The Cognitive Structure of Emotions ANDREW ORTONY University of Minots at Urbana-Champaign GERALD L. CLORE University of Htinois at Urbana-Champaign ALLAN COLLINS. Bolt, Beranek, and Newman, Inc Cambridge, Massachuseus CAMBRIDGE UNIVERSITY PRESS Cambridge New York Port Chester Melbourne SydneyIniroduction ‘The Study of Emotion Emotion is one of the most central and pervasive aspects of human experi- ence, Normal people experience a wide range of emotions, from the quiet satisfaction of completing a relatively mundane task to the grief at the death of a loved one. Yet while emotions color, deepen, and enrich human experience, they can also cause dramatic disruptions in judgment and per- formance. Such disruptions can have profound and sometimes terrible con- sequences for individuals and society as, for example, in crimes of passion, suicides, and mental illness. This fact is clearly recognized by creators of literature, which thrives on the imagined emotions of its characters, The basic recipe is very simple: The writer describes a situation that read recognize as being important to a character in the sense that it has impor- tant implications with respect to the goals, standards, or attitudes that the character is known or assumed to have. Then, the character is portrayed as correctly of incorrecily construing the situation as good or bad relative to these goals or standards or attitudes, and typically is described as having, or is assumed to have, z valenced (i.c., @ positive or negative) reaction to the situation. Finally, the construal together with the reaction usually results in some sort of change in the character's judgment or behavior. Consider, for example, the main plot of Othello. We start with the assumption that the maintenance of Desdemona’s love and fidelity is important for Othello. He then (incorrectly) construes Cassio’s (presumed) actions as a threat to this goal and becomes consumed with anger and jealousy, The result is a dra- ‘matic deterioration in judgment and a correspondingly drastic action in which he kills both Desdemona and himself. As readers, a certain suspen: sion of disbelief is required, but only up to a point, The essential ingredi- cents have to be believable. If literature is a microcosm of the real world, it hhas to be recognizable as such. Itis apparent that writers can reliably produce in readers an awareness of a character's affective states by characterizing a situation whose construal is assumed to give rise to them. This suggests that writers use an implicit theory that individual emotions can be specified in terms of personal or interpersonal situational descriptions that are sufficient to produce them. Thus, writers do not always have to state what emotions a character is experiencing because if the described situation contains the eliciting condi- tions for a particular emotion, the experience of that emotion can be in- ferred. The fact that millions of readers, often over decades or even centu- ties, all infer similar emotions from the described situations suggests that this implicit theory cannot be too far wrong.4 The Cognitive Structure of Emotions If the eliciting conditions of an emotion are to be elfective. the experiene ing individual must encode the relevant situation in a particular way. In other words, ifan emotion such as distress is a reaction to some undesirable event, the event itself must be construed as undesirable, and because con- Struing the world is a cognitive process, the eliciting conditions of emotions embody the cognitive representations that result from such construals Perhaps one of the most obvious cases of the major contribution that ‘cognition through construals makes to emotion is afforded by the reactions of players and fans at sports events. When one observes the reactions of the players to the outcome of an important game (tor example, the final of the World Cup, or the NCAA basketball championship) it is clear that those on the winning team are elated while those on the losing team are devas- tated. Yet, in a very real sense, both the winners and losers are reacting to the same objective event. It is their consiruals of the event that are differ- cent, The victors construe it as desirable, the losers as undesirable, and itis these construals that drive the emotion system. The emotions are very real and very intense, but they still issue from cognitive interpretations imposed, oon external reality, rather than directly from reality itself. It is in this sense that we claim that there is an essential and profound cognitive basis for ‘emotions. Before we start, it is important to make clear that some emotions, (¢.8.. lisgust), involve much less cognitive processing and structure than others (c.g., shame). Interestingly, however, those that involve relatively little cognition usually have metaphorical analogs that involve much more, whereas the converse is not true. Thus the emotion that one might experi- ence in response to, for example. some totally inappropriate and unaccept- able social behavior might well be called “disgust.” On the other hand, itis \ifficult to imagine a cognitively impoverished analog of a social emotion Such as embarrassment. We should also make clear at the outset that our claim that emotions always involve some degree of cognition is not the ‘same as asserting that the contribution of cognition is necessarily conscious. To say that emotions arise from cognition isto say that they are determined by the structure, content, and organization of knowledge representations und the processes that operate on them, These representations and pro- cesses might sometimes be available to consciousness. but there is no rea son to suppose that they necessarily are so. Although we doubt that William James would have approved of our characterization of emotion in general, it i interesting to note that our claim that some emotions involve more cognition than others has a parallel in James's famous paper Whar és an emotion? (James, 1884). There. James restricted his discussion to emotions having “a distinct bodily expression’ Introduction 5 in which “a wave of bodily disturbance of some kind accompanies the perception of the interesting sights or sounds. or the passage of the exciting train of ideas. Surprise. curiosity, rapture. fear. anger, lust. greed, and the like, become then the names of the mental states with which the person is possessed” (p. 189). Emotions like these, James called the “standard emo- tions.” He seems to have considered the “standard emotions” to involve little or no cognition, arguing that “in advance of all experience of ele phants no child can but be frightened if he suddenly finds one trumpeting ‘and charging upon him” (p. 191). However, James acknowledged that there can be more complex emotion-inducing perceptions, ones which. in ‘modern terms, would have to be described as involving @ relatively high degree of cognition, such as events having to do with the violation of social conventions: “Most occasions of shame and many insults are purely conven- tional, and vary with the social environment.” In considering these as potential counterexamples to his theory, James goes on, rhetorically: “In these cases, at least it would seem that the ideas of shame, desire, regret, etc., must first have been attached by education and association to these conventional objects before the bodily changes could possibly be awak- ened” (p. 195), James's answer to this apparent threat is to assert that the nature of the emotior-inducing perception is not the issue: rather, the issue is that, once triggered, the perception gives rise to the bodily response that is the emotion, However, like it or not (and James is now in no position to object), James had essentially characterized a range of cognitive content for the emotion-producing perception from low (e.g. a mother's delight at the sight of her beautiful baby) to high (¢.g., the delight of receiving @ national honor) Modern theories of cognition have relatively little to say in the way of specific proposals about affeet and emotion (Norman, 1981; Zajonc, 1980), It is quite possible that the root cause of the dissociation between cognitive theories and emotion theories lies in the emphasis that has been placed in recent years on the computer metaphor of “human information process: ing.” This approach to cognition has been ss noticeable in its failure 10 ‘make progress on problems of affect as it has been for its success in making progress on problems of cognition. Given the abundance of psychological ‘evidence that cognitions can influence and be influenced by emotions (¢.8.. Bower, 1981; Isen, Shalker, Clark, & Karp, 1978: Johnson & Tversk} 1983; Ortony. Turner. & Antos. 