Effects of Emotion Displays on Social Identification

David R. Heise
Department of Sociology
Indiana University
Bloomington, IN 47405

Published in Social Psychology Quarterly, 52 (1989): 10-21. See publication for final text.

Emotion displays influence character assessments, as when an offender avoids stigmatization by exhibiting shame over a deviant act. This article develops a model of the process within the framework of affect control theory. Analyses suggest that expression of appropriate affect can foster positive characterizations of the participants in an event and that expression of inappropriate affect instigates condemnation. Emotion displays by an actor influence character assessments most, but emotions displayed by the recipient of action also can alter outcomes.

In this article I extend affect control theory (Heise, 1977; 1979; 1986; Smith-Lovin and Heise, 1988)--integrating the theory's labeling model and emotion model--in order to deal with this issue. The next section provides a brief overview of affect control theory. Then I outline the logic of the new ideas. The theoretical work is mathematical and is presented as an Appendix. After examining some direct implications of the theoretical product, I present results for a variety of examples to show how the formulation translates into the language of everyday life.

The work in this article is wholly theoretical. However, the theoretical formulations are grounded in past empirical research (Heise, 1979; Smith-Lovin and Heise, 1988), and the model offers bearings for future empirical studies focusing on how emotion displays influence labeling processes ⁽¹⁾

.

Affect control theory proposes that definitions of situations invoke social identities and other cultural categories, and people generate experiences to confirm sentiments which are bound to the categories. Emotions signal how well the confirmation process is going. When the process is not going well, people engage in corrective action, or they adjust their definitions of the situation to fit what has happened.

Sentiments are measured in terms of three dimensions scaled from about -4 to +4: Evaluation (scaled from "bad" to "good"), Potency (from "powerless" to "powerful"), and Activity (scaled from "quiet, still" to "noisy, lively"). The EPA dimensions are universals of human judgment according to studies in more than 20 cultures (Osgood, May and Miron, 1975).

Transient impressions of people--produced by events--also can be assessed in terms of the EPA dimensions. For example, good people ordinarily act nicely toward children and thereby create positive impressions of themselves and the children. An actor who engages in deep, significant acts and who avoids being the recipient of others' actions generates an appearance of potency. An actor who performs nice, quiet acts like consoling someone seems slower and more mature than someone who chatters or who behaves violently. EPA impressions of an actor, recipient of action, and behavior can be predicted with considerable precision using empirically-derived impression-formation equations (Heise, 1979; Smith-Lovin, 1987b) which define outcome impressions from pre-event impressions of actor, behavior, and object.

Differences between fundamental sentiments and the transient impressions resulting from an event measure how well the event confirms identities. People try to construct events that will minimize the difference between sentiments and impressions by engaging partners whose identities offer resources for self-confirmation and by conduct that produces desired impressions. For example, maintaining a valued identity requires interacting with partners who also have valued identities, and it requires engaging in respectable acts with these people. Roles associated with social identities get enacted as sociocultural logic constrains the sequencing of events such that specific outcomes result as each person operates to experience subjective confirmation of sentiments (Spiro, 1961; Fararo and Skvoretz, 1984).

Emotion registers the impression of self created by an event and also how that impression compares to the sentiment for one's identity (Heise, 1987; Averett and Heise, 1987). Creating a good impression produces pleasant emotion and creating a bad impression produces unpleasant emotion. Creating an impression of powerfulness instigates emotions of mastery, creating an impression of powerlessness generates emotions of vulnerability. Being lively after an event translates to emotional activation, and being quieted by an event translates to languid emotions. However, one's identity also enters into the emotion process by setting a standard of comparison. For example, despondency can result when events create a good, potent, lively impression which, however, is not good enough, powerful enough, lively enough to confirm an exalted identity: the sweetheart who receives a mere card on his birthday is favored, but not enough and therefore feels unhappy.

One way to deal with disconfirming events is to manufacture further events that repair impressions: punish major transgressions, forgive minor ones, and remunerate overly zealous behavior by others in order to restore feelings about interactants to their proper places. In effect, such behavioral reactions are claims that a prevailing definition of the situation is correct and disconfirming events are aberrations requiring no cognitive shifts and no revision of social relationships.

Another way to deal with disconfirming events is to relinquish the current definition of the situation in favor of another definition that fits events better. An actor who behaves evilly can be cast into a stigmatized identity which accounts for evil behavior; an actor who performs heroically can be promoted to a status where excellence is the norm. Recipients of action can be reidentified, too, in order to acquire an interpretation of reality that fits with what is happening: those who are victimized can be viewed as repugnant characters who deserve the abuse they get; those who seem excessively favored can be viewed as charming in a way that elicits special benevolence from others.

