Sensory perception in autism

Abstract

Autism is a complex neurodevelopmental condition, and little is known about its neurobiology. Much of autism research has focused on the social, communication and cognitive difficulties associated with the condition. However, the recent revision of the diagnostic criteria for autism has brought another key domain of autistic experience into focus: sensory processing. Here, we review the properties of sensory processing in autism and discuss recent computational and neurobiological insights arising from attention to these behaviours. We argue that sensory traits have important implications for the development of animal and computational models of the condition. Finally, we consider how difficulties in sensory processing may relate to the other domains of behaviour that characterize autism.

Full text

The ability to reflect on our own and others’ thoughts
and emotions (that is, theory of mind) is a defining
characteristic of human cognition. Children with
autism spectrum conditions (ASCs; henceforth ‘autism’)
show delays in the development of this capacity1, with
knock-on consequences for cognitive empathy2 across
the lifespan. Interestingly, these alterations in social
cognition are accompanied by a very different percep-
tual experience of the world. Atypical sensory experi-
ence is estimated to occur in as many as 90% of autistic
individuals3,4 and to affect every sensory modality:
taste5, touch6,7, audition8, smell9,10 and vision11. A central
challenge of autism research is to identify the common
thread that unites these various aspects of cognition and
sensation. What neurobiological alterations might affect
processes as disparate as social cognition and sensory
perception?
This challenge is highlighted by the latest interna-
tional diagnostic criteria for autism, which now include
sensory sensitivities as a core diagnostic feature12.
Although sensory symptoms were noted in early reports
of the condition13, they have historically been construed
as secondary aspects of autistic cognition rather than as
primary phenotypic markers (see Supplementary infor-
mationS1 (box)). As well as having clinical implications
for creating autism-friendly environments, understand-
ing the importance of sensory differences in autism is
crucial for neurobiological accounts of the condition.
Because the neural computations underlying sensory
processing are relatively well understood in typically
developing individuals and are conserved between
humans and other animals, studies of sensory behaviour
have considerable potential for shedding light on autistic
neurobiology14. Further, as precursors to developmental
milestones in social cognition, sensory symptoms could
potentially serve as early diagnostic markers.
However, the issue of primacy is key. Is autism, as
often posited, a disorder of the ‘social brain’ (REF.15), with
sensory differences representing either secondary con-
sequences after a lifetime of reduced social interaction
or alterations in domain-general mechanisms (such as
attention) that affect both social processing and sensory
processing? Or are the sensory differences primary in
terms of both development and neurobiology?
Here, we explore whether sensory traits are, in fact,
core phenotypic markers of autism. To do this, we apply
four tests of core phenotypic status, by asking whether
autistic sensory traits are present in early development,
substantially improve diagnostic accuracy when included
in diagnostic assessments, reflect alterations to neural cir-
cuitry in sensory-dedicated regions of the brain, and are
evident in genetic animal models of the condition.
The evidence we review suggests that the autistic cortex
is affected by distinct, low-level changes in neural circuitry
that is dedicated to perceptual processing (including pri-
mary sensory areas). Further, perceptual symptoms in
individuals with autism are evident early in development,
account for independent variance in diagnostic criteria of
the condition, and show a persistent relationship to clini-
cal measures of higher-order social cognition and behav-
iour. We suggest thatan understanding of the perceptual
symptoms in autism may provide insight into signature
differences in canonical neural circuitry that might under-
pin multiple levels of autistic features, and may thus help
to elucidate autistic neurobiology. We also discuss how
primary sensory changes might relate to higher-order
aspects of cognition in autism.
Sensory processing in autism
Sensory symptoms have been clinically documented as
early as 6months of age in infants later diagnosed with
autism16,17 — considerably earlier than children reach key
1Harvard Society of Fellows,
Harvard University,
Cambridge, Massachusetts,
02138, USA.
2McGovern Institute for Brain
Research, Massachusetts
Institute of Technology,
Cambridge, Massachusetts,
02139, USA.
3Department of Psychological
and Brain Sciences,
Dartmouth College,
Hanover, New Hampshire,
03755, USA.
4Autism Research Centre,
Department of Psychiatry,
Cambridge University,
Cambridge, CB2 8AH, UK
*Correspondence to C.E.R
carolinerobertson@fas.
harvard.edu
doi:10.10 38 /n rn. 201 7. 112
Published online 28 Sep 2017
Cognitive empathy
The ability to understand and
respond appropriately to
others’ mental states and
emotions (unlike affective
empathy, the ability to respond
with an appropriate emotion to
others’ mental states or
feelings).
Sensory perception in autism
Caroline E.Robertson1,2,3* and Simon Baron-Cohen4
Abstract | Autism is a complex neurodevelopmental condition, and little is known about its
neurobiology. Much of autism research has focused on the social, communication and cognitive
difficulties associated with the condition. However, the recent revision of the diagnostic criteria
for autism has brought another key domain of autistic experience into focus: sensory processing.
Here, we review the properties of sensory processing in autism and discuss recent computational
and neurobiological insights arising from attention to these behaviours. We argue that sensory
traits have important implications for the development of animal and computational models of
the condition. Finally, we consider how difficulties in sensory processing may relate to the other
domains of behaviour that characterize autism.
NATURE REVIEWS
|
NEUROSCIENCE ADVANCE ONLINE PUBLICATION
|
1
REVIEWS
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ

Joint attention
An early-developing
cornerstone of social cognition:
the child’s ability to use
another person’s gestures and
gaze to direct his or her
attention to objects or events
in the environment.
Broader autism phenotype
Mild autistic traits (in both
social and sensory processing
domains) often observed in
relatives of individuals with
autism in multiplex families.
Multiplex families
Families in which multiple
individuals have an autism
diagnosis; family members
may carry shared genetic
risk factors.
Crowding
The breakdown of visual
recognition of peripheral
stimuli in cluttered visual
environments.
developmental milestones in social cognition, such as joint
attention (14–18months)18. Sensory symptoms not only
precede17 but a lso are pre dictive of s ocial- communication
deficits19 and rep etitiv e behav iours i n childho od20, as well
as eventual diagnostic status19. Assessments of sensory
traits in the broader autism phenotype suggest a genetic
component to these symptoms: the parents and sib-
lings of individuals with autism show higher levels of
self- reported sensory traits relative to the general pop-
ulation21,22. Importantly, greater atypicalities in sensory
processing are observed in families that are thought to
have higher genetic liability for autism (multi plex families)
than in families with a single individual diagnosed with
autism (simplex families), in which the genetic basis
of autism is likely to be denovo21.
Taken together, these findings suggest that such traits
represent early markers of autism. Yet, are these traits pri-
mary, or do they simply reflect secondary outcomes of
alterations in domain-general neural mechanisms, such
as attention? In this section, we briefly review laboratory-
based characterizations of autistic sensory behaviour,
drawing particular attention to replicated findings in the
literature (for in-depth reviews, see REFS11,23–27), before
approaching this question.
Visual detection
Individuals with autism have been characterized as ‘see-
ing the trees, but not the forest’: attuned to details of
the perceptual world at the expense of the global per-
cept they compose28. This framework for understanding
autistic sensory experience emphasizes that perceptual
processing cannot simply be characterized as a talent or a
deficit29 or as re fle ctin g hyp erse nsiti vity or hypo sensitivity.
Rather, perceptual representation in autism exhibits a rel-
ative bias towards local over global features of a sensory
scene, which can be more or less advantageous depending
on task demands30.
This detail-focused perceptual style is well captured
by two studies of autistic visual behaviour (FIG.1). First,
individuals with autism often show faster detection of sin-
gle details (targets) embedded in cluttered visual displays
(that is, among distractors) and a relative insensitivity
to the number of distractors in the display31. This visual
search superiority in autism has been widely replicated31–37
and extended as a promising early marker in toddlers
through eye-tracking38,39. Second, machine-learning
approaches have shown that gaze patterns from individ-
uals with autism during passive viewing of naturalistic,
complex scenes favour scene regions that rank high in
pixel-level saliency (for example, regions that are sali-
ent in terms of contrast, colour or orientation) com-
pared with object-level saliency (for example, relating
to the size, density or contour complexity of objects)
or semantic-level saliency (for example, of text, tools or
faces), which drive gaze biases in neurotypical individu-
als40. This data-driven approach provides a compelling
demonstration of detail-focused visual preferences in
autism, even in the context of naturalistic viewing.
One prediction from these demonstrations would be
that individuals with autism might have superior detec-
tion or discrimination thresholds for static stimuli41.
However, perplexingly, basic measures of visual sen-
sitivity such as visual acuity37,42, contrast discrimina-
tion43,44, orientation processing, crowding45,46 and flicker
detection47,48 have all been shown to be typical in autism,
leaving unresolved the question of how the autistic brain
gives rise to rapid and accurate perception of detail.
There are some replicated atypicalities in low-level
visual processing in autism, particularly in the domain of
high-spatial-frequency stimuli49,50, but these are unlikely
to account for the full magnitude of autistic superiority
in visual search, where stimuli are not necessarily of high
spatial frequency.
Te m p o ra l s y n t h e s i s o f s e n s o r y s i g n a l s
If basic visual detection thresholds for static, local stim-
uli are typical in autism, why do individuals with autism
display altered local–global processing? One possibility is
that perceptual processing in autism may be marked not
by an overall bias towards enhanced local perception but
rather by a shift in the temporal pattern of local–global
processing towards slower global processing51. This may
particularly affect dynamic visual representations, which
are by their nature built up over time. This hypothesis
rests on evidence from research suggesting that temporal
processing of local sensory signals is slower and/or nois-
ier in individuals with autism in the domains of visual,
tactile, auditory and multisensory processing.
Visual motion processing. Unlike with static stimuli,
individuals with autism often exhibit atypical process-
ing of dynamic (social or non-social) visual stimuli52–54
(FIG.1). Although detection thresholds for local motion
are typical55 or even superior in autism56,57, individuals
with autism often struggle with global motion percep-
tion: that is, the ability to discern the global direction
(for example, rightward or leftward motion) of a ‘cloud’
of local visual motion signals (for instance, moving
dots)58,59. These deficits are predictive of the severity of
higher-order autistic symptoms58,59 and are particularly
pronounced when the motion signal is weak or the time
to integrate is short58,59, suggesting that global motion
processing in autism is not disrupted perse but evolves
more slowly overtime.
Ta c t i l e p e r c e p t i o n . As in the visual domain, evidence for
alterations in basic tactile detection thresholds in autism
is mixed — with some studies finding typical60,61 and oth -
ers reporting superior62 or reduced sensitivity compared
with controls7 — although the tactile paradigms used
in these studies vary. One difference in autistic tactile
perception is well replicated: whereas control individuals
present worse detection thresholds for stimuli that grad-
ually increase in amplitude over time into a detectable
range (reflective of dynamic thresholds) relative to acute
stimuli (which require static thresholds), dynamic pres-
entation does not impair tactile sensitivity in individuals
with autism7,63. This difference is proposed to stem from
reduced feedforward inhibition in the autistic sensory
cortex7,64, consistent with magnetoencephalography
findings65, and again suggests alterations in the temporal
features of sensory processing inautism.
REVIEWS
2
|
ADVANCE ONLINE PUBLICATION www.nature.com/nrn
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ

Auditory percept ion. Similar disruptions in the tempo-
ral envelope of sensory processing have been observed
in the domain of auditory processing in autism. Children
with autism often show difficulty discerning the rela-
tive presentation order of two closely occurring tones66
and show delayed evoked neural responses to auditory
tones compared with typically developing children67,68.
This latency in auditory responses predicts autism symp-
tom severity69 and is observed in response to pure tones
as well as to complex, social stimuli (such as speech
sounds)70, raising the hypothesis that this difference
might precipitate higher-order autistic difficulties in
communication71,72.
Multisensory binding. Converging evidence suggests
a deficit in multisensory integration in autism, both in
humans69,70,73–78 and in animal models76,79. Specifically,
individuals with autism demonstrate an elongated win-
dow of audio–visual temporal binding: relative to control
individuals, they are less able to discern the presentation
b Individuals with autism show atypical perception
of global motion
a Individuals with autism show higher pixel-level saliency
c Individuals with autism show weaker binocular rivalry
Where do you naturally look in a scene?
Face, house or mixture?
40%
coherence
Global direction of motion?
Le Right Time
Pixel level
(e.g. colour, intensity,
orientations)
Object level
(e.g. size, solidity,
convexity, eccentricity
of objects)
Semantic level
(e.g. tactile contact
between people and
objects, actions,
text, faces)
Figure 1 | Trade -off in visu al p erce pti on i n au tism . a|eee
reveal greater preferences for scene regions with high pixel-level saliency (for example, regions of high contrast, colour or
orientation) at the expense of regions rich in semantic-level saliency (for example, regions including tactile contact
between people and objects, actions, text and faces)40. The photograph has been modified as an example to highlight
these various levels of image features. b|eeeee
times and higher signal-to-noise ratios to determine the general direction of dynamic stimuli (in the example, a set of dots
moving generally to the right with 40% coherence)53,59. c|eeee
images, one presented to each of an individual’s eyes, alternate back and forth in perception as each is suppressed in turn
eeeeeeeeeee
between the inputs to their left and right eyes, as well as a reduced strength of perceptual suppression (when one image is
fully suppressed from visual awareness). This replicated behavioural signature of autism in vision is predictive of the
severity of social cognition symptoms measured using the Autism Diagnostic Observation Schedule (ADOS)127,132,133.
REVIEWS
NATURE REVIEWS
|
NEUROSCIENCE ADVANCE ONLINE PUBLICATION
|
3
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ

