Abstract
The generalizability theory of Cronbach et al. (1972) is illustrated by three reliability studies conducted in clinical settings. Data are presented for the Koppitz‐Bender emotional indicators, for the WAIS verbal subtests, and for ratings in the study of a single case. It is argued that classical reliability theory is not adequate to deal with many of the questions which clinicians are led to ask: the question may have greater complexity than the classical view readily admits; or the questions may have practical and procedural implications which the classical theory does not recognize. Generalizability theory offers a more direct expression for questions having clear operational meaning.