disciplines, 41, 369–379.
Baron-Cohen, S., Wheelwright, S., Skinner, R., Martin, J., & Hall, J. A. (1984). Non-verbal sex differences: Communication,
Clubley, E. (2001). The Autism-Spectrum Quotient (AQ): accuracy and expressive style. Baltimore; London: Johns
Evidence from Asperger syndrome/high-functioning autism, Hopkins University Press.
males and females, scientists and mathematicians. Journal of Happe, F. (1994). An advanced test of theory of mind: Under-
Autism and Developmental Disorders, 31, 5–17. standing of story characters’ thoughts and feelings by able
Baron-Cohen, S. (1995). Mindblindness: An essay on autism and autistic, mentally handicapped, and normal children and
theory of mind. Boston: MIT Press/Bradford Books. adults. Journal of Autism and Developmental Disorders, 24,
Baron-Cohen, S. (2003). The Essential Difference: Men, women and 129–154.
the extreme male brain. London: Penguin. Harris, P. L. (1989). Children and emotion: The development of
Baron-Cohen, S., Golan, O., Hill, J. J., & Wheelwright, S. (sub- psychological understanding. Oxford: Blackwell.
mitted). Developmental changes in the emotion lexicon: a Hobson, R. P. (1986a). The autistic child’s appraisal of expressions
research note. of emotion. Journal of Child Psychology and Psychiatry, 27,
Baron-Cohen, S., Golan, O., Wheelwright, S., & Hill, J. J. (2004). 321–342.
Mindreading: The interactive guide to emotions. London: Jes- Hobson, R. P. (1986b). The autistic child’s appraisal of expressions
sica Kingsley Limited (www.jkp.com). of emotion: A further study. Journal of Child Psychology and
Baron-Cohen, S., Hill, J. J., Golan, O., & Wheelwright, S. (2002). Psychiatry, 27, 671–680.
Mindreading Made Easy. Cambridge Medicine, 17, 28–29. Hobson, R. P., Ouston, J., & Lee, A. (1988). Emotion recognition
Baron-Cohen, S., Jolliffe, T., Mortimore, C., & Robertson, M. in autism: co-ordinating faces and voices. Psychological
(1997). Another advanced test of theory of mind: evidence Medicine, 18, 911–923.
from very high functioning adults with autism or asperger Howard, M. A., Cowell, P. E., Boucher, J., Broks, P., Mayes, A.,
syndrome. Journal of Child Psychology and Psychiatry, 38, Farrant, A., & Roberts, N. (2000). Convergent neuroana-
813–822. tomical and behavioural evidence of an amygdala hypothesis
Baron-Cohen, S., Ring, H. A., Bullmore, E. T., Wheelwright, S., of autism. Neuroreport, 11, 2931–2935.
Ashwin, C., & Williams, S. C. (2000). The amygdala theory of Jolliffe, T., & Baron-Cohen, S. (1999). The Strange Stories Test: A
autism. Neuroscience and Biobehavioral Reviews, 24, 355–364. replication with high-functioning adults with autism or As-
Baron-Cohen, S., Spitz, A., & Cross, P. (1993). Can children with perger Syndrome. Journal of Autism and Developmental Dis-
autism recognize surprise? Cognition and Emotion, 7, 507–516. orders, 29, 395–406.
Baron-Cohen, S., Wheelwright, S., Hill, J. J., Raste, Y., & Plumb, Kilts, C. D., Egan, G., Gideon, D. A., Ely, T. D., & Hoffman, J. M.
I. (2001). The ‘‘Reading the Mind in the Eyes’’ Test revised (2003). Dissociable neural pathways are involved in the rec-
version: a study with normal adults, and adults with Asperger ognition of emotion in static and dynamic facial expressions.
syndrome or high-functioning autism. Journal of Child Psy- Neuroimage, 18, 156–168.
chology and Psychiatry, 42, 241–251. Kleinman, J., Marciano, P. L., & Ault, R. L. (2001). Advanced
Bormann-Kischkel, C., Vilsmeier, M., & Baude, B. (1995). The theory of mind in high-functioning adults with autism. Journal
development of emotional concepts in autism. Journal of Child of Autism and Developmental Disorders, 31, 29–36.
