Recent studies have demonstrated that cognitive loading aggravates the gait impairments that are typically seen in Parkinson's disease (PD). To better understand the relationship between cognition and gait in PD, we evaluated 30 subjects with PD who suffer from motor response fluctuations. The subjects were clinically and cognitively assessed using standard clinical (e.g., Unified Parkinson's Disease Rating Scale) and cognitive tests while in the "ON" period of the medication cycle. In addition, the subjects wore force-sensitive insoles to quantify the timing of the gait cycles during 80-m walks at a self-selected, comfortable pace during three randomly presented gait conditions: (1) usual-walking, (2) dual tasking (DT), performing serial 3 subtractions (DT_S3), and (3) DT_S7. Stride length, gait speed, gait variability and bilateral coordination of gait were affected by DT, compared to the usual-walking (P < 0.001) as was gait asymmetry (P = 0.024). Stepwise regression analyses showed that a subset of the cognitive performance scores accounted for the changes seen in the gait parameters during DT, e.g., set shifting capabilities as expressed by the Trial Making Test Scores (P < 0.001). Affect (e.g., anxiety) was not associated with DT-related gait changes. For most gait features, DT had a large impact on the DT_S3 condition with only minimal additional effect in the DT_S7 condition. These results demonstrate that the complex cognitive-motor interplay in the control of gait in patients with PD who suffer from motor response fluctuations has a profound and marked effect during DT conditions on gait variability, asymmetry and bilateral coordination, even in the "ON" state when patients are likely to be most active, mobile and vulnerable to the negative effects of dual tasking.