state-trait anger theory and the utility of the trait ... · such as the buss-durkee hostility...

Journal of Counseling Psychology1996, Vol. 43, No. 2, 131-148

Copyright 1996 by the American Psychological Association, Inc.0022-0167/96/S3.00

State-Trait Anger Theory and the Utility of the Trait Anger Scale

Jerry L. Deffenbacher, Eugene R. Oetting, Gregory A. Thwaites, Rebekah S. Lynch, Deborah A. Baker,Robert S. Stark, Stacy Thacker, and Lora Eiswerth-Cox

Colorado State University

Eight studies present support for state-trait anger theory. In Studies 1-3, high-anger partic-ipants reported (a) greater anger in many different provocations, in their most angeringongoing situations, and in daily life, (b) greater anger-related physiological arousal, (c)greater state anger and dysfunctional coping in response to a visualized provocation, and (d)greater use of suppression and outward negative expression of anger. Only heart rate in thevisualized provocation did not support predictions. In Studies 4-5, high-anger individualssuffered more frequent and intense anger consequences. In Studies 6-8 , trait anger hadhigher correlations with dimensions of anger than with other emotions, cognitions, andbehaviors. Few gender differences were found across studies. Results were discussed in termsof state-trait theory, convergent and discriminant validity for the Trait Anger Scale, angerexpression, gender, and the implications for counseling.

Although research on anger and anger reduction haslagged behind research on other emotional problems such asanxiety and depression, anger is beginning to receive greaterattention in applied psychology. For example, an increasingamount of published research suggests that anger is a factorin health problems such as cardiovascular disease (e.g.,Siegman & Smith, 1994), domestic violence (e.g., Leonard& Blane, 1992; Maiuro, Cahn, Vitaliano, Wagner, &Zegree, 1988; Pan, Neidig, & O'Leary, 1994), and sub-stance abuse (e.g., Leibsohn, Oetting, & Deffenbacher,1994) and that anger reflects a pattern of emotional distressassociated with elevated angry emotionality, negative per-sonal consequences, and other psychosocial problems (Def-fenbacher, 1992). There is also a growing body of treatmentresearch (e.g., Deffenbacher & Stark, 1992; Deffenbacher,Thwaites, Wallace, & Oetting, 1994; Moon & Eisler, 1983;Novaco, 1975) showing that general anger can be reduced.Anger research, however, has often suffered from theoreti-cal, conceptual, and measurement confusion (Spielberger,Jacobs, Russell, & Crane, 1983). For example, too often theoverlapping constructs of anger, hostility, and aggressionhave been blurred and used interchangeably, or anger as anemotional, experiential construct has not been separatedfrom the behaviors or modes through which anger is ex-pressed. Yet, if both theoretical and applied endeavors are toprosper, our theories need to articulate constructs clearly, toelucidate how these might be validated empirically, and to

Jerry L. Deffenbacher, Eugene R. Oetting, Gregory A. Thwaites,Rebekah S. Lynch, Deborah A. Baker, Robert S. Stark, StacyThacker, and Lora Eiswerth-Cox, Department of Psychology, Col-orado State University.

Preparation of this article was funded, in part, by the Tri-EthnicCenter for Prevention of Drug Abuse, National Institute on DrugAbuse Grant P50DA07074.

Correspondence concerning this article should be addressed toJerry L. Deffenbacher, Department of Psychology, Colorado StateUniversity, Fort Collins, Colorado 80523.

develop reliable and valid means of assessing and evaluat-ing these constructs.

In attempts to clarify and refine the understanding ofanger, Spielberger and his colleagues (e.g., Spielberger,1988; Spielberger et al., 1983; Spielberger, Krasner, &Solomon, 1988; Spielberger, Reheiser, & Sydeman, 1995)adapted state-trait personality theory to anger. State angerrefers to a transitory emotional-physiological conditionconsisting of subjective feelings and physiological activa-tion. Affectively, state anger is experienced along a contin-uum from little or no anger through mild to moderateemotions such as irritation, annoyance, and frustration tohighly emotionally charged states such as fury and rage.Physiologically, state anger varies from little or no changein physiological arousal to marked sympathetic arousal,increased tension in facial and skeletal muscles, and releaseof adrenal hormones. State anger is thus an emotional-physiological condition that occurs in response to an imme-diate situation, varies in intensity, and fluctuates over shortperiods. Trait anger, on the other hand, refers to a stablepersonality dimension of anger proneness or the tendency toexperience state anger. That is, high trait anger individualsexperience more frequent and more intense state anger(Spielberger, 1988; Spielberger et al., 1983, 1988). Al-though recent writings on state-trait anger theory do notaddress response duration, earlier state-trait anxiety theory(e.g., Spielberger, 1966, 1972) suggests high trait angerindividuals experience more lengthy state anger arousal aswell. Trait anger, therefore, is thought to be a relativelystable individual difference in frequency, intensity, andduration of state anger.

In postulating trait anger as a broad personality disposi-tion toward anger, state-trait anger theory leads to fivegeneral theoretical predictions, (a) Trait anger reflects atendency to become more easily angered (the elicitationhypothesis; i.e., high-anger individuals should be more eas-ily angered, which should be reflected in greater numbers ofthings that anger them and in greater frequencies of dailyanger), (b) Trait anger reflects a tendency to respond with

131

132 DEFFENBACHER ET AL.

more intense anger when provoked (the intensity hypothe-sis; i.e., high-anger individuals should experience strongeranger reactions), (c) Because of greater intensities and fre-quencies of anger reactivity, high trait anger individuals arepredicted to cope less well with anger and to express them-selves in less positive, less constructive ways. That is, traitanger reflects a tendency to express anger in less adaptiveand less functional ways (the negative expression hypothe-sis), which should be reflected in more frequent angersuppression and outward, negative expression of anger andless frequent application of constructive coping, (d) Be-cause of greater frequencies and intensities of anger andbecause of less positive coping, high trait anger individualsare more likely to experience negative anger-related conse-quences. That is, trait anger reflects a tendency to experi-ence more frequent or severe anger-related consequences(the consequence hypothesis), (e) If trait anger reflects aunique personality disposition toward anger and not otheremotional traits, then trait anger should relate to anger-related constructs more powerfully than to constructs thatdo not involve anger (the discrimination hypothesis).

This article presents eight studies that address these fivehypotheses. The purpose of this article, therefore, is to testcentral predictions of state-trait anger theory and to presentevidence of convergent and discriminant validity for theTrait Anger Scale (TAS). The first five studies were alsodesigned specifically to attend to the clinical relevance andutility of the TAS, that is, the value of the TAS in discrim-inating clinically meaningful anger groups by creating ahigh-anger counseling analog. This was done by employinga high-anger group that was not only normatively high ontrait anger but also reported a personal problem with angerand an interest in seeking help for anger reduction. If theTAS has clinical utility, then the TAS should document thatsuch a high-anger, help-seeking group, compared to low-anger individuals who report few problems with anger,should experience significantly more (a) anger, (b) mal-adaptive ways of expressing anger, and (c) negative conse-quences stemming from their anger expression. Moreover,in order to have high clinical utility, the TAS should also beshown to differentiate anger from other emotional and be-havioral issues so that it is demonstrably a measure of anger,not a measure of general emotional problems and distress.

