Stress and adaptational outcomes. The problem of confounded measures.

" Confounding is a major source of uneasiness among many who do research on the relationship between stress and adaptational outcomes such as psychological symptoms and somatic health. A commonly proposed solution, illustrated by a recent article by Dohrenwend, Dohrenwend, Dodson, and Shrout, is to purify the independent variable, stress, by focusing on its environmental aspects and by making it independent of psychological response variables such as perceptions or appraisals. Such a solution, however, is neither possible nor desirable, and it obviates relational, cognitive theories of psychological stress such as our own. In this article, we closely examine the problem of confounding and circularity in stress research and provide new data. We argue that the appraisal process should not and cannot be removed in the measurement of psychological stress, and therefore some confounding is inevitable. Like emotion, stress is best regarded as a complex rubric consisting of many interrelated variables and processes rather than as a simple variable that can be readily measured and correlated with adaptational outcomes. No issue in the psychology of health is of greater interest and importance than whether and how stress influences adaptational outcomes such as well-being, social functioning, and somatic health. This issue has fueled extensive research on stressful life events (see Thoits, 1983, for a recent review) and more recently on an alternative type of stress variously referred to by the terms microstressors (McLean, 1976; Monroe, 1983), chronic role strains (Pearlin, 1983), and daily hassles (DeLongis, Coyne, Dakof, Folkman, & Lazarus, 1982; Kanner, Coyne, Schaefer, & Lazarus, 1981; Lazarus, 1984, in press: Lazarus & DeLongis, 1983). This article was prompted by an article by Dohrenwend, Dohrenwend, Dodson, and Shrout (1984) that dealt with the issue of confounding between measures of stress and health outcomes. We address the issues posed by these authors and, based on their findings, provide some reanalyses of data that bear importantly on them. However, the issue of confounding raises questions that go far beyond parochial concerns, and it warrants a broad and searching examination. In this article we explore these questions and respond in terms of our theory of psychological stress and our measurements of daily hassles. In the critique and research reported by Dohrenwend et al. (1984), clinical psychologists evaluated the items on Holmes and Rahe's Life Events Scale (1967), Lin, Dean, and Ensel's Instrumental-Expressive Social Support Scale (1981), and the Berkeley Stress and Coping Project's Daily Hassles Scale (DeLongis et al., 1982; Kanner et al., 1981) with respect to the likelihood that, in addition to what they were supposed to measure, their contents also reflected symptoms of psychological disorder. Their ratings indicated that the scales contained substantial though varying amounts of psychopathology, thus producing a confounding between what these scales were supposed to measure and what they were often used to predict, namely, psychological problems. The authors concluded that more care should be given to how stressful life events, social supports, and hassles are conceptualized and used in the study of psychological stress and health. They emphasized the positive value of treating stressors as environmental inputs that are independent of the reaction or state of mind of the person. In contrast, our view is that stress lies not in the environmental input but in the person's appraisal of the relationship between that input and its demands and the person's agendas (e.g., beliefs, commitments, goals) and capabilities to meet, mitigate, or alter these demands in the interests of well-being (Folkman & Lazarus, 1984; Lazarus, 1981; Lazarus & Folkman, 1984; Lazarus, Kanner, & Folkman, 1980; Lazarus & Launier, 1978). The searching question that guides this article is whether and how relational, cognitive approaches to psychological stress such as ours, which draw on subjectively defined assessments of stress, can overcome the dangers of confounding and circularity. The Problem of Confounded Measures That confounding is a major source of uneasiness among those who do research on stress and adap770 July 1985 • American Psychologist Copyright 1985 by the American Psychological Association, Inc. 0003-066X/85/$00.75 "Col. 40, No. 7, 770-779 tational outcomes is indicated by the following comment by Kasl (1978): Unfortunately, this convergence of theoretical formulations [about the role of individual differences in appraisal] has led to a self-serving methodological trap which has tended to trivialize a good deal of the research on work stress or role stress: the measurement of the "independent" variable (e.g., role ambiguity, role conflict, quantitative overload, etc.) and the measurement of the "dependent" variable (work strain, distress, dissatisfaction) are sometimes