Key points The human gastrocnemius muscle is slack at its shortest lengths in vivo. We hypothesised that when the muscle is lengthened slack is taken up progressively, first in some muscle fascicles, then in others. A new, quasi‐three‐dimensional ultrasound imaging method was used to test this hypothesis directly. Contrary to our prediction, the slack was taken up nearly simultaneously in muscle fascicles located throughout the gastrocnemius muscle. Dynamic three‐dimensional reconstructions indicated that, over the range of lengths that was investigated, passive changes in the length of gastrocnemius muscle–tendon units are due to changes in the length of both muscle fascicles and tendinous structures, in approximately equal measure. Changes in the alignment of muscle fascicles and the curvature of aponeuroses have a negligible contribution to change in muscle length. These findings give insights into how skeletal muscles change their lengths in vivo. AbstractThe mechanisms by which skeletal muscles lengthen and shorten are potentially complex. When the relaxed human gastrocnemius muscle is at its shortest in vivo lengths it falls slack (i.e. it does not exert any passive tension). It has been hypothesised that when the muscle is passively lengthened, slack is progressively taken up, first in some muscle fascicles then in others. Two‐dimensional imaging methods suggest that, once the slack is taken up, changes in muscle length are mediated primarily by changes in the lengths of the tendinous components of the muscle. The aims of this study were to test the hypothesis that there is progressive engagement of relaxed muscle fascicles, and to quantify changes in the length and three‐dimensional orientation of muscle fascicles and tendinous structures during passive changes in muscle length. Ultrasound imaging was used to determine the location, in an ultrasound image plane, of the proximal and distal ends of muscle fascicles at 14 sites in the human gastrocnemius muscle as the ankle was rotated passively through its full range. A three‐dimensional motion analysis system recorded the location and orientation of the ultrasound image plane and the leg. These data were used to generate dynamic three‐dimensional reconstructions of the architecture of the muscle fascicles and aponeuroses. There was considerable variability in the measured muscle lengths at which the slack was taken up in individual muscle fascicles. However, that variability was not much greater than the error associated with the measurement procedure. An analysis of these data which took into account the possible correlations between errors showed that, contrary to our earlier hypothesis, muscle fascicles are not progressively engaged during passive lengthening of the human gastrocnemius. Instead, the slack is taken up nearly simultaneously in all muscle fascicles. Once the muscle is lengthened sufficiently to take up the slack, about half of the subsequent increase in muscle length is due to elongation of the tendinous structures and half is due to elongation of muscle fascicles, at least over the range of muscle–tendon lengths that was investigated (up to ∼60 or 70% of the range of in vivo lengths). Changes in the alignment of muscle fascicles and flattening of aponeuroses contribute little to the total change in muscle length.