As the relationship between the ratings of perceived exertion (RPE, Borg 1998), oxygen uptake (VO₂) and heart rate (HR) during a graded exercise test (GXT) is high (r>0.90), the RPE is often used to prescribe various levels of exercise intensity. The relationship between RPE and VO₂ may also facilitate the prediction of maximal oxygen uptake (VO₂max). However, it is unknown if it is possible to predict VO₂max from sub-maximal effort production levels. This would be a useful application, particularly as the production of a given RPE is often used as a guide to exercise intensity regulation. The purpose of this study was to assess the validity of predicting VO₂max from sub-maximal VO₂ values elicited during a perceptually-regulated GXT. We hypothesized that the strong relationship between RPE and VO₂ would enable VO₂max to be predicted and that this would improve with practice. Ten physically active men (23.6±2.1 years, 79.6±10.2 kg, 1.81±0.05 m) performed four exercise tests: a GXT to establish VO₂max (starting at 50 W and increasing by 50 W every 3 min until exhaustion), and three sub-maximal RPE production trials on an electromagnetically braked cycle ergometer, each separated by 48 h. The resistance on the cycle was manipulated within ±1 W, independent of pedal speed. VO₂ and HR were monitored continuously. Physiological and exercise intensity information were concealed from the participant. In each production trial, participants exercised at five self-regulated RPE levels (9, 11, 13, 15 and 17) prescribed in an incremental fashion for 4 min. The RPE Scale was mounted in full view of the participant. At the beginning of each RPE increment, 2-3 min was allowed to adjust the resistance until the participant was sure that the resistance equated with the prescribed RPE. No further intensity adjustments were made after the initial 2-3 min. Each bout was separated by 50 W active recovery for four more minutes, before commencing the next RPE production. Cadence was maintained at 50 —80 rpm. Correlations between RPE and VO₂ were in the range 0.92 to 0.99 across the three trials. There were no significant differences between measured VO₂ (mean ± SD, 48.8 ± 7.0 ml kg^-1 min^-1) and predicted VO₂max values (47.3 ± 10.0, 48.6 ± 8.1 and 49.9 ± 7.4 ml kg^-1 min^-1, for trials 1, 2 and 3, respectively) when VO₂max was predicted from RPE values of 9—17 (P>.05) The same was observed when VO₂max was predicted from RPE 9—15. Limits of agreement (LoA) analysis on actual and predicted VO₂max values (from RPE 9—17) were (bias ±1.96 x SDdiff) 1.5±7.3, 0.2±4.9 and −1.2±5.8 ml kg^-1 min^-1, for trials 1, 2 and 3, respectively. Corresponding LoA values for actual and predicted VO₂max (from RPE 9—15) were 5.4±11.3, 4.4±8.7 and 2.3±8.4 ml kg^-1 min^-1, respectively. The data suggest that a sub-maximal, perceptually-guided, GXT provides acceptable estimates of VO₂max which are further improved with practice in fit young males. The method may have potential in groups where maximal tests are not desirable.