Copyright 1992 by the American Psychological Association, Inc. 0882-7974/92/S3.00
Psychology and Aging 1992, Vol. 7, No. 4, 551-561
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Type A Behavior Pattern and Emotion Expression in hunger and Older Adults Ruth Jonas
Carol Malatesta-Magai Long Island University
New York University Medical Center
Beth Shepard Yale University
L. Clayton Culver New School for Social Research
Eighty younger (less than 50 years, M = 28 years) and 80 older (more than 50 years, M = 69 years) Type A and Type B Ss were evaluated for Type A behavior pattern using the Structured Interview (SI) and given personality tests for anxiety, depression, anger, aggression, hostility, and anger-inanger-out. Ss also underwent an emotion induction procedure. Videotapes of the emotion induction procedure (N= 160) and the SI (N= 80) were coded for facial expression of emotion. Type As did not differ from Bs on anxiety or depression but did on anger and aggression. Type As showed anger inhibition and anger bound to shame, as predicted by emotion socialization theory. The greatest number of differential effects were observed between age groups. Older individuals, in general, were more emotionally expressive than younger Ss across a range of emotions. Women appeared more conflicted about anger expression than men, and Type A women more so than Type Amen.
Research on personality and its effects on illness and aging has come of age in the past two decades (Carstensen & Neale, 1989). One of the most fertile areas of research has emerged from the literature concerning Type A behavior pattern (TABP) and its relation to cardiovascular disease (CHD)—a disease that is the leading cause of death in older adults. Three recent reviews of the now extensive literature on the subject confirm that TABP is a significant risk factor in cardiovascular disease (Booth-Kewley & Friedman, 1987; Lang & Shedler, 1987; Matthews, 1988). Booth-Kewley and Friedman's quantitative meta-analysis revealed that TABP is related to CHD with an average effect size that is equivalent to that of other disease risk factors such as obesity and smoking. Type A individuals are twice as likely to develop heart disease and experience heart attacks as Type Bs (Booth-Kewley & Friedman, 1987; Rosenman, Brand, Jenkins, Friedman, & Straus, 1975). In effect, this increases the risk factor from one third of 7% to two thirds of 7%, a small but important elevation. In recent years, efforts have been directed at trying to determine whether the individual components of TABP (hard-driving competitiveness, time urgency, hostility) are all equally important in predicting disease or whether some components are more etiologically significant than others. There is now fairly conclusive evidence that anger and hostility figure importantly in the syndrome and in the prediction of disease (Chesney &
Rosenman, 1985; Dembroski, MacDougall, Williams, Haney, & Blumenthal, 1985). Williams (1984) reported that both TABP and scores on the Cook-Medley Hostility Scale (Cook & Medley, 1954) are independently related to coronary atherosclerosis; the latter measure also predicted clinical coronary disease in a 25-year follow-up of medical students (Barefoot, Dahlstrom, & Williams, 1983). Booth-Kewley and Friedman's (1987) quantitative review found that anger-hostility was a reliable predictor of CHD, with an average effect size of. 14, making it a significant risk factor. In the course of the growing volume of research on TABP a number of questions and anomalies have presented themselves, namely issues of emotion specificity, inhibition versus expression, age, gender, and socialization to TABR The present study was designed to address each of these issues. The first issue encompasses the debate over whether the emotional configuration of TABP is specific to anger and hostility or involves other emotions as well. Booth-Kewley and Friedman's (1987) meta-analysis of the association between TABP and emotion found that TABP was associated with anger-hostility as well as depression and anxiety, leading them to conclude that the Type A coronary-prone individual is a person with one or more negative emotions as a prominent feature of personality. However, as Malatesta (1988) subsequently pointed out, the issue of emotion specificity or generality—whether there is a disease-specific-emotion-specific association versus a general negative emotion neuroticism linked to disease—cannot be resolved by meta-analysis. Because meta-analysis rests on the assumptions of independence of effect sizes and probability levels, Booth-Kewley and Friedman could permit each sample to contribute only one effect size per personality variable, thereby eliminating the possibility that this kind of review could resolve
This study was supported by National Institutes of Health Grant 1RO1HL343444 to Carol Malatesta-Magai and a release time grant from Long Island University. Correspondence concerning this article should be addressed to Carol Malatesta-Magai, Department of Psychology, Long Island University, 1 University Place, Brooklyn, New York 11201. 551
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
552
MALATESTA-MAGAI, JONAS, SHEPARD, AND CULVER
the issue of specificity. What is needed are analyses that focus on patterns within the same investigation. In the present study, we assessed anger-hostility, depression, and anxiety in the same subjects; moreover, unlike previous studies that relied almost exclusively on self-report inventories, we used standard self-report personality measures as well as direct behavioral assessment. The second issue engages the question of whether the more unhealthy aspects of TABP relate to emotion expression or inhibition. Despite the great volume of literature, it is still unclear whether Type As are prone to experience certain emotions more frequently or intensely, suppress them more, or manage them in a different and more maladaptive fashion than Type Bs. The sometimes conflicting results accrue in part from conceptual confusion, which is reflected in the variety of measures of both TABP and anger-hostility that different investigators have used. In addition, measures of emotion have typically consisted of self-report instruments. The issue of emotion expression versus inhibition also relates to questions concerning early emotion socialization and its products. Tomkins (1963) suggested that children learn to inhibit certain emotions when their parents or other social agents shame them for displaying those emotions. Because shame readily attaches to any emotion with which it is repetitively associated, children acquire affect-shame binds. The consequence of the affect-shame bind is that, when the emotion that has been