Abstract
How does confidence track temporal attention? Visual confidence is a second order estimation of a primary decision and could be assimilated to the subjective probability of being correct in a task. Metacognitive sensitivity concerns the variance in accuracy that confidence could account for, after controlling for bias. Temporal attention has been studied through the "Attentional Blink" (AB) phenomenon: when two targets (T1 and T2) are displayed too close in time in a RSVP stream, the second target is often missed. This effect is most pronounced for the second and third items after T1 in the stream. Yet, when T2 is presented just after T1, the AB is not present – an effect called "lag-1 sparing". Here, we ask whether observers are metacognitively aware of such non-monotonous fluctuations of their performance during the AB. To do so, we engaged participants (N=34) in a RSVP identification task: they had to report two target letters in a stream, and then evaluate their confidence about these reported letters. Discrimination accuracy replicated the AB and Lag-1 sparing phenomena. Confidence about T2, however, was dissociated from accuracy: confidence was comparable for Lag-1 and Lag-2, while accuracy was much greater at Lag-1. To distinguish between metacognitive bias and metacognitive sensitivity, the distribution of errors around T2 during Lag-1 was analysed. Despite the strong under-confidence, participants' confidence still discriminated between large and small temporal selection errors. However, comparing Lag-1 to Lag-6 that showed similar accuracy levels, metacognitive sensitivity was reduced for Lag-1, suggesting that only part of the confidence evidence was preserved at Lag-1. At the peak of the AB, in the near-absence of attention (Lag-2), metacognitive sensitivity was nearly absent. These results suggest that confidence blink, which seems mainly due to metacognitive bias, has a temporal structure that is distinct from both temporal attention and metacognitive sensitivity.
Meeting abstract presented at VSS 2018