摘要：The congruency sequence effect (CSE) describes the finding that congruency effects in classic probes of selective attention (like the Stroop, Simon, and flanker tasks) are smaller following an incongruent than following a congruent trial. The past two decades have generated a large literature on determinants and boundary conditions for the CSE and similar, congruency-proportion based modulations of congruency effects. A prolonged and heated theoretical discussion has been guided primarily by a historically motivated dichotomy between “top-down control” versus “associative bottom-up” explanations for these effects. In the present article, I attempt to integrate and contextualize the major empirical findings in this field by arguing that CSEs (and related effects) are best understood as reflecting a composite of multiple levels of learning that differ in their level of abstraction. Specifically, learning does not only involve the trial-by-trial encoding, binding, and cued retrieval of specific stimulus-response associations, but also of more abstract trial features, including the spatial and temporal context in which a stimulus occurs, as well as internal states, like the experience of difficulty, and the attentional control settings that were employed in dealing with the stimulus. From this perspective, top-down control and bottom-up priming processes work in concert rather than in opposition. They represent different levels of abstraction in the same learning scheme and they serve a single, common goal: forming memory ensembles that will facilitate fast and appropriate responding to recurring stimuli or events in the environment.
