In evaluating energy cost (EC) of walking, referred to as walking efficiency, the use of net measurement protocols (i.e. net=grossβresting) has recently been recommended. However, nothing is known about the comparative reproducibility of net protocols and the commonly used gross protocols. Ten minutes of resting and 5 minutes of walking at a selfβselected speed were used to determine gross and net EC in 13 children with spastic cerebral palsy (CP; seven males, six females; mean age 8y 7mo [SD 3y 4mo], range 4y 1moβ13y) and in 10 children (three males, seven females) with typical development. In the former, their Gross Motor Function Classification System levels ranged from Level I to Level III; and seven had hemiplegia and six diplegia. There were four repeated sessions on different days, with periods of 1 week between sessions. Reproducibility was assessed for speed, and gross and net EC, by using the standard error of measurement. The results of this preliminary study showed that EC measurements were more variable for children with CP than for children with typical development. Furthermore, in both groups there was considerably more variability in the net measurements than in the gross measurements. We conclude that, on the basis of the methodology used, the use of gross EC, rather than net EC, seems a more sensitive measure of walking efficiency to detect clinically relevant changes in an individual child with CP.