Conference on Cognitive Computational Neuroscience

While several models of decision-making allow for evidence-accumulation to continue after choice commitment and influence subsequent confidence judgements, they disagree regarding the underlying computations. According to some models, the accumulation process continues uninterrupted after initial commitment, while others hold that it is re-framed to accumulate evidence toward an ‘error detection bound’. We tested these predictions by transiently modulating the stimulus contrast within a trial towards the correct/incorrect option (termed positive/negative pulses) before or after an initial choice, and measuring their impact on a neural signature of evidence accumulation with EEG (centro-parietal positivity; CPP). Positive pre-choice pulses increased CPP build-up and improved accuracy, RT, and confidence, while negative pulses had the opposite effect on behaviour. Confirmatory post-choice pulses increased CPP build-up more than disconfirmatory ones, even after correct initial-choices, contradicting error-accumulation predictions. Post-choice pulses had less effect if they arrived later after the initial choice, suggesting that accumulation was terminated while evidence remained available. We will compare simulations and fits of different post-decision accumulation models to explain our data.