New Bayesian model identifies three distinct failure modes in pain location diagnosis

Adam Y Shavit's new paper on arXiv (2607.26297) argues that the diagnostic utility of patient-reported pain location is not a single gradient but a combination of three epistemically distinct failures. Anatomical multiplexing occurs when many structures share the same anatomical location—a non-identifiable inverse problem. Delocalized amplification, clinically known as central sensitization or nociplastic pain, represents a change in the generative model where centrally driven pain behavior replaces the peripheral source. Referred or atypical displacement describes a systematic, person-dependent shift in pain location, as seen in referred pain and atypical presentations—a covariate-dependent bias.

The research unifies these failures under one Bayesian inference framework, where failures occur at different nodes: the likelihood, model class, prior, and loss. The paper's main formal contribution is a spatial Bayesian model of 'where pain is said to be' distinct from 'where it is felt.' Observation over time increases recoverable information. Additionally, Shavit re-examines published accuracy data and finds that the reported 'high-utility' band relies on overstated specificity, meaning the real gradient is flatter than commonly drawn. This synthesis sits on a 'why-location-fails' axis, distinct from the existing nociceptive/neuropathic/nociplastic taxonomy, and offers a more precise diagnostic approach for clinicians.