Although serial peak expiratory flow (PEF) measurements are often used to assess the variability of airflow obstruction, the range of values to be expected in the general population has never been defined, nor is there any consensus as to how PEF variability should be expressed. We have compared PEF recordings made by 121 subjects selected at random from the population of a small town (Group A) and 221 subjects selected because of wheeze in the last year (Group B). Subjects were asked to record PEF every 2 h during waking hours for 7 days using a mini-Wright peak flow meter. Seven indices of PEF variability were derived for each subject and the range for each index determined. All indices showed a positively skewed distribution in the random sample. Two variability indices, standard deviation percent mean and amplitude percent mean, provided the greatest separation between subjects with both a diagnosis of asthma and wheeze in the last year and subjects with neither feature and also provided the highest intra-class correlation coefficients. We conclude that amplitude percent mean and standard deviation percent mean provide the best means of expressing PEF variability for epidemiological purposes, but that amplitude percent mean is more easily derived and appears to be the index of choice. PEF variability has a continuous distribution in the general population and no clear-cut division between asthmatic and nonasthmatic subjects can be defined.