Limited information is available regarding the physiological responses to prolonged load carriage. This study determined the energy cost of prolonged treadmill walking (fixed distance of 12 km) at speeds of 1.10 m.s-1, 1.35 m.s-1, and 1.60 m.s-1, unloaded (clothing mass 5.2 kg) and with external loads of 31.5 and 49.4 kg. Fifteen male subjects performed nine trials in random order over a 6-week period. Oxygen uptake (VO2) was determined at the end of the first 10 min and every 20 min thereafter. A 10-min rest period was allowed following each 50 min of walking. No changes occurred in VO2 over time in the unloaded condition at any speed. The 31.5 and 49.4 kg loads, however, produced significant increases (ranging from 10 to 18%) at the two fastest and at all three speeds, respectively, even at initial exercise intensities less than 30% VO2max. In addition, the 49.4 kg load elicited a significantly higher (P less than 0.05) VO2 than did the 31.5 kg load at all speeds. The measured values of metabolic cost were also compared to those predicted using the formula of Pandolf et al. In trials where VO2 increased significantly over time, predicted values underestimated the actual metabolic cost during the final minute by 10-16%. It is concluded that energy cost during prolonged load carriage is not constant but increases significantly over time even at low relative exercise intensities. It is further concluded that applying the prediction model which estimates energy expenditure from short-term load carriage efforts to prolonged load carriage can result in significant underestimations of the actual energy cost.