Pergamon

PII: S0021-9630(96)00052-2

J. Child Psychol. Psychiat. Vol. 37, No. 7, pp. 873-877, 1996 19% Association for Child Psychology and Psydiiatry Published by Elsevier Science Ltd Printed in Great Britain. All rights reserved 0021-9630/% $15.00 + 0.00

Studying Weak Central Coherence at Low Levels: Children with Autism do not Succumb to Visual Illusions. A Research Note
Francesca G.E. Happe
MRC Cognitive Development Unit, London, U.K. While anecdotal reports of abnormal perceptual experiences in autism abound, there have been to date no experimental studies showing fundamental perceptual peculiarities. The present paper reports results from a first study of low-level visual integration in autism. Twenty-five subjects with autism, 21 normal 7- and 8-year-olds, and 26 children with learning difficulties were asked to make simple judgements about six well-known visual illusions. Two conditions were used, in an attempt to explore group differences; standard two-dimensional black and white line drawings, and the same figures augmented with raised coloured lines. The subjects with autism were less likely to succumb to the twodimensional illusions than were the other groups, and were less aided by the threedimensional 'disembedded' condition. These striking results are discussed with reference to the 'central coherence' account of autism. Copyright 1996 Association for Child Psychology and Psychiatry. Keywords: Autism, central coherence, visual illusions, perception Abbreviations: BPVS, British Picture Vocabulary Scale; DSM, Diagnostic and Statistical Manual of the American Psychiatric Association; MLD, moderate learning difficulties; VMA, verbal mental age; WISC-R, Wechsler Intelligence Scale for ChildrenRevised

Introduction Perceptual abnormalities are a common feature of the autobiographical accounts of individuals with autism (Grandin, 1984; Williams, 1992; White & White, 1987). Hypo- and hyper-sensitivity to sound, touch or vision are described, as well as fragmented perception and intense experience of normally unnoticed aspects of the environment. To date, however, these possible peculiarities have not been systematically explored. Frith and Baron-Cohen (1987), reviewing research in this area, concluded that there was no empirical evidence against the hypothesis that low-level perceptual processes are intact in autism. Current psychological theories of autism (reviewed in Happe, 1994a, b) focus mainly on the social impairments, and have relatively little to say about the nonsocial deficits and assets seen in autism. Non-social features which are clinically prominent include the restricted repertoire of interests (necessary for diagnosis in the third revised version of the Diagnostic and Statistical Manual, DSM-III-R, of the APA, 1987), the obsessive desire for sameness (one of two cardinal

Requests for reprints to: Francesca G. E. Happe, MRC Cognitive Development Unit, 4 Taviton Street, London WCIH OBT, U.K. 873

features for Kanner & Eisenberg, 1956), the islets of ability (an essential criterion in Kanner, 1943), the idiot savant abilities (striking in one in 10 individuals with autism; Rimland & Hill, 1984), the excellent rote memory (emphasized by Kanner, 1943) and the preoccupation with parts of objects (a diagnostic feature in DSM-IV, APA, 1994). The 'central coherence' theory presented by Frith (1989) and further developed by Frith and Happe (1994) and Happe (1994a, c) is one exception to this general trend. These authors suggest that the non-social features of autism, both assets and deficits, may follow from a tendency to process local vs global information, and a failure to process incoming stimuli in context. One question concerns the level at which coherence is hypothesized to be lacking. Findings that subjects with autism are better than controls at spotting embedded figures (Shah & Frith, 1983), and that they are better able to reproduce unsegmented block designs (Happe, 1994c; Shah & Frith, 1993) suggest problems in coherence at a relatively early perceptual or attentional stage. By contrast, demonstrations of failure to process information in context during sentence reading (Frith & Snowling, 1983; Happe, in press) point to deficits in coherence at higher levels of meaning extraction. The present study represents a first attempt to establish the level at which central coherence is weak in subjects with autism. Visual illusions were chosen as

874
Titchener circles

F. G. E. HAPPE

oo oOo oo
Muller-Lyer figures

Control figure

Are these two circles different sizes or the same size?

Control figure

Are these two lines the same length or different lengths?

Ponzo illusion Poggendorf illusion Hering iUusion

Kanisza triangle

Figure 1.

