Objective: To obtain intersession estimates of error for temporal and distance (TD) parameters of gait in a sample of stroke patients undertaking inpatient rehabilitation. Design: Thirty-one stroke patients were measured with an instrumented footswitch system (after a median of 46 days poststroke; interquartile range = 26 to 63) walking over a 10-meter distance a total of four times on 3 consecutive days. Two familiarization walks provided intrasession retest data. Results: Metric estimates of systematic and random error have been provided for obtained TD parameters. Proportional indices of reliability (ICC [2, 1] and Pearson's r) were generally high, ranging from .72 to .94. Conclusion: By quantifying systematic and random error associated with the process of repeated measurements, criteria have been provided for evaluating change in TD variables during rehabilitation. Although error for gait velocity was small relative to individual differences in the stroke group, it was large relative to levels of change derived from measurements reported during typical periods of rehabilitation. Serial measurements of gait during rehabilitation may be better than two consecutive measurements. This study highlights the need to interpret estimates of error according to the purpose of measurement.