Running head: BEYOND VALENCE
Beyond valence: Toward a model of
emotion-specific influences on judgment and choice
Jennifer S. Lerner Dacher Keltner
Carnegie Mellon University and University of California, Berkeley
University of California, Los Angeles
Correspondence address through June, 1999
Jennifer S. Lerner
Psychology Department, UCLA
1285 Franz Hall, Box 951563
Los Angeles, CA 90095-1563
Phone: (310) 206-9651
Facsimile: (310) 206-5895
Most theories of affective influences on judgment and choice take a valence-based approach, contrasting the effects of positive versus negative feeling states. These approaches have not specified if and when distinct emotions of the same valence have different effects on judgment. In this article, we propose a model of emotion-specific influences on judgment and choice. We posit that each emotion is defined by a tendency to perceive new events and objects in ways that are consistent with the original cognitive-appraisal dimensions of the emotion. To pit the valence and appraisal-tendency approaches against one another, we present a study that addresses whether two emotions of the same valence but differing appraisals -- anger and fear -- relate in different ways to risk perception. Consistent with the appraisal-tendency hypothesis, fearful people made pessimistic judgments of future events whereas angry people made optimistic judgments. In the discussion we expand the proposed model and review evidence supporting two social moderators of appraisal-tendency processes.
Beyond valence: Toward a model of
emotion-specific influences on judgment and choice
Once an exclusively cognitive enterprise, research on judgment and choice increasingly addresses the powerful influence of affect (for reviews, see Bodenhausen, 1993; Bower, 1991; Clore, Schwarz & Conway, 1994; Forgas, 1995; Loewenstein, 1996; Schwarz & Clore, 1996; Zajonc, 1998). Despite the recent flowering of research on affect and cognition, relatively few theories have systematically addressed the influences of specific emotions upon judgment and choice. Rather, the majority of studies in this tradition have been motivated by a valence-based approach, contrasting the effects of positive versus negative feeling states on judgment and choice (for reviews reaching this conclusion, see Elster, 1998; Forgas, 1995; Higgins, 1997). As a result, questions about whether and how different emotions of the same valence -- such as anger, fear, and sadness -- influence judgment and choice remain largely unaddressed (for notable exceptions, see Bodenhausen, Sheppard & Kramer, 1994; Niedenthal, Covert & Barlow, 1998; Weiner, Graham & Chandler, 1982).
The purpose of the present article is to present an emotion-specific framework for studying affective influences on judgment and choice. To do so, we first briefly review valence approaches to affect and judgment. We then draw upon cognitive-appraisal theories of emotion to outline how specific emotions influence judgment and choice. To pit these two approaches against one another, we present a study that addresses whether two emotions of the same valence but differing appraisals -- anger and fear -- are differentially related to risk perception. Our concluding section addresses theoretical implications of the proposed framework.
Valence-Based Approaches to the Study of Affect, Judgment, and Choice
Researchers have been concerned with two general kinds of affective influences upon judgment and choice. Studies of integral affect document the influences of subjective experiences that are relevant to present judgments and choices. For example, anticipated regret when evaluating a gamble has been shown to influence how much one is willing to gamble (Larrick & Boles, 1995; Loomes & Sugden, 1982; Mellers, Schwartz, Ho & Ritov, 1997). Studies of incidental affect -- the concern of this paper -- focus on the sometimes puzzling influence of subjective emotional experiences that should be irrelevant to present judgments and choices. For example, affect produced watching movies, enjoying sunny weather, or experiencing stressful exams has been shown to influence judgments of unrelated topics and objects (for reviews, see Bodenhausen, 1993; Clore et al., 1994; Forgas, 1995; Forgas & Bower, 1988; Schwarz, 1990; Schwarz & Clore, 1996).
Studies of incidental affect have examined both direct and indirect mediational mechanisms (Forgas, 1995). Indirect mechanisms include affect-related cognitive processes that influence subsequent judgments. For example, people will selectively retrieve mood-congruent information from memory and then use that information in unrelated judgments (Bower, 1981; Bower, 1991; Isen, Shalker, Clark & Karp, 1978). According to associative network models, this process explains why people in good moods make optimistic judgments and people in bad moods make pessimistic judgments (Kavanagh & Bower, 1985; Wright & Bower, 1992).
