Conditional deviant repetition in the oddball paradigm modulates processing at the level of P3a but not MMN.

The auditory system has an amazing ability to rapidly encode auditory regularities. Evidence comes from the popular oddball paradigm, in which frequent (standard) sounds are occasionally exchanged for rare deviant sounds, which then elicit signs of prediction error based on their unexpectedness (e.g., MMN and P3a). Here, we examine the widely neglected characteristics of deviants being bearers of predictive information themselves; naive participants listened to sound sequences constructed according to a new, modified version of the oddball paradigm including two types of deviants that followed diametrically opposed rules: one deviant sound occurred mostly in pairs (repetition rule), the other deviant sound occurred mostly in isolation (non‐repetition rule). Due to this manipulation, the sound following a first deviant (either the same deviant or a standard) was either predictable or unpredictable based on its conditional probability associated with the preceding deviant sound. Our behavioral results from an active deviant detection task replicate previous findings that deviant repetition rules (based on conditional probability) can be extracted when behaviorally relevant. Our electrophysiological findings obtained in a passive listening setting indicate that conditional probability also translates into differential processing at the P3a level. However, MMN was confined to global deviants and was not sensitive to conditional probability. This suggests that higher‐level processing concerned with stimulus selection and/or evaluation (reflected in P3a) but not lower‐level sensory processing (reflected in MMN) considers rarely encountered rules. Our results indicate that processing at the level of sensory analysis (MMN) is foremost shaped by the frequency of rule encounters (global probability) rather than their precision (conditional probability). Yet, higher‐level processes (P3a) appear to be sensitive to conditional probability even when associated with rarely encountered deviants. Deviants are powerful tools to elicit prediction errors, yet our findings suggest that deviants can be relevant predictors themselves. [ABSTRACT FROM AUTHOR]