SYNOPSIS. Seven subjects walked on a programmable treadmill both at constant (3.5 ±0.0 and 5.0 ±0.0 km/hr) and oscillating speeds (±0.5, ±1.0, ±1.5, ±2.0 km hr~'), set to sinusoidally change between the two limits in 3 sec. In each condition oxygen consumption measurements were taken. The same experimental protocols were replicated on a walkway by asking subjects to adapt their stride frequency to an audio signal corresponding to the sinusoidal stride frequency changes measured on the treadmill. Differently from what expected, only the ±2.0 km hr~' oscillation resulted to be metabolically different from the constant speed walking, both for the treadmill and the walkway conditions. The time course of the mechanical energy of the body centre of mass could reveal that a strategy devoted to benefit from the usual energy fluctuations occurring at constant speed, is likely to be used to cope with speed varying sequences. From the energy curve observed at constant speed, it is possible to derive an energetically equivalent curve by cumulating acceleration portions, and deceleration ones, of a group of strides as to produce a single acceleration and a single deceleration phase, as it is observed in oscillating speed walking. Being aware of the bias introduced by using a non- , inertia! frame (the treadmill protocol), we are replicating the experiments with a laser beam projected on a wide radius circular path at oscillating speeds, that the subjects have to follow. The preliminary data seem to confirm the invariance of the metabolic requirements in oscillatory walking up to ±1.5 km hr~'.