Actions a person initiates provide a better guide to character than do actions which the person elicits from others because an actor's identity exerts more control over event construction than does the identity of the recipient of action. Thus, once we embark on reidentification in order to understand an event, we generally prefer to redefine the actor, attributing the event to his or her character. However, sometimes--especially if self is the actor--it may seem preferable to reidentify the object person in an event in order to interpret how an unaccountable behavior was elicited.

Affect control theory provides mathematical representations for all of the processes discussed above. The models for emotion and for reidentification--the central issues in this paper--are summarized below.

Emotion Model

Equations for predicting emotions interpret the difference between the transient impression of a person and the impression which is expected due to the person's identity (Heise, 1987; Averett and Heise, 1987). The emotion equations are derived mathematically from impression-formation equations (Heise, 1987, p. 25), and an interaction term in impression-formation equations1. ⁽²⁾

(Heise and Thomas, 1988) complicates the solution. However, the emotion equations can be interpreted as linear in form with changes in coefficients for every identity that is considered. To illustrate, here is the structure for a very valued identity (I_e=+2.0):

E_e = .51 + 1.21T_e + .68T_p - .56I_e - .41I_p + .07I_a ( 1)

E_p = .42 + .34T_e + 1.73T_p - .14I_e - 1.02I_p ( 2)

E_a = .20 - .17T_p + 1.89T_a + .13I_p - 1.20I_a ( 3)

and here is the structure for a very negative identity (I_e=-2.0):

E_e = 1.20 + 2.83T_e + 1.59T_p - 1.31I_e - .96I_p + .16I_a ( 4)

E_p = .61 + .79T_e + 1.98T_p - .35I_e - 1.17I_p ( 5)

E_a = .23 + .12T_e - .13T_p + 1.89T_a + 0.11I_p - 1.19I_a ( 6)

T stands for a measure relating to transient impressions, E is a measure relating to emotion, I is an identity measure, and the letters EPA in lowercase identify which scale of measurement is involved.

Suppose, for example, that a person has a self-identity corresponding to the EPA profile, 2 0 0--something like a Lady. If events support this identity perfectly (i.e., T_e T_p T_a also is 2 0 0) then equation (5) yields 1.8--felt emotion is quite good. If events create a less than ideal impression of 1 0 0, then equation (5) gives 0.6--the emotion still is slightly positive. Now suppose that a person has a self identity corresponding to the EPA profile, -2 0 0--something like a Snob. With perfect confirmation, equation (8) gives -1.8--a quite unpleasant feeling. With events creating a less than ideal impression of -3 0 0, equation (8) gives -4.7--despair. Note that, given the same degree of disconfirmation, the Lady's emotion declines 1.2 on evaluation, and the Snob's emotion drops 2.9 on evaluation; in the language of psychiatry, people with disvalued selves have more emotional lability than people with positive self concepts.

The potency component of emotion arises mainly from the impression of potency which is created by events, adjusted for the ideal potency defined by a person's identity, and the process is similar for people with valued and disvalued identities except that "looking good" enhances feelings of power more for people with disvalued selves than for people with valued identities. For example, a Snob who creates a self impression 1.0 higher than usual on evaluation goes up 0.8 units in emotion potency, whereas a Lady who achieves the same degree of success goes up 0.4 units in emotion potency.

Activation processes are mostly the same regardless of whether one's identity is valued or disvalued. That is, the equations imply that emotion activation arises mainly from the impression of liveliness created by events, referenced to the level of liveliness accorded to one's identity.

Reidentification Model

Equations for predicting reidentifications characterize the kind of actor who would perform a given act on a person with a given identity. The equations are too convoluted to present in other than matrix format, but they have the following general form.

A_e = function of (B_e B_p B_a , O_e O_p O_a, M) ( 7)

A_p = function of (B_e B_p B_a , O_e O_p O_a, M) ( 8)

A_a = function of (B_e B_p B_a , O_e O_p O_a, M) ( 9)

A stands for a measure of the actor's new identity, B is a measure relating to the given behavior, O is a measure characterizing the object person's identity, the letters EPA in lowercase identify which scale of measurement is involved, and M stands for a matrix of prediction coefficients. (See Heise, 1987, for detailed specifications.)

For example, suppose that someone engages in a bad, weak, lively act (EPA profile -2 -2 2) on a person who is good, powerful, and lively (2 2 2). Plugging these values into each equation and solving gives an actor EPA profile of -1.1 -1.3 1.7. Thus the equations predict that someone doing such an act toward such a person would be viewed by observers as slightly bad, sightly weak, and active.

Another example: if someone behaves in a nice, potent, quiet way (2 2 -2) toward a good, weak, lively object person, then the equations yield an EPA profile of 3.1 1.1 -1.9. The actor should be viewed by observers as extremely good, slightly potent, and quiet, according to predictions.

