Intuition as Metacognitive Dissociation: A Subcortical Weak-Signal Integration Framework

Intuitive judgments feature high subjective confidence despite inability to articulate evidential basis—a dissociation between confidence and accessible justification that challenges existing accounts of rapid decision-making (19,33,39). We propose intuition constitutes a metacognitive phenotype arising from preconscious integration of weak interoceptive and contextual signals within subcortical-cortical circuits centered on the amygdala-insula-vmPFC axis (8,10,37,38).We present empirical validation using heartbeat detection data (N=234) from an openly available dataset (MacCormack et al., 2024; OSF: osf.io/z7c2a) (7,24). Results demonstrate significant dissociation between objective performance (M=0.597, SD=0.124) and subjective confidence (M=0.522, SD=0.168), with systematic underconfidence (M=+0.075, t(233)=+6.65, p<.001, d=0.43). Critically, 67.9% exhibited performance exceeding confidence, confirming impaired metacognitive access despite intact discrimination (11,12). Weak performance-confidence correlation (r=.326, R²=10.6%) and independence from conscious interoceptive awareness (MAIA: r=−.062, p=.346; BAQ: r=−.098, p=.134) support subcortical preconscious mechanisms (10,37,38). Robustness analyses—quality filtering (d=0.45 after outlier removal), latent awareness factor extraction (r=-.072, p=.27), TOST equivalence testing (p<.05), Bayesian analysis (BF₀₁≈4.2), and null simulation (p<.000001)—confirm findings are not artifacts of measurement choice, outliers, or sampling noise.Convergent evidence from OpenNeuro intracranial EEG (ds003374) (20) reveals amygdala dynamics consistent with transient integration: block-level decoding at chance (40%, p=.76) but time-resolved analysis shows early gamma-driven peak (4–8s post-stimulus), indicating phase-specific processing incompatible with sustained categorical representations. HCP-derived fusiform face area maps (27,51,52) demonstrate categorical information can be focally organized, constraining 'diffuse storage' alternatives.We introduce an operational three-component dissociation protocol and modular five-test empirical roadmap yielding discriminating predictions: temporal precedence (subcortical→cortical), interoceptive timing effects (systole→threat bias), content-specific integration versus nonspecific gain modulation, lesion double-dissociations, and pharmacological specificity. The framework is falsifiable: null results in any of five critical tests would necessitate revision. While neural mechanisms remain inferential (no causal manipulation, lesion data, or purpose-built fMRI), the demonstrated behavioral phenotype, robust independence from awareness, convergent neural dynamics, and pre-specified falsification criteria establish a testable foundation for mechanistic investigation.