More direct influences of incidental affect upon judgment have been summarized in the affect-as-information model. According to this model, people rely on their present feelings in heuristic fashion to make complex judgments, as long as the experienced feelings are perceived as relevant to the object of judgment (Clore, 1992; Schwarz, 1990; Schwarz & Clore, 1983). For example, when asked to rate overall life satisfaction, participants do not go to the trouble of calculating estimates on a number of life dimensions; they simply ask themselves, how am I feeling? Participants in a positive mood give higher ratings of life satisfaction than participants in a negative mood. Importantly, if participants attribute their feelings to a source that is irrelevant to the judgment at hand (e.g., the current weather), the feelings are no longer considered informative, and exert little or no influence on judgment (Keltner, Locke & Audrain, 1993; Schwarz, 1990; Schwarz & Clore, 1983).
These various research traditions, although differing in claims about mediating mechanisms, share a common feature. They base predictions on the valence of the affect (Forgas, 1995, p. 61). Positive and negative moods are experimentally induced or observed naturalistically, and these general feeling states are expected to lead to more positive or negative judgments respectively. Indeed, readers of the affect-judgment literature could easily conclude that "the only relevant aspect of the emotions is their valence" (Elster, 1998, p. 64, emphasis added).
Valence-based approaches face one obvious shortcoming, however. They fail to specify whether different emotions of the same valence differentially influence judgments and choices. In fact, given the centrality of valence to emotion, valence-based approaches might by default predict that distinct emotions of the same valence, such as sadness, anger, and fear, would exert similar influences on judgment and choice. Yet this general prediction immediately encounters intuitive counterexamples (e.g., one would expect a highly fearful leader to make different decisions than an angry one). This general valence-based prediction is also out of step with current research on emotion, which indicates that emotions of the same valence differ in their antecedent appraisals (Smith & Ellsworth, 1985), facial expressions (Keltner & Ekman, in press), autonomic physiology (Levenson, Ekman & Friesen, 1990), and central nervous system physiology (Panksepp, 1982). Valence-based approaches may sacrifice specificity in the service of parsimony (c.f. Higgins, 1997). To assess the significance of this sacrifice, research needs to examine whether specific emotions of the same valence differentially influence judgment and choice outcomes.
The Appraisal-Tendency Approach: Emotion-Specific Influences on Judgment and Choice
How might specific emotions influence judgment and choice? Two broad theoretical approaches provide a framework for answering this question: cognitive-appraisal theories of emotion and functional (evolutionary) theories of emotion. From cognitive-appraisal theories we borrow the idea that a range of cognitive dimensions (rather than just valence) usefully differentiates emotional experience and effects.
Of the different accounts of cognitive-appraisal processes (e.g., Lazarus, 1991b; Ortony, Clore & Collins, 1988; Roseman, 1984; Scherer, 1988; Smith & Ellsworth, 1985; Weiner; 1980;1986), we draw most directly on that of Smith and Ellsworth (1985) to make predictions concerning the influences of specific emotions upon judgment. Through empirical examination of all appraisal dimensions identified in the literature, Smith and Ellsworth (1985) identified the six cognitive dimensions that best define the patterns of appraisal underlying different emotions: certainty, pleasantness, attentional activity, control, anticipated effort, and responsibility. In their research, participants recalled past emotional experiences and rated the emotion-eliciting events along these six dimensions of appraisal (e.g., Smith & Ellsworth, 1985). Each emotion was found to be defined by central dimensions, which characterize its core meaning or theme (Lazarus, 1991a; Smith & Ellsworth, 1985). For example, certainty, control, and responsibility are the central dimensions which distinguish anger from other negative emotions. Anger arises from appraisals of: (a) other-responsibility for negative events, (b) individual control, and (c) a sense of certainty about what happened (Averill, 1983; Betancourt & Blair, 1992; Smith & Ellsworth, 1985; Weiner et al., 1982).
From functional approaches to emotion, we borrow the idea that emotions serve an impressive coordination role; they trigger a set of responses (physiology, behavior, experience, and communication) that enable the individual to deal quickly with encountered problems or opportunities (Frijda, 1986; Levenson, 1994; Oatley & Johnson-Laird, 1996). Of particular importance, emotion-related cognition interrupts ongoing cognitive processes and directs attention, memory, and judgment to address the emotion-eliciting event (Johnson-Laird & Oatley, 1992; Lazarus, 1991a; Schwarz, 1990; Simon, 1967; Tooby & Cosmides, 1990). Interestingly, an emotionís ability to focus cognition may be so strong that the emotion not only directs thoughts relevant to the initial emotion-eliciting event but also to unrelated events. For example, anger triggered in one situation automatically elicits blame cognitions in other situations (Quigley & Tedeschi, 1996).