Similar prediction equations are available for predicting the characterization of a person who is the object of a given kind of behavior enacted by a given kind of actor.

The standard reidentification equations in Affect Control Theory include no separate terms representing the emotions of either actor or object. That is the purpose of this article: to elaborate the reidentification model so that it does include terms for emotions.

Example

A computer program (Heise, 1988), based on affect control theory's mathematical formulation and provisioned with a large number of empirical sentiment measures, allows one to analyze social situations as follows2. ⁽³⁾

.

A man entering a school to pick up his child may take the role of Father which has an EPA profile of 1.8, 2.1, -0.7: this means that the sentiment associated with Father is quite good, quite powerful, and slightly quiet. He may identify the child as Daughter which is 1.7, -0.5, 1.5: in words, a Daughter is felt to be quite good, a bit powerless, and lively. He thereupon anticipates events which will create impressions confirming these sentiments. Solving impression-formation equations for the behavior profiles that would accomplish this reveals that the father has to engage in acts which are quite good, potent, and neither too lively nor too quiet (EPA profile: 1.7, 1.8, 0.1)--acts like Praising or Assisting his daughter. The daughter has to perform acts that are good, lacking in potency, and lively (1.6, 0.1, 1.5)--like Hailing and Acclaiming her father.

Suppose that the father assists his daughter (EPA profile for assists: 1.8 1.6 0.0). This is almost perfectly confirming of his Father identity, creating an impression of him as 2.0, 1.7, -0.7: merely a bit weaker than he should be. His emotion--the subjective transformation of his identity into his current impression of self--is quite positive according to prediction equations: 2.3, 1.7, -0.3, corresponding to words like Warm and Affectionate. The father assisting his daughter also confirms the Daughter identity nearly perfectly, producing an impression of her as 1.5, -0.8, 1.3. The emotion profile that would describe the relation between this transient impression and her identity is 1.2, -0.3, 0.9 which translates to words like Lighthearted and Glad.

Now suppose that as soon as she sees him, the daughter begins cajoling her father about something (EPA profile: -0.6 0.1 0.8). This is somewhat disconfirming of his Father identity: it makes him seem less good and considerably less potent than he should be (1.0, 0.6, -0.7), and his implied emotional transformation has a profile of 0.3, -0.5, 0.0: he feels emotionally neutral rather than affectionate. Cajoling also disconfirms the Daughter identity to some degree, giving a transient impression of her as insufficiently nice: 0.3, -0.4, 1.6. If she's acting in role, then she must be feeling -0.1, 0.0, 1.4: Anxious or Restless.

The normal response of the father would be corrective action in the direction of reconfirming identities. That is, he has to construct an event that transforms the disconfirming impressions of himself and his daughter into impressions that are closer to the sentiments associated with their identities. According to impression-formation equations, his behavior has to be normally good but exceptionally potent and quiet (1.9, 2.4, -1.5), and this corresponds to acts like Consoling and Calming.

The father is not likely to redefine the situation merely because his daughter cajoles him, but suppose that he did. The reidentification equations, now solved for the kind of actor who would cajole a father, yield an identity profile of -0.7 -0.5 0.8. The man could implement this new sentiment toward the girl by labeling her as something like a Spendthrift, or he could think of her as an Irritable Daughter. Under some conditions he might try to understand the circumstances by redefining himself. The appropriate sentiment toward someone who is cajoled by a daughter is -0.5, -0.3, 1.5; and to implement this sentiment toward himself he conceivably could activate some deviant self-conception like Gambler or Flirt.

Affect control theory's established approach to reidentification assumes that new identities are sought for interactants in order to make a past event maximally confirming. The formulation does not use information about emotion displays because it implicitly assumes that the labeled person would feel the emotion which arises from confirming the new identity. Emotion is something that might be inferred from labeling, not an input. (In the example just given, the girl labeled as a Spendthrift is confirmed well by the act of cajoling a father, and the emotion of a Spendthrift in this event would be 0.1, 0.2, 0.9, corresponding to words like Anxious or Emotional).

This formulation has applications because people do sometimes judge others on the basis of conduct alone. Legal trials deliberately focus attention on bare facts; and narratives about actions rarely convey much information about participants' emotions because language--English, at least--provides a crude vocabulary for describing expressive displays (e.g., "smiling" covers numerous emotions, and so does "crying"). Affect control theory's established approach to reidentification supplies an appropriate model in these cases when character assessments are made at a distance from actual events.