Population r ecep tive fields
A model-driven quantitative
measurement of the average
size and shape of receptive
fields contained within a single
functional MRI voxel.
Cross-activation
Activation of one
sensory-dedicated cortical
region by sensory stimulation
of another modality.
Synaesthesia
The cross-activation of one
sensory modality by
stimulation of another.
Cortical minicolumns
Basic anatomical units of the
neocortex, in which neurons
are arranged in vertical
columns across cortical layers
of the brain.
order of a tone and flash at close temporal offsets and
are more likely to perceive asynchronous events as syn-
chronous74,80. Further, whereas control individuals are
faster to detect a visual stimulus when presented with
an auditory tone as opposed to when presented alone,
this behavioural benefit is reduced in autism, paralleled
by a reduction in multisensory facilitation measured
using electroencephalography81. Deficits in multisensory
binding are particularly observed with audiovisual speech
paradigms74,80,82,83 and may be developmental cornerstones
of deficits in language and communication84 (see below).
A temporal processing problem? In sum, altered tem-
poral processing of sensory stimuli is seen in several
sensory modalities in autism. Specifically, in autism,
local stimuli often elicit delayed evoked responses in the
auditory domain, and integration of multiple local stim-
uli into a global percept often requires a wider window
of temporal binding. These differences may particularly
tax multisensory processing, in which stimuli must be
integrated from two sensory modalities85, and dynamic
perception, in which signals are built up overtime.
Ye t a r e t h e s e p r o c e s s i n g d i f f e r e n c e s a c t u a l l y d i f f e r -
ences in sensation, or could they result from atypical
modulation of sensory processing by higher-order
cognitive mechanisms? For example, superiority dur-
ing conjunctive search could arise from differences in
parallel processing86, deficits in judging global motion
in two-alternative forced-choice tasks could arise from
altered decision criteria87, and reductions in multi-
sensory binding could arise from differences in the
cognitive mechanisms involved in drawing causal infer-
ences88. In the next section, we discuss neuro imaging
findings that demonstrate differences in the low-level
primary sensory areas of the autisticbrain.
Neuroimaging evidence
Consistent with the psychophysical evidence indicating a
low-processing-level origin of the local–global perceptual
style in autism, neuroimaging evidence strongly suggests
that autistic sensory traits are indeed low-level in origin
(FIG.2). Atypical responses in primary sensory cortices
have been observed in autism, across sensory modalities
and during multimodal perception.
Global-motion perception tasks (FIG.1b) involve both
sensory and decision-making processes and have there-
fore been particularly useful in determining whether
autistic perceptual differences are truly sensory in
origin89–91. The slower integration of local motion signals
into a global percept observed in autism58,59,92,93 (discussed
above) could be caused either by an atypical representa-
tion of local motion signals in early visual cortex (in
the primary visual area (V1) and the primary motion
area (MT)) or by alterations in the decision criteria by
which these signals are integrated (in the intraparietal
sulcus (IPS)) over time into a global percept58. Functional
MRI (fMRI) studies have revealed that whereas the IPS
response is typical in autism in these tasks, V1 and MT
show reduced responses to low-strength motion signals
(that is, with short durations and/or low coherence)
in autism — presumably limiting the rate at which
motion signals can be integrated into a global percept
at higher-order processing stages59. Atypical V1 and
MT responses in autism have been observed in several
motion-processing studies94–97, although whether they
can account for deficits in perceiving biological motion
in the condition, or simply contribute to differences in
processing non-social global motion stimuli, is debated98.
In further support of a low-level origin of autistic sen-
sory differences, a robust signature of autistic sensory cor-
tices is an increase in the inter-trial (within- individual)
variability of evoked responses99–101 (FIG.2b). This repli-
cated difference affects the visual, somatosensory and
auditory cortices of individuals with autism (with some
exceptions102) and differentiates people with autism from
individuals with schizophrenia103 (BOX1). This finding
may reflect a disruption of the excitatory–inhibitory
balance (E–I balance), which typically modulates the
trial-by-trial reliability of evoked sensory responses, in
the autistic cortex104. Alterations in the functional archi-
tecture of sensory cortex have been observed as well:
larger population receptive fields have been measured in
extrastriate regions of the autistic visual cortex, includ-
ing MT, and these co-vary with autistic traits105 (FIG.2b).
Another persistent finding in neuroimaging studies of
autism is unexpected cross-activation of visu al cor tex dur-
ing auditory tasks77 — potentially reflecting auditory–
visual synaesthesia, which is more common in people
with autism than in the general population106.
To ge t h er , t h e se f i nd i n gs i n di c at e t ha t ne u r al s i gn a -
tures of autism are evident in early sensory processing
— as early as in primary sensory regions of the autistic
brain. Granted, attention modulates neural responses
in these early sensory regions107,108; thus, it is difficult to
attribute group differences in primary sensory areas
to local changes in sensory signalling rather than to top-
down attentional modulation, especially given that direct
manipulations of attentional load are lacking in the fMRI
studies described above. However, neuro anatomical
changes in low-level primary sensory regions of the autis-
tic brain suggest local alterations in the circuitry of sensory
cortex. For example, cortical minicolumns are reported to be
wider in both the primary auditory cortex and higher-
order association areas in autism109 (but see REF.110 ).
Moreover, be havioura l studies s uggest that aty pical atten-
tional deployment is unlikely to explain the detail-focused
visual perception in autism: although people with autism
show sharper enhancement of visual performance around
a cued location than do control individuals111,112 and have
difficulties tracking multiple moving objects regardless of
object speed113, these individuals show typical measures
of visual performance at the peak of a cued location46,114.
Overall, this pattern of findings is compatible with
the hypothesis that sensory differences are core pheno-
typic markers of autism. Higher-order neural processes
that govern how sensory representations are modified
by attention, integrated towards decision criteria, or
influenced by task demands and expectation may also
be altered in autism. However, given the evidence for
alterations in primary sensory cortex during perceptual
processing in autism, higher-order differences alone
are unlikely to account for the perceptual experience of
REVIEWS
4
|
ADVANCE ONLINE PUBLICATION www.nature.com/nrn
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ

individuals with autism. With this in mind, we consider
some putative alterations in low-level neural circuitry that
may characterize sensory regions of the autisticbrain.
Circuit-level insights
As the neural mechanisms of sensory perception have
been well characterized using electrophysiology and psy-
chophysical approaches, sensory symptoms may offer
concrete insights into circuit-level differences in the
autistic brain14. Indeed, our understanding of the neuro-
biology of autism has undergone many advances owing
to tests of neural circuitry theories of the condition in
the domain of sensory perception. Here, we focus in par-
ticular on the hypothesis that the autistic sensory cortex
might be marked by differences in GABAergic signalling,
as this hypothesis has been tested using neuroimaging
approaches as well as computational approaches.
Reduced GABAergic signalling
E–I imbalance is posited to be a central characteristic of the
neurobiology of autism, inspired in part by the high preva-
lence of seizures (perhaps as high as 1 in 3 by adolescence)
Figure 2 | Neuroimaging evidence for low-level origin of visual symptoms in autism. Atypical representations in
primary sensory areas have been observed in autism in different sensory modalities. a|eee
organizational properties of visual areas are typical in terms of the surface area devoted to each early visual cortical region
(V1, V2, V3 and V4); the cortical magnification function (that is, the ratio in the cortical area dedicated to foveal versus
peripheral representations; not shown); and retinotopic maps, the cortical area dedicated to each part of the visual field,
assessed in terms of polar angle (upper and lower visual fields) or eccentricity (distance from the fovea)105. b|ee
distinct changes in the neurochemical composition, functional architecture and signalling fidelity of early visual cortex
are observed in autism. Specifically, magnetic resonance spectroscopy (MRS) measurements implicate GABA in visual
suppression deficits in autism127. Control individuals evidence a tight linkage between the strength of visual suppression
and GABA levels in visual cortex, but this link is absent in autism (upper graphs). Measurements of the size of population
recepti ve fields in the visual co rtex find larg er populati on receptive f ields in auti sm105 (shown schematically in the middle
graphs). Cortical responses evoked by sensory stimuli (including moving dots, auditory tones and tactile stimuli), as
eeeeee100,103 (schematic responses shown here in the
lower two graphs). The upper two graphs in part b are adapt ed with pe rmission from REF.127, Cell Press/Elsevier.
Nature Reviews | Neuroscience
a Gross organization of visual cortex
is typical in autism b Neurochemical and functional properties
are atypical in autism
Polar angle
Eccentricity
Upper
Meridian
Lower
Periphery
Fovea
Periphery
Periphery
Fovea
Periphery
e
GABA levels do not regulate visual function
Leeeee
Less-reliable evoked potentials
Visual suppression
GABA GABA
Amplitude
Stimulus onset
Time
Stimulus onset
Time
Autism Control
V3
V4
V2
V1
Sizes of visual regions
e
Calcarine
sulcus
REVIEWS
NATURE REVIEWS
|
NEUROSCIENCE ADVANCE ONLINE PUBLICATION
|
5
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ

Binocular rivalry
A visual phenomenon in which
two images, presented
simultaneously to the two eyes,
alternate in perception as
neuronal pools coding for each
eye’s percept compete for
perceptual dominance.
Spatial suppression
A visual phenomenon in which
motion discrimination is
counter- intuitively attenuated
at larger, instead of smaller,
stimulus sizes, probably owing
to suppressive interactions
(centre–surround antagonism
or inhibitory feedback).
Critical period
A developmental period during
which a neural system (such as
vision) is particularly plastic
and sensitive to environmental
influence.
among people with autism115. GABA receptor per-
turbations have been associated with autism through
genetic116–121 and histological studies122, and GABAergic
signalling is disrupted in several different mouse models
of autism123,124. The pivotal roles of GABA in canonical
cortical computations125 and neurodevelopment126 indi-
cate that the GABAergic signalling pathway is key to the
neurobiology of autism12.
Magnetic resonance spectroscopy (MRS) stud-
ies have linked disruptions in autistic visual process-
ing126,127 to GABA concentrations in early visual cortex
(including V1). Specifically, binocular rivalry — a basic
visual function that depends on the strength of inhib-
itory interactions in visual cortex128–131 — is weaker in
autism127,132,133, and this deficit is associated with reduced
GABAergic action in early visual cortex127 (FIGS1c,2b).
This replicated behavioural signature of autism is also
predictive of the severity of social cognition symptoms
measured using the Autism Diagnostic Observation
Schedule (ADOS)127,132. Two further MRS studies have
reported reduced GABA levels in auditory and soma-
tosensory cortex of autistic individuals127,134, suggesting
that reduced inhibition may characterize several corti-
cal regions and perhaps underpin several sensory traits
inautism.
Several behavioural and neuroimaging findings
regarding autistic visual perception have been theo-
retically linked to altered inhibitory neurotransmission in
the brain. Findings of decreased spatial suppression134,135,
atypical representations of motion signals58,59, more
within-individual variability in evoked responses100,136
and expanded population receptive fields105,137 each
recapitulate the effects of blocking GABAergic sensory
signalling in animal studies104,138,139. Yet, which part of
the GABAergic pathway might be atypical in autism
remains unclear. Mixed evidence implicates the avail-
ability of GABA itself 140, the prevalence or integrity of
GABA receptors141–144, the polarity of GABAergic action
(which shifts from excitatory to inhibitory during the
critical period of de velopm ent)145, and the density of cortical
inhibitory interneurons123.
Moreover, various other neurotransmitters and
neuro modulators of GABAergic signalling may have a
role in autistic sensory symptoms. For example, given
that excitatory and inhibitory signalling typically exhibit
homeostatic coupling during sensory development146
and learning147, alterations in GABAergic signalling in
autism might be expected to be accompanied by alter-
ations in excitatory signalling. Indeed, higher levels
of glutamate in blood plasma148 and higher glutamate
Box 1 | Comparison with other psychiatric conditions
Although much progress has been made in characterizing differences in sensory processing in autism, less is known
about which of these differences are unique to autism or are seen in other neurodevelopmental conditions. This point is
crucial for the early identification and translational potential of sensory behavioural assays. Survey-based studies have
detected higher rates of sensory abnormalities in autism compared with other developmental disabilities, such as Down
syndrome219. However, these questionnaire-based observations can only measure the magnitude of sensory sensitivities
in a condition, rather than the characteristics of sensory processing. Below, we review key empirical findings that
highlight similarities and differences in sensory function between autism and other psychiatric conditions. This
evidence of patterns of sensory-processing differences in Rett syndrome, schizophrenia and dyslexia that are distinct
from those in autism lends support to the notion that specific deficits in autistic sensory behaviour may indeed be able
to serve as selective, objective markers of autism.
Rett syndrome
Until the recent revision of the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-V)12, Rett
syndrome (RTT) was included under the diagnostic umbrella of the autism spectrum, as individuals with RTT have many
phenotypic similarities to individuals on the autism spectrum. However, the sensory profile of individuals with RTT is
distinct from that of individuals with autism. Notably, individuals with RTT exhibit differences to control individuals even
in basic visual acuity paradigms171,220, whereas similarly basic measures of low-level visual function (including visual
acuity, contrast sensitivity and flicker detection) are typical in individuals with autism42.
Schizophrenia
Given the evidence for genetic overlap between schizophrenia and autism221, common sensory paradigms have been
used to investigate these conditions. Importantly, such paradigms have revealed distinct patterns of sensory behaviour
differences in autism and schizophrenia. For example, whereas neural responses evoked by sensory stimuli are more
variable in autism100, individuals with schizophrenia show typical response variance and lower-amplitude evoked
responses103. Second, although reduced surround suppression is consistently observed in schizophrenia in many
perceptual tasks222–224, similar deficits are only seen in autism at low stimulus contrasts134. Last, whereas a robust
reducti on in perceptual suppression during binocular ri valry has been observed in au tism127,132,133, the opposite finding
— increased perceptual suppression — is reported in schizophrenia225. Together, these findings illustrate distinct
profiles of alterations in sensory processing in people with autism and individuals with schizophrenia.
Dyslexia
Individuals with dyslexia, similar to individuals with autism, often demonstrate deficits in global-motion perception
compared with controls226 and reduce d activit y in the pri mary moti on area in n euroimagi ng studies227. However,
evidence suggests that global-motion-processing deficits in dyslexia are secondary to reduced time spent reading,
rather than bei ng primary to the cond ition228: deficits are not observed when individuals with dyslexia are compared
with reading-matched typical controls and are ameliorated by reading training229.
REVIEWS
6
|
ADVANCE ONLINE PUBLICATION www.nature.com/nrn
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ

