Visual confidence depends on serial task difficulty and explicit report of confidence judgments

In a series of hundreds of trials, confidence on a visual task demonstrates serial dependence. However, it remains unclear what serial factors give rise to such serial dependence. In the present study, we addressed this question by building a series of motion-discrimination tasks based on trial sequences with manipulated properties. In every trial, participants performed a left-right, direction-discrimination task on a random-dot motion pattern, and simultaneously indicated their confidence on a four-point scale. We first calibrated task difficult to individual participants’ discrimination sensitivity. Then, we manipulated serial task difficulty by controlling motion coherence, so that a trial with medium difficulty level was preceded by either easy trials or hard trials. In Experiment 1, we found that confidence rating for the medium-difficulty trial was higher when it was preceded by easy trials than when preceded by difficult trials, although direction-discrimination accuracy remained constant. In Experiment 2, such serial dependence on task difficulty was found to weaken significantly when participants were instructed to judge motion speed instead of to give confidence judgment in the preceding trial. Our findings suggest that both task difficulty and explicit report of confidence in preceding trials contribute to the serial dependence of judgments on visual confidence.