Study 1 employs a variety of methods to evaluate theelicitation, intensity, and negative expression hypotheses,that is, it tests whether high-anger individuals will experi-ence more frequent and more intense anger reactions thanlow-anger individuals and whether they will express them-selves in more dysfunctional ways. Studies 2 and 3 replicateand extend the findings of Study 1 by assessing effects ofdifferent provocations. Studies 4 and 5 extend theory testingto the area of anger consequences and test the propositionthat high-anger individuals also experience more frequentand severe consequences from their anger (consequencehypothesis). Studies 6-8 continue to evaluate basic theo-retical predictions, but also test the discrimination hypoth-esis that trait anger predicts anger-related phenomena betterthan it predicts other nonanger phenomena.

Studies in this article were conducted over an 8-yearperiod, and the designs of each study emerged from ques-tions, confounds, and issues stemming from prior studies.That is, as potential confounds and alternative theoreticalpremises were clarified, studies were developed to addressthose issues. All samples are independent of one another,except for the possibility that some of the students from theabnormal psychology class in Study 7 may have partici-pated in one of the earlier studies that used introductorypsychology students. The numbers of such students, how-ever, is anticipated to be small and should introduce nosignificant bias. Participant pools from which samples weredrawn were predominantly Caucasian (i.e., <5% non-Cau-casian), and conclusions should be limited to this popula-tion. Additionally, given the large number of comparisonswithin some of these studies, in order to reduce experiment-wise error, alpha level was set at .05 for multivariate anal-yses of variance (MANOVAs), but post hoc between-groupcomparisons for the MANOVAs and various other between-group comparisons were adjusted with the Bonferroni cor-rection. In this article, we present and briefly discuss theresults for each study individually, with the major integra-tion of findings in the general discussion at the end of thearticle. Moreover, because there is considerable overlap inthe instrumentation and methodology of some studies, theMethod sections of several studies have been condensedwith reference to earlier sections in order to reduce redun-dancy and facilitate readability.

Study 1

The elicitation, intensity, and negative expression hypoth-eses were tested by comparing a group of high trait angercollege students who also self-identified as having personalanger problems and as being interested in seeking counsel-ing for anger reduction with low-anger students who did notreport personal anger problems. Specifically, it was pre-dicted that high-anger individuals would show (a) a height-ened vulnerability to elicitation of anger, which would bereflected in greater anger across a wide range of potentiallyprovocative situations; (b) greater anger in ongoing situa-tions that are uniquely angering to the individual; (c) greaterfrequency and intensity of anger in daily living; (d) moresevere anger-related physiological symptoms; greater (e)physiological arousal, (f) state anger, and (g) dysfunctionalcoping when provoked; and (h) greater general tendenciesto suppress anger and to express anger in an outward,negative manner. Resting blood pressure and pulse levels,and trait anxiety were also assessed.

Method

Participants

Participants were 89 introductory psychology students (Mdnage = 19) who were from the upper quartile of the TAS (TAS >22) and who identified themselves as having personal problemswith anger and as desiring counseling for those problems, or whowere from the lower quartile of the TAS (TAS < 16) and who

SPECIAL SECTION: TRAIT ANGER THEORY 133

indicated no problem with anger. By condition, 23 low-angerwomen, 22 low-anger men, 22 high-anger women, and 22 high-anger men participated. They received one of three required re-search credits for participation.

Instruments

Trait Anger Scale. The TAS (Spielberger, 1988) is a 10-item,Likert-type (1 = almost never to 4 = almost always) scale onwhich participants reported how angry they generally felt. TASinternal consistency reliabilities range from .81 to .91 with highestreliabilities for college students (Spielberger, 1988). The TAScorrelates positively with a variety of anger and hostility measuressuch as the Buss-Durkee Hostility Inventory and with various stateanger measures (Spielberger, 1988), and discriminates high- fromlow-anger groups (Lopez & Thurman, 1986; Spielberger, 1988).

Instruments were selected to evaluate each of the eight predic-tions derived from state-trait theory, (a) Anger across a wide rangeof provocative situations was assessed by the 90-item AngerInventory (AI; Novaco, 1975), on which participants indicate on a5-point scale (1 = not at all to 5 = very much) how much angerwould be experienced if the situation were encountered. The AIhas internal consistency reliabilities ranging from .94 to .96, cor-relates moderately with other anger scales, and discriminates morefrom less angry groups (Deffenbacher, Demm, & Brandon, 1986;Novaco, 1975). (b) Anger in ongoing, personally angering situa-tions was measured by the Anger Situation measure (Deffen-bacher, Demm, & Brandon, 1986), which has participants describein detail their most angering, ongoing situation and rate the inten-sity of anger in response to that situation on a 0-100 (0 = no angerto 100 = maximal anger ever experienced) scale. The AngerSituation has a 10-week test-retest reliability of .81 (Deffenbacher,Story, Brandon, Hogg, & Hazaleus, 1988) and modest correlationswith other anger indexes (Deffenbacher, Demm, & Brandon, 1986;Deffenbacher & Sabadell, 1992). (c) Intensity and frequency ofdaily anger was assessed by the Anger Log (Deffenbacher, Demm,& Brandon, 1986), on which participants, every day for a week,described in detail and rated the intensity of the most provocativeexperience of the day (same 0-100 scale as Anger Situation).Three measures were derived from the Anger Log: mean dailyintensities, the number of incidents during the week rated > 30(i.e., the frequency of days with at least mildly angering events),and mean daily intensities for days rated > 30 (score of 30 wasbased on an earlier study of Hazaleus & Deffenbacher, 1986,which suggested this score indicated at least mild anger for mostparticipants). The Anger Log correlates moderately with otherperson-specific anger measures (Deffenbacher, 1992). (d) Anger-related physiological arousal was evaluated by the Anger Symp-tom Index (Deffenbacher, Demm, & Brandon, 1986), on whichparticipants describe their most common anger-related physiolog-ical reaction (e.g., shakiness or sweating) and rate symptom inten-sity on a 0-100 scale (0 = no problem at all to 100 = symptomextremely severe). The Anger Symptom Index has 10-week test-retest reliability of .85 (Deffenbacher et al., 1988) and significantcorrelations with other state and trait anger measures (Deffen-bacher, Demm, & Brandon, 1986). (e) Physiological arousal fol-lowing provocation was assessed by a self-monitored, 15-s pulseon the nondominant wrist or carotid artery immediately followingthe experimental provocation. Reliability for self-monitored pulsehas been established through high correlations (rs = .90-.93) withmechanically recorded pulse rates (Hazaleus & Deffenbacher,1986; Southard & Katahn, 1967). Validity for heart rate as ameasure of arousal is found in significant correlations with stateanger (r = .31; Deffenbacher, Demm, & Brandon, 1986) and in