Divisive normalization
A canonical neural
computation in which the
activity of a neuron is divided
by the total activity of
neighbouring neurons to reflect
context-dependent responses.
Pre-pulse inhibition
A sensory phenomenon in
which the behavioural
response to a strong sensory
stimulus is dampened by a
weak preceding stimulus,
probably through feedforward
inhibition.
receptor expression149 have been observed in individu-
als with autism, although empirical links with autistic
symptoms have not yet been reported. Other neuro-
modulators, such as testosterone and oxytocin, modu-
late GABAergic signalling150–152 and are associated with
autistic traits153–155. Future research is needed to establish
the role of these, and other, molecules in modulating
inhibitory signalling in regions of the autisticbrain.
Computational accounts
To da te , mo s t c i rc ui t -level computational accounts
of autism build on a seminal theory that proposed an
excitation-dominant imbalance of neurotransmission
in the autistic brain156. On the basis of this theory, two
computational models that attempted to recapitulate
specific aspects of autistic sensory behaviour support the
hypothesis of reduced inhibition relative to excitation in
autistic visual circuitry157,158.
In the first account, Vattikuti and Chow157 demon-
strate that an excitation-dominant circuit could simulate
reports of less-precise saccadic targeting (dysmetria) and
reduced saccadic velocity (hypometria)159–162 in autism.
The model predicts an increase in recurrent excitatory
activity in the autistic cortex. In turn, this increase is pre-
dicted to reduce the spatial specificity of the neural pop-
ulation code for a saccadic target (leading to dysmetria)
and to dampen sensitivity to activity from outside the
self-excitatory system, therefore decreasing the rate at
which saccadic switching between targets can occur
(leading to hypometria). A second model158 implicates
reduced inhibition in a specific neural computation in
autism. The authors propose that reducing the spatial
spread of inhibition during divisive normalization may
recapitulate two behavioural results in the autism liter-
ature: reduced spatial suppression134 and sharper spatial
processing 112 (but see REF.14).
Computational approaches draw together disparate
findings under a unifying framework that, when informed
by circuitry-level models of neural function, may reveal
generalizable principles of neural differences in autism.
However, a key limitation of such approaches is that they
are developed post hoc to recapitulate select behavioural
deficits and thus risk losing explanatory and predictive
power. We recommend that future computational studies
of autistic behaviour be coupled with empirical tests of
their predictions in novel experimental paradigms.
Overall, converging evidence from neuroimaging,
psychophysics and computational modelling supports
the long-held hypothesis that altered GABAergic inhi-
bition may underpin visual symptoms in autism. Given
the pivotal roles of GABA in canonical cortical compu-
tations125 and neurodevelopment126, future work will
need to interrogate whether the neural changes to the
circuitry in the visual system also characterize other
regions of the autisticcortex.
Tran slation al resea rch
Two l in es o f re se ar ch s up po rt t he n ot io n th at i nve st ig a-
tions into sensory behaviours might provide promising
translational tools for autism research (BOX2; FIG.3). First,
denovo mutations associated with autism converge on
pathways that influence synaptic connectivity, signal-
ling and plasticity163 and therefore would be predicted
to affect wide-ranging neural processes such as sensory
perception that are not necessarily confined to the ‘social
brain’. Second, increasing evidence suggests that sensory
traits are present in common genetic models of autism.
For example, mice with mutations in Mecp2, Gabrb3,
Shank3 or Fmr1 (which encode methyl-CpG-binding
protein 2, GABA type A receptor (GABAAR) subunit-β3,
SH3 and multiple ankyrin repeat domains protein 3, and
fragile X mental retardation protein 1, respectively) all
demonstrate tactile hypersensitivity, as measured in
pre-pulse inhibition para digms 124,164. These sensory traits are
specifically linked to the loss of GABAAR-mediated inhi-
bition in both Mecp2-mutant mice and Gabrb3-mutant
mice124, suggesting that disrupted GABAergic neuro-
transmission is a common feature in multiple genetic
models of autism123,124,165,166 (but see also REF.167).
Genetic animal models of autism have also been
shown to exhibit deficits in multisensory perception.
Mice harbouring genetic mutations in Gad2 (also known
as Gad65; encoding glutamate decarboxylase 2), Shank3
or Mecp2 show reduced electrophysiological signatures
of multisensory integration, which are again specifi-
cally linked to reduced GABAergic signalling in neu-
ral regions implicated in integrating cross-modal input
Box 2 | Genetic animal models of autism
Genetically modified animals represent a powerful tool for discovering circuit-level
alterations in autistic neurobiology. The contribution of genetics to autism is well
established: autism heritability is as high as 54–88% for monozygotic twins, compared
with 10–33% for dizygotic twins230,231, and many genetic risk factors for autism have
been identified through copy number variant, genetic-linkage and genome-wide
association studies232,233. Notably, gene variants that confer high penetrance for autism
occur in a small subset of the autism population — fewer than 2% of individuals with
the condition234,235 — suggesting that the genetic aetiology of autism is complex and
polygenic. Nevertheless, diverse genetic mutations may have converging downstream
effects on specific biological pathways236. Thus, studying neural development in
single-gene mutant animal models of autism may shed light on the aetiology of the
condition by identifying common neurobiological pathways affected by different
autism-associated mutations, along with their contributions to autistic-like traits
in animals.
Animal models of human psychiatric conditions are typically held to three standards
of validity: construct validity (whereby the condition is caused by the same biological
alteration as in humans), face validity (the behaviour of the animal bears a strong
resemblance to human behaviour), and predictive validity (the responses to therapy are
likely to translate into humans)237. Genetic models of autism are exemplars of the first of
these standards, construct validity, as they model a specific genetic mutation that is
found in people with autism.
However, a major challenge for animal research is the lack of face and predictive
validity. Behavioural symptoms in animals rarely present a compelling analogue to
human experience, in part because most core features associated with autism in
humans manifest in social cognitive functions, such as theory of mind or language
comprehension, which are arguably human-specific. Behavioural assays in animal
models of autism have traditionally focused on analogues for repetitive behaviours and
social anxiety, such as marble burying and sociability237 — traits that are not specifically
related to autism but that also manifest in models of obsessive–compulsive disorder
and social anxiety.
Translatable behavioural assays in autism research would facilitate the discovery of
generalizable principles about neural circuitry that can move from animals to humans.
Measures of sensory behaviour represent promising avenues for such translational
assays, given the conserved nature of neural computations involved in sensory
processing between animals and humans14.
REVIEWS
NATURE REVIEWS
|
NEUROSCIENCE ADVANCE ONLINE PUBLICATION
|
7
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ

(including the insular cortex), and which can be rescued
by pharmacologically enhancing GABAergic signalling
in early development76. These atypical electrophysio-
logical and behavioural markers of sensory processing
across multiple monogenic models of autism inde-
pendently implicate reduced GABAergic inhibition in
the aetiology of these features.
Interestingly, one animal model study argues for a
feedforward developmental role for sensory deficits in
the broader autism phenotype. Specifically, targeted pre-
natal (but not adult) silencing of GABAergic inhibition
in peripheral somatosensory neurons, sparing the CNS,
produced tactile sensitivities and social interaction defi-
cits in mice124. Importantly, this finding suggests that sen-
sory symptoms in development alone may be sufficient to
produce social deficits. Further work is needed to isolate
the subcircuits by which this GABAergic deficit in soma-
tosensory pathways led to these social- processing defi-
cits, as well as the relevance of these findings tohumans.
Despite these parallel findings in human and animal
studies, a lack of directly comparable behavioural para-
digms currently limits the potential of studies of sensory
behaviour to serve translational research (BOX2). One
naive assumption in animal research is that sensory per-
ception in autism can be ubiquitously characterized as
‘hypersensitive’ — a generalization that, as we have seen,
does not capture the nuances of sensory behaviour in
people with autism. Standard tests of ‘hypersensitivity’
used in the animal literature, such as startle response
or pre-pulse inhibition, show mixed results in people
with autism168–170. Thus, we suggest that a standardized
set of sensory paradigms that demonstrate clear psycho-
physical differences in humans and that are suitable for
translation into animals would enable validation of
animal models of autism and further research into the
neuro biology of the condition.
An elegant example of work that paves such a clear
translational path can be found in the Rett syndrome
(RTT) literature. Here, neural and behavioural sensory
phenotypes established in animal and human models
strongly correspond: Mecp2-knockout mice and indi-
viduals with RTT both exhibit a reduction in the ampli-
tude of visually evoked responses in V1 and reduced
spatial visual acuity165,171. Interestingly, in the case of
the Mecp2-knockout mice, these differences stem from
a reduction in the activity of GABAergic parvalbumin-
expressing interneurons, which affects both excitatory
and inhibitory responses165. Current candidate para-
digms and measures for translational autism research may
include binocular rivalry127,132,133, multisensory percep-
tion69,70,73–78 and latency of auditory responses67,68 (FIG.3).
Linking with higher-order traits
The co-occurrence of autistic differences in both low-level
perceptual behaviour and high-level social- cognitive pro-
cessing is a central puzzle of autism research. One clue
Nature Reviews | Neuroscience
Theories Humans
Animals
Genetic models
(e.g. Shank3, Gabrb3 or
Mecp2 knockout, or 16p11.2 deletion)
•
•
Environmental risk models
(e.g. MIA model or VPA model)
Computational accounts of
e
(e.g. divisive normalization,
Bayesian inference, lateral inhibition)
•
•
Circuit-level accounts of
alterations in animals (e.g. reduced PV+
neuron- or GABAAR-mediated inhibition)
Neurochemical assays
(e.g. MRS)
Neuroimaging of activity
(e.g. fMRI, EEG and MEG)
Behaviour
(e.g. psychophysics and
clinical observation)
Figure 3 | Sensory symptoms as translational behavioural markers of autism. Given the well-characterized and
evolutionarily conserved nature of sensory circuitry in the brain, perceptual symptoms represent a clear tool for
translational research. Animal-level research can motivate the design of sensory paradigms in humans and,
conversely,aid in delineating alterations in neural circuitry that produce behavioural traits in humans as well as
downstream neural targets that merit investigation in humans. When animal-level research implicates a specific
neurotransmitter pathway, human neuroimaging methods (magnetic resonance spectroscopy (MRS) or positron
emission tomography (PET)) can probe the integrity of this pathway in humans and test whether it underpins behavioural
differences. Computational and circuit-level theories of human and animal-level findings can help to uncover principles
of neural function that might generalize across sensory paradigms, across sensory modalities and even to other domains
of autistic traits. A crucial next step is to test whether genetic models of autism recapitulate the observed human-level
eeeeeeeeee
GABAAeeeeeeee+, parvalbumin-
expressing; VPA, valproic acid.
REVIEWS
8
|
ADVANCE ONLINE PUBLICATION www.nature.com/nrn
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ

