D’-tecting the Beat: Refining d’ in Heartbeat Detection Tasks

Cardiac interoception—the ability to perceive and interpret one’s own heartbeat—is increasingly recognized as fundamental to emotion regulation, cognition, and allostasis, as well as a potential transnosologic biomarker. Interoceptive accuracy—performance on objective behavioral tests of heartbeat detection—can be assessed using behavioral sensitivity indices derived from experimental tasks that leverage temporal information, such as d’ and mean distance. Despite the richness of temporal dynamics captured during the measurement of interoception, they are often overlooked in current behavioral indices to assess performance. As such, current analytical approaches are unrefined and insufficiently tailored to capture the full complexity of interoceptive processing, impeding our theoretical and clinical understanding of interoception. For instance, while d’ is theoretically suited for measuring sensitivity, its application is problematic in heartbeat detection tasks (HBT) as it lacks a clear signal-versus-noise distinction. To overcome this limitation, we refined d’ for HBT by 1) modifying how signal detection theory (SDT) outcomes—used to calculate d’—are extracted, and 2) implementing a pooled z-score approach to avoid dubious statistically assumptions. Thirty-three healthy participants performed a HBT, consisting of two exteroceptive blocks (tapping to a recorded heartbeat sound) followed by two interoceptive blocks (tapping to their own heartbeat), with concurrent electrocardiogram heartbeat recordings. We calculated d’ using the original design which uses temporal windows for SDT outcome extraction—, a window free design, and the window free design with pooled z-scoring, alongside mean distance—a response frequency-based interoceptive accuracy measure. Compared to the standard approach, our revised metric exhibited a more balanced influence of SDT outcomes on the pooled z-score d’ (p less than 0.001, d greaterthan 1), sensitivity differences across the three d’ approaches (p less than 0.001), stronger correlations between d’ and mean distance (p less than 0.001), and a more robust, assumption-light framework for interoceptive accuracy assessment. This refined d’ enhances the precision of accuracy measurements in future interoception studies—enabling more sensitive detection of sensory processing differences in clinical populations and advances the methodological rigor of cardiac interoception research. Furthermore, new metrics combining both d’ and mean distance are now possible and should be pursued, given their shared underlying sensitivity factor.