Modeling congruency sequence effects with the revised diffusion model for conflict tasks.

Abstract

The congruency sequence effect (CSE), or Gratton effect, describes diminished congruency effects (i.e., faster responses for <<<<< vs. <<><< stimuli in a flanker task) on trials following an incongruent trial compared to those following a congruent trial. Traditionally, the CSE is regarded as an index of conflict adaptation. Accounts of the CSE have typically emphasized either top-down (cognitive control) or bottom-up (associative) processes. To disentangle top-down and bottom-up contributions to the CSE, we compared performance on versions of Simon and flanker tasks that control for memory and learning effects present in the standard versions of these tasks with the standard tasks. We analyzed the data using a recently developed model that explains conflict effects in terms of attention-shifting dynamics, the revised diffusion model for conflict tasks. Our modeling analyses suggest the CSEs in both Simon tasks are mainly driven by across-trial changes in the way attention is loaded onto distractor information, consistent with top-down control. In contrast, for both flanker tasks, our model-based analysis does not provide clear evidence favoring either top-down or bottom-up contributions to the CSEs overall. Furthermore, we fit the standard version of these tasks while separately analyzing trials based on whether the response was repeated or alternated across successive trials. Model fits suggest a combination of both influences, with different patterns for congruent and incongruent trials. Performance benefits on congruent trials are driven by adjusting how attention is loaded onto distractor information based on the previous congruency types (i.e., top-down influence). However, on incongruent trials, asymmetrical repetition benefits are mainly due to faster memory retrieval (i.e., bottom-up influence) and reactive attentional control. Our findings highlight that a more nuanced and integrated theoretical framework is necessary to capture the interplay between top-down and bottom-up processes in explaining control mechanisms when analyzing different response sequences. (PsycInfo Database Record (c) 2025 APA, all rights reserved).