McGurk effect
A perceptual illusion in which
a sound (for example, of the
syllable /ba/) paired with a
visual signal (a mouth
pronouncing /ga/) produces
athird percept (voice and
mouth /da/).
Pragmatics
The ability to use the social
context of an utterance to
inform and communicate
meaning.
is that perceptual differences in individuals with autism
often predict the severity of higher-order autistic tra
its36,38,52,69,105,112,127,172–177 in the laboratory setting. Large-
scale studies also demonstrate covariance between
questionnaire-based measures of sensory sensitivities
and autistic traits in the general population11,177,178, in
both Western and Japanese cultures179. This correlation
presents a strong argument for a relationship between
sensory and social-cognitive processing in autism (BOX3).
Neurobiolo gical accounts of how and why these lower-
and higher-order symptoms might be related in autism are
largely divided into two camps. ‘Sensory-first’ accounts
posit that social-cognitive symptoms may be downstream
effects of atypical sensory processing in early develop-
ment, whereas ‘top-down’ accounts posit that symptoms
in sensation and social cognition might co-arise from
alterations in domain-general mechanisms (such as atten-
tion, decision-making or causal inference) that affect both
levels of information processing in thebrain.
In this section, we discuss evidence for and against
these accounts. There are, of course, many theories of
autism that offer elegant accounts of one domain — but
not multiple domains — of autistic features11,41,180,181; how-
ever, here we focus only on theories that offer a unifying
account of diverse domains of autistic behaviour. Finally,
we highlight a third approach, which we call the ‘canonical
micro-circuitry view’, that posits that disparate levels of
autistic features share common neural mechanisms.
Sensory-first accounts
Sensory-first accounts, which are motivated in part by
studies of sensory deprivation during child institutional-
ization182,183, hold that atypical sensory processing during
early development causally stunts typical development
of social cognition, in a feedforward manner. After all,
dynamic sensory information is the medium of social
communication: subtle fluctuations in the pitch of spo-
ken language cue prosody, coordinated motions of the
face communicate emotions and cues relevant to empa-
thy184, and the preparatory motions of a person’s body
relative to other objects in the world communicate inten-
tions and requests185. Thus, a child who struggles to inte-
grate dynamic sensory information may also struggle to
build social information into meaningful representations
or, alternatively, may find social information confusing
and therefore self-select away from exposure or engage-
ment with social information186,187. As discussed above,
this hypothesis seems consistent with recent findings
in animals124.
Elegant research on the relationship between multi-
sensory binding deficits and language processing in
autism supports such a feedforward causal link. The abil-
ity to perceptually bind sensory signals across auditory
and visual senses is fundamental to language perception,
as it facilitates the integration of vocal and facial cues188.
As discussed above, individuals with autism often show
reduced multisensory binding66, particularly for social
stimuli (such as faces and voices)72,83. Furthermore,
altered audiovisual binding thresholds in autism pre-
dict less-robust integration of visual and auditory sig-
nals of spoken language in tests of the McGurk effect189 as
well as a lower ability to accurately perceive speech in a
noisy auditory environment84. These studies clearly illus-
trate how differences in basic sensory processing might
affect the development of higher-order functions such as
language perception.
Sensory-first accounts have strong merits but also
shortcomings. First, among verbal individuals with
autism, differences in language processing compared
with that in neurotypical controls particularly peak in
the domain of pragmatics190, but why sensory differences
would particularly affect this feature is not clear. Second,
the neural basis of theory of mind is comparable in blind
and sighted individuals, suggesting its development
does not depend on typical sensory experience191. Thus,
although sensory-first views may account for difficulties
in certain aspects of language development, it is difficult
to explain top-level autistic deficits in theory of mind
from cascading difficulties in building stable sensory
representations.
To p - d ow n a c c o u n t s
Top-down accounts posit that a centralized deficit in
domain-general cognitive processes (such as attention,
decision-making or causal inference) underpins defi-
cits in both sensory and social-cognitive processing
in autism. One instantiation of this theory is the ‘weak
central coherence’ hypothesis, which posits that autistic
neurobiology is characterized by a centralized pertur-
bation of neural processes that aggregate information
Box 3 | Time to give up on a unified account of autism?
Some have argued that it is time to give up on a centralized account of autistic traits
and that perhaps the disparate categories of autistic symptoms each have
independent genetic causes and neural origins238. This argument largely rests on
studies of autistic personality traits in large normative twin samples (including more
than 3,000 twin pairs), which suggest that the degree of parent-reported autistic-like
trait severity in social, communicative and repetitive behaviours are only modestly
genetically related in typically developing children239–241. Further, autistic-like trait
severity in typically developing children sometimes ‘peaks’ in single trait areas: it is
estimated that 10% of children show autistic traits in only one symptom domain238.
By contrast, studies of individuals with autism suggest stronger genetic overlap
between autistic symptom domains. One small twin sample of autistic individuals
found that common genetic factors represent the primary drivers of both social-
communication symptoms and repetitive behaviours, with high heritability242.
Furthermore, if autistic symptom domains are indeed fractionable, it remains
perplexing that autistic perceptual symptoms often strongly co-vary with
social-cognitive symptoms both at the population level175–178,214 and in the
laboratory36,38,52,105,112,127,172–174, as well as with clinical assessments of repetitive
behaviours20,243,244.
Resolving the question of whether autistic behavioural domains are indeed
fractionable will require overcoming three limitations of past studies. First, strong
phenotypic measures of autistic behaviour across symptom domains that can be
adapted for large-scale studies are needed to directly probe the relationship between
symptom domains, rather than relying on measures of parental report, which show
only modest test–retest reliability240. Second, genotyping individuals with autism may
provide more clarity regarding genetic factors shared by different autistic symptom
domains than twin studies afford245. Last, sensory behaviours should be assessed
in genetic studies of the autistic phenotype, as they are now included in the
diagnostic criteria of autism and show clear experimental links to social and
communicative traits12.
In the meantime, we suggest it may be premature to give up on the hypothesis
that symptoms of autism in different domains spring from common neurobiological
and genetic origins.
REVIEWS
NATURE REVIEWS
|
NEUROSCIENCE ADVANCE ONLINE PUBLICATION
|
9
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ

Neural motifs
Stereotyped, local neural
circuits that are found in
multiple regions of the brain
and participate in common
canonical computations (such
as habituation, response
normalization or biased
competition).
Bayesian perceptual
inference
A model of perception in which
prior knowledge about a
stimulus is combined with
noisy, stimulus-evoked sensory
signals to infer the percept and
generate prediction errors.
Ambiguity resolution
The ability to impose meaning
on ambiguous sensory
information. Two or more
interpretations may be equally
viable (as in bistable visual
phenomena, such as in
binocular rivalry) or can be
disambiguated using
contextual information.
(sensory or cognitive) into coherent percepts or cogni-
tions192. In this account, sensory signalling is presumed
to be unaffected in the autistic brain; rather, a higher-
order mechanism that integrates these representations
is altered. A neurobiological realization of such a theory
would be, for example, deficits in association areas of
the brain, where multimodal sensory representations are
integrated with task demands.
However, top-down accounts such as the weak cen-
tral coherence hypothesis are not immediately com-
patible with the empirical findings of alterations in
low-level sensory cortex discussed above. It is particu-
larly challenging to use top-down accounts to explain
the neuroanatomical observations of altered mini-
column architecture in not only the association cortices
but also the primary auditory cortex of post-mortem
autistic brains109. It is therefore unlikely that centralized
cognitive accounts will be able to provide a unifying fac-
tor for autistic symptoms in sensory and higher-order
cognition.
Canonical micro-circuitry view
We turn now to a third hypothesis, which we call the
canonical micro-circuitry view. This view is largely
inspired by genetic studies in autism that implicate
changes in synaptic connectivity, signalling and plas-
ticity in the condition163. Such low-level changes would
not necessarily be confined to particular cortical regions
(such as the ‘social brain’) but would be predicted to
affect basic components of neural circuits throughout
the brain163. Given that many cortical regions share neural
motifs14,193 that participate in common canonical compu-
tations, genetic disruption of neural motifs might affect
many regions of the brain and produce structurally sim-
ilar behavioural traits in various perceptual and cognitive
domains194.
In addition to divisive normalization (mentioned
above), another candidate neural motif-mediated com-
putation that has recently been implicated in autism
is Bayesian perceptual inference. Bayesian inference has
been shown to be implemented in every neural domain,
including sensory perception195, motor planning196,
language197, social cognition198 and proprioception199.
Individuals with autism have been posited to have per-
ceptual representations in which bottom-up sensory
input is weighted more than top-down predictions181.
Other theories challenge this hypothesis, holding that
autistic perception could instead be characterized by
imprecise sensory representations200, aberrant weight-
ing of sensory prediction errors201 or aberrant updating
of priors202. Compellingly, one recent study shows a
reduced reliance on implicitly learned priors when dis-
criminating sensory representations in a volatile envi-
ronment, an effect that is reflected in reduced measures
of surprise, as derived from pupil dilation, in individu-
als with autism203. Further empirical studies of autistic
behaviour are needed to disentangle these hypotheses
and to specify the levels of cortical processing at which
Bayesian inference might be altered in autism. However,
the rubric of Bayesian inference presents the opportu-
nity to test whether systematic failures of a common
computational principle might account for different
domains of autistic symptomatology. For example, can
weaker priors aptly describe autistic performance on
sensory, pragmatic- language and social-cognitive tasks?
How might we go about identifying altered neural
motifs in autism? We suggest that this is a two-part
endeavour that involves both human and animal model
research. In human research, we might start by identify-
ing behavioural paradigms in which similar differences
in autism can be observed across different domains of
processing (for example, in perception, language and
cognition) and that might therefore engage a common
neural motif. One example of such a task might be
ambiguity resolution. In visual perception, individuals
with autism are slower than controls to resolve low-
level perceptual ambiguity of two conflicting inputs
presented to the two eyes (binocular rivalry)127,132,133.
Similarly, in pragmatic language, when presented
with sentences containing words that could have two
meanings (for example, homographs such as ‘bow’ or
‘bass’), children with autism struggle to resolve this
ambiguity, failing to use the sentence context to inform
their pronunciation and often defaulting to the more
common pronunciation204,205. Ambiguity resolution in
both domains of representation — in visual perception
and in language — may plausibly rely on neural motifs
consisting of reciprocal inhibitory competitive interac-
tions between neural populations that vie for perceptual
representation206,207.
Such a motif may even be a neural substrate of
theory- of-mind challenges in autism1: during theory-
of-mind judgements, the child must co-activate and
flexibly alternate between two, sometimes conflicting,
representations of the world — their own understand-
ing and another person’s — to navigate social interac-
tions. Interestingly, the ability to resolve ambiguity in
perception (during perceptual bistability), in language
(homo graph understanding) and in social cognition
(theory of mind) tends to develop at approximately the
fourth year of life208,209. Furthermore, individual dif-
ferences in the onset of these abilities correlate across
domains209,210, suggesting that ambiguity resolution
across these processing domains may be linked. Notably,
in children with autism, individual differences in percep-
tual bistability predict theory-of-mind performance and
ADOS scores127,132,209.
Once a potential neural motif has been identified in
humans, animal model research may help to identify spe-
cific disruptions in neural circuitry that underpin this
motif (FIG.3). When animal model research implicates a
specific neurotransmitter pathway, human neuroimaging
studies using MRS or positron emission tomography can
probe the integrity of this pathway in humans and test
whether it underpins behavioural differences. Finally, a
crucial step will be to test whether genetic animal models
of autism recapitulate the observed human-behavioural
differences. Given the relative ease of translating sen-
sory behavioural findings between humans and animal
models, further research into symptoms of autism in
the sensory domain may lead to promising translational
opportunities.
REVIEWS
10
|
ADVANCE ONLINE PUBLICATION www.nature.com/nrn
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ

Looking forward
This Review posits that sensory symptoms are core,
primary characteristics of the neurobiology of autism.
Specifically, sensory processing differences in autism are
visible early in development211–213, as early as infancy16, and
are predictive of diagnostic status later in childhood19,38,39.
They predict higher-order deficits in social and cognitive
function in adults178,214 and explain independent variance
in social and communication symptoms in diagnostic
assessments215. Moreover, autism-associated sensory
symptoms reflect alterations in sensory-dedicated neural
circuitry59,100, including neuro molecular and anatomical
changes in primary sensory regions of the brain64,109,127,
rather than secondary consequences of alterations in
higher-order cognitive processes. These differences
manifest in both humans and genetic animal models of
autism, in which GABAergic signalling is often commonly
affected64,76,123,124,127, holding promise for translational
biomarkers of the condition.
This conclusion marks a revolutionary shift in our
conception of autism from its early diagnostic charac-
terizations13 and calls into question modern ‘social brain’
theories15, in which sensory deficits are hypothesized to
be epiphenomenal to core deficits in social processing.
Moving forward, neurobiological theories of autism
must account for atypical processing in both social and
sensory domains.
One of the biggest challenges to formulating neuro-
biological theories of autism has been the persistent
difficulty of documenting robust, replicable differ-
ences between individuals with autism and controls,
even with simple tests of sensory processing. Given the
genetic heterogeneity of the autistic population, one
promising contribution of sensory paradigms may be
the ability to stratify the autistic population into more
homogeneous subgroups of individuals who share com-
mon underlying neurobiological alterations, such as on
the basis of sensory differences that are associated with
certain genetic polymorphisms216,217. Indeed, sensory
subtypes are often reported in children with autism in
clinical surveys218. Identifying and characterizing such
subgroups in the laboratory setting will require the
analysis of larger samples than are typically used. In
the meantime, replications in independent samples of
participants and a number of statistical practices must be
used to ensure meaningful between-group comparisons,
including using nonparametric statistics when data vio-
late assumptions of normality, bootstrapping statistical
comparisons to minimize the effects of outliers, match-
ing groups on relevant psychophysical factors, and eye
tracking when retinal position is a relevant variable for
task performance.
Auti sm affects e very domain of human exp erience:
from sensation and perception to motor behaviour,
emotion, communication and cognition. A central
challenge of autism research is to understand how
these disparate domains might be related. We sug-
gest that research on sensory symptoms may be able
to help untangle this complexity, shedding light on
circuit-level alterations in the brain that might affect
various domains of cortical processing in autism and
offering avenues for translational research.
1. Baron-Cohen,S., Leslie,A.M. & Frith,U. Does the
autistic child have a ‘theory of mind’? Cognition 21,
37–46 (1985).
2. Baron-Cohen,S. & Wheelwright,S. The empathy
quotient: an investigation of adults with Asperger
syndrome or high functioning autism, and normal sex
differences. J.Autism De v. Diso rd. 34, 163–175
(2004).
3. To m c h e k , S.D. & Dunn,W. Sensory processing in
children with and without autism: a comparative study
using the Short Sensory Profile. Am. J.Occup. Ther.
61, 190–200 (2007).
4. Tavassoli,T., Miller,L.J., Schoen,S.A.,
Nielsen,D.M. & Baron-Cohen,S. Sensory over-
responsivity in adults with autism spectrum
conditions. Autism 18, 428–432 (2014).
5. Ta va s s o l i , T. & B a r o n - C o h e n , S. Taste identification in
adults with autism spectrum conditions. J.Autism Dev.
Disord. 42, 1419–1424 (2012).
6. Marco,E.J. etal. Children with autism show reduced
somatosensory response: an MEG study. Autism Res.
5, 340–351 (2012).
7. Puts,N.A.J., Wodka,E.L., Tommerdahl,M.,
Mostofsky,S.H. & Edden,R.A.E. Impaired tactile
processing in children with autism spectrum
disorder. J.Neurophysiol. 111, 1803–1811
(2014).
8. Bonnel,A. etal. Enhanced pitch sensitivity in
individuals with autism: a signal detection analysis.
J.Cogn. Neurosci. 15, 226–235 (2003).
9. Galle,S., Courchesne,V., Mo ttr on ,L. & Frasnelli,J.
Olfaction in the autism spectrum. Perce ptio n 42,
341–355 (2013).
10. Rozenkrantz,L. etal. A mechanistic link between
olfaction and autism spectrum disorder. Curr. Biol. 25,
1904–1910 (2015).
11. Simmons,D.R. etal. Vision in autism spectrum
disorders. Vision Res. 49, 2705–2739 (2009).
12. American Psychiatric Association. Diagnostic and
Statistical Manual of Mental Disorders (DSM-5) 5th
edn (American Psychiatric Association, 2013).
13. Kanner,L. Autistic disturbances of affective contact.
Nerv. Child 2, 217–250 (1943).
14. Heeger,D.J., Behrmann,M. & Dinstein,I. Vision as
abeachhead. Biol. Psychiatry 81, 832–837 (2016).
15. Pelphr ey,K.A., Shultz,S., Hudac,C.M. & Vander
Wyk,B. C. Research review: constraining
heterogeneity: the social brain and its development in
autism spectrum disorder. J.Child Psychol. Psychiatry
52, 631–644 (2011).
16. Baranek,G.T. etal. Hyporesponsiveness to social and
nonsocial sensory stimuli in children with autism, children
with developmental delays, and typically developing
children. Dev. Psychopathol. 25, 307–320 (2013).
17. Estes,A. etal. Behavioral, cognitive, and adaptive
development in infants with autism spectrum disorder
in the first 2years of life. J.Neurodev. Disord. 7, 24
(2015).
This longitudinal study of infants at risk of autism
diagnoses shows that high-risk infants have higher
levels of sensorimotor symptoms at 6months of
age and that these symptoms predict subsequent
diagnostic status..
18. Brooks,R. & Meltzoff,A.N. The importance of eyes:
how infants interpret adult looking behavior. Dev.
Psychol. 38, 958–966 (2002).
19. Tur n e r - B r o w n , L.M., Baranek,G.T., R e z n i c k , J.S.,
Wats on ,L.R. & Crais,E.R. The First Year Inventory:
alongitudinal follow-up of 12-month-old to 3-year-old
children. Autism 17, 527–540 (2013).
20. Boyd,B.A. etal. Sensory features and repetitive
behaviors in children with autism and developmental
delays. Autism R es. 3, 78–87 (2010).
21. Donaldson,C.K., Stauder,J.E.A. & Donkers,F.C.L.
Increased sensory processing atypicalities in parents
of multiplex ASD families versus typically developing
and simplex ASD families. J.Autism De v. Diso rd. 47,
535–548 (2017).
22. Glod,M., Riby,D.M., Honey,E. & Rodgers,J. Sensory
atypicalities in dyads of children with autism spectrum
disorder (ASD) and their parents. Autism R es. 10,
531–538 (2017).
23. Mikkelsen,M., Wodka,E.L., Mostofsky,S.H. &
Puts,N.A.J. Autism spectrum disorder in the scope
of tactile processing. Dev. Cogn. Neurosci. http://dx.
doi.org/10.101 6/j.d cn .2 016.12 .0 05 (2017) .
24. Haesen,B., Boe ts, B. & Wage man s,J. A review of
behavioural and electrophysiological studies on
auditory processing and speech perception in autism
spectrum disorders. Res. Autism Spectr. Disord. 5,
701–714 ( 2011).
25. Baum,S.H., Stevenson,R.A. & Wallace,M.T.
Behavioral, perceptual, and neural alterations in
sensory and multisensory function in autism
spectrum disorder. Prog. Neurobiol. 134, 140–160
(2015).
26. Stevenson,R.A. etal. Keeping time in the brain:
autism spectrum disorder and audiovisual temporal
processing. Autism Res. 9, 720–738 (2016).
27. Wallac e,M.T. & S t e ve ns o n , R.A. The construct of the
multisensory temporal binding window and its
dysregulation in developmental disabilities.
Neuropsychologia 64, 105–123 (2014).
28. Shah,A. & Frith,U. An islet of ability in autistic
children: a research note. J.Child Psychol. Psychiatry
24, 613–620 (1983).
29. Happe,F., Br i s k ma n , J. & Frith,U. Exploring the
cognitive phenotype of autism: weak ‘central
coherence’ in parents and siblings of children with
autism. Experimental tests. I. J.Child Psychol.
Psychiatry 42, 309–316 (2001).
30. Baron-Cohen,S., Ashwin,E., Ashwin,C., Tavassoli,T.
& Chakrabarti,B. Talent in autism: hyper-systemizing,
hyper-attention to detail and sensory hypersensitivity.
Phil. Trans. R.Soc. Lond. B Biol. Sci. 364,
1377–1383 (2009).
31. Plaisted,K., O’Riordan,M. & Baron-Cohen,S.
Enhanced visual search for a conjunctive target in
autism: a research note. J.Child Psychol. Psychiatry
39, 777–783 (1998).
This foundational study shows faster target
detection during conjunctive visual search in autism
— a finding now widely replicated.
REVIEWS
NATURE REVIEWS
|
NEUROSCIENCE ADVANCE ONLINE PUBLICATION
|
11
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ

32. Plaisted,K., O’Riordan,M. & Baron-Cohen,S.
Enhanced discrimination of novel, highly similar
stimuli by adults with autism during a perceptual
learning task. J.Child Psychol. Psychiatry 39,
765–775 (1998).
33. O’Riordan,M.A., Plaisted,K.C., Driver,J. & Baron-
Cohen,S. Superior visual search in autism. J.Exp.
Psychol. Hum. Percept. Perform. 27, 719–730
(2001).
34. Keehn,B., Br enn er,L., Palmer,E., Lincoln,A.J. &
Müller,R.-A. Functional brain organization for visual
search in ASD. J.Int. Neuropsychol. Soc. 14,
990–1003 (2008).
35. Baldassi,S. etal. Search superiority in autism within,
but not outside the crowding regime. Vision Res. 49,
2151– 2156 (2009 ).
36. Joseph,R.M., Keehn,B., Con nol ly,C., Wolfe,J.M. &
Horowitz,T.S. Why is visual search superior in autism
spectrum disorder? Dev. Sci. 12, 1083–1096 (2009).
37. Kéïta ,L., Mottron,L. & Bertone,A. Far visual acuity is
unremarkable in autism: do we need to focus on
crowding? Autism R es. 3, 333–341 (2010).
38. Gliga,T. , B e d fo r d , R., Charman,T. & J o h n s o n ,M.H.
Enhanced visual search in infancy predicts emerging
autism symptoms. Curr. Biol. 25, 1727–1730 (2015).
39. Kaldy,Z., Kraper,C., Carter,A.S. & Blaser,E.
Toddlers with autism spectrum disorder are more
successful at visual search than typically developing
toddlers. Dev. Sci. 14, 980–988 (2011).
40. Wang,S. etal. Atypical visual saliency in autism
spectrum disorder quantified through model-based
eye tracking . Neuron 88, 604–616 (2015).
This data-driven analysis of gaze patterns during
natural scene-viewing in individuals with autism
shows that gaze patterns in autism are biased
towards scene regions that are salient in terms of
low-level pixel features (such as contrast, colour
and orientation).
41. Mottron,L., Dawson,M., Soulières,I., Hubert,B. &
Burack,J. Enhanced perceptual functioning in autism:
an update, and eight principles of autistic perception.
J.Autism Dev. Disord . 36, 27–43 (2006).
42. Ta v a s s o l i , T., L a t h a m , K., Bach,M., Dakin,S.C. &
Baron-Cohen,S. Psychophysical measures of visual
acuity in autism spectrum conditions. Vision Res. 51,
1778–1780 (2011).
43. de Jonge,M.V. etal. Visual information processing in
high-functioning individuals with autism spectrum
disorders and their parents. Neuropsychology 21,
65–73 (2007).
44. Koh, H.C., Milne,E. & Dobkins,K. Spatial contrast
sensitivity in adolescents with autism spectrum
disorders. J.Autis m Dev. Disor d. 40, 978–987 (2010).
45. Freyberg,J., Robertson,C.E. & Baron-Cohen,S.
Typical magnitude and spatial extent of crowding in
autism. J.Vis. 16, 17 (2016).
46. Grubb,M.A. etal. Endogenous spatial attention:
evidence for intact functioning in adults with autism.
Autis m Res. 6, 108–118 (2013).
47. Bertone,A., Mottron,L., Jelenic,P. & Fa u b e r t, J.
Enhanced and diminished visuo-spatial information
processing in autism depends on stimulus complexity.
Brain 128, 2430–2441 (2005).
48. Pellic an o, E., Gibson,L., Maybery,M., Durkin,K. &
Badcock,D.R. Abnormal global processing along the
dorsal visual pathway in autism: a possible mechanism
for weak visuospatial coherence? Neuropsychologia
43, 1044–1053 (2005).
49. Kéïta,L., Guy,J., Berthiaume,C., Mottron,L. &
Bertone,A. An early origin for detailed perception in
autism spectrum disorder: biased sensitivity for high-
spatial frequency information. Sci. Rep. 4, 5475 (2014).
50. Caplette,L., Wicker,B. & G oss eli n,F. Atypical time
course of object recognition in autism spectrum
disorder. Sci. Rep. 6, 35494 (2016).
51. Van de r Ha ll en, R. Evers,K., Brewaeys,K.,
Van de n No ort ga te,W. & Wa gem ans ,J. Global
processing takes time: a meta-analysis on local–global
visual processing in ASD. Psychol. Bull. 141,
549–573 (2015).
52. Blake,R., Turner,L.M., Smoski,M.J., Pozdol,S.L. &
Stone,W.L. Visual recognition of biological motion is
impaired in children with autism. Psychol. Sci. 14,
151–157 (2003).
53. Spencer,J. etal. Motion processing in autism:
evidence for a dorsal stream deficiency. Neuroreport
11, 2765–2767 (2000).
54. Koh,H.C., Milne,E. & Dobkins,K. Contrast sensitivity
for motion detection and direction discrimination in
adolescents with autism spectrum disorders and their
siblings. Neuropsychologia 48, 4046–4056 (2010).
55. Koldewyn,K., Whitney,D. & Rivera,S.M.
The psychophysics of visual motion and global form
processing in autism. Brain 133, 599–610 (2010).
56. Manning,C. etal. Enhanced integration of motion
information in children with autism. J.Neurosci. 35,
6979–6986 (2015).
57. Chen,Y. etal. Enhanced local processing of dynamic
visual information in autism: evidence from speed
discrimination. Neuropsychologia 50, 733–739
(2012).
58. Robertson,C.E., Martin,A., Baker,C.I. & Baron-
Cohen,S. Atypical integration of motion signals in
autism spectrum conditions. PLoS ONE 7, e48173
(2012).
59. Robertson,C.E. etal. Global motion perception
deficits in autism are reflected as early as primary
visual cortex. Brain 137, 2588–2599 (2014).
60. Güçlü,B., Tanid ir,C., Mukaddes,N.M. & Ünal,F.
Tactile sensitivity of normal and autistic children.
Somatosens. Mot. Res. 24, 21–33 (2007).
61. O’Riordan,M. & Passetti,F. Discrimination in autism
within different sensory modalities. J.Autism Dev.
Disord. 36, 665–675 (2006).
62. Blakemore,S. etal. Tactile sensitivity in Asperger
syndrome. Brain Cogn 61, 5–13 (2006).
63. Ta v a s s o l i , T. etal. Altered tactile processing in children
with autism spectrum disorder. Aut ism R es. 9,
616– 620 (2016) .
64. Puts,N.A.J. etal. Reduced GABA and altered
somatosensory function in children with autism
spectrum disorder. Autism Res. 10, 608–619 (2017).
65. Khan,S. etal. Somatosensory cortex functional
connectivity abnormalities in autism show opposite
trends, depending on direction and spatial scale.
Brain 138, 1394–1409 (2015).
66. Kwakye,L.D., Foss-Feig,J.H., Cascio,C.J.,
Stone,W.L. & Wallace,M.T. Altered auditory and
multisensory temporal processing in autism spectrum
disorders. Front. Integr. Neurosci. 4, 129 (2011).
67. Edgar,J.C. etal. Neuromagnetic oscillations predict
evoked-response latency delays and core language
deficits in autism spectrum disorders. J. Autis m Dev.
Disord. 45, 395–405 (2013).
68. Roberts,T.P. L. etal. MEG detection of delayed
auditory evoked responses in autism spectrum
disorders: towards an imaging biomarker for autism.
Autis m Res. 3, 8–18 (2010).
69. Brandwein,A.B. etal. Neurophysiological indices of
atypical auditory processing and multisensory
integration are associated with symptom severity in
autism. J.Autism Dev. Dis ord. 45, 230–244 (2014).
70. Roberts,T. P. L. etal. Auditory magnetic mismatch
field latency: a biomarker for language impairment in
autism. Biol. Psychiatry 70, 263–269 (2011).
71. Oram Cardy,J.E., Flagg,E.J., Roberts,W. &
Roberts,T.P. L. Auditory evoked fields predict
language ability and impairment in children.
Int. J.Psychophysiol. 68, 170–175 (2008).
72. Stevenson,R.A., Segers,M., Ferber,S.,
Barense,M.D. & Wallace,M.T. The impact of
multisensory integration deficits on speech perception
in children with autism spectrum disorders. Front.
Psychol. 5, 379 (2014).
This is a recent review of evidence for robust multi-
sensory-integration deficits in autism and their
implications for difficulties in language
development.
73. Bebko,J.M., Weiss,J.A., Demark,J.L. & Gomez,P.
Discrimination of temporal synchrony in intermodal
events by children with autism and children with
developmental disabilities without autism. J.Child
Psychol. Psychiatry 47, 88–98 (2006).
74. Stevenson,R.A. etal. Multisensory temporal
integration in autism spectrum disorders. J.Neurosci.
34, 691–697 (2014).
75. Stevenson,R.A. etal. Evidence for diminished
multisensory integration in autism spectrum
disorders. J.Autis m Dev. Disor d. 44, 3161–3167
(2014).
76. Gogolla,N., Takesian,A.E., Feng,G., Fagiolini,M. &
Hensch,T.K. Sensory integration in mouse insular
cortex reflects GABA circuit maturation. Neuron 83,
894–905 (2014).
77. Jao Keehn,R.J. etal. Impaired downregulation of
visual cortex during auditory processing is associated
with autism symptomatology in children and
adolescents with autism spectrum disorder. Autism
Res. 10, 130–143 (2017).
78. Collignon,O. etal. Reduced multisensory facilitation
in persons with autism. Cortex 49, 1704–1710
(2013).
79. Siemann,J.K. etal. An autism-associated serotonin
transporter variant disrupts multisensory processing.
Tr a n s l P s y c h i a t r y 7, e1067 (2017).
80. Noel,J.P., D e N i e a r , M.A., Stevenson,R., Alais,D. &
Wall ac e,M.T. Atypical rapid audio-visual temporal
recalibration in autism spectrum disorders. Autism
Res. 10, 121–129 (2017).
81. Brandwein,A.B. etal. The development of
multisensory integration in high-functioning autism:
high-density electrical mapping and psychophysical
measures reveal impairments in the processing of
audiovisual inputs. Cereb. Cortex 23, 1329–1341
(2013).
82. Stevenson,R.A. etal. Multisensory speech perception
in autism spectrum disorder: From phoneme to whole-
word percept ion. Aut ism Res. 10, 1280–1290
(2017).
83. Foxe,J.J. etal. Severe multisensory speech
integration deficits in high-functioning school-aged
children with autism spectrum disorder (ASD) and
their resolution during early adolescence. Cereb.
Cortex 25, 298–312 (2015).
84. Stevenson,R.A. etal. The cascading influence of
multisensory processing on speech perception in
autism. Autism http://dx.doi.
org/10.117 7/136 23 6131 770 441 3 (2017).
85. Brock,J., Brown,C.C., Boucher,J. & Rippon,G.
The temporal binding deficit hypothesis of autism.
Dev. Psychopathol. 14, 209–224 (2002).
86. Nakayama,K. & Silverman,G.H. Serial and parallel
processing of visual feature conjunctions. Nature 320,
264–265 (1986).
87. Churchland,A.K., Kiani,R. & Shadlen,M.N.
Decision-making with multiple alternatives.
Nat. Neurosci. 11, 693–702 (2008).
88. Magnotti,J.F., Ma , W.J. & Beauchamp,M.S. Causal
inference of asynchronous audiovisual speech. Front.
Psychol. 4, 798 (2013).
89. Shadlen,M.N. & Newsome,W.T. Motion perception:
seeing and deciding. Proc. Natl Acad. Sci. USA 93,
628–633 (1996).
90. Mikami,A., Newsome,W.T. & W u r t z , R.H. Motion
selectivity in macaque visual cortex. I. Mechanisms of
direction and speed selectivity in extrastriate area MT.
J.Neurophysiol. 55, 1308–1327 (1986).
91. Snowden,R.J., Treue,S., Erickson,R.G. &
Andersen,R. The response of area MT and V1
neurons to transparent motion. J.Neurosci. 11,
2768–2785 (1991).
92. Zaidel,A., Goin-Kochel,R.P. & A n g e l a k i , D.E. Self-
motion perception in autism is compromised by visual
noise but integrated optimally across multiple senses.
Proc. Natl Acad. Sci. USA 112 , 6461–6466 (2015).
93. Manning,C., Charman,T. & Pe l li c a n o ,E. Processing
slow and fast motion in children with autism spectrum
conditions. Autism Res. 6, 531–541 (2013).
94. Ta k a r a e, Y. , Lu n a, B., Min she w,N.J. & Sweeney,J.A.
Visual motion processing and visual sensorimotor
control in autism. 20, 113–122 (2014).
95. Peiker,I. etal. Stronger neural modulation by visual
motion intensity in autism spectrum disorders. PLoS
ONE 10, 1–17 (2015).
96. Herrington,J.D. etal. The role of MT+/V5 during
biological motion perception in Asperger Syndrome:
an fMRI study. Res. Autism Spectr. Disord. 1, 14–27
(2007).
97. Greimel,E. etal. The temporal dynamics of coherent
motion processing in autism spectrum disorder:
evidence for a deficit in the dorsal pathway. Behav.
Brain Res. 251, 168–175 (2013).
98. Kold ewyn ,K., Whitney,D. & Rivera,S.M. Neural
correlates of coherent and biological motion perception
in autism. Dev. Sci. 14, 1075–1088 (2011).
99. Milne,E. Increased intra-participant variability in
children with autistic spectrum disorders: evidence
from single-trial analysis of evoked EEG. Front.
Psychol. 2, 51 (2011).
100. Dinstein,I. etal. Unreliable evoked responses in
autism. Neuron 75, 981–991 (2012).
This key study shows that, although the magnitude
of evoked responses to sensory stimuli is typical in
autism, the signal-to-noise ratio of sensory
responses is poor, owing to higher intra-subject
variance in the respo nse ampl itude in autism.
101. Haigh,S.M., Minshew,N., Heeger,D.J., Dinstein,I.
& Behrmann,M. Over-responsiveness and greater
variability in roughness perception in autism. 9,
393–402 (2016).
102. Butler,J.S., Molholm,S., Andrade,G.N. & Foxe,J.J.
An examination of the neural unreliability thesis of
autism. Cereb. Cortex 27, 1–16 (2017).
REVIEWS
12
|
ADVANCE ONLINE PUBLICATION www.nature.com/nrn
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ

103. Haigh,S.M. etal. Differential sensory fMRI signatures
in autism and schizophrenia: analysis of amplitude
and trial-to-trial variability. Schizophr. Res. 175,
12–19 (2016).
104. Shen,W., McK eow n,C.R., Demas,J.A. & Cline,H.T.
Inhibition to excitation ratio regulates visual system
responses and behavior invivo. J.Neurophysiol. 106,
2285–2302 (2011).
105. Schwarzkopf,D.S. etal. Larger extrastriate
population receptive fields in autism spectrum
disorders. J.Neurosci. 34, 2713–2724 (2014).
106. Baron-Cohen,S. etal. Is synaesthesia more common
in autism? Mol. Autism 4, 40 (2013).
107. Kast ner,S., De Weerd,P. , D es i m o n e, R. &
Ungerleider,L. Mechanisms of directed attention in
the human extrastriate cortex as revealed by
functional MRI. Science 282, 108–111 (1998).
108. Silver,M.A., Ress,D. & Heeger,D.J. Neural
correlates of sustained spatial attention in human
early visual cortex. J.Neurophysiol. 97, 229–237
(2007).
109. Mckavanagh,R., Buckley,E. & Chance,S.A. Wider
minicolumns in autism: a neural basis for altered
processing? Brain 138, 2034–2045 (2015).
This recent, large post-mortem study observed
wider minicolumns in the brains of individuals; this
alteration to a basic building block of neural
architecture affects both primary sensory cortex
(A1) and higher-order association areas.
110 . Casanova,M.F. etal. Minicolumnar abnormalities
in autism. Acta Neuropathol. 112, 287–303
(2006).
111. Ronconi,L., Gori,S., Ruffino,M., Molteni,M. &
Facoe tti,A. Zoom-out attentional impairment in
children with autism spectrum disorder. Cortex 49,
1025 –1033 (2013 ).
112 . Robertson,C.E., Kravitz,D.J., Freyberg,J., Baron-
Cohen,S. & Baker,C.I. Tunnel vision: sharper
gradient of spatial attention in autism. J.Neurosci.
33, 6776–6781 (2013).
113 . Koldewyn ,K., Weigelt,S., Kanwisher,N. & Jiang,Y.
Multiple object tracking in autism spectrum disorders.
J.Autism Dev. Disord . 43, 1394–1405 (2013).
114 . Grubb,M.A. etal. Exogenous spatial attention:
evidence for intact functioning in adults with autism
spectrum disorder. J.Vis. 13, 9 (2013).
115 . Canitano,R. Epilepsy in autism spectrum disorders.
Eur. Child Adolesc. Psychiatry 16, 61–66 (2007).
116 . Sanders,S.J. etal. Multip le recurren t de novo CNVs,
including duplications of the 7q11.23 Williams
syndrome region, are strongly associated with autism.
Neuron 70, 863–885 (2011).
117 . Griswold,A.J. etal. Evaluation of copy number
variations reveals novel candidate genes in autism
spectrum disorder-associated pathways. Hum. Mol.
Genet. 21, 3513–3523 (2012).
118 . Shao,Y. etal. Fine mapping of autistic disorder to
chromosome 15q11-q13 by use of phenotypic
subtypes. Am. J.Hum. Genet. 72, 539–548 (2003).
119 . Ma,D.Q. etal. Identification of significant association
and gene-gene interaction of GABA receptor subunit
genes in autism. Am. J.Hum. Genet. 77, 377–388
(2005).
120. Chen,C.H. etal. Genetic analysis of GAB RB3 as a
candidate gene of autism spectrum disorders. Mol.
Autis m 5, 36 (2014).
121. Piton,A. etal. Analysis of the effects of rare variants
on splicing identifies alterations in GABAA receptor
genes in autism spectrum disorder individuals. Eur.
J.Hum. Genet. 21, 749–756 (2013).
122. Fatem i,S.H., Reutiman,T.J., Folsom,T.D. &
Thuras,P.D. GABAA receptor downregulation in brains
of subjects with autism. J.Autism Dev. Disord . 39,
223–230 (2009).
123. Gogolla,N. etal. Common circuit defect of exc itatory–
inhibitory balance in mouse models of autism.
J.Neurodev. Disord. 1, 172–181 (2009).
This study provides insights into common
circuit-level differences observed in multiple mouse
models of autism with distinct genetic origins, by
reporting a reduction in parvalbumin-expressing
inhibitory neurons across parietal and occipital
cortex.
124. Orefice,L.L. etal. Peripheral mechanosensory
neuron dysfunction underlies tactile and behavioral
deficits in mouse models of ASDs. Cell 166, 299–313
(2016).
This new study demonstrates that several mouse
models of autism exhibit tactile hypersensitivities,
which arise from disruptions to GABAAR-mediated
inhibition.
125. Tu r k h e i m e r, F. E., Leech,R., Expert,P., L o r d , L.-D. &
Ver no n, A.C. The brain’s code and its canonical
computational motifs. From sensory cortex to the
default mode network, a multi-scale model of brain
function in health and disease. Neurosci. Biobehav.
Rev. 55, 211–222 (2015).
126. Hensch,T.K. Critical period plasticity in local cortical
circuits. Nat. Rev. Neurosci. 6, 877–888 (2005).
127. Robertson,C.E., Ratai,E.M. & Kanwisher,N.
Reduced GABAergic action in the autistic brain.
Curr. Biol. 26, 80–85 (2016).
This recent study links a replicated autistic visual
deficit in binocular rivalry to disrupted GABAergic,
but preserved glutamatergic, action in the autistic
visual cortex.
128. Laing,C.R. & Chow,C.C. A spiking neuron model for
binocular rivalry. J.Comput. Neurosci. 12, 39–53
(2002).
129. Seely,J. & Chow,C.C. Role of mutual inhibition in
binocular rivalry. 106, 2136–2150 (2012).
130. Said,C.P. & H e e g e r, D.J. A model of binocular rivalry
and cross-orientation suppression. PLoS Comput. Biol.
9, e1002991 (2013).
131. Van Loo n,A.M. etal. GABA shapes the dynamics of
bistable perception. Curr. Biol. 23, 823–827 (2013).
132. Robertson,C.E., Kravitz,D.J., Freyberg,J., Baron-
Cohen,S. & Baker,C.I. Slower rate of binocular
rivalry in autism. J.Neurosci. 33, 16983–16991
(2013).
133. Freyberg,J., Robertson,C.E. & Baron-Cohen,S.
Reduced perceptual exclusivity during object and
grating rivalry in autism. J.Vis. 15, 11 (2015).
134. Foss- Feig, J.H., Tadin,D., Schauder,K.B. &
Cascio,C.J. A substantial and unexpected
enhancement of motion perception in autism.
J.Neurosci. 33, 8243–8249 (2013).
135. Sysoeva,O.V., Ga lu ta ,I.A., Davletshina,M.S.,
Orekhova,E.V. & S tr og an ova ,T. A. Abnormal size-
dependent modulation of motion perception in
children with autism spectrum disorder (ASD). Front.
Neurosci. 11, 164 (2017).
136. Haigh,S.M., Heeger,D.J., Dinstein,I., Minshew,N. &
Behrmann,M. Cortical variability in the sensory-
evoked respon se in autism. J. Auti sm Dev. Disord. 45,
117 6 –11 9 0 (2 01 4) .
137. Schauder,K.B., Park ,W.J., Tadin,D. & Bennetto,L.
Larger receptive field size as a mechanism underlying
atypical motion perception in autism spectrum
disorder. Clin. Psychol. Sci. http://dx.doi.
org/10.117 7/2167 702617 7077 33 (2017).
138. Thiele,A., Herrero,J.L., Distler,C. & Hoffmann,K.P.
Contribution of cholinergic and GABAergic
mechanisms to direction tuning, discriminability,
response reliability, and neuronal rate correlations in
macaque middle temporal area. J.Neurosci. 32,
16602–16615 (2012).
139. Lund,J.S., Wu,Q., Hadingham,P.T. & L e v i t t , J.B.
Cells and circuits contributing to functional properties
in area V1 of macaque monkey cerebral cortex: bases
for neuroan atomically realistic mode ls. J.Anat. 187,
563–581 (1995).
140. Yip,J., Soghomonian,J.J. & Blatt,G.J. Decreased
GAD67 mRNA levels in cerebellar Purkinje cells in
autism: pathophysiological implications. Acta
Neuropathol. 113 , 559–568 (2007).
141. Blatt,G.J. etal. Density and distribution of
hippocampal neurotransmitter receptors in autism: an
autoradiographic study. J.Auti sm Dev. Dis ord. 31,
537–543 (2001).
142. Fatem i,S.H., Folsom,T.D., Reutiman,T.J. &
Thuras,P.D. Expression of GABAB receptors is altered
in brains of subjects with autism. Cerebellum 8,
64–69 (2009).
143. Collins,A.L. etal. Investigation of autism and GABA
receptor subunit genes in multiple ethnic groups.
Neurogenetics 7, 167–174 (2006).
144. Menold,M.M. etal. Association analysis of
chromosome 15 GABAA receptor subunit genes in
autistic disorder. J.Neurogenet. 15, 245–259
(2001).
145. Ben-Ari,Y. Is birth a critical period in the pathogenesis
of autism spectrum disorders? Nat. Rev. Neurosci. 16,
498–505 (2015).
146. Dorrn,A.L., Yuan,K., Barker,A.J., Schreiner,C.E. &
Froemke,R.C. Developmental sensory experience
balances cortical excitation and inhibition. Nature
465, 932–936 (2010).
147. Shibata,K. etal. Overlearning hyperstabilizes a skill
by rapidly making neurochemical processing
inhibitory-dominant. Nat. Neurosci. 20, 470–475
(2017).
148. Zheng,Z. etal. Blood glutamate levels in autism
spectrum disorder: a systematic review and meta-
analysis. PLoS ONE 11, e0158688 (2016).
149. Fatem i,S.H., Folsom,T.D., Kneeland,R.E. &
Liesch,S.B. Metabotropic glutamate receptor 5
upregulation in children with autism is associated with
underexpression of both fragile X mental retardation
protein and GABAA receptor beta 3 in adults with
autism. Anat. Rec. (Hoboken) 294, 1635–1645
(2011).
150. Zhang,L. etal. The expression of GABAA recepto r α2
subunit is upregulated by testosterone in rat cerebral
cortex. Neurosci. Lett. 265, 25–28 (1999).
151. Harden,S.W. & Fra zie r,C.J. Oxytocin depolarizes
fast-spiking hilar interneurons and induces GABA
release onto mossy cells of the rat dentate gyrus.
Hippocampus 26, 1124–1139 (2016).
152. Owen,S.F. etal. Oxytocin enhances hippocampal
spike transmission by modulating fast-spiking
interneurons. Nature 500, 458–462 (2013).
153. Baron-Cohen,S. etal. Why are autism spectrum
conditions more prevalent in males? PLoS Biol. 9,
e1001081 (2011).
154. Baron-Cohen,S. etal. Elevated fetal steroidogenic
activity in autism. Mol. Psychiatry 20, 369–376
(2015).
155. Gordon,I. etal. Oxytocin enhances brain function in
children with autism. Proc. Natl Acad. Sci. USA 110,
20953–20958 (2013).
156. Rubenstein,J.L.R. & Merzenich,M.M. Model of
autism: increased ratio of excitation/inhibition in key
neural systems. Brain 2, 255–267 (2003).
157. Vat ti ku ti ,S. & Chow,C. A computational model for
cerebral cortical dysfunction in autism spectrum
disorders. Biol. Psychiatry 67, 672–678 (2010).
158. Rosenberg,A., Patterson,J.S. & Angelaki,D.E. A
computational perspective on autism. Proc. Natl
Acad. Sci. USA 112 , 9158–9165 (2015).
159. Rosenhall,U., Johansson,E. & Gillberg,C.
Oculomotor findings in autistic children. J.Laryngol.
Otol. 102, 435–439 (1988).
160. Ta ka r a e , Y., Mi ns h ew, N.J., Luna,B., Kri sk y,C.M. &
Sweeney,J.A. Pursuit eye movement deficits in
autism. Brain 127, 2584–2594 (2004).
161. Ta ka r a e , Y., Mi ns h ew, N.J., Luna,B. & Sw een ey,J.A.
Oculomotor abnormalities parallel cerebellar
histopathology in autism. J.Neurol. Neurosurg.
Psychiatry 75, 1359–1361 (2004).
162. Luna,B., Dol l,S.K., Hegedus,S.J., Minshew,N.J. &
Sweeney,J.A. Maturation of executive function in
autism. Biol. Psychiatry 61, 474–481 (2007).
163. De Rubeis,S. etal. Synaptic, transcriptional and
chromatin genes disrupted in autism. Nature 515,
209–215 (2014).
164. DeLorey,T.M. etal. Somatosensory and sensorimotor
consequences associated with the heterozygous
disruption of the autism candidate gene, Gabrb3.
Behav. Brain Res. 216, 36–45 (2011).
165. Banerjee,A. etal. Jointly reduced inhibition and
excitation underlies circuit-wide changes in cortical
processing in Rett syndrome. Proc. Natl Acad. Sci.
USA 113 , E7287–E7296 (2016).
166. Selimbeyoglu,A. etal. Modulation of prefrontal cortex
excitation/inhibition balance rescues social behavior in
CNTNAP2-deficient mice. Sci. Transl Med. 9,
eaah6733 (2017).
167. Ta b u c h i , K. etal. A neurolig in-3 mutation implicated in
autism increases inhibitory synaptic transmission in
mice. Science 318, 71–76 (2007).
168. McAlonan,G.M. etal. Brain anatomy and
sensorimotor gating in Asperger’s syndrome. Brain
125, 1594–1606 (2002).
169. Madsen,G.F., B i l e n be r g ,N., Cantio,C. & Oranje,B.
Increased prepulse inhibition and sensitization of the
startle reflex in autistic children. Autism Re s. 7,
94–103 (2014).
170. Kohl, S. etal. Prepulse inhibition of the acoustic
startle reflex in high functioning autism. PLoS ONE 9,
e92372 (2014).
171. LeBlanc,J.J. etal. Visual evoked potentials detect
cortical processing deficits in Rett syndrome.
Ann. Neurol. 78, 775–786 (2015).
172. Sutherland,A. & Crewther,D.P. Magnocellular
visual evoked potential delay with high autism
spectrum quotient yields a neural mechanism for
altered perception. Brain 133, 2089–2097
(2010).
173. Flevaris,A.V. & M ur ray, S.O. Orientation-specific
surround suppression in the primary visual cortex
varies as a function of autistic tendency. Front. Hum.
Neurosci. 8, 1–9 (2015).
REVIEWS
NATURE REVIEWS
|
NEUROSCIENCE ADVANCE ONLINE PUBLICATION
|
13
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ

