The conventional method used to estimate the change in mean body temperature (dMBT) is by taking X% of a body core temperature and (1−X)% of weighted mean skin temperature, the value of X being dependent upon ambient temperature. This technique is used widely, despite opposition from calorimetrists. In the present paper we attempt to provide a better method. Minute-by-minute changes in dMBT, as assessed using calorimetry, and 21 (20 if esophageal temperature was unavailable) various regional temperatures (dRBTs), as assessed using thermometry, including 6 subcutaneous measures, were collected from 7 young male adults at 6 calorimeter temperatures. Since a calorimeter measures only changes in heat storage, which can be converted to dMBT, all body temperatures are expressed as changes from the reasonably constant pre-exposure temperatures. The following three aspects were investigated. (1) The prediction of dMBT from the 21 (or 20) dRBTs with multi-linear regression analysis (MLR). This yields two results, model A with rectal temperature (dTre) alone, and model B with dTre and esophageal temperature (dTes). (2) The prediction of dMBT from dTre with or without dTes and 13 skin surface temperatures combined to one weighted mean skin temperature (dT¯sk), using MLR. This results in models C and D. Six more models (E–J) were added, representing the above two sets in various combinations with four factors. (3) The conventional method calculated with four values for X. Model A predicted better than 0.3 °C in 70% of the cases. Model I was the best amongst the models with 13 weighted skin temperatures (better than 0.3 °C in 60% of the cases). The conventional method was erratic.