Results of our previous studies have shown that the slow, shuffling gait of Parkinson's disease patients is due to an inability to generate appropriate stride length and that cadence control is intact and is used as a compensatory mechanism. The reason for the reduced stride length is unclear, although deficient internal cue production or inadequate contribution to cortical motor set by the basal ganglia are two possible explanations. In this study we have examined the latter possibility by comparing the long-lasting effects of visual cues in improving stride length with that of attentional strategies. Computerized stride analysis was used to measure the spatial (distance) and temporal (timing) parameters of the walking pattern in a total of 54 subjects in three separate studies. In each study Parkinson's disease subjects were trained for 20 min by repeated 10 m walks set at control stride length (determined from control subjects matched for age, sex and height), using either visual floor markers or a mental picture of the appropriate stride size. Following training, the gait patterns were monitored (i) every 15 min for 2 h; (ii) whilst interspersing secondary tasks of increasing levels of complexity; (iii) covertly, when subjects were unaware that measurement was taking place. The results demonstrated that training with both visual cues and attentional strategies could maintain normal gait for the maximum recording time of 2 h. Secondary tasks reduced stride length towards baseline values as did covert monitoring. The findings confirm that the ability to generate a normal stepping pattern is not lost in Parkinson's disease and that gait hypokinesia reflects a difficulty in activating the motor control system. Normal stride length can be elicited in Parkinson's disease using attentional strategies and visual cues. Both strategies appear to share the same mechanism of focusing attention on the stride length. The effect of attention appears to require constant vigilance to prevent reverting to more automatic control mechanisms.