Download as pdf or txt
Download as pdf or txt
You are on page 1of 24

Britisb J o m I 01Drvelopmmtal Psyrbology (1991).

9, 39-16 Printed in Grtat Britain 393


0 1991 The British Psychological Society

Skill, content and generative strategies in


autistic children’s drawings

Vicky Lewis*
Department of Psychology, University of Warwick, Coventry C V 4 7 A L

Jill Bouchet
Department of Pycbology, University of Shefleld

The drawing skill, picture content and strategies for generating ideas for drawings
were compared in 20 drawings collected over the course of one year from 12
relatively able autistic children, not selected for drawing ability, and 12 learning-
impaired children. The drawing skills of the two groups and the content of their
pictures were very similar. However, the autistic children generated ideas which
were more closely related than those of the controls. The implications of the
findings for our understanding of autism are discussed.

It is well known that a small proportion of autistic people have unusual drawing
ability which is sufficiently striking to place them in the category of ‘idiot savant’
(Sacks, 1985; Selfe, 1977; Selfe, 1983). This ability is not correlated with overall
intelligence, language ability, o r social behaviour (O’Connor & Hermelin, 1988;
Selfe, 1983). However, it has been suggested that unusually good visual imagery and
visual memory may underlie unusual drawing ability (Selfe, 1983), although
O’Connor & Hermelin (1987a,b) suggest that the association is with graphomotor
memory rather than with visual memory.
Wide publicity has been given to this small minority of autistic people with
unusual drawing ability, but very little attention has been paid to the drawings of
autistic individuals in general. In particular, it is not clear whether o r not drawing
ability is a peak skill, relative to other abilities, in autistic people in general. In the
present study the drawings of a group of relatively able autistic children, not selected
for drawing ability, were compared with those of a group of learning-impaired
children of comparable chronological and mental ages.
We were interested not only in drawing skill, but also in the content of the
children’s drawings. In addition, and related to this, we were interested in the
strategies used to generate ideas o r topics for drawings.

*Requests for reprints.


394 V . Lewis and J. Borrcher
Drawing skiII
In order to assess drawing skill in a non-selected group of autistic children we
examined three aspects of their drawings. First we looked at the use of devicesgiving
photographic realism to drawings. Use of these devices, which include (a) proportion-
ality, (6) diminishing size with distance, (c) occlusion and (d) representation of three
dimensions, has been reported as occurring by age 4 in autistic children with
exceptional drawing ability (Selfe, 1980). However, in normal children these devices
do not appear until around 8 years of age (Freeman, 1980; Lewis, 1990), and in the
majority of learning-impaired children use of these devices is related to mental age,
rather than to chronological age (Harris, 1963; Selfe, 1983).
Selfe (e.g. 1985) also points out that the pictures of the autistic children with
exceptional drawing ability are drawn as if from a single viewpoint unlike the
drawings of young normal children. Drawing as if from a single viewpoint alters the
spatial rehtionships of the drawing. This was the second aspect of drawing skill which
we examined. In particular, we looked at the spatial relationships (a) between objects
within a picture and (6) between objects and the edge of the page. If a child draws
two or more objects within a picture from a single viewpoint the objects will be
meaningfully related spatially whereas they will not be related spatially in a
meaningful way if each object is drawn as if from a different viewpoint. Selfe (1983)
reports that all the autistic children with exceptional drawing ability that she
studied were capable of representing realistic proportions between objects at
different distances in their drawings, in one case before the age of 5 years. Further,
when drawing an object which does not fit the page, the child who does not draw
from a single viewpoint is likely to distort the object to fit the available space,
whereas the child who draws from one viewpoint is likely to truncate lines at the
edge of the page. In support of this, Selfe reports that the autistic children with
exceptional drawing ability did not distort their drawings to fit the page.
Our final measure of drawing skill concerned the amorrnt of &tad represented. It is
clear that the drawings by the autistic children with exceptional drawing ability
studied by Selfe are extremely detailed, being composed of many lines (Selfe, 1980).
This detail is far greater than that observed in the drawings of normal children of
equivalent developmental level (e.g. Selfe, 1977).
Since the autistic children in our study were not selected on grounds of having
exceptional drawing ability, we predicted that their drawings would not differ from
those of controls in any of three measures of drawing skill described above.

Drawing content
We were interested in two aspects of drawing content. Our first interest was the
compIexi9 of the children’s drawings. Complexity was defined in terms of the number
of main objects and actions represented in a picture, regardless of detail or skill.
Complexity so defined is likely to correlate with drawing skill. However, it may also
be related to creativity, in the sense of spontaneous productivity, and this was the
question of interest here. Lack of creativity and imagination is sometimes said to be a
defining feature of autism (Wing & Gould, 1979). It seemed possible that the autistic
Skill, content and generative strategies in autistic children’s drawings 395
children’s drawings might be less complex than those of controls, despite the fact that
n o difference between the groups’ drawing skill was predicted.
The second aspect of drawing content which interested us was the range of ideas
o r topics represented in autistic children’s drawings, in particular the representation of
(a) people, (6) objects and (c) unusual topics. Selfe’s observations suggest that autistic
children with exceptional drawing ability are likely to draw human figures less often
than other children, to draw certain kinds of objects, in particular buildings, roads
and road signs more frequently than other children, and to include unusual material,
such as words and numbers, o r schematic material, in their drawings. We predicted
that the autistic children in our study would show these abnormalities of content. To
test this prediction, the number of pictures containing people, the number repres-
enting objects in different taxonomic categories, and the number containing unusual
material were compared in autistic children and controls. In addition, we assessed the
developmental level of the children’s representations of people, using scoring criteria
from the Draw-a-Man and Draw-a-Woman Scales of the Goodenough-Harris
Drawing Test (Harris, 1963), predicting that the autistic children’s representations of
people would be less mature than those of controls. Since it is also widely believed
that autistic people have a less well-developed sense of themselves than do non-
autistic people (Bettelheim, 1967; Ryce-Menuhin, 1988) we also compared the
numbers of children in each group who represented themselves in their drawings.

