Untangling brain and behavioural measures of visual statistical learning: A longitudinal study in infancy.

Statistical learning in infancy is typically studied using post-exposure behavioural paradigms, which can yield variable outcomes and pose interpretative challenges. Electrophysiological measures, recorded online during learning, offer a complementary approach. In this longitudinal study, we examined visual statistical learning in infants across three timepoints: 3 months (T1, n = 30), 6 months (T2, n = 27), and 9 months (T3, n = 23). At each session, infants participated in an EEG frequency-tagging paradigm involving 20-second sequences of shapes presented at 6 Hz, with deterministic doublets occurring at 3 Hz. Following familiarization, infants viewed familiar and novel doublets while looking times were recorded. Neural responses indicated sensitivity to the doublet structure in occipital regions as early as T1, with the learning index (entrainment at 3 Hz relative to 6 Hz) significantly increasing by T3. Behavioural looking times did not differ across ages, and a link between neural entrainment and behaviour was observed only in the older age group. These findings suggest that the infant brain can detect visual statistical regularities from early in life, with neural markers of learning becoming more robust and predictive of behaviour by 9 months. EEG provides a sensitive and continuous index of learning, capturing effects that may not always be evident in behaviour.