Offset analgesia (OA) represents a disproportionately large decrease of pain perception after a brief, temporary increment of thermal pain stimulus and was reported attenuated in patients with neuropathic pain. We examined whether OA depends on the increment duration before offset, and whether individual features of OA distinguish patients with chronic pain and healthy controls. We used a Peltier-type thermal stimulator and OA paradigms including 5-, 10-, or 15-s duration of 1°C-increment (T2) over 45°C. We first examined OA response, on the left volar forearm, at 3 different T2's in 40 healthy volunteers, and OA and constant stimulus responses in 12 patients with chronic pain and 12 matched healthy controls. We measured magnitude of OA ([INCREMENT]OA) and maximum visual analogue scale (VAS) latency (time to peak VAS) during constant stimulus for each individual. Pain perception kinetics were compared with analysis of variance and sought for correlations with psychophysical parameters with a significance threshold at P < 0.05. In healthy controls, longer T2 at 10 or 15 seconds resulted in larger [INCREMENT]OA compared with T2 at 5 seconds (P = 0.04). In patients, [INCREMENT]OA was significantly smaller than controls at T2 = 5 or 10 seconds (P < 0.05) but grew comparable at T2 = 15 seconds with controls. Maximum VAS latency was longer in patients than in controls and negatively correlated with [INCREMENT]OA in patients. An OA index ([INCREMENT]OA/[maximum VAS latency]) proved diagnostic of chronic pain with an area under the receiver operating characteristic curve at 0.897. Patients with chronic pain showed impairment of OA and reduced temporal sharpening of pain perception, which might imply possible disturbance of the endogenous pain modulatory system.