On the Evaluation of Personality Changes as Measured by Psychometric Test Profiles

A wide variety of research, especially in clinical psychology, involves the measurement of behavior change. Studies dealing with treatment effects, developmental changes, and trait stability all require evaluation of changes occurring over time. In order to cope with the complexity of behavior change, current researchwhether experimental or survey-frequently employs multiple rather than single measures on each S. In particular, the multiple indicants may be standardized to form a profile; the Minnesota Multiphasic Personality Inventory, and the Wittenborn Psychiatric Rating Scales are examples. Moreover, current usage of such multiple measure tests tends to emphasize the configuration of scores. Research designs employing multiple indicant tests frequently involve groups of profiles (hereafrer called Treatment Groups) which in the simplest experimental situation would consist of two groups, experimental and control. In the case of a survey, such as the example referred to below, it is convenient to maintain the terminology, recognizing that here Treatment Groups refers to population strata differentiated on the basis of some designated attribute, such as male versus female, different instirutional settings, etc. Once groups of individual profiles are formed, the critical, practical, and theoretical issue arises of determining the kind and extent of behavior change among the Treatment Groups2 as revealed in contrasts between the initial and subsequent mean profiles. To the clinician, as well as to the experimentalist working in the personality area, rhe assumption is suspect rhat the attributes being measured are invariant over time, so that, unlike the classical concept of test-retest reliability, test changes may not be regarded simply as so called error vasiance (Loevinger, 1957; Chassan, 1957). In fact, changes in profile patterning in repeated testings are realistically viewed as more a measure of trait variance than of reliability of scales, at least In the case of instruments such as the MMPI (McKinley & Hathaway, 1944). Thus, to the excent that profile patterns reflect