Does the Treadmill Support Valid Energetics Estimates of Field Locomotion?

Quantifying animal energy expenditure during locomotion in the field is generally based either on treadmill measurements or on estimates derived from a measured proxy. Two common proxies are heart rate (ƒH) and dynamic body acceleration (accelerometry). Both ƒH and accelerometry have been calibrated extensively under laboratory conditions, which typically involve prompting the animal to locomote on a treadmill at different speeds while simultaneously recording its rate of oxygen uptake (V̇o2) and the proxy. Field estimates of V̇o2 during locomotion obtained directly from treadmill running or from treadmill-calibrated proxies make assumptions about similarities between running in the field and in the laboratory. The present study investigated these assumptions, focusing on humans as a tractable species. First we investigated experimentally if and how the rate of energy expenditure during treadmill locomotion differs to that during field locomotion at the same speeds, with participants walking and running on a treadmill, on tarmac, and on grass, while wearing a mobile respirometry system. V̇o2 was substantially higher during locomotion in both of the field conditions compared with on a level treadmill: 9.1% on tarmac and 17.7% on grass. Second, we included these data in a meta-analysis of previous, related studies. The results were influenced by the studies excluded due to particulars of the experiment design, suggesting that participant age, the surface type, and the degree of turning during field locomotion may influence by how much treadmill and field locomotion V̇o2 differ. Third, based on our experiments described earlier, we investigated the accuracy of treadmill-calibrated accelerometry and ƒH for estimating V̇o2 in the field. The mean algebraic estimate errors varied between 10% and 35%, with the ƒH associated errors being larger than those derived from accelerometry. The mean algebraic errors were all underestimates of field V̇o2, by around 10% for fH and varying between 0% and 15% for accelerometry. Researchers should question and consider how accurately a treadmill-derived proxy calibration of V̇o2 will estimate V̇o2 during terrestrial locomotion in free-living animals.