anger producing greatest change in cardiovascular measures com-pared to other emotions (Schwartz, Weinberger, & Singer, 1981).(f) State anger when provoked was measured by the 10-item StateAnger Scale (SAS; Spielberger, 1988), on which participants rateitems o n a l - 4 ( l = not at all to 4 = very much so) scale based onfeelings present in the experimental provocation (score = 10-40).Alpha reliabilities range from .88 to .95 with highest reliabilitiesfor college students (Spielberger, 1988; Spielberger et al., 1983).The SAS correlates minimally with the TAS under neutral condi-tions (Spielberger, 1988) but to a greater degree under provocativeconditions (r = .55; Deffenbacher, Demm, & Brandon, 1986;Deffenbacher & Sabadell, 1992). The SAS gathered when theindividual is provoked also correlates with other anger measuressuch as the AI and Anger Symptom Index (rs = .60 and .37,respectively; Deffenbacher, Demm, & Brandon, 1986). (g) Copingtendencies when provoked were assessed by the 6-item CopingStrategies measure (CS; Novaco, 1975), on which individualsindicate the probability (1-7 scale from low to high probability) ofcoping in the manner described in the item (two items eachmeasure physical antagonism, verbal antagonism, and constructivecoping). These two-item measures correlate with each other andstate anger, and to a lesser degree with other anger measures(Deffenbacher, Demm, & Brandon, 1986; Deffenbacher & Sa-badell, 1992). (h) General styles of expressing anger were mea-sured by the Anger-In (AX-I) and Anger-Out (AX-O) scales of theAnger Expression Inventory (AX; Spielberger, 1988). AX-I andAX-O are 8-item, Likert-type (1 = almost never to 4 = almostalways) measures of how individuals generally handle their angerby suppressing or holding anger in (AX-I) and by expressing angeroutwardly (AX-O), typically in negative ways such as cursing orthrowing things. AX alpha reliabilities range from .73 to .84(Spielberger, 1988). AX-I is essentially uncorrelated with AX-O(Spielberger, 1988), and discriminant validity for the two measuresis reflected in different patterns of correlations with anger, person-ality, and physiological variables (Deffenbacher, 1992; Lopez &Thurman, 1986; Spielberger, 1988).

Resting blood pressure and pulse. Resting blood pressure andpulse measures were taken on the Astropulse Model 90 AutomaticDigital Electronic Blood Pressure/Pulse Monitor manufactured byMarshall Electronics, Skokie, Illinois. Readings are accuratewithin 3 mm Hg for blood pressure and 5% of pulse rates. Tworeadings were taken for each participant and averaged.

Trait Anxiety Inventory. The Trait Anxiety Inventory (TAI;Spielberger, Gorsuch, & Lushene, 1970) is a 20-item, Likert-type(1 = almost never to 4 = almost always) questionnaire on whichparticipants report how generally anxious they feel. TAI has in-ternal consistency reliabilities of .89-.90, test-retest reliabilities of.73-.86 over 3-week to 3-month intervals, and positive correlationswith many other anxiety indexes (Spielberger et al., 1970).

Procedure

The screening questionnaire was administered in five introduc-tory psychology classes, each of approximately 150-200 students.This single-page questionnaire included the TAS, a place to indi-cate whether students believed they had a personal problem withanger and desired help for that problem, a place to indicate thatstudents thought they had no personal problem with anger, andplaces for name, address, and telephone number should studentswish to participate in studies on anger. Interested students scoringin the upper quartile of the TAS quartile and indicating a personalproblem with anger and desiring help, or in the lower quartile ofthe TAS and indicating no personal problem with anger wereassessed in groups of 10-20 in a small university classroom. After


informed consent procedures, students completed, in order, theAX, AI, TAI, Anger Symptom Index, and Anger Situation. Whilestudents were completing questionnaires, the experimenter, anadvanced female graduate student who was unaware of the partic-ipant's anger level, randomly drew students to a back corner of theroom and obtained resting pulse and blood pressure. When all hadcompleted questionnaires and resting physiological measures, theywere instructed in obtaining arterial pulse on the nondominantwrist or carotid artery. They practiced for 15-s intervals until allcould do so proficiently. Then they were instructed to listen to anaudiotape of a provocative scene, and taped instructions asked theparticipants to imagine themselves in the situation as if it werehappening to them at the moment. The scene was modeled afterNovaco's (1975) current political event scene and depicted anintense interaction at a party during which the individual is criti-cized and put down for his or her views on a current political topic(i.e., nuclear arms). Following a 2-min visualization of the scene,participants monitored and recorded their pulse (provoked heartrate). Then they completed the SAS and the CS regarding theirreactions during the imagined provocation. Following the comple-tion of state measures, participants were given the Anger Log andinstructed in how to record the intensity of the most angering eventof the day and how to return the log in a week and receive researchcredit.

Results and Discussion

A 2 X 2 (Gender X Anger) MANOVA revealed signifi-cant multivariate effects for level of anger and gender,Fs(18, 68) = 26.01 and 2.70, respectively, ps < .01, but notfor the interaction, F(18, 68) = 1.29. Significant univariategender effects were found on two variables; men had highersystolic blood pressure and lower trait anxiety (Ms =125.56 and 38.00, respectively) than women (Ms = 111.76and 40.87, respectively), Fs(l, 85) = 27.71 and 10.54,respectively, ps < .01. Remaining findings are summarizedin terms of anger level (Table 1). Theoretically predicted,

significant univariate anger main effects were found on allbut heart rate after provocation and Anger Log frequencygreater than 30. Moreover, even though the Anger SymptomIndex only called for rating one symptom, a significantlyhigher proportion of high- (79%) than low- (21%) angerindividuals spontaneously reported more than one symp-tom, z = 10.00, p < .001, further reflecting greater per-ceived physiological involvement for high-anger individu-als. In terms of anger expression style, high-angerparticipants reported greater tendencies to both suppressanger and negatively express anger. High-anger participantsreported significantly higher trait anxiety as well.

