Most falls occur after a loss of balance following an unexpected perturbation such as a slip or a trip. Greater understanding of how humans control and maintain stability during perturbed walking may help to develop appropriate fall prevention programs. The aim of this study was to examine changes in spatiotemporal gait and stability parameters in response to sudden mechanical perturbations in medio-lateral (ML) and anterior-posterior (AP) direction during treadmill walking. Moreover, we aimed to evaluate which parameters are most representative to quantify postural recovery responses. Ten healthy adults (mean = 26.4, SD = 4.1 years) walked on a treadmill that provided unexpected discrete ML and AP surface horizontal perturbations. Participants walked under no perturbation (normal walking), and under left, right, forward, and backward sudden mechanical perturbation conditions. Gait parameters were computed including stride length (SL), step width (SW), and cadence, as well as dynamic stability in AP- (MoS-AP) and ML- (MoS-ML) directions. Gait and stability parameters were quantified by means, variability, and extreme values. Overall, participants walked with a shorter stride length, a wider step width, and a higher cadence during perturbed walking, but despite this, the effect of perturbations on means of SW and MoS-ML was not statistically significant. These effects were found to be significantly greater when the perturbations were applied toward the ML-direction. Variabilities, as well as extremes of gait-related parameters, showed strong responses to the perturbations. The higher variability as a response to perturbations might be an indicator of instability and fall risk, on the same note, an adaptation strategy and beneficial to recover balance. Parameters identified in this study may represent useful indicators of locomotor adaptation to successfully compensate sudden mechanical perturbation during walking. The potential association of the extracted parameters with fall risk needs to be determined in fall-prone populations.