Prediction error coding as the computational basis for nocifensive and nocifensive-like behaviors

Nocifensive behavior (NB) is a protective response to noxious stimuli that threaten tissue damage. However, similar motor responses, termed nocifensive-like behavior (NLB), can be evoked by unexpected innocuous stimuli. This observation challenges strict "labeled-line" models of pain, raising a fundamental question: how does the nervous system discriminate true threats from false alarms? We review evidence suggesting NB and NLB exist on a shared behavioral continuum, where defensive responses aren't determined solely by sensory input but by the brain's integrated threat assessment. This assessment computes the probability of harm by weighing somatosensory input against contextual factors like prior experience and multisensory cues. We propose this process is governed by a threat prediction error (TPE) mechanism, which is computationally analogous to the reward prediction error (RPE) mechanism encoded by the dopaminergic system. Under this framework, defensive responses are scaled to the magnitude of the TPE - the discrepancy between expected and actual sensory outcomes. Critically, this means the surprise of a benign touch in a dangerous environment can produce a larger TPE - and a stronger withdrawal - than the anticipation of a noxious pinprick in a safe environment. Furthermore, while NLB represents an adaptive response that can be permanently resolved as the stimulus is learned to be non-threatening, NB represents an innate response, permitting only transient suppression due to the real risk of injury. This model positions defensive behaviors as dynamic perceptual decisions arising from probabilistic inference, offering a unified theory for how context and expectation gate the expression of protective motor programs.