Communication, Computation, and Cognitive Processes (СССР3)
Cognitive control includes two basic aspects: maintenance of task-specific processes (including attention), and non-specific control of the motor threshold. In line with this distinction, performance errors may be explained by two alternative mechanisms: fast errors are believed to result from inappropriate action impulses that were allowed due to a failure in keeping the motor threshold sufficiently high, while slow errors may be caused by failures in specific task-relevant information processing, leading to uncertainty. Correct responses are not uniform as well – they may be preceded by subthreshold attempts to execute an erroneous response, thus evidencing uncertainty and resulting in increased reaction time (RT). We expected that RT might be a valid proxy that will allow distinguishing trials with high and low level of uncertainty. We hypothesized that on fast-RT trials an internal outcome detection would occur following response commission, while on slow-RT trials the low level of certainty would hinder the internal detection, and only an external feedback signal would lead to processing the trial outcome. EEG data were collected from 50 healthy right-handed adults. We used an auditory condensation tasks involving a two-alternative choice. For each participant, responses were classified into fast and slow relative to individual median RTs. Visual feedback was given 500 ms after the response. We found a much greater error-related theta power increase at fronto-central sites on fast-RT trials compared with slow-RT trials. Only on slow-RT correct trials, positive feedback induced a feedback-related increase in prefrontal beta power. Late parietal suppression of alpha oscillations was also prominent on slow-RT erroneous trials. Our findings are compatible with the view that fast responses involve little internal uncertainty, thus allowing internal error detection. Slow responses occur in conditions of higher internal uncertainty, making the external feedback unpredictable and thus informative in terms of reinforcement learning.