Vagal Tone and Predictive Coding

Vagal Tone and Predictive Coding: A Neurophysiological Model for Interoceptive Inference Contemporary review evidence provides only modest support for predictive coding, and active inference still requires stronger empirical validation against competing accounts (Hodson et al., 2024). Current review evidence provides only modest empirical support for predictive coding, while active inference requires stronger validation against competing models (Hodson et al., 2024).

J.Michelle

Abstract

The integration of autonomic nervous system function with higher-order cognitive processes remains a central challenge in affective neuroscience. This manuscript proposes a theoretical framework linking vagal tone—indexed by respiratory sinus arrhythmia and heart rate variability—to predictive coding mechanisms in the brain. Vagal tone reflects parasympathetic modulation of cardiac function and has been associated with emotional regulation, social engagement, and adaptive responding to environmental demands. Predictive coding theory posits that the brain continuously generates predictions about sensory input and updates these predictions based on prediction errors. We argue that vagal tone serves as both an index and a mechanism of interoceptive predictive coding, wherein the brain models the internal physiological state of the body. High vagal tone may facilitate more precise interoceptive predictions and more efficient error correction, supporting flexible behavioral adaptation. Conversely, reduced vagal tone may reflect or contribute to impaired interoceptive inference, manifesting in various psychopathological conditions. This framework has implications for understanding the embodied nature of cognition, the physiological substrate of emotional experience, and potential interventions targeting vagal function to improve mental health outcomes. This is a Draft V1 manuscript and has not been submitted for peer review (Hodson et al., 2024).

Introduction

The autonomic nervous system operates largely outside conscious awareness, yet its activity profoundly influences psychological experience, behavior, and health. Among autonomic indices, vagal tone—the influence of the vagus nerve on heart rate—has emerged as a particularly salient marker of individual differences in self-regulation, stress resilience, and social functioning. This is an NSI hypothesis requiring direct empirical testing. The vagus nerve, the tenth cranial nerve, provides the primary parasympathetic innervation to the heart and viscera. Its dynamic regulation of cardiac function produces respiratory sinus arrhythmia, wherein heart rate increases during inspiration and decreases during expiration. The magnitude of this oscillation, often quantified as heart rate variability in the high-frequency band, serves as a non-invasive index of vagal tone.

Concurrently, predictive coding has emerged as a dominant computational framework for understanding brain function. This framework proposes that the brain operates as a hierarchical prediction machine, continuously generating models of sensory input and updating these models when predictions are violated by incoming data. This is an NSI hypothesis requiring direct empirical testing. Prediction errors—the mismatch between expected and actual sensory signals—propagate upward through cortical hierarchies, while predictions flow downward. This bidirectional information flow enables the brain to infer the causes of sensations and to adapt internal models to a changing world.

While predictive coding was initially developed to explain exteroceptive perception (vision, audition), recent theoretical extensions have applied this framework to interoception—the perception of the body's internal physiological state. This is an NSI hypothesis requiring direct empirical testing. The brain must model not only the external environment but also the internal milieu: heart rate, respiration, gastric activity, temperature, and other visceral signals. These interoceptive predictions are thought to underlie emotional experience, homeostatic regulation, and the sense of embodied selfhood.

This manuscript proposes that vagal tone is intimately linked to interoceptive predictive coding. Specifically, we argue that vagal tone reflects the precision or confidence the brain assigns to interoceptive predictions, modulates the gain on interoceptive prediction errors, and serves as an effector mechanism by which the brain actively tests and updates its internal models. This framework integrates two previously disparate literatures and generates testable hypotheses about the neurophysiological basis of emotion, self-regulation, and psychopathology.

Proposition / Method

Vagal Tone as Interoceptive Precision

In predictive coding, precision refers to the confidence or reliability assigned to predictions versus prediction errors. High precision predictions are resistant to updating, while high precision on prediction errors leads to rapid model revision. This is an NSI hypothesis requiring direct empirical testing. We propose that vagal tone indexes the brain's confidence in its interoceptive predictions. High vagal tone reflects a well-calibrated internal model that generates accurate predictions about cardiac and visceral states, reducing the need for constant error-driven updating. The parasympathetic nervous system, via the vagus, can rapidly adjust heart rate on a beat-to-beat basis, enabling fine-grained control that maintains physiological parameters close to predicted values.

