To determine, in a cohort of ambulatory older adults, whether spatial-temporal measures of foot placement during gait can predict the likelihood of future falls or whether these measures are more likely to be indicative of adaptations associated with pre-existing fear of falling. Prospective cohort study. Baseline gait measurements were performed in a gait and balance laboratory; subsequent history of falling was monitored prospectively for 1 year in two self-care facilities. Fourteen male and 61 female consecutive volunteers (mean age = 82, SD = 6) who were independent in activities of daily living and able to walk 10 m unaided. Spatial gait parameters were derived from digitized "footprints"; temporal parameters were derived using footswitches. A clinical activity-based gait assessment was also performed. The dependent variables were pre-existing fear of falling (reported at baseline) and future falling (experiencing one or more falls during the 1-year follow-up). Reduced stride length, reduced speed, increased double-support time, and poorer clinical gait scores were associated with fear but showed little evidence of an independent association with falling. Conversely, increased stride-to-stride variability in stride length, speed, and double-support was associated independently with falling but showed little evidence of relationship to fear. Increased stride width showed some evidence of association with both falling and fear. Stride-to-stride variability in speed was the single best independent predictor of falling. Changes in gait cited previously as risk factors for falling, i.e., decreased stride length and speed and prolonged double support, may in fact be stabilizing adaptations related to fear of falling. Stride-to-stride variability in the control of gait is an independent predictor of falling and may be a useful measure for identifying high-risk individuals and evaluating preventive interventions. Stride width may also be a useful outcome measure. Contrary to common expectation, a wider stride does not necessarily increase stability but instead seems to predict an increased likelihood of experiencing falls.