The time taken to achieve peak tidal expiratory flow as a proportion of total expiratory time (t PTEF/t E) during tidal breathing (TB) is used as a parameter of airway obstruction in children with asthma. Curve selection bias is one of the most important limitations to the method. This study evaluates three curve selection methods, including a computer program, which on-line selects and analyses TB curves (Masterscreen Paediatric; Jaeger, Germany). TB analysis was performed in 26 children (aged 4-7 yrs) with asthma, before and after methacholine provocation and after subsequent bronchodilatation. Levels and stability of TB parameters derived from computer-selected, unselected and unbiased eye-selected curves were compared. t PTEF/t E ratios of the computer-selected curves agreed well with the unbiased eye-selected curves (limits of agreement -4.8 and +5.8%), but were significantly different from the ratios of unselected curves. Computer-derived t PTEF/t E ratios had the highest level of stability: the reliability coefficient of baseline measurements was 0.96 for computer selection, 0.84 for eye selection and 0.87 for no selection (reliability index = 1 at maximal stability). Tidal volume, respiratory rate, inspiratory and expiratory time were also assessed accurately by the computer program. The mean t PTEF/t E ratio (computer selection) dropped after methacholine provocation (from 30 +/- 9 to 22 +/- 9% at provocative dose at which forced expiratory volume in one second had dropped > or = 20% from baseline (FEV1-PD20 level), p < 0.001) and was restored after bronchodilation (30 +/- 6%; p < 0.001). We conclude that on-line computer analysis is preferable to no selection and to by-eye selection. The use of the program avoids curve selection bias and enhances the applicability of tidal breathing analysis as a measure of airflow obstruction in young children.