Probing cognitive processes through the structure of event-related potentials during learning: an experimental and...

Data reporting correlated changes, due to learning, in the amplitudes and chronometry of several event-related potentials (ERPs) are compared to neural explanations and predictions of the adaptive resonance theory. The ERP components processing negativity (PN), early positive wave (P120), N200, and P300 covary with model processes of attentional priming and top-down expectancy learning, matching of bottom-up input patterns with learned top-down expectations, mismatch-mediated activation of the orienting subsystem, reset by the orienting subsystem of recognition codes in short-term memory, and direct activation of recognition codes via a bottom-up adaptive filter. These model mechanisms enable a recognition code to be learned in a self-stabilizing fashion in response to an input environment of arbitrary complexity. Thus spatiotemporal correlations among several ERPs during learning provide important evidence in support of postulated neural mechanisms for self-stabilizing self-organization of cognitive recognition codes.