Metacognition and Computational Psychiatry Group

Head: Steve Fleming

We have focused on two key lines of research, on the affective and perceptual dimensions of human metacognition.

First, we have developed empirical tools to study metacognitive evaluation across different temporal and "spatial" (domain) scales. Specifically, we have pursued the relationship between “local” confidence in individual decisions or actions, and global features of self-evaluation such as self-efficacy, and asked how these variables generalise across different domains (e.g. does evaluating ourselves negatively on task X affect how we think about our skill on task Y). This has led to the development of novel tasks that ask people to reflect on their performance at different timescales (Rouault et al., 2019 Nat Comms), allowing us to study how one level (e.g. local confidence judgments) informs long-run global self-performance estimates, and vice-versa. We have used fMRI to identify a neural nexus for local-global interactions (Rouault et al., 2020 PNAS), and shown how global metacognition changes as a function of ageing (McWilliams et al., 2021 psyArXiv), self-esteem (Rouault et al., 2022 Trans Psychiatry) and functional disorders (Bhome et al., 2022 Brain Comms). In ongoing work, we are seeking to formulate and test models of how external feedback, attention, emotion and episodic memory contribute to self-performance estimates, and how such computations may be altered in disorders of mental health (Seow et al., 2021 Biol Psychiatry). 

Second, we have developed models of how self-monitoring the (global) precision of a perceptual generative model can provide a principled basis for understanding how and why subjects become aware of perceptual content, whereas in other circumstances information remains unconscious (Fleming, 2020 Neurosci Consc). Such a model makes specific predictions about the temporal and spatial hierarchical organisation of prediction errors on content and precision (awareness), which we are testing with fMRI and MEG in ongoing work. In addition, this framework can account for why people find it hard to distinguish between imagination and reality, as both can lead to high-precision perceptual content (Dijkstra et al., 2022 Neurosci & Biobehav Rev). We have found such striking failures of reality monitoring – erroneously assigning imagined content to be real – in novel high-N, low-trial task designs involving surreptitious one-trial presentations of “real” stimuli (Dijkstra & Fleming, 2021 psyArXiv).

Key publications 

1. Dijkstra, N., Kok, P., & Fleming, S. M. (2022). Perceptual reality monitoring: Neural mechanisms dissociating imagination from reality. Neuroscience & Biobehavioral Reviews, 104557.
2. Fleming, S. M. (2020). Awareness as inference in a higher-order state space. Neuroscience of Consciousness, 2020(1), niz020.
3. Seow, T. X., Rouault, M., Gillan, C. M., & Fleming, S. M. (2021). How local and global metacognition shape mental health. Biological Psychiatry, 90(7), 436-446.