Interoceptive predictions in the brain

Lisa Feldman Barrett is University Distinguished Professor of Psychology at Northeastern University, where she has run the Interdisciplinary Affective Science Laboratory since her arrival in 2008; she previously held positions at Penn State and Boston College. Her 2017 trade book How Emotions Are Made (Houghton Mifflin Harcourt) extended her theory of constructed emotion to a general audience. W. Kyle Simmons was at the Laureate Institute for Brain Research and the University of Tulsa Department of Community Medicine at the time of the paper — the same Simmons who co-authored the 2018 Roadmap — and brought the visceromotor-cortex anatomy expertise. Their collaboration sits across affective science and clinical imaging.

Published in Nature Reviews Neuroscience in May 2015, the piece is a synthesising review proposing a specific architectural model — the Embodied Predictive Interoception Coding (EPIC) model — that integrates corticocortical anatomy with Bayesian active- inference principles. The argument is that the agranular visceromotor cortices — including the anterior insula but also the anterior cingulate cortex and ventromedial prefrontal cortex — generate top-down predictions about expected body state, which are then compared against ascending visceral afferents; the prediction error is what reaches awareness. The model places the locus of interoception across a wider cortical network than Craig's insula-centric frame and reads the insula as one node in a predictive system rather than as the seat of feeling.

The piece sits as the empirical-anatomical companion to Seth's 2013 theoretical statement, with which it shares the predictive-processing frame but adds specific architectural commitments grounded in cortical- connectivity data. It is also the principal Barrett-camp counter- statement to Craig's interoception model: the two frames coexist in the field, with the Roadmap treating them as legitimate alternatives rather than declaring a winner. Where Craig sees the right anterior insula as the substrate of subjective feeling, Barrett and Simmons see feeling as constructed across a distributed visceromotor network, with the insula playing a comparator role rather than a feeling- generator role. The anatomical foundation it leans on is the kind of cytoarchitectonic synthesis Augustine 1996 provided.

The stake is theoretical and disciplinary. Barrett has built the theory of constructed emotion across a long programme of work and has substantial stakes in its scientific reception; her trade book and public-facing argument-making (TED talks, popular essays in the New York Times and elsewhere) give the position visibility beyond the journal literature. Simmons's stakes are more methodological, anchored in the corticocortical-connectivity data the EPIC model rests on. None of the work has direct commercial application, but the constructed-emotion frame is the principal alternative-to-Craig position within the field, and the alignments of researchers behind one or the other frame have become legible in the post-2015 literature. Read it as the second major theoretical option, not as a synthesis of the field.