This study investigated whether sprint triathlon performance can be adequately predicted from laboratory tests. Ten triathletes [mean (SEM), age 21.8 (0.3) years, height 179 (2) cm, body mass 67.5 (2.5) kg] performed two graded maximal exercise test in random order, either on their own bicycle which was mounted on an ergometer or on a treadmill, to determine their peak oxygen consumption ( VO(2)peak). Furthermore, they participated in two to three 30-min constant-load tests in both swimming, cycling and running to establish their maximal lactate steady state (MLSS) in each exercise mode. Swim tests were performed in a 25-m swimming pool (water temperature 27 degrees C). During each test heart rate (HR), power output (PO) or running/swimming speed and blood lactate concentration (BLC) were recorded at regular intervals. Oxygen uptake ( VO(2)) was continuously measured during the graded tests. Two weeks after the laboratory tests all subjects competed in a triathlon race (500 m swim, 20-km bike, 5-km run) [1 h 4 min 45 s (1 min 38 s)]. Peak HR was 7 beats.min(-1) lower in the graded cycle test than in the treadmill test ( p<0.05) at similar peak BLC (approximately 10 mmol.l(-1)) and VO(2)peak (approximately 5 L.min(-1)). High correlations were found between VO(2)peak during cycling ( r=-0.71, p<0.05) or running ( r=-0.69, p<0.05) and triathlon performance. Stepwise multiple regression analysis showed that running speed and swimming speed at MLSS, together with BLC in running at MLSS, yielded the best prediction of performance [1 h 5 min 18 s (1 min 49 s)]. Thus, our data indicate that exercise tests aimed to determine MLSS in running and swimming allow for a precise estimation of sprint triathlon performance.