Neural Sensitivity without Behavioural Evidence: Frequency-Based Statistical Learning in Fast Periodic Visual Stimulation

Humans are sensitive to statistical cues in their perceptual input, including associations between stimuli (transitional probability) and their absolute frequency (unimodal information). To investigate the mechanisms behind the acquisition of this latter cue, both at an implicit and explicit level, we combined Fast Periodic Visual Stimulation (FPVS) with EEG recordings and two-alternative forced choice (2AFC) tasks. Skilled readers viewed rapid sequences (6 Hz) of words and pseudofonts, in which low-frequency items appeared regularly at a slower periodic rate (1.2 Hz) among high-frequency items. EEG frequency-domain analyses revealed robust neural responses at the oddball frequency, indicating sensitivity to frequency-based regularities and implicit grouping. Despite these clear signs of neural learning, no discrimination between low- and high-frequency items emerged in our participants’ behaviour. This pattern of results was confirmed in a replication experiment. Together, these findings suggest that frequency-based regularities are robustly encoded at the neural level, even when behavioural expression remains limited. They also suggest that neural entrainment and behavioural learning might track different aspects of the complex neural and cognitive machinery that makes us sensitive to the statistics of our environment.