1983; Schwarz & Clore, 1983). the ab- sence of a viahle account of the emotions compatible with a general theory ‘of cognition renders existing theories of both inadequate Just as few theories of cognition hase much to say about emotion, so theories of emotion tend to be unacceptably vague about evietly what role6 The Cognitive Sructure of Emotions cognition plays in emotion. Psychologists (e.g.. Arnold, 1960: Lazarus. Kanner, & Folkman, 1980; Mandler, 1975: 1984) and philosophers (e.g. Lyons. 1980) frequently acknowledge that cognition plays an essential role inemotion, but for the most part they have not provided detailed proposals about exactly how this happens (but see, e.g, Abelson, 1983; Mandler, 1984). One of the clearer accounts is that offered by Mandler, who claims that what he calls “cognitive interpretation” or “meaning analysis” (ie appraisal) is the “cold” part of emotion. The “heat” is provided by arousal, which aecording to Mangler. is normally oecasioned by the interruption of plans or action sequences. We find Mandler’s account more attractive than other arousalappraisal theories because of its specificity with respect to the appraisal aspect of emotion (see, for example, Mandler, 1982), and be- cause of its explicit recognition of the importance of plans, goals, and knowledge representations. However, Mandler's account has little to say bout specific emotions, especially positive ones, and it offers no system- atic account of the relation among different emotions Another problem with the arousal/appraisal theories is that they offer no secount of how arousal and appraisal interact to produce emotion. Our approach to this problem is to postulate an arousal-producing mechanism that, at the same time, registers valence. This obviates the need to postu late distinet mechanisms corresponding to arousal and appraisal, thus elimi nating the need to explain how such mechanisms interact for the ordinary experience of emotion. We believe such an approach is viable even though, under special circumstances, itis possible to produce one in the absence of the other (e.g., Schachter & Singer, 1962). Our initial discussion of these issues can be found in Chapter 3 From a global perspective, it seems that past research on emotion con- verges on only two generalizations. One is that emotion consists of arousal and appraisal (e.g., Arnold, 1960; Lazarus, Averill, & Opton, 1970; Mandler, 1975; Schachter & Singer. 1962). The other, emerging from the scaling literature (e.g... Abelson & Sermat, 1962; Block, 1957: Davitz, 1969; Engen, Levy. & Schlosberg, 1958; Russell, 1980), is that any dimen sional characterization of emotions is likely to include at least the 1w0 dimensions of activation and valence. But, on eloser inspection, even these isto generalizations appear to be merely two sides of the same coin: The activity dimension can be viewed as the reflection of arousal, and the valence dimension as the reflection of appraisal. Many of the studies that have discovered such relatively simple dimensional structures have been based on judgments about emotion words. We suspect, however. that the uncritical use of scaling techniques with emotion words is inappropriate, of at least premature, The problem is that judgments about (the similaity Introduction a between) emotion words depend on various, usually uncontrolled (and often uncontrollable) aspects of the stimuli~ aspects such as intensity of the corresponding emotions, types of antecedents, types of consequences, and so on. Without knowing to which of these (or other) aspects someone is attending, judgments of similarity are largely uninterpretable. It is partly for this reason, no doubt, that the plethora of multidimensional scaling and factor analytic studies that have been conducted seems only to agree that the major descriptive dimensions of emotions are valence and arousal, We find this conclusion isas uninformative as it is unsurprising, There have, of course, been numerous attempts to characterize the struc ture of emotions, They have been developed in different ways, often for different purposes. Theories have been proposed based on all kinds of variables; for example, biological/evolutionary variables (c.g... Plutchik, 1962; 1980), phenomenal variables (e.g., de Rivera, 1977), behavioral variables (e-g., James, 1890), facial expression variables (c.g., Ekman, 1982), and cognitive variables (¢.g., Roseman, 1984), Authors have con- sidered such variables to represent the primitive ingredients of human nature, and thus of human emotions. For example, some theorists argue that there is a fundamental opposition between fear and anger because of the underlying approachavoidance difference. Notice, however, that this difference is rooted in the typical response to these emotions rather than in their causes; when viewed from a causal perspective there may or may not be reason to believe that they are opposed in an important way. An almost universal characteristic of these approaches to emotion is the postu lation of a small number of basic emotions (typically fewer than ten). Our ‘own view is that the search for and postulation of basic emotions is not a profitable approach. One of our many reasons for saying this is that there seems to be no objective way to decide which theorist’s set of basic emotions might be the right one (for a more detailed discussion of this issue, see Chapter 2) Apart from scaling and arousalappraisal approaches to emotion, the other main approackes have studied the physiology of emotions and facial expressions. The visceral sensations accompanying emotions and the ex: pressive manifestations of emotions are perhaps the two characteristics that ‘most set emotions apart from other psychological states and events. This may explain why so much research has been concerned with them. The physiological research (see. e.g... Grings & Dawson, 1978: Levi. L975) is valuable and interesting and may he important for understanding the func: tions of emotions. However. it docs not adkdress questions about the cog tive origins of emotiens. Such questions are also Finessed by the research on c research om facial expressions, with which some of the most impress8 The Cognitive Structure of Emotions ‘emotion has been concerned (e.g.. Ekman, Friesen. & Ellsworth. 