socialized through shaming is activated, shame is experienced and the affect is inhibited. Thus, at least according to theory, indicators of shame can signal the presence of an inhibited emotion. Because the TABP is thought to involve inhibited emotion (anger), at least according to some studies, it should also be marked by shame and shame bound to anger. Another unresolved question in the TABP literature centers on the variable of age and its relation to emotional expressivity. Williams (1984) and others have observed that the relation between TABP and heart disease appears to obtain only for younger samples of subjects. We propose that the relation may disappear in older subjects because Type A survivors are different in some way to begin with. Friedman, Hall, and Harris (1985) examined the expressive behavior of a sample of men from the Harvard Multiple Risk Factor Intervention Trial longitudinal study. They were able to distinguish between healthy and unhealthy Type As differentiated by the degree to which subjects were overt or inhibited in their expression of emotion. The unhealthy, illness-prone Type As (with more symptoms of cardiovascular disease) were repressed, tense, and less expressive, whereas healthier Type As were talkative, charismatic, and more emotionally expressive. These findings are provocative, especially in light of other research suggesting a more general link between emotional inhibition and illness (Pennebaker & Traue, in press). Collectively, these data suggest that Type As might differ in terms of illness proneness and in the expression of emotion, including the expression of anger. Indeed, it may be that the anger and hostility feature that is associated with TABP results in coronary risk only for those Type As who repress their anger. Friedman et al. did not look at anger expressivity per se, nor did they perform an age analysis. If expressivity confers a health benefit in Type As who are able to express their emotion, older Type As should be more emotionally expressive, even with
respect to the emotion of anger, which is diagnostic for Type A classification. We predicted that older Type As would be more overtly expressive, particularly with respect to anger, than younger Type As. We also reasoned that, if there were a general health benefit related to overt expressivity, older subjects as a group would appear to be more emotionally expressive over a range of positive and negative emotions. Finally, we were interested in ascertaining whether the emotional profile of the TABP syndrome expressed itself in the same way in men and women. Although cardiovascular disease incapacitates or kills more men than women (Eaker & Castelli, 1988) and begins afflicting men at an earlier age than it does women, it is also the chief cause of death in women. Despite this, almost all of the major longitudinal studies tracing the origins and course of the disease over time have relied exclusively on male subjects (the Framingham heart study being a notable exception). Without comparative data on women, we have no way of knowing whether the relation between hostility and development of heart disease in men can be generalized to women. A recent study of recurrent heart attacks (Thoresen & Low, 1990) indicated the need for caution. In that study, recurrence of heart attacks in men was associated with hostility, whereas in women, it was associated with anxiety. Given this finding, one would expect that the premorbid emotional profiles of Type A women might also show heightened anxiety, and that they would not be particularly distinguished with respect to anger, at least in relation to Type A men. Regarding the emotional profile of women in general, irrespective of type, research indicates that there may be a number of differences. For example, King and Emmons (1990) found that women experience more emotional conflict or ambivalence than men, which may be why women have been observed to engage in more expressive masking and dissimulation than men (Bugental, Love, & Gianetto, 1971). Popular cultural stereotype has it that women are less comfortable with anger than men, although Averill's (1982) large-scale survey study of men's and women's reports of their anger episodes found no significant gender differences. Many of the hypothesized gender differences in affectivity have rarely been systematically studied, nor have they been subjected to detailed behavioral analysis. In the present study, we had the opportunity to explore hypothesized differences in affect expression using both self-report measures and measures of actual behavior. In summary, the specific hypotheses of the present study are as follows. First, we predicted the emotional profile of Type As to be specific to anger, that is, Type As would show elevations on anger but not on other measures of negative emotion, such as depression or anxiety. Second, we predicted that Type As would show an elevation in their experience of anger but that they would exercise a greater degree of behavioral inhibition of this emotion, as measured by facial expressivity, than their Type B counterparts. Third, in accord with emotion socialization theory and Tomkins's (1962,1963) affect theory, we predicted that Type As would display more shame during anger challenge (ije., when anger is provoked) and more shame bound to inhibited anger. Fourth, in accord with the thesis that it is generally healthier to express than to inhibit emotional feelings (Pennebaker & Traue, in press), we predicted that our older subjects, a long-lived healthy sample, would show a pattern of more overt
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
EMOTION EXPRESSION IN YOUNGER AND OLDER ADULTS expressivity than our younger subjects; more specifically, we expected older Type As to show more overt signs of anger during anger arousal than younger Type As. Finally, we predicted that the emotional profile of Type A women would overlap with that of Type A men, but that Type A women would be higher on anxiety than Type A men, and that Type A men would be higher on anger and inhibited anger. Women, in general (irrespective of type), were predicted to be characterized by greater conflict around the expression of emotion, especially anger, as reflected in affective masking, defined behaviorally by Malatesta and Izard (1984) as the expression of joy or positive blend expressions (joy mixed with interest or surprise) under conditions of negative affect arousal. To explore the issues discussed above, we had a sample of older and younger men and women complete personality measures of several emotion traits, undergo a mood induction procedure, and be evaluated for TABP. We videotaped sessions and subjected them to detailed behavioral coding.