However, when we are present at the events that stimulate reidentifications or when we observe a person narrating his or her own behavior, we have access to the expressive signaling system that our species has evolved (Ekman, 1984), and we obtain information about what emotions the person feels as a result of the events. No longer is it appropriate to assume that the person feels whatever emotion follows from a character reassessment because we can see what emotion he or she does, in fact, feel. The question is not merely what kind of person would engage in such conduct, but rather who would engage in such conduct and feel the way this person feels. Reassessment of the person's character has to take account of the emotion displayed because the expressive behavior may cue us that the other feels disconfirmed by events rather than confirmed.

Analytically, the problem of actor reidentification with added information on emotion amounts to this. We have a specified event (say, the actor maligned a friend). We also have a reading of the actor's emotion as a result of the event which can be expressed approximately by a term labeling subjective emotion (say, ashamed). The problem is to find an actor identity which combines with the specified event and produces an outcome impression of the actor that is offset from the unknown identity sentiment in just such a way that the observed emotion term corresponds to the difference. As usual, the impression-formation equations are solved for an actor EPA profile, but now the solution is obtained under the assumption that the event provides a specified kind of disconfirmation rather than confirmation.

The Appendix to this article shows how the problem can be formulated mathematically and shows that the solution (equation 40) constitutes a more general form of the prior reidentification model. That is, reidentification without information on emotions is a special case of the new formulation in which emotion displays are a factor in assessing someone's character from the person's conduct.

Reidentification of an object person proceeds the same way. For inputs we have a specified event (say, a colleague criticized someone) and the observed emotion of the recipient (say, flustered). The problem is to find an identity for the object person such that the given event produces an impression of the object person which is offset from the unknown identity in just such a way that the observed emotion would be stimulated. Equation (40) in the Appendix defines this solution also if quantities relating to the object are substituted for those relating to the actor (i.e., subscripts are changed to ).

The emotion of alter--the interactant who is not being reidentified--implicitly enters into these analyses because obtaining a solution requires constraining impressions of alter. Ordinarily we assume that the redefinition gives optimal confirmation to alter's identity, which amounts to saying that alter should feel an emotion corresponding to identity confirmation. On the other hand, we can specify alter's emotion explicitly as another input in the puzzle. In this case, we require that the identity being sought has to combine with the given event in such a way as to produce a particular impression of alter as well as a particular impression of the party being reidentified. Indeed, once a model is set up, we also can consider cases in which alter's emotion is specified but the emotion of the person being reidentified is assumed to be the emotion which corresponds to confirmation of the unknown identity. The mathematical adjustments for these cases are defined in the Appendix by equation (41) and equation (42).

Properties of the Model

The Appendix formulates these ideas within the framework of affect control theory and derives a mathematical model for analyzing how observed emotions influence reidentification processes. The model has complexities and subtleties which have to be explored through examples, but direct examination of the equations as is done in the Appendix does reveal some general properties of the solution.

One implication is that displays of appropriate affect foster enhancements of status and displays of inappropriate affect foster stigmatization. I use "appropriate" to refer to matching of emotion with the kinds of evaluative outcomes produced by an event. For example, consider an event that produces evaluatively positive impressions: hedonically positive emotions (contented, pleased, excited, etc.) are the appropriate emotions; hedonically negative emotions (mad, frightened, ashamed, unhappy, etc.) are inappropriate. Just the opposite applies for an event producing evaluatively negative impressions: hedonically negative emotions are appropriate, and positive emotions are inappropriate.

For example, an actor who engages in a gentle act toward a lovable object creates a positive impression and should feel an appropriate positive emotion. Politicians notoriously use just such an event--kissing and cuddling babies--to get themselves characterized favorably by the public. If the camera shows them beaming happily during the event they gain; were the camera to catch them grimacing nauseously they might end up discredited.

An actor who engages in a cruel act toward a child creates a negative impression, and any sort of positive emotion is inappropriate. Thus, if an actor expresses nonchalance or even pleasure about hurting a child, then he is viewed as evil--a Child Abuser. Yet the actor in such an event actually could gain positive credit by showing that the event overwhelms him with appropriate shame and distress, thereby convincing observers that the behavior was unintentional and uncharacteristic of his true identity.

The model also suggests that characterizations of a person may be more extreme when information about emotions is taken into account. For example, "the mother harmed her child" produces a negative characterization of the mother, but not as negative as "the mother harmed her child and was happy about it". The model further suggests that characterizations generally are more extreme when the observed emotion is hedonically negative rather than positive. Thus, theoretically, "the mother harmed her child and was mortified about it" actually might gain the mother more moral credit than "the mother saved her child and was happy about it".

Now I turn to a systematic set of examples in order to consider the implications of this solution in more detail and in order to see how emotion displays might influence potency and activity aspects of labeling as well as evaluation.

