Some of the earliest signs of autism appear not in how babies speak or move, but in how they look at the world. Research using eye-tracking technology has revealed that differences in visual attention can emerge within the first months of life, well before the social and language milestones that parents and pediatricians typically watch for. A landmark study found that infants later diagnosed with autism showed a measurable decline in eye contact starting around two months of age, even though their eye-looking began at the same level as other infants.1PubMed Central. Attention to eyes is present but in decline in 2-6-month-old infants later diagnosed with autism That finding upended the assumption that reduced eye contact in autism is present from birth, and it opened a window into a whole constellation of visual differences that researchers have since mapped in detail.
Eye Looking Starts Normally, Then Fades
The most striking discovery in this area comes from a prospective study that tracked infants from two to six months of age. Babies who were later diagnosed with autism looked at eyes at typical rates during their first weeks of life. Between two and six months, though, their fixation on eyes gradually declined, a pattern not seen in infants who developed normally. This trajectory is significant because it suggests the difference is not a matter of being “born without” interest in faces. Something changes during a critical developmental window.1PubMed Central. Attention to eyes is present but in decline in 2-6-month-old infants later diagnosed with autism
The subcortical visual system may help explain why. In early life, before the brain’s cortical areas are fully wired, a fast pathway running through deeper brain structures supports a newborn’s basic attention to faces and movement. This rapid pathway is thought to give all babies an initial orientation toward social stimuli like eyes and facial expressions.2Frontiers in Integrative Neuroscience. The “Primitive Brain Dysfunction” Theory of Autism: The Superior Colliculus Role As cortical circuits mature and take over more of the visual processing load, infants who go on to develop autism may not build the same reinforcing loops that keep typically developing babies locked onto social information. The initial “starter kit” for social looking works, but the system that is supposed to build on it does not ramp up in the same way.
Sticky Attention and Trouble Shifting Gaze
Another visual pattern that shows up in the first year is sometimes called “sticky attention,” a delay in disengaging from one thing to look at something else. Typically developing babies get faster and more flexible at shifting their gaze as they grow. In studies that tracked high-risk infants (those with an older sibling already diagnosed with autism), the ones who later received an autism diagnosis showed notably prolonged latency to disengage their visual attention by 12 months, and this pattern continued at 15, 18, and 24 months.3PubMed. Prospective examination of visual attention during play in infants at high-risk for autism spectrum disorder: a longitudinal study from 6 to 36 months of age
A separate study found a similar result, observing that by 14 months, infants later diagnosed with autism showed longer latencies to disengage compared to both low-risk controls and other high-risk infants who did not develop autism. Between 7 and 14 months, these infants did not show the typical gains in the speed and flexibility of visual orienting that other babies made.4PubMed Central. Disengagement of Visual Attention in Infancy is Associated with Emerging Autism in Toddlerhood Interestingly, the same study noted that some infants who had other developmental concerns but did not meet criteria for autism also showed a similar lag in disengagement, suggesting this visual pattern is not exclusive to autism but is a broader marker of atypical development.
For caregivers, this can look like a baby who stares at a spinning fan or a pattern on the carpet for long stretches without responding when called. It is not that the baby cannot see other things; it is that the visual system seems to have difficulty releasing attention from what it is currently locked onto.
A Preference for Geometric Patterns Over Faces
One of the more unusual findings in this field involves what researchers call “geometric preference.” When shown a screen split between social content (people interacting, children playing) and abstract geometric patterns (spinning shapes, moving grids), a subset of toddlers with autism fixated on the geometric side for more than 69% of the viewing time. Using that threshold, the pattern had very high specificity at 98%, meaning almost no typically developing children showed it, though its sensitivity was only about 21%, meaning most children with autism did not show it either.5PubMed Central. Eye Tracking Reveals Abnormal Visual Preference for Geometric Images as an Early Biomarker of an Autism Spectrum Disorder Subtype Associated With Increased Symptom Severity
The children who did show this strong geometric preference tended to have worse cognitive, language, and social skills compared to children with autism who preferred the social images. This has been replicated: toddlers with autism as a group spent more time viewing geometric images than any other diagnostic group, and greater time on geometric images correlated with higher symptom severity.6PubMed Central. The geometric preference subtype in ASD: identifying a consistent, early-emerging phenomenon through eye tracking The pattern seems to identify a particular subtype within autism rather than being a universal marker. If your toddler is captivated by spinning objects and repeating visual patterns while showing little interest in faces on a screen, that is worth flagging with a pediatrician, though most children with autism will not show such an extreme preference.
