Whereas several animal studies have indicated the important role of the motor cortex in the control of voluntary gait modifications, little is known about the effects of cortical lesions on gait adaptability in humans. Obstacle avoidance tasks provide an adequate paradigm to study the adaptability of the stepping pattern under controlled, experimental conditions. In the present study, an exploratory assessment was made of the failure rate, the preferred stepping strategies (step lengthening vs step shortening), and the spatiotemporal stride characteristics (percentage increases in stride length, duration, and velocity of the crossing and postcrossing strides) during obstacle avoidance in 11 hemiplegic stroke patients and seven healthy controls. Patients were less successful in avoiding obstacles than controls (14% failure rate vs 0.5% in controls), independent of whether the affected or the unaffected leg led the obstacle avoidance. The number of failed trials increased systematically when the available response time became shorter. During successful trials, lengthening of the step was generally preferred over shortening. This bias towards step lengthening was more pronounced in stroke patients (step lengthening in 91% of the trials vs 75% in controls), irrespective of the side of obstacle presentation. For both groups, overall strategy preference did not adhere to a principle of minimal foot displacement, since step lengthening was used even if it would be more spatially efficient to shorten the step. No statistically significant group differences were found for the increases in length, duration, and velocity of the crossing and postcrossing strides. However, for a subgroup of more slowly walking patients, large percentage increases were found in crossing stride length, duration, and velocity. Similar results were obtained for the postcrossing stride, indicating that, for this subgroup of patients, restoration of the normal walking cadence was more difficult. Overall, no systematic differences were found between the affected and the unaffected leg in stroke patients with respect to failure rates, stepping strategies, or spatiotemporal measures of obstacle avoidance. The present findings suggest that the ability to adequately modify the stepping pattern in response to imposed spatiotemporal constraints is impaired in persons with stroke, especially when modifications have to be performed under time pressure. In addition, the stepping strategies employed by subjects with stroke are different from those found in controls, possibly to reduce the complexity of the avoidance maneuver and to enhance safety. Finally, unilateral cortical damage results in an impaired ability to avoid obstacles on both sides of the body, suggesting that the reduced ability of stroke patients to negotiate obstacles may be related to problems of a more general coordinative nature.