The purpose of this investigation was to determine the effects of dual task performance on postural instability in subjects with idiopathic Parkinson's disease (PD) compared with healthy elderly people. In particular, we aimed to divert attention to a secondary task so the full extent of balance disturbance could be revealed without compensation by attentional mechanisms. Forty-five subjects were tested: 15 PD subjects with a past history of falls; 15 PD subjects with no history of falls; and 15 unimpaired individuals. Groups were matched for age and sex and subjects with PD were tested at peak dose in the levodopa medication cycle. Each subject was tested on their ability to maintain stability in three conditions: (1) steady standing (feet apart, feet together, step stance, tandem stance, single leg stance); (2) in response to perturbations generated by self-initiated movements (arm raise test, step test); and (3) in response to an unexpected external perturbation in upright stance, the shoulder tug test. The concurrent task was verbal-cognitive and required subjects to recite the days of the week backwards. The concurrent task produced a significant deterioration in performance for the arm raise test in all groups, the step test for the PD fallers and controls and for tandem stance in the PD fallers. Ceiling effects were evident for timed tests with feet apart and feet together resulting in poor discriminative validity for these tests. The external perturbation test showed differences between the three groups for both unitask and concurrent task conditions, yet similar rates of change from unitask to dual task conditions. Because PD fallers had a more severe initial deficit than controls, deterioration placed them in that part of the balance continuum at high risk of losing equilibrium.