Implicit Statistical Learning in Fast Periodic Visual Stimulation


Society for the Neurobiology of Language (SNL) Annual Meeting, virtual edition, October 21-24, 2020


October, 2020


Mara De Rosa, Maria Ktori, Yamil Vidal, Roberto Bottini and Davide Crepaldi

Research on the neural underpinnings of linguistic representations has recently received a methodological boost with an approach that capitalises on the principle of neural entrainment by combining Fast Periodic Visual Stimulation (FPVS) and electrophysiological recordings (e.g., Lochy, Van Belle, & Rossion, 2015). In this paradigm, streams of visual stimuli (i.e., base stimuli, e.g., consonant strings) that are presented at a frequency rate F, are interleaved with oddballs (i.e., stimuli from another category, e.g., words), which are periodically inserted at fixed intervals (i.e., every n items), appearing thus at a slower frequency rate (F/n). Neural entrainment is indexed by a response at the base stimulation frequency, while an additional neural response at the pre-defined oddball stimulation frequency reflects the brain's ability to discriminate between the two types of stimulus categories, and is selective to the dimension that differentiates the oddball from the base stimuli. The present study investigated (a) whether such category-selective response can be generated by incidental implicit learning even when base and oddball stimuli are of the same kind, and (b) whether it can be modulated by stimulus familiarity. Skilled readers (N = 30; native Italian speakers) were asked to monitor the color change of a central fixation cross while exposed to sequences of four types of linguistic items with decreasing familiarity: (1) existing Italian words (e.g., ombra, shadow); (2) pronounceable, but not attested letter strings (e.g., barmo, swhoad); (3) unpronounceable consonant strings (e.g., qnlvd), and (4) strings of non–alphabetic characters that share low-level visual features with letters. Crucially, these sequences were made of stimuli belonging to the same category, in which the only distinction between base and oddball items was the frequency of individual tokens within a stream. Stimuli were presented at a frequency rate of 6Hz, with oddball items inserted every fifth item (i.e., 6/5 = 1.2Hz). Since individual base tokens appeared four times more often than their oddball counterparts, a neural entrainment to oddball items would reflect an effect of token frequency, while a response in any of these categories would index which level of stimulus familiarity (i.e., whole-word level, co-occurring letter clusters, letters, letter features) can give rise to such implicit statistical learning. Results revealed a significant response at the oddball frequency and its harmonics in all conditions, suggesting the emergence of two distinct classes of items purely informed by token frequency. Cross-condition comparisons indicated that the effect is independent of stimulus familiarity, and arises across a wide span of stimuli, from non–alphabetic characters that were never experienced to fully fledged, frequent words. The implications of our findings are twofold. On a theoretical level, we observe an online neural index of fast implicit statistical learning, a mechanism that might account for the bootstrapping of linguistic categories. On a methodological level, we show that sensitivity to statistical regularities can contaminate any category-selective response in FPVS-oddball designs calling for future studies to take this phenomenon under serious consideration.