Generally, results supported state-trait anger theory(Spielberger, 1988; Spielberger et al., 1983). For example,evidence for the elicitation hypothesis was found. The vul-nerability to anger elicitation was reflected in the fact thathigh-anger participants reported that many different poten-tial provocations elicited anger. The intensity hypothesiswas also supported. High-anger individuals reported greateranger across many potential provocations, greater anger inidiographically defined ongoing situations, more intenseanger on a day to day basis, more intense anger-relatedphysiological symptoms, and elevated state anger in theexperimental provocation. Only the hypothesized differenceon heart rate when provoked was not found. The negativeexpression hypothesis also received support. High-angerparticipants became more verbally and physically antago-nistic and engaged in less constructive coping in response tothe imagined provocation and were more likely to out-wardly and negatively express anger and to suppress theiranger. Thus, many different findings converged to providesupport for state-trait anger theory and validity for the TAS.Moreover, the clinical value of the TAS was demonstratedbecause not only were these predictions confirmed, but theywere also confirmed in a group whose members were both

Table 1Means and Standard Deviations for High- and Low-Anger Groups in Study 1

Measure

Anger InventoryAnger SituationAnger LogAnger Log-Frequency > 30Anger Log-Intensity > 30Anger Symptom IndexProvoked pulse/15 sState Anger ScaleVerbal antagonismPhysical antagonismConstructive copingAnger Expression-InAnger Expression-OutResting pulse/minSystolic blood pressureDiastolic blood pressureTrait Anxiety Inventory

Low

M

247.5166.5831.443.35

35.7249.8417.5312.184.633.42

10.4216.3913.4269.51

115.3379.9534.42

Group

anger

SD

51.2018.5922.36

2.8925.3920.82

2.713.552.711.622.154.242.739.07

12.449.717.85

High

M

329.1588.1548.354.61

64.4173.4317.8019.548.046.268.35

18.9520.0867.04

118.6376.3344.78

anger

SD

34.649.44

19.001.90

16.8714.952.988.182.873.282.794.543.77

10.8912.919.549.52

Univariateanger effect,

F(l, 85)

77.97*38.62*17.32*7.85

39.43*33.56*0.06

26.10*28.20*22.30*9.92*

11.25*133.73*

2.750.022.66

37.02** p < .05 (alpha level has been adjusted for experimentwise error).


high on anger and also perceived themselves as having apersonal problem with anger and wanting help for thatproblem (i.e., a client analog).

Three other findings are worthy of note. First, high- andlow-anger groups did not differ on resting pulse and bloodpressure. This suggests that differences on the Anger Symp-tom Index and state anger during the provocation were notdue to higher resting physiological arousal for high-angerstudents, that is, basal differences were unlikely to accountfor the findings. Second, gender was relatively unrelated tofindings. There were no gender differences on anger mea-sures, and there were no Gender X Anger interactions,suggesting that high-anger men and women did not differsystematically on anger reactivity or anger expression.Third, high-anger participants reported more trait anxiety, afinding that suggests a closer link between trait anger andtrait anxiety than might be expected.

Study 2

Study 1 employed only a single situation in the experi-mental provocation, limiting the generalizability of resultsin terms of the type of provocations that elicit the theoreticaldifferences between high- and low-anger individuals, thatis, reactions to experimental provocation might, theoreti-cally, be limited to a specific type of provocation. Toaddress this specific limitation in generalizability, Study 2replicated only the analog provocation portion of Study 1but employed a very different type of provocation. If find-ings could be replicated with a different type of provocation,then generalization and confidence would be enhanced. Itwas predicted that high-anger individuals would showgreater state anger, more verbal and physical antagonism,less constructive coping, and perhaps greater changes inheart rate than low-anger participants.

Method

Thirty-six (19 men and 17 women) low-anger and 37 (18 menand 19 women) high-anger students (operationally defined as inStudy 1) received one research credit for participation. With twoexceptions, screening, recruitment, and provocation procedureswere identical to the experimental provocation portion of Study 1.First, a different scene similar to Novaco (1975) was employed and

involved a rude grocery store customer who pushes past theindividual and knocks his or her groceries to the floor withoutapology. Second, heart rate was self-monitored both prior to andfollowing the experimental provocation. Otherwise, trait and stateanger, self-monitored heart rate, and coping measures were iden-tical to Study 1.


A Gender X Anger MANOVA revealed a significantmultivariate effect for anger, F(6, 64) = 9.82, p < .001, butnot for gender or the interaction, Fs(6, 64) = 1.74 and 1.08,respectively. Univariate anger effects (see Table 2) werefound on all but physiological variables, replicating Study 1.High-anger participants again reported more state angerarousal, more physically and verbally hostile responses, andlessened facilitative responding. Only heart rate did notsupport predictions of state-trait theory. Again, both gen-ders showed the same anger reactivity. Thus, this studydemonstrated that the results of the prior study were notlimited to a specific type of provocation, providing addi-tional support for state-trait anger theory and the TAS.

Study 3

Results of the first two studies and prior research (e.g.,Deffenbacher, Demm, & Brandon, 1986) have shown thathigh-anger individuals report more frequent and more in-tense anger experiences, which state-trait anger theory saysoccurs because they possess a personality disposition torespond with anger to potentially frustrating or provocativesituations. A personality variable, however, might not ac-count for the findings, as it is possible that differences infrequency and intensity of anger in daily life occur notbecause high-anger individuals tend to react more withanger but because they encounter significantly differentprovocations. For example, if high-anger individuals areinvolved in intense antagonistic interpersonal relationshipsmore often than low-anger individuals, then the frequenciesand intensities of their anger could be higher as a result offrequent, anger-engendering encounters rather than a per-sonality disposition toward anger. Similarly, high- and low-anger individuals may respond differently. For example,earlier results showed that they tended to cope with and

Table 2Means and Standard

Measure

Deviations for

Low

M

High- and Low-Anger Groups in

anger

SD

Group

High anger

> M SD

Study 2

Univariateanger,

F(l, 69)

Resting pulse/15 sProvoked pulse/15 sState Anger ScalePhysical antagonismVerbal antagonismConstructive coping

* p < .05 (alpha level has been adjusted for experimentwise error).

18.0018.8621.69

3.865.569.28

2.542.516.581.912.382.48

17.4918.6828.707.22

10.326.08

3.323.417.703.512.942.43

0.660.11

16.86*26.79*56.51*29.94*


express their anger in different ways, and they might differin the types of physiological symptoms experienced as well.Study 3 explored these possibilities. Responses of high- andlow-anger individuals on the Anger Situation, Anger Log,and Anger Symptom Index were analyzed to see if thesituational sources of anger or the nature of reported phys-iological manifestations of anger were different for high-and low-anger students.

Method

One hundred ninety-nine (82 men and 117 women) low-angerand 168 (83 men and 85 women) high-anger (operationally definedas in Study 1) students were recruited over four semesters. Eachreceived one research credit for participation for completing theAnger Situation, Anger Symptom Index, and Anger Log (seeStudy 1 for description) in groups of approximately 20 in a smalluniversity classroom. Anger Situations, Logs, and Symptoms werecategorized by one of the initial raters (Lora Eiswerth-Cox) ac-cording to a coding scheme developed by Deffenbacher, Eiswerth,and Stark (1986). Situational sources of provocation for AngerSituation and Anger Log measures were coded into one of thefollowing nine categories: (a) family—anger stemming from aninteraction with a parent, stepparent, sibling or other member ofthe individual's family; (b) interpersonal nonfamily—anger arisingfrom interactions with people other than family members (e.g.,roommates, friends, significant other, etc.) and with people notinvolved in school or work settings; (c) self—anger arising fromthe individual's own behavior, attitudes, or characteristics (e.g.,performing poorly at some task and anger at self for oversleeping);(d) school—anger arising from interactions with fellow students,professors, and university personnel; from university policies andprocedures; or from school-related responsibilities such as home-work, timing of assignments, etc.; (e) work—anger elicited in theworkplace involving work tasks or interactions with employers,supervisors, coworkers, or customers; (f) driving—anger arisingfrom situations involving operating or being a passenger in a motorvehicle or riding a bicycle; (g) object—anger elicited in interac-tions with inanimate objects (e.g., malfunctioning car or comput-er); (h) other—a miscellaneous category for situations not fittingany of the above situations (e.g., anger at a pet or the weather); and(i) no anger reported—participants reported that they did notexperience anger on a given day on the Anger Log. For some

