Fourteen firemen exposed to polychlorinated biphenyls (PCBs) and their byproducts generated in a transformer fire and explosion had neurophysiological and neuropsychological tests 6 mo after the fire. They were re-studied 6 wk later after undergoing 2-3 wk of an experimental detoxification program consisting of medically supervised diet, exercise, and sauna. A case-control comparison with firemen matched from the same department, but who did not participate in controlling the transformer fire, had shown significant impairment of memory for stories, visual images, and digits backwards. Cognitive function was impaired for block design, identifying embedded figures, and design association and recognition using Culture Fair. Making of trails and choice reaction time, which measured cognitive function and perceptual motor speed, were also impaired. These signs of protracted neurobehavioral impairment were attributed to PCBs and heat-produced byproducts. No relationship, however, was found between the firemen's serum or fat levels of PCBs as Arochlor 1248 and their type or degree of neurobehavioral impairment. Retesting following the detoxification program showed significantly improved scores on: three memory tests, block design, trails B, and embedded figures. Thus, there was significant reversibility of impairment after the detoxification interval. However self-appraisal scores for depression, anger, and fatigue--which were initially elevated--and for vigor--which was reduced--did not change across this interval.