Method Subjects We recruited subjects for a study of "emotions across the lifespan" using posters and newspaper ads; word of mouth; and announcements in undergraduate and graduate classes, in classes for continuing education, and at senior centers. Potential subjects were screened and excluded if they had a history of heart disease or hypertension not currently controlled, because we were interested in profiling the emotional status of the TABP unconfounded by emotions generated by the person's concern about heart disease. A total of 240 subjects were recruited and run. We deleted 6 from the sample base because of physical or mental impairment that interfered with completion of the task, and 3 because they fell into the relatively rare Type X pattern. We chose 160 subjects by random means for inclusion in a balanced factorial analysis of variance (ANOVA) design (80 younger subjects, 80 older subjects, with equal numbers of men and women and Type As and Type Bs in each age group). We chose a balanced ANOVA design using this restricted number of subjects, against the alternative of using all 231 subjects in a multiple regression design, because emotions coding is labor intensive, often requiring two hours of coding per minute of videotape. Once the 160 subjects had been selected, their videotapes were coded, as described later. The mean age of the younger group was 28.09 years (SD = 6.97) and the older was 69.27 (SD = 8.18). The mean level of education for the 160 subjects was 1.6 years of college. ANOVA tests indicated no difference on amount of education by age, gender, or type.
Materials and Equipment Before coming to the laboratory, subjects filled out a set of personality measures. These consisted of measures of anger (described below), anxiety (trait anxiety scale of Spielberger's, 1986, State-Trait Personality Inventory), and depression (Beck Depression Inventory; Beck, Ward, Mendelson, Mock, & Erbaugh, 1961). To measure anger, we used Spielberger's (1990) State-Trait Anger Expression Inventory. This scale contains 24 items that, when taken together, provide a general index of how often anger is aroused (overall anger). An anger-in subscale assesses individual differences in the tendency to experience anger but hold it in, and an anger-out subscale assesses the tendency to express anger overtly. Additional measures of anger management were the seven questions used in the original Framingham study (Eaker & Castelli, 1988; i.e., subjects were asked how
553
likely, when angered, they were to "act like nothing happened," "keep it to the self," "apologize though right," "take it out on others," "blame it on others," "get it off the chest," or "talk to friends"). Two anger-relevant constructs—aggression and hostility—were measured by the trait aggression scale from the State-Trait Personality Inventory (Spielberger, 1986) and the Cook-Medley Hostility Scale (Cook & Medley, 1954). Items of the Cook-Medley scale were intermixed with items of the Crowne-Marlowe Social Desirability Scale (Crowne & Marlowe, 1960). Equipment included a Sony videorecording camera fitted with a zoom lens, a microphone, a recording and playback unit, and a video monitor. A visual time display was superimposed on the tape for coding timed behaviors.
Procedure Upon the subject's arrival at the laboratory, the goals and procedures of the study were reviewed and informed consent was obtained. Subjects first underwent an emotion induction procedure, which was followed by the Structured Interview (conducted by a different experimenter). Neither experimenter knew the hypotheses of the study. The experimental procedures were unobtrusively videotaped with the subject's knowledge and consent. Emotion induction procedure. This procedure, developed by Malatesta and Izard (1984) and described in detail by Malatesta, Izard, Culver, and Nicolich (1987), was used to generate emotional expressive behavior for subsequent affect coding. Subjects are asked to recount events in the recent past that generated strong emotion: In the present study, the target emotions were interest-excitement, sadness, fear, anger, and affection. Affection was included to restore the subject to a positive mood at the conclusion of the mood procedure and was not included in analysis. Four orders were created for the administration of the task, such that each of the four target emotions appeared once in the first and last positions, in order to assess the possibility of primacy or fatigue effects. The four orders were randomized across subjects. The emotion induction procedure is designed to maximize the subjects' ability to recollect and recount emotionally charged material so that their emotional expressivity has high authenticity and immediacy. After the subject had recounted each event, the experimenter interjected, "And once again, in one sentence, how did the experience really make you feel." This was done to generate a condensed, concentrated summary of the emotional experience for coding purposes. It also met our need for a sample of expressive behavior that would be somewhat comparable in length across individuals; the longer narratives were more variable, ranging from 20 s to almost 5 min in length. When the subject had finished recounting the first episode, he or she proceeded to the next four on prompt from the experimenter. After each episode, the subject was asked to rate the type and intensity of the emotion experienced during the inductions (the rating scale is described below). To facilitate return to a neutral state between episodes, subjects were instructed to count backwards from 20 to 1 before proceeding to the next emotion. As a manipulation check on induced emotion, subjects were asked to complete a modified version of Izard's (1972) Differential Emotions Scale (DES) immediately after each episode. In this procedure, the subject rated each of 10 emotion terms on a 9-point scale of intensity. Structured Interview. The Structured Interview (SI; Rosenman, 1978) and Jenkins Activity Scale (JAS; Jenkins, Rosenman, & Friedman, 1967) are the two most frequently used procedures for assessing TABP. Booth-Kewley and Friedman's (1987) meta-analysis of the Type A literature indicated that the SI is superior to the JAS as a predictor of CHD, and thus the Structured Interview was the typing device we used in the present study. The SI takes about 20 min to administer and is designed to elicit
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
554
MALATESTA-MAGAI, JONAS, SHEPARD, AND CULVER