Two-dimensional stimuli for visual illusions study and examples of control stimuli.

figures devoid of 'higher level' meaning, which nonetheless appear to require the integration of perceptual features. Gregory (1967) has pointed out that in many illusions 'induced' and 'inducing' parts can be easily distinguished. Taking this suggestion of a figure-ground separation, illusions appear to present examples of the effects of context on perception at very low levels. For example, the central circles in the Titchener illusion (Fig. 1) are easily perceived as the same size when seen in isolation (as in the control figure. Fig. 1), but when placed in the inducing context (of surrounding small or large circles) are judged to be of different sizes. In this illusion the central circles presumably become part of a whole figure gestalt which changes subjects' perception of these parts. If subjects with autism are able to process the induced parts separately from the inducing contextthat is, if they fail to integrate all elements of the illusionsthen it

might be predicted that they will make more accurate judgements, and fail to succumb to the common illusory effects. If their better performance on the illusions is due to this weak coherence, then a manipulation which artificially segments out the parts to be judged (through colour and depth) should help the controls to perform as accurately as the subjects with autism, who should themselves be little aided by this pre-segmentation.

Method Subjects
Twenty-five children with autism took part in the study. All were attending special schools for autism, and all had received a diagnosis of autism by experienced clinicians using the guidelines of standard criteria such as DSM-III, DSM-IH-R (APA, 1987). Control subjects of two types were also tested. Twenty-one young normal children approximately matched as

WEAK CENTRAL COHERENCE AT LOW LEVELS

Table 1. Subject Characteristics BPVS standardized scoreAVISC-R VIQ 58.7 14.2 (40-92) 67.3 11.0 (46-85) 106.0* 13.3 (78-132)

875

Group Autism 25 Mean SD Range Mean SD Range Mean SD Range

Age(y:m) 13:1 2:4 (8:1-16:8) 12:0 2:2 (6:6-14:9) 7:9* 0:4 (7:5-8:4)

VMA 7:1 2:3 (3:7-13:0) 7:9 2:0 (4:8-11:1) 8:6 1:6 (5:7-11:6)

MLD

26

Normal

21

V<.001.
a group to the mean verbal mental age (VMA) of the autism group were tested. Twenty-six children with moderate (nonspecific) learning difficulties (MLD), but without autism, were also tested. This group was of mixed aetiology and attended special schools in London. Most subjects were assessed with the British Picture Vocabulary Scale (BPVS; Dunn, Dunn, Whetton, & Pintilie, 1982) to establish a receptive vocabulary age equivalent. For six subjects with autism and eight subjects in the MLD group, however, verbal IQ and VMA were calculated from the Wechsler Intelligence Scale for ChildrenRevised (WISC-R; Wechsler, 1974). Subject characteristics are shown in Table 1. The autism and MLD groups did not differ in any subject variable. that each subject had sufficient language to comprehend and answer the test questions, and was motivated and able to make simple line length and orientation judgements with nonillusory figures. Subjects who were unable to answer questions about the control stimuli were excluded from the sample. The three-dimensional illusions were created by adding brightly coloured plastic strips/shapes over the original black and white line drawings. In each case the raised forms highlighted those parts of the figure about which the subject was asked to make a judgement. So, for example, in the Muller-Lyer illusion, the two straight line segments were overlaid with raised plastic lines of the same dimensions as the original lines. The coloured plastic forms, which were blue, red or yellow, stood up approximately 4mm from the surface of the card.

Materials
A selection of six well-known illusions was used, taken from illustrations in standard psychology textbooks. These comprised the Ponzo illusion, Titchener circles, Kanisza triangle, Muller-Lyer figures, Hering illusion and the Poggendorf illusion (see Fig. 1). The illusions varied in size from approximately 6 x 6cm to 10 x 7cm. Each illusion was presented in two conditions. In the two-dimensional condition the illusions were shown in black and white on a laminated card. In addition, control figures were presented for each illusion (see examples in Fig. 1). These were used to ensure

Procedure
Each subject was tested in a quiet place in his/her school. The standard two-dimensional illusions, with their control figures, were shown to the subject first. The experimenter shuffled the cards, presenting illusions and control stimuli interspersed in random order. The three-dimensional illusions were shown to the subject on a second visit, between 1 and 3 weeks later. Subjects were asked the following questions, as appropriate to the illusion: 'Are these two lines the same length or different lengths?' (Ponzo and Muller-Lyer illusions), 'Are these two circles different sizes or the same size?' (Titchener circles), 'How many triangles can you see here?' (Kanisza), 'Does line A join up with line B or line C?' (Poggendorf), 'Are these two lines straight or curvy?' (Hering illusion). Order of alternatives in the test question was counterbalanced, and where subjects answered 'Different', they were asked to point out the longer line/larger circle.

