Autistic Artists give Clues to Cognition


Allan W. Snyder and Mandy Thomas
Centre for the Mind
Australian National University
Canberra, Australia

Certain autistic children whose linguistic ability is virtually non-existent can draw natural scenes from memory with astonishing accuracy. In particular their drawings display convincing perspective as illustrated in Figure 1. In contrast, normal children of the same preschool age-group and even untrained adults draw primitive schematics or symbols of objects which they can verbally identify. These are usually conceptual outlines devoid of detail as shown in Figure 2. We argue here that the difference between autistic child artists and normal individuals is that autistic artists make no assumptions about what is to be seen in their environment. They have not formed mental representations of what is significant and consequently perceive all details as equally important. Equivalently, they do not impose visual or linguistic schema - a process that we believe is necessary for rapid conceptualisation in a dynamic existence, especially when the information presented to the eye is incomplete.

Naturalistic drawing skills are not essential to survival, but it is nonethless a surprising fact that normal individuals can not draw unless they are taught the tricks and schema to do so or equivalently unless they learn these tricks by trial and error or by using drawing aids. The reason why this is so unexpected is that our brains obviously possess all the information required to draw, but we are unaware of this. For example, our brains perform the necessary calculations for us to recognize and label objects and to recall sufficient detail to perceive subtle changes in familiar scenes. Indeed, it is clear that human vision itself reveals the artist's touch , yet we are not consciously aware of how our brain does so. This fact has frequently intrigued philosophers and historians of art. For example, Gombrich says, 'One thing stands out from this story and demands a psychological explanation. It is that this imitation of visual reality must be very complex and indeed a very elusive affair, for why should it otherwise have taken so many generations of gifted painters to learn its tricks?' One trick, pertinent to our story, is to view the subject from an unconventional perspective, that is, one which has not been overlearned.

Why is it then that we cannot draw natural scenes, not only from our rich visual memories but even when we are also confronted with a stationary landscape? Indeed, teachers of art admonish us for not observing what lies before our very eyes. Children especially draw not what they 'see' but, rather, their own internalised schema for objects, schema which are surprisingly invariant across cultures. There is an appealing answer to this enigma. Our drawing difficulties may well be the by-product of strategies which have evolved to efficiently utilise a finite brain for rapid decision-making, particularly, as so often happens, when the information is incomplete.

Indeed, it has been argued that scenes presented to the eye are not an unlimited succession of surprises, but rather conform to certain patterns. Thus, cognition would become more rapid and efficient if the brain had expectations about what is to be seen, and it would be plausible for these expectations to become incorporated into perceptual mechanisms. Now natural scenes, as distinguished from the infinity of contrived scenes, are composed of discrete objects. We would expect then that objects of particular importance or familiarity be granted privileged mental representations. Such visual representations are acquired through the process of overlearning as are the complex thought processes required for expertise within our niche, including fluency in language, complex social skills and even driving a car. Because all naturally occuring objects share common attributes, those studied through the ages by artists, we might also expect these more intrinisic properties to be structured into visual processing. This has been discussed at length elsewhere. So, in summary, we hypothesize that the brain has mental representations that embody the salient or ecologically significant aspects of the environment. These allow for automatic complex actions.

We can make some revealing predictions about a brain that accelerates its formulation of the percept by incorporating expectations about familiar or important objects. Firstly, this particular strategy of visual processing would reveal itself by falling prey to illusions or prejudices in contrived or unnatural situations. Indeed, errors in cognition are ubiquitous. Even assumptions about the more intrinsic attributes in common with all objects can cause illusions, as discussed fully elsewhere. Secondly, and most importantly, it is the object labels or symbolic identifications which are of ultimate importance, so we need not be aware of all (and thus unable to draw) the attributes and characteristics which are processed by the brain to formulate the label. For example, we are not consciously aware of why one face is recognizable over another, nor how we derive shape from shading, perspective from gradients of texture, size invariance with distance, etc. This strategy of visual processing is mirrored also in early linguistic skills whereby children tend to linguistically label objects long before they are conscious of the semantic nuances which differentiate one object from another.

If the brain places emphasis upon expectations acquired through mental representation that is suggested above and, in particular, reports only the ecological aspects of objects to our conscience, then we have advanced a long way in explaining our innate difficulty in drawing. But this raises a fascinating question. What are the characteristics of a brain that makes no assumption about what is to be seen? Equivalently, what would it be like if our minds did not possess mental representation that embody expectations? Suppose, instead that we were to make no assumptions whatsoever.