1982), Insofar as such research is indeed concerned with the expression of any particular emotion, it presupposes that the emotion already exists, leaving unaddressed the problem of how it came to be there in the first place. ‘Types of Evidence for Theories of Emotion ‘There are four kinds of evidence to which one might appeal in attempting to understand the emotions. Fitst, there is the language of emotions, which comes replete with ambiguity, synonymy (or near synonymy), and an abun- dance of lexical gaps and linguistic traps. Of course, emotions are not themselves linguistic things, but the most readily available nonphenomenat access we have (o them is through language. Thus, in order to specify the domain of a theory of emotion it is difficult to avoid using natural language words and expressions that refer to emotions. However, a theory of emo- tion must not be confused with a theory of the language of emotion. Consid- crable care needs to be taken in the use to which natural language is put in developing a theory of emotions. Not all distinct emotion types necessarily have associated words in any particular language. and not all the emotion words that refer to emotions in some particular language necessarily refer to distinct ones. The absence of a word in one language to designate the particular emotion that might be referred to by a word in another does not ‘mean that people in cultures using the first Janguage cannot and do not experience that emotion (Wierzbicka, 1986). Such linguistic gaps can be filled through catachresis and metaphorical descriptions, although the lat ter are often used even in cases where the language does provide a word for the particular category of emotion, but where one seeks to communicate the particular quality of an instance of the category (Fainsilber & Ortony, 1987). For some categories of emotions, a language like English provides a relatively large number of tokens, thus reducing the need for metaphorical descriptions of emotional quality. In such cases, it becomes necessary to identify one of the words in the category as the unmarked form or category label. For example, fear has lexical realizations that mark special cases such as very strong fear (“terrified”), very weak fear (“worried”), typical fear- induced behaviors (“cowering”) and so on. Thus, it may be helpful to think of the word “fear” as a relatively neutral word for an emotion type, fear. In other words, one can view the word “fear” as designating a distinct emotion lype (whereas the word “terrified” does not). This is quite consistent with a subsequent fine-grained analysis that might examine what exactly the differ cent tokens for the same emotion type do distinguish, and why. However. our ultimate goal is not to define emotion words such as “fear” but t0 Iniroduction 9 specify. in as language-neutral a manner as possible, the characteristics of distinct emotions. Language. therefore, is a source of evidence that has to be used with considerable care, We will assume that the words in our common language reflect « number of important distinctions, that they reflect a number of not so important distinctions, and that sometimes they fail to reflect important distinctions at all. Some of these issues will be discussed in Chapter 9. We have also discussed them at length elsewhere (e-g., Clore, Ortony, & Foss, 1987; Ortony & Clore, 1981; Ortony, Clore, & Foss, 1987). The second kind of evidence is evidence from self reports of experienced emotions. There is as yet no known objective measure that can conclusively establish that a person is experiencing some particular emotion, just as there is no known way of establishing that a person is experiencing some particular color. In practice, however, this does not normally constitute a problem because we are willing to treat people's reports oftheir emotions as valid. Because emotions are subjective experiences, like the sensation of color or pain, people have direct access to them, so that if a person is, experiencing fear, for example, that person cannot be mistaken about the fact that he or she is experiencing fear. This is not to deny that the person might be mistaken about some relevant aspect of the world that isthe cause of the fear (for example, about the threat that the feared event poses), or that the person may not be able to express the emotion in words (as in the case of a small child frightened by a dog). Yet, in normal cases, we treat self reports of emotions as valid. To be sure, we sometimes evaluate the reported emotions of others, but when we do, we evaluate them as being appropriate or inappropriate, or justifiable or unjustifiable, not as being tue or false, Furthermore. these evaluations are invariably based on our ‘own intuitions about the conditions under which different emotions can and do normally arise. Consequently, in the scientific study of emotions it is not unreasonable to appeal to our intuitions about what emotional states ‘are typically produced by situations of certain kinds, Clearly, it is possible to determine whether or not such intuitions are shared by others, even if verifying the empirical accuracy of such intuitions. widely shared as they may be, is more problematical. In some cases. therefore, one has to take note, albeit cautiously. of shared intuitions about emotions. OF course if a person does not share the consensual meanings of emotion terms. or is emotionally abnormal, or is simply being deceitful. then his or her self reports may well be invalid, but such exceptions presuppose a background of reliable reporting for their relevance This raises the question of the diffe theories of emotions. It might be argued that just as people’s naive theories snee between scientific and folk10 The Cognitive Structure of Emotions of certain physical phenomena bear no necessary relation to the scientific theories that account for those phenomena, so the folk theories of emotion that underlie our evaluations of the emotions of others bear no necessary relation to a scientific theory of emotion, However, the study of emotions is not like the study of physics. The phenomena that naive physics orga- nizes are phenomena external to the judging system, but the phenomena that a folk theory of emotions organizes are subjective experiences that are part of the judging system-their veridicality is not an issue. A person who is afraid ordinarily knows that he is afraid, and he ordinarily knows that his fear is caused by the prospect of some sort of threat. That is what fear isin ‘our culture. If this is a naive theory of fear, it has a quite different status from, for example, a naive theory of motion, an example of which would be the belief that a projectile emitted from a coiled tube will continue in a spiral trajectory that gradually straightens out (McCloskey, 1983), The status of the fear theory is different not merely because the fear theory is essentially correct while the motion theory is not, but because its correct- ness is guaranteed in much the same way as is the correctness of gram- maticality judgments of native speakers of a language. Linguists and psycholinguists assume that native speakers have a racit knowledge of the grammar of their language that is difficult or impossible to articulate. An important part of the linguist’s job is to discover the grammar by making explicit the implicit principles embodied in the normal Linguistic experience of native speakers. Similarly, we maintain that an important part of the psychologist’s job is to discover the “grammar” of emotions by making explicit the implicit principles embodied in normal emotional experience, So far, we have reviewed two kinds of evidence available to emotion theorists: language and self reports. We turn now to the third kind, namely, behavioral evidence. We shall play down this aspect of emotions for rea- sons that relate {o the nature of our goals. Perhaps initially stimulated by William James's claim that the emotions are the bodily responses, research ‘on emotions during the behaviorist and postbehaviorist era has been domi- nated by approaches based on the characteristic behavior associated with the emotions. Our view is that although itis important to identify character- istic behaviors associated with individual emotions, itis not often that these behaviors actually constitute an emotion, although. as has been observed by bother theorists (¢-g.. Friida, 1987; Lang, 1984). in many cases action renden- cies might be properly construed as part of a total emotional experience We prefer the view that actual behavior isa response to an emotional state in conjunction with the particular initiating event. All kinds of factors, many having litle it anything to do with the emotions, determine whether some particular behavior actually occurs, For example. people often reject Introduction. nl possible courses of action in response to an emotion if they believe that the intended goal will no: in fact be achieved by them. On the other hand, we consider it important to emphasize that emotions are important determi crs of motives. Since. ordinarily, there is a nonrandom connection between motives and behavior, our theory ought to be capable of accounting for certain classes of behaviors in terms of certain emotions (together with other determining factors). This. however. is not the same as predicting. specific behaviors in connection with specific emotions. In general. the problem with concentrating on behavior when considering the emotions is that the same behavior can result from very different emotions (or even from no emotion at all), and that very different behaviors can result from the same emotion Whereas we see serious problems associated with the general use of specific behaviors as evidence in the study of emotions, we are more sympa- thetic to the notion (e.g. Arnold, 1960); Frijda. 