reactivity in persons prone to impatience and irritation. In the present study the SI was administered by one experimenter and observed by another simultaneously on a TV monitor. Both were trained in SI administration and scoring by Jean Landau, who had originally trained with Ray Rosenman. Subjects were typed (A or B) immediately after the conclusion of the procedures. Interrater reliability was 92%, with discrepancies resolved by discussion and consensus. After the SI, subjects were asked to rate their degree of selfconsciousness during the emotion induction procedure on a 1-9 scale and also to rate what emotions they experienced during the SI using the DBS rating scale. They were then debriefed and thanked for their participation in the study. Coding procedures. Facial expressions of emotion were coded on a second-to-second basis using an adaptation of the Maximally Discriminative Facial Movement Coding System (Max), a theoretically based, anatomically linked objective component coding scheme developed by Izard (1979) and adapted by Malatesta and Izard (1984) for use with older faces. All material was coded by a team of eight coders who had been trained on the adult-modified coding system and who had achieved at least 80% reliability on each of the coding categories with both the training material and each other. Intercoder reliabilities (formulas following Izard's, 1979, procedure), sampled at random (15% of the material), indicated that coders reached and maintained sufficient reliability (78% to 99% for each category of expression) and that they did not drift from the original procedures. In the case of the emotion induction procedure, 20 s of the emotion event narrative and the entire length of the summary episode were coded for facial expressivity to yield sufficient material for parametric analysis. The summary episodes were less variable in length across subjects than the preceding narratives and thus provided a more standardized sampling of expressive behavior; we desired additional coding material for statistical purposes and chose the last 20 s of the narratives. We assumed that even though subjects had narratives that varied in length, they would all be roughly at the same psychological juncture at the end because they would all be bringing their narratives to a resolution. We used the end rather than the beginning or middle portion of the narrative because we expected this portion to be less contaminated by self-consciousness and to contain more overt affectivity. The entire sample of 160 subjects was coded. Because emotion coding often requires 2 hr or more of coding per 1 min of material (Izard, 1979), restrictions on amount of coding are frequently mandated. In the present study, because our primary focus was on testing the thesis of emotion specificity, for which we required the four emotion inductions, and because the Structured Interviews were on average twice as long as the inductions (20 min versus 2 min for each of 4 episodes), we limited the SI coding to one half of the subjects. We used a random numbers program to select 80 subjects, half Type As and half Type Bs, balanced for gender and age within type. The entire interviews of these subjects were coded.
Results General Design There were four factors in the study: Age (two levels, younger than 50 years, and older than 50 years), Gender, Type (A or B), and Order (four randomly assigned standard orders). ANGM\s indicated that order did not affect any of the dependent measures, which is consistent with previous research (Malatesta & Izard, 1984). All subsequent analyses used a three-way factorial design with age, gender, and type as the independent variables. Personality dependent variables were the scores on the several personality measures: overall anger, anger-in, anger-out, anger management items 1-7, hostility, trait aggression, trait anxiety,
and social desirability. Facial dependent variables (see Selection of MAX Codes, below) were total number of seconds spent in each category of emotion, proportional ized for segment length. The proportionalized data did not require transformation because residuals were normal, based on examination of normal probability plots. The alpha level for all a priori hypotheses was set at .05. Post hoc tests were Fisher's least significant difference tests. We used a more conservative cutoffof p < .01 for the reporting of effects that were not predicted a priori. Preliminary Analyses and Manipulation Checks Social desirability and self-consciousness. ANOVAs involving the Crowne-Marlowe social-desirability scores and the selfconsciousness ratings (administered after the videorecording sessions) disclosed two significant age effects. The older subjects scored higher on social desirability than did the younger subjects (M = 16.6 vs. 12.4), and younger subjects scored higher on the self-consciousness rating than did older subjects (M = 4.7 vs. 3.1). In light of these unanticipated findings, all analyses (behavioral as well as self-report variables) were subjected to both ANOVA and analysis of covariance (ANCOVA) with social desirability and self-consciousness as covariates. However, results indicated that at no point did ANCOVA results alter the pattern of findings disclosed by ANOVA, and thus only ANOVAs based on our a priori model are reported in this article. Effectiveness of the mood induction procedure. A manipulation check was made to establish that emotions appropriate to the specific induction conditions were elicited. For this we examined the DBS ratings of subjective emotional experience. Scale values could range from a low of 1 to a high of 9. The mean for fear under the fear induction condition was 6.8 (SD = 2.8); the mean for anger under the anger induction condition was 7.6 (SD = 2.3); the mean for sadness under the sadness induction was 7.2 (SD = 2.6); and the mean for interest under the interest induction was 6.7 (SD = 2.8). Statistical tests using difference scores between the subjects' emotion rating of the emotion congruent with the specific induction (e.g., fear under fear induction) and each of the ratings for the other emotions (anger, sadness, etc.) were conducted. One-sample t tests applied to difference scores computed for each condition and all emotions revealed that, for each of the four induction conditions, the mean ratings for the intended emotions were significantly higher in magnitude than any of the other emotions (ps ranged from
Psychology and Aging 1992, Vol. 7, No. 4, 551-561
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Type A Behavior Pattern and Emotion Expression in hunger and Older Adults Ruth Jonas
Carol Malatesta-Magai Long Island University
New York University Medical Center
Beth Shepard Yale University
L. Clayton Culver New School for Social Research
Eighty younger (less than 50 years, M = 28 years) and 80 older (more than 50 years, M = 69 years) Type A and Type B Ss were evaluated for Type A behavior pattern using the Structured Interview (SI) and given personality tests for anxiety, depression, anger, aggression, hostility, and anger-inanger-out. Ss also underwent an emotion induction procedure. Videotapes of the emotion induction procedure (N= 160) and the SI (N= 80) were coded for facial expression of emotion. Type As did not differ from Bs on anxiety or depression but did on anger and aggression. Type As showed anger inhibition and anger bound to shame, as predicted by emotion socialization theory. The greatest number of differential effects were observed between age groups. Older individuals, in general, were more emotionally expressive than younger Ss across a range of emotions. Women appeared more conflicted about anger expression than men, and Type A women more so than Type Amen.