analyses, driving and object categories were combined into theother category because of low sample size. Physiological re-sponses on the Anger Symptom Index were coded into one of thefollowing 10 categories: (a) sweating, (b) shaking (reports ofshakiness or shaky feelings), (c) clenched jaw, (d) shoulder-neck-back tension (reports of tension in one or more of these areas), (e)fast pulse or heart pounding (reports of increased heart rate,palpitations, or both), (f) headaches, (g) high blood pressure, (h)flushed face (reports of blushing or feeling hot in the face), (i)clenched fist, and (j) other (symptoms not in prior categories, e.g.,cold hands or upset stomach). If more than one symptom wasmentioned, the symptom with the highest rating was coded. If twoor more symptoms were listed with equally high ratings, the firstsymptom was coded. In the first semester, two advanced graduatestudents were trained to use the situational coding scheme for theAnger Situation and Anger Log and the symptom coding schemefor the Anger Symptom Index. Interrater agreement for categori-zations was 98% for anger situations, 96% for log entries, and 99%for symptoms. During the last three semesters of data collection,one of the original raters (Lora Eiswerth-Cox) coded all protocols.Each semester, 25 protocols were randomly drawn and rated byJerry L. Deffenbacher. Because rates of interrater agreement wereas high as or higher than in the initial phase, rater reliabilitythroughout the study was judged as satisfactory.


A series of preliminary analyses revealed that gender wasunrelated to findings, and data therefore were collapsedacross gender. The general outcomes for the Anger Situa-tion, the Anger Log, and the Anger Symptom Index repli-cated findings from Study 1. Specifically, high-anger stu-dents reported significantly more intense anger in their mostangering, ongoing provocation, higher daily anger levels,and more severe anger-related physiological symptoms(Ms = 79.79, 45.78, and 71.82, respectively) and than low-anger students (Ms = 60.76, 28.51, and 49.33, respective-ly), Fs(l, 363) = 72.73, 143.54, and 89.45, respectively,ps < .001.

When Anger Situations were broken down by type ofsituation (Table 3), high- and low-anger participants did notdiffer by type of situation in which anger was experienced,

Table 3Means, Standard Deviations, and Percentages of Sample Endorsing Each Type of MostAngering Situation in Study 3

Group

Type ofsituation

Interpersonal-NFFamilySelfSchoolWorkOther3

%

52.316.111.67.03.59.5

Low anger

M

63.6164.3651.5045.3672.8656.74

SD

22.1220.8223.9524.1425.7425.74

%

53.614.38.95.49.58.3

High anger

M

81.5281.5474.4067.7878.9480.29

SD

15.9414.9823.7529.9013.7421.20

Angereffect,

F

40.38*11.78*8.20*3.930.927.80

Note. Interpersonal-NF = interpersonal nonfamily.a Driving and object categories were collapsed into the other category because of an insufficientnumber of responses.* p < .05 (alpha level has been adjusted for experimentwise error).


X*(5, N = 367) = 6.63, but did differ in the level of angerexperienced in those situations (Table 3). High-anger par-ticipants reported significantly higher anger in interpersonalnonfamily, family, and self-relevant contexts. High- andlow-anger groups did not differ in intensity of anger relatedto work and school or "other" contexts, but these wereendorsed by a relatively small number of participants. Thatis, for their most angering ongoing situations, high- andlow-anger individuals did not differ in the types of situa-tions reported, but high-anger individuals experienced moreintense anger in nearly every type of situation.

When day-to-day angering situations were broken out bytype of situation, the percentages of different sources ofdaily anger for high- and low-anger participants, respec-tively, were as follows: interpersonal nonfamily (40.2% vs.31.5%), family (4.9% vs. 4.2%), self (17.4% vs. 21.3%),school (4.9% vs. 7.5%), work (3.4% vs. 2.5%), driving(2.7% vs. 2.2%), objects (4.3% vs. 4.2%), other (14.0% vs.17.6%), and no anger logged on that day (8.1% vs. 15.5%).Given the very high numbers involved (i.e., a large numberof participants with seven daily entries for each), evenrelatively small differences were statistically significant,^ ( 8 , N = 2543) = 60.72, p < .001. However, inspection ofthe data suggest that, in general, high- and low-anger indi-viduals tended to report about the same proportions ofdifferent sources of daily provocation and that apparentdifferences primarily were due to more low-anger partici-pants reporting no anger for the day and more high-angerindividuals reporting anger in interpersonal, nonfamily sit-uations. High-anger participants, however, reported signif-icantly more intense anger in all categories than low-angerindividuals (Table 4). That is, except for more low-angerindividuals who reported no anger-involving incidents,high- and low-anger groups did not differ greatly in thekinds of provocative situations they encountered, but high-anger individuals experienced more anger in every type ofsituation. This finding is consistent with state-trait angertheory in that low-anger individuals are expected to expe-rience lower levels of anger and therefore might be expected

to report more days in which no situation aroused a notice-able level of anger. High-anger participants were morelikely to report interpersonal provocations that did not in-volve family, but otherwise, the types of situations thataroused anger were similar for high- and low-anger indi-viduals, and high-anger participants reported more intenseanger in nearly all types of day-to-day situations.

High- and low-anger groups did not differ significantly inthe types of anger-related physiological symptoms de-scribed, ^ ( 8 , N = 367) = 12.01, but did differ on thestrength of their responses on many different symptoms(Table 5). High-anger participants reported more intensesweating, shakiness, clenched jaws, shoulder-neck-backtension, and elevated heart rate and heart palpitations. Be-cause of the low response rates, the large mean differencesin intensity of responses for reports of flushed face andelevated blood pressure were not significantly different.Serendipitously, the Anger Symptom Index provided addi-tional information. Even though the measure asked for onlyone physiological reaction, some participants spontaneouslygave and rated two or more symptoms. A significantlyhigher proportion of high- (33%) than low- (19%) angerparticipants reported more than one symptom, ^ ( 1 , N =367) = 8.30, p < .01, and the intensity of these secondarysymptoms was also significantly greater for high-anger par-ticipants (M = 72.88) than low-anger individuals, M =54.82, F(l, 92) = 15.73, p < .001. Overall, high- andlow-anger participants did not differ in the frequency withwhich they report different types of anger-related physio-logical responses but did differ in response intensity, withhigh-anger students reporting greater physiological involve-ment.

