Not long after the publication of the Staw, Bell, and Clausen (1986) study, more sophisticated measures of affect appeared in the literature.1 Watson, Clark, and Tellegen (1988) composed scales of positive and negative affect (PA and NA) designed to tap individual mood states. They also proposed that these scales, with small changes in the time frame, could measure long-term affective states, thus operationalizing the concept of affective disposition or temperament (Watson, 1988; Watson & Clark, 1992).
Both Judge (1992) and Davis-Blake and Pfeffer (1989) questioned the conceptual meaning of the affective disposition scale used by Staw, Bell, and Clausen (1986). Although this scale (actually a selection of Block’s Q-sort items) was internally consistent and appeared to be face valid in terms of content, it had not been previously used in research studies on affect.
Watson’s positive and negative affect scales (PANAS) have been widely used by organizational researchers. For example, George (1989) found measures of PA and NA to be significantly associated with positive and negative mood at work. Levin and Stokes (1989) found that NA was inversely related to job satisfaction, and that this relationship held after controlling for several characteristics of the job such as autonomy and skill variety. In a longitudinal study, Watson and Slack (1993) also found that PA was positively related to overall job satisfaction at two periods of time. More recently, Thoresen and Judge (1997) reviewed 29 studies exploring the PA–job satisfaction link and 41 in which the NA–job satisfaction link was examined, finding overall true score correlations of.52 and ?.40, respectively.
Though research has strongly supported the relationship between affect and job satisfaction, some questions about this linkage can be raised. One concern is common method bias, because both the PANAS and measures of job satisfaction contain evaluative overtones and are usually measured on a single questionnaire. A second problem concerns the theoretical meaning of the PANAS. When most people think of affect, adjectives such as happy, sad, cheerful, or pleased are brought to mind. Yet none of these pleasantness items are included in PANAS. Included instead are items that are higher in activation as well as positive/negative evaluative tone. For example, the adjectives for positive affect (the PA scale) include items such as interested, excited, enthusiastic, and attentive. The adjectives for negative affect (the NA scale) contain items such as upset, distressed, nervous, and scared. Thus, as Larsen and Diener (1992) and Russell and Barrett (1999) noted, what Watson and his colleagues have termed positive and negative affect are rather stylized (or complex) representations of emotion rather than a direct scaling of pleasant and unpleasant states.2
A third and perhaps more fundamental issue raised by research on positive and negative affect concerns the dimensionality of emotion itself. Watson and his colleagues have argued for a bivariate model of affect in which PA and NA are relatively independent dimensions rather than two ends of a single bipolar scale. Their position has been supported by a number of empirical studies using questionnaire measures such as PANAS (Watson & Tellegen, 1985; Watson & Tellegen, 1999). Also bolstering the bivariate position are neuropsychological studies showing that positive (appetitive) and negative (aversive) processing take place in different locations of the brain, thereby consisting of physiologically separable processes (e. g., Gray, 1994).
2For a simple solution to this problem, it is possible to add pleasantness items as an addendum to the PANAS items or use other questionnaire scales such as Tellegen’s well-being scale.
Two arguments have been leveled against the bivariate model of affect. First, with respect to studies using questionnaire data, Green, Salovey and Truax (1999) have noted that much of the empirical evidence supporting the independence of positive and negative affect is likely the result of measurement error and scaling artifacts. With regard to neurological research on affect, Cacioppo and his colleagues have noted that there may be important differences between physiological and psychological sources of affect (Cacioppo & Berntson, 1999; Cacioppo, Gardner, & Berntson, 1999). Just because positive and negative stimuli may be processed by different physiological systems does not mean that they cannot be cognitively integrated into a single evaluative process. They argue that all combinations of positive and negative activation are cognitively combined into a net predisposition toward a particular object, and that there may be evolutionary importance for people to aggregate preferences and organize action toward (and away from) external objects.