Maintaining balance following gait perturbations is difficult and still not well addressed in gait assistive devices. A challenge is in identifying perturbations, and whether and which responses are required to reestablish balance and walking. Here, we investigate the timing of changes in the kinematic and muscle activation patterns of unimpaired subjects to external perturbations. We used the ETH Knee Perturbator to lock the knee at different points of swing phase, and identified changes in the gait pattern with Statistical Parametric Mapping, adjusted for data containing perturbations. We show that kinematic patterns differ within approximately 100 ms of the perturbation, and that muscle activity changes later, much closer to foot-strike. Our results suggest that mechanical (joint angles and velocities) sensors are best suited to identify external perturbations, devices should change their behavior in response to such perturbations, and responses may not need to be initiated immediately following the perturbation.