Appraisal tendencies. Drawing on evidence that each specific emotion (a) is defined by a set of central dimensions and (b) directs cognition to address specific problems or opportunities, we hypothesize that each emotion activates a cognitive predisposition to appraise future events in line with the central-appraisal dimensions that triggered the emotion -- what we call an appraisal tendency. In short, appraisal tendencies are goal-directed processes through which emotions exert effects upon judgment and choice until the emotion-eliciting problem is resolved.
We believe that appraisal-tendency processes apply to the effects of both momentary and dispositional emotions. Whereas dispositional emotions refer to the tendency to react with specific emotions across time and situations, momentary emotions refer to immediate affective reactions to a particular target (Gross, Sutton & Ketelaar, 1998; Larsen & Ketelaar, 1991; Lazarus, 1994; Malatesta, 1990). Recent empirical research indicates that dispositional emotion resembles momentary emotion in important ways, and thus, should yield similar effects upon judgment. For example, people dispositionally prone to fear report experiencing more fear at a variety of points in time and across situations (Gross et al., 1998), they report higher levels of state fear in response to negative affect inductions (Gross et al., 1998), and they display more fear in the face (Keltner, 1996).
Some initial evidence supports the appraisal-tendency proposal. For example, incidental anger increases tendencies to perceive other individuals as responsible for subsequent events (Keltner, Ellsworth & Edwards, 1993), and to make punitive judgments of other individuals, both related and unrelated to the original source of anger (Goldberg, Lerner & Tetlock, in press; Lerner, Goldberg & Tetlock, 1998). Incidental sadness, in contrast, increases the tendency to perceive situational factors (such as fate or circumstances) as responsible for ensuing events (Keltner et al., 1993). These emotional carryover effects are consistent with the underlying appraisal patterns of each emotion. Although both anger and sadness are highly negative, anger arises from appraisals of individual control of negative events whereas sadness arises from appraisals of situational control of negative events.
For attempts to gather further evidence, the appraisal- tendency model points to a clear empirical strategy: research should compare emotions that are highly differentiated in their appraisal themes on judgments/choices that relate to that appraisal theme. For example, because the responsibility dimension shares a conceptual theme with blame, researchers interested in studying integral or incidental emotion effects on blame could contrast emotions on opposite poles of the responsibility dimension, such as shame (self-responsibility) and anger (other-responsibility) (see Weiner et al., 1982). Implicit in this strategy is the idea that emotions of the same valence should sometimes influence judgment in opposite ways -- a proposition that contradicts predictions from valence accounts and therefore provides a useful point for comparing valence and appraisal-tendency approaches.
To illustrate this appraisal-tendency approach, Figure 1 compares predictions for the influences of two negative emotions -- fear and anger -- on risk perception (left side) and two positive emotions -- surprise and pride -- on attribution (right side). In the top panel of the figure, the left column contains six cognitive-appraisal dimensions (e.g., certainty) that differentiate emotions (see Roseman, 1984; Smith & Ellsworth, 1985). For each of the six dimensions, entries indicate the relative position of each emotion (for precise scale values of each emotion on the relevant dimension, see Smith & Ellsworth, 1985). If an emotion is relatively high or low on a given dimension, the dimension is considered central to the definition of that emotion and likely to exert influences on subsequent judgments or choices. In the middle panel, entries indicate the appraisal tendency that is likely to be associated with each emotion. Finally, in the bottom panel, entries indicate predictions for emotion influences on the outcome of interest.
As illustrated in the left side of the figure, fear is defined by three central appraisal themes that are conceptually related to risk perception: uncertainty, unpleasantness, and situational control (e.g., Lazarus, 1991a; Smith & Ellsworth, 1985). Drawing on fearís appraisal structure, the model predicts that fear will be associated with the tendency to perceive uncertainty and situational control in new situations and that fearful people will -- as a consequence of that appraisal tendency -- perceive greater risk across new situations. Anger, by contrast, will be associated with the tendency to perceive certainty and individual control in new situations and -- as a consequence -- to perceive less risk across new situations. As illustrated in the right side of the figure, pride is defined by the central appraisal themes of self-responsibility and pleasantness. The model predicts that pride will therefore be associated with the tendency to perceive the self as responsible for positive events, even in new situations. Surprise, by contrast, will be associated with the tendency to perceive others as responsible, even in new situations.
The above predictions are only a small sample of the ways in which an appraisal-tendency perspective systematically links specific emotions to specific judgment and choice outcomes. In the discussion section, we provide some additional examples of this approach. In the study below, we provide an initial test of the differential predictions for fear and anger on risk perception.
