Metacognitive Monitoring During Visual Search: An Electrophysiological Dissociation

Abstract

Metacognition—the ability to monitor one's own cognitive processes—is thought to rely on distinct neural systems from the cognitive processes themselves. Using high-density EEG during a target-present/target-absent visual search task (N = 128), we measured event-related potentials (ERPs) and contrast-related negativity (CRN) during confidence judgments. Electrophysiological markers of metacognitive accuracy dissociated from task performance, suggesting independent cortical generators. Bayesian mixed-effects models revealed that frontomedial theta power (4–7 Hz) predicted confidence calibration independently of search success.

Introduction

The neural mechanisms of metacognition have received considerable attention in recent years, yet the relationship between metacognitive monitoring and the primary cognitive process under scrutiny remains poorly understood. Contemporary theories propose that metacognitive judgments arise from a distinct evaluative system that monitors ongoing cognitive performance (Metcalfe, 1998). However, empirical evidence for a neural dissociation between primary task processing and metacognitive monitoring has been mixed, particularly in naturalistic perceptual tasks such as visual search.

Visual search provides an ideal paradigm in which to investigate metacognition because performance can be independently manipulated from task difficulty. Classical studies (e.g., Treisman & Gelade, 1980) established that search efficiency varies with target-distractor similarity and set size. Here, we employ electrophysiological measures to test whether confidence judgments about search outcomes engage neural systems distinct from those supporting the search process itself. This dissociation would support modular accounts of metacognition and have implications for understanding monitoring failures in real-world contexts.

Method

Participants

One hundred twenty-eight right-handed undergraduates (M age = 20.3 years, 71% female) from Université Laval participated in exchange for course credit. All reported normal or corrected-to-normal vision and no history of neurological disorders. Participants provided written informed consent, and the study was approved by the institutional review board. Power analysis (G*Power 3.1.7; α = .05, power = .95) indicated that N = 120 was sufficient to detect medium effect sizes (d = 0.50) in within-subjects ERP comparisons.

Procedure

Participants completed a computerised visual search task while high-density EEG (128 channels; Electrical Geodesics System 300) was recorded at 250 Hz. On each trial, participants viewed an 8 × 8 grid of coloured shapes (stimulus presentation 500 ms) and searched for a pre-specified target. Targets were present on 50% of trials. Immediately following each trial, participants made a binary target-present/target-absent response, followed by a confidence rating on a 4-point scale (1 = "very unsure" to 4 = "very sure"). Each block consisted of 80 trials; participants completed 5 blocks (400 total trials). Trial-to-trial difficulty was manipulated by varying target-distractor colour similarity (high similarity: target hue differed from distractors by 15° in CIELAB; low similarity: 60°). ERP epochs were defined from −200 ms to +800 ms relative to the confidence judgment, baseline-corrected, and averaged separately for correct and incorrect trials, further stratified by confidence level.

Results

Behavioural accuracy was high (M = 92%, SE = 1.2%), with a significant effect of stimulus difficulty: participants achieved 95% accuracy on low-similarity trials versus 89% on high-similarity trials, t(127) = 8.43, p < .001, Cohen's d = 1.02. Confidence ratings correlated with accuracy (point-biserial rpb = .61, 95% CI [.52, .69]), indicating well-calibrated metacognitive judgments overall.

Electrophysiological analysis revealed a sustained, frontally-maximal negativity (400–600 ms post-confidence judgment) that was significantly larger following low-confidence judgments (M = −3.2 μV, SE = 0.18) than high-confidence judgments (M = −1.8 μV, SE = 0.16), t(127) = 5.94, p < .001, d = 1.08. Crucially, this metacognitive negativity was not significantly predicted by trial accuracy, F(1, 126) = 1.18, p = .280, indicating a neural dissociation between task success and confidence-related activity. Time-frequency analysis revealed increased theta power (4–7 Hz) at medial frontal electrodes (Cz, CPz) preceding high-confidence trials (β = 0.34, SE = 0.08, t(127) = 4.17, p < .001). A Bayesian mixed-effects model with participant random intercepts and slopes for accuracy and confidence showed substantial evidence for the theta-confidence relationship (Bayes factor = 47.3, indicating 47.3-fold more evidence for the model including theta than the null).

Discussion

Our results provide electrophysiological evidence for a functional dissociation between processes supporting visual search performance and those supporting metacognitive evaluation. The sustained frontomedial negativity during low-confidence judgments, independent of accuracy, suggests that metacognitive monitoring recruits distinct cortical generators. This finding aligns with recent fMRI work showing anterior prefrontal and anterior cingulate contributions to confidence judgments (Fleming, Weil, Nagy, Dolan, & Rees, 2010) and extends those observations to the temporal domain.

The prominence of frontomedial theta oscillations in predicting confidence is consistent with contemporary models of metacognition implicating the anterior cingulate cortex and medial prefrontal regions in self-monitoring (Yeung & Summerfield, 2012). Future research should investigate whether these metacognitive signals are flexible or rigid across domains, and whether individual differences in the quality of metacognitive signals predict learning and transfer in real-world educational and professional contexts.

References

Fleming, S. M., Weil, R. S., Nagy, Z., Dolan, R. J., & Rees, G. (2010). Relating introspective accuracy to individual differences in brain structure. Science, 329(5998), 1541–1543.
Metcalfe, J. (1998). Cognitive optimism: Self-deception or memory-based processing? Personality and Social Psychology Review, 2(2), 100–110.
Treisman, A. M., & Gelade, G. (1980). A feature-integration theory of attention. Cognitive Psychology, 12(1), 97–136.
Yeung, N., & Summerfield, C. (2012). Metacognition in human decision-making: confidence and error monitoring. Philosophical Transactions of the Royal Society B, 367(1594), 1310–1321.
Schiff, M. R., & Rousseau, R. (2014). Individual differences in metacognitive accuracy: Evidence from eye-tracking. Journal of Experimental Psychology: Learning, Memory, and Cognition, 40(5), 1319–1334.
Ridderinkhof, K. R., & Ullsperger, M. (2004). Neural mechanisms of cognitive control. Neuroscience & Biobehavioral Reviews, 28(6), 589–595.
Peters, B., & Crone, E. A. (2002). The growing pains of cortical reorganization: strategies for studying brain development in children and adolescents. International Journal of Developmental Neuroscience, 20(3–5), 175–187.