Why predictive coding is a natural fit for interoception and the bodily self
Predictive coding is the idea that the brain constantly generates expectations about what it will sense and then updates those expectations based on the difference between what it expected and what actually happened. This works for signals from inside the body (interoception) just as it does for sights and sounds. The insula, a brain region that receives information about internal organs and bodily states, is thought to play a key role in estimating how reliable those internal signals are and in adjusting the strength of top-down predictions accordingly [1]. In this view, bodily self-awareness arises when the brain successfully integrates predictions about the body with incoming sensory feedback, minimizing the mismatch between them.
This framework naturally explains how interoception could support multisensory integration: the brain combines internal signals with external ones to create a coherent sense of the body as one's own. For example, in the rubber-hand illusion, where a person feels a fake hand as part of their body, interoceptive signals appear to directly influence the experience of ownership through multisensory integration, a finding that supports interoceptive predictive coding models of selfhood [2].
Body illusions show interoception shaping ownership and agency
The rubber-hand illusion is a classic demonstration that the feeling of body ownership depends on integrating what we see, feel, and know about our body. The finding that interoceptive signals modulate this illusion suggests that the brain uses internal bodily information as part of its predictive model of the self [2]. This is consistent with predictive coding, where the brain weighs sensory evidence against prior expectations, and interoceptive signals contribute to those expectations.
In anorexia nervosa, a condition marked by distorted body experience, predictive coding models help explain why patients may prioritize their own internal sensations over actual sensory input. One study found that individuals with anorexia nervosa relied more on their confidence in interoceptive signals (metacognitive awareness) than on bottom-up bodily information, leading to persistent body distortions even after rehabilitation [3]. This suggests that when interoceptive predictions become overly precise or rigid, they can dominate multisensory integration and lock the person into a distorted bodily self-awareness. Such findings support the idea that predictive coding can account for both typical and atypical bodily self-experience.
The model is promising but not yet proven better than alternatives
Much of the support for predictive coding comes from its ability to explain existing observations rather than from direct tests that compare it against other theories. For instance, the proposal that aphantasia—the inability to voluntarily form mental images—arises from reduced interoceptive precision in the insula is explicitly described as a theoretical model, with limited direct empirical evidence linking interoception and imagery in aphantasia [1]. Similarly, the rubber-hand illusion study supports the involvement of interoception but does not directly test predictive coding against competing accounts [2].
Other work discusses predictive coding as a mechanistic account of bodily self-awareness and multisensory integration, but often in the context of broader reviews or clinical applications rather than head-to-head comparisons [4][5]. This means that while predictive coding offers a unified and intuitively appealing framework, it has not yet been shown to outperform alternatives such as simple associative learning or other integration models. The evidence is consistent with predictive coding, but it does not uniquely confirm it.
What would make the predictive-coding explanation more convincing
To establish that predictive coding explains interoceptive contributions to multisensory integration and bodily self-awareness better than alternatives, researchers would need experiments that directly manipulate interoceptive precision and measure the consequences for multisensory integration and self-experience. For example, studies could test whether altering interoceptive signals (such as through breathing or heartbeat perception tasks) changes the strength of body illusions in ways that predictive coding uniquely predicts. They could also use neuroimaging to see whether the insula's predicted role in modulating prediction gain is actually observed during tasks that require integrating internal and external signals.
The anorexia nervosa study provides a hint of how this could work: it showed that interoceptive metacognition was linked to embodiment even when sensory signals were mismatched, suggesting that patients' prior expectations overrode bottom-up input [3]. If similar patterns are found in healthy people under controlled manipulations, and if they can be explained only by predictive coding, the case would be stronger. Until then, predictive coding remains a leading candidate but not a proven superior explanation.
Sources used in this answer
Interoception, insula, and agency: a predictive coding account of aphantasia
Proposes a predictive coding model in which reduced interoceptive precision in the insula disrupts the gain of top-down predictions, impairing mental imagery and contributing to aphantasia, but notes the model is theoretical with limited direct empirical evidence.
Multisensory integration across exteroceptive and interoceptive domains modulates self-experience in the rubber-hand illusion
Shows that interoceptive signals directly influence the experience of body ownership in the rubber-hand illusion via multisensory integration, supporting interoceptive predictive coding models of selfhood.
Mental health meets computational neuroscience: A predictive Bayesian account of the relationship between interoception and multisensory bodily illusions in anorexia nervosa
Finds that individuals with anorexia nervosa prioritize interoceptive metacognitive processes over bottom-up bodily inputs, leading to persistent body distortions, consistent with a Bayesian predictive account of interoception and multisensory integration.
Beyond Awareness: Visual Perception, Body Illusions, and Cognitive Biases in Anorexia Nervosa and Obesity
Discusses visual perception, body illusions, and cognitive biases in anorexia nervosa and obesity, aligning with predictive coding models of the bodily self and multisensory integration.
Developing a self from the inside out
Explores predictive coding as a mechanistic account of bodily self-awareness and multisensory integration, but does not explicitly detail interoceptive contributions to multisensory integration.