Generative strategies
O u r interest in autistic children’s generative strategies arose from the evidence that
autistic children may have specific memory difficulties which affect their ability to
voluntarily retrieve information from longer term memory (Boucher, 1981 ; Boucher
& Lewis, 1989; Boucher & Warrington, 1976). If an individual is unable to retrieve
specified information at will from long-term memory, it seems possible that the
ability to generate unspecified information voluntarily may also be impaired. In a
study of word fluency, generative ability was tested by asking relatively able autistic
children to generate lists of words with o r without a category cue (Boucher, 1988).
The autistic children were significantly less able than the controls to generate words
without cueing. This kind of impaired generative ability would be consistent with
the autistic person’s characteristic lack of spontaneity and initiative. In the present
study it was predicted that autistic children would be less able than learning-impaired
controls to generate varied ideas, o r topics, for drawings. In particular, it was
predicted first that the autistic children would be more likely than controls to copy
visible objects rather than generate their own ideas. Second, it was predicted that there
would be a greater degree of relatedness among topics generated by the autistic children
than those generated by the controls, in terms of such things as category member-
ship, physical similarity o r shared semantic features. It was assumed that a high
degree of relatedness amongst topics would reflect restricted use of generative
strategies.
396 V.Lewis and J . Boucber
Method
Subects
T w o groups of 12 subjects, an autistic and a learning-impaired control group, took part in this study.
All the autistic children, of whom there were eight boys and four girls, were of secondary school age and
were attending special schools o r units at which some formal education is given. All had some
functional spoken language, and some (but not all) of the children had non-verbal intelligence within the
normal range. Relatively able autistic children were used as subjects following the argument that
essential characteristics of autism will emerge more clearly in this group than in a group with overall
severe mental retardation. All had been diagnosed autistic according to Rutter's (1978) criteria, and
currently showed the criteria1 patterns of abnormal social and communicative behaviours and repetitive
behaviours of various kinds. However, as is usual in such a group, individual children varied greatly in
the severity of their specifically autistic behaviours, as well as in the presence or absence of additional
problems. None of the children were physically disabled.
The control group contained the same number of girls and boys as the experimental group. The
control group was selected so as to be of similar age and verbal and non-verbal intelligence to the
autistic group, though children were not individually matched. Given the discrepancy which by
definition occurs between verbal and non-verbal ability in autistic children, a control group was selected
with a similar discrepancy between verbal and non-verbal ability. Children with poorer verbal than non-
verbal abilities are commonly found in special schools, where they are not usually classed as having
specific language disorders, their impaired language frequently resulting from a combination of low
overall ability and disadvantage. The majority of children in the control group used in the present study
fell into this category. However, one boy was a clear case of moderate to severe developmental
dysphasia (predominantly expressive). Another boy had a repaired cleft lip and palate, and moderate
hearing loss as well as a past history of dysphasic language difficulties. Tests used for assessing verbal
and non-verbal ability were, respectively, the Renfrew Word Finding Test (Renfrew, 1972) and the
Coloured Progressive Matrices (Raven, 1963). Subject details are shown in Table 1.

Table 1. Means and ranges of chronological age (CA), Matrices and Renfrew Word
Finding Test (RWF) scores for the two groups of subjects

CA Matrices score" RWF scoreb

Autistic
Mean 13,l 29.6 48.4
Range (10.1 1-15,O) (22-35) (42-59)
Learning-impaired
Mean 13,3 29.2 47.2
Range (10,ll-15,4) (23-35) (41-55)

' A matrices score of 29 is equivalent to a mental age of approximately 1 1 years 3 months.


' A score of 47 on the Renfrew Word Finding Tcst is equivalent to a mental age of approximately 7 years 4 months.

Procedure
The drawings test was one of a battery of tests included in a study of some possible effects of impaired
information retrieval and generation on autistic children's behaviour in semi-naturalistic situations. The
study was carried out over the course of a year, individual children being seen at closely spaced intervals
for some of the tests, and then perhaps not again for a period of weeks, when different tests were given.
Children were seen individually, in their own schools.
The drawings test took place over several sessions spread throughout the year. Each child was
required to draw 20 pictures in total, a maximum of three pictures being drawn in any one session. Each
Skill, content and generative strategies in autistic children’s drawings 397
picture was given a title by the child, with help from the tester, if needed. Titles were written on to the
pictures by the child after completion, or by the tester if the child could not write. Each child’s pictures
were accumulated in a folder and, at the start of each drawing session, the child was shown his o r her
previously drawn pictures and reminded of the titles.
Instructions were as follows. For the first picture, children were asked to draw anything they liked.
The same instruction was given for pictures 2-10 with the additional request that each picture was
different from the previous pictures. The notion of ‘something different’ was concretized, if necessary,
by pointing out when an idea for a picture was the same as something already drawn, or by saying ‘Not a -;
not a -’etc., while going through the child’s folder. N o specification or restriction was given as to the
kind of pictures which might be drawn, apart from the fact that each one must be different. A few
children asked questions such as ‘Can I draw a scene?’, o r ‘Can I copy that picture on the wall?’, and
were told that they could. Approval for all pictures was given, regardless of the content o r quality of the
drawing, so long as the picture was not a repeat of a previous picture. N o time limit was given, either for
generating ideas or for drawing pictures. A note was made of pictures which were either copied o r
traced.
For pictures 11-20, the instructions were similar, except that the children were told that they must not
copy anything they could see. This instruction was more difficult to convey, and produced a few bizarre
responses (see Results). Phrases such as ‘Something you can’t see’, ‘Something which isn’t in the room’,
‘Something out of your head’ were used interchangeably.
One o r two children were reluctant to draw anything at first, o r became negative towards the end of
the test. In these few instances, the child was allowed to generate an idea which the tester then drew,
minimally, but not schematically. In a few cases, therefore, less than 20 pictures drawn by a child were
available for assessment of drawing skill. However, 20 topics, or titles, of pictures were available (for all
but one child) for the assessment of content and of generative ability. Inadvertently, only 19 pictures
were requested from one child in the control group.

