Distinct outcome processing in deterministic and probabilistic reversal learning

• Published in: Cortex Vol. 141; pp. 224 - 239
• Main Authors: Habiby Alaoui, Selim, Adam-Darqué, Alexandra, Ptak, Radek, Schnider, Armin
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.08.2021
• Subjects: Behavioral extinction; Judgment; Monitoring; Orbitofrontal reality filtering; Outcome processing; Prediction error; Orbitofrontal reality filtering; Behavioral extinction; Outcome processing; Monitoring; Prediction error; Judgment
• DOI: 10.1016/j.cortex.2021.04.008
• ISSN: 0010-9452; 1973-8102

Anticipations that fail to happen are important drivers of behavioral adaptation. Their processing appears to depend on the context. In a deterministic environment, where a stimulus unequivocally predicts the outcome, processing of absent outcomes involves the posterior orbitofrontal cortex (OFC). Failure has been linked to reality confusion with confabulations and disorientation. In a probabilistic environment, absent outcomes appear to be processed by the anterior cingulate cortex (ACC) rather than the OFC. Failure has been associated with poor decision making and schizophrenia. These data suggest different mechanisms depending on the context. Here, healthy human subjects made two formally similar reversal learning tasks, but one with deterministic, the other with probabilistic instructions. Brain activity was monitored using high-density electroencephalography. We found that in the deterministic task, negative outcomes, which unequivocally call for a behavioral switch, induced a distinct frontal potential at 200–300 msec. Computational modeling indicated a strong association of evoked potentials with prediction error, surprise, and behavioral adaptation. In the probabilistic task, where behavioral adaptation follows the cumulated processing of outcomes, negative outcomes evoked potentials that were associated with prediction error and surprise, but had a weak link with subsequent behavior. Outcome processing in the probabilistic task induced stronger activation than the deterministic task of an extended network including the ACC, OFC and striatum at 300–400 msec. In both tasks, negative outcomes were processed differently from positive outcomes at 400–600 msec, possibly reflecting updating of the outcome record. We conclude that the brain disposes of at least two distinct systems processing outcomes with unequivocal or ambiguous behavioral significance. These systems differ along behavioral, clinical, electrophysiological and anatomical dimensions.