Ordin, M. 1, 2 , Polyanskaya, L. 1 , Molinaro, N. 1, 2 & Soto, D. 1, 2
1 BCBL - Basque Centre for Cognition, Brain and Language
2 Ikerbasque - Basque Foundation for Science
Assume that the listeners are exposed to tri-syllabic sequences with high transitional probability so that syllable B follows syllable A and is followed by syllable C. So, the sequence ABC is segmented from continuous stream of concatenated syllables and is learnt as a whole constituent. Also, the acoustic stream contains syllables XY in some units, and YZ in different units. Both XY and YZ sequences are presented often, the sequence XYZ is never presented. We want to find out 1) the neural pathways and substrates that underlie extraction of statistical regularities; 2) whether XYZ sequences (items that are statistically congruent yet not physically presented during encoding) emerge as perceptual units; 3) the neural pathways and substrates underling recognition of old sequences (words), statistically congruent new sequences (phantoms), and statistically incongruent new sequences (e.g., BYX or CZB, nonwords). To address these issues, we set up an fMRI experiment. We analyzed the neural changes during encoding, as habituation to the stream of recurrent statistical items continues. Also, we compared activation of neural substrates and active pathways between presentation of words and random concatenation of the same syllables, interspersed with structured habituation stream at different times. Following the habituation, we set up the recognition test (fast event-related design) presenting a 3-second sequence of either words, phantoms, or nonwords. Participants had to respond whether the sequence is coming from the habituation stream or not, and indicate their confidence. Participants reliably endorsed words, rejected nonwords, and were at chance with phantoms. Endorsed phantoms received high confidence rating, like correctly endorsed words and rejected nonwords. Rejected phantoms received low confidence rating, like incorrectly rejected words and endorsed nonwords. This suggests that endorsing phantoms are processed as correct responses. Comparing activation patterns on endorsed phantoms and words should reveal a novel nondelarative memory mechanism supporting statistical learning