Freezing of gait in Parkinson's disease can be difficult to study in the laboratory. Here we investigate the use of a variable-width doorway to provoke freeze behavior together with new objective methods to measure it. With this approach we compare the effects of anti-parkinsonian treatments (medications and deep-brain stimulation of the subthalamic nucleus) on freezing and other gait impairments. Ten "freezers" and 10 control participants were studied. Whole-body kinematics were measured while participants walked at preferred speed in each of 4 doorway conditions (no door present, door width at 100%, 125%, and 150% of shoulder width) and in 4 treatment states (offmeds/offstim, offmeds/onstim, onmeds/offstim, onmeds/onstim). With no doorway, the Parkinson's group showed characteristic gait disturbances including slow speed, short steps, and variable step timing. Treatments improved these disturbances. The Parkinson's group slowed further at doorways by an amount inversely proportional to door width, suggesting a visuomotor dysfunction. This was not improved by either treatment alone. Finally, freeze-like events were successfully provoked near the doorway and their prevalence significantly increased in narrower doorways. These were defined clinically and by 2 objective criteria that correlated well with clinical ratings. The risk of freeze-like events was reduced by medication but not by deep-brain stimulation. Freeze behavior can be provoked in a replicable experimental setting using the variable-width doorway paradigm, and measured objectively using 2 definitions introduced here. The differential effects of medication and deep-brain stimulation on the gait disturbances highlight the complexity of Parkinsonian gait disorders and their management.