The purpose of this study was to assess the reproducibility over four trials of perceptually regulated exercise intensity during short-term cycle ergometry. Recent research has suggested that an improvement in the reproducibility (better agreement) of the exercise output would be observed with a repeated practice using regulatory tools such as Borg's 6 - 20 rating of perceived exertion (RPE) scale. Eighteen healthy active volunteers (nine males, mean age (+/- SD) 24.7 +/- 3.4 yr, and nine females 27.6 +/- 5.4 yr) completed four identical intermittent effort production trials on a cycle ergometer over a period of two - three weeks, with all trials being between three and five days apart. After warm-up, the volunteers were asked to produce four x three-minute bouts of exercise at RPE levels: 13, 15, 9, and 17 (in this order). Power output (W), percentage maximum heart rate reserve (%MHRR), and oxygen consumption (VO(2); ml x kg(-1) x min(-1)) were recorded in the final minute of each bout. Analysis revealed that the 95 % limits of agreement (LoA) between repeated trials did not decrease for the objective markers of exercise intensity, remaining wide throughout. In the worst case comparisons the LoA represented changes (expressed as a proportion of the mean of two trials) of up to 58.3 % in power output (T2 vs. T3 at RPE 9), 65.5 % in % MHRR (T1 vs. T2 at RPE 13) and 36.5 % in VO(2) (T3 vs. T4 at RPE 17). These findings question the use of ratings of perceived exertion to regulate exercise effort. That the reproducibility of effort is also not seen to improve with practice raises doubts about the validity of using the RPE scale to provide training intensities for this type of exercise.