Freezing of gait (FOG) is an episodic gait disturbance that is commonly seen in Parkinson's disease (PD). To date, treatment efficacy is limited. We tested the hypothesis that an intervention that utilizes motor learning provided through intensive cueing can alleviate this symptom. Fifteen subjects with PD suffering from FOG participated in a 6 week progressive motor learning program (three training sessions per week--open trial). A training session included FOG-provoking situations (e.g., turns). Prior to each presumed FOG provocation (e.g., just before a turn), rhythmic auditory stimulation (RAS) was elicited and the subject was trained to walk rhythmically, coordinate left-right stepping and to increase step size, utilizing the RAS cueing. Net training duration increased from week to week and secondary cognitive tasks while walking were added to increase FOG propensity. FOG symptom burden was assessed before, immediately, and 4 weeks after the training period. The mean number of FOG episodes (±SEM) per 10 m of walking in a standardized gait assessment decreased from 0.52 ± 0.29 in the pre-testing to 0.15 ± 0.04 in the post-testing (p < 0.05). The duration of FOG episodes decreased from 4.3 ± 2.1 to 2.6 ± 0.6 s (p < 0.05). Additional measures (e.g., FOG questionnaire, gait speed) varied in their responsiveness to the treatment. These effects were retained 4 weeks after the training. The results of this open label study support the possibility that a motor learning-based intervention is apparently effective in reducing FOG burden, suggesting that RAS can deliver 'anti-FOG' training.