Measuring Attitudes Toward Self and Others in Society: State of the Art

Abstract

This study investigates the correlation between 31 social life tests within and across groupings by title, by concept, and by factor weights; also the principal dimensions of the 107 items appearing in those tests. Unadjusted test correlations tend to be larger within groupings than across them; however, correlations adjusted for overlapping items show little or no difference within and between groupings. Item correlations contain eight significant (as defined) dimensions; corresponding factor scores were interpreted to be measures of pessimism, depression, cynicism, anxiety, fatalism, job morale, life satisfaction, and personal morale. The findings on the whole suggest that the number of social life attitude tests in use in sociology is excessive and that a smaller number would suffice. Since the 1930s sociologists have constructed a number of objective tests to measure attitudes and feelings toward self and others in society. These tests go by different but related titles: morale, normlessness, estrangement, etc. Since they contain closely similar items, by that criterion they all belong to the same domain. Consequently, one would expect all tests to be positively correlated with one another but not necessarily to the same degree. Tests bearing the same title might be expected to be more closely correlated than tests bearing different titles; tests representing the same construct might be expected to be more closely correlated than tests representing different constructs. This study addresses these issues by analyzing first the correlations between 31 separate tests-hereafter referred to as the tests-and second the correlatons between the 107 items appearing in those tests. It seeks to answer the following specific questions: (1) Do tests with the same title correlate more closely than tests with different titles? (2) Do tests purporting *This research was supported directly by a U.S. Public Health Service Grant (MH22294) and indirectly by a U.S. Public Health Service Training Grant (MH10577). We wish to acknowledge our debt to Norman Bradburn, Melvin Kohn, and Melvin Seeman, who read a preliminary draft of this paper and made suggestions for its improvement; also our debt to Robin Stryker for her careful examination of writings pertaining to test constructs.