Results
Drawing skill
Drawing skill in all the pictures was assessed by the first author. The second
author independently assessed drawing skill in each child’s 5th and 15th pictures.
Tests of inter-observer reliability are reported below. The mean number of pictures
drawn by the autistic and the control group respectively were 9.92 and 9.33 for
drawings 1-10, and 9.42 and 8.5 for drawings 11-20.

( i ) Devices giving photographic realism. Every drawing was rated for (a) proportion-
ality, (6)diminishing size with distance, (c) occlusion and ( d ) representation of three
dimensions.
(u) Proportionality. This was rated on a four-point scale: 0-not proportional; 1-
realistic proportions within single objects; 2-realistic proportions within and
between objects drawn in the same plane or on the same baseline; 3-appropriate
proportions between objects at different distances.
(6) Diminishing size with distance. This was rated on a three-point scale: 0-
objects on a single baseline or free standing; 1-more distant objects towards the
top of the page; 2-distant objects proportionally smaller than near objects.
(6) Occlusion. This was rated on a three-point scale: 0-objects segregated on the
page; 1-objects overlap, but both complete; 2-object in the foreground partially
occludes object in the background (hidden line elimination).
(d) Representation of three dimensions. This was rated on a three-point scale: 0-
398 V . Lewis and J. Bou-her
one side portrayed; 1-more than one side of an object represented, but no evidence
of perspective; 2-use of perspective when representing more than one side.
For individual drawings the rating given on each scale was the highest noted even
if there was only a single example at that level in the drawing. Examples of the ratings
given for individual drawings by different children are reproduced in Appendix 1.
On all the measures of drawing skill the majority of the ratings were 0 or 1. Inter-
observer agreements on proportionality, diminishing size with distance, occlusion
and representation of three dimensions for pictures 5 and 15 combined were 84, 87,
71 and 82 per cent respectively.
A total score for each rating scale was calculated for every child by multiplying
each rating point by the number of drawings demonstrating that level and adding
together the resulting figures. This total score was then divided by the number of
pictures drawn by the child, to allow for differences between children in the number
of pictures drawn. The median scores for the two groups on the four scales are
shown in Table 2. This table also shows the number of children in each group who
exhibited some evidence of each drawing device.

Table 2. The medians for the two groups for depiction of (a) proportionality
(maximum 3), (6) diminishing size with distance (maximum 2), (t) occlusion
(maximum 2) and (d) representation of three dimensions (maximum 2). The table also
shows the numbers of children in each group achieving a rating greater than 0 for at
least one drawing on each scale

Proportionality Diminishing size Occlusion Three dimensions


Median N Median N Median N Median N

Autistic 0.55 12 0.05 6 0.43 12 0.00 5


Learning-
impaired 0.62 11 0.00 4 0.73 12 0.11 9

It can be seen that every child demonstrated the ability to use occlusion. In fact all
but three autistic children and two control children produced at least one example of
hidden line elimination. All but one control child showed some use of proportion-
ality, although only three autistic children and one control child gained the maximum
rating of three for at least one drawing. Less than half the children showed
evidence of diminishing size with distance, with only four autistic children and two
control children receiving the maximum rating for at least one drawing. On
representing objects in three dimensions, three autistic children and six control
children gained the maximum rating for at least one drawing. Mann-Whitney U tests
(two-tailed) showed there to be no differences between the groups in their use of any
of the devices giving photographic realism to drawings, namely proportionality,
diminishing size with distance, occlusion and representation of three dimensions.

(ii) Spatial relationships. (a)Between objects within a picture. Each drawing of two or
more objects was assigned to one of two categories on the basis of whether or not the
objects were drawn in a meaningful spatial relationship to one another. Inter-rater
Skill, content and generative strategies in autistic children’s drawings 399
reliability tests were carried out on ratings of whether or not a drawing contained
more than one object and, if so, whether or not at least two objects were related
meaningfully. The majority of the drawings contained only one object. Inter-
observer agreement on pictures 5 and 15 combined was 87 per cent.
The number of drawings showing a meaningful relationship, as a proportion of
the number of drawings showing two or more objects, was calculated for each child.
All the autistic children and 10 of the control children produced at least one drawing
containing two or more objects, and of these 11 autistic children and all 10 control
children produced at least one drawing showing two objects in a meaningful
relationship. The median proportion scores (maximum 1.OO)for the autistic and the
control groups were 0.81 and 1.00 respectively. This group difference, using the data
from the 12 autistic children and the 10 control children who produced at least one
drawing containing two or more items, was not significant (Mann-Mhitney U = 34.5,
p = .077, two-tailed).
(6)Between objects and the edge of the page. Each drawing was assigned to one of
three categories on the basis of the relationship between the drawing and the edge of
the page: A-the drawing fits the page; B-the drawing is distorted to fit the page;
C-the lines of the drawing are truncated at the edge of the pages. Six examples of
drawings in category A and two examples of drawings in category C are reproduced
in Appendix 1. The large majority of the pictures were assigned to category A. Inter-
observer agreement on the relationship between objects and the edge of the page for
pictures 5 and 15 combined was 87 per cent.
The number of drawings in each category for each child was divided by the
number of drawings produced by the child. All the children produced at least one
drawing which fitted the page, five autistic children and one control child showed
distortion in at least one drawing, and eight autistic and five control children
produced at least one drawing with lines truncated at the edge of the page. The
medians of the resulting scores (maximum 1.00) for the autistic and the control
groups respectively were: category A 0.93 and 1 .O; category B 0.0 and 0.0; category C
0.05 and 0.0. The group difference for category A was non-significant (Mann-Mhit-
ney U = 42, p = .073, two-tailed). The relative infrequency of truncation and
distortion made statistical analysis of these categories inappropriate.