1987) that different emo- tions involve different associated action rendencies. However, we remain unconvinced that this is a characteristic of all emotions, and if itis not, then action tendencies weuld turn out merely to be concomitants (albeit high frequency ones) of emotions. Our reservations over the necessity to ineor- porate action tendencies as part of emotions is that we find the analysis quite strained in the case of many positive emotions (e.g., happiness. re- lief), and even in the case of some negative ones (¢.g., grief). We should make it clear that we do not intend to deny that, for example, smiling frequently accomparies happiness, or that weeping frequently accompa- nies grief. Our point does not pertain to involuntary refleslike expressions of emotions but rather to the more voluntary actions that follow on the: heels of emotions. We think the attention to action and action tendencies is a legacy from treating anger and fear and their associated reflexes of fight land flight as prototypes of emotions. In the general case, we think that action tendencies ar2 neither necessary nor sufficient for emotions. We think they are not necessary because emotions (such as admiration) that are rooted in standards, for example, rather than in goals are not coher ently characterized in terms of action tendencies in relation to changing. goal priorities. We think they are not sufficient because it is perfectly possible for there to 3¢ action tendencies without associated emotions. For example. realizing tat a house plant needs to be watered may lead to the action (tendency) to water it, but the antecedent perception is hardly an ‘emotion. Our conclusion, therefore, is that action tendencies may be typi cal, and for some emotions even normal, but they cannot be constitutive of all emotions The fourth kind of evidence is physiological. Our stew iy that whereasR The Cognitive Structure of Emotions the physiological concomitants of emotional experiences are of indisput able importance, they throw relatively little light on the cognitive compo- nents of emotion, which is the focus of our work. Thus we do not feel obliged to take sides in the debate about whether there is a unique pattern of physiological activity (of the sympathetic nervous system) associated with each specific emotion, because patterns of physiological activity are not directly relevant to the cognitive antecedents of emotions. There may ‘or may not be unique patterns; the issue remains an open and empirical question (see Lazarus, Kanner, & Folkman, 1980). To believe, however, that the importance of the cognitive determinants of emotions is in any way contingent on the final resolution of this issue (should there ever be one) is to misunderstand the nature of the cognitive claim. Some Goals for a Theory of the Cognitive Structure of Emotions We share Abelson’s (1983) view that an analysis of emotion must go be- yond differentiating positive from negative emotions to give a systematic account of the qualitative differences among individual emotions such as fear, envy, anger, pride, relief, and admiration. One way of assessing the various approaches to emotion is to determine how adequately they accom- plish this task. A number of interesting studies have focused on individual emotions or on small groups of emotions. For example, Averill (1982) studied anger, Mowrer (1960) looked at hope, fear, disappointment, and relief, and Epstein (1967) and Spielberger (1972) represent but two of the many who have studied anxiety. Such accounts, however, tend not to consider the emotions they investigate in the context of a larger, more comprehensive system, and they have not led to widely accepted conctu- sions about emotions in general. The more system-level arousal/appraisal theories (e.g., Arnold, 1960; Lazarus, Kanner, & Folkman, 1980; Man- ler, 1975; 1984) have a better chance of success but generally tend to be weak when faced with the problem of differentiating distinct emotions. Furthermore, they are often based on a narrow range of (frequently only negative) emotions. However, such problems are not, we think. endemic to system-level approaches, Indeed, our own theory is essentially an arousal! appraisal theory, yet we think it does not suffer from these limitations, (One of our main goals is to present an approach to the study of emotion that explains how people's perceptions of the world~their construals ‘cause them to experience emotions. We consider nyo questions to be central to this enterprise. The first is “What is the cognitive structure of the emo: tional system as a whole?” The second main question iy "What is the eogni: tive structure of individual emotions?” Our approach to the first of these Introduction i: {questions is to try to show the relationships that exist among the individu emotions in groups of related emotions. The general answer we propose is that the emotions are best represented asa set of substantially independent groups based on the nature of their cognitive origins. Our response to the second question is based on the assumption that the particular emotion a person experiences onsome occasion is determined by the way he construes the world or changes in it. Thus we attempt to specify both the eliciting conditions for the distinct emotions and the variables that influence their intensity. Insofar asa definition of emotion presupposes a theory. the ade- {quacy ofthe definition we propose will ultimately depend upon the adequacy of the theory. Our working characterization views emotions as valenced reactions to events, agents, or objects, with their particular nature bing deter- ‘mined by the way in which he eliciting situation is construed. We believe that ourtwo key questions concerning the overall structure of the emotion system ard the nature ofthe individual emotions are related in important and hitherto unexplored ways. In order to address them, itis important to distinguish between those affective states, and other mental conditions, that are genuinely emotional and those that are not (Clore, Ortony, & Foss, 1987; Ortony, Clore, & Foss, 1987). This issue is ne- lected by most existing theories, perhaps because itis not appropriately considered as part of a theory of emotion as such. Be that as it may, we consider a treatment of the issue to be essential for any theory of emotion ‘because it delimits the range of phenomena that need to be explained. ‘We have structured this book in a way that presents first the ideas that constitute the core of the theory that we are proposing. Thus we startin Chapter 2 with a presentation of the overall structure of the theory. We argue that there are three broad classes of emotions that result from focus- ing on one of thee salient aspects of the world events and their conse- quences, agents and their actions, of objects, pure and simple. Having laid out this’ general structure we turn in Chapter 3 to a discussion of the appraisal mechanism. This requires us to make some proposals concerning the macrostructure of people's knowledge representation. In particular, we discuss the nature of an organization scheme that might allow one