Abstract
Theoretical models of action control assume that abstract task-set settings regulate lower-level stimulus/response representations. However, we know little about the degree to which task-set representations can actually explain variability in performance, or about the dynamic interplay between task-set and lower-level representations. Addressing such questions requires time-resolved information about the strength of the different representations on a trial-by-trial basis. Using a cued task-switching paradigm, we show that information about both task sets and lower-level stimulus/response features can be extracted through decoding analyses from the scalp electrophysiological signal (EEG) with high temporal resolution. The representational strength of the different goal-relevant aspects—indexed through decoding accuracy—proceeds from superficial task cues, to stimulus locations, to features/responses. Importantly, these lower-level representations are accompanied by task-set representations that are prominent throughout almost the entire processing cascade. Moreover, trial-by-trial information about representational strength allows a detailed analysis of when and to what degree (a) different representations predict performance and (b) are related to each other. Indeed, the strength of abstract task sets emerges as a remarkably strong and consistent predictor of both within-individual and across-individual variability in performance. Also, task-set strength is related to target representation strength at an early period and to the strength of feature/response representations at a later period, consistent with the notion that task-sets coordinate successive, lower-level representations in a concurrent manner. These results demonstrate a powerful approach towards uncovering the different stages of information processing and their relative importance for performance.