In summary, the predictions of state-trait anger theoryheld up, and the alternative possibilities did not. Althoughthere may be differences not captured in the coding schemeemployed, high- and low-anger individuals did not differgreatly in the types of situations that aroused anger. That is,high-anger participants did not seem to live in environmentsthat presented different provocations but did appear to differ

Table 4Average Daily Anger Arousal as a Function of Type of Situation and Anger Status inStudy 3

Type ofsituation M

Low anger

SD

Group

High

M

anger

SD

Angereffect,

Interpersonal-NFFamilySelfSchoolWorkOther"

36.1536.3030.0132.1530.096.54

21.8424.0421.0724.4221.5614.25

54.9752.6847.2150.7346.6519.49

20.4423.6224.3224.5724.3130.79

61.67*9.21*

35.15*15.36*33.50*14.54*

Note. Interpersonal-NF = interpersonal nonfamily."When participants had more than one entry per category in a week's time, intensities wereaveraged over the number of entries and a single score was entered into the analyses, such that noperson entered analyses more than once. b Driving and object categories were collapsed into theother category.* p < .05 (alpha level has been adjusted for experimentwise error).


Table 5Means, Standard Deviations, and Percentages of Sample Endorsing Type of StrongestAnger-Related Physiological Symptom in Study 3

Type of symptom

SweatingShakinessClenched jawSNB tensionFast pulseHeadacheHigh blood pressureFlushed faceOther

%

6.57.5

13.118.119.611.14.09.0

11.1

Low anger

M

36.9247.3347.6450.5854.1348.6446.6345.2854.32

Group

SD

21.2624.8526.8322.2726.6422.5814.3522.3929.58

%

7.713.715.513.115.58.94.84.2

16.7

High anger

M

67.0878.0073.9269.8275.3967.4766.1457.1472.68

SD

18.9112.6418.5022.0119.4220.8621.0421.3824.00

Angereffect,

F

14.59*25.28*16.84*10.28*12.20*6.594.201.455.89

Note. SNB = shoulder-neck-back.* p < .05 (alpha level has been adjusted for experimentwise error).

from low-anger individuals primarily in the level of angerexperienced. The same was true for reported physiologicalsymptoms; that is, high-anger individuals appeared to differprimarily in terms of the intensity of the physiologicalarousal, not in the type of responses.

Study 4

Studies 1-3 provided evidence suggestive of greater dif-ficulties and negative consequences for high-anger individ-uals. For example, because high trait anger individualsexperienced more frequent and intense anger, it is reason-able to predict that they may also experience more frequentand more severe negative outcomes or consequences fromtheir anger. Their tendencies to cope with provocation viaverbal and physical antagonism and with fewer constructiveactions could exacerbate negative interpersonal interactions,thereby increasing the odds of interpersonal aggression andother negative outcomes. Greater tendencies toward nega-tive outward expression of anger could result in greaterprobabilities of negative expressions and potential conse-quences to self, others, and property. Heightened psycho-physiological responses to provocation and greater angersuppression could increase the probability of negativehealth consequences as well. In sum, it is reasonable topredict that the high trait anger individual's more intenseand frequent anger, arousal and more dysfunctional ways ofexpressing and coping with anger could eventuate in morenegative outcomes. However, it is also possible that al-though high-anger individuals experience more frequentand intense anger, they do not experience any differentresults or outcomes of their anger, that is, they might expe-rience differences in emotional arousal and expression butnot in the consequences of either the arousal or the expres-sion thereof.

There is, however, very little research that directly ad-dresses this negative consequence hypothesis of state-traitanger theory. Averill (1983) and Kassinove and Sukhodol-sky (1995) reported that many individuals rated the out-comes of anger as positive. However, their findings were for

people in general and were not related to the person's traitanger level. Another study (Hazaleus & Deffenbacher,1986) reported that high trait anger individuals experiencedfairly severe negative consequences; however, participantswere clients in a treatment study, and there were no com-parative base rates of anger consequences in low-angerindividuals. Desnoes and Deffenbacher (1995) included alow-anger comparison group, but findings were reported interms of greater general life interference rather than adverseconsequences per se. We found no research that addressedthe frequency of fairly common anger consequences. Spe-cifically, do high-anger individuals experience more nega-tive anger-related consequences on a day-to-day basis?Study 4 addressed this question (consequence hypothesis)by comparing the frequencies with which high and low traitanger individuals' anger made them do, feel, or experiencedifferent consequences over a 2-month period. Addressingthis question is not only important in terms of testing atheoretical extrapolation of state-trait anger theory but alsoimportant clinically in terms of the utility of the TAS as atool that identifies individuals whose anger leads to adverseoutcomes and who therefore may be in need of psycholog-ical assistance.

Method

Seventy-one (31 men and 40 women) low-anger and 72 (36 menand 36 women) high-anger students (operationally defined as inStudy 1) received one research credit for completing an inventoryof anger-related consequences constructed for this study, an in-strument that was constructed because no measures of the frequen-cies of anger-related consequences were found in the literature.Based on interviews with high-anger participants in a pilot study,an inventory of 33 different anger-related outcomes (e.g., becamedepressed, hit someone, damaged a friendship, got drunk, etc.) wasconstructed. Participants completed this inventory of consequencesby circling a number from 0 to 4 or more (a score of 4 or more wasanalyzed as a 4) regarding the number of times in the last 2 monthsthat their anger led them to experience the consequence describedby that item. Participants completed the inventory in groups ofapproximately 20 in a small university classroom.



A TRYSYS key cluster variable analysis (Dean & Win-ters, 1991; Tryon & Bailey, 1966) on the consequencemeasure led to eight consequence clusters. Cluster content,items, and alpha reliabilities were as follows: (a) four itemsinvolving drug and alcohol use (drink alcohol, get drunk,and use other drugs; a = .87); (b) four items involvingphysical assault on others (feel like hurting someone, getinto a physical fight, hurt someone, and hit someone; a =.91); (c) three items involving physical assault on objects(feel like breaking something, hit a wall or something, andbreak something; a = .89); (d) three items involving hurt-ing self or being physically ill (hurt self physically, feel likekilling self, and feel physically ill; a — .71); (e) two itemsinvolving verbal assault (get into an argument and say nastythings; a = .63); (f) six-items about experiencing negativeemotions (feel bad about self, feel ashamed, depressed,withdrawn, feel dumb, and feel embarrassed; a = .88); (g)three items involving relationship problems (make friendsmad at me, damage a friendship, and make friends afraid ofme; a = .77); and (h) three items involving legal or voca-tional difficulties (trouble with the law, hurt my job, and getinto an accident; a = .76). Six items had communalities lessthan .20 (e.g., fighting with family and overeating) and didnot enter consequence clusters. Correlations among conse-quence clusters revealed that 22 of 28 formed modest,significant, positive correlations with rs from .15 to .40.Three correlations were significantly larger than this, that is,correlation of physical assault on others with physical as-sault on objects and with verbal assault (rs = .61 and .50)and of negative emotions with hurting self or feeling phys-ically ill (r = .47). Three correlations were nonsignificant,that is, the correlation of negative emotions with physicalassault on others and on objects and with legal or vocationalconsequences (rs = .01, .08, and .12, respectively). Overall,these correlations suggest a moderate relationship amongthe frequencies of negative anger-related consequences, yetindependence enough among types of consequences to al-low meaningful comparisons as a function of gender andanger level.

