Quanta Podcast: Predictive Categorization in the Brain with Connor Feeley

Overview

The Quanta podcast explores a shift in neuroscience from a filing-cabinet view of categorization to a prediction-based framework that ties perception to bodily states and goals. Connor Feeley explains how the brain actively predicts and labels inputs, using interoceptive signals and allostatic regulation to guide behavior.

Key insights

• Categories are context dependent and dynamically shaped by bodily state and goals.
• Prediction, rather than mere feature matching, drives perception and labeling of the world.
• Interoception, the brain’s sense of the body, informs how we experience emotions and actions.
• Two leading neuroscientists Lisa Feldman Barrett and Earl Miller frame categorization as a brain-wide predictive process that spans the nervous system.

Introduction

The podcast launches with a simple everyday example: an identical apple can be perceived as food or trash depending on the eater’s context. This leads into a discussion of categorization in neuroscience and how modern theories view categories as dynamic constructs shaped by bodily state and goals rather than static feature lists.

Traditional vs Predictive Views of Categorization

Historically, perception was thought to passively extract sensory features and match them to stored templates in memory. The new perspective argues perception builds a predictive model of the world, constantly updated by incoming signals and past experience, with the brain sacrificing energy by focusing on prediction errors that are most informative.

Interoception and Allostasis

The hosts introduce the concept of interoception—the brain’s continual monitoring of internal bodily signals like hunger, heart rate, and respiration. Allostasis describes the brain’s predictive regulation of energy use to keep physiological systems running smoothly. Emotions emerge as category-like constructs that shape action by interpreting bodily and environmental context to drive behavior.

The Barrett Miller Framework

Two prominent scientists, Lisa Feldman Barrett and Earl Miller, are highlighted for integrating prediction and bodily needs into categorization. Barrett emphasizes allostasis and emotion as context-dependent labels we apply to bodily states to guide action, while Miller focuses on the neural circuitry of predictive feedback and feedforward processing across the cortex. Together, they propose that categorization is a core organizing principle of the nervous system, enabling efficient, context-sensitive behavior.

Evidence and Neural Mechanisms

The podcast discusses evidence from neuroscience showing abundant feedback connections in the brain, with the visual cortex dominated by predictive signals and compression of sensory input as it travels deeper into the cortex. Surprising or context-violating stimuli trigger stronger feedforward activity and prediction errors, which the brain uses to refine its model. Interoceptive signals travel via pathways like the vagus nerve to influence perception, memory, and decision-making at broad neural scales.

Implications and Reflections

Beyond theory, the discussion touches on how mislabeling bodily states—such as interpreting arousal as anxiety when it might be excitement—can lead to maladaptive responses. The conversation also reflects on how changing our understanding of categories can reshape our thinking about biology, perception, and even everyday experiences.

Conclusion

The podcast closes with a personal takeaway: categories and boundaries are human constructs that are meaningful but flexible, shaped by context and bodily needs as we navigate the world.