Similar computational noise for perceptual decision making with confidence, expectation, and reward

Mounting evidence suggests that a critical source of corruption in perceptual decision making is computational noise. However, it remains unclear whether similar computational noise appears in different types of perceptual decisions. Here we examined whether metacognitive noise – a type of noise hypothesized to only affect type-2 confidence ratings – also appears in type-1 decisions biased by expectations or rewards. Subjects (N=271) performed a simple perceptual discrimination task and in different conditions provided confidence, or incorporated expectation or reward cues. Metacognitive noise manifests as a decrease in sensitivity (d') when computed based on confidence criteria compared to the decision criterion. We first replicated this drop in sensitivity with confidence ratings. Critically, we found equivalent drops in sensitivity for both expectation and reward cues, regardless of whether the cues were presented before or after the stimulus. A computational model with a single type of computational noise that underlies confidence, expectation, and reward reproduced these results. These findings suggest the existence of similar computational noise in perceptual decision making involving confidence, expectation, and reward, thus suggesting common computational principles across type-1 and type-2 decisions.