Decoding Stimulus-Response Representations and Their Stability Using EEG-Based Multivariate Pattern Analysis.

Goal-directed actions require proper associations between stimuli and response. This has been delineated by cognitive theory, for example, in the theory of event coding framework, which proposes that event files represent such bindings. Yet, how such event file representations are coded on a neurophysiological level is unknown. We close this gap combining temporal electroencephalography (EEG) signal decomposition methods and multivariate pattern analysis (MVPA). We show that undecomposed neurophysiological data is unsuitable to decode event file representations because different aspects of information coded in the neurophysiological signal reveal distinct and partly opposed dynamics in the representational content. This is confirmed by applying MVPA to temporal decomposed EEG data. After intermixed aspects of information in the EEG during response selection have been separated, a reliable examination of the event file's representational content and its temporal stability was possible. We show that representations of stimulus-response bindings are activated and decay in a gradual manner and that event file representations resemble distributed neural activity. Especially representations of stimulus-response bindings, as well as stimulus-related representations, are coded and reveal temporal stability. Purely motor-related representations are not found in neurophysiological signals during event coding.