Bronchoalveolar lavage (BAL) is used to obtain inflammatory cells from the lung. for clinical research, parametric statistics are frequently used to compare cells present in BAL of patients with lung disease with cells present in BAL of normal subjects. To determine if these populations can be compared in this manner we performed BAL on 111 never-smoking, normal volunteers and determined: (1) the mean, median, standard deviation, and range of the cells in BAL; (2) whether the data are normally distributed and satisfy the criteria for use of parametric statistical analysis. The BAL cellularity was expressed as a percentage of total cells, cells per milliliter return, and total cells per lavage. Regardless of the means of expression, no measure of BAL cellularity (total cells, macrophages, lymphocytes, neutrophils, or eosinophils) conformed to the normal (bell-shaped) distribution when tested for goodness of fit with the G statistic (all p < 0.001). The lack of fit to the normal distribution was not substantially altered by either the method of expressing the data (i.e., cells per milliliter, total cells, or percent of cells) or log transformation of the data. The poor fit in all cases resulted from clumping of the data about the mean and large tails. The percent of cells were, therefore, tested for goodness of fit to the Poisson distribution, a distribution of discrete variables. The neutrophil and eosinophil percentages resulted in an excellent fit to the Poisson distribution, but macrophage and lymphocyte percentages did not. The observations that the BAL data are not normally distributed indicate that nonparametric statistics should be used for comparing BAL cell data between normal subjects and patients with lung disease.