The aim of the study was to examine reproducibility of primary and secondary hyperalgesia in a psychophysical model of human inflammatory pain. Mild burns were produced on the crura of 12 volunteers with a 50 x 25 mm thermode (47 degrees C, 7 min). Assessments of (i) cold and warm detection thresholds, (ii) mechanical and heat pain thresholds, (iii) pain to heat (43 degrees C and 45 degrees C, 5 s), (iv) secondary hyperalgesia, and (v) skin erythema were made 1.75 and 0.5 h before, and 0, 1, 2, 4, and 6 h after a burn injury. Sensory thresholds and hyperalgesia to heat and mechanical stimuli were examined by contact thermodes and von Frey hairs, and pain intensity was rated with a visual analog scale (0-100). To describe between-day reproducibility, the subjects were examined three times at intervals of 21 days. Within-day comparisons showed that a 20% change could be detected as significant for all variables with fewer than 12 subjects in a cross-over design (2alpha = 5% and power = 80%). Between-day comparisons demanded up to 25 subjects to detect changes of the same magnitude. The burns caused mild to moderate pain (VAS: mean 29, SD 14) and the subjects (all right-handed) were more sensitive to heat pain on their left side (P < 0.03). Hyperalgesia was induced instantaneously by the burn and outlasted the study period (6 h). However, no spontaneous pain was observed after the injury, and a brief period of hypoesthesia to warm and cold stimuli was induced by the burn. The painful measurements themselves evoked hyperalgesia to heat and mechanical stimuli on the arm, but only to mechanical stimuli on the legs. including secondary hyperalgesia. Hyperalgesia evoked by the measurements was significantly less intense than that induced by injury. Habituation to the painful stimuli was demonstrated by significantly higher pain thresholds and lower pain responses on the second and third day of the study. The burn model is a sensitive psychophysical model of acute inflammatory pain, when cross-over designs and within-day comparisons are used, and the model is suitable for double-blind, placebo-controlled studies of analgesics. In similar models, we recommend that analgesic and placebo are evenly divided between right and left sides and study days.