The clinical balance tests presently used cannot predict falls in Parkinson's disease (PD), perhaps because they probe fairly isolated "components" of postural control. The Multiple Tasks Test (MTT) is a new balance test that simultaneously assesses multiple components of postural control. We investigated whether this MTT can detect postural abnormalities in PD patients. Fifty young controls (mean age 27.6 years), 20 elderly controls (mean age 62.5 years), and 20 PD patients (mean age 61.8 years, mean Hoehn and Yahr stage 2.2) participated. The MTT consisted of eight separate tasks of increasing complexity, which were executed sequentially. These tasks were composed of several motor components (standing up, walking, avoiding obstacles, touching the floor, turning around, and sitting down) and one cognitive component (answering serial questions). Four additional components included carrying an empty or loaded tray, wearing slippery shoes, and reduced illumination. All components within each task had to be performed simultaneously or directly sequentially. Errors were defined as Hesitations (slowed performance) or Blocks (complete cessation), which were scored separately for execution of the motor and cognitive components. Speed of performance was not stressed, but we did measure the time taken to complete all tasks. The complete MTT was performed by all subjects, except for a subgroup of seven patients and seven elderly controls who performed a shortened version, with only three of the eight sequential tasks (simple, intermediate, and most difficult). The number of subjects that produced Hesitations or Blocks for the motor components differed between the three groups [two-way repeated measures MANOVA, F(2.7) = 20.56; P < 0.001], patients making more errors than young and elderly controls. Furthermore, the number of subjects that made motor errors increased as the tasks became more complex [F(2.7) = 6.69; P < 0.001]. This increase differed across the three groups [significant interaction effect; F(2.7) = 3.31; P < 0.001] because particularly patients produced motor errors during the more complex tasks. In both control groups, 62% performed all eight consecutive tasks without errors in the motor components. In contrast, only 8% of the patients completed all tasks without motor errors (log rank test, P < 0.0001). This difference between patients and controls disappeared if the cognitive component was also scored, because more controls made cognitive errors during complex tasks than patients. Controls apparently gave priority to execution of the motor components, which they performed significantly faster than the patients. Both patients and controls made more errors during the shortened MTT, suggesting that learning effects (gain in performance through practice) influenced performance on the complete test. The MTT is a new balance test that clearly discriminates between healthy subjects and PD patients. Unlike controls, PD patients lend less priority to motor tasks over cognitive tasks. In addition, impaired motor learning may partially explain the higher error rate in PD. Future studies must determine if impaired MTT performance can predict actual falls in daily life.