Abstract

Humans often face decisions between multiple alternatives. In these contexts, some evidence suggests that only the alternative with highest evidence is represented by the decision system. However, other findings indicate that information from unchosen alternatives remains accessible for decision computations. Furthermore, the amount of information that reaches metacognitive levels remains unexplored. By fitting competing computational models to data from two pre-registered experiments (N=32), we found that noisy evidence from unchosen alternatives reaches the decision system. We extended this by fitting the models to two previously published datasets involving different stimuli (N=17) and found concordance evidence. Moreover, allowing a separate channel of evidence not corrupted by noise for metacognitive computations improves model fitting, which suggests that the metacognitive system has access to more evidence than the decision system. These results act as a bridge between previous conflicting findings in multialternative decisions and extend them by also modelling the metacognitive stage.