The effect of confidence on visual perceptual learning

Abstract

Listen

Translate article

Visual perceptual learning can occur in the absence of external feedback provided in the environment (Sasaki et al., 2010, Nat Rev Neuro). Here, we are testing whether confidence about the validity of our perceptual decisions could be this internal mechanism that can affect perceptual learning when there is no external feedback. We asked human participants to decide about the direction of motion of a random-dot kinematogram (RDK) stimulus over five consecutive days. Every two decisions, participants reported their confidence about the accuracy of their perceptual decision using the confidence forced-choice (CFC) method. In order to investigate the effect of confidence on perceptual learning, we manipulated confidence by intertwining two conditions in which the signal-to-noise ratio is similar but the absolute evidence differs. The two conditions consisted in high- and low-density stimuli, where the number of both signal and noise dots was higher in the high than in the low-density condition. To separate the learning procedures in the two conditions, the direction of motion in the RDK stimulus was along different axes in high- vs. low-density conditions. Thresholds were estimated for %65 and %85 correct at the beginning of each day, using the method of constant stimuli (day 1) or an Accelerated Stochastic Approximation (ASA) staircase (subsequent days). In line with previous studies (Maniscalco et al., 2020, PsyArXiv), our results showed that increasing the absolute evidence while keeping the signal-to-noise ratio constant leads to overconfidence in the high-density condition. Additionally, in the low-density condition, as compared to the high-density, participants exhibited a greater improvement in both confidence efficiency and sensory threshold across days. Altogether, our results suggest that in the absence of external feedback, confidence can act as the only available feedback system, and can facilitate visual perceptual learning.