Previous investigations have reported that accelerometer counts plateau during running at increasingly faster speeds. OBJECTIVE To assess whether biomechanical and/or device limitations cause this phenomenon and the feasibility of generating oxygen uptake (.VO2) prediction equations from the combined use of accelerometry and heart rate during walking and running. METHODS : Sixteen endurance-trained subjects completed two exercise tests on a treadmill. The first was a continuous incremental test to volitional exhaustion to determine ventilatory threshold and peak .VO2. The second was a discontinuous incremental exercise test while walking (3, 5, and 7 km.h(-1)) and running (8, 10, 12, 14, 16, 18, and 20 km.h(-1), or until volitional exhaustion). Subjects completed 3 min of exercise at each speed, followed by 3-5 min of recovery. Activity counts from uni- and triaxial accelerometers, heart rate, and gas exchange were measured throughout exercise. RESULTS All accelerometer outputs rose linearly with speed during walking. During running, uniaxial accelerometer outputs plateaued, whereas triaxial output rose linearly with speed up to and including 20 km.h(-1). Prediction of .VO2 during walking and running using heart rate (R2 = 0.42 and 0.59, respectively), accelerometer counts (R2 = 0.48-0.83 and 0.76, respectively), the combined methodologies (R2 = 0.54-0.85 and 0.80, respectively), and the combined methodologies calibrated with individual data (R2 = 0.99-1.00 and 0.99, respectively) was completed by linear regression. CONCLUSIONS Uni- and triaxial accelerometer outputs have a linear relationship with speed during walking. During running, uniaxial accelerometer outputs plateau because of the biomechanics of running, whereas triaxial accelerometer output has a linear relationship. The combined methodologies predict .VO2 better than either predictor alone; a subject's individually calibrated data further improves .VO2 estimation.