Invited Symposium 1 - Cognitive neuroscience in infants

Sunday, March 8, 2026, 10:00 am – 12:00 pm PDT, Salon ABCD

Chair: Nick Turk-Browne¹; ¹Yale University
Presenters: Rhodri Cusack, Heather L. Kosakowski, Cameron Ellis, Ghislaine Dehaene-Lambertz

How do infants experience and understand the world around them? Much of what we know about early perceptual and cognitive development has come from behavioral studies and from infant-friendly neuroscientific methods like EEG, MEG, and sleeping fMRI. The past decade has seen a rising wave of fMRI studies in awake infants inspired by the progress and advantages of task-based fMRI in older children and adults. These infant studies provide unique opportunities to link stimuli and behaviors to functional activity throughout the infant brain, including subcortical, medial, and ventral structures, to obtain spatially precise localization and patterns of activity, and to adopt advanced multivariate and model-based analysis approaches. This symposium includes talks on the latest research by four leaders in this emerging field: Rhodri Cusack will present work on the computational development of the visual system. Heather Kosakowski will present evidence for early selectivity in the face network. Cameron Ellis will share an ongoing dense longitudinal case study. Ghislaine Dehaene-Lambertz will discuss the availability of a prosodic hierarchy in the pre-syntactic language system. These presentations will advance our theoretical understanding of the developmental foundations of cognitive neuroscience and reveal the remarkable possibilities of awake infant fMRI.

Presentations

The origin of rich visual categories in early infancy

Rhodri Cusack¹; ¹Trinity College Dublin

The development of object vision in infancy is shaped by both experience and intrinsic brain structure. To probe this developmental process, we acquired awake fMRI data in 2-month-old (n=112) and 9-month-old (n=52) infants. We presented a rich stimulus set of 36 images and used representational similarity analysis to characterise the visual features represented in the brain. Early visual cortex was dominated by perceptual representations, whereas in ventral occipitotemporal cortex, categorical visual representations were already present by 2 months of age and were further strengthened and refined by 9 months. To further unpack the continuum from perceptual to categorical features, we used the layers of a deep neural network (DNN) model and found a hierarchy of visual feature complexity along the ventral visual stream even at 2 months. These representations were best captured by DNNs trained on blurred stimuli, suggesting that infants represent broader spatial scales more strongly than standard DNNs. Baseline models—untrained DNNs or DNNs trained on retinal waves—were less similar to infants, suggesting tuning is specific to the visual features that discriminate visual categories. To characterise how intrinsic brain structure contributes to these early rich representations, we have examined the long-range structural connectivity of these ventral visual regions in neonates using diffusion MRI and tractography. In sum, our results suggest an interplay between intrinsic brain structure and early experience, and demonstrate how awake fMRI combined with computational modelling can provide a rich characterisation of early visual development.

Face selectivity in the human infant brain

Heather L. Kosakowski¹; ¹University of Southern California

Philosophers and psychologists have long debated the relative roles of built-in structure versus learning in the developing human mind. It is only recently that whole-brain measurements from awake infants have become available to inform these debates. In adult brains, face perception is supported by the fusiform face area (FFA) while social interaction perception is supported by regions in the superior temporal sulcus (STS) and abstract representations that support social inferences are computed in the medial prefrontal cortex (MPFC). Despite having different functions, FFA, STS, and MPFC are face selective. Do these regions develop face-selective responses sequentially, from low-level regions close to sensory input first to high-level abstract regions farthest from sensory input last? To investigate the sequence of functional development in the cortex, we used fMRI to measure whole-brain responses from awake infants while they watched videos of faces, bodies, toys, and landscapes. We increased data quality by (1) developing MR-safe infant headphones, (2) designing a new infant coil that accommodated headphones and increased SNR, and (3) optimizing a BOLD acquisition sequence to image infants’ brains. These data showed that infants, like adults, have face-selective responses in FFA, STS, and MPFC. Further, face selectivity is present in even the youngest infants, those that were 2-5 months, suggesting that cortical function develops in parallel. In the last part of my talk, I will briefly suggest a potential neural mechanism that could support parallel cortical development.

Continuity and discontinuity in infant brain development: Dense longitudinal awake fMRI in the first year of life

Cameron Ellis¹; ¹Stanford University

The first year of life is marked by rapid cognitive change. Whether this change is smooth and continuous or sudden and discontinuous remains a central question in developmental psychology. Some historical and modern perspectives align with continuity (e.g., behaviorism, core knowledge), and other perspectives align with discontinuity (e.g., Freudian psychosexual stages, Theory Theory). A comparable debate happened for whole body growth: although height appears to increase continuously when measured every 3 months, only when measurements were taken every two weeks was it learned that height changes in 'spurts'. Thus, to understand the rate of infant cognitive development, we collected anatomical and task-based fMRI data from the same infant every two weeks from 2 to 12 months. The tasks included movie-watching and event-related image presentation of diverse categories. With this uniquely frequent and large dataset — averaging 59.7 minutes of task data after motion censoring per month — we can precisely measure the change in brain anatomy and function. Preliminary analyses show a continuous trajectory of brain volume growth: a logarithmic fit to age explains more than 95% of variance in whole-brain volume in this individual, providing no evidence for growth spurts at the sampled cadence. The implications for the prioritization of brain development are discussed. Ongoing analyses measure changes in the functional response over development. These results demonstrate the value of precise and dense longitudinal neuroimaging to address debates in cognitive development.

From syllables to sentences: Tracing the prosodic hierarchy in the infant brain

Ghislaine Dehaene-Lambertz¹; ¹Centre National de la Recherche Scientifique

Because prosody mirrors the hierarchical organization of syntax, it may provide infants with an early scaffold to discover grammatical structure before they know any words. We investigated this hypothesis in 15 healthy 3.5-month-old infants using fMRI. Participants listened to two successive sequences of eight syllables, either repeated or not, and organized into different levels of the prosodic hierarchy: isolated syllables, bisyllabic “words,” intonational phrases, or full sentences. Preliminary analyses reveal an increase in functional connectivity with suprasylvian regions as prosodic units become larger, suggesting a progressive recruitment of higher-order integrative areas. We also examined BOLD response variability across subjects and cortical regions to assess the robustness and maturation of these hierarchical networks in early infancy. Beyond inter-subject and inter-regional differences, we observed notable temporal fluctuations within the same voxels, which may explain why classical HRF-based analyses often show limited sensitivity in the infant brain. Together, these findings suggest that the infant brain already organizes speech according to a hierarchical prosodic structure, which may pave the way for later syntactic learning.