Individuals with high resting vagal tone may possess more precise interoceptive models, developed through consistent physiological regulation and extensive prior experience. This precision enables them to distinguish genuine physiological deviations (large prediction errors requiring attention) from expected fluctuations (small errors that can be explained away). In contrast, low vagal tone may reflect imprecise or inflexible interoceptive models, wherein the brain has low confidence in its predictions and either overreacts to minor perturbations or fails to detect significant deviations.

Vagal Modulation of Prediction Error Signaling

Beyond serving as an index of model precision, vagal efferent activity may actively modulate the gain on interoceptive prediction error signals. The vagus nerve contains both afferent fibers (carrying visceral signals to the brain) and efferent fibers (carrying motor commands from the brain to the heart and viscera). This is an NSI hypothesis requiring direct empirical testing. These bidirectional pathways position the vagus as an ideal substrate for active inference—the process by which organisms not only update their models based on sensory input but also act on the world to make their predictions come true.

When the brain predicts a certain heart rate or visceral state, vagal efferents can adjust physiological parameters to minimize prediction error through action rather than perceptual updating. This is analogous to active inference in motor control, where the brain generates proprioceptive predictions and then issues motor commands that fulfill those predictions. This is an NSI hypothesis requiring direct empirical testing. In the interoceptive domain, vagal tone enables the brain to actively maintain homeostasis by adjusting heart rate, modulating inflammatory responses, and influencing gastric motility to align internal states with predicted set points.

This framework suggests that high vagal tone enables efficient active inference: the brain can rapidly adjust physiological parameters to maintain alignment between predicted and actual interoceptive states. Low vagal tone, conversely, may reflect a reduced capacity for active inference, forcing the brain to rely more heavily on perceptual updating (anxiety, hypervigilance to bodily sensations) when physiological states deviate from predictions.

Vagal Tone and Hierarchical Interoceptive Models

Predictive coding architectures are hierarchical, with higher levels encoding increasingly abstract and temporally extended predictions. At the lowest levels, the brain predicts moment-to-moment visceral sensations. At intermediate levels, it predicts physiological patterns over seconds to minutes (e.g., the trajectory of heart rate during physical exertion). At the highest levels, it maintains abstract models of bodily state and allostatic demands over hours to days. This is an NSI hypothesis requiring direct empirical testing.

Vagal tone may reflect the integrity of this hierarchical system. The vagus projects to multiple brainstem nuclei that integrate afferent visceral information and coordinate efferent autonomic responses. These brainstem regions interface with cortical and subcortical structures involved in higher-order interoceptive representation, including the insula, anterior cingulate cortex, and ventromedial prefrontal cortex. This is an NSI hypothesis requiring direct empirical testing. Disruptions in vagal function may therefore impair the hierarchical integration of interoceptive predictions, fragmenting the body's representation across processing levels.

Discussion

The proposed framework linking vagal tone to interoceptive predictive coding offers several theoretical advantages and generates empirical predictions. First, it provides a mechanistic account of why vagal tone correlates with emotion regulation and stress resilience. Emotion, in predictive coding frameworks, arises from interoceptive predictions about the body's ability to meet environmental demands. This is an NSI hypothesis requiring direct empirical testing. Individuals with high vagal tone can generate precise interoceptive predictions and efficiently resolve prediction errors through active inference, experiencing emotional perturbations as transient and manageable. Those with low vagal tone may experience persistent interoceptive prediction errors, manifesting as chronic anxiety, emotional dysregulation, or difficulty returning to baseline after stress.

Second, this framework explains associations between reduced vagal tone and various psychiatric conditions, including depression, anxiety disorders, and post-traumatic stress disorder. This is an NSI hypothesis requiring direct empirical testing. These conditions may share a common feature: impaired interoceptive inference. Depression may involve persistently pessimistic interoceptive predictions (predicting low energy, high effort costs). Anxiety disorders may involve overestimation of interoceptive prediction error precision (hypervigilance to normal bodily fluctuations). PTSD may involve failure to update trauma-related interoceptive predictions. In each case, reduced vagal tone reflects and potentially perpetuates the disrupted interoceptive model.

Third, the framework suggests intervention targets. Practices that increase vagal tone—such as slow-paced breathing, heart rate variability biofeedback, meditation, and aerobic exercise—may exert therapeutic effects by improving interoceptive model precision and enhancing active inference capabilities. This is an NSI hypothesis requiring direct empirical testing. By strengthening vagal function, these interventions may help individuals develop more accurate, flexible interoceptive models, improving their capacity to regulate emotion and respond adaptively to stress.