to distin- Buish three major ingredients of appraisal. These are goals, standards, and attitudes, and they constitute respectively the criteria for evaluating events, the actions of agents, and objects. Because our analysis of individual emo- tions involves specifying the variables that influence their intensity. the Second part of Chapter 3 and all of Chapter 4 are devoted to a discussion of the key factors that we see as affecting the intensity of emotions. This is an important issue for us because « major part of our elaims about the overall Structure of emotions is that groups of emotions tend to have the sumeu The Cognitive Structure of Emotions variubles affecting their intensity. and that many of these variables are local in the sense that they do not affect the intensity of emotions in other groups. Having provided the background concepts. Chapters 5 and 6 are concerned with our detailed analyses of the Event-based emotions, that is, ‘of the emotions for which the underlying appraisal is based on goals. Chap- ter 7 is devoted to the Attribution emotions, namely those that attribute responsibility to the agents of actions in terms of standards, and Chapter 8 provides a brief survey of some of the issues surrounding the Attraction emotions, which are ultimately rooted in attitudes. Finally, Chapter 9 at- tempts to provide some criteria for the question of what is to count as an emotion and discusses a number of implications of the theory including, in particular, a discussion of how our proposals might, in principle, be formal- ized into a computationally tractable model. Summary Our goal is to construct a cognitive theory concerning the origins of the emotions. In particular, we want to specify the global structure interrelat- ing different emotions as well as the characteristics of individual emotions The global analysis breaks emotions into three general classes: reactions 10 events, agents, and objects. The analysis of individual emotions specifics the eliciting conditions for euch emotion and the variables that affect the intensity of each emotion, There are four main kinds of evidence about the emotions: language, self reports, behavior, and physiology. The latter two kinds of evidence con- cern the consequences or concomitants of emotional states, but not their origins, which we think are based upon the cognitive construal of events. For this reason we largely ignore behavioral and physiological evidence, focusing instead on language and self reports. Although we do take linguis- tic evidence into account, our analysis of emotions is not to be taken as a theory about emotion words. Rather, it attempts to specify different types of emotions. In many cases ordinary language has several words that refer to different aspects of the same underlying emotion type. For example. English has many words that refer to different levels of intensity or to the behavioral consequences of fear, but ultimately, all refer to the same under lying emotion type. Our account of emotions is in terms of classes of ‘emotion types, stich as these, and not in terms of specific words. It is an exercise in theoretical psychology, but one that we think can be tested ‘empirically 2 The Structure of the Theory In discussing the merits and dangers of using linguistic evidence, the last chapter introduced the notion of an emotion type. An emotion type is a distinct kind of emotion that can be realized in a variety of recognizably related forms. The example we gave was of the emotion type, fear, which can be manifested in varying degrees of intensity (marked in English by ‘words such as “concern,” “fright,” “petrified.” and so on), and for which there can be various subtle shifts in emphasis (for example, an emphasis on a particular kind of associated behavior is captured by the word “cow- ring,” and an emphasis on the object of fear as being psychological rather than physical is captured by the word “anxiety”). This notion of an ‘emotion type is ceniral to the theory. In order to provide a coherent account of the emotions, it is essential to reduce somehow the infinitude of phenomenally possible emotions to manageable proportions. We do this by focusing on the distinet emotion types rather than on the multitude of discriminable emotional states. The question that we then have to address is: What are the distinct emotion types and how are they related to one another? It seems to us that the distinct emotion types cannot be arranged informa- tively into any single space of reasonably low dimensionality. Rather, we suspect that the emotions come in groups for which the intragroup struc lure is quite simple even though the intergroup structure is not (Fillenbaum & Rapaport, 1971). We shall therefore proceed by trying to identify and characterize representative groups or clusters. Each group is structured so. that the definition of each cell in the group provides a specification of an emotion type that incorporates the eliciting conditions for the emotion in that cell. That is. it incorporates a situational description of the conditions, ‘under which the emotion can be triggered The groups of emotions that we identify have two important characteris tics. First, emotions in the same structurally related. For examp soup have eliciting conditions that are ‘one of the groups that we propose. the Is16 The Cognitive Structure of Emotions “Attribution Group.” contains four emotion types, each of which depends ‘on whether the attribution of responsibility to some agent for some action is positive or negative, and on whether the agent is the self or some other person (see Chapter 7). The eliciting conditions for each of the four emo- tion types are thus structurally related, involving only different bindings of the variables for identity of agent and valence of attribution. The second important characteristic of emotion groups is that each distinct emotion type represented in them is best thought of as representing a family of closely related emotions. The emotions in each family are related by virtue of the fact that they share the same basic eliciting conditions, but differ in terms of their intensity and sometimes in terms of the weights that are assigned to different components or manifestations of the emotions (e.g. behavioral components) ‘Another aspect of the theory is that the characterization of each emotion type includes a specification of the principal variables that affect its inten- sity. These variables are local to particular groups, and have to be distin- guished from what we call “global” variables, which can influence the intensity of emotions in all groups. An important guiding principle in devel- oping the theory was that it be sufficiently specific to permit empirical testing. Although we shall not discuss them in this book, there are two ‘general methods of testing that we use. One involves the use of experimen- tal studies relying primarily on evidence from self reports. The other in- volves the use of computer simulations. Experimental studies allow us 10 determine the validity of the groups that we propose. They allow us to determine whether, as they take on different values, the principal variables represented in the eliciting conditions really do give rise to phenomenally distinct emotions, and they allow us to determine whether or not the local intensity variables that we hypothesize as being related to individual emo- tions really do affect the intensity of those emotions. Many of the studies that we conduct to investigate these issues are studies in which we tap people's intuitions about the emotions they would expect others to experi cence under various conditions (e.g., Clore, Foss, Levine, & Ortony, in preparation). In such studies we systematically manipulate the underlying characteristics of the described conditions and the variables thought 10 influence intensity and then look for intersubject agreement about which emotions they think would be experienced, and with what intensity. It is relatively easy to collect data on a full range of emotions using such tech rniques because it is not necessary t0 create or wait for appropriate real- world situations to arise—for some emotions. such as grief. there are other factors that mitigate against using real-world emotions. However. it is ‘The Structure of the Theory 0 clearly desirable to show that the proposed theory also describes ade~ quately the emotions of individuals as they experience them. To this end some of our investigations have explored emotions under conditions that are ecologically more valid ‘Two particular lines of our empirical research have been concerned with the emotions experienced by individuals as opposed to the emotions they impute to others. One of these uses an automated diary technique. In one application of this method (see Turner. 