Postersession 3
Poster #: 57
Topic: Error signals
Friday, Sep 11, 2015
15:30-17:00
1st floor

Implicit learning of predictable sound sequence modulates mismatch responses at different levels of the auditory hierarchy

Françoise Lecaignard, Olivier Bertrand, Jérémie Mattout, & Anne Caclin

DYCOG Team, CRNL, Bron, France
francoise.lecaignard@inserm.fr

Within the framework of predictive coding, deviance processing is part of an inference process where prediction errors (the mismatch between incoming sensations and predictions established through experience) are minimized. In this view, the Mismatch Negativity (MMN) is a measure of prediction error. This provides new hypothesis on MMN modulation by experimental factors, whose testing could refine our understanding of the cognitive processes underlying mismatch responses. In particular, the MMN should decrease as the occurrence of a deviant stimulus becomes more predictable.

We conducted a passive oddball EEG study and manipulated the predictability of deviance occurrence by means of stimulus sequences with different temporal structures. Importantly, our design departs from previous studies that compared violations of different time-scale regularities.

Evoked responses revealed (1) a modulation of the MMN amplitude by deviance predictability as expected, and (2) an earlier deviance response (around 50 ms) that was also reduced with sound predictability. This twofold effect of predictability supports the view that both deviance responses reflect prediction errors and belief updates computed at different levels of the auditory hierarchy. Furthermore, as none of the participants had been aware of the sound sequence structure, we conclude that it could be encoded through an implicit learning process implemented within the hierarchy. We propose that large time-scale regularities could induce high-level predictions that modulate both the content and the precision of lower-level ones. Our findings hence substantiate predictive coding and provide formal constraints for emerging generative neurocognitive models of (mismatch) evoked responses in the brain.