(iii) Amount of detail represented. Each drawing was assigned to one of four
categories, based on the number of lines drawn: category a-a single line when the
pen/pencil has not been removed from the paper; category b-between two and five
lines; category c-between six and 10 lines; category d-11 or more lines. Examples
of drawings in category d and one drawing in category b are reproduced in Appendix 1.
The large majority of the pictures were assigned to category d. Inter-observer
agreement on the amount of detail represented for pictures 5 and 15 combined was 78
per cent.
The number of drawings by each child which fell into each category was divided
by the number of pictures drawn by the child. There were no significant differences
between the groups for any of the categories, using Mann-Whitney tests.
400 V . Lewis and J . Boucher
Drawing content
For the analysis of drawing content (and for the analysis of generative strategies-see
below), the children’s ideas for drawings, rather than their execution of these ideas,
were of main interest. Topics suggested by children which were then drawn by the
tester were therefore included in these analyses. Content of all the pictures was
assessed by the second author.

( i ) Complexity. Complexity was defined in terms of the number of objects and


actions represented in a drawing, regardless of detail. Only whole objects were
included in the score. Their constituent parts or qualities were not counted. No
inferences were made about the possible content of the drawings. Thus, if a child
drew clouds but did not in any other way indicate the sky, only the clouds were
scored. This can be seen in Appendix 1, drawing ii in which five objects were
identified. These were spade, bucket, sand, clouds and horizon. Although drawing ii
contained very little detail these five objects were clearly represented. By contrast a
single object was represented in drawing v in Appendix 1, but in great detail.
When the number of main objects and actions had been noted for each picture,
pictures were assigned to one of four complexity categories on the basis of the
number of main objects and actions represented: I. ‘Simple’-one object or action; 11.
‘Intermediate7- two to four objects and/or actions; 111. ‘Complex’-five to nine
objects and/or actions; IV. ‘Elaborate7- 10 or more objects and/or actions. Pictures
drawn by the tester of topics suggested by a child were rated as ‘simple’ in all but two
cases. In these two cases the suggested topics consisted of two main objects, and the
pictures were rated as ‘intermediate’. Examples of pictures by children in each group in
each complexity category are reproduced in Appendix 1. In order to examine inter-
observer reliability on this measure the first author assessed the drawings 5 and 15 of
all the children for complexity. In contrast to the ratings of drawing skill all four
categories of complexity were widely represented in the drawings of both groups.
Therefore Pearson product moment correlation was used to assess inter-observer
reliability. Pearson product moment correlation for pictures 5 and 15 combined was
.89.
The numbers of pictures in each complexity category, and the numbers of subjects
producing at least one picture in any category, are shown for the two groups in Table
3. From the data in Table 3 it is clear that the groups did not d 8 e r significantly in

Table 3. The numbers of pictures (Np) in each complexity category. The numbers of
children (N> in each group producing at least one picture (or idea for a picture) in any
category are also shown

Simple (I) Intermediate (11) Complex (111) Elaborate (IV)


Np N Np N Np N Np N

Autistic 163 12 50 11 23 6 4 2
Learning-impaired 178 12 32 11 23 4 6 2
Skill, content and generative strategies in autistic children’s drawings 40 1
either the number of pictures assigned to each category, nor in the numbers of
subjects producing at least one picture in any complexity category.
T o examine whether or not complexity correlated with drawing skill, the pictures
drawn by the tester were excluded. For each child’s remaining pictures an average
skill score was calculated by summing the scores on the four measures of propor-
tionality, diminishing size with distance, occlusion and representation of three
dimensions and dividing this by the number of pictures drawn by the child. For the
complexity measure, category I was awarded one point, category I1 two and so on.
Then an average complexity score was calculated for the pictures drawn by each child
by summing the complexity scores and dividing this by the number of pictures drawn
by the child. The resulting measures for complexity correlated significantly with the
summed measure for drawing skill in the autistic group (r,) = .73. p < .Ol) but not in
the controls (rs=.5, n.s.). However, with such small sample sizes the difference
between the two correlation coefficients was not significant.

