Foreign and soluble gas rebreathing methods are attractive for determining cardiac output (Q(c)) because they incur less risk than traditional invasive methods such as direct Fick and thermodilution. We compared simultaneously obtained Q(c) measurements during rest and exercise to assess the accuracy and precision of several rebreathing methods. Q(c) measurements were obtained during rest (supine and standing) and stationary cycling (submaximal and maximal) in 13 men and 1 woman (age: 24 +/- 7 yr; height: 178 +/- 5 cm; weight: 78 +/- 13 kg; Vo(2max): 45.1 +/- 9.4 ml.kg(-1).min(-1); mean +/- SD) using one-N(2)O, four-C(2)H(2), one-CO(2) (single-step) rebreathing technique, and two criterion methods (direct Fick and thermodilution). CO(2) rebreathing overestimated Q(c) compared with the criterion methods (supine: 8.1 +/- 2.0 vs. 6.4 +/- 1.6 and 7.2 +/- 1.2 l/min, respectively; maximal exercise: 27.0 +/- 6.0 vs. 24.0 +/- 3.9 and 23.3 +/- 3.8 l/min). C(2)H(2) and N(2)O rebreathing techniques tended to underestimate Q(c) (range: 6.6-7.3 l/min for supine rest; range: 16.0-19.1 l/min for maximal exercise). Bartlett's test indicated variance heterogeneity among the methods (P < 0.05), where CO(2) rebreathing consistently demonstrated larger variance. At rest, most means from the noninvasive techniques were +/-10% of direct Fick and thermodilution. During exercise, all methods fell outside the +/-10% range, except for CO(2) rebreathing. Thus the CO(2) rebreathing method was accurate over a wider range (rest through maximal exercise), but was less precise. We conclude that foreign gas rebreathing can provide reasonable Q(c) estimates with fewer repeat trials during resting conditions. During exercise, these methods remain precise but tend to underestimate Q(c). Single-step CO(2) rebreathing may be successfully employed over a wider range but with more measurements needed to overcome the larger variability.