Throughout I translate numerical profiles to words in order to interpret predictions in everyday terms. This approach warrants a forewarning. Verbal results can seem inaccurate for a variety of reasons aside from the theory being wrong. (1) Cultural variations in ratings. Verbal predictions derive from EPA ratings of identities, behaviors, and modifiers, and if the ratings are different than those that you would provide, then results may describe a culture which is foreign to you. (2) Errors in ratings. The measurements used are averages computed over data from 25 or more males to reduce the possibility of chance errors, but 25 is a small sample by social research standards so errors due to measurement inaccuracies surely arise on occasion. (3) Errors in equations. Human psychology is being modeled through complicated equations which, though defined through research, are subject to errors in parameter estimation. (4) Lexical errors. Some errors arise because semantic rules governing word usage are not fully implemented. For example, merely matching words to profiles might lead to the prediction that a male interactant is a "sister" which is bizarre because "sister" should not be used as a noun describing males. Related to this is the problem that predictions are being made without fully defining a context. (5) Misconceptions. One function of theory is to offer new insights and to correct fallacies, so nonintuitive results could suggest an idea worth serious consideration.

Actor Emotions

First consider a situation in which a man and a woman are together, and the man's act of KISSING the woman instigates labeling of the actor by a male observer. A variety of cases are obtained by setting the man's emotion in turn to "cheerful", "calm", "angry", "disgusted", "nervous", and "ashamed"--a selection which represents the range of possible emotions (Morgan and Heise, 1988)--in order to see the effects of expressed emotion on reidentifications. The woman's emotion is unspecified, implying that reidentification of the man should confirm the woman's identity as much as possible.

Cheerful. Kissing a woman and feeling cheerful about it leads to positive identification of the actor. The EPA profile produced by the model in this case is 3.4, 0.3, -0.1, corresponding to identities like "gentleman", "pal", "mate". A man gains quite a bit of esteem in going from an evaluation of 1.1 (the evaluation of "man") to 3.4. The conception of the actor's power, however, declines from 1.1 (the potency rating for "man") to 0.3, and activity declines, too (the activity rating for "man" is 0.6).

Calm. The labeling profile in this case is 1.0, 1.5, -1.1 which corresponds to identities like "grown-up" and "advisor". Kissing a woman and feeling calm about it leaves a man's evaluation and power unchanged but requires a conceptualization of the actor as less lively than a "man".

Angry. In this case, the model produces a profile of -1.9, -0.1, -0.3, which corresponds to identities like "killjoy", "snob". The inappropriate affect involved in kissing a woman angrily causes a drastic reduction in evaluation of a man, a decline in conceptualization of power, and a decrease in activation.

Disgusted. Here the labeling profile is -2.4, 0.8, -0.5, and corresponding identities still include "killjoy" and "snob". An actor who is disgusted as he kisses a woman is stigmatized and also is conceptualized as less activated than a "man". Power does not change much.

Nervous. The labeling profile is -1.7, 0.0, -0.7, and identities with profiles like this include "grouch" and "killjoy". Kissing a woman and displaying nervousness reduces a man's evaluation and power and causes a decline in activity.

Ashamed. The labeling profile is -2.4, 0.7, -1.0, like a "miser" or "killjoy". Being ashamed while kissing a woman leads to a decline in evaluation of a man, and a reconceptualization of him as less active than a "man".

These examples indicate that reidentifications of an actor vary directly with the pleasantness of an actor's displayed emotion when the actor is engaging in positive behavior. Theoretically, the more unpleasant the actor's emotion, the greater the stigmatization. Emotional activation controls potency of characterizations: the livelier the emotional response, the more powerless the actor's new identity; the more torpid the displayed feeling, the more powerful is the label for such an actor. An interaction effect also seems operative: evaluation is especially high when the actor displays pleasant, potent, activated emotion.

Now consider the effects of emotion on labeling when a man engages in a disapproved act toward a woman: the man SCOLDS the woman. In an ordinary labeling analysis without consideration of emotions the man would be characterized as -1.3, 1.2, 0.3, a profile corresponding to identities like "bigshot", "critic". The results of taking the actor's displayed emotion into account are as follows.

Cheerful. The model gives a profile of -3.8, 0.5, 0.3 for someone who scolds a woman while displaying cheerfulness; the closest identity is "assassin". Thus the mildly deviant act accompanied by very inappropriate affect generates a reconceptualization of the actor such that a man is extremely stigmatized.

Calm. The computed profile in this case is -3.9, 2.4, 0.0; "ogre" and "mobster" are identities which are closest. Scolding a woman and being calm about it again requires reconceptualization of the actor as of lower value and activation than a "man", but power increases.

