Abstract
Recent studies suggest that metacognitive judgments (type-2) rely on information that is not accessible to a perceptual decision-making process (type-1). To explore what kind of information type-2 tasks rely on, we asked participants to perform both type-1 and type-2 tasks while manipulating bottom-up attention; They decided whether the right edge of a target was above or below the horizontal meridian (type-1 task) and simultaneously judged their confidence in the decision (type-2 task). The stimuli were oriented bars (25°-65° either clockwise or counter-clockwise from the horizontal meridian) presented to the left and right visual fields in a rectangular (Exp. 1) or circular (Exp. 2) grid. Among them, a randomly selected bar was cued as the target. Before the stimuli appeared, a rectangular (Exp. 1) or arc-shaped (Exp. 2) frame was briefly flashed in either of visual fields to draw attention. Type-1 sensitivity, type-2 sensitivity, and the ratio between them (metacognitive efficiency) were calculated both when the target was attended and unattended. Metacognitive efficiency signifies how much of the information used in the type-1 task is available to the type-2 task. We found that type-1 sensitivity was higher in the attended condition, but metacognitive efficiency was similar (Exp. 1) or even lower (Exp. 2). Furthermore, confidence ratings were higher in the attended condition regardless of the correctness of the perceptual decision. Metacognition might have reflected whether attention was allocated to the target, which would increase the subjective sense that one is correct. Further analysis of the target orientation confirmed this idea. Participants were more confident about targets with a larger orientation difference from the horizontal meridian, while their type-1 performance was similar. These results suggest that metacognition on perceptual discrimination exploits proxy evidence that influences a subjective sense of correctness, which is independent from the perceptual decision itself.
Meeting abstract presented at VSS 2018