( i i ) Topics represented. (a) People. The numbers of pictures in which people were
represented were noted for each of the two groups. Eleven of the 12 autistic subjects
drew between them a total of 54 pictures which included people. Ten learning-
impaired children drew a total of 45 such pictures. There was no group difference in
the numbers of pictures which included people.
In order to examine the developmental level of figures of people drawn by the
children, each figure was rated according to the Goodenough-Harris Draw-a-Man
and Draw-a-Woman Scales, and the best raw score achieved by each child was noted.
The median maximum scores for the 11 autistic and the 10 learning-impaired children
were respectively 18.5 (mental age equivalent 5 to 6 years) and 24 (mental age
equivalent 6 to 8 years). There was no significant difference between the groups
(Mann-Whitney U = 28.5, p = .149).
Seven of the autistic children represented themselves in at least one drawing
compared with three of the controls.
( b ) Objects. For this analysis, only pictures at the simple and intermediate
complexity levels were included, since it was not always easy to determine what
objects were drawn in the background of the complex and elaborate pictures.
Pictures at the simple and intermediate levels constituted the large majority of
pictures drawn by children in both groups (see Table 3). Drawings of unusual topics,
as well as drawings of people, were excluded from this analysis.
Objects drawn by the children were subdivided into 16 categories. Most of the
categories were conventional taxonomic categories. However, a few were more
loosely defined, for example ‘road related’ (e.g. a bus stop, traffic lights), or ‘building
related’ (a door, a wall). (See Appendix 2). The categories of objects were suggested
by the drawings themselves, a category being noted if three or more children
produced pictures falling into that category. A small number of unclassified pictures
remained (13 in the autistic group and five in the control group). A few objects were
assigned to more than one category [e.g. teddy bear was assigned to the categories
‘animals’ and ‘toys/games/sports and leisure’; clock to the categories ‘appliances
(electrical/mechanical)’and ‘household items’]. This was only done when more than
one category seemed appropriate in terms of the primary definition of an objects. By
402 V.Lewis and J . Boucber
primary definition was meant that category to which an object most clearly and
readily belonged. Nine pictures drawn by autistic children, and 11 drawn by controls,
were assigned to two categories. The numbers of pictures in eight of the 16
categories was less than 10 in both groups. These scores were too small to make
comparisons meaningful. The numbers of pictures in each of the remaining eight
categories for both groups are shown in Table 4.
Mann-Whitney tests of the data in Table 4 showed that the Ieaming-impaired
children drew significantly more household items than the autistic children (V= 28,
p = .006), and that the autistic children drew significantly more toys/games/sports
and leisure objects than the learning-impaired children (U=35, p = .024). None of
the other group differences were significant.
(c) Unusual topics. Unusual topics for drawings included the following: maps;
diagrams; aerial views; pictures of charts or notices; things traced or drawn around;
scribble; patterns; letters, numbers o r other symbols as the main topic of a drawing.
The number of pictures with an unusual topic, and the number of subjects drawing at
least one such picture, were noted for the two groups. Seven autistic children drew a
total of 23 pictures of unusual topics, compared with four controls’ total of 12
pictures of unusual topics. Since most of the pictures of unusual topics were drawn
by only one or two children from each group, there was no significant group
difference on this measure.
In a separate analysis, unusual content was assessed by comparing the numbers of
pictures produced by the two groups in which writing (letters, numbers or other
conventional symbols) appeared, whether as the main topic (for instance, one child
drew a large, elaborate figure seven), or as redundant labelling (such as writing
‘record player’ on a picture of a record player or ‘shop’ on a picture of a shop) or as an
integral part of an object (e.g. the figures on a watch; the name of a shop). Three
autistic children together drew a total of four pictures with numbers o r letters as the
main topic, compared with only one such picture drawn by a control subject. Five
autistic children produced a total of 19 pictures with redundant labelling included,
compared with an absence of redundant labelling by controls. Eleven autistic
children produced a total of 34 pictures in which letters or numbers appeared as an
integral part of an object, compared with nine control subjects’ total of 13 such
pictures. The difference between the groups in the number of pictures containing
written symbols was significant (Mann-Whitney U = 34, p = 34,p = .022, one-tailed).
As mentioned in the Method section, the instruction for drawings 11 to 20, i.e. that
children must not copy anything visible, produced some bizarre responses. One child
in the autistic group drew a picture of some books which were missing from the rack
of books which was visible in the room. Another child in the autistic group named as
picture topics three objects which were out of sight in visible containers which he
drew-the objects were books in a cupboard, a tracksuit in a bag and some fish flakes
in a tin. One child in the control group drew a circle with ears on the side described
by the child as ‘out of my head’. This same child also produced a picture of ‘an
invisible man’ and another of ‘an invisible woman’. However, these latter pictures
were disallowed as he had already drawn a man’s face and a woman’s face.
e 4. The numbers of pictures in each of the eight commonly occurring categories (Np). The numbers of children (N> in each
producing at least one picture in any category are also shown

Vehicles Buildings Plants Food Appliances Household Toys Animals


items etc.
Np N Np N Np N Np N Np N Np N Np N Np N
ic 19 7 19 7 8 6 13 6 24 8 5 3 19 10 16 5
ing-impaired 31 9 13 9 14 7 12 6 24 9 22 10 6 4 32 8
404 V.Lewis and J. BoMcher
Generative strategieJ
(i) Copy visible obects. The strategy of copying was only available for drawings 1 to
10, since ideas for drawings 11 to 20 could not be of anything visible. In their
drawings 1-10, the autistic group drew a total of 32 pictures of visible objects, eight
of the 12 children contributing to this total. In addition two of the autistic children
each produced one picture which was traced. The control group produced a total of
34 pictures of visible objects, 10 of the 12 children contributing to this total. One
control child produced two pictures which were traced. There was no significant
difference between the groups in the use of a copying strategy as a means for
generating ideas for drawings.

