For a physiologically realistic range of joint motion and therefore range of muscle fiber lengths, only part of the force-length curve can be used in vivo; i.e., the section of the force-length curve that is expressed can vary. The purpose of this study was to determine the expressed section of the force-length relationship of the gastrocnemius for humans. Fourteen male and fourteen female subjects aged 18-27 performed maximal isometric plantar flexions in a Biodex dynamometer. Plantar flexion moments were recorded at five ankle angles:-15 degrees , 0 degrees , 15 degrees , 30 degrees , and 40 degrees , with negative angles defined as dorsiflexion. These measurements were repeated for four randomly ordered knee angles over two testing sessions 4 to 10 days apart. The algorithm of Herzog and ter Keurs (1988a) was used to reconstruct the force-length curves of the biarticular gastrocnemius. Twenty-four subjects operated over the ascending limb, three operated over the descending limb, and one operated over the plateau region. The variation found suggests that large subject groups should be used to determine the extent of normal in vivo variability in this muscle property. The possible source of the variability is discussed in terms of parameters typically used in muscle models.