Neurally-informed modelling unravels a single evidence accumulation process for choices and subsequent confidence reports

Subjective confidence in perceptual choices depends on computations occurring prior to and after choice commitment. However, the nature of these computations remains unclear. Current models disagree on two fundamental questions: what stopping-rule is applied to post-choice evidence sampling and to what degree do choice and confidence reports rely on shared versus distinct evidence accumulation processes? These models have proven difficult to dissociate because they often make similar behavioural predictions. Therefore, we used a neurally-informed modelling approach to jointly model initial choices and subsequent confidence reports while leveraging the additional constraints offered by human electrophysiological signatures of evidence accumulation. Participants made self-paced confidence reports after indicating perceptual choices on a random dot motion task, with stimulus presentation continuing during the interval between these responses. Model comparison showed that boundary-based stopping rules provided superior fits to the behavioural data than time-based stopping rules, as the latter could not account for relationships between confidence and confidence-RT. The behavioural fits were inconclusive when comparing models with a single evidence accumulation process dictating choices and confidence reports compared to models invoking a distinct post-choice process. However, the single-process model markedly outperformed the other models in terms of its ability to recapitulate the observed neural evidence accumulation patterns associated with choice confidence. Our study thus demonstrates that choice behaviour, subjective confidence, and associated neural decision signals can be jointly explained by a model invoking a single process of evidence accumulation with separate boundaries for the initial choice and subsequent confidence reports.