Psychology and Psychiatry, 36, 1243–1259. Loveland, K. A., Tunali Kotoski, B., Chen, R., & Brelsford, K. A.
Buitelaar, J. K., Van derVan der Wees, M., Swabb Barneveld, H., (1995). Intermodal perception of affect in persons with autism
& Van derVan der Gaag, R. J. (1999). Theory of mind and or Down syndrome. Development and Psychopathology, 7,
emotion-recognition functioning in autistic spectrum disorders 409–418.
and in psychiatric control and normal children. Development Loveland, K. A., Tunali Kotoski, B., Chen, Y. R., Ortegon, J.,
and Psychopathology, 11, 39–58. Pearson, D. A., Brelsford, K. A., & Gibes, M. C. (1997).
Calder, A. J., Keane, J., Manes, F., Antoun, N., & Young, A. W. Emotion recognition in autism: Verbal and non-verbal infor-
(2000). Impaired recognition and experience of disgust fol- mation. Development and Psychopathology, 9, 579–593.
lowing brain injury. Nature Neuroscience, 3, 1077–1078. Ozonoff, S. (1995). Reliability and validity of the Wisconsin Card
Calder, A. J., Lawrence, A. D., & Young, A. W. (). Neuropsy- Sorting Test in studies of autism. Neuropsychology, 9, 491–500.
chology of fear and loathing. Nature Neuroscience, 2, 352–363. Ozonoff, S., Pennington, B. F., & Rogers, S. J. (1991). Executive
Capps, L., Yirmiya, N., & Sigman, M. (1992). Understanding of function deficits in high-functioning autistic individuals:
simple and complex emotions in non-retarded children with Relationship to theory of mind. Journal of Child Psychology
autism. Journal of Child Psychology and Psychiatry, 33, 1169– and Psychiatry, 32, 1081–1105.
1182. Phillips, M. L., Young, A. W., Scott, S. K., Calder, A. J., Andrew,
De Waal, F. B. M. (1998). Chimpanzee politics: power and sex C., Giampietro, V., Williams, S. C., Bullmore, E. T., Bram-
among apes (Rev. ed.). Baltimore, London: Johns Hopkins mer, M., & Gray, J. A. (1998). Neural responses to facial and
University Press. vocal expressions of fear and disgust. Proceedings of the Royal
Ekman, P., & Friesen, W. (1971). Constants across cultures in the Society Series B-Biological Sciences, 265, 1809–1817.
face and emotion. Journal of Personality and Social Psychol- Rutherford, M. D., Baron-Cohen, S., & Wheelwright, S. (2002).
ogy, 17, 124–129. Reading the mind in the voice: A study with normal adults
 The Cambridge Mindreading (CAM) Face-Voice Battery 183

and adults with Asperger syndrome and high functioning Wellman, H. M. (1992). The child’s theory of mind. Cambridge,
autism. Journal Autism and Developmental Disorders, 32, 189– Mass/London: The MIT Press.
194. World Health Organisation. (1994). ICD-10—International classi-
Sodian, B., & Frith, U. (1992). Deception and sabotage in autistic, fication of diseases (10th ed.). Geneva, Switzerland: World
retarded and normal children. Journal of Child Psychology and Health Organisation.
Psychiatry, 33, 591–605. Yirmiya, N., Sigman, M. D., Kasari, C., & Mundy, P. (1992).
Wechsler, D. (1999). Wechsler Abbreviated Scale of Intelligence Empathy and cognition in high-functioning children with
(WASI), The Psychological Corporation. UK. autism. Child Development, 63, 150–160.

View publication stats