We compared the central-chemoreflex sensitivities estimated from steady-state tests with those estimated from rebreathing tests in five subjects. In one laboratory, each subject underwent nine dynamic end-tidal forcing experiments. Three repetitions of 3, 6 and 9 mmHg step changes in the end-tidal partial pressure of carbon dioxide, from a pre-step partial pressure 1.5 mmHg above resting, were used to establish four points of the steady-state ventilatory response to carbon dioxide. In another laboratory, each subject underwent two rebreathing experiments, one using Read's rebreathing technique and the other a modified rebreathing method which included a prior hyperventilation. The central-chemoreflex sensitivities, estimated from the slopes of the ventilatory responses to carbon dioxide using different combinations of the four steady-state points. were compared to those estimated from the slopes of the rebreathing responses. The steady-state sensitivities were significantly lower than the Read rebreathing sensitivities. The ratio of modified rebreathing sensitivities to steady-state sensitivities was closest to one when steady-state sensitivities were estimated from the two middle points of the ventilatory responses. The mean (SE) ratio of the sensitivities was 1.22 (0.21) in this case. We identify a number of factors that may affect the estimation of central-chemoreflex sensitivity using each technique. These include a maximum limit of the ventilation response at high partial pressures of carbon dioxide, an inability to sustain high ventilation for the duration of the steady-state tests and the inclusion of parts of the ventilatory response whose carbon dioxide partial pressures lie below the central-chemoreflex threshold. We conclude that the modified rebreathing method provides the best estimate of central-chemoreflex sensitivity of the three methods.