Research on personality and its effects on illness and aging has come of age in the past two decades (Carstensen & Neale, 1989). One of the most fertile areas of research has emerged from the literature concerning Type A behavior pattern (TABP) and its relation to cardiovascular disease (CHD)—a disease that is the leading cause of death in older adults. Three recent reviews of the now extensive literature on the subject confirm that TABP is a significant risk factor in cardiovascular disease (Booth-Kewley & Friedman, 1987; Lang & Shedler, 1987; Matthews, 1988). Booth-Kewley and Friedman's quantitative meta-analysis revealed that TABP is related to CHD with an average effect size that is equivalent to that of other disease risk factors such as obesity and smoking. Type A individuals are twice as likely to develop heart disease and experience heart attacks as Type Bs (Booth-Kewley & Friedman, 1987; Rosenman, Brand, Jenkins, Friedman, & Straus, 1975). In effect, this increases the risk factor from one third of 7% to two thirds of 7%, a small but important elevation. In recent years, efforts have been directed at trying to determine whether the individual components of TABP (hard-driving competitiveness, time urgency, hostility) are all equally important in predicting disease or whether some components are more etiologically significant than others. There is now fairly conclusive evidence that anger and hostility figure importantly in the syndrome and in the prediction of disease (Chesney &
Rosenman, 1985; Dembroski, MacDougall, Williams, Haney, & Blumenthal, 1985). Williams (1984) reported that both TABP and scores on the Cook-Medley Hostility Scale (Cook & Medley, 1954) are independently related to coronary atherosclerosis; the latter measure also predicted clinical coronary disease in a 25-year follow-up of medical students (Barefoot, Dahlstrom, & Williams, 1983). Booth-Kewley and Friedman's (1987) quantitative review found that anger-hostility was a reliable predictor of CHD, with an average effect size of. 14, making it a significant risk factor. In the course of the growing volume of research on TABP a number of questions and anomalies have presented themselves, namely issues of emotion specificity, inhibition versus expression, age, gender, and socialization to TABR The present study was designed to address each of these issues. The first issue encompasses the debate over whether the emotional configuration of TABP is specific to anger and hostility or involves other emotions as well. Booth-Kewley and Friedman's (1987) meta-analysis of the association between TABP and emotion found that TABP was associated with anger-hostility as well as depression and anxiety, leading them to conclude that the Type A coronary-prone individual is a person with one or more negative emotions as a prominent feature of personality. However, as Malatesta (1988) subsequently pointed out, the issue of emotion specificity or generality—whether there is a disease-specific-emotion-specific association versus a general negative emotion neuroticism linked to disease—cannot be resolved by meta-analysis. Because meta-analysis rests on the assumptions of independence of effect sizes and probability levels, Booth-Kewley and Friedman could permit each sample to contribute only one effect size per personality variable, thereby eliminating the possibility that this kind of review could resolve
This study was supported by National Institutes of Health Grant 1RO1HL343444 to Carol Malatesta-Magai and a release time grant from Long Island University. Correspondence concerning this article should be addressed to Carol Malatesta-Magai, Department of Psychology, Long Island University, 1 University Place, Brooklyn, New York 11201. 551
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
552
MALATESTA-MAGAI, JONAS, SHEPARD, AND CULVER
the issue of specificity. What is needed are analyses that focus on patterns within the same investigation. In the present study, we assessed anger-hostility, depression, and anxiety in the same subjects; moreover, unlike previous studies that relied almost exclusively on self-report inventories, we used standard self-report personality measures as well as direct behavioral assessment. The second issue engages the question of whether the more unhealthy aspects of TABP relate to emotion expression or inhibition. Despite the great volume of literature, it is still unclear whether Type As are prone to experience certain emotions more frequently or intensely, suppress them more, or manage them in a different and more maladaptive fashion than Type Bs. The sometimes conflicting results accrue in part from conceptual confusion, which is reflected in the variety of measures of both TABP and anger-hostility that different investigators have used. In addition, measures of emotion have typically consisted of self-report instruments. The issue of emotion expression versus inhibition also relates to questions concerning early emotion socialization and its products. Tomkins (1963) suggested that children learn to inhibit certain emotions when their parents or other social agents shame them for displaying those emotions. Because shame readily attaches to any emotion with which it is repetitively associated, children acquire affect-shame binds. The consequence of the affect-shame bind is that, when the emotion that has been socialized through shaming is activated, shame is experienced and the affect is inhibited. Thus, at least according to theory, indicators of shame can signal the presence of an inhibited emotion. Because the TABP is thought to involve inhibited emotion (anger), at least according to some studies, it should also be marked by shame and shame bound to anger. Another unresolved question in the TABP literature centers on the variable of age and its relation to emotional expressivity. Williams (1984) and others have observed that the relation between TABP and heart disease appears to obtain only for younger samples of subjects. We propose that the relation may disappear in older subjects