We have studied intrasubject variability with respect to counts of immunostained inflammatory cells in snap frozen endobronchial biopsies from a group of patients with clinically mild to moderate stable asthma. Fiberoptic bronchoscopy was used to obtain endobronchial biopsies from the upper and lower lobes in 12 volunteer subjects with asthma on two separate occasions 1 mo apart. During this period there was no significant change in asthma treatment, symptom scores, pulmonary function, or airway hyperresponsiveness. With the aid of immunohistochemistry and interactive image analysis, subepithelial counts for T lymphocytes, T-lymphocyte subsets, and eosinophils were made. There was wide intrasubject variability between anatomic site and with time for all the inflammatory cell counts. In each case the intrasubject variability with time was greater than lobar differences. Power calculations were made to establish the optimal sample size required for each marker. We conclude that even in stable asthma considerable biologic variability compounds sampling variability. Such sources of variability need to be borne in mind when calculating the likely power of intervention studies using biopsy data as end points. Power calculations suggest that approximately 15 subjects would be an optimal number with little advantage in increasing beyond this.