Table 2 Group Results. Number of Illusions 'Succumbed' (max = 6)

to Which Subjects Threedimensional illusions 1.80 0-3 .91 2.00 0-5 1.20 2.09 0-5 1.34 test, autism

Group Autism

N 25 Mean Range SD Mean Range SD Mean Range SD

Twodimensional illusions 2.24* 1-5 .93 3.42 2-5 1.03 4.09 2-6 1.09

Results The mean number of illusions by which subjects were 'fooled' (i.e. gave typical incorrect judgements) is shown in Table 2. As can be seen, the autism group succumbed to fewer illusions in the standard two-dimensional form than did the control groups. Control subjects, as predicted, made significantly more accurate judgements in the three-dimensional condition than with the twodimensional illusions. In the autism group, however, the difference between two- and three-dimensional illusion performance was not significant. Interestingly, in the

MLD

26

Normal

21

*F(2, 69) = 20.08, p < .000; Tukey group < MLD and normal, p < .001.

876

F. G. E. HAPPE Table 3 Frequency Data for Individual Illusions, Percentage of Subjects who Succumbed to Each Illusion Group Autism Two-dimensional Three-dimensional MLD Two-dimensional Three-dimensional Normal Two-dimensional Three-dimensional Ponzo 28.0 24.0 34.6 19.2 52.4 28.6 Kanisza 8.0 12.0 42.3 42.3 57.1 61.9 Titchener Muller-Lyer 32.0 28.0 61.5 34.6 71.4 23.8 88.0 60.0 88.5 30.8 80.9 28.6 Hering 40.0 24.0 69.2 46.2 76.2 33.3 Poggendorf 28.0 36.0 57.7 26.9 61.9 38.1

three-dimensional condition the autism and control groups did not differ (F(2,69) = .40, p = .67). On multiple regression ANOVA there was a significant group by condition interaction (F(2,69) = 9.95, p = .000). In terms of frequency data, only seven of the 25 subjects with autism (i.e. 28%) succumbed to three or more of the six two-dimensional illusions, compared with 20 (77%) of the MLD and 20 (95.2%) of the young normal controls. For the three-dimensional illusions, the comparable frequencies were seven subjects with autism (28%), eight (31%) subjects with MLD and nine (42.9%) young normal subjects succumbing to half or more of the illusions. Most subjects in the control groups were aided by the disembedded three-dimensional condition in making accurate judgements, i.e. most subjects succumbed to fewer illusions in the three-dimensional form than in the two-dimensional form. This was not true for the subjects with autism. While less than half (44%) of the autism group was aided by the three-dimensional disembedded condition, almost all (95%) the normal controls and most (73%) of the MLD group were more accurate in the three-dimensional than the two-dimensional condition. Chi-square analysis showed that the autism group differed significantly in this respect from both control groups (autism vs normal, x^ = 13.67, p < .001; autism vs MLD, x^ = 4.44, p < .05). The pattern of results for the individual illusions is shown in Table 3. Of the six illusions, two showed somewhat different results. The Muller-Lyer figure stood out from the other illusions as the most 'persuasive' to all groups. Eighty-one per cent of the normal children, 88% of MLD subjects and, interestingly, 88% of the autism group succumbed to this illusion, erroneously judging the lines to be of different lengths. This was the only illusion which 'tricked' a significant proportion of the autism group. The other illusion which did not follow the pattern of result reported above, was the Kanisza triangle which differed in being relatively unaffected by the three-dimensional 'disembedded' condition (i.e. as many errors were made on the three-dimensional version of the triangle as on the two-dimensional version). Discussion The present study was a first exploration of illusory figure perception in autism. The results are striking and somewhat surprising; subjects with autism made more