because Type A survivors are different in some way to begin with. Friedman, Hall, and Harris (1985) examined the expressive behavior of a sample of men from the Harvard Multiple Risk Factor Intervention Trial longitudinal study. They were able to distinguish between healthy and unhealthy Type As differentiated by the degree to which subjects were overt or inhibited in their expression of emotion. The unhealthy, illness-prone Type As (with more symptoms of cardiovascular disease) were repressed, tense, and less expressive, whereas healthier Type As were talkative, charismatic, and more emotionally expressive. These findings are provocative, especially in light of other research suggesting a more general link between emotional inhibition and illness (Pennebaker & Traue, in press). Collectively, these data suggest that Type As might differ in terms of illness proneness and in the expression of emotion, including the expression of anger. Indeed, it may be that the anger and hostility feature that is associated with TABP results in coronary risk only for those Type As who repress their anger. Friedman et al. did not look at anger expressivity per se, nor did they perform an age analysis. If expressivity confers a health benefit in Type As who are able to express their emotion, older Type As should be more emotionally expressive, even with
respect to the emotion of anger, which is diagnostic for Type A classification. We predicted that older Type As would be more overtly expressive, particularly with respect to anger, than younger Type As. We also reasoned that, if there were a general health benefit related to overt expressivity, older subjects as a group would appear to be more emotionally expressive over a range of positive and negative emotions. Finally, we were interested in ascertaining whether the emotional profile of the TABP syndrome expressed itself in the same way in men and women. Although cardiovascular disease incapacitates or kills more men than women (Eaker & Castelli, 1988) and begins afflicting men at an earlier age than it does women, it is also the chief cause of death in women. Despite this, almost all of the major longitudinal studies tracing the origins and course of the disease over time have relied exclusively on male subjects (the Framingham heart study being a notable exception). Without comparative data on women, we have no way of knowing whether the relation between hostility and development of heart disease in men can be generalized to women. A recent study of recurrent heart attacks (Thoresen & Low, 1990) indicated the need for caution. In that study, recurrence of heart attacks in men was associated with hostility, whereas in women, it was associated with anxiety. Given this finding, one would expect that the premorbid emotional profiles of Type A women might also show heightened anxiety, and that they would not be particularly distinguished with respect to anger, at least in relation to Type A men. Regarding the emotional profile of women in general, irrespective of type, research indicates that there may be a number of differences. For example, King and Emmons (1990) found that women experience more emotional conflict or ambivalence than men, which may be why women have been observed to engage in more expressive masking and dissimulation than men (Bugental, Love, & Gianetto, 1971). Popular cultural stereotype has it that women are less comfortable with anger than men, although Averill's (1982) large-scale survey study of men's and women's reports of their anger episodes found no significant gender differences. Many of the hypothesized gender differences in affectivity have rarely been systematically studied, nor have they been subjected to detailed behavioral analysis. In the present study, we had the opportunity to explore hypothesized differences in affect expression using both self-report measures and measures of actual behavior. In summary, the specific hypotheses of the present study are as follows. First, we predicted the emotional profile of Type As to be specific to anger, that is, Type As would show elevations on anger but not on other measures of negative emotion, such as depression or anxiety. Second, we predicted that Type As would show an elevation in their experience of anger but that they would exercise a greater degree of behavioral inhibition of this emotion, as measured by facial expressivity, than their Type B counterparts. Third, in accord with emotion socialization theory and Tomkins's (1962,1963) affect theory, we predicted that Type As would display more shame during anger challenge (ije., when anger is provoked) and more shame bound to inhibited anger. Fourth, in accord with the thesis that it is generally healthier to express than to inhibit emotional feelings (Pennebaker & Traue, in press), we predicted that our older subjects, a long-lived healthy sample, would show a pattern of more overt
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
EMOTION EXPRESSION IN YOUNGER AND OLDER ADULTS expressivity than our younger subjects; more specifically, we expected older Type As to show more overt signs of anger during anger arousal than younger Type As. Finally, we predicted that the emotional profile of Type A women would overlap with that of Type A men, but that Type A women would be higher on anxiety than Type A men, and that Type A men would be higher on anger and inhibited anger. Women, in general (irrespective of type), were predicted to be characterized by greater conflict around the expression of emotion, especially anger, as reflected in affective masking, defined behaviorally by Malatesta and Izard (1984) as the expression of joy or positive blend expressions (joy mixed with interest or surprise) under conditions of negative affect arousal. To explore the issues discussed above, we had a sample of older and younger men and women complete personality measures of several emotion traits, undergo a mood induction procedure, and be evaluated for TABP. We videotaped sessions and subjected them to detailed behavioral coding.
