We examined the effects of within-breath changes in compliance (C) upon the accuracy of measurements of compliance and resistance (R) by linear regression analysis and by Mead and Wittenberger's method. These effects were illustrated by a computer model and by lung models with linear and nonlinear pressure/volume relationships, and were also studied in 14 normal spontaneously breathing premature infants (mean +/- SD, BW 1,290 +/- 200 g, GA 29.9 +/- 2.7 weeks, age 7.4 +/- 2.1 days). Flow was measured by pneumotachography and tidal volume was derived as digitally integrated flow, and transpulmonary pressure as airway minus esophageal pressure. We found that C and R calculated from the equation of motion is accurate only if C and R remain constant throughout the respiratory cycle. Calculated compliance depends more on C at the end than at the beginning of inspiration. A decreasing C leads to underestimation or R, while an increasing C leads to an overestimation of inspiratory R. Calculated total R may be accurate, but with low r values for measurement points. Mead and Wittenberger's method and the regression method are similarly affected by changing C; however, since the regression method is based on many more measurement points and therefore allows the detection and analysis of within-breath changes of C and R, it is less prone to erroneous results secondary to signal artifacts than Mead and Wittenberger's method.