174. Keehn ,B. & Jo sep h,R.M. Exploring what’s missing:
what do target absent trials reveal about autism
search superiority? J.Autism De v. Disord. 46,
1686–1698 (2016).
175. Mayer,J.L. The relationship between autistic traits
and atypical sensory functioning in neurotypical and
ASD adults: a spectrum approach. J.Autism Dev.
Disord. 47, 1–12 (2016).
176. Ta ka y a m a , Y. etal. Standardization of the Japanese
version of the Glasgow Sensory Questionnaire (GSQ).
Res. Autism Spectr. Disord. 8, 347–353 (2014).
177. Horder,J., Wilson,C.E., Mendez,M.A. &
Murphy,D.G. Autistic traits and abnormal sensory
exper iences i n adult s. J. Autism Dev. Disord. 44,
1461–1469 (2013).
178. Tav a s s o l i , T. , H o e k s t r a , R.A. & Baron-Cohen,S. The
Sensory Perception Quotient (SPQ): development and
validation of a new sensory questionnaire for adults
with and without autism. Mol. Autism 5, 29 (2014).
179. Ujiie,Y. & W ak ab a ya sh i ,A. Psychometric properties
and overlap of the GSQ and AQ among Japanese
university students. Int. J.Psychol. Stud. 7, 195–205
(2015).
180. Mottron,L. & Burack,J.A. in The Development of
Autis m Perspect ives fro m Theory and Res earc h (eds
Burack,J.A., Charman,T., Y i r m i y a , N. & Zelazo,P.R.)
131–148 (Erlbaum, 2001).
181. Pell ic ano,E. & Burr,D. When the world becomes ‘too
real’: a Bayesian explanation of autistic perception.
Tr e n d s C o g n . S c i . 16, 504–510 (2012).
182. Cermak,S.A. & Daunhauer,L.A. Sensory processing
in the postinstitutionalized child. Am. J.Occup. Ther.
51, 500–507 (1997).
183. Nelson,C.A.A. Neurobiological perspective on early
human deprivation. Child Dev. Perspect. 1, 13–18
(2007).
184. Bassili,J.N. Facial motion in the perception of faces
and of emotional expression. J.Exp. Psychol. Hum.
Perc ept. Perfo rm. 4, 373–379 (1978).
185. Crane,E. & Gross,M. in Affective Computing and
Intelligent Interaction (eds Paiva ,A.C.P., P r a d a , R. &
Picard,R.W.) 95–101 (Springer Berlin Heidelberg,
2007).
186. Markram,H., Rinaldi,T. & M a r k r a m , K. The intense
world syndrome — an alternative hypothesis for
autism. Front. Neurosci. 1, 77–96 (2007).
187. Brock,M.E. etal. Temperament and sensory features
of children with autism. J.Autism Dev. Disord . 42,
2271–2284 (2012).
188. Kuhl, P. K. & Meltzoff,A.N. The bimodal perception of
speech in infancy. Science 218 , 1138–1141 (1982).
189. McGurk,H. & MacDonald,J. Hearing lips and seeing
voices. Nature 264, 746–748 (1976).
190. Ta g e r - F l u s b e rg , H. On the nature of linguistic
functioning in early infantile autism. J.Autism Dev.
Disord. 11, 45–56 (1981).
191. Bedny,M., Pascual-Leone,A. & Saxe,R.R. Growing
up blind does not change the neural bases of Theory
of Mind. Proc. Natl Acad. Sci. USA 106,
113 12 – 113 17 ( 20 0 9) .
192. Happé,F. & F r it h , U. The weak coherence account:
detail-focused cognitive style in autism spectrum
disorders. J.Autis m Dev. Disor d. 36, 5–25 (2006).
193. Mountcastle,V.B. The columnar organization of the
neocortex. Brain 120, 701–722 (1997).
194. Reynolds,J.H. & Heeger,D.J. The normalization
model of attention. Neuron 61, 168–185 (2009).
195. Lee,T.S. & Mumford,D. Hierarchical Bayesian
inference in the visual cortex. J.Opt. Soc. Am. A Opt.
Image Sci. Vis. 20, 1434–1448 (2017).
196. Shipp,S., Adams,R.A. & Friston,K.J. Reflections on
agranular architecture: predictive coding in the motor
cortex. Tre n d s N e u r o s c i . 36, 706–716 (2013).
197. Frank,M.C. & Goodman,N.D. Predicting pragmatic
reasoning in language games. Science 336, 998
(2012).
198. Koste r-Ha le,J. & Saxe,R. Theory of Mind: a neural
prediction problem. Neuron 79, 836–848 (2013).
199. Barrett,L.F. & Simmons,W.K. Interoceptive
predictions in the brain. Nat. Rev. Neurosci. 16,
419–429 (2015).
200. Brock,J. Alternative Bayesian accounts of autistic
perception: comment on Pellicano and Burr. Trends
Cogn. Sci. 16, 573–574 (2012).
201. Van de Cru ys,S., Van der Hallen,R. & Wagemans,J.
Disentangling signal and noise in autism spectrum
disorder. Brain Cogn. 112 , 78–83 (2016).
202. Haker,H., Schneebeli,M. & Stephan,K.E. Can
Bayesian theories of autism spectrum disorder help
improve clinical practice? Front. Psychiatry 7, 107
(2016).
203. Lawson,R.P., M a t h y s, C. & Rees,G. Adults with
autism overestimate the volatility of the sensory
environment. 2017, 4–6 (2017).
204. Frith,U. & Snowling,M. Reading for meaning and
reading for sound in autistic and dyslexic children.
Br. J.Dev. Psychol. 1, 329–342 (1983).
205. D’Souza,D., Booth,R., Connolly,M., Happé,F. &
Karmi loff-Sm ith ,A. Rethinking the concepts of ‘local
or global processors’: evidence from Williams
syndrome, Down syndrome, and Autism Spectrum
Disorders. Dev. Sci. 19, 452–468 (2016).
206. Rayomnd,W. & Gib bs, Jr. A new look at literal
meaning in understanding what is said and implicated.
J.Pragmat. 34, 457–486 (2002).
207. Desimone,R. & Duncan,J. Neural mechanisms of
selective visual attention. Annu. Rev. Neurosci. 18,
193–222 (1995).
208. Gopnik,A. & Rosati,A. Duck or rabbit? Reversing
ambiguous figures and understanding ambiguous
representations. Dev. Sci. 4, 175–183 (2001).
209. Sobel,D.M., Capps,L.M., Gopnik,A., Capps,L.M. &
Gopnik,A. Ambiguous figure perception and theory of
mind understanding in children with autistic spectrum
disorders. Br. J.Dev. Psychol. 23, 159–174 (2005).
210. Doherty,M.J. & Wimmer,M.C. Children’s
understanding of ambiguous figures: which cognitive
developments are necessary to experience reversal?
Cogn. Dev. 20, 407–421 (2005).
211. Rogers,S.J., Hepburn,S. & Wehner,E. Parent reports
of sensory symptoms in toddlers with autism and
those with other developmental disorders. J.Autism
Dev. Disord. 33, 631–642 (2003).
212. Kern ,J.K. etal. The pattern of sensory processing
abnormalities in autism. Autism I nt. J.Res. Pract. 10,
480–494 (2006).
213. Baranek,G.T. Autism during infancy: a retrospective
video analysis of sensory-motor and social behaviors
at 9–12months of age. J.Autism De v. Diso rd. 29,
213– 224 (1999 ).
214. Robertson,A.E. & Simmons,D.R. Sensory
sensitivities in autism spectrum disorders:
a qualitative analysis [Abstract]. International Meeting
for Autism Re search (2010).
This large-scale survey, now twice replicated, shows
that trait measures of sensory sensitivities and
social interaction difficulties co-vary in the general
population.
215. Frazier,T.W. etal. Validatio n of propos ed DSM-5
criteria for autism spectrum disorder. J.Am. Acad.
Child Adolesc. Psychiatry 51, 28–40.e3 (2012).
216. Goodbourn,P. T. etal. Variants in the 1q21 risk region
are associated with a visual endophenotype of autism
and schizophrenia. Genes Brain Behav. 13, 144–151
(2014).
217. Goodbourn,P. T. etal. Cortical patterning genes are
associated with individual differences in visual
orientation perception. Invest. Ophthalmol. Vis. Sci.
55, 6396 (2014).
218. Ausd erau ,K.K. etal. Sensory subtypes in children
with autism spectrum disorder: latent profile
transition analysis using a national survey of sensory
features. J.Child Psychol. Psychiatry. 55, 935–944
(2014).
219. Rogers,S.J. & Ozonoff,S. Annotation: what do we
know about sensory dysfunction in autism? A critical
review of the empirical evidenc e. J.Child Psychol.
Psychiatry. 46, 1255–1268 (2005).
220. Te t z c h n e r, S. etal. Vision, cognition, and
developmental characteristics of girls and women with
Rett syndrome. Dev. Med. Child Neurol. 38, 212–225
(1996).
221. Lee,S.H. etal. Genetic relationship between five
psychiatric disorders estimated from genome-wide
SNPs. Nat. Genet. 45, 984–994 (2013).
222. Tibber,M.S. etal. Visual surround suppression in
schizophrenia. Front. Psychol. 4, 88 (2013).
223. Yoo n , J.H. etal. Diminished orientation-specific
surround suppression of visual processing in
schizophrenia. 35, 1078–1084 (2009).
224. Ta d i n , D. etal. Weakened center-surround interactions
in visual motion processing in schizophrenia.
J.Neurosci. 26, 11403–11412 (2006).
225. Stanley,J., Park,S., Blake,R. & Carter,O. Binocular
rivalry dynamics and mixed percept in schizophrenia.
Front. Hum. Neurosci. http://dx.doi.org/10.3 389/c on f.
fnhum.2015.217.00325 (2015).
226. Cornelissen,P., R i c h a r d s o n , A., Mason,A., Fowler,S.
& Stein,J. Contrast sensitivity and coherent motion
detection measured at photopic luminance levels in
dyslexics and controls. Vision Res. 35, 1483–1494
(1995).
227. Eden,G.F. etal. Abnormal processing of visual motion
in dyslexia revealed by functional brain imaging.
Nature 382, 66–69 (1996).
228. Goswami,U. Sensory theories of developmental
dyslexia: three challenges for research. Nat. Rev.
Neurosci. 16, 43–54 (2015).
229. Olulade,O.A., Napoliello,E.M. & Eden,G.F.
Abnormal visual motion processing is not a cause of
dyslexia. Neuron 79, 180–190 (2013).
230. Sandin,S. etal. The famili al risk of a utism. JAM A 311,
1770 (2014).
231. Hallmayer,J. etal. Genetic heritability and shared
environmental factors among twin pairs with
autism. Arch. Gen. Psychiatry 68, 1095
(2011).
232. Ronemus,M., Iossifov,I., Levy,D. & Wigler,M. The
role of denovo mutations in the genetics of autism
spectrum disorders. Nat. Rev. Genet. 15, 133–141
(2014).
233. Chen,J.A., Peñagarikano,O., Belgard,T.G.,
Swarup,V. & Geschwind,D.H. The emerging picture
of autism spectrum disorder: genetics and
pathology. Annu. Rev. Pathol. 10, 111–144
(2015).
234. Moessner,R. etal. Contrib ution of SHANK3 mutations
to autism spectrum disorder. Am. J.Hum. Genet. 81,
1289–1297 (2007).
235. Yi,F. etal. Autism-associated SHANK3
haploinsufficiency causes Ih channelopathy in human
neurons. Science 352, aaf2669 (2016).
236. Bourgeron,T. From the genetic architecture to
synaptic plasticity in autism spectrum disorder.
Nat. Rev. Neurosci. 16, 551–563 (2015).
237. Silverman,J.L., Yang,M., Lord,C. & Crawley,J.N.
Behavioural phenotyping assays for mouse models
of autism. Nat. Rev. Neurosci. 11, 490–502
(2010).
238. Happé,F., R o n a ld , A. & Plomin,R. Time to give up on
a single explanation for autism. Nat. Neurosci. 9,
1218–1220 (2006).
239. Ronald,A., Happé,F. & P l o m in , R. The genetic
relationship between individual differences in social
and nonsocial behaviours characteristic of autism. 5,
444–458 (2005).
240. Ronald,A. etal. Genetic heterogeneity between the
three components of the autism spectrum: a twin
study. J.Am. Acad. Child Adolesc. Psychiatry 45,
691–699 (2006).
241. Robinson,E.B. etal. A multivariate twin study of
autistic traits in 12-year-olds: testing the fractionable
autism triad hypothesis. Behav. Genet. 42, 245–255
(2012).
242. Frazier,T.W. etal. A twin study of heritable and
shared environmental contributions to autism.
J.Autism Dev. Disord . 44, 2013–2025 (2014).
243. Lidstone,J. etal. Relations among restricted and
repetitive behaviors, anxiety and sensory
features in children with autism spectrum
disorders. Res. Autism Spectr. Disord. 8, 82–92
(2014).
244. Boyd,B.A., McBee,M., Holtzclaw,T., Baranek,G.T.
& Bodfish,J.W. Relationships among repetitive
behaviors, sensory features, and executive
functions in high functioning autism. Res. Autism
Spectr. Disord. 3, 959–966 (2009).
245. Lynch,M. & Walsh,B. Genetics and analysis of
quantitative traits (Sinauer, 19 98).
Ackno wle dgement s
Th e au th ors th an k M. Coh en , A. Spi eg el, G. C ho i,
N.Kanwisher, A. J. Haskins, and M. Sur for comments on
sections of this manuscript and helpful discussion.
Autho r co ntr ibu tio ns
C.E.R. researched the article. C.E.R. and S.B.-C. equally con-
tributed to discussions of the content, writing the article and
to review and/or editing of the manuscript before
submission.
Competing interests statement
The authors declare no competing interests.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
SUPPLEMENTARY INFORMATION
See online article: S1 (box)
ALL LINKS ARE ACTIVE IN THE ONLINE PDF
REVIEWS
14
|
ADVANCE ONLINE PUBLICATION www.nature.com/nrn
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ
ǟɥƐƎƏƗɥ!,(++-ɥ4 +(2'#12ɥ(,(3#"Ʀɥ/13ɥ.$ɥ/1(-%#1ɥ341#ƥɥ++ɥ1(%'32ɥ1#2#15#"ƥ

