Visual confidence accurately tracks increasing internal noise with eccentricity in peripheral vision.

Area of Science:

• Cognitive Neuroscience
• Psychology
• Computational Neuroscience

Background:

• Sensory representations are noisy, impacting decision-making.
• Perceptual confidence, the evaluation of decision quality, is crucial but its accuracy in tracking internal noise is debated.
• Peripheral vision offers a unique environment to study this, but prior research yielded inconsistent findings.

Purpose of the Study:

• To investigate whether perceptual confidence accurately reflects internal sensory noise, particularly in peripheral vision.
• To test the Bayesian-confidence hypothesis, proposing confidence derives from the posterior probability distribution of perceptual estimates.
• To resolve discrepancies in previous studies by employing a normative Bayesian framework and incentivized confidence measurements.

Main Methods:

• Utilized a normative Bayesian framework to model perceptual confidence.
• Conducted two perceptual tasks (spatial localization, orientation estimation) with varying stimulus eccentricity.
• Measured confidence using post-decision wagering, where participants set a range around estimates for rewards based on accuracy and range width.
• Estimated sensory noise, assuming it increases linearly with eccentricity, to predict confidence.

Main Results:

• The Bayesian ideal-observer model provided the best prediction of confidence across both tasks.
• Human confidence judgments accurately reflected the increased sensory noise observed with greater stimulus eccentricity.
• Incentivized confidence measurements aligned with normative predictions.

Conclusions:

• Humans possess the metacognitive ability to accurately monitor internal noise, even in peripheral vision.
• Perceptual confidence serves as a reliable indicator of sensory evidence quality.
• Findings support the Bayesian-confidence hypothesis and suggest optimal use of sensory information for decision-making.