accurate judgements of illusory figures and less often succumbed to the typical mis-judgements. That their success on this task was due to failure to integrate the induced lines and the inducing context, is indicated by the results from the three-dimensional condition. Where the induced lines were artificially disembedded from the inducing context, through colour and depth, control subjects performed like subjects with autism, and the group differences disappeared. This effect is akin to Shah and Frith's (1993) finding with the block design task, where pre-segmenting the designs brought control, performance to the level of the autism group's success, but did not aid the subjects with autism. The theory that people with autism have weak central coherence predicts and explains the present finding, and it is not apparent that any other current theory of the disorder (e.g. deficits in executive function, theory of mind) can readily account for these results. Differences between the visual illusions used also suggest that weak coherence may provide a good explanation for the success of the autism group. The only illusion to which a significant proportion of the autism group succumbed was the Muller-Lyer. This illusion cannot easily be split into induced figure and inducing context, since the arrow heads in the version used are connected to the judged lines and form part of the same 'object'. It is not clear exactly why perception of the Kanisza triangle was so little affected by the threedimensional condition, but it is worth noting that this illusion is unlike the others used in that it induces perception of illusory contours. The present findings are surprising in so far as the visual illusions appear to require integration at very low levels, and it seemed unlikely a priori that subjects would show weak coherence at this level. It will, of course, be necessary to replicate these results with other samples of subjects and different stimuli. More sophisticated techniques are available for estimating the magnitude of illusory effects (e.g. asking subjects to adjust lines to match lengths), but it is impressive that even using quite gross judgements and very simple methods, significant group differences could be found. If subjects with autism do fail to integrate visual infortnation at this low level, it may be necessary to think again about perceptual abnormalities, reports of fragmented perception, and how people with autism 'see the world'.

WEAK CENTRAL COHERENCE AT LOW LEVELS

Acknowledgements^The data reported in this paper were collected with the help of the following schools, to whom I am extremely grateful: Helen Allison School, Sybil Elgar School, Springhallow School, Chalgrove School, Fairway School, DoUis School and Northway School. I would also like to thank Nicole Berenson and Melanie Bernitz for their help with collecting the data, and Uta Frith, and other colleagues at the CDU, for invaluable advice and discussion.

877

Gregory, R. L. (1967). Eye and brain. New York: World University Library. Happe, F. G. E. (1994a). Annotation: Psychological theories of autism. Journal of Child Psychology and Psychiatry, 35,

215-229
Happe, F. (1994b). Autism: an introduction to psychological theory. Lx)ndon: UCL Press. Happe, F. G. E. (1994c). Wechsler IQ profile and theory of mind in autism: a research note. Journal of Child Psychology and Psychiatry, 35, 1461-1471 Happe, F. (in press). Central coherence and theory of mind in autism: reading homographs in context. British Journal of Developmental Psychology. Kanner, L. (1943). Autistic disturbances of affective contact. Nervous Child, 2, 217-250 Kanner, L. & Eisenberg, L. (1956). Early infantile autism 1943-1955. American Journal of Orthopsychiatry, 26, 5 5 65 Rimland, B., & Hill, A. L. (1984). Idiot savants. In Wortis, J. (Ed.), Mental retardation and developmental disabilities. Vol. 1 (pp. 155-169). New York: Plenum Press. Shah, A. & Frith, U. (1983). An islet of ability in autistic children: a research note. Journal of Child Psychology and Psychiatry, 24, 613-620 Shah, A. & Frith, U. (1993). Why do autistic individuals show superior performance on the Block Design task? Journal of Child Psychology and Psychiatry, 34, 1351-1364 Wechsler, D. (1974). Wechsler Intelligence Scale for ChildrenRevised. New York: The Psychological Corporation. Williams, D. (1992). Nobody nowhere. London: Doubleday. White, B. B. & White, M. S. (1987). Autism from the inside. Medical Hypotheses, 24, 223-229 Accepted manuscript received 29 February 1996

References
American Psychiatric Association (1987). Diagnostic and statistical manual of mental disorders (DSM-III-R) (3rd ed., rev.). Washington, DC: American Psychiatric Association. American Psychiatric Association (1994). Diagnostic and statistical manual of mental disorders (DSM-IV) (4th ed.). Washington, DC: American Psychiatric Association. Dunn, L. M., Dunn, L. M., Whetton, C , & Pintilie, D. (1982). British Picture Vocabulary Scale. Windsor, U.K.: NFERNelson. Frith, U. (1989). Autism: explaining the enigma. Oxford: Blackwell Scientific Publications. Frith, U., & Baron-Cohen, S. (1987). Perception in autistic children. In D. J. Cohen, A. Donnellan, & R. Paul (Eds.), Handbook of autism and pervasive developmental disorders (pp. 85-102). New York: John Wiley. Frith, U. & Happe, F. (1994). Autism: beyond 'theory of mind'. Cognition, 50, 115-132 Frith, U. & Snowling, M. (1983). Reading for meaning and reading for sound in autistic and dyslexic children. Journal of Developmental Psychology, 1, 329-342 Grandin, T. (1984). My experiences as an autistic child and review of selected literature. Journal of Orthomolecular Psychiatry, 13, 144-175