The objective of this study is to investigate the value of dual-task performance for the prediction of falls inpatients with Parkinson’s disease (PD). Two hundred sixty three patients with PD (H&Y 1–3, 65.2 ± 7.9 years)walked two times along a 10-m trajectory, both under single-task and dual-task (DT) conditions (combined with an auditory Stroop task). To control for a cueing effect, Stroop stimuli were presented at variable or fixed 1- or 2-s intervals. The auditory Stroop task was also performed alone. Dual-task costs were calculated for gait speed, stride length, stride time, stride time variability, step and stride regularity, step symmetry and Stroop composite scores(accuracy/reaction time). Subsequently, falls were registered prospectively for 1 year (monthly assessments). Patients were categorized as non-recurrent fallers (no or 1 fall) or recurrent fallers ([1 falls). Recurrent fallers (35%) had a significantly higher disease severity, lower MMSE scores, and higher Timed ‘‘Up & Go’’ test scores than non recurrent fallers. Under DT conditions, gait speed and stride lengths were significantly decreased. Stride time, stride time variability, step and stride regularity, and step symmetry did not change under DT conditions. Stroop dual-task costs were only significant for the 2-s Stroop interval trials. Importantly, recurrent fallers did not show different dual-task costs compared to non-recurrent fallers on any of the gait or Stroop parameters. These results did not change after correction for baseline group differences. Deterioration of gait or Stroop performance under dual task conditions was not associated with prospective falls in this large sample of patients with PD.