How do people arrive at assessments of their own life quality? This investigation begins from the hypothesis that social indicators of perceived well-being will, like other attitudes, reflect two basic types of influence: affect and cognition. In addition, the indicators were expected to include two other components: unique variance (mainly random measurement error) and correlated measurement error. The extent to which each of these variance components was reflected in assessments of life-as-awhole from recent national quality of life surveys was estimated by a confirmatory factor analysis model. In addition, a series of models was developed to provide an interpretation of the way the factors of cognition and affect operate along with evaluations of specific life concerns (domains) in the perception of well-being. In the preferred model, it was found that the domain evaluations had no direct impact on life-as-a-whole assessments-the contribution of the domains was indirect by way of their associations with cognition and affect.
The notion of subjective well-being or perceived quality of life takes on specific meaning in terms of the measures used to operationalize it. It remains nevertheless an abstract concept capable of diverse interpretations. Different measures can carry different meanings. In a series of secondary analyses (Andrews & McKennell, 1980;McKennell, 1978; McKennell & Andrews, 1980), some of the measures that have been tried out in recent national quality of life surveys were re-examined. The results of this re-examination lead to a greater understanding of the measures that should be found useful to designers of studies of subjective well-being in communities.
The focus of the secondary analyses was the distinction between cognition and affect which previous work indicated was important but had left unclarified. Cognition refers here to the process whereby relativism enters into satisfaction judgments. Satisfactions, it is commonly held, are tied to the aspirations and standards of comparison in terms of which current circumstances are being evaluated. It is possible, for example, for people in relatively disadvantaged circumstances, because of their lower aspirations, to report higher than average levels of satisfaction, and for the more privileged, because of their higher aspirations, to be dissatisfied. Afect, on the other hand, refers to the individual's immediate feeling state that is not anchored, or not tied to the same extent, to cognitive frames of reference. (For a more extended review of the general characteristics and some possible implications of the distinction between cognition and affect in the perception of well-being, see McKennell, 1978.) This kind of distinction is sometimes made, at a semantic level, between "happiness" and "satisfaction." The scales developed by Bradburn (1969) to measure affect do correlate more highly with happiness than with life-satisfaction ratings, but only marginally. McKennell (1978), in an initial paper, provided some evidence to show that currently used rating-scale indicators of perceived well-being are compounds of both affect and cognition, differing only marginally in the extent to which they loaded on one or the other factor. One analysis focused on the area of nonoverlap between happiness and life-satisfaction ratings by dichotomizing and cross-tabulating the two indicators. By comparing appropriate cells of the 2 X 2 table, it was possible to contrast more purely the affective component in the happiness ratings with the cognitive component in the *Reprint requests should be sent to author, 56 The Pryors, East Heath Rd., London, N.W. 3, United Kingdom.