The infant brain rapidly entrains to visual statistical regularities during stimulus exposure.

Statistical learning (SL) has been studied quite extensively in infancy. Still, most evidence relies on post-exposure behavioral tasks whose directionality (familiarity vs. novelty effects) may not be straightforward to predict nor to interpret. In addition, these tasks do not tell anything about the online learning dynamics and may be influenced by memory effects. In this work, we investigated online SL mechanisms by tracking neural entrainment to visual regularities in a group of 4- to 6-month-old infants exposed to a stream of shapes presented at 6 Hz. Shapes were either organized in doublets or presented randomly. Results revealed that entrainment at the doublet frequency of 3 Hz and harmonics varied across conditions and trials. Infants showed greater occipital entrainment to the doublet frequency in the deterministic condition than in the random one, especially over the first trials of exposure. This suggests that the brain can detect visual regularities from early infancy. Further, this sensitivity emerged early over the exposure period and did not show a learning curve when the evolution of the doublet-level signal-to-noise ratio (SNR) was assessed in relation with the base-level SNR over time. Hence, considering its time course and the brain regions involved, neural entrainment at the doublet frequency seems to primarily reflect a bottom-up detection mechanism rather than the full expression of learning. These findings are crucial to better understand how infants extract regularities during stimulus exposure and what neural entrainment can reveal in a visual SL task.