Lateral Glances, Downward Gaze, and Other Atypical Visual Behaviors
Parents and clinicians sometimes notice that young children with autism use their vision in physically unusual ways. One well-documented behavior is the “lateral glance,” where a child turns their head and looks at objects out of the corner of their eye rather than straight on. A study of children aged roughly 3 to 6 years found that this was the most frequent atypical visual behavior in the autism group and was more common than in typically developing children of similar age. The researchers proposed that lateral vision filters out fine detail and enhances the perception of movement, so the behavior might be the child’s way of managing visual input that feels overwhelming.7PubMed. Lateral glances toward moving stimuli among young children with autism: Early regulation of locally oriented perception?
In naturalistic settings, children with autism also tend to direct their gaze slightly downward compared to typically developing peers, and they make more extensive use of their lateral (side) field of view.8PLoS ONE. Investigating Gaze of Children with ASD in Naturalistic Settings Brain-imaging research has found that these unusual fixation patterns correspond to differences in how the visual cortex maps peripheral space, suggesting the behavior is tied to underlying neural organization rather than being purely a habit or preference.9PubMed Central. Atypical cortical representation of peripheral visual space in children with an autism spectrum disorder
A case report of a 15-month-old with autism illustrated another dimension of this: the infant’s visual fixation during video scenes was driven by the physical properties of the stimuli, like movement and contrast, rather than by their social meaning. When watching point-light animations of human action, the infant attended to the motion itself, not to the fact that it represented a person doing something.10PubMed. Altered face scanning and impaired recognition of biological motion in a 15-month-old infant with autism This captures a core theme across the research: it is not that visual abilities are impaired in autism. In fact, there is evidence of enhanced perceptual abilities in some domains, such as better detection of visual targets when they appear near the center of a display.11SURFACE at Syracuse University. Visual Processing Across Space and Time in Children with Autism Spectrum Disorder The difference is in what the visual system prioritizes.
Pupillary and Autonomic Responses to Faces
The pupil of the eye does not just respond to light. It also dilates in response to emotional arousal, cognitive effort, and social engagement. Researchers have used pupil size as a window into how babies with autism risk process social information differently. In one study, 9-month-old infants at high risk for autism (because of an older affected sibling) showed larger pupil dilation when viewing emotional faces compared to low-risk infants. Greater arousal in these high-risk babies was linked to worse social-communicative functioning at 18 months, suggesting that heightened autonomic reactivity to faces might be an early mechanism behind the social difficulties that emerge later.12PubMed Central. Greater Pupil Size in Response to Emotional Faces as an Early Marker of Social-Communicative Difficulties in Infants at High Risk for Autism
Children with autism also process the social content of facial expressions differently. A study comparing children with autism and typically developing controls found that both groups showed similar overall pupillary responses to faces. But the groups diverged when it came to sensitivity to gaze direction on happy faces: typically developing children showed increased pupil dilation when a happy face looked directly at them versus looking away, while children with autism did not distinguish between the two. This difference was specific to happy faces with direct gaze and did not appear for negative or neutral expressions.13PubMed Central. Abnormal social reward processing in autism as indexed by pupillary responses to happy faces The implication is that the social reward value of a smile directed at you, the feeling of “that person is happy to see me,” may not register the same way in the autonomic nervous system of children with autism.
Beyond social stimuli, the basic pupillary light reflex itself may differ. Infants at high risk for autism showed a stronger constriction of the pupil in response to light flashes, with a mean relative constriction of about 48% compared to roughly 40% in low-risk infants, and the reflex kicked in faster.14PubMed Central. Hypersensitive pupillary light reflex in infants at risk for autism This hypersensitive reflex points to broader differences in sensory processing that go beyond social cognition.
How the Brain Responds to Faces
When babies look at faces, their brains produce measurable electrical signals that researchers can detect using electrodes on the scalp. Two of these signals, known by their technical labels N290 and P400, are especially relevant to face processing in infancy. A meta-analysis of studies comparing infants at elevated likelihood for autism with average-likelihood infants found that those who later received an autism diagnosis showed a slower N290 response and a smaller P400 response. These patterns point to atypical face processing that is detectable in infancy, potentially serving as early neural markers.15Research in Autism. Face processing in infants at elevated likelihood for autism: A systematic review and meta-analysis
A study of 12-month-old infants looked at brain responses when babies were shown images of their mother versus a stranger. Across all groups, the average brain response difference between mother and stranger was similar, meaning infants at high risk for autism did recognize their mother’s face at a neural level. But in the subgroup that later received an autism diagnosis, the size of this brain response to mother versus stranger was tied to language ability at 12 months and to social symptoms at 18 months. A stronger differentiation between mother and stranger in the brain predicted better language and milder social difficulties.16PubMed Central. Neural correlates of face processing associated with development of social communication in 12-month infants with familial risk of autism spectrum disorder This is an important nuance: the infant brain in autism is not blind to faces. It processes them, but the way it processes them seems to have different downstream consequences for social and language development.
