Apparent sequential effects due to trial-to-trial fluctuations in decision criterion

Decision making is shaped by sequential effects, whereby preceding stimuli and responses influence current decisions. Computational models typically explain these effects by assuming that decision parameters are causally updated as a function of experimental history. Here, we challenge this assumption by showing that several hallmark sequential effects can instead emerge from slow fluctuations in the decision criterion. Using simulations based on Signal Detection Theory (SDT), we demonstrate that trial-to-trial criterion fluctuations, independent of experimental history, generate three commonly observed sequential effects: the repulsive effect of the previous stimulus (i.e., a tendency to alternate), the influence of previous response (i.e., choice history bias), and the modulation of choice history bias by previous confidence. To disentangle causal effects from effects driven by criterion fluctuations, we used the Hierarchical Model for Fluctuations in Criterion (hMFC), which simultaneously estimates both processes. Applying hMFC to 14 empirical datasets revealed that the repulsive effect of the previous stimulus was fully explained by criterion fluctuations. Accounting for criterion fluctuations also substantially altered choice history bias estimates: attractive biases disappeared in all four affected datasets, while three datasets revealed a previously masked tendency to alternate and two showed stronger alternation. In contrast, the modulation of choice history bias by previous confidence was largely unaffected by concurrent estimation of criterion fluctuations. Collectively, these findings demonstrate that criterion fluctuations account for a substantial portion of sequential effects and highlight the importance of explicitly modeling them to avoid confounded interpretations of behavioral and neuroimaging data.