Cognitive Control and Metacognition: The Effect of Interference, Working-Memory Load, and Shifting on Confidence in Response Correctness and Partial Error

Background/Objectives: Confidence is typically studied in simple perceptual decisions, although everyday decisions also require cognitive control and monitoring of action. Across two experiments, we examined whether response conflict, stimulus complexity, and sequence affect task performance and retrospective metacognitive judgments concerning different aspects of performance: confidence in response correctness and confidence that a covert partial error occurred. Methods: Participants performed a modified Stroop task in which they responded to word colour and, in compound trials, also reported the position of an accompanying line. In Experiment 1 (N = 66), participants rated confidence in the correctness of their responses. In Experiment 2 (N = 47), electromyography was used to identify covert partial errors, and participants rated confidence that a partial error had occurred. Results: Conflict reduced response accuracy, prolonged response times, and was associated with a small reduction in confidence in correct responses. In Experiment 1, the effect of sequence on confidence depended on stimulus complexity: confidence was higher in simple-repeated than in simple-switched trials, whereas the opposite pattern was observed for compound trials. In Experiment 2, partial errors occurred more frequently under conflict but less frequently in compound trials. Greater incorrect-burst surface and longer correction time were associated with higher partial-error confidence, and both relationships were weaker in compound trials. The association with incorrect-burst surface remained reliable after adjustment for response time. Exploratory analyses also showed lower metacognitive sensitivity under conflict and complexity and lower metacognitive efficiency in compound trials. Conclusions: Cognitive-control demands were associated with retrospective metacognitive judgments concerning both completed response outcomes and partial-error events occurring during response generation. Conflict was associated with a small reduction in confidence in correct responses, whereas partial-error confidence was associated with measurable characteristics of the partial-error event. The weaker relationship between EMG-derived response-process characteristics and confidence during compound processing suggests that current task demands affect how strongly internal response-related information is related to metacognitive judgments.