The aim of this study was to determine the reproducibility of the maximal accumulated oxygen deficit and the associated exercise time to exhaustion during short-distance running. Fifteen well-trained males (mean +/- s: VO2max = 58.0+/-4.6 ml x kg(-1) x min(-1)) performed the maximum accumulated oxygen deficit test at an exercise intensity equivalent to 125% VO2max. The test was repeated at the same time of day on three occasions within 3 weeks. There was no significant systematic bias between trials for either maximum accumulated oxygen deficit (man +/- s: trial 1 = 69.0+/-13.1; trial 2 = 71.4+/-12.5; trial 3 = 70.4+/-15.0 ml O2 Eq x kg(-1); ANOVA, F = 0.70, PP= 0.51) or exercise time to exhaustion (trial 1 = 194 + 31.1; trial 2 = 198 + 33.2; trial 3 = 201 + 36.8 s; F= 1.49, P = 0.24). In addition, other traditional measures of reliability were also favourable. These included intraclass correlation coefficients of 0.91 and 0.87, and sample coefficients of variation of 6.8% and 5.0%, for maximum accumulated oxygen deficit and exercise time to exhaustion respectively. However, the '95% limits of agreement' were 0+/-15.1 ml O2 Eq (1.01 multiply/divide 1.26 as a ratio) and 0+/-33.5 s (1.0 multiply/divide 1.18 as a ratio) for maximum accumulated oxygen deficit and exercise time to exhaustion respectively. We estimate that the sample sizes required to detect a 10% change in exercise time to exhaustion and maximum accumulated oxygen deficit after a repeated measures experiment are 10 and 20 respectively. Unlike the results of previous maximum accumulated oxygen deficit studies, we conclude that it is not a reliable measure.