Method Subjects We recruited subjects for a study of "emotions across the lifespan" using posters and newspaper ads; word of mouth; and announcements in undergraduate and graduate classes, in classes for continuing education, and at senior centers. Potential subjects were screened and excluded if they had a history of heart disease or hypertension not currently controlled, because we were interested in profiling the emotional status of the TABP unconfounded by emotions generated by the person's concern about heart disease. A total of 240 subjects were recruited and run. We deleted 6 from the sample base because of physical or mental impairment that interfered with completion of the task, and 3 because they fell into the relatively rare Type X pattern. We chose 160 subjects by random means for inclusion in a balanced factorial analysis of variance (ANOVA) design (80 younger subjects, 80 older subjects, with equal numbers of men and women and Type As and Type Bs in each age group). We chose a balanced ANOVA design using this restricted number of subjects, against the alternative of using all 231 subjects in a multiple regression design, because emotions coding is labor intensive, often requiring two hours of coding per minute of videotape. Once the 160 subjects had been selected, their videotapes were coded, as described later. The mean age of the younger group was 28.09 years (SD = 6.97) and the older was 69.27 (SD = 8.18). The mean level of education for the 160 subjects was 1.6 years of college. ANOVA tests indicated no difference on amount of education by age, gender, or type.
Materials and Equipment Before coming to the laboratory, subjects filled out a set of personality measures. These consisted of measures of anger (described below), anxiety (trait anxiety scale of Spielberger's, 1986, State-Trait Personality Inventory), and depression (Beck Depression Inventory; Beck, Ward, Mendelson, Mock, & Erbaugh, 1961). To measure anger, we used Spielberger's (1990) State-Trait Anger Expression Inventory. This scale contains 24 items that, when taken together, provide a general index of how often anger is aroused (overall anger). An anger-in subscale assesses individual differences in the tendency to experience anger but hold it in, and an anger-out subscale assesses the tendency to express anger overtly. Additional measures of anger management were the seven questions used in the original Framingham study (Eaker & Castelli, 1988; i.e., subjects were asked how
553
likely, when angered, they were to "act like nothing happened," "keep it to the self," "apologize though right," "take it out on others," "blame it on others," "get it off the chest," or "talk to friends"). Two anger-relevant constructs—aggression and hostility—were measured by the trait aggression scale from the State-Trait Personality Inventory (Spielberger, 1986) and the Cook-Medley Hostility Scale (Cook & Medley, 1954). Items of the Cook-Medley scale were intermixed with items of the Crowne-Marlowe Social Desirability Scale (Crowne & Marlowe, 1960). Equipment included a Sony videorecording camera fitted with a zoom lens, a microphone, a recording and playback unit, and a video monitor. A visual time display was superimposed on the tape for coding timed behaviors.
Procedure Upon the subject's arrival at the laboratory, the goals and procedures of the study were reviewed and informed consent was obtained. Subjects first underwent an emotion induction procedure, which was followed by the Structured Interview (conducted by a different experimenter). Neither experimenter knew the hypotheses of the study. The experimental procedures were unobtrusively videotaped with the subject's knowledge and consent. Emotion induction procedure. This procedure, developed by Malatesta and Izard (1984) and described in detail by Malatesta, Izard, Culver, and Nicolich (1987), was used to generate emotional expressive behavior for subsequent affect coding. Subjects are asked to recount events in the recent past that generated strong emotion: In the present study, the target emotions were interest-excitement, sadness, fear, anger, and affection. Affection was included to restore the subject to a positive mood at the conclusion of the mood procedure and was not included in analysis. Four orders were created for the administration of the task, such that each of the four target emotions appeared once in the first and last positions, in order to assess the possibility of primacy or fatigue effects. The four orders were randomized across subjects. The emotion induction procedure is designed to maximize the subjects' ability to recollect and recount emotionally charged material so that their emotional expressivity has high authenticity and immediacy. After the subject had recounted each event, the experimenter interjected, "And once again, in one sentence, how did the experience really make you feel." This was done to generate a condensed, concentrated summary of the emotional experience for coding purposes. It also met our need for a sample of expressive behavior that would be somewhat comparable in length across individuals; the longer narratives were more variable, ranging from 20 s to almost 5 min in length. When the subject had finished recounting the first episode, he or she proceeded to the next four on prompt from the experimenter. After each episode, the subject was asked to rate the type and intensity of the emotion experienced during the inductions (the rating scale is described below). To facilitate return to a neutral state between episodes, subjects were instructed to count backwards from 20 to 1 before proceeding to the next emotion. As a manipulation check on induced emotion, subjects were asked to complete a modified version of Izard's (1972) Differential Emotions Scale (DES) immediately after each episode. In this procedure, the subject rated each of 10 emotion terms on a 9-point scale of intensity. Structured Interview. The Structured Interview (SI; Rosenman, 1978) and Jenkins Activity Scale (JAS; Jenkins, Rosenman, & Friedman, 1967) are the two most frequently used procedures for assessing TABP. Booth-Kewley and Friedman's (1987) meta-analysis of the Type A literature indicated that the SI is superior to the JAS as a predictor of CHD, and thus the Structured Interview was the typing device we used in the present study. The SI takes about 20 min to administer and is designed to elicit
This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
554