(ii) Degree o j relatedness among topics. In order to assess this, every picture was
examined in relation to all subsequent pictures drawn by each child. From this a set of
possible relationships was identified. The relationshipswere mainly those categories used
in the assessment of picture content. However, relationships of other kinds were also
apparent. For example, some children drew a succession of pictures of objects or
people in their school (e.g. classmates, playground objects), or of their own named
possessions (my pen, my bracelet). Other pictures were related in that they derived
from TV or fairy stories (e.g. Little Red Riding Hood; The Three Bears) or from
traditional occasions (e.g. fireworks; a Christmas tree; an Easter egg). Quite different
kinds of relationships were identified in pictures where a child started repeatedly
from a single shape (e.g. a circle), apparently using this or other shapes to suggest
ideas (e.g. a ball, a balloon, a sun, then a sausage, a worm). The full list of identified
relationships is shown in Appendix 2.
Drawings could obviously be related in more than one way, for example a drawing
of a wind gauge and a drawing of a temperature gauge were related as belonging to
the categories ‘measuring devices’, ‘appliances’, and by being ‘utterance related’ and
‘shape related’. Further, it seemed clear that some relationships were stronger than
others. Therefore a weight of 2 was given where the relationship was strong, and a
weight of 1 was given to all others. In the example given above the wind gauge and
the temperature gauge were strongly related as measuring devices, appliances and as
being utterance related, but less strongly related in their shape. Thus the total
weighted score for the above example was 7 (i.e. 2 + 2 + 2 + 1).
The degree of relatedness among topics was assessed by both authors, one of
whom was unaware of the group to which the children belonged. Complete sets of
pictures from four children in each group were scored by both authors indepen-
dently. Scores from these eight sets were assessed for reliability, Pearson product
moment correlations for each of the eight children ranging from .742 to .959. The
remaining sets of pictures from each group of children were divided equally between
the authors for independent scoring.
Table 5 shows the extent of the relationships which were found between the
drawings of the children in the two groups. From this table it can be seen that the
control children drew more drawings which bore no relationship to previous
pictures than did the autistic children, and also that more autistic children than
controls produced pictures which were strongly related to previous pictures. In order
to examine these data each child’s weighted scores were summed to give an overall
Skill, content and generative strategies in autistic children’s drawings 405
Table 5. The medians of the numbers of pairs of drawings showing no relationship
(0) or relationships with weights of 1, 2, 3 , 4 , 5, 6, and 7 o r more for the two groups
of children. The table also shows the numbers of children (N> in each group with at
least one example of the different relationship weightings and the median overall
weighted scores for the two groups.

Learning-impaired Autistic
N Median N Median

0 12 141.5 12 130.5
1 12 14.0 12 18.0
2 12 20.5 12 23.5
Weight of 3 8 1 .o 11 5.0
relationship 4 12 5.0 11 6.5
5 4 0.0 7 1 .o
6 2 0.0 8 3.0
27 1 0.0 4 0.0
Overall weighted
score 109.5 154.5

weighted score for each child. The difference between the overall weighted scores in
the two groups was significant (Mann-Whitney U=29, p = .013).

Discussion
Findings concerning the three areas of interest in this study, namely drawing skill,
content and generative strategies will be discussed in turn.

Drawing skill
It was predicted that autistic children selected randomly from special schools and
units at which some formal education is given would not show exceptional drawing
skill, relative to controls. This prediction contrasts with, although it does not
contradict, the reports by Selfe and others of individual autistic people with
exceptional drawing skill. The prediction was supported, the autistic children’s skill
being comparable to that of controls on all the measures. Many of the children in
both groups used devices giving photographic realism at a level commensurate with
their non-verbal ability; almost all the children showed the ability to relate objects to
one another, and to the page; the drawings of each group did not differ in the amount
of detail and were, on average, quite detailed.

Drawing content
We predicted that the autistic children’s drawings might be less complex than those
of controls, reflecting a lack of spontaneous creativity. This prediction was not
406 V.Lewis and J . Boucher
supported. It was also thought possible that complexity might be related more
strongly to drawing skill in the autistic group than in the control group. The
correlation was significant for the autistic group but not for the controls, as
predicted. However, the difference between the groups was not significant. Thus
there is no firm evidence to suggest that there is any lack of complexity in the
drawings of autistic children which is related to lack of spontaneous creativity.
We also predicted, following Selfe, that the autistic children would draw fewer
pictures of people (and especially of themselves) than controls, but that they would
draw more pictures of buildings, roads and road signs. These predictions were not
confirmed. The autistic children actually drew more (though not significantly more)
pictures of people than did controls, including more pictures of themselves. This
might be explained in terms of what is taught in special schools and units for autistic
children, where drawings are frequently used as a means of teaching the children
about themselves and other people. The findings might also owe something to the
fact that several children in the autistic group drew a string of stereotyped drawings
of people but identified them as pictures of different named individuals (e.g.
classmates and/or members of staff at the child’s school). It is not clear why our
finding on the representation of people differs from that of Selfe. However, her
autistic subjects were more artistically talented but on average less intellectually able
than the children we tested. Her comparison data also differed from ours. It is
possible that one or other of these factors caused the difference between the findings.
We further predicted that the autistic children’s drawings of people would be less
mature, as measured on the Goodenough test, than those of controls. This prediction
was not supported. This suggests that autistic children’s visual image, or concept, of
a person is not impaired, relative to controls.
The range of objects (as opposed to people, or unusual topics) represented in the
drawings of the autistic children was on the whole comparable to that of controls.
However, within the range of categories of objects represented, the autistic children
drew significantly fewer household items and more toys/games/sports and leisure
objects than controls. It seems likely that this reflects differences in the ways in which
autistic and non-autistic children spend their leisure time, the autistic children being
less involved in day-to-day household activities than the controls, and more likely to
be occupied with standard play materials.
Unusual topics in the autistic children’s drawings did not include an excess of
buildings, roads and road signs, as observed by Selfe. Only one of the autistic
children we tested drew road signs. However, the autistic children did include
significantly more written symbols in their drawings than did controls, especially in
the form of redundant labelling of objects, and in the form of including symbols as an
integral part of their drawings. In addition, the majority of the autistic children’s
drawings of unusual topics were schematic, including maps and cross-sectional or
aerial views. Drawings of unusual topics by the controls included only two schematic
representations, the rest being unusual in other respects, such as drawing round an
object, or drawing the numbers on a label.
The findings on the range of topics, or ideas, represented in the drawings of the
autistic children tend to support other evidence suggesting that autistic children’s
semantic knowledge, or knowledge base, is not dissimilar to that of other children of
Skill, content and generative strategies in autistic children’s drawings 407
similar developmental age (Tager-Flusberg, 19850,b). However, the findings on the
abnormal inclusion of written symbols, and on the inclusion of schematic drawings
of various kinds suggest that what autistic children understand by ‘drawing’ may
differ in some way from what is understood by other children. Alternatively, autistic
children may include symbolic and schematic material in their drawings because of an
obsessional interest in this type of material.
T w o autistic children and one control child produced bizarre responses to the
instruction not to draw anything visible. These responses share the characteristic of a
literal interpretation of the instruction. Interpreting language literally, especially
abstract language, was noted by Kanner (1943) in his original account of autism and
therefore it is not surprising to find some literal interpretation amongst the present
group of children. The control child who interpreted the instructions literally was an
older and quite sophisticated boy who was well aware of the humour of his
responses.