Angry. The labeling profile is 2.4, 0.7, 0.0, like a "Christian" or "gent". Scolding a woman while displaying anger seems normal because the affect is appropriate. Therefore reconceptualization of the actor involves no stigmatization. The labeling process would provide little loss of power for a man, but the actor seems less active than a "man".

Disgusted. Here the labeling profile is 0.2, 2.4, -0.2, corresponding to identities like "boss" or "supervisor". A man who displays disgust while scolding a woman loses some esteem and gains power; he seems still or old relative to the identity of "man".

Nervous. The computed profile is 2.9, 2.3, -0.4, corresponding to identities like "gentleman" or "physician". Scolding a woman and feeling nervous about it gains evaluation and power for a man. However, he also gets reconceptualized as less active than a "man".

Ashamed. The labeling profile is 1.8, 3.7, -0.8, which is close to "father" or "judge". A man who is ashamed in scolding a woman gains some status, much power, and seems inhibited.

The examples show that the emotion display of an actor in response to his own deviant action operates inversely on reidentifications as far as evaluation is concerned. Displaying pleasant emotions produces stigmatization; displaying unpleasant emotions can make deviance into a basis for moral credit. Emotions that signal activation and vulnerability during deviant behavior encourage judgments of one's character that are more positive in evaluation, while emotional languor encourages judgments that one is a powerful and quiet person.

Object Emotions

The model suggests that emotion displays by the recipient of action can affect characterization of the actor. As it turns out, the effect is tiny if we assume that an actor's emotion is unknown and therefore let the actor's emotion be whatever would best confirm the reidentified actor in the event. However, some interesting results occur for KISSING if the actor is observed displaying the natural emotion of a man kissing a woman which is approximately a feeling of happy friendliness. Then the displayed emotions of the woman who receives the kiss influence characterization of the man as follows.

Cheerful. If the woman is cheerful as she is kissed then the reidentification profile for the actor is 2.9, 2.4, 0.4: like a "truelove" or "friend".

Calm. When the woman is calm, the actor is reidentified as 1.4, 2.9, 0.3, corresponding to "hero", "bodyguard".

Angry. An actor whose kiss makes a woman angry might be characterized as -0.6, 2.4, 0.4: "supervisor", "boss".

Disgusted. If the woman is disgusted by a kiss, then the actor might be reidentified as -1.0, 3.0, 0.4: a "mafioso", "ringleader".

Nervous. The woman seeming nervous as she's kissed makes the actor look like -0.6, 3.0, 0.4: again like a "mafioso", "ringleader".

Ashamed. If the kiss makes the woman ashamed, then the actor seems to be -1.4, 3.7, 0.3: "slavedriver", "devil".

Now consider again the case of SCOLDING, first letting the actor feel appropriate affect--say, angry, as a man might feel when doing this. Then the following predictions occur as the woman's emotions vary.

Cheerful. If the woman is cheerful as she is scolded by an angry man then the reidentification profile for the actor is 0.2, 0.6 0.5 corresponding to "evangelist", "lookout".

Calm. When the woman is calm, the angry actor is reidentified 2.3, -0.3, 0.2; "neighbor", "worker".

Angry. An angry actor whose scolding makes a woman angry might be characterized as 3.4, -2.9, -0.1; no identities match.

Disgusted. If the woman is disgusted by a scolding, then the angry actor might be reidentified as 4.1, -2.3, -0.1; no identities match.

Nervous. The woman seeming nervous as she's scolded makes the angry actor look like 3.6, -1.8, 0.0; no identities match.

Ashamed. If the scolding makes the woman ashamed, then the angry actor seems to be 4.6, -1.3, 0.0; no identities match (the closest would be "darling").

On the other hand, if the man displays inappropriate affect for SCOLDING--if, say, he seems proud of it--then the woman's emotions have the following effects.

Cheerful. If the woman is cheerful as she is scolded then the reidentification profile for the proud actor is -3.0, 0.5, 0.6, a "thief", "cutthroat".

Calm. When the woman is calm, the proud actor is reidentified -2.6, -0.6, 0.6, a "traitor", "bigot".

Angry. An actor whose proud scolding makes a woman angry might be characterized as -2.4, -3.3, 0.3; "slave" is the closest match.

Disgusted. If the woman is disgusted by a scolding that makes the actor proud, then the actor might be reidentified as -2.3, -2.8, 0.4: something like a "slave", "addict".

Nervous. The woman seeming nervous because she's scolded makes the proud actor look like -2.4, -2.2, 0.4: "addict", "hypocrite".

Ashamed. If the scolding makes the woman ashamed, then the proud actor seems to be -2.2, -1.8, 0.5: "fag", "addict".

