Knowledge about human brain temperature is still very limited, despite evidence demonstrating the critical influence of mild increases in temperature on the ischaemic brain. It has been suggested that in passive and exercise hyperthermia the brain may be protected against thermal damage by a mechanism of selective brain cooling (SBC). It is said to bring about suppression of the temperature of the brain, rendering it significantly lower than trunk and arterial blood temperature. Yet very little is known about the possible role of this mechanism in fever, a condition fundamentally different from "physiological" hyperthermia, especially when it occurs in brain-damaged patients. In our investigation we retrospectively analysed the results of direct recordings of cerebral temperature within the subdural space (Tsd) and within the brain parenchyma (Tbr-16 cases) in 63 unanaesthetized patients following neurosurgical procedures, including 23 with fever > 38 degrees C. The difference between trunk temperature, measured in the rectum (Tre) or in the oesophagus (Tes), and the intracranial temperature, were calculated in all subjects. A statistically significant reduction of these differences, in step with increasing fever, would be compatible with demonstrating a process of selective brain cooling. The offsets Tre-Tsd, Tre-Tbr, and Tes-Tsd were plotted against Tre over a wide range of body temperature and near zero correlation was found. This finding suggests that brain temperature in fever was not selectively suppressed by any specific thermolytic mechanism and that dissipation of the main bulk of cerebral metabolic heat both in normothermia and in fever depends on heat uptake by arterial blood. The results suggest that the brain in fever can be seriously jeopardized by heat stress and no specific cooling mechanism exists, to reduce it below body temperature in feverish neurosurgical patients. Tbr and/or Tsd remained the highest body temperature in 14 out of the 23 patients during fever.