When Eyes and Ears Do Not Line Up
Visual development does not happen in isolation. Babies learn partly by connecting what they see with what they hear, and this audiovisual integration appears to work differently in infants at risk for autism. Research has found that difficulty detecting the correspondence between a face’s lip movements and the sounds being produced may contribute to both the language delays and the social challenges commonly seen in autism. If a baby’s brain is not efficiently linking the sight of a moving mouth with the sound of speech, the foundation for language learning gets shakier from early on.17PLoS ONE. Atypical Audiovisual Speech Integration in Infants at Risk for Autism The inability to detect the match between facial expressions and vocal tone could also play a role in the social-emotional reading difficulties that become apparent later. This means some of what looks like a “visual” difference is really an integration problem across the senses.
Eye Tracking as a Screening Tool
Much of this research depends on eye-tracking technology, which precisely measures where a baby looks, for how long, and how quickly their gaze shifts. That same technology is being developed as a practical screening tool. A systematic review and meta-analysis found that machine learning algorithms trained on gaze data achieved sensitivity between 78% and 89% and specificity between 74% and 86% for distinguishing children with autism. The entire assessment can be completed in under ten minutes, making it feasible for large-scale screening in pediatric settings.18PubMed Central. Eye-Tracking as a Screening Tool in the Early Diagnosis of Autism Spectrum Disorder: A Systematic Review and Meta-Analysis
A 2024 study tested this in primary care clinics with children averaging about two and a half years old. When a composite biomarker from the eye tracker was used alone, sensitivity was about 78% and specificity about 77%. When the eye-tracking results were combined with the pediatrician’s clinical judgment and diagnostic certainty, accuracy jumped: sensitivity reached about 91% and specificity about 87%.19JAMA Network Open. Eye-Tracking Biomarkers and Autism Diagnosis in Primary Care This combination approach is promising because it does not ask the technology to replace clinical expertise. Instead, it gives the clinician an objective data layer to complement their own observations.
These tools are still emerging, and no eye-tracking test has been approved as a standalone diagnostic. But they could eventually fill a real gap. The average age of autism diagnosis in many countries is still around four or five, well past the window when early intervention is most effective. A brief, objective screening that can flag risk in infancy or toddlerhood would be a meaningful advance.
Cultural Differences in Visual Attention Patterns
One complication that gets less attention is cultural variation in gaze. A cross-cultural study comparing participants in Hong Kong and the United States found that both autistic and non-autistic individuals from Hong Kong spent more time looking at body regions and contextual background information and less time looking at faces compared to their American counterparts. The autism-related differences in gaze were most apparent only in more complex stimuli with ambiguous facial expressions.20PubMed Central. A cross-cultural study of visual attention in autism spectrum disorder This matters because many of the gaze-based markers described in the research were developed and validated primarily in Western samples. Reduced face-looking in a child from a cultural background where direct face-gazing is less common could be misinterpreted if the norms are based entirely on American or European data. As eye-tracking screening moves toward clinical use, establishing culturally appropriate baselines will be essential.
Targeting Visual Engagement Through Early Intervention
Identifying visual differences early matters because early intervention can build on whatever social attention a child does have. Joint attention, the ability to share focus on an object or event with another person by following their gaze or pointing, is one of the core social skills affected in autism. Interventions targeting joint attention in toddlers have shown real effects. A randomized controlled trial of a caregiver-mediated intervention found that toddlers in the treatment group made significant improvements in their responsiveness to joint attention bids and in the diversity of their play acts, with medium to large effect sizes, and these gains held up a year after the intervention ended.21PubMed Central. Randomized controlled caregiver mediated joint engagement intervention for toddlers with autism
A separate trial confirmed that a joint engagement intervention could enhance coordinated joint engagement, with improvements observed at a three-month follow-up.22PubMed. Efficacy of caregiver-mediated joint engagement intervention for young children with autism spectrum disorders Not every approach works equally well, though. A review of developmentally based interventions found that two out of three programs studied did not produce treatment effects on social communication or language, while one specific joint attention intervention did improve both the child’s ability to respond to adult-initiated joint attention and their receptive language.23PubMed Central. Methods to improve joint attention in young children with autism: a review The takeaway for parents is that caregiver-led interventions, where you learn strategies from a therapist and practice them at home, have the best evidence for this age group, but the specific program matters. Asking a provider about interventions with randomized trial evidence behind them is reasonable.
The broader promise of understanding visual development in autism is that it connects the very first months of life to practical action. A baby who is beginning to look away from eyes at three months is not destined for any particular outcome. But catching that shift early and responding with the right kind of enriched social interaction can nudge the developmental trajectory in a more connected direction, long before anyone would have suspected a diagnosis based on behavior alone.