An Empirical Test of the Valence and Appraisal-Tendency Perspectives:
The Influences of Fear and Anger on Risk Perception
Risk perception has been the focus of several valence-based studies of affect and judgment (e.g., Johnson & Tversky, 1983; Wright & Bower, 1992), making it an interesting outcome variable on which to compare appraisal and valence-based predictions. In one of the most widely-cited studies of affect and risk perception, Johnson and Tversky (1983) gave participants newspaper stories designed to induce positive or negative affect and then asked participants to complete a risk questionnaire. The risk questionnaire listed a variety of potential causes of death (e.g., heart attack) and instructed participants to estimate the annual number of fatalities associated with each potential cause. Consistent with a valence-based approach, participants who received the negative-mood induction offered more pessimistic estimates (i.e., they estimated higher frequencies of death) than participants who received the positive-mood induction.
To derive predictions for the influences of specific emotions on risk perception, we followed the appraisal-tendency strategy briefly outlined above. First we identified appraisal dimensions that are conceptually related to risk perception. We predicted that differences in the certainty and control dimensions would influence risk perception because these dimensions map directly onto the two cognitive meta-factors in the risk literature that reliably determine risk assessments: "unknown risk" (defined at the high end by hazards judged to be uncertain), and "dread risk" (defined at the high end by perceived lack of individual control) (McDaniels, Axelrod, Cavanagh & Slovic, 1997; Slovic, 1987; Slovic, 1994; Slovic, Fischhoff & Lichtenstein, 1986). We then selected emotions that fall at opposite ends of the certainty and control dimensions, namely fear and anger. As previously noted, fear arises from appraisals of profound uncertainty -- a sense that even such basic needs as safety are uncertain -- as well as appraisals of situational control -- a sense that factors beyond oneís control shape outcomes (Smith & Ellsworth, 1985). By contrast, anger arises from appraisals of certainty and individual control. Finally, we chose to investigate the influence of dispositional fear and anger because some evidence supporting an appraisal-tendency perspective already exists for momentary emotions (see Keltner et al., 1993). In sum, to pit the valence and appraisal-tendency perspectives against one another, we compared the risk assessments made by fearful and angry individuals.
Competing Research Hypotheses
According to valence approaches, negative feelings lead to negative judgments. Fearful and angry people should, therefore, make relatively pessimistic risk assessments (Johnson & Tversky, 1983; Wright & Bower, 1992). According to the appraisal- tendency hypothesis, an emotion's underlying appraisal theme dictates its influence upon subsequent judgments. Whereas the sense of uncertainty and situational control that defines fear should lead fearful people to make relatively pessimistic risk assessments, the sense of certainty and individual control that define anger should lead angry people to make relatively optimistic risk assessments. In sum, if the valence approach is correct, then both fear and anger will be positively related to pessimistic risk assessments, as defined by making higher frequency estimates for deaths. If the appraisal-tendency approach is correct, then only fear will be positively related to pessimistic risk assessments; anger will be negatively related to pessimistic risk assessments.
Participants and procedural overview. Ninety-seven undergraduates (28 males, 69 females) at the University of California participated in return for course credit. Participants expected to participate in several short, unrelated studies. Specifically, they were told that in order to make use of the full hour available different researchers had pooled together their respective questionnaire packets. The first packet, a "Self-Evaluation Questionnaire," contained measures of baseline state emotions and dispositional emotions. After completing the packet, participants received a separate questionnaire containing the dependent measure (risk perception) followed by a variety of filler questionnaires on unrelated topics (e.g., potential causes for various events). Following completion of all packets, participants were fully debriefed.
Measures of dispositional fear and anger. Participants completed two measures that assess dispositional fear. First, they completed a 12-item version of the Fear Survey Schedule-II, which assesses the degree of fear, if any, participants feel in response to 12 specific situations or objects (e.g., enclosed places, snakes) (see Bernstein & Allen, 1969; Geer, 1965; Suls & Wan, 1987). Participants made their assessments on a Likert-scale that ranged from 0 (none) to 4 (terror). Second, participants completed Spielberger's (1983) 20-item trait-anxiety scale, which assesses the frequency with which participants feel "nervous" or "anxious" on a Likert-scale that ranged from 1 (almost never) to 4 (almost always). The Pearson correlation between these two scales was reasonably high (r =.57, p <.01). To combine the two measures into one composite index of dispositional fear, we used principal components analysis and imposed a one-factor solution that retained all items (Eigenvalue = 10.20). We then calculated regression-factor scores for each participant. The composite dispositional-fear scale achieved an alpha-level of .91.
