Tony Charman (1), John Swettenham (2), Simon Baron-Cohen(3),
Antony Cox (4), Gillian Baird (4) and Auriol Drew (4)
(1) Department of Psychology, University College London.
(2) Department of Psychology, Goldsmith's College, London.
(3) Departments of Experimental Psychology and Psychiatry, University of Cambridge.
(4) Department of Child and Adolescent Psychiatry, Bloomfield Clinic and
Newcomen Centre, Guy's and St. Thomas' Hospital Medical
School, London.
Introduction
Research into the impaired social communicative abilities
of children with autism has mostly been conducted with school-age
children, adolescents or young adults. This research has demonstrated
autism-specific impairments in the development of infant social
communicative abilities, such as pretend play, joint attention
and imitation. Furthermore, various theoretical accounts of the
development of autism (such as Baron-Cohen and Leslie's theory
of mind account, Hobson's affect responsivity account; Mundy and
Sigman's joint attention account; and Meltzoff and Gopnik's and
Rogers & Pennington's imitation account) have emphasised the
critical role of infant social, cognitive and affective factors.
To date, no experimental work has been conducted
with infants with autism, since the disorder is rarely diagnosed
before the age of 3. The present research takes advantage of
a recent prospective method of identification of autism (Baron-Cohen
et al., 1996) to investigate the development of social communicative
abilities in infants with autism.
Prospectively screening for infants with autism
We employed a new instrument, the Checklist for Autism
in Toddlers (CHAT), a screening instrument for childhood autism.
The CHAT checks for the presence of pretend play and joint attention
behaviours, as well as unrelated developmental accomplishments
such as rough and tumble play, in 18-month-old infants. The CHAT
was employed on a large population of 18-month-olds (see Baron-Cohen
et al., 1996 and conference paper by Swettenham et al. yesterday).
This enabled prospective identification of infants with autism,
as well as comparison groups of children with developmental delay
and normally-developing children.
In the present study, these infants (aged 20 months)
were tested on the following experimental measures: interest in
and empathic response to a display of distress, pretend play,
joint attention behaviours and imitation. Since work with school-age
children with autism has demonstrated autism-specific impairments
in these abilities, we would expect infants with autism to be
impaired also. However, two questions arise about the developmental
course of these social-communicative impairments in autism:
First, will infants with autism show specific
impairments, in comparison to language and mental-age matched
children with developmental delay but without autism?
Second, will closely-related behaviours such as functional
play and the non-social use of gaze (which have been shown to
be relatively intact in school-age children with autism) also
be intact in infants with autism?
The subjects
Children identified by the screen were allocated
into one of three experimental groups by the application of standardised
diagnostic and psychometric instruments. The groups for the present
experimental study were: 10 children with autism (all boys), 16
children (9 boys, 7 girls) with developmental delay, and 22 (19
boys, 3 girls) normal children.
The children were approximately 20 months old when
seen, the clinical groups had Non-Verbal Mental Age (NVMA) of
approximately 17 months and were 1SD behind the normal controls
on the Reynell Language Age (LA) measure. The normal subjects
had a higher NVMA and LA than the subjects with autism and those
with developmental delay. However, importantly there were no
differences between the autism and developmental delay groups
on CA, NVMA, or LA.
The measures
Empathic response
A measure of affective and attentional response to
a display of distress by an adult, based on earlier work by Sigman
et al. (1 992) was employed. The experimenter played jointly
with the child, with a plastic pounding toy and hammer. During
this joint play the experimenter pretended to hurt themselves
by hitting their thumb with the hammer. For 10 seconds the experimenter
displayed facial and vocal expressions of distress (i.e. cries
of pain), without using words, and stopped touching the toy.
After a further 10 second period of neutral affect the experimenter
showed the child that their finger did not hurt any more, and
resumed playing with the toy.
It was recorded whether during the first 10 seconds
of the trial the child: looked to the experimenter's face; looked
to the experimenter's hand; moved to help; and stopped playing
with, or touching, the toy. In addition, the child's own facial
affect was coded as either: (i) concerned/upset; (ii) indifferent/neutral;
or (iii) positive.
