Basic Overview

As you read the words on this page, your brain produces a conscious experience – a subjective ‘what it is like’ to be reading. Simultaneously, your brain also produces a metacognitive sensation of confidence, keeping you informed of how well you understand this text. This metacognitive ability to monitor your experiences is a crucial part of decision making, because it allows you to learn from your mistakes and make better decisions in the future. In the ECG, we study how these processes are influenced by incoming signals from the visceral organs of the body, and how these process can go awry in mental illness. For example, using psycho-physics, computational modelling, and neuroimaging, we study how fluctuations in heartbeats, arousal, and respiration alter  the neural computations underlying perceptual decision-making and metacognitive learning. This work takes place in two interacting research units, with a core basis at Aarhus University where we conduct large-scale brain imaging in healthy participants, and a clinical lab at Cambridge Psychiatry, where we apply our tasks and models to brain imaging in a variety of clinical and psychiatric populations. This work takes the form of several interlocking projects, which you can read more about below.

An Embodied Predictive Processing Approach


Our approach to understanding brain-body interaction is grounded in the application of mechanistic, computational models to understand the inferences our brain makes when we make decisions. To this end, we develop computational models to try and understand how the brain combines visceral (interoceptive) and perceptual (exteroceptive) signals during conscious perception and metacognitive-self monitoring. At the core of this is the idea that your brain and body work together to support hierarchical Bayesian inferences, and that we can model processes like subjective confidence as arising from how the brain integrates inferences about the body and world. These models will in turn enable us to ask better, more focused questions about what happens when these mechanisms go awry; for example, if anxiety relates to an overly strong influence of visceral information (‘gut feelings’) on confidence or perception. You can read more about these models here

Hierarchical Modelling of Self-Awareness


We take a multi-level approach to understanding the psychological mechanisms underlying perception and metacognitive learning. To do so, we use multivariate and hierarchical modelling techniques to try to understand how perception and metacognition interact across different modalities such as perception, memory, and interoception. Here, the key goal is to understand the domain generality vs domain specificity of metacognition in these domains, and how visceral inferences shape the modularity of metacognition. For example, is there a general system for confidence in the brain across memory, bodily sensations, and visual perception? Or are these systems relatively isolated from one another as distinct cognitive modules? By modelling these systems as interacting hierarchies, we hope to uncover latent variables which predict behavior across a variety of domains and may thus have maximal trans-diagnostic utility for understanding psychiatric illness and symptoms. You can read more about this approach here.

Embodied Computational Psychiatry


What happens to the mind when the body becomes unreliable? Or vice versa; when we are anxious or unhappy, how do these symptoms relate to our function and awareness of bodily processes? Currently our approach to mental illness operates as if the only thing that matters for a healthy mind is what is ‘above the neck’. However, this conflicts with what we know from epidemiology and psychiatry; many if not all psychiatric and personality disorders involve a variety of bodily symptoms such as hyper/hypo-arousal, sleep issues, eating issues and/or weight gain or loss,  sexual disruption, and hyper-sensitivity to bodily signals. Our goal in this project is to use computational modelling and specially designed decision-making tasks to identify how specific computational phenotypes (e.g., a bias towards prior information) interact with visceral processes, and to relate these to mental illness and emotional function. This project takes two forms; in Denmark we conduct large scale phenotyping projects relating brain-body traits to sub-clinical psychiatric symptoms in the general population. In Cambridge, we work together with our clinical collaborators to investigate how surgical, pharmacological, and genetic disruption of the visceral body influences decision-making and self-awareness.

Causal Manipulation of Visceral Prediction Errors

A key aspect of our approach is to go beyond merely modelling brain-body correlations, to actually manipulating bodily states so we can observe their effects. This is important both for verifying the predictions of our models, and for building a basis for future intervention-based research aiming to re-mediate disordered visceral inferences. In this line of research we’re developing a variety of pharmacological and physiological techniques to manipulate visceral states; for example by exploring how beta-blockers influence metacognition, and by developing novel paradigms which directly control the predictability of arousal and visceral signals. You can read more about this work here.

An Open Science Approach

Our lab strives to embodied the principles of open and reproducible science. You can follow our research over at our blog and twitter. As we progress through our different projects, we’ll be sharing all of our code, data, and tasks through our github page and associated data repository. And we’re also open to collaboration – if you have an interesting patient population and would like to use some of our methods to explore how embodied computations may underlie their symptoms, get in touch!