Participants also completed two measures of dispositional anger. First, participants completed Spielberger's (1996) 10-item trait-anger scale, which assesses tendencies to react with sudden and intense anger to a variety of life situations. Participants made these assessments on a Likert-scale which ranged from 1 (almost never) to 4 (almost always). Second, participants completed a 10-item face-valid anger scale that was written for this study; it addressed the chronic tendency to experience various forms of less intense anger. For each of 10 statements describing various kinds of chronic anger, participants were asked to indicate the extent to which the description was "true of them." Response options on a Likert-scale ranged from 1 (not at all true of me) to 7 (very true of me). The Pearson correlation between the two measures of dispositional anger was reasonably high (r = .70, p <.01). To combine the two measures into one composite index of dispositional anger, we used principal-components analysis and imposed a one-factor solution that retained all items (Eigenvalue = 6.53). We then calculated regression factor scores for each participant. The composite anger scale achieved an alpha-level of .81.
Measures of state affect. We have proposed that the appraisal tendency hypothesis applies to dispositional and momentary (state) emotions because we assume that a close correspondence exists between these two kinds of emotion. To address whether dispositional emotion does predict state emotion, which we have implicitly assumed, we assessed baseline- state emotions among participants at the start of the study. We measured state-fear with Spielberger's (1983) state-anxiety scale, which consists of twenty statements that evaluate the extent to which respondents feel anxious (i.e., tense, frightened and worried) "right now." We assessed baseline state-anger with Spielberger's (1996) state-anger scale, which consists of ten statements that evaluate the extent to which respondents feel intensely angry (e.g., furious, burned up, like breaking things) "right now." For both scales, response options ranged on a Likert scale from 1 (not at all) to 4 (very much so).
Risk perception. Participants completed Johnson and Tversky's (1983) "Perception of Risk Questionnaire," which presented participants with 12 events that lead to a certain number of deaths each year in the United States (e.g., brain cancer, strokes, floods). The measure asked participants to estimate the number of annual fatalities due to each event, based on the knowledge that 50,000 people in the United Stated die in car accidents each year. Following procedures from Johnson and Tversky, participants were also instructed to: (a) be as accurate as possible, (b) check their answers for consistency, and (c) feel free to erase and change answers to make the relative frequencies of the entire set consistent with their best opinions.
Preliminary analyses. Before testing our hypothesis, we conducted two preliminary analyses. First, we assessed the relationship between the two composite emotion dispositions. Consistent with the fact that fear and anger share a common valence, a significant correlation emerged between the composite dispositional scales for fear and anger (r =.48, p < .05). This correlation implied that inferential analyses would need to control for the influence of one emotion to ascertain the independent relationship between the other emotion and risk perception.
A second preliminary analysis addressed variance in the risk assessment measure. As in Johnson & Tversky (1983), frequency estimates spread across several orders of magnitude and produced skewed distributions. Following Johnson & Tversky's procedure, we calculated a logarithmic transformation of the data and then submitted the transformed scores to a confirmatory principal components analysis. This procedure generated a normally distributed composite factor for risk perception (alpha = .86).
Inferential analyses. Recall that a valence approach to affect and judgment predicted that fear and anger would be positively related to pessimistic risk assessments. An appraisal tendency perspective, in contrast, predicted that fear would be positively related to pessimistic risk assessments, and anger, in spite of its negative valence, would be negatively related to pessimistic risk assessments.
To ascertain the independent influence of each emotion disposition on judgment, we simultaneously entered each emotion disposition in one regression equation with the measure of perceived risk as the outcome measure. Figure 2 presents the results of this regression analysis, which supported the appraisal tendency hypothesis rather than the valence hypothesis. Fear was positively related to perceived risk, t (94) = 2.39, p < .05, and anger was negatively related to perceived risk, t (94) = -2.00, p < .05. Although fear and anger are both negative emotions, they exerted unique influences on judgments -- systematically shaping risk perception in a manner consistent with their underlying appraisal structures.
We also tested our assumption concerning the relation between dispositional and state emotion. Recall that if the appraisal-tendency model applies to both dispositional and momentary emotion, a systematic correspondence should exist between these two kinds of emotion. Specifically, dispositional emotion should predict emotion states across time and situations (see Gross et al., 1998). Consistent with this prediction, high scores (more than one standard deviation above or below the mean) on the dispositional emotion scales predicted participants' baseline state emotions at the beginning of the study. Participants low in dispositional fear felt less baseline fear than did participants high in dispositional fear, respective means = -1.01 and 1.24; t (33) = -8.65, p <.05. Similarly, participants low in dispositional anger felt marginally less baseline anger than did participants high in dispositional anger, respective means = -.28 and .11; t (23) = -1.39, p = 09. Finally, the sex of participants did not qualify any of the findings; we observed the same patterns for males and females.
