A basic scientific requirement of measuring instruments is that they measure reliably. Some psychologists take exception to this premise, arguing that reliability considerations are irrelevant or inappropriate to the evaluation of clinical techniques (see Holzberg, 1960, pp. 361-362).' The standard definitions of test reliability challenge the tenability of such a position, however. According to Cronbach (1949a, p. 59), for example, a reliability coefficient expresses "the accuracy of measurement." Gulliksen (1950, p. 22) defines the index of reliability as the correlation of true and observed scores. Viewed from these definitions, disavowal of reliability concerns in psychodiagnostic testing is equivalent to disregard for whether the tests accurately measure personality

488 Conclusion functioning, that is, whether they yield scores that closely approximate the subject's "true" scores. Disregard for accuracy is not easily reconcilable with usual scientific aims.