The distribution of ventilation per unit volume (Va/Vol) in the lung is often studied using the multibreath N2 washout test. Most approaches have focused on the determination of a single Va/Vol distribution (consisting of from two compartments to a continuum) compatible with the data, but because the potential number of lung units greatly exceeds the number of breaths measured, many distributions are usually compatible with a given set of washout data. Interpretation of single distributions therefore requires evaluation of the variability among all such compatible distributions. The technique of linear programing is well suited to evaluation of these compatible distributions and its application to the N2 washout is explained. Twelve examples of washout data are analyzed in this way. The results indicate that using the first 20 or so breaths of the washout, narrow distributions of Va/Vol are well specified. Distributions with up to four distinct modes are also well specified, but for broad distributions the shape cannot be adequately defined. Good resolution in the region of high Va/Vol is found to be critically dependent on measuring the first few breaths of the washout.