Babies’ Gaze Between Eyes and Mouth Shifts With Speech Rhythm, Autism Family History and Language

Where do babies look when they watch a face? For decades, developmental scientists have debated whether infants at elevated genetic risk for autism scan faces differently from their peers, and the literature has produced a frustrating tangle of contradictory findings. A new longitudinal study, published in the Journal of Autism and Developmental Disorders, suggests that part of the confusion may come from two overlooked sources of variability: the speed at which a mouth moves, and the presence of attention deficit hyperactivity disorder (ADHD) in the family. By following 160 infants from five months to three years of age, the research team, led by Mary Agyapong of King’s College London together with colleagues at Birkbeck, University of London, has produced one of the most detailed pictures yet of how early face-scanning behaviour unfolds in children with and without a family history of autism or ADHD.

The study drew on the STAARS cohort, the Studying Autism and ADHD Risk in Siblings project, recruited through the Birkbeck Centre for Brain and Cognitive Development. Families were enrolled if they had a first-degree relative with a clinical diagnosis of autism, a clinical or probable research diagnosis of ADHD, or if they had typically developing older children and no first-degree relatives with either condition. The final sample comprised 79 infants with an elevated likelihood of autism, 31 with an elevated likelihood of ADHD, 21 with elevated likelihood of both conditions, and 29 with typical likelihood for neither. Each child visited the lab at five, ten and fourteen months for eye-tracking sessions, and returned for developmental and language assessments at 36 months, allowing the researchers to connect infant gaze patterns with later language outcomes.

The experimental task was elegantly simple. At each visit, infants watched pairs of ten-second silent videos of a woman’s face presented side by side on a screen. In each pair, the mouth flashed open and closed at one of three temporal frequencies: a slow one hertz, a mid-range six hertz, or a fast ten to fifteen hertz, depending on the age of the infant. The mid-frequency condition is scientifically crucial because fluent speech carries most of its energy in the theta band, roughly four to eight hertz, the rhythm at which syllables are produced. Neural oscillations in this theta range are thought to be central to how the brain segments and processes speech, and previous work has shown that even newborns show theta-frequency neural activity related to speech perception. The researchers wanted to know whether babies’ eyes are drawn to mouths moving at this language-relevant rhythm.

Gaze was recorded with a Tobii TX-300 infrared eye tracker sampling at 120 hertz while infants sat on a parent’s lap. The team computed an eye-mouth looking index, or EMI, defined as the proportion of face-directed looking time spent on the eyes relative to the eyes plus the mouth combined. An index above 0.5 signals a preference for the eyes; below 0.5, a preference for the mouth. Trials with fewer than 25 percent valid gaze samples were discarded, and the longitudinal data were analysed with generalised linear mixed models, a statistical framework well suited to the missing data and bounded proportions that characterise infant eye-tracking studies.

The developmental pattern was clear. Across the whole sample, the eye-mouth index declined significantly between five and fourteen months, meaning that as babies grew older they shifted their attention progressively away from the eyes and toward the mouth. This replicates a well-established trajectory: typically developing infants look predominantly at the eyes before about five months of age and then increasingly favour the mouth through the first year, precisely the period during which they begin to comprehend and produce their first words. Interestingly, the researchers did not observe the re-orienting back toward the eyes around twelve months that some earlier studies reported, and they note that other work has found mouth attention persisting in both monolingual and bilingual children up to age five, supporting the idea that watching a moving mouth is adaptive for language learning.

Temporal frequency also mattered. Infants looked more at the eyes relative to the mouth when watching the slow one hertz faces than when watching the six hertz, speech-like faces. In other words, mouths moving at the rhythm of syllabic speech captured more attention than slowly moving mouths, providing partial behavioural support for the idea that human infants, like rhesus monkeys in earlier work by Asif Ghazanfar and colleagues, are perceptually tuned to theta-frequency facial movement. The effect was not exclusive, however: there was no significant difference in looking between the mid-frequency and high-frequency conditions, suggesting that the tuning is not a narrow bandpass filter but perhaps reflects a broader preference shaped by arousal and stimulus salience, which the authors flag as a target for future research measuring physiological arousal.