Generative strategies
We predicted that the autistic children might have difficulty in generating a variety of
ideas for drawings. In particular we predicted, first, that they would be more likely
than controls to copy visible objects (either exactly, o r their own version of the
object) when this strategy was permitted (i.e. in drawings 1-10). This prediction was
not supported, children in both the groups using copying (or tracing) to a
comparable extent.
Copying is an immediately obvious strategy to use when asked to draw a range of
different things. It is also a strategy which can be objectively identified. Another
obvious strategy, which can also be objectively identified, is to draw pictures of items
within a single clear-cut taxonomic category, for example animals o r vehicles. Other
objectively identifiable strategies of a different kind include drawing simple geomet-
ric shapes, such as circles o r squares, embellishing them slightly, and labelling them
as different objects; o r drawing a series of identical people and labelling them as
different named persons. Such clearly identifiable strategies were used by children in
both the groups, and were reflected in what we have called ‘relatedness’ between
pictures. Many pictures within the sets of drawings appeared to be related, but in
ways which could not so safely be ascribed to objectively verifiable strategies.
For example, some children drew a succession of fictional characters, o r a series of
objects which they named as being their own; quite often children drew a part of
some previously drawn object. We therefore used relatedness (the product) as a
measure of generative ability (the process), assuming that a high degree of relatedness
amongst drawings reflects restricted generative ability, and vice versa. Autistic
children’s drawings were, as predicted, significantly more related to each other than
were the drawings of the control subjects. We therefore concluded that autistic
children either have impaired generative ability, o r fail to use their generative ability
to produce varied sets of drawings.
It might be suggested that the high degree of relatedness amongst individual
children’s drawings is caused by impoverished knowledge, rather than by impaired
generation of ideas from the knowledge base (Ornstein & Naus, 1985). However,
408 V.Lewis and J. Bouther
our analysis of the content of the drawings shows that, as a group, the autistic
children drew as wide and varied a range of topics as the controls. Moreover, the
groups were equated for vocabulary naming ability. It therefore seems unlikely that
impoverished knowledge (i.e. the potential stock of ideas for drawings) can explain
the high degree of relatedness we observed among the autistic children’s drawings.
It is interesting to speculate whether autistic children cannot voluntarily (i.e.
without cues) generate varied information from their knowledge base, or whether
they can do this, but do not in fact do so. It is unsafe to conclude from the fact that an
autistic person does not do something, that they cannot do it (Lewis 8t Boucher,
1988). There was no restriction in the present experiment on drawing related items. It
is not therefore possible to discriminate between these two types of explanation. This
could be examined in subsequent experiments by varying the conditions under which
the drawings are produced. For example, it would be interesting to examine a series
of drawings produced by learning-impaired children and autistic children with no
request that the drawings should differ from one another. We would predict that the
difference in relatedness between the drawings of the children in the two groups
would increase. Such a result would support an explanation of the present findings in
terms of autistic children’s failure to make spontaneous use of available strategies.

Conclusion
This study has demonstrated that, despite the relatively well-known finding that a
small group of autistic children have exceptional drawing skill, unselected autistic
children have drawing skills which are in no way above average or different from
those of non-autistic learning-impaired children. Such autistic children also draw a
very similar range of subject matter to that drawn by learning-impaired children.
However, individual autistic children draw less varied pictures than controls, and
this may be a result of impaired generative ability or failure to utilize a range of
generative strategies.

Acknowledgements
The authors would like to thank the Mental Health Foundation for financial support, Rosemary
Woolfitt and Glyn Collis of the Department of Psychology, University of Wanvick. for’assistance with
data analysis and statistical advice respectively and all the schools and children who took part in the
experiment.

