Evidence accumulation modeling of learning, repetition, and control adaptation processes in Stroop and contingency learning
Year: 2026
Type: preprint
Abstract: Sequential effects in the Stroop task arise from multiple adaptive processes, including conflict adaptation and contingency learning. Standard analyses typically operationalize these processes using regressors encoding events such as recent or frequent congruency and response contingencies. However, because adaptation unfolds over multiple trials, such analyses provide only a partial characterization and may produce spurious effects when multiple processes operate simultaneously. We address this problem using a Bayesian generative model of evidence accumulation that tracks multiple latent expectations. Iterative cycles of model criticism and extension reveal that sequential effects in the contingency learning task are best described as a combination of processes, including stimulus-response contingency learning in tandem with a more strategic signaling effect. We show that these processes can mimic conflict adaptation effects found in the Stroop task. However, the model’s trial‑level characterization of sequential dynamics helps dissociate these processes from pure conflict adaptation processes, revealing two additional updating mechanisms—sequential and item‑specific conflict adaptation—which are detectable even in a highly controlled confound minimized design. More broadly, we argue that analyses of multi-trial sequential data should be accompanied by posterior predictive model criticism, as significance tests on history-dependent regressors cannot by themselves rule out misspecified accounts of the observed dynamics.
Authors: Sven Wientjes, Clay B. Holroyd
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7169761853