Means and standard deviations for consequence clustersare presented in Table 6. A Gender X Anger MANOVA onthe consequence clusters revealed significant multivariategender, anger, and interaction effects, Fs(8, 132) = 5.87,12.61, and 4.36, respectively, ps < .001. Univariate Gen-der X Anger interactions were found on consequence clus-ters involving drugs and alcohol, physical assault on people,physical assault on property, hurting one's self, and verbalassault, Fs(l, 139)= 5.11, 22.49, 15.22, 8.29, and 4.16,respectively, ps < .05, and were explored further withTukey post hoc tests. When the alpha level was adjusted forexperimentwise error, differences involving drugs, alcohol,and verbal assault were no longer significant. For conse-quences involving hurting one's self, high-anger men re-ported negative consequences more often than low-angerwomen, whereas both high and low anger women did notdiffer significantly from either male group. For conse-quences involving physical assault on others and property,high-anger men reported a significantly higher frequency ofthese consequences than other groups, which did not differamong themselves. Gender effects for assault on personsand property primarily were due to the effects of high-angermen, as noted previously. Women more frequently reportednegative emotions stemming from their anger than did men.Univariate anger effects were found for all consequencecluster scores, save hurting one's self and legal or voca-tional consequences (Table 6). That is, high-anger studentsreported that their anger led them to greater use of drugs andalcohol, more frequent physical assault on others and onproperty, greater verbal assault, more negative emotions,and more relational difficulties. As noted above, however,anger's main effects on physical assault on others andproperty were heavily influenced by the higher levels re-ported by high-anger men. Even though the item "fight withfamily" did not enter a consequence cluster, data for thisitem are presented because of their implications for familydistress and turmoil. A univariate anger effect was found,F(l, 139) = 14.70, p < .001, with high-anger individuals(M = 1.92 for women and 2.31 for men) reporting more

Table 6Means and Standard Deviations for Frequency of Anger Consequences in Last 2Months for High- and Low-Anger Groups in Study 4

Consequence

Drugs-alcoholAssault-peopleAssault-objectsHurt selfAssault-verbalNegative emotionsRelationshipLegal-school

Low

Men

M

4.524.684.003.394.00

10.393.543.23

SD

2.931.141.610.801.634.031.670.62

anger

Group

Women

M

4.954.703.754.234.75

14.453.653.05

SD

2.801.381.791.492.226.011.170.22

High

Men

M

8.3610.178.254.977.75

15.335.723.53

SD

4.544.704.012.692.307.382.963.05

anger

Women

M

6.085.644.424.147.08

17.505.283.13

SD

3.722.662.791.271.996.332.880.42

Univariateangereffect,

F(l, 139)

17.23*44.88*28.67*6.66

76.63*15.13*24.29*

1.13' p < .05 (alpha level has been adjusted for experimentwise error).


fights with family members than low-anger participants(M = 1.23 for women and 1.60 for men).

High-anger individuals experienced significantly morefrequent adverse consequences as a result of their anger.Moreover, these were robust findings, as anger effects werefound for nearly all consequences, confirming the negativeconsequences hypothesis of state-trait anger theory. Theseconclusions, however, must be qualified by the significantgender by anger interactions for consequences related tophysical assault on people and on property, in which high-anger men experienced greater numbers of adverse conse-quences than other groups. Overall, however, results sup-ported the extension of state-trait anger theory to thefrequency of negative consequences as high-anger individ-uals suffered significantly more frequent adverse outcomesfrom their anger.

Study 5

Study 4 found that high-anger individuals experiencedmore frequent anger-related consequences but did not de-termine whether they experienced more severe negativeconsequences. Study 5 addressed this possibility by assess-ing the severity of anger consequences in the individual'stwo worst anger incidents in the last year. The worst anger-related incidents in the last year were chosen for two rea-sons. First, choosing the most significant anger-involvedevents provided for the necessary range of outcome sever-ities, which might not be found for more frequently occur-ring consequences. Second, a year period provided an op-portunity for significant events to occur and yet allowedparticipants to be able to recall detail with reasonableaccuracy.

Method

Ninety-two (40 men and 52 women) high-anger and 102 (48men and 54 women) low-anger students (operationally defined asin Study 1) received one research credit for completing the Angerin the Last Year Questionnaire, which was developed for thisstudy. In a pilot study, 175 students in abnormal psychologydescribed, in writing, their two worst anger-related incidents in thelast year and listed consequences resulting from their anger. Seventypes of consequences were identified from these lists: (a) physicaldamage to self (anger resulted in bodily damage or disease tohimself or herself), (b) physical damage to others (anger resultedin bodily damage or disease to another person), (c) physicaldamage to objects (anger led to physical destruction of or damageto property), (d) relationship damage (anger interfered with, dam-aged, or terminated one or more of his or her relationships, exceptfor relationships in school or work settings), (e) vocational orschool problems (anger led to difficulties with relationships orperformance at school or work), (f) legal consequences (angerculminated in involvement with police, courts, or quasi-legal sys-tems such as campus disciplinary processes), and (g) damage toself-esteem (anger resulted in negative feelings about or interpre-tations of self). Because consequence descriptions in the pilotstudy were not sufficiently detailed to allow for ratings of theirseverity, the format was modified to solicit more thorough conse-quence descriptions. In the revised Anger in the Last Year Ques-

tionnaire, participants first described, in detail, the worst angerevent in the last year. Then for each type of consequence listedabove, they indicated whether it had occurred or not, and if it had,they described the consequences in detail. After responding to allspecific consequences, participants rated on a 5-point scale (0 =no cost, 1 = a little, 2 = somewhat, 3 = very, 4 = extremely) howcostly the incident was, all things considered. This process wasthen repeated for the second most anger-involved incident in thelast year.

Consequences were coded for severity on a 0-3 scale (0 = noconsequence, 1 = mild, 2 = moderate, 3 = severe), except fordamage to self-esteem, for which moderate and severe categorieswere collapsed because of the inability to distinguish clearlybetween them based on information supplied by participants. Cat-egories were anchored to specific characteristics or parameters(e.g., ratings for physical damage to self or others involved thenumber of symptoms suffered, amount of pain reported, the levelof medical attention involved, and the nature and lasting charac-teristics of physical problems). To increase reliability and stan-dardization of judgments, (a) consequence categories were asmutually exclusive as possible, (b) only actual outcomes wererated (i.e., even though behavior could potentially lead to moresevere consequences, only the consequences actually experiencedwere coded), and (c) when consequences were vague, incomplete,or open to multiple interpretations, raters coded the less severeconsequence level. Interrater reliability (Pearson r of one rater'sseverity score with another rater's severity score for the conse-quence category) was established by having three people, two ofthe authors and a female political science undergraduate studentwho was unfamiliar with the project, independently code 100questionnaires from the present data set (25 drawn randomly fromeach of the four male-female, high-low anger combinations). Theundergraduate rater was given the Consequence Rating Systemmanual and approximately three hours of training by one of theauthors. Interrater reliabilities were high for all consequence cat-egories, ranging from .88 to 1.00 (Mdn r = .94). Leibsohn et al.(1994) found similar interrater reliabilities when applying the samecoding system to alcohol-involved consequences. Because reli-abilities were uniformly high in this and other studies, one author(Gregory A. Thwaites), who was one of the coders in the estab-lishment of interrater reliability and who was unaware of partici-pant anger level and gender, coded the remaining consequenceprotocols within the next week's time.