so close operationally that they appear to be simply two similar measures of a single concept. (p. 13) Dohrenwend et al. (1984) and Kasl (1978; see also Dohrenwend, 1978, and Thoits, in press) posed an issue that must be confronted by those who define stress in relational and cognitive terms. I f the relationship being studied is entirely circular, that is, the same process is being measured in the independent and dependent variables, then we would have to agree that probably nothing of importance will be learned about the functional connection between stress as a condition of life and the health outcome. That circularity or confounding is a problem in present-day research is illustrated by an article by Cohen, Kamarck, and Mermelstein (1983). The authors presented what they called a "global measure of perceived stress" (p. 385). Cohen et al. (1983) reported that their global measure correlates well with outcome measures of symptomatology. However, an examination of the 14 items of this scale, which assess general, negativ~ feelings and reactions over the last month, suggests that it is yet another measure of psychopathology or distress. These items deal with reactions such as being upset, having or lacking control over things, feeling nervous and stressed, feeling or not feeling effective, and being overwhelmed by difficulties. In this strategy of stress measurement, the antecedent and consequent measures seem to overlap entirely, making it questionable whether the correlation provides any gains in knowledge. Confounding or circularity refers to redundancy among variables: "The causes contain all that is contained in the effect; the effect contains nothing that is not contained in the causes" (Oxford English Dictionary, 1970, Vol. 11, p. 116). It can occur for a variety of reasons. For example, the components Work on this article was supported in part by grants from the MacArthur Foundation and grant 1R01 DAO2976 from the National Institute on Drug Abuse. We would like to thank Philip E. Tetlock, Seth Roberts, and Guy Swanson for their helpful comments and suggestions on an earlier draft of this article. Requests for reprints should be sent to Richard S. Lazarus, Department of Psychology, University of California, Berkeley, California 94720. making up the variables themselves can be redundant or identical, as when there are duplicated or similar items. This is the problem with the research of Cohen et al. (1983) cited above. Perhaps it would be even better to speak of overlapping measures (see Nicholls, Licht, & Pearl, 1982, for an analysis of overlapping item content in personality questionnaires). Alternatively, the variables as measured can reflect some third, underlying variable in common, even though there is no explicit redundancy in the form of duplicated or similar items. Psychologists have been debating the problems inherent in circular or tautological reasoning for a long time. Some would argue that circularity is inevitable in any scientific explanation. For example, Pratt (1939) said: The aversion to circular arguments is the last stronghold of that faith which believes that science can furnish for any given array of well-established facts a why which is more than a mere statement of how the facts go together. In certain quarters the faith still lingers on that the laws of nature are something more than the events in nature themselves. The events obey the laws, which can not be the same as saying that the events obey themselves. If the statement that nature is subject to law really means that nature is subject to nature, then the so-called laws of science explain nothing. They are pompous tautologies, vicious circles. The majority of present-day scientists, particularly the correlationists, have indeed come to realize that a scientific why can be no more than an exact how, but the corollary is frequently overlooked, namely, that all arguments in support of any given hypothesis must of necessity be circular. (pp. 149-150) Pratt went on to give a historical example of how circular concepts can lead to gain. Shortly after the planet Uranus was found, it became evident that its orbit could not be accounted for by the known motions of Jupiter and Saturn. After it was realized that inaccurate observations were not involved, it was suggested that the deviations in orbit must be due to the action of a still more distant, as yet undiscovered, planet. Calculations were made of the orbit such a new planet would have to have to produce the positions observed for Uranus, and this hypothesized planet was regarded as a reasonable explanation for the perplexing behavior of Uranus. Such an explanation, noted Pratt, is a perfect illustration of circularity, in which "the formal properties of the calculated planet were derived from the material properties, that is, the observed positions of Uranus, and then these same formal properties were used to explain the behavior of Uranus" (Pratt, 1939, p. 153). This circular reasoning led to the discovery o