Play tasks
Spontaneous play task.-
When the child entered the room the following sets of toys were
available: a toy teaset; a toy kitchen stove with miniature pots
and pans, spoon, pieces of green sponge; and junk accessories
(e.g. brick, straw, rawplug, cottonwool, cube, box) and conventional
toy accessories (toy animals, cars etc.).
Each different play act produced by the child during
the 5 minute session was coded into the following four mutually
exclusive categories: sensorimotor; ordering; functional play;
and pretend play.
Structured play task.
A series of structured play tasks designed to evaluate the effects
of scaffolding on both functional and object substitution play,
based on the earlier work of Fein (1975) and Charman and Baron-Cohen
(in press), were conducted. For the functional trials the child
was presented with a toy telephone with a banana in place of a
receiver, a doll and a toy spoon, and a doll and a toy cup. For
the object substitution trials the child was presented with a
doll and a metal rod, a doll and a toy cup, and a doll and a wooden
brick.
The following series of scaffolding prompts was given until a response was made: First, an open
prompt was given ("What
can you do with these?"). Next, a specific prompt
was given ("Let's pretend. Give the doll a drink of juice.").
Finally the functional or pretend play action was modelled
and a specific prompt was given ("Let's pretend. Give the
doll a drink of juice.").
On each trial the first response made was scored
as functional play if the spoon/cup was placed onto the doll's
mouth in a feeding/drinking motion and object substitution if
the rod/brick was placed onto the doll's mouth in a feeding/drinking
motion, or if the child picked up the banana and put it to its
ear in the manner of a telephone receiver.
Joint attention tasks
A series of three active toy tasks based on those
described by Butterworth and Adamson-Macedo (1987) were conducted.
The child stood or sat between their mother and the experimenter.
A series of mechanical toys, designed to provoke an ambiguous
response - that is, to provoke a mixture of attraction and uncertainty
in the child - were placed one at a time onto the floor of the
room 1 to 2 metres from where the child was sitting, or standing.
The toys were a robot, which flashed and beeped and moved around
in circular sweeps; a car which followed a circular path around
the room; and a pig which made "oinking" noises and
shunted backwards and forwards. The toys were controlled by the
experimenter via a control box and an electrical lead which ran
from the box to the toy. They were active for a period of 1 minute,
during which time they stopped and restarted twice.
The following actions were scored as either present
or absent for each trial: (i) infant switched gaze between toy
and adult (experimenter or parent) and back to toy, (ii) infant
looked to control box, (iii) infant pointed to target object,
(iv) infant reached towards target object, (v) infant vocalised.
Second, a series of goal detection tasks, as described
by Phillips et al. (1992) were conducted at different times throughout
the testing session: The blocking task. When the child
was manually and visually engaged with a toy, the experimenter
covered the child's hands with their own, preventing the child
from further activity. The teasing task.- The experimenter
offered the child a toy. When the child looked at the toy and
began to reach for out it, the experimenter withdrew the toy and
held it out of reach.
The key behaviour recorded was whether the child
looked up towards the experimenter's eyes during the 5 second
period immediately after the block or tease.
Imitation
The materials and method for the procedural imitation
task followed those employed with normallydeveloping infants by
Meltzoff (1988ab), and employed with older subjects with autism
by Charman and Baron-Cohen (1994). The child sat opposite the
experimenter. Four actions were modelled, all on objects designed
to be unfamiliar to the child: a dumbbell-shaped toy was pulled
apart and put back together again; a hinge was unfolded to its
maximum angle and returned to its flat position; a recessed button
on a box which produced a mechanical beeping sound was pushed;
and the experimenter leant forward to touch the top panel of box
with his forehead, which illuminated the top panel of the box.
At the end of the modelling period, the objects were
placed, in turn, in front of the child. One non-specific prompt
("What can you do with this?") was given if the child
failed to pick up or manipulate the object at once.
Analysis
For some tasks, such as the empathic response task
and the play tasks, single behaviours were recorded by their presence
or absence, and the data was analyzed non-parametrically. For
other tasks where multiple trials were conducted, such as the
joint attention and imitation tasks, group differences were analyzed
by ANCOVA, with CA, NVMA, and LA entered as covariates. For brevity
no actual statistical test data will be presented here, but all
results reported are significant at the p < 0.05 level.
Results
On the empathy task, fewer children with autism
looked to the experimenter's face and none expressed facial concern,
in response to feigned distress. There was no difference in the
proportion of children in each group who looked to the experimenter's
"injured" hand.