The present study assessed the relative merits of two approaches to the study of affect and judgment. One widely influential approach assumes that valence constitutes the basis for predicting influences of affective states upon judgment (see Bower, 1981; Bower, 1991; Isen et al., 1978). The valence approach predicted that fear and anger would have similar influences on judgment, both leading to pessimistic risk perception (see Johnson & Tversky, 1983; Wright & Bower, 1992). An appraisal tendency approach assumes that underlying appraisal themes define the influences of different emotions upon judgment. Because anger and fear sharply diverge on appraisals of uncertainty and control, they should exert different influences upon risk assessments. Whereas fear (defined by great uncertainty and situational control) should predict pessimistic assessments, anger (defined by certainty and individual control) should predict optimistic assessments.
Consistent with the appraisal-tendency view, fearful and angry individuals indicated strikingly different assessments of the level of risk in the environment: fear predicted higher risk assessments; anger predicted lower risk assessments. Notably, dispositional fear and anger led to different risk assessments in spite of the fact that they are both high in negative valence and involve heightened sympathetic autonomic nervous system arousal (Levenson et al., 1990). In combination with previous evidence indicating that sadness and anger influence causal attributions in highly distinct ways (Bodenhausen, et al., 1994; Keltner et al., 1993), the present study suggests that negative emotions are likely to influence a variety of judgments in highly differentiated ways.
The present study is significant in one other general way. The present study may be the only study to date that compares the influence of two dispositional emotions on judgment and choice among healthy, non-disordered participants. Almost all studies of dispositional affect have either: (a) treated self-reports of emotion as the outcome measure (e.g. Larsen & Ketelaar, 1991; Watson & Clark, 1984; Watson & Tellegen, 1985), (b) addressed affective influences on cognition in disordered, clinical samples (e.g., Butler & Mathews, 1983; Rapee, 1986), or (c) focused on only one kind of dispositional affect (e.g. Butler & Mathews, 1987; Mathews, 1990). Given recent findings that emotion dispositions are: (a) reflected in relatively stable differences in underlying neurochemical systems (Davidson, 1998), (b) heritable (Gabbay, 1992), and (c) stable across the life course (Helson & Klohnen, 1998), it is increasingly important to systematically link differences in dispositional emotion to the extensive judgment and choice literature.
Limitations and implications for future research. The present study has certain limitations and raises questions that warrant further research. First, our study could not test possible mechanisms of the influence of dispositional fear and anger upon risk assessment. Fearful and angry people, given their baseline differences in state affect, may have recalled different memories in making their risk estimates, as affect- priming theories might argue (Bower, 1981; Bower, 1991). Another possibility, perhaps complementary to the appraisal-tendency view, is that fearful and angry people used their current sense of certainty and control as information in making assessments, as the affect-as-information perspective might suggest (Schwarz, 1990; Schwarz & Clore, 1983). Future research should test these explanations for the differing risk assessments of fearful and angry people.
Although the present study sought to test competing hypotheses for the relationship between emotion dispositions and risk assessment, and did not seek to test causal paths, a potential third-variable cause merits note. If fearful people have actually experienced greater levels of risk in their lives than angry people, this could then influence judgments of future risk. We should note, however, that the evidence indicates that it is in fact anger-prone people who lead risky lives rather than fear-prone people (see Caspi, Elder & Bem, 1987; Heaven, 1994; Leith & Baumeister, 1996; Pfefferbaum & Wood, 1994; Wills, Vaccaro & McNamara, 1994). Indeed, the present evidence suggests that the tendency for angry people to take risks and behave recklessly may be partially mediated by a systematic misperception of risk. It is important, nevertheless, to generalize the findings from the present study to studies that manipulate momentary feelings of fear and anger, to more clearly establish causal relations between emotion and judgment.
Finally, the present study raises the intriguing question of whether dispositional and momentary emotions exert different or similar influences upon judgment. We have argued that the effects of momentary and dispositional emotion upon judgment are analogous in content, but we offered no claims about the magnitude of such effects. One might argue that because momentary emotions are likely to be more intense than dispositional emotions, they would exert greater influences upon judgment. This simple notion encounters certain problems. First, momentary emotions are likely to be consciously linked to a cause of emotion, which reduces its effects upon judgments of other objects and events (Schwarz, 1990). Second, several theorists have speculated that dispositional emotions play a larger role in shaping judgment and choice than do on-line state emotions, because dispositional emotions emerge early in life, remain stable over the life course, and function as chronic schemas for organizing and interpreting events (Damasio, 1994; Gasper & Clore, 1998; Malatesta, 1990).
