It is generally assumed that exercise and shivering are analogous processes with regard to substrate utilisation and that, as a consequence, exercise can be used as a model for shivering. In the present study, substrate utilisation during exercise and shivering at the same oxygen consumption (VO2) were compared. Following an overnight fast, eight male subjects undertook a 2-h immersion in cold water, designed to evoke three different intensities of shivering. At least 1 week later they undertook a 2-h period of bicycle ergometry during which the exercise intensity was varied to match the VO2 recorded during shivering. During both activities hepatic glucose output (HGO), the rate of glucose utilisation (Rd), blood glucose, plasma insulin, free fatty acid (FFA) and beta-hydroxybutyrate (B-HBA) concentrations were measured. The VO2 measured during the different levels of shivering averaged 0.49 l.min-1 (level 1: low), 0.6 l.min-1 (level 2: low-moderate), and 0.9 l.min-1 (level 3: moderate), and corresponded closely to the levels measured during exercise. HGO and Rd were greater (P < 0.05) during exercise than during shivering at the same VO2 (9.5% and 14.7%, respectively). The average (SD). HGO during level 3 exercise was 3.0 (0.91) mg.kg-1.min-1 compared to 2.76 (1.0) mg.kg-1.min-1 during shivering. The values for Rd were 3.06 (0.98) mg.kg-1.min-1 during level 3 exercise and 2.68 (0.82) mg.kg-1.min-1 during shivering. Blood glucose levels did not differ between conditions averaging 5.4 (0.3) mmol.l-1 over all levels of shivering and 5.2 (0.3) mmol.l-1 during exercise. Plasma FFA and B-HBA were higher (P < 0.01) during shivering than during corresponding exercise (12.3% and 33.3%, respectively). FFA averaged 0.61 (0.2) mmol.l-1 over all levels of shivering and 0.47 (0.16) mmol.l-1 during exercise. The figures for L-HBA were 0.44 (0.13) mmol. l-1 during all levels of shivering and 0.32 (0.1) mmol.l-1 during exercise. Plasma insulin was higher (P < 0.05) during level 2 and 3 shivering compared to corresponding exercise; at these levels the average value for plasma insulin was 95.9 (21.9) pmol.l-1 during shivering and 80.6 (16.1) pmol.l-1 during exercise. On the basis of the present findings it is concluded that, with regard to substrate utilisation, shivering and exercise of up to 2 h duration should not be regarded as analogous processes.