The first test of the motor hypothesis for self-monitoring was to define the monitoring system for decisions that we make compared to decisions that we observe. We reasoned that confidence in overt decisions should derive from both the quality of perceptual evidence and the monitoring of action signals associated with overt decisions. In contrast, observed decisions should not involve action signals. Healthy participants made and observed decisions while we recorded their brain activity to identify the neural signals associated with confidence. Our results showed that metacognitive performance was better for committed than for observed decisions, in particular because of the involvement of sensorimotor brain regions. We pursued this line of work by conducting a follow-up study comparing confidence judgments and metacognitive performance between conditions that differed only in the sensorimotor information available for the decision, but were indistinguishable in terms of perceptual evidence. We found that response-related sensorimotor activity modulated confidence judgments over and above the strength of internal perceptual signals. Taken together, these results revealed which aspects of motor behaviour during overt decisions specifically contribute to metacognition. We also investigated the neural underpinnings of such contributions using intracranial EEG in people with epilepsy.
Regarding the clinical aspects of this project, we quantified metacognition in schizophrenia to test whether the metacognitive deficits described in this psychiatric disorder stem from deficient sensorimotor signaling. To our surprise, we found that metacognitive performance itself was not impaired in schizophrenia. However, we discovered that sensorimotor parameters associated with decision-making were less closely related to confidence in patients than in matched controls. The fact that metacognitive performance was preserved in patients was unexpected and contradicted most clinical and scientific literature. To better quantify the putative metacognitive deficits in schizophrenia, we conducted a meta-analysis, suggesting that these deficits may have been overestimated in previous research due to pervasive cognitive impairments that are not specific to metacognition. The results provided strong support for our hypothesis, indicating that metacognitive deficits were twice as small in studies that controlled for cognitive performance. We concluded that efforts should be made to develop experimental protocols accounting for reduced cognitive performance in schizophrenia before evaluating potential deficits. This information informed our second aim, which was to develop and assess a method for optimizing metacognitive judgments.
Having shown that perceptual metacognition is preserved in schizophrenia, we explored metacognitive deficits in more complex tasks involving memory. As with perception, we found that metamemory remains intact in schizophrenia, provided that the observed memory deficits are accounted for. We expanded our investigations to obsessive-compulsive disorder (OCD), revealing that individuals with OCD experience a metacognitive deficit in the perceptual domain
We also attempted to replicate and extend a procedure for enhancing metacognitive monitoring. In this procedure, we identified two potential confounds that might have led to an artificial increase in metacognitive performance. Consequently, we conducted a pre-registered conceptual replication. Importantly, after addressing these two confounds, we found moderate evidence for a lack of metacognitive training. Furthermore, we sought to alleviate the metacognitive deficits observed in OCD using deep brain stimulation. Our results indicated that acute changes in DBS were insufficient to restore metacognitive performance. Thus, our findings suggest that previous claims regarding metacognitive training are premature and highlight the need for further research into how individuals can be trained to monitor their own performance.
In addition to the motor hypothesis for self-monitoring, we considered other possible mechanisms to explain confidence and metacognition. In particular, we developed a dynamic model of confidence based on evidence accumulation, which proposes that participants rate their confidence in a given choice in proportion to the maximum evidence supporting that choice. This model successfully reproduced confidence ratings for both perceived and unperceived stimuli, which is highly relevant given the close links between metacognition and perceptual consciousness. We used this model to elucidate subjective aspects of perceptual experience, including subjective duration and confidence. Finally, we employed evidence accumulation models and intracranial EEG to explain changes of mind during perceptual decision-making.