The examples indicate that emotion displayed by the recipient of action moderates labeling of the actor as follows. If the instigating event is positive then the actor is viewed as worse and more powerful as the object person's emotion gets more unpleasant; emotion signaling vulnerability in the object person may make the actor look more potent; and emotional activation in the object person enhances evaluation of the actor somewhat. On the other hand, if the instigating event is deviant, then stigmatization of the actor increases when the object person displays pleasant feelings (an actor is worse when the object person enjoys the deviant event); displays of unpleasant emotion by the victim make the deviant actor seem less malevolent but also weaker. Potency in the object person's emotion may make the actor seem somewhat less powerful; activity may make the actor more stigmatized.

Object Labeling

Recipients of action can be labeled as well as actors, and examples were computed to examine the kinds of solutions which result when emotions are taken into account. The complete set of analyses is not worth presenting in detail because the numerical results often strayed outside the range of EPA profiles for real words, so the benefit of seeing predictions in plain language was lost3. ⁽⁴⁾ Instead, I present a summary of patterns which appeared in the numerical results.

The recipient of action--when a target for reidentification--appears to be weak regardless of what emotions he or she displays, but unpleasant emotions in response to a positive act make the recipient seem bad and even weaker--so much so that computed profiles go beyond the range of actual words. On the other hand, displaying positive feeling when one has been the object of a deviant act makes the recipient look very bad (as Smith-Lovin, 1987a, implied), and so does displaying feelings of potency. Showing quietness in emotional response to a deviant act gets one reidentified in a relatively quiet way, whereas emotional activation implies a very active labeling.

The actor's emotional display also can influence how the recipient is labeled. If an actor feels pleasant about a positive action then the recipient seems good and lively; if the actor has unpleasant feelings about his positive act then the recipient seems bad and quiet. In the case of negative actions, displays of actor emotion generally serve to give the recipient identities which are either lower in evaluation or lower in potency. The recipient of a deviant act is less stigmatized when an actor feels bad about the act, but also the worse the actor feels the more the victim seems impotent and quiet.

Starting with the premise that people do, in fact, take account of emotion displays when judging others' character, this article develops a model of the process within the framework of affect control theory. According to this analysis, displays of hedonically appropriate affect promote positive characterizations, and displays of inappropriate affect foster stigmatization. Emotion displays by an actor have the most impact, but emotions displayed by the recipient of action also may have consequences for labeling.

Example analyses suggest that the activation dimension in emotion might influence character assessments in complex ways. An actor who displays lively emotions may be judged as powerless, yet lively emotion by an actor may moderate stigmatization based on deviant conduct. On the other hand, display of lively emotion by the recipient of action may increase stigmatization of an actor, and it may make the recipient seem a more lively character if the recipient is reidentified. The potency dimension in emotion seems still more complex--sometimes a display of potent emotion enhances character assessments, sometimes it leads to greater stigmatization--but effects of emotion potency are difficult to separate from positivity of emotion since the two dimensions are strongly correlated in the emotion domain (Morgan and Heise, 1988).

The theoretical analyses indicate that an actor might exert quite a bit of control over the way he is characterized were he to manipulate his emotional expressions, raising the question of why actors do not always employ "emotion work" (Hochschild, 1979) in order to avoid derogation. One reason is that expressions of natural emotion may slip out before actors realize that they are at risk of being labeled, and thereafter observers discount reversals in displayed affect even if actors succeed in manufacturing them. Alternatively, actors may insist on the appropriateness of their emotion as a form of rhetoric to support their ideologies (Averill, 1983; Hochschild, 1979): for example, a deviant's refusal to show shame serves as a claim that his behavior is not deviant. Still another possibility is that some actors may seek stigmatization to support a negative self-concept (Swann, 1987), even to the point of displaying inappropriate affect in order to earn just desserts.

The examples suggest that the potential for influencing labeling outcomes through emotion displays is not as great for the recipient of action as for the agent of action. Indeed, social consequences for a victim of deviant behavior might be especially dire if disclosure proceedings are not kept emotion-free with a focus on the morality of the actor's behavior: analyses suggest that knowing about emotional responses--of either the actor or the object--could lead to severe stigmatization of the victim.

At least some features of this theory have intuitive appeal, and examples exploring the theory produced verbal outcomes which often seemed to accord with common sense. Much of this face validity stems from the empirically-derived impression formation equations which incorporate a variety of subtle psychological processes into the model. Additionally, though, achieving credibility in results fortifies the theoretical premises which are at work in the model: that people seek experiences to confirm sentiments, and that judgments of others may depend on the emotions others display as well as on others' conduct.

Mathematical appendix not reproduced here. See the printed article.

Averett, C. P. and D. R. Heise 1987. Modified social identities: Amalgamations, Attributions, and Emotions. Journal of Mathematical Sociology 13: 103-132.

Averill, James R. 1983. Studies on anger and aggression: Implications for theories of emotion. American Psychologist 38: 1145-1160.