MALATESTA-MAGAI, JONAS, SHEPARD, AND CULVER
reactivity in persons prone to impatience and irritation. In the present study the SI was administered by one experimenter and observed by another simultaneously on a TV monitor. Both were trained in SI administration and scoring by Jean Landau, who had originally trained with Ray Rosenman. Subjects were typed (A or B) immediately after the conclusion of the procedures. Interrater reliability was 92%, with discrepancies resolved by discussion and consensus. After the SI, subjects were asked to rate their degree of selfconsciousness during the emotion induction procedure on a 1-9 scale and also to rate what emotions they experienced during the SI using the DBS rating scale. They were then debriefed and thanked for their participation in the study. Coding procedures. Facial expressions of emotion were coded on a second-to-second basis using an adaptation of the Maximally Discriminative Facial Movement Coding System (Max), a theoretically based, anatomically linked objective component coding scheme developed by Izard (1979) and adapted by Malatesta and Izard (1984) for use with older faces. All material was coded by a team of eight coders who had been trained on the adult-modified coding system and who had achieved at least 80% reliability on each of the coding categories with both the training material and each other. Intercoder reliabilities (formulas following Izard's, 1979, procedure), sampled at random (15% of the material), indicated that coders reached and maintained sufficient reliability (78% to 99% for each category of expression) and that they did not drift from the original procedures. In the case of the emotion induction procedure, 20 s of the emotion event narrative and the entire length of the summary episode were coded for facial expressivity to yield sufficient material for parametric analysis. The summary episodes were less variable in length across subjects than the preceding narratives and thus provided a more standardized sampling of expressive behavior; we desired additional coding material for statistical purposes and chose the last 20 s of the narratives. We assumed that even though subjects had narratives that varied in length, they would all be roughly at the same psychological juncture at the end because they would all be bringing their narratives to a resolution. We used the end rather than the beginning or middle portion of the narrative because we expected this portion to be less contaminated by self-consciousness and to contain more overt affectivity. The entire sample of 160 subjects was coded. Because emotion coding often requires 2 hr or more of coding per 1 min of material (Izard, 1979), restrictions on amount of coding are frequently mandated. In the present study, because our primary focus was on testing the thesis of emotion specificity, for which we required the four emotion inductions, and because the Structured Interviews were on average twice as long as the inductions (20 min versus 2 min for each of 4 episodes), we limited the SI coding to one half of the subjects. We used a random numbers program to select 80 subjects, half Type As and half Type Bs, balanced for gender and age within type. The entire interviews of these subjects were coded.
Results General Design There were four factors in the study: Age (two levels, younger than 50 years, and older than 50 years), Gender, Type (A or B), and Order (four randomly assigned standard orders). ANGM\s indicated that order did not affect any of the dependent measures, which is consistent with previous research (Malatesta & Izard, 1984). All subsequent analyses used a three-way factorial design with age, gender, and type as the independent variables. Personality dependent variables were the scores on the several personality measures: overall anger, anger-in, anger-out, anger management items 1-7, hostility, trait aggression, trait anxiety,
and social desirability. Facial dependent variables (see Selection of MAX Codes, below) were total number of seconds spent in each category of emotion, proportional ized for segment length. The proportionalized data did not require transformation because residuals were normal, based on examination of normal probability plots. The alpha level for all a priori hypotheses was set at .05. Post hoc tests were Fisher's least significant difference tests. We used a more conservative cutoffof p < .01 for the reporting of effects that were not predicted a priori. Preliminary Analyses and Manipulation Checks Social desirability and self-consciousness. ANOVAs involving the Crowne-Marlowe social-desirability scores and the selfconsciousness ratings (administered after the videorecording sessions) disclosed two significant age effects. The older subjects scored higher on social desirability than did the younger subjects (M = 16.6 vs. 12.4), and younger subjects scored higher on the self-consciousness rating than did older subjects (M = 4.7 vs. 3.1). In light of these unanticipated findings, all analyses (behavioral as well as self-report variables) were subjected to both ANOVA and analysis of covariance (ANCOVA) with social desirability and self-consciousness as covariates. However, results indicated that at no point did ANCOVA results alter the pattern of findings disclosed by ANOVA, and thus only ANOVAs based on our a priori model are reported in this article. Effectiveness of the mood induction procedure. A manipulation check was made to establish that emotions appropriate to the specific induction conditions were elicited. For this we examined the DBS ratings of subjective emotional experience. Scale values could range from a low of 1 to a high of 9. The mean for fear under the fear induction condition was 6.8 (SD = 2.8); the mean for anger under the anger induction condition was 7.6 (SD = 2.3); the mean for sadness under the sadness induction was 7.2 (SD = 2.6); and the mean for interest under the interest induction was 6.7 (SD = 2.8). Statistical tests using difference scores between the subjects' emotion rating of the emotion congruent with the specific induction (e.g., fear under fear induction) and each of the ratings for the other emotions (anger, sadness, etc.) were conducted. One-sample t tests applied to difference scores computed for each condition and all emotions revealed that, for each of the four induction conditions, the mean ratings for the intended emotions were significantly higher in magnitude than any of the other emotions (ps ranged from