In our remaining discussion, we focus on two general issues. First, we will briefly address how other emotions might influence other judgment and choice domains. Second, we will consider potential boundary conditions for the influence of emotion-related appraisal tendencies on judgment and choice.
Applying an Appraisal-Tendency Framework to Other Specific Emotions
The benefits of systematic comparisons between different emotions extend beyond simply pointing out limitations to the valence-based approaches. By illuminating the cognitive processes associated with different emotions, they also bring emotion into the study of judgment and decision making in systematic ways. The appraisal-tendency approach provides a flexible yet specific framework for developing a host of testable hypotheses concerning affect, judgment, and decision making.
To date, only two judgment domains have been explored from an appraisal-tendency perspective: the effects of specific negative emotions upon causal attributions (Keltner et al., 1993) and risk assessments (the present study). Research examining the effects of other emotions upon other kinds of judgments will illuminate the more general role of emotion in judgment and decision making, and may lead to refinements of previous hypotheses. We have already outlined differential predictions for emotions of the same valence on attributions. Many other differential predictions may also be derived from the appraisal-tendency model. For example, our analysis suggests that incidental anger and sadness -- two negatively valenced emotions -- should exert different effects on unrealistic illusions of control (see Langer, 1975). Whereas anger (characterized by attributions of personal control) should amplify this illusion, sadness (characterized by attributions of situational control) should attenuate this illusion. Indeed, evidence from depressed and nondepressed individuals supports the idea that depressed individuals are "sadder but wiser", in that they are less likely to overestimate their control over outcomes (Alloy, Abramson & Viscusi, 1981). As another example, the appraisal-tendency model suggests that incidental desire and disgust should exert different effects on subsequent motivation to pursue a task. Whereas desire (characterized by devout attention to a person or object) should increase attention to a task, disgust (characterized by strong unwillingness to attend to a person or object) should decrease attention given to a task.
Boundary Conditions for the Influences of Emotion Upon Judgment and Choice
Studies of the influences of affect upon judgment inevitably raise the question of boundary conditions: When do people make judgments or choices independent of their current emotion? This question has motivated centuries of philosophical discussion concerning the interplay between passion and reason, and more recently, several important lines of research (e.g., Schwarz, 1990). An appraisal-tendency perspective points to at least two kinds of social moderators of the influence of emotion upon judgment: goal attainment and cognitive awareness processes.
Drawing on the idea that emotions guide specific judgments and perceptions to respond to significant problems or opportunities , the goal-attainment hypothesis asserts that appraisal tendencies will be deactivated when an emotion-eliciting problem is solved or opportunity responded to, even if the emotion persists experientially (see Frijda, 1988). Consistent with this hypothesis, a recent study found that anger led to increased punitive judgments of unrelated cases, but only when the perpetrator of the original anger-inducing crime went unpunished due to a technicality (Goldberg et al., in press). If the perpetrator of the crime had been punished, and the goal of anger served, the emotion no longer influenced subsequent judgments. Researchers may develop and test similar goal-attainment hypotheses by drawing on the appraisal literature, which has identified essential goals of emotions (see, Lazarus, 1991).
Drawing on the idea that initial emotion-related appraisals are automatic in nature (Ekman, 1992; Lazarus, 1991a; LeDoux, 1996), the cognitive-awareness hypothesis asserts that appraisal tendencies will be deactivated when individuals become aware of their own judgment process. Specifically, conscious monitoring of oneís judgment process will lead individuals to focus on judgment-relevant information and discount such judgment-irrelevant information as incidental affect. Several recent studies support this claim. The tendency for incidental happiness to increase reliance on stereotypes was attenuated when participants expected to be accountable for their judgments (Bodenhausen, Kramer & Süsser, 1994). In another study, accountability attenuated the tendency for incidental anger to increase punitiveness in unrelated tort cases (Lerner et al., 1998). In both studies, accountable participants discounted their present feelings as a function of increased attention to their judgment process. These results are consistent with evidence that certain kinds of accountability encourage individuals to carefully scrutinize the relevance of any cues used in forming an opinion (see Lerner & Tetlock, in press).
In this article, we have addressed two questions. How do specific emotions influence different judgments? And what social factors moderate the influences of different emotions upon judgment? Concurring with Forgas' (1995, p. 61) conclusion that "appraisal theories present a rich and largely untapped source of hypotheses about the judgmental consequences of affect," we have drawn upon the appraisal literature to propose that emotions activate appraisal tendencies, which are relatively automatic processes that guide subsequent perception and judgment. This approach generated specific predictions concerning how and when specific emotions influence different judgments. Moreover, an initial test of this approach involving the influence of dispositional fear and anger on risk perception proved it to be a better predictor of outcomes than the historically dominant valence approach. Our hope is that the appraisal-tendency approach outlined here will encourage research addressing the systematic influences of specific emotions on judgment and choice and the social factors that moderate those influences.
