Prediction formulae of shivering metabolism (Mshiv) are critical to the development of models of thermoregulation for cold exposure, especially when the extrapolation of survival times is required. Many such formulae, however, have been calibrated with data that are limited in their range of core temperatures (Tc), seldom involving values of less than 36 degrees C. Certain recent studies of cold-water immersion have reported Tc as low as 33.25 degrees C. These data comprise measurements of Tc (esophageal) and mean skin temperature (Ts), and metabolism from 14 males [mean (SD); age = 28 (5) years; height = 1.78 (0.06) m; body mass = 77.7 (6.9) kg; body fat (BF) = 18.4 (4.5)%] during immersion in water as cold as 8 degrees C for up to 1 h and subsequent self-rewarming via shivering under dry blanketed conditions. The data contain 3343 observations with mean (SD) Tc and Ts of 35.92 (0.93) degrees C and 23.4 (8.9) degrees C, respectively, and have been used to re-examine the prediction of Mshiv. Rates of changes of these temperatures were not used in the analysis. The best fit of the formulae, which are essentially algebraic constructs with and without setpoints, are those with a quadratic expression involving Ts. This is consistent with the findings of Benzinger (1969) who demonstrated that the thermosensitivity of skin is parabolic downwards with temperature peaking near a value of 20 degrees C. Formulae that included a multiplicative interaction term between Tc and Ts did not predict as well. The best prediction using 37 degrees C and 33 degrees C as the Tc and Ts setpoints, respectively, was found with BF as an attenuation factor: Mshiv (W x m(-2)) = [155.5 x (37- Tc) + 47.0 x (33 - Ts) - 1.57 x (33 - Ts)2]/(%BF)(0.5).