On the play tasks, few infants in either
the autism or the developmental delay group produced spontaneous
pretend play, although many produced some functional play. Thus,
any potential group differences in the production of spontaneous
pretend play may have been masked by floor effects since object
substitution only emerges at around 18 months of age, and remains
somewhat infrequent and fragile until later in the second year.
In contrast, on the structured play task
the developmentally delayed infants, but not the infants with
autism, produced object substitutions.
On the joint attention tasks, infants with
autism produced fewer gaze switches of attention in response to
both ambiguous toys and ambiguous actions. In contrast, they
produced as many "nonsocial" looks at the box that controlled
the toys. Both the autism and developmental delay groups produced
little protodeclarative pointing, and any autism-specific impairments
in pointing may be masked by these floor effects.
Infants with autism also produced less imitation
than the developmental delay controls, and both clinical groups
produced less imitation than the normally-developing children.
Discussion
Thus, on some measures 20-month-old infants with
autism showed specific impairments relative to a control group
of developmentally delayed children, on others the autism group
did not differ from the developmental delay controls, whilst on
a third set of measures floor effects make interpretation of the
present findings difficult.
The picture that emerges from this series of experiments
is that some differences between children with autism and other
developmentally delayed children are clear by the end of infancy
- children with autism do not share interest in objects or actions
by making eye contact with others, they attend to and respond
less to the emotional reactions of others, and are impaired in
their imitation skills. These social communication abilities,
on which clear autism-specific impairments were apparent, are
all abilities which emerge soon after the end of the first year
of life in normal development.
In contrast, autism-specific impairments were not
apparent at age 20 months on abilities which emerge later during
the second year of normal development - such as spontaneous pretend
play and protodeclarative pointing - since developmentally delayed
children without autism were also delayed in these areas.
A developmental picture emerges of the course of
the specific impairments which children with autism show, in comparison
to children with general developmental delay but without autism.
This developmental story is reinforced by the finding that across
all three groups older and more able children tended to produce
more of the target behaviours.
This is consistent with the picture that emerges
from studies with older school-age children with autism. For
example, although there is some evidence for intact basic-level
pretend play (at least under structured, or prompted, conditions)
and imitation in autism, these basic-level early-emerging abilities
are present only in older and more able children with autism,
and only at a very much later age than for children with developmental
delay without autism.
There are several important limitation to the present
study, resulting from the design of the larger epidemiological
study by which the present sample was identified. First, the
sample of children with autism studied is not wholly representative
of the autistic population, since it excluded cases with severe
developmental delay. Thus, the present sample constitute a relatively
high-functioning sample of children with autism, whose performance
may not be generalizable to the more disabled total autistic population.
Second, on some measures - most notably the spontaneous pretend
play task and the production of pointing on the joint attention
task - floor effects may mask differences between the autism and
developmental delay groups. In addition, whilst the autism and
developmental delay groups were matched on measures of CA, NVMA,
and LA, the normally-developing children were matched on CA only.
Thus, no normative data is available on normal infants with an
equivalent mental age to the clinical groups.
Do the findings inform our understanding of the various
theoretical accounts of the basic autistic deficit, such as Baron-Cohen
and Leslie's theory of mind account, Mundy and Sigman's account,
Hobson's affective responsivity account, the executive dysfunction
account, and Frith and Happé's central coherence account?
By itself, the present study does not discriminate
between these competing theories, as all would predict
affective and cognitive impairments - similar to those demonstrated
here - by late infancy. Such questions cannot be answered by
the present cross-sectional research design, which can only access
online relationships. Only longitudinal studies can inform
us about developmental causal relationship. The present
sample has recently been followed-up at aged 42 months and clearer
evidence of the developmental importance of the pattern of intact
and impaired abilities found in this young sample of children
with autism will hopefully emerge.
Detailing the specificity of the impairments
in infants with autism, in comparison to those that are present
in other children with general developmental and language delays
but without autism, has implications for aiding early clinical
diagnosis. Although children with developmental delay may be
delayed in their development of some aspects of joint attention
and pretend play skills, the present results show that in other
aspects of joint attention, pretend play, imitation and empathic
responding these impairments are more severe in 20 month-olds
with autism than in children with general developmental delay
but without autism.