Independent Component Analysis Reveals Atypical
Electroencephalographic Activity During Visual
Perception in Individuals with Autism
Elizabeth Milne, Alison Scope, Olivier Pascalis, David Buckley, and Scott Makeig
Background: Individuals with autistic spectrum disorder (ASD) experience atypical visual perception, yet the etiology of this remains
unknown. The aim of this study was to investigate the neural correlates of visual perception in individuals with and without ASD by carrying
out a detailed analysis of the dynamic brain processes elicited by perception of a simple visual stimulus.
Methods: We investigated perception in 20 individuals with ASD and 20 control subjects with electroencephalography (EEG). Visual
evoked potentials elicited by Gabor patches of varying spatial frequency and stimulus-induced changes in - and -frequency bands of
independent components were compared in those with and without ASD.
Results: By decomposing the EEG data into independent components, we identified several processes that contributed to the average
event related potential recorded at the scalp. Differences between the ASD and control groups were found only in some of these processes.
Specifically, in those components that were in or near the striate or extrastriate cortex, stimulus spatial frequency exerted a smaller effect on
induced increases in - and -band power, and time to peak -band power was reduced, in the participants with ASD. Induced -band
power of components that were in or near the cingulate gyrus was increased in the participants with ASD, and the components that were in
or near the parietal cortex did not differ between the two groups.
Conclusions: Atypical processing is evident in individuals with ASD during perception of simple visual stimuli. The implications of these
data for existing theories of atypical perception in ASD are discussed.
Key Words: power, autism, cingulate gyrus, power, perception,
spatial frequency, visual cortex
I
n addition to impairments in social behavior and communi-
cation, atypical visual perception is also recognized as being
part of the autistic spectrum disorder (ASD) phenotype.
However, the degree to which visual abnormalities reflect an
enhancement or a reduction of perceptual sensitivity is unclear.
For example, individuals with ASD respond faster when detect-
ing visual targets than their typically developing counterparts
(1,2), yet they show impairment in tasks that involve perceptual
integration such as detecting random dot motion in noise (3).
This contradiction is mirrored by the weak central coherence
account of autism (4) that, on the one hand, suggests superior
dis-embedding and, on the other, highlights reduced integration
in individuals with ASD.
No model has yet been able to explain parsimoniously both
the superior and inferior perceptual skills shown by individuals
with ASD, although it has been suggested that “reduced effi-
ciency of neuro-integrative mechanisms” (5; page 2431) might
give rise to atypical perception in ASD or that area V1 (primary
visual cortex) might be hyper-active in individuals with ASD (6).
However, these suggestions have been neither supported nor
refuted by experimental evidence, because there have been few
attempts to measure the neural correlates of visual perception in
individuals with ASD.
The technique of electroencephalography (EEG) is well
suited to the investigation of neural integration, because changes
in EEG power provide an index of (partial) synchronization of
neuronal field potentials. Evoked increases in EEG power are
modulated by stimulus characteristics, including spatial fre-
quency (7,8). Gratings of various spatial frequencies therefore
provide an idea stimulus with which to investigate perception in
ASD.
A recent study compared event related potentials (ERPs)
elicited by the onset of low and high spatial frequency gratings
in children with ASD and a control group and found reduced
ERP amplitude in the participants with ASD (9). However, EEG
measured at the scalp is the sum of many electrical processes,
including those with neural or muscular origin (for an elegant
example see [10], page 106). The far-field potential from each
of these sources is recorded, to a greater or lesser degree, by
each scalp electrode. This raises the uncomfortable possibility
that when comparing EEG response averages between groups
there could be many reasons why differences might or might
not be found. Independent Component Analysis (ICA) has
been successfully applied to EEG data to separate these mixed
signals into temporally independent processes, thus providing
a more functionally relevant analysis of brain dynamics (11).
Therefore, in the following study we recorded EEG while
participants with and without ASD viewed the sudden onset of
Gabor patches, the properties of which form a good representa-
tion of the profiles of simple cell receptive fields in area V1 (12).
We present analysis of the ERP and the changes in - and -band
From the Department of Psychology (EM, AS, OP), The University of Shef-
field, Academic Unit of Ophthalmology and Orthoptics (DB), School of
Medicine & Biomedical Sciences, Sheffield, United Kingdom; and the Swartz
Center for Computational Neuroscience (EM, SM), Institute for Neural Com-
putation, University of California San Diego, La Jolla, California.
Address reprint requests to Elizabeth Milne, Ph.D., Department of Psychol-
ogy, Western Bank, Sheffield, South Yorkshire, S10 2TN, UK; E-mail:
E.Milne@Sheffield.ac.uk.
Received January 23, 2007; revised July 13, 2008; accepted July 24, 2008.
BIOL PSYCHIATRY 2009;65:22–30 0006-3223/09/$36.00
doi:10.1016/j.biopsych.2008.07.017 © 2009 Society of Biological Psychiatry