1987), subjects responded on an ‘almost daily basis for two months to a detailed computer-administered questionnaire about emotions experienced in the preceding 24-hour pe- riod. This enabled us to examine a wide range of questions about variables affecting the elicitation and intensity of over 40 emotion tokens. A second. approach investigates the experienced emotions of fans witnessing sports, events. In these studies we ask subjects to identify the emotions that they are experiencing from time to time while watching basketball games. We also solicit from them information about the intensity of their reported emotions and about factors that we predict ought to be influencing what particular emotions they are experiencing, and with what intensity. As the data from such studies come in, it is quite possible that we shall find ourselves being forced to change certain aspects of the theory, and to the extent that this is $0, we do not view what we present in the following pages as the last word. It may well be that the theory will require serious repair, or even that it will have to be abandoned. Yet, we think that even in this worst case, it will have been worth presenting. The point of what we Propose is not that we think its the right answer, but rather that we think it is the right approach. There must be some cognitive principles underlying the experience of emotions, and we are simply proposing an approach to thinking about what they might be. A quite different technique for exploring the validity of the kind of theory we are proposing is to employ the tools of Artificial Intelligence in ‘an attempt to model the theory. The point of such an enterprise is not 10 create machines with emotions—we think that such an endeavor would be Pointless and futile but to create a computer model that can “understand” What emotions people would be likely to experience under what conditions. Such a system should be able to predict and explain human emotions, not have them. To the degree that the predictions and explanations of a com- Puter system match those of humans one can have confidence that the System embodies a reasonable mode! of the cognitive origins of emotions. Another virtue of exploring @ theory such as ours through the use ot a ‘vomputer model is that one can experiment with the model by manipulat-18 The Cognitive Structure of Emotions ing its parameters and examining the consequences of such manipulations to see whether of not they are reasonable. This degree of control is difficult, and sometimes impossible to achieve in experimental laboratory settings ‘The Organization of Emotion Types In presenting the overall structure of our account of emotion types, we start with the least contentious assumption that we can, an assumption not about ‘emotions as such, but about the ways in which people (at least in our culture) can perceive the world. The assumption is that there are three major aspects of the world, or changes in the world, upon which one can focus, namely, events, agents, or objects. When one focuses on events one does so because one is interested in their consequences, when one focuses ‘on agents, one does so because of their actions, and when one focuses on ‘objects, one is interested in certain aspects or imputed properties of them qua objects. Central to our position is the notion that emotions are valenced reactions, and that any particular valenced reaction is always & reaction to one of these perspectives on the world. We are perfectly willing to admit that different organisms, or people in different cultures, might ve up the world in different ways. If they do, then there would be corresponding changes in the emotion types that they could experience ‘Thus if, for example, certain animals are biologically and cognitively lim- ited to attending only to events and their consequences. then the only emotion types that they will be able to experience are those associated with reactions to events. Before discussing the overall structure of the theory. it will be helpful (0 clarify what we have in mind in distinguishing agents, objects, and events. ‘Our conception of events is very straightforward—events are simply peo ple’s construats about things that happen, considered independently of any beliefs they may have about actual or possible causes. Our notion of focus ing on objects is also quite simple, Objects are objects viewed qua objects ‘This leaves us with agents, which are things considered in light of their actual or presumed instrumentality or agency in causing or contributing 1© events. Agents are not limited to people, even though they are the most ‘usual manifestations. Agents can be nonhuman animate beings, inanimate ‘objects or abstractions, such as institutions, and even situations. provided they are construed as causally efficacious in the particular context, When objects are construed as agents, they are just thatobjects construed as agents. So, for example. a person who buys a new car that turns out te be constant source of trouble might blame the car for his series af misfortunes In doing so, however. he would be (reating the car as though it were a” The Structure of the Theory = 7 -, 5 =a SSeowen tononen Meco mess | eal Jy Le ——7 ree | conrtaeo | mow “i i) =) | Figure 2.1, Global traciore of emotion types. agent, rather than simply as an object. In treating it as an agent. he could disapprove of it, Were he to treat it only as an object, his affective reaction to it would be one of disike The overall structure that we propose is illustrated in Figure 2.1. where the three main branches correspond to the three ways of reacting to the world. It is important to realize that this structure is intended to he inter Preted asa logical de sition. not as a temporal one, Each branch, that is.2 The Cognitive Structure of Emotions cach of the three kinds of things to which one can have valenced reactions, is associated with a broad class of affective reactions. Whether or not these affective reactions are experienced as emotions depends upon how intense they are. which is one of the reasons why it is important to know what factors affect the intensity of what emotions. This issue is addressed in detail in the second part of Chapter 3 and in Chapter 4. The first broad lass comprises all the emotion types indicated in the left hand branch of the figure. This general class of affective reactions we have indicated as those of being pleased and displeased. We should say immediately that the choice of these words here (and of comparable words elsewhere in the structure, indicated in lower case letters) is not critical. They are intended only as convenient reminders for the corresponding locations in the struc: ture. Thus, they really serve the function of technical terms that are defined by the role they play in the proposed structure. For example, the particular words “pleased,” and “displeased,” simply represent the best we can do to find relatively intensity-neutral English words that refer (only) to the undif- ferentiated affective reactions one can have to events and their conse quences. These affective reactions arise when a person construes the conse quences of an event as being desirable