In hyperthermic subjects Deklunder et al. (1991) have found "no experimental evidence that an efficient thermal counter-current heat exchange occurred in the human brain". If that conclusion is true, it should be validated by both temperature measurements and Doppler flow techniques, but the experiments described seem deficient in several ways, namely: 1. In their experiment, the temperature gradient between the trunk and the intracranium has been investigated by the measurement of rectal temperature and tympanic temperature (Try). It is well known that rectal temperature, because of its time lag behind central blood temperature, is a poor index of trunk temperature during thermal transients (Eichna et al. 1951). This time lag would not have been seen if oesophageal temperature (Toes) had been measured instead of rectal temperature. The time lag has been found to be negligible for Try when the latter is measured accurately (Brinnel and Cabanac 1989). Accuracy cannot be presumed in the study of Deklunder et al. (1991) since Try was considerably lower than rectal temperature (mean, 1.03 ° C) at the beginning of the hyperthermia sessions. When measured accurately, Tty has been shown to be higher than trunk temperature in normothermic conditions (Brinnel and Cabanac 1989). Therefore, the use of rectal temperature, and the poor accuracy of the Try measuring technique did not offer to these authors a valid thermal argument against the existence of selective brain cooling in humans. 2. Deklunder et al. (1991) used Doppler signals to explore the potential influence of cool blood inflow through the angularis vein (AV) on Tty. In the Discussion (p. 347, left hand column, line 17), they quote theoretical calculations by Wenger (1987) stating that "the heat transfer required to decrease brain temperature by 1 ° C implies an arterio-venous temperature difference of 119°C '' in the cavernous sinus. Nagasaka et al. (1990) have shown that the compression of both AV (right and left) for 8 min, in conditions of mild hyperthermia, thereby mechanically obstructing blood flow through AV and partially inhibiting counter-current heat exchange in the cavernous sinus, could lower the Toes-Try gradient by 0.12°C from 0.44 (SEM 0.12) °C to 0.32 (SEM 0.06) ° C (means, n = 8). In the same subjects, mechanical compression of the alinasal facial veins increased the Toes-Try gradient by 0.06 ° C and Doppler recordings of AV during facial vein compression confirmed that, during mild hyperthermia, emissary venous blood flow through AV is not at its maximum. The results of Nagasaka et al. (1990), showing that mere mechanical manipulation of blood flow through AV influence intracranial temperature by as much as 0.18°C during mild hyperthermia, have clearly demonstrated the existence of selective brain cooling through AV, and they remind us that sound experiments are better than theoretical calculations. In conclusion, Deklunder et al. (1991) have confirmed the dramatic increase in inward bound AV blood velocity during hyperthermia, the sine qua non condition of selective brain cooling, but they fall short of drawing the rational conclusion from their results. Instead, they have cited the absence of a reversal in the direction of blood flow as a reason for drawing the opposite conclusion. Yet the reversal of blood flow as a function of core temperature has never been a condition for our hypothesis which cannot, therefore, be rejected by such a weak argument as this. Unfortunately, strong arguments, such as those developed above, have not been discussed in Deklunder et al. (1991).