In 1991 the West Midlands Pulmonary Function Audit Group examined the consistency between pulmonary function laboratories in the West Midlands. Three healthy subjects visited 22 centres and performed a standard set of pulmonary function tests. Demographic data on nine hypothetical subjects was also supplied for the laboratories to produce predicted values. Equipment was checked for accuracy using standard methods. The 1991 audit revealed significant inter-laboratory variability. Sources of error were identified and after consultation, recommendations were made to improve consistency. In addition, national and regional training workshops were organized for laboratory staff. In 1995 the audit was repeated using the same three subjects. Significant differences continued for all predicted results except for residual volume (RV) and forced vital capacity (FVC) and for all measured results except for functional residual capacity (FRC). However, improvements in the coefficient of variation were seen compared with 1991 for predicted forced expiratory volume (FEV1), total lung capacity (TLC), gas transfer (TLCO), FVC, FRC and RV. Similar improvements were seen in measured results for FEV1 and FVC. Increased variation was seen for predicted corrected transfer factor (KCO) and actual RV. The majority of variables in the 1995 audit had a coefficient of variation of less than 5% between laboratories. Analysis of the predicted results from the hypothetical subjects continued to show unacceptable variation reflecting continuing computer algorithm inconsistency. The improvements seen are encouraging and suggest that a regular audit programme is worthwhile.