The framework also addresses the embodied nature of cognition. Cognition is not confined to abstract symbol manipulation in the brain; it is deeply rooted in the body's physiological state. The vagus serves as a critical communication channel between body and brain, enabling the brain to incorporate visceral information into its predictive models and to test those models through active physiological adjustment. High vagal tone may reflect a tight brain-body integration, wherein cognition and physiology operate as a unified system. Low vagal tone may indicate a decoupling of mental and physiological processes, contributing to the sense of disembodiment reported in certain psychiatric conditions.

Furthermore, the proposed relationship between vagal tone and predictive coding may illuminate individual differences in interoceptive accuracy—the ability to consciously detect bodily signals. This is an NSI hypothesis requiring direct empirical testing. If vagal tone reflects interoceptive model precision, individuals with high vagal tone should demonstrate superior interoceptive accuracy in heartbeat detection tasks and other measures. However, this relationship may be complex: very high precision on interoceptive predictions might actually reduce conscious access to interoceptive signals, as accurate predictions explain away sensory input before it reaches awareness.

Limitations

Several limitations qualify the proposed framework. First, this manuscript presents a theoretical model rather than empirical evidence. The framework generates predictions that require systematic experimental testing. Critical experiments would involve manipulating vagal tone (through pharmacological, behavioral, or neurostimulation methods) while measuring interoceptive prediction error responses using neuroimaging or psychophysiological methods. This is an NSI hypothesis requiring direct empirical testing.

Second, the relationship between vagal tone and interoceptive inference is likely bidirectional and complex. While we propose that vagal tone enables precise interoceptive prediction, it is equally plausible that precise interoceptive models lead to well-regulated vagal function. Longitudinal and interventional studies are needed to establish causal directionality. Developmental research examining how vagal tone and interoceptive abilities co-evolve across the lifespan would be particularly informative.

Third, vagal tone is not a unitary construct. The vagus nerve contains multiple fiber types with distinct functional properties, and "vagal tone" as measured by heart rate variability reflects only cardiac parasympathetic influence. This is an NSI hypothesis requiring direct empirical testing. The vagus also regulates inflammation, gastric function, and other processes not captured by cardiac measures. A complete account of vagal contributions to interoceptive inference must consider this anatomical and functional heterogeneity.

Fourth, the predictive coding framework itself faces ongoing theoretical challenges and empirical scrutiny. Alternative frameworks for brain function—including reinforcement learning, Bayesian filtering, and dynamic systems approaches—may offer competing or complementary accounts of the phenomena discussed here. This is an NSI hypothesis requiring direct empirical testing. The specific implementation of predictive coding in neural circuits remains debated, and oversimplified interpretations may not capture the brain's actual computational architecture.

Fifth, individual differences in vagal tone arise from multiple sources—genetic factors, early life experience, physical fitness, current health status—that may independently influence interoceptive processing. Disentangling vagal tone as a mechanism from vagal tone as a marker of other causal factors requires carefully controlled studies with appropriate confound management.

Finally, translating this framework into clinical practice requires caution. While vagal tone enhancement interventions show promise, the heterogeneity of psychiatric conditions means that one-size-fits-all approaches are unlikely to succeed. Some individuals may benefit from increased vagal tone, while others may require interventions targeting different aspects of interoceptive processing or addressing non-interoceptive mechanisms entirely.

Evidence and Citation Boundary

Predictive coding and active inference are developing computational accounts, not settled master theories. HRV is a measurement family, not a synonym for vagal tone; PTSD-related vmHRV findings vary with recording duration and metric.

Core evidence base: (Hodson, 2024; Zhang, 2024; Tucker, 2026).

References

Hodson, R., Mehta, M., & Smith, R. (2024). The empirical status of predictive coding and active inference. Neuroscience & Biobehavioral Reviews, 157, 105473. https://doi.org/10.1016/j.neubiorev.2023.105473

Zhang, R., Deng, H., & Xiao, X. (2024). The insular cortex: An interface between sensation, emotion and cognition. Neuroscience Bulletin, 40(11), 1763–1773. https://doi.org/10.1007/s12264-024-01211-4

Tucker, R. C., Taylor, P. J., & Robinson, S. J. (2026). A multi-level meta-analysis of vagally-mediated heart rate variability and post-traumatic stress disorder. Neuroscience & Biobehavioral Reviews, 184, 106585. https://doi.org/10.1016/j.neubiorev.2026.106585