or undesirable, so that judged desirability (including undesirabilty) is the most important, or the central, riable that affects the intensity of all these Event-based emotions. This ‘means that desirability (which, as will be discussed in Chapter 4, also has to be taken as a technical term) is the main criterion for evaluation. The second general class of affective reactions is shown in the middle branch of the figure as being those of approving and disapproving. When these reac- tions are sufficiently intense they lead to a group of emotions that we call the Attribution emotions. These are caused by reactions to the actions of agents, when they are viewed as being either praiseworthy or blameworthy. making judged praiseworthiness (which we take to include blameworthi ness) the primary basis for evaluation, Finally, the third general class com- prises the essentially unstructured affective reactions of liking and dislik- ing. The associated emotions here are the Attraction emotions, which are caused by reactions to objects, or aspects of objects. in terms of theit appealingness. As mentioned above. we are using a broad sense of “object” here, a sense in which an object can be animate or inanimate, concrete oF abstract To sce how. in general. focusing on events, agents, and objects leads t0 different classes of emotional reactions, we shall consider a highly oversim: plified example of the reactions a person might have upon learning that bis neighbor isa merciless child-heater. If such a person focuses only on the neighbor's role as the agenr of child-beating, judging it as blameworthy The Structure of the Theory 2 because of its violatior of certain standards. his valenced reaction towards the neighbor could be realized as an Attribution emotion such as reprouch ‘or contempt. The person could also focus on one or more aspects of a child beating event. If he focuses only on its undesirability it might cause him jstress. He could also focus on the plight of his neighbor's children and experience pity. Finally, the person might focus on his neighbor qua (unap- pealing) object, giving rise to an Attraction emotion such as hatred Clearly, this is a highly oversimplified picture. In reality, the person is likely to experience # mixture of emotions resulting from considering the situation from these different perspectives at different moments so that some of the resulting emotions may cooccur and some will occur in se- quences. However, we have presented this example only in order to expli cate the effects of focusing on different aspects of an emotion-inducing situation, It was not presented for the purpose of analyzing exactly what ‘emotions would occur in what mixtures or sequences. We can now return to discussing Figure 2.1. In interpreting the figure, notice that some of the labels are in upper case and some in lower case. Labels in upper case represent structural elements, whereas those in lower case represent emotional, or potentially emotional, states. Individual groups of emotion types are enclosed in boxes with the name we have ‘chosen for each group indicated in the panel at the bottom of each box. Representative names for the emotion types are shown in lower case, In all ‘cases where we have indicated emotion types, the particular words have been chosen only as suggestive labels for entries for that position in the Structure. ‘Thus, the scructural descriptions that they represent are not intended to be taken as definitions of them. We start by introducing the distinctions and associated emotions relating {0 reactions to events and their consequences. This branch includes emo- tion types that can be loosely thought of as pleased and displeased, joy and distress, hope and fear, relief and disappointment, pity and resentment. All Of these emotions, reachable from the CONSEQUENCES OF EVENTS branch. are reactions of the experiencing person to the implications of vents for the person experiencing the emotion. This branch of the emotion {ree in the figure is the most complex. The first potentially emotional states {0 be encountered are shown, before the branch divides. as being pleased and displeased. Thus being pleased and displeased are intended to refer to Telatively undifferentiated affective states that are nonspecific in the sense {hat they are nothing more than valenced reactions to evems (as opposed 60 {ems OF objects). A consequence of this i that all emotion nodes elow {he Pleased displeased mode ate ditferentiated instances nt being pleased or “spleased. Being pleased and displeased scem to be rather pale when22 The Cognitive Structure of Emotions considered as labels for emotions. One reason for this i that they lack the specificity that is characteristic of typical emotional states. A second reason is that we have purposely chosen rather neutral terms to indicate these states~that i, terms that do not strongly imply a particular level of emo- tional intensity ‘The structure that falls below the pleased/displeased node divides firs according to whether the person who experiences the emotions is reacting to the consequences of the focal event with respect only to himself, or aso with respect to some other person. This distinction is represented by the two main branches labeled CONSEQUENCES FOR SELF and CONSE. QUENCES FOR OTHER. Consider first the CONSEQUENCES FOR SELF branch, This branch leads to two groups of emotions. For one of these groups the consideration of the prospect of an event is erucial and for the other it is irrelevant. ‘The emotions for which the consideration of the prospects irrelevant are simply those that result from positive or negative reactions to the events that affect one. They reflect upon one’s well-being. and are in fact simply default cases of being pleased and being displeased Such emotions are usually referred to in English by words like “happiness.” “joy.” “unhappiness.” “sadness.” and “distress.” However, as will be dis cessed in Chapter 5, where these emotions are dealt with in detail under the rubric of the “Well-being” emotions, the situation is somewhat more complicated than this. Specifically. a number of emotions will be asigned to this category that are quite particular in content—emotions such as eref and regret. The reason we consider such emotions to be specific kinds of distress is that they appear to differ from distress only in that they involve ‘more specific events about which the experiencing person is displeased. So, for example, grief is not just a generalized form of being displeased at an undesirable event; the undesirable event is a more specific one comprising the loss of a loved one. The branch marked PROSPECTS RELEVANT includes frst the emotions of hope and fear. These emotions result from reacting to the prospect of positive and negative events respectively. Four additional emotions arise depending upon whether the prospect of a positive or negative event is believed to have been confirmed or disconfirmed. These four emotions are shown as “satisfaction.” “disappointment,” “relief,” and an emotion that for want of a better name we call “fears-confirmed.” This entire group of six emotion types we refer to as the “Prospect-hased” emotions, They ae discussed in detail in Chapter 6. The final soup of emotions that we propose under the Event-hased emo tions are those appearing under the CONSEQUENCES FOR OTHER branch. These are emotions that result from reacting to the consequenees The Structure of the Theory 2 of events when focusing on the consequences for others. This group of Emotions we refer to as the “Fortunes-oF-others” group. and it contains four distinct emotion types. The emotions represent the reactions to events, that a person can have when the events are desirable or undesirable relative to the goals and interests of another person. The branch marked DESIR- ABLE FOR OTHER leads to emotion types that we refer to as “happy-for ‘and “resentment,” while the branch labeled UNDESIRABLE FOR OTHER leads to “gloating” and “pity.” Thus. for