The aim of this study was to: (1) determine directly the energy cost of shuttle running (C Sh) and (2) compare it to the values indirectly estimated from kinematic data. C Sh over distances of ≈10 or ≈20 m was determined on 65 subjects (group 1) from gas exchange measurements over 155 trials, or indirectly estimated on 10 subjects (group 2) from the time course of the speed as follows. (1) The cost to accelerate from zero to peak speed was estimated assuming a 25 % efficiency and added to that of constant-speed running, as obtained on subjects of group 1. (2) Since (i) accelerated running on flat terrain is equivalent to running uphill at constant speed, on a slope dictated by the forward acceleration (di Prampero et al. in J Exp Biol 208:2809-2816, 2005), and (ii) the energy cost of running uphill is known, C Sh was obtained from the time course of the acceleration. C Sh increased with the average speed, at any given speed being significantly greater for the shorter distances; e.g., at ≈4 m/s over 10 m, it amounted to ≈14 J/(kg m), i.e., 3.5-fold larger than that at constant speed. The two indirect methods yielded results not significantly different from C Sh over the longer (≈20 m), but underestimated it over the shorter (≈10 m) distances. From our results, over ≈20 m C Sh can be obtained with sufficient accuracy from actual measurements of peak speed alone, thus, greatly simplifying the experimental procedure. The so-obtained data can then be utilized to assess the athletic status of any subject, as well as to plan appropriate training strategies.