The most striking findings concerned family history. Infants with an elevated likelihood of autism showed a significantly larger eye-mouth index than infants with typical likelihood, meaning they looked relatively more at the eyes and less at the mouth across infancy. The authors interpret this as a possible developmental delay in the usual shift from eyes to mouth, since prolonged eye preference is normally a feature of younger infants. Crucially, this effect was attenuated when there was also a family history of ADHD: the interaction between autism and ADHD familial likelihood was significant, and at fourteen months the EL-autism group showed a larger index than the EL-autism-plus-ADHD group. A family history of ADHD on its own was also associated with more eye relative to mouth looking at fourteen months, a novel result in a population that has been almost entirely neglected by the eye-tracking literature. The authors argue that future family-history studies of autism should systematically map ADHD in the family, because its presence may mask or modify the effects researchers are trying to detect, potentially explaining some of the heterogeneity in previous reports.

Language added a further layer of nuance. At fourteen months, infants who had already said their first words, as measured by an item on the Mullen Scales of Early Learning, showed a smaller concurrent eye-mouth index, that is, more mouth looking, than infants who had not yet produced words, with a medium-to-large effect. This fits the experience-dependent view of face attention: once babies begin talking, watching mouth movements may help them refine and imitate the shapes of words they hear from speakers. Yet the long-term picture was more surprising. When the researchers controlled for fourteen-month word status, higher receptive language scores at 36 months were associated with greater looking to the eyes, not the mouth, across infancy. This ran counter to the authors’ prediction and to some previous findings linking mouth preference to better later language. One interpretation, drawing on recent work by Viktorsson and colleagues, is that looking at the eyes supports gaze following and joint attention, which are themselves foundational for word learning. The takeaway, the authors suggest, is that both eyes and mouth matter: mouth looking serves emerging language skills in the moment, while eye engagement relates to the interpersonal foundations of language that come to fruition in toddlerhood.

Not every hypothesis was confirmed. There were no significant differences between likelihood groups in parent-reported comprehension or production vocabulary at fourteen months on the MacArthur-Bates Communicative Development Inventory, although girls comprehended significantly more words than boys, consistent with the broader literature on early sex differences in language. Nor did the frequency effect interact with word status, and neither expressive nor receptive language at 36 months related to infant EMI until first-word status was taken into account. The authors are candid about the limitations of their design. The silent videos isolated mouth movement in a way naturalistic speech does not, arousal was not measured, and the cohort, though prospective and well characterised, was underpowered to examine actual autism or ADHD diagnoses as outcomes. Causal conclusions are therefore off the table.

Even with those caveats, the study delivers a message that resonates well beyond specialist circles: the way a baby looks at a face is not a single fixed trait but a moving target shaped by age, by the rhythm of the face itself, by the constellation of conditions in the family, and by the child’s own blossoming language. For researchers hunting early markers of neurodevelopmental conditions, the implication is that heterogeneous findings may partly reflect unmeasured heterogeneity in the samples, particularly the under-appreciated role of familial ADHD. For parents, the study offers no cause for alarm but a fascinating window into the infant mind, revealing that from the earliest months, babies are already calibrating their gaze to the rhythms of human communication, sampling eyes and mouths in a dynamic balance that foreshadows the conversational partners they will one day become.

Subject of Research: Infant eye-tracking study of visual attention to eyes versus mouths in relation to familial autism and ADHD likelihood and language development

Article Title: Infant Attention to Eyes Versus Mouth Varies With Mouth Movement Speed, Family History of Autism and Language Development

Article References: Agyapong, M., Mason, L., Begum Ali, J., Johnson, M. H., Charman, T., Jones, E. J. H., the STAARS Team, Bazelmans, T., Dafner, L., Ersoy, M., Gliga, T., Goodwin, A., Haartsen, R., Halkola, H., Hendry, A., Holman, R., Kalwarowsky, S., Kolesnik, A., Lloyd-Fox, S., … Taylor, C. (2026). Infant Attention to Eyes Versus Mouth Varies With Mouth Movement Speed, Family History of Autism and Language Development. Journal of Autism and Developmental Disorders. https://doi.org/10.1007/s10803-026-07516-9

Image Credits: AI Generated

DOI: 10.1007/s10803-026-07516-9

Keywords: autism, ADHD, infancy, eye-tracking, language development, visual attention, face scanning, theta frequency, family history, longitudinal study, social attention, developmental psychology