example, under suitable condi- jons, when an event is undesirable for some other person but is for that reason desirable for the person experiencing the emotion, that person can experience the emotion of gloating, or Schadenfreude, whereas if the expe- rience of that undesirable event by the other person is (also) undesirable for the experiencer, the possibilty of the emotion of pity arises. The way in ‘which the figure should be interpreted with respect to, for example, gloat- ing is that gloating is the valenced reaction to an event characterized by being pleased that some event is undesirable for another person. This illustrates the way in which we view the emotions below the pleased! displeased node to be differentiated cases of being pleased or displeased. All the emotions in this Fortunes-of-others group are discussed in detail in the second part of Chapter 5 The second main set of emotions are those emanating from the mid- dle branch of the figure. This branch, Tabeled simply, ACTIONS OF AGENTS, represents emotions having to do with people's reactions to the ‘agency that they attribute to agents. Basically, these emotions are differen. tiated forms of the affective reuctions of approving and disapproving of an Agent's actions. The figure shows these “Attribution” emotions spliting into two, depending on whether the approval or disapproval focuses on the Self as agent (labeled SELF AGENT) or on some other as agent (OTHER AGENT). When the formal agent isthe se, the emotion types of pride and shame can arise. When some other person is the formal agent. the emotion ‘ypes of admiration and reproach can arise. A crucial aspect of these emo- tions has to do with the way in which the notion of the self is conceptual- ized. In order to account for the fact that Attribution emotions. such as Pride, can result from the actions of others, an extended notion of the sels fequired in which the self can be the formal agent while the actual agent is. ‘Some other person with whom one views oneself as being in a cognitive unit (Helder, 1958), This issue, and others concerned with the Atribution emo- is. discussed in etal Chapter? re cites branch of the figure shows a structureless group of emo: NS resulting from reactions to objects qua objects. ‘These emotions which we call the f all variations of the affective traction” emotions24 The Cognitive Structure of Emotions reactions of liking and disliking. Thus. they represent undifferentiated af fective and aesthetic reactions to objects, for which love and hate ate good examples. We do not wish to imply that objects construed in this way are necessarily construed independently of their agency. but rather that iis the object qua object, not the object qua agent, that is the focus of the evalua tion. In fact, because of the psychological difficulty of separating agents from their actions, there is a strong tendency for Attraction emotions such as hatred to cooccur with compatible Attribution emotions such as con: tempt. It is important to emphasize at this point that we view the Attraction ‘emotions as being momencary in nature, rather than enduring dispositions towards objects. Our intention is to focus on the momentary state of, for example, loving or hating somebody or something. The way in which dispositional evaluation is constructed out of individual experiences is pi marily a problem to be handled by a theory of impression formation and personal attraction, not by a theory of emotion, although we do discuss it briefly in Chapter 8. The figure also shows a branch leading out of the Well-being emotions and a branch leading out of the Attribution emotions both converging on 1 group of emotions labeled “Well-being/ Attribution compounds.” These compound emotions include those labeled as gratification, gratitude, re- ‘morse, and anger, and they have the property that they arise from simulta- neously focusing on both the action of an agent and the resulting event and its consequences. We call them compounds because they involve more than the mere cooceurrence of their corresponding constituent emo tions. Gratitude is not simply the cooecurrence of admiration for an agent and happiness at a resulting desirable outcome. It is a unified emotion in Which the constituents need not necessarily be independently exper enced. A detailed discussion of these emotions is presented in the second section of Chapter 7. ‘A major feature of the scheme that we have outlined is that some of the factors that affect the intensity of emotions are specific to particular groups of emotions. In general, as we shall discuss in more detail in Chapter 4 progress down the structure shown in the figure tends to result in the introduction of variables that affect the intensity of all the emotions lower down (see Figure 4.1). We mentioned three such variables in introducing the three main branches when we said that events and their consequences are evaluated in terms of their desirability, actions of agents in terms of their praiseworthiness. and aspects of objects in terms of their appealing ness. Desirability. praiseworthiness, and appealingness are all variables that influence the intensity of all the emotions below the point at which they are introduced, so that. although each is central to a distinct class of The Structure of the Theory 2s emotions (Event-based, Attribution, and Attraction emotions. respec tively) their effects are local to collections of emotions. At the same ts will also be discussed in Chapter 4. there are a number of variables such las unexpectedness that bave global rather than local effects on int ‘One final aspect of the overall structure that we are proposing is that there seems to be an interesting tendency for emotional reactions to de- velop in a left-to-right manner with respect to the structure shown in the figure. To the extent that this is true, it might be because the most salient initial experience is that some event transpires; this would constrain the ‘emotional reaction to an Event-based one. At the same time, however, one may seek to understand the causal origins of the event, so that an Attribu- tion emotion becomes a possibility. Finally, because inferences about the properties of (animate) objects are made on the basis of their actions, Attribution emotions may give rise to momentary reactions of liking or disliking of agents qua objects, A more detailed discussion of this conjec ture is presented in Chapter 8 Although we embrace the notion of emotion types, we are inclined to Feject the idea that there is a sct of “basic” emotions such that they. ‘ogether with their combinations, account for all the emotions (views aptly referred to by Scherer, 1984, as “palette theories”). At first sight, the thesis that there are basic emotions is not an implausible one. Surely there are too many emotions for all to be distinet and equally basic. As in the physical world of, say, chemical elements versus chemical compounds, it would not be unreasonable to suppose that the phenotypical emotions are based on a smaller number of gerotypical or primary emotions. Such a conception seems parsimonious. Having said that, however, one has to be able to answer a number of difficult questions. First, it is by no means obvious What the claim that there are basic or primary emotions is supposed to nga. Does it mean that such emotions ate universal? Does it mean that the) ,2tm emotion compounds or blends, or emotion mixes? Does it mean Tons) ae choUKd appear developmentally before other (nonbasic) emo: suck, A Second problem is that whatever answers one might propose for Hom tions. it does not folfow that one needs a concept of basic emo- perfects emotions. such as happiness, sadness, anger. and fear, could being 2 atel! Be (and probably are) found in all cultures, without their bar ae i any other sense (toe nals might be found in all cultutes tov. cmotnat WOuld not be sufficient to render them anatomically base). Such "ons could also combine with other emotions without that necessitit