References
Bettelheim, B. (1967). The E m f 9 Fortress. New York: Free Press.
Boucher. J. (1981). Memory for recent events in autistic children. Journal of Autism and Deve/opmental
Disorders, 11, 293-302.
Boucher, J. (1988). Word fluency in high-functioning autistic children. Journal of AHtism and
Developmental Disorders, 18, 637-646.
Boucher. J. & Lewis, V. (1989). Memory impairments and communication in relatively able autistic
children. Journal of Child Psychology and Psychiatry, 30. 99-122.
Boucher, J. & Warrington, E. K. (1976). Memory deficits in early infantile autism: Some similarities to
the amnesic syndrome. British Journal of Pychlogy, 67, 73-87.
Skill, content and generative strategies in autistic children’s drawings 409
Freeman, N. H. (1980). Strategies of Representation in Young Children: Anahsis of Spatial Skills and Drawing
Processes. London and New York: Academic Press.
Harris, D. B. (1963). Children’s Drawings as Measures of IntelIectual Maturify. New York: Harcourt, Brace
& World.
Kanner, L. (1943). Autistic disturbances of affective contact. Nervous Child, 2, 217-250.
Lewis, V. (1990). Young children’s painting of the sky and the ground. International Journal of Behavioral
Development, 13, 4 9 4 5 .
Lewis, V. & Boucher, J. (1988). Spontaneous, instructed and elicited play in relatively able autistic
children. British Journal of Dcuelopmental Psychology, 6, 325-339.
O’Connor, N. & Hermelin, B. (1987~).Visual memory and motor programmes: Their use by idiot-
savant artists and controls. British Journal of Prychology, 78, 307-323.
O’Connor, N. & Hermelin, B. (1987b). Visual and graphic abilities of the idiot-savant artist. Psychological
Medicine, 17, 79-90.
O’Connor, N. & Hermelin, B. (1988). Low intelligence and special abilities. Journal of Child Psychology
and Psychiatry, 29, 391-396.
Ornstein, P. A. & Naus, M. J. (1985). Effects of knowledge-base on children’s memory strategies. In
H. W. Reese (Ed.), Advances in Child Development and Behaviour. Orlando, FL: Academic Press.
Raven, J. (1963). The Coloured Progressive Matrices. London: H. K. Lewis.
Renfrew, C. (1972). Word Finding Vocabulary Scale. C. E. Renfrew, North Place, Headington, Oxford.
Rutter, M. (1978). Diagnosis and definition of childhood autism. Journal of Autism and Childhood
Schixophrenia, 8, 139-161.
Ryce-Menuhin, J. (1988). The Serf in Ear4 Childhood. London: Free Association Books.
Sacks, 0. (1985). The Man who Mistook his Wifefor a Hat. London: Picador.
Selfe, L. (1977). Nadia: A Case of Extraordinary Drawing A b i l i g in an Autistic Child. London and New
York: Academic Press.
Selfe, L. (1980). A review of current theories in psychology of children’s drawings. British Journal of
Aesthetics, 20, 160-1 64.
Selfe, L. (1983). Normal and Anomalous Representational Drawing Ability in Children. London and New
York: Academic Press.
Selfe, L. (1985). Anomalous drawing development: Some clinical studies. In N. H. Freeman & M. V.
Cox (Eds), Visual Order: The Nature and Development of Pictorial Representation. Cambridge:
Cambridge University Press.
Tager-Flusberg, H. (1985~).Basic level and superordinate level categorization by autistic, mentally
retarded, and normal children. Journal of Experimental Child Psychology, 40,45W69.
Tager-Flusberg, H. (1985b). The conceptual basis for referential word meaning in childen with autism.
Child Development, 56, 1 167-1 178.
Wing, L. & Gould, J. (1979). Severe impairments of social interaction and associated abnormalities in
children: Epidemiology and classification. Journal of Autism and Developmental Disorders, 9, 11-29.

Received 25 October 1989; revised version received 26 June 1990


410 V.Lewis and J. Boncber
Appendix 1
Drawings by four children in the autistic group and four children in the control
group. The table below shows the scores for each drawing for devices giving
photographic realism (proportionality, diminishing size with distance, occlusion,
representation of three dimensions), spatial relationships between objects and the
edge of the page, amount of detail, and complexity.

Drawings
1 11 iii iv V vi vii
...
Vlll

Proportionality 2 0 2 2 1 1 0 1
Diminishing size 2 0 1 1 0 0 0 0
Occlusion 2 0 2 1 2 0 0 0
Three dimensions 0 0 0 2 0 0 0 0
Relationship to page C A A A A A C A
Detail (number of
lines) d d d d d b d d
Complexity category I11 I11 IV IV I I I1 I1
Noh. Before the drawings were photographed any comments written on the paper by the experimenter after the child
had completed her/his drawing were covered up. The photographic quality of the reproductions varies since some
children drew in biro (red or blue), others in pencil. In addition some of the paper had been used previously for other
purposes on the reverse side.

JQ .---
I
Drawing i. ‘Shandy and the hot air balloon’ by a b o y in the learning-impaired group.
Skill, content and generative strategies in autistic children’s drawings 41 1

Drawing ii. ‘Bucket and spade at the seaside’ by a boy in the autistic group.

Drawing iii. ‘King Harold and his men’ by a boy in the learning-impaired group.
412 V.Lcwh and J. Boucher

Drawing iv. 'The band' by a girl in the autistic group.


Skill, content and generative strategies in autistic children’s drawings 413

Drawing v. ‘Wind gauge’ by a boy in the autistic group.

s
Drawing vi. ‘Balloon’ by a girl in the learning-impaired group.
414 V.Lewis and J. Bonder
Skill, content and generative strategies in autistic children's drawings 415

Drawing viii. 'Road signs' by a boy in the autistic group.


416 V.L w i s and J. B o d w
Appendix 2
Relationships identified between drawings. Those relationships which were also used
as categories in the analysis of content are asterisked.

People Utterance related


Body parts Schematic content related
Containers Named place related
Measuring devices * Animals
Monsters Buildings
Brushes Building related
Geometric shapes * Vehicles
Money related Road related
Things which fly * Toys/games/sports and leisure
War related * Appliances (electrical/mechanical)
Theatre related Plants
Train related * Food
School related * Household items
Personal possessions * Furniture
Fiction related * Clothing and accessories
Occasion related * Musical instruments
‘Invisible’ things * Tools
Things copied/traced * Stationery
Shape related * Landscape objects
Part-whole

You might also like