For helpful comments on the risk perception study, we thank Robert MacCoun and Christina Maslach; for helpful comments on the manuscript, we thank the members of Shelley Taylorís research group -- particularly Regan Areesh Raj Gurung, Laura Klein, Sonja Lyubormirsky, Greg Gold and Inna Rivkin. Preparation of this paper was partially supported by a National Institute of Mental Health postdoctoral fellowship at UCLA (MH15750).
Correspondence may be sent to Jennifer Lerner at the Department of Social and Decision Sciences, Carnegie Mellon University, Pittsburgh, Pennsylvania, 15213-3890. Electronic mail may be sent via the Internet to: email@example.com.
1 Although this special issue addresses affect and decision-making, we also address affective influences on judgment because decisions typically depend on initial judgments about a given situation.
2 Appraisal-tendencies are conceptually related to what Frijda (1986, p. 70) calls "action tendencies." Whereas action-tendencies are "states of readiness to execute a given kind of action," appraisal tendencies are the perceptual processes through which emotions color the interpretation of stimuli. The appraisal-tendency sequence (appraisal-emotion-appraisal tendency) is also conceptually related to what Weiner (1980; 1986) calls an emotion-attribution-action sequence. Whereas Weinerís sequence primarily addresses attributional differences among emotions, the appraisal-tendency sequence can address all cognitive differences among emotions.
3 Although Figure 1 focuses on comparing the effects of two negative emotions on risk perception and two positive emotions on attribution, it is worth noting that anger should increase attributions to others and surprise should increase perceptions of risk.
4 Risk researchers have found that frequency, rather than probability, estimates are superior indicators of risk perception (Lichtenstein, Slovic, Fischhoff, Layman & Combs, 1978).
5 Even extensive demand awareness interviews with participants have shown that this cover story elicits little suspicion among participants (Goldberg et al., in press; Lerner et al., 1998).
6 The ten items in this measure were: I rarely get pissed off at my friends; I am often mad at someone or something; I often find myself feeling angry; I am rarely frustrated by other people; I often blame others before blaming myself; A lot of people annoy me; I get mad easily; It s rare for me to get enraged; Other drivers on the road infuriate me; I d like to tell people how much they piss me off.
7 Follow-up interviews with the research participants indicate that this finding should not be interpreted as a purely cognitive effect, wherein people react more punitively in the tort cases because they deem harsher punishment as a rational response in a world where violent perpetrators evade punishment. Rather, participants firmly believed that judgments they made in the tort cases were uninfluenced by the justice feedback in the ostensibly separate study.
Figure 1. Two illustrations of the appraisal-tendency approach, each comparing emotions that are highly differentiated in their central appraisal themes on a judgment that relates to those appraisal themes.
Figure 2. Whereas dispositionally fearful people make pessimistic assessments of future events, dispositionally angry people make optimistic assessments.
Illustration with Illustration with
negative emotions positive emotions
Anger Fear Pride Surprise
Certainty High Low Medium Low
Pleasantness Low Low High High
Attentional Medium Medium Medium Medium
Anticipated Medium High Medium Medium
Control High Low Medium Medium
Responsibility High Medium Low High
Appraisal Perceive negative Perceive negative Perceive positive Perceive positive
Tendency events as predictable, events as events as brought events as unpredictable
under human control, unpredictable & about by self & brought about by
& brought about under situational others
by others control
Influence on risk perception Influence on attribution
Relevant Perceive low risk Perceive high risk Perceive self Perceive others
Outcome as responsible as responsible
Notes. Certainty is the degree to which future events seem predictable and comprehensible (high) versus unpredictable and incomprehensible (low). Pleasantness is the degree to which one feels pleasure (high) versus displeasure (low). Attentional activity is the degree to which something draws oneís attention (high) versus repels oneís attention (low). Control is the degree to which events seem to be brought about by individual agency (high) versus situational agency (low). Anticipated effort is the degree to which physical or mental exertion seems to be needed (high) versus not needed (low). Responsibility is the degree to which someone or something other than oneself (high) versus oneself (low) seems to responsible. We refer interested readers to Smith & Ellsworth (1985) for comprehensive descriptions of each dimension and each emotionís scale values along the dimensions.