To investigate the genetic susceptibility to asthma, we developed an algorithm to classify the phenotype of each family member enrolled in a family study on the genetics of asthma. This algorithm was applied to 92, two- and three-generation families, identified through a subject (proband) with asthma first diagnosed 25 yr previously. The algorithm consisted of five classes based on the presence or absence of bronchial hyperresponsiveness (BHR), respiratory symptoms, smoking, airways obstruction, and bronchodilator reversibility. All family members were classified as: (1) definite asthma; (2) probable asthma; (3) unclassifiable airway disease; (4) chronic obstructive pulmonary disease (COPD); (5) unaffected (without clinical evidence of asthma and COPD). Thirteen of the 92 probands (14%) could not be classified as asthmatic when retested 25 yr later because of loss of BHR, loss of bronchodilator reversibility, or a current history of cigarette smoking. Of the 265 first-degree offspring, 49 (18%) were classified as having definite asthma (Class 1), and 22 (8%) as probable asthma (Class 2). A large number of offspring with clinical evidence of asthma did not have a prior physician's diagnosis of asthma, and offspring in Class 1 (definite asthma), with and without a physician's diagnosis, had similar clinical and physiologic characteristics. These results support the usefulness of this approach to classify subjects with asthma for genetic epidemiologic studies and show that reliance on a prior physician's diagnosis may result in misclassification or underdiagnosis. Characterization of the offspring in this family study showed that there is familial clustering, which supports the presence of a hereditary component in asthma.