Box S1 | How does sensation fit into the history of autism?
1943: First descriptions of autism include sensory features.
Kanner1 describes atypical sensory behaviours (including heightened sensitivity to noise and touch,
attraction to visual patterns and spinning objects, finger-stimming in front of the eyes) in seven of the
eleven children with autism whom he observed, but he considered these symptoms to be
epiphenomenal to core deficits in interacting with the external world: “intrusion comes from loud
noises and moving objects […] Yet it is not the noise or motion itself that is dreaded. The
disturbance comes from the noise or motion that intrudes itself, or threatens to intrude itself, upon
the child’s aloneness” (p. 245).
1947 and 1960–1970: Origins of ‘sensory-first’ theories of autism.
Bergman and Escalona (1947)2 hypothesize that sensory sensitivities have an early developmental
onset in autism and cause the child to withdraw from the social world, proposing that: “these
children start life with a high degree of sensitivity against which they eventually succeed in building
some defenses” (p. 345).
Eveloff (1960)3 hypothesizes that altered sensory processing interferes with the development of self-
representations in autism: “Briefly, a sense of self apparently depends a good deal on the integration
of the body image with properly cathected, well-functioning distance receptors of sound, sight, and
smell” (p. 102).
Hermelin and O’Connor (1970)4 posit that individuals with autism show an atypical preference for
specific sensory modalities over others (e.g. visual and tactile over auditory), perhaps leading to
difficulties with language development.
1967: Sensory processing abnormalities are posited as underlying impairments in autism.
Psychologist Lorna Wing notes the “detail-oriented” perceptual style in autism: “attending to minor
or trivial aspects of people or objects instead of attending to the whole (p. 50)”. She proposes this
feature along with “abnormal responses to sensory experiences (p. 49)” as diagnostic criteria in the
first scheme for the clinical description of autism5. These features, however, are not included among
the first DSM diagnostic criteria for autism in 19806.
1978: Robustness of sensory features in autism is questioned.
Richer7 firmly argues that disruptions of sensory processing in autism are red-herrings: sensory
processing is not affected in autism and any theory based on sensory processing in autism: “is as
incoherent as it is untestable”. Similarly, Rutter8 argues that “visuo-spatial perceptual defects do not
play any essential role in the development of infantile autism” (p. 91), forwarding the hypothesis
language impairment is at the heart of the condition.
1982: Experimental demonstrations of visuospatial talent in autism.
Using the Embedded Figures Test, Shah and Frith9 demonstrate measurable visuospatial processing
talents in autism, which they posit are the result of a cognitive reduction in the drive for meaning:
“Perhaps they [the children with autism] were able to locate the target figure so easily because the

overall meaning of the complex figure (or the embedding context) was not relevant or dominant for
them” (p. 618).
1983: Origins of ‘social-perception’ theories of autism.
Rutter10 concludes that sensory symptoms are byproducts of primary deficits in social cognition:
“The data suggest that the deficit does not lie in the processing of stimuli of any particular sensory
modality or, indeed, of stimuli that are defined in terms of any particular sensory qualities. Rather, it
appears that the stimuli that pose difficulties for autistic children are those that carry emotional or
social ‘meaning’” (p. 528).
1987: First cognitive-based theories of sensory symptoms in autism.
Frith and Baron-Cohen11 later posit that: “low-level perceptual processes are intact in autistic
children. Instead, [talents in visuospatial processing] indicate a central cognitive deficit as the
primary problem area” (p. 99).
2013: Sensory processing are first included among international diagnostic criteria of autism.
The 2013 revision of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5)12 includes
sensory symptoms, specifically “hyper-or hypo-reactivity to sensory input or unusual interest in
sensory aspects of environment”, among the diagnostic criteria of autism.
References
1. Kanner, L. Autistic disturbances of affective contact. Nerv. Child 2, 217–250 (1943).
2. Bergman, P. & Escalona, S. K. Unusual Sensitivities in Very Young Children. Psychoanal.
Study Child 3, 333–352 (1947).
3. Eveloff, H. H. The Autistic Child. Arch. Gen. Psychiatry 3, 66–81 (1960).
4. Hermelin, B. & O’connor, N. Psychological experiments with autistic children. (1970).
5. Wing, L. & Wing, J. K. Early childhood autism: Clinical, educational, and social aspects.
(Pergamon, 1967).
6. Association, A. P. Diagnostic and Statistical Manual (DSM-III). (1980).
7. Richer, J. in Autism 47–61 (Springer US, 1978). doi:10.1007/978-1-4684-0787-7_3
8. Autism: A reappraisal of concepts and treatment. (Plenum Press, 1978).
9. Shah, A. & Frith, U. An islet of ability in autistic children: a research note. J. Child Psychol.
Psychiatry. 24, 613–620 (1983).
10. Rutter, M. Cognitive Deficits in the pathogenesis of autism. J. Child Psychol. Psychiatry 24,
513–531 (1983).
11. Frith, U. & Baron-Cohen, S. Perception in Autistic Children. Handbook of Autism and Disorders
of Atypical Development. 85–102 (1987). doi:10.1111/j.1469-7610.1987.tb00658.x
12. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, 5th
Edition (DSM-5). (American Psychiatric Association, 2013).