Feelings as Information: The “How-Do-I-Feel-About-It?” Heuristic

Feelings as Information: The “How-Do-I-Feel-About-It?” Heuristic

Schwarz and Clore (1983), drawing on earlier suggestions by Wyer and Carlston (1979), proposed that individuals may use their apparent affective reactions to the target as a basis of judgment. In fact, some evaluative judgments refer, by definition, to our feelings about the stimulus. For example, when asked how “likeable” a person is, we may base this judgment on our own feelings toward the person rather than on a review of the person’s features. Other judgment tasks may not pertain directly to our feelings but may pose a task that is complex and demanding. In this case, we may simplify the task by assessing our apparent affective reactions to the target, essentially asking ourselves, “How do I feel about this?” Because we have only one window on our immediate experiences, however, we may mistake feelings due to a preexisting mood state as a reaction to the target, resulting in more positive evaluations under happy than under sad moods. This feelings-as-information assumption generates a number of predictions that cannot be derived from the assumption that mood effects on evaluative judgments are mediated by mood-congruent recall or encoding, of which I address only two (see Schwarz, 1990; Schwarz & Clore, 1988, 1996, for more detailed discussions).

Perceived Informational Value

The most crucial prediction is that the impact of feelings on evaluative judgments is a function of the feelings’ perceived informational value. If individuals attribute their current feelings to a source that is irrelevant to the evaluation of the target, the informational value of their feelings should be discredited and their feelings should not influence the evaluative judgment. According to models of mood-congruent recall, however, the impact of affective states should depend only on the evaluative implications of the information retrieved from memory, rather than on information provided by the affective state itself. Therefore, models of mood-congruent recall predict that manipulations of the informational value of one’s current mood will not influence its impact on evaluative judgments.

In line with the feelings-as-information hypothesis, Schwarz and Clore (1983) observed that the impact of moods on judgments of life-satisfaction was eliminated when subjects attributed their current feelings either correctly (Experiment 2) or incorrectly (Experiment 1) to a transient source. For example, subjects reported higher life satisfaction and a more elated current mood in telephone interviews when called on sunny rather than rainy days. In some conditions, however, the interviewer, who pretended to call from out of town, asked, “By the way, how’s the weather down there, where you live?” This question, presented as part of a private aside, was introduced to draw subjects’ attention to the weather as a plausible transient cause of their current mood. As expected, this manipulation eliminated the otherwise observed difference in life-satisfaction: Once subjects realized that they may feel bad because of the rainy weather, they did not draw on their feelings in evaluating their life satisfaction. Subjects’ reports of their current mood, on the other hand, were not influenced by the weather question, indicating that this question did not change their current feelings but only the perceived informational value of these feelings. Similarly, recalling a sad life event did not influence subjects’ judgments of life satisfaction when they could misattribute the resulting sad feelings to the alleged impact of the experimental room (Experiment 1). In addition, current mood, as assessed at the end of the experiment, was more strongly correlated with judgments of life satisfaction when subjects’ attention was not directed to a transient source of their feelings than when it was. Conceptual replications of these findings have been reported by Keltner, Locke, and Audrain (1993); Schwarz, Servay, and Kumpf (1985); Savitsky, Medvec, Charlton, and Gilovich (1998); and Siemer and Reisenzein (1998), among others.

In combination, these findings indicate that individuals may use their current feelings as a basis of judgment unless the diagnostic value of their feelings for the judgment at hand is called into question. This discounting effect (Kelley, 1972) is incompatible with predictions based on mood-congruent recall because the attributional manipulations only discredit the implications of one’s current feelings but not the implications of valenced information about one’s life, recalled from memory: Attributing one’s bad mood to the weather, for example, only undermines the informational value of one’s mood, but does not affect the implications of a recalled negative event, such as failing an exam. Given these findings, it is not surprising that mood effects on evaluative judgments have been observed in the absence of any evidence for mood effects on the recall of relevant information from memory (e.g., Fiedler, Pampe, & Scharf, 1986).

Relative Contributions of Mood and Thought Content

In addition, the predictions of both models differ with regard to the contributions of mood at the time of judgment and the specific content by which the mood is induced. From the perspective of an associative network model of memory, judgments should be most consistent with a person’s current mood when the mood is induced by a technique that itself increases the accessibility of mood-congruent material that is relevant to the judgment at hand. For example, a depressed mood that is induced through thoughts about a serious disease should affect judgments about diseases more strongly than a depressed mood that is induced by other thoughts, because information about diseases would be activated by both the content of one’s thoughts and one’s depressed mood. According to the feelings-as-information hypothesis, however, the nonemotional content of the mood-inducing stimulus should be irrelevant, unless it influences the apparent informational value of the accompanying feelings.

In line with the latter hypothesis, several studies indicate that mood effects on evaluative judgments are largely independent of the specific content by which the mood was induced (e.g., Johnson & Tversky, 1983; Mayer, Gaschke, Braverman, & Evans, 1992). For example, Johnson and Tversky (1983) observed that reading descriptions of negative events, which presumably induced a depressed and slightly anxious mood, increased judgments of risk across a wide set of targets. Most important, the impact of mood was independent of the object of judgment or the content by which the mood was induced. Reading about cancer, for example, affected judgments of the risk of cancer, but had equally strong effects on judgments of the risk of accidents and divorce. Such thoroughly generalized effects, undiminished over dissimilar content domains, are incompatible with models of mood congruent recall. However, they are consistent with the feelings-as-information hypothesis. From this perspective, participants presumably simplified the difficult task of evaluating unknown risks by consulting their depressed and anxious feelings, resulting in judgments of higher risk.