A conceivable answer, we suggest, comes from the study of autistic children. Autism is believed to be a developmental disorder and thus its behavioural ramifications vary with age. In its purest form, autistic individuals confront life as a series of surprises. A common feature of autistic children is an impoverished linguistic ability, for example less than ten words by age four. They appear to give attention to parts at the expense of the whole. Indeed, they can be very good at spotting abstract shapes embedded in a coherent pattern but have an inability to integrate or to make links. They observe the world without interpretation. They cannot work out what others think. Put simply, autistic children do not have expectations about this world and its dynamic interactions. Their mental representations have yet to be crystalized.

Figure 1. Autistic child's drawing at about three and a half years. (taken from Reference 1.)

Now recall our arguments above which explain the difficulty of drawing as due to a by-product of our assumptions and expectations about the objects in our visual environment. Such expectations are incorporated by our mental representations. If this is true, then it would appear that natural drawing skills could potentially be found among autistic children because they apparently do not make such assumptions. Figure 1 shows that, this is the case. This particular drawing was performed from memory by Nadia, a child of about three and a half. It is initiated at a seemingly arbitrary position and executed quickly without hesitation. No attempt is made to fit such drawings onto a page, they can end abruptly at the border. Furthermore, as the linguistic abilities of this (and other) autistic artist improved, her naturalistic drawing skills deteriorated. All of this is consistent with the view that autistic preschool children can draw because they lack fixed mental representations. Indeed, there is evidence that Nadia’s mental images are occasionally mixed, allowing for subjects to be drawn from different angles. Although rare, similar drawings are characteristic of other preschool autistic child artists. The rarity is due in part to the fact that autism often exists in parallel with other mental disorders and these mitigate against drawing. Autistic preschool children are also notoriously difficult to motivate. They do not draw on request. Finally, no normal preschool child has been known to draw naturalistically, figure 2 being representative of their drawings. Autism is apparently a necessary condition for a preschool child to draw an accurate detail of natural scenes.

Two Horses

Figure 2. Representative drawings of normal children, each at age four years and two months. (Emma and Teneal, Parents on Campus Preschool).

Some older autistic children (with language ability) have also been known to have exceptional drawing skills, but so do many normal children of the same age. In fact, childrens drawings after age 6 are like adults in that they are all governed to a degree by conventions. Accordingly, we have deliberately restricted our study to younger preschool children.

We have hypothesized that the brain has mental representations which embody the salient or ecologically significant aspects of the environment. As an example, normal preschool children draw rather schematic or symbolic representations, stripped of; shape from shading, perspective from gradients of texture, size invariance with distance, etc. These sketches convey highly imposed or internalised meanings. They, along with early linguistic skills, are consistent with a mind that is conscious of the objects themselves and not the object characteristics processed by the brain to formulate the label or symbolic identification. All of this fits neatly into a strategy for making rapid decisions, particularly when confronted with incomplete information. Such strategies mitigate against naturalistic drawing. In contrast, autistic preschool children do not appear to impose expectations on their world and consequently, to them, every detail is of equal importance. Consequently, they are sometimes capable of curious skills such as drawing naturalistically, but at the cost of being able to make rapid decisions. Finally, our discussion may give telling insight into machine vision. A machine that possesses complete information about the world is useless unless assumptions about what is deemed important are built in.

Acknowledgements: We much appreciate penetrating discussions with Max Coltheart, Ian Gold, D. John Mitchell and Gerrard Vassiliou.


1. Selfe, L. 1977 Nadia; A Case of Extraordinary Drawing Ability in an Autistic Child. Academic Press.

2. Selfe, L. 1983 Normal and Anomolous Representational Drawing Ability in Children. Academic Press.

3. Arnheim, R. 1976 Art and Visual Perception. University of California Press; To the Rescue of Art. Cambridge: University of Cambridge Press. pp.155-163.

4. Freeman, N. H. 1987 Childrens drawings of human figures, in the Oxford Companion to the Mind, ed. R. L. Gregory, Oxford.

5. Gombrich, E. H. 1960 Art and Illusion. London: Phaidon Press.

6. Snyder, A. W. and Barlow, H., 1987 'Human Vision Reveals the Artist's Touch', Nature 331, 117-118.

7. Frith, U., 1989 Autism; Explaining the Enigma. London: Blackwell; 1992 Autism and Asperger's Syndrome. Cambridge: Cambridge University Press.

8. Leslie, A. 1991 The theory of mind impariment in autism. In Natural theories of mind, ed. A. Whiten, Blackwell.

9. Baron-Cohen, S. 1995 Mindblindness, MIT Press.

10. Park, C. 1978, Book review of "Nadia", Autism and Childhood Schizophrenia 8, 457-472.

11. Hermelin, B. and O’Connor, N. 1990 "Art and accuracy: the drawing ability of idiot-savants".

12. Wiltshire, S. 1987 "Drawings". London J.M. Dent and Sons Ltd.

13. Damasio, A.R. and Damasio, H. 1992 Brain and Language, Scientific American Sept, 89-110.