The aims of this study were to quantify the inter-test reliability of several voluntary, non-invasive measures of respiratory muscle function and to determine the implications of these data for studies using a repeated-measures design. Systematic measurement differences were found for 50% of the variables ( P</=0.05, t-tests). Nevertheless, 95% ratio limits of agreement for most measures proved acceptable and similar to those reported elsewhere. The random error component of the agreement ratios ranged from 1.047 to 1.149 for measures of pulmonary function in healthy subjects ( n=46), 1.045 to 1.056 for maximum static respiratory pressures ( n=24), 1.062 to 1.173 for measures relating to the maximum pressure-flow-power relationship ( n=16-22), and 1.036 to 1.071 for measures relating to maximum incremental inspiratory muscle performance ( n=12). The judgement that the limits of agreement were acceptable is supported by the sample-size calculations. Estimated sample sizes based upon an alpha level of 0.05 and a statistical power of 0.9 were mostly </=11 for a repeated-measures experimental design, particularly for the larger effect sizes (>/=5%). However, peak expiratory flow, the maximum rate of pressure development and the time constant of relaxation, require larger sample sizes to detect small within-group changes. In conclusion, the described protocols provide reliable measurements for most parameters of respiratory muscle function in healthy subjects. Furthermore, experiments utilising a within-subjects design lasting up to 3 weeks can be conducted with feasible sample sizes (</=11 per group) where substantial (>/=5%) changes are expected.