Ekman, P. 1984. Expression and the nature of emotion. Pp. 319-343 in Approaches to Emotion, edited by K. Scherer and P. Ekman. Hillsdale, NJ: Lawrence Erlbaum.

Fararo, Thomas J., and J. Skvoretz 1984. Institutions as production systems. Special issue of the Journal of Mathematical Sociology 10: 117-182.

Hawkins, R., and G. Tiedeman 1975. The Creation of Deviance: Interpersonal and Organizational Determinants. Columbus OH: Charles E. Merrill.

Heise, D. R. 1977. Social action as the control of affect. Behavioral Science 22:163-77.

Heise, D. R. 1979. Understanding Events: Affect and the Construction of Social Action. New York: Cambridge University Press.

Heise, D. R. 1985. Affect control theory: Respecification, estimation, and tests of the formal model. Journal of Mathematical Sociology 11: 191-222.

Heise, D. R. 1986. Modeling symbolic interaction. Pp. 291-309 in Approaches to Social Theory, edited by S. Lindenberg, J. S. Coleman, and S. Nowak. New York: Russell Sage Foundation.

Heise, D. R. 1987. Affect control theory: Concepts and model. Journal of Mathematical Sociology 13: 1-33.

Heise, D. R. 1988. Programs INTERACT and ATTITUDE. Raleigh, NC: National Collegiate Software Clearinghouse.

Heise, D. R. and E. M. Lewis 1988. Introduction to INTERACT. Raleigh, NC: National Collegiate Software Clearinghouse.

Heise, D. R. and N. MacKinnon 1987. Affective bases of likelihood judgments. Journal of Mathematical Sociology 13: 133-151.

Heise, D. R., and L. Thomas 1988. Predicting impressions created by emotion-identity combinations. Xeroxed, Department of Sociology, Indiana University.

Hochschild, A. R. 1979. Emotion work, feeling rules, and social structure. American Journal of Sociology 85: 551-575.

Lazowski, L. 1987. Speakers' nonverbal expressions of emotion as moderators of listeners' reactions to disclosures of self harm and social harm. Unpublished dissertation, Department of Psychology, University of California--Santa Barbara.

Morgan, R. and D. R. Heise 1988. Structure of Emotions. Social Psychology Quarterly 51: 19-31.

Osgood, C. H., W. H. May, and M. S. Miron 1975. Cross-Cultural Universals of Affective Meaning. Urbana: University of Illinois Press.

Smith-Lovin, L. 1979. Behavior Settings and Reactions to Social Scenarios: The Impact of Settings on the Dynamics of Interpersonal Events. Unpublished Dissertation, Department of Sociology, University of North Carolina, Chapel Hill.

Smith-Lovin, L. 1987a. A quantitative model of affect control. Paper presented at the meetings of the American Sociological Association, Chicago.

Smith-Lovin, L. 1987b. Impressions from events. Journal of Mathematical Sociology 13: 35-70.

Smith-Lovin, L. and D. R. Heise 1988. Analyzing Social Interaction: Advances in Affect Control Theory. New York: Gordon and Breach Science Publishers. (Reprint of a special issue of the Journal of Mathematical Sociology: 13.)

Spiro, M. E. 1961. Social systems, personality, and functional analysis. Chapter 2 in Studying Personality Cross-Culturally, edited by B. Kaplan. New York: Harper and Row.

Swann, William, Jr. 1987. Identity negotiation: Where two roads meet. Journal of Personality and Social Psychology 53: 1038-1051.

Note: I am grateful to Russell Fazio, Diane Felmlee, Linda Lazowski, David Schneider, and to anonymous reviewers for comments that helped to improve this work.

1. Hawkins & Tiedeman (1975) offer an organized discussion of the labeling approach in sociology.

2. The Heise and Thomas (1988) impression-formation equations predict how judges form impressions of emoting people. One of the predictive terms is the product of emotion evaluation with identity evaluation, and the significance is that displaying either a hedonically negative emotion or a stigmatized identity fosters impressions which are nearly as negative as a combination of negative emotion and stigmatized identity.

3. Throughout this article I use a revised version of program INTERACT which incorporates findings by Heise & Thomas (1988). Translations from words to profiles, and vice versa, use dictionaries of EPA profiles obtained from Southern U.S. undergraduates (first reported in Smith-Lovin, 1979, and reprinted in Heise & Lewis, 1988). Though the dictionaries give separate profiles for male raters and for female raters, in this article I present only the male viewpoint for the sake of clarity and brevity; similar results are obtained with female data.

4. That computed profiles strayed outside the range of actual words might mean that there are limitations on object-person reidentifications which incorporate information about the object person's emotion.