The reproducibility of peak cardiopulmonary exercise responses and the first (VT1) and second [VT2) ventilation thresholds was studied in sixteen endurance-trained male cyclists (mean +/- SD peak oxygen uptake [VO2 peak] = 63.3 +/- 7.1 ml x kg(-1) x min(-1)) during duplicate 30 W x min(-1) ramp cycling protocols. Expired gas sampled from a mixing chamber was analysed on-line and VT1 and VT2 were determined by computerised V-slope analysis and visually by two evaluators (test-retest reliability) and again by one of the evaluators 12 months later (intra-evaluator reliability) from 20-s-average respiratory data. The results demonstrated high intra-evaluator reliability (r = 0.91-0.97, P < 0.0001) for repeat determinations of VO2, work rate (WR) and heart rate (HR) at VT1 and VT2. No significant differences were observed between Tests 1 and 2 for any of the measured variables (P > 0.05). Test-retest intraclass reliability coefficients ranged from 0.86 to 0.98 (P < 0.0001) for VO2 peak, peak pulmonary ventilation (VE), carbon dioxide output (VCO2), HR and WR values, and measurements of VO2 and WR at VT2, and from 0.67 to 0.80 (P < 0.01) for measurements of VO2 and WR at VT1. The reliability of VT1 and VT2 was reduced when the thresholds were expressed as relative (%VO2 peak) (r = 0.67-0.70, P<0.01) rather than absolute (l x min(-1)) (r = 0.77-0.93, P<0.001) VO2 values. It was concluded that VO2 peak, peak VE, VCO2. HR and WR values, and VT2 are highly reproducible in trained cyclists using a 30 W x min(-1) ramp exercise function. However, determinations of VT1 are less reliable. Additionally, ventilation thresholds are more reliably described using absolute rather than relative VO2 values.