Resting-state EEG dissociates heartbeat counting from heart-rate discrimination: Cortical signal entropy tracks counting accuracy, default-mode connectivity tracks exteroceptive discrimination.

Bidirectional communication between the heart and the brain shapes emotion, self-regulation, and adaptive behaviour, yet how the brain's intrinsic organisation supports cardiac interoception remains unclear. Two assays dominate the literature - heartbeat counting (HBC) and heart-rate discrimination (HRD) - but whether they engage separable resting-brain networks at high spatial density is untested. We analysed 256-channel resting EEG from 47 healthy adults (5 min eyes-open, 5 min eyes-closed) with the Schandry HBC task (Schandry, 1981; Garfinkel et al., 2015) and the psychophysical HRD task (Legrand et al., 2022), which includes a tone-versus-tone exteroceptive control block. Source-space connectivity was evaluated with the network-based statistic (NBS), complemented by modified multi-scale sample entropy. HBC accuracy was unrelated to HRD sensitivity in either modality (d', slope), though it covaried with the HRD interoceptive threshold, which indexes bias rather than precision (ρ= +0.41, p = .004). Entropy yielded the most robust association: higher eyes-closed source-space sample entropy co-occurred with better HBC accuracy across a broad multi-network cortical mantle (50 of 100 parcels at q<0.05); no other behaviour produced an FDR-significant parcel. This entropy-accuracy association survived adjustment for alpha power and the aperiodic (1/f) exponent, and all headline associations were unchanged after controlling pulse-derived heart rate and heart-rate variability and after excluding reversed responders identified by influence diagnostics. NBS returned four family-wise-significant subnetworks. Three were anchored on default-mode parcels but tracked d' in the exteroceptive rather than the interoceptive block; the fourth, a disjoint theta-band dorsal-attention-somatosensory cluster, tracked HBC awareness. Because 128 NBS families were screened and the reported clusters re-test a pre-selected subset of them, this yield does not clearly exceed chance expectation; the connectivity results are therefore reported as hypothesis-generating. The findings support a behavioural and neural dissociation between HBC and HRD and caution against pooling them as interchangeable indices of interoceptive accuracy.