Are affective expectations modulated by the interplay between visceral signals and uncertainty of the sensory environment?

Alexandrina Vasilichi, Micah Allen, Peter Dayan

Many papers highlight an important role for interoceptive inference and visceral rhythms in shaping our exteroceptive perception. Furthermore, there is a growing consensus in perception science that expectations shape how we see the world. Here, we consider whether and how expectations in the affective domain might be shaped by the cardiac signal as well as its interaction with uncertainty about the sensory environment.

We administered the Confidence Weighted Task (CWT), a novel perceptual and affective reversal learning task, while recording cardiac signals. We will test whether an interaction between sensory uncertainty and prior expectations of the sensory stimuli influences behaviour, here operationalized as reaction times of stimulus decisions, decision confidence, and decision accuracy. Then we will test whether this interaction is dependent on the phases of the cardiac cycle. Subsequently, we also plan to use a computational model called the Hierarchical Gaussian Filter (HGF) to model the beliefs subjects might form about the stimuli and cues during a trial, as well as how they change with accumulating evidence throughout the task.

The dataset was already collected by my secondary supervisor Micah Allen in Aarhus, Denmark. To date, I have been curating the dataset and writing scripts to bidsify both the behaviour and physiology files. At the moment, I am using NeuroKit to clean the cardiac signal and extract the cardiac events indices throughout the cardiac phase. Subsequently, I will calculate where in the cardiac phase stimuli occurred. Then, I will do the behavioural and physiological analysis, followed by HGF modelling.

The overall aim of this project is to increase understanding of the potential (computational) mechanisms by which interoceptive signals (here, the cardiac signal) control affective expectations and affective learning processes.