Interoception, the perception of internal bodily states, has been increasingly recognized as an important construct in substance use behavior, and research in this area has grown in recent years. However, empirical validation of key theoretical propos...
Interoception, the perception of internal bodily states, has been increasingly recognized as an important construct in substance use behavior, and research in this area has grown in recent years. However, empirical validation of key theoretical proposals remain limited, and existing findings are difficult to integrate due to inconsistencies across studies and methodological constraints. This thesis addresses these gaps by providing a systematic review of interoception in substance use disorders (SUDs) and an empirical investigation in smokers that was designed to overcome the limitations identified in the review.
This thesis is structured into three chapters. Chapter 1 presents a systematic review that synthesizes 39 empirical studies examining interoceptive dimensions, including accuracy, sensibility, and awareness, across individuals with SUDs and related substance use conditions. The review found reduced interoceptive accuracy in individuals with SUDs, with this impairment being most consistent in alcohol use disorder and less consistent across other substance use populations. In contrast, findings for interoceptive sensibility were highly heterogeneous, and studies of interoceptive awareness were notably scarce. The review also identified key methodological limitations, including an emphasis on only two interoceptive dimensions (accuracy and sensibility), overreliance on cardiac measures at the expense of other modalities, widespread use of the heartbeat counting task despite well-documented validity concerns, and inadequate reporting and control of smoking status as a potential confound.
To address these limitations, Chapter 2 reports an empirical study comparing cardiac and respiratory interoception between smokers and nonsmokers, while also examining smokers under both nicotine-satiated and deprived states. Smokers (n = 55) and nonsmokers (n = 55) completed three interoceptive tasks, including the heartbeat tapping task, heart rate discrimination task, and respiratory resistance sensitivity task, along with self-report questionnaires. To move beyond accuracy-based summaries and infer latent interoceptive mechanisms, Bayesian computational modeling was used to estimate interoceptive precision and prior beliefs, psychometric parameters (slope, threshold), signal detection indices (d′, c), and metacognitive sensitivity and efficiency (meta-d′, M-ratio). In the heartbeat tapping task, deprived smokers exhibited higher prior beliefs during guess trials (p = 0.004) and lower interoceptive precision during breath-hold trials (p = 0.004), indicating altered top-down expectations alongside impaired dynamic precision modulation. In the heart rate discrimination task, deprived smokers showed a perceptual bias toward perceiving faster heart rates (p = 0.018), while satiated smokers demonstrated reduced metacognitive sensitivity (p = 0.01), suggesting that smoking affects both low-level perceptual processes and high-level metacognitive monitoring of cardiac interoception. In contrast, respiratory interoception showed no significant differences between groups or smoking states, demonstrating robust modality-specificity of nicotine's effects on interoceptive processing. In contrast, respiratory interoception did not differ by group or state, providing evidence for modality-specificity of nicotine's effects on interoception.
Chapter 3 discusses these findings within predictive coding and active inference frameworks, outlines limitations, and considers clinical implications Predictive coding conceptualizes interoception as precision-weighted integration of visceral signals and predictions; nicotine deprivation was characterized by a failure to dynamically upweight sensory precision, resembling a computational signature reported across chronic substance use. Active inference extends this account by proposing that interoceptive predictions function as allostatic setpoints that motivate regulatory action. Consistent with this view, impaired interoceptive precision predicted withdrawal severity, and anticipatory tapping patterns predicted time to first cigarette, linking perceptual inference to motivated substance seeking behavior. Moreover, the modality-specific pattern challenges unitary models of interoception and aligns with hierarchical, multidimensional accounts in which different bodily axes rely on partially distinct mechanisms. Clinically, the results suggest a hierarchical dual-pathology model in which deprivation is associated with low-level perceptual bias, whereas satiation is associated with impaired metacognitive insight, implying that interventions may need to target both levels. Overall, this thesis demonstrates that interoceptive alterations in smokers are complex, hierarchical, modality-specific, and state-dependent, underscoring the need for multidimensional assessment grounded in computational theory.