Stride intervals measured during steady-state walking are irregular. These stride interval fluctuations are not random but exhibit long-range power-law correlation (alpha) such that a given stride interval is 'influenced' by earlier variations in the stride intervals. To estimate alpha, one requires a minute long sequence of right or left side stride interval data. However, to obtain a reliable alpha point estimate, the minimal stride sequence length is unknown. Additionally, it is unknown if the right and left side alpha are equivalent. In this study, the within-day and the right and left side reliabilities of alpha point estimates were examined in 23 volunteers performing three 8-min treadmill walks. In addition, eight volunteers were retested on three additional days to estimate between-day reliability. The standard error of measurement (S.E.M.) and the within- and between-day intraclass correlation (ICC) values, and their 95% confidence intervals, each calculated using the combined right and left leg 8-min alpha estimates were acceptable [0.047 (0.044-0.051); 0.914 (0.882-0.932) and 0.769 (0.689-0.815), respectively]. The left alpha (0.688 +/- 0.93) was greater than the right alpha (0.664 +/- 0.094), albeit this finding was underpowered (0.55). The alpha point estimates obtained from the full 8-min walks provided minimal S.E.M. and maximal within- and between-day ICCs. However, the minimal S.E.M. was statistically indistinguishable from the 6- and 7-min walk durations and all of the within-day and between-day ICCs were similar except for the 3- and 8-min between-day ICCs. This study suggests that data from four 3 min, three 6 min or two 8 min walk duration trials provide reliable alpha point estimates from a short series of short treadmill walks.