An exponentially increasing incremental work rate protocol can be used to accommodate a wide range of peak exercise capacities. (Northridge et al. 1990; Riley et al. 1992). Theoretical analysis suggests that peak gas exchange would not differ between exponential and linear protocols, but that the more complex kinetics with exponential increments might possibly affect the determination of the gas exchange anaerobic threshold (θ). (Fukuba et al. 2000). This study aims to compare exponential and linear protocols for peak [vdot]O₂ and θ measurements using bicycle ergometer and treadmill. Cardiopulmonary exercise tests were carried out using breath by breath gas analysis (Sensormedics Vmax 229) on six healthy male subjects aged 20-24 years. We used a standardised exponential exercise protocol (Northridge et al. 1990) in which the work rate was incremented each minute by 15% of the previous work rate on a bicycle ergometer (Lode Corival) and treadmill (Marquette Series 2000) and compared with linear protocols (25 W increments each minute on the ergometer) or equal grade increments at constant speed on the treadmill (modified Balke protocol). The modes of exercise were also compared. Subjects were blind to the protocol and randomly allocated to protocol in an orthogonal cross-over design. An interval of at least 24 hours was allowed between tests. Peak [vdot]O₂ was taken as the average [vdot]O₂ in the last 30 s of exercise and θ from the accelerated [vdot]CO₂ by the V-slope method, determined blind to both the protocol and mode of exercise. Intraclass correlation and analysis of variance were used to determine the reliability and significance of the results. All subjects exercised to exhaustion on every test. There was no significant carry-over effect between tests. The mean [SD] peak [vdot]O₂ was 3.2 [0.6] l min^-1 with the linear protocol and 3.3 [0.7] with the exponential protocol (95% CL of difference -0.1 to +0.3; p=0.5). The mean [SD] θ was 1.61 [0.43] l min^-1 with the linear protocol and 1.71 [0.44] with the exponential protocol (95% CL of difference -0.04 to +.24) (p=0.3). Intraclass correlation coefficients between protocols were 0.83 (95% CL 0.2 to 0.97) (p=0.01) for peak [vdot]O₂ and 0.90 (95% CL 0.57 to 0.99) (p=0.001) for θ with a scatter close to the line of identity (Fig. 1). Comparing modes of exercise, the mean peak [vdot]O₂ was higher on the treadmill, 3.4 l min^-1, than the ergometer, 3.0 (95%CL of difference +0.2 to +0.6; p=0.01). The mean θ was also higher on the treadmill (1.76 l min^-1) than the ergometer (1.56; 95% CL of difference +0.06 to +0.34; P=0.05). Thus there was agreement within confidence limits between exponential and linear incremental work protocols for both peak [vdot]O₂ and θ measurements. Differences between the two modes of exercise confirm results from previous studies. (Shimizu et al. 1991; Riley et al. 1992)