The detection of cognitive state transitions by stability changes in event-related cortical field potentials

Cognitive tasks are characterized by sequences of stable states, which are thought to re#ect distinct steps of neurocognitive processing. Here, we investigate a stability measure, derived from adaptive multivariate autoregressive (AMVAR) modeling of cortical "eld potentials, as an index for relating changes in large-scale neural activity to changes in cognitive state. We show that this stability measure can be used to decide the optimal window length for AMVAR modeling and to detect state transitions related to external sensory or motor events. By using this approach, we demonstrate clear di!erentiation of GO and NO-GO processes in a macaque monkey performing a visuomotor pattern discrimination task. Moreover, we are able to identify regional di!erences in state transitions, apparently re#ecting regional information processing di!erences. 2001 Elsevier Science B.V. All rights reserved.