Separate 2 X 2 (Gender X Anger) MANOVAs for eachanger incident (Table 7) revealed significant multivariatemain effects for anger, Fs(7, 185) = 7.99 and 5.12, respec-tively, ps < .001, but not significant interactions, Fs(7,185) = 1.03 and 0.57, respectively. The multivariate gendereffect was significant for the worst anger incident, F(7,185) = 2.80, p < .01, but not for the second incident, F(7,185) = 1.87. Univariate analyses of the gender effect on themost anger-involved incident revealed no significant differ-ences, however. Univariate analyses of the multivariateanger effects (Table 7) showed that in the worst incident,high-anger participants reported more severe consequencesin all categories, except for work or school problems, legalconsequences, and self-esteem. In the second most anger-involved situation, high-anger individuals experienced moredamage to relationships, self-esteem, and overall costs. Le-


Table 7Severity of Anger Consequences by Gender and Anger Status in Study 5

Consequence

Physical-selfPhysical-otherPhysical-objectsRelationshipsWork-schoolLegalSelf-esteemOverall cost

Physical-selfPhysical-otherPhysical-objectsRelationshipsWork-schoolLegalSelf-esteemOverall cost

Low

Men

M SD

anger

Group

Women

M SD

High

Men

M

Most anger-involved incident ii0.340.090.230.800.500.020.480.75

0.810.470.710.980.950.150.700.87

Second most0.160.000.250.860.230.000.230.61

0.480.000.751.070.680.000.420.81

0.270.040.061.250.440.000.730.95

0.170.190.231.020.810.000.711.03

0.600.520.661.790.740.130.811.72

SD

anger

Women

M

l last year0.840.991.091.131.090.590.811.32

0.490.190.331.930.740.050.861.72

: anger-involved incident in last year0.240.070.091.200.420.000.350.82

0.770.420.481.030.880.000.550.88

0.340.190.511.360.380.040.591.64

0.850.621.051.210.880.280.801.49

0.260.090.301.840.720.000.841.77

SD

0.850.590.781.181.090.310.801.24

0.660.430.641.041.160.000.811.77

Univariateangereffect,

F(l, 191)

10.58*10.26*10.19*28.85*

3.642.434.60

28.57*

0.942.684.59

12.93*2.980.84

19.55*36.78*

* p < .05 (alpha level has been adjusted for experimentwise error)

gal consequences were seldom reported and were droppedfrom subsequent analyses.

From the above analyses, it is difficult to determine ifanger effects were due to a few high-anger individualsreporting very severe consequences, to more high-angerindividuals reporting mild consequences, or both. To ex-plore these issues, the proportions (Table 8) of individualsreporting some type of consequence in a category (i.e.,persons with ratings > 0) were compared, and the intensityof these consequences analyzed by 2 X 2 (Gender X Anger)analyses of variance (ANOVAs; Table 9). For the worstanger-involved incident, greater proportions (Table 8) ofhigh-anger participants experienced a negative consequencein all categories, except for work or school problems anddamage to self-esteem. For the second incident, signifi-cantly higher proportions of high- than low-anger partici-pants experienced consequences in all categories, except forwork or school consequences and physical damage to self.When participants who did not report a consequence weredropped from analyses for the worst incident, ANOVAsrevealed no significant gender or gender by anger interac-tions, but demonstrated anger effects on relationship dam-age and overall cost consequences (Table 9) where high-anger individuals (Ms = 2.31 and 2.17, respectively)experienced more severe consequences than low-anger stu-dents (Ms = 1.76 and 1.55, respectively). For the secondincident, although no gender or interaction effects werefound, significant anger effects (Table 9) were found onself-esteem and overall cost ratings where high-anger par-ticipants (Ms = 1.45 and 2.33, respectively) experiencedmore severe consequences than low-anger individuals(Ms = 1.08 and 1.41, respectively).

As hypothesized, in situations involving high levels ofanger, high-anger individuals experienced greater conse-quences in terms of physical damage to self, others, andproperty, relationship damage, lowered self-esteem, andoverall costs. Examination of findings, however, showedthat the differences between groups generally occurred be-cause greater numbers of high-anger participants experi-

Table 8Proportions of High- and Low-Anger ParticipantsExperiencing Consequences in Study 5

Group

ConsequenceLowanger

Highanger

Difference inproportions test (z)

Most anger-involved incident in last yearPhysical-selfPhysical-otherPhysical-objectsRelationshipsWork-schoolSelf-esteemOverall cost

Second mostPhysical-selfPhysical-otherPhysical-objectsRelationshipsWork-schoolSelf-esteemOverall cost

.17

.04

.08

.60

.25

.49

.55

.45

.20

.26

.80

.35

.58

.79

anger-involved incident in.10.02.07.58.17.27.52

.17

.08

.21

.75

.23

.48

.73

6.22*4.95*4.85*4.55*2.201.964.80*

last year2.332.84*2.75*3.71*1.424.30*4.20*

* p < .05 (alpha level has been adjusted for experimentwise error).


Table 9Anger Consequences Dropping High- and Low-Anger Participants Not Reporting aConsequence in Study 5

Consequence

Physical-selfPhysical-otherPhysical-objectsRelationshipsWork-schoolSelf-esteemOverall cost

Physical-selfPhysical-otherPhysical-objectsRelationshipsWork-schoolSelf-esteemOverall cost

Low

Men

M SD

anger

Group

Women

M SD

High

Men

M

Most anger-involved incident ii1.182.002.001.752.001.311.43

0.831.411.000.640.770.480.66

Second most1.400.002.001.902.001.001.42

0.550.000.840.720.710.000.61

1.671.001.001.771.711.251.63

0.870.000.000.740.610.440.83

1.452.002.062.212.051.432.22

SD

anger

Women

M

I last year0.670.880.900.800.780.501.06

1.621.831.752.442.131.422.11

: anger-involved incident in last year2.602.002.501.781.921.121.41

0.551.410.710.710.790.330.71

2.001.672.452.122.221.482.49

1.001.030.690.810.670.511.12

1.572.001.442.082.381.442.17

SD

0.800.890.890.700.640.501.02

0.790.000.530.850.650.510.90

Univariateanger

effect, F

1.120.351.18

18.24*1.582.25

14.45*

0.440.001.603.242.02

13.56*30.17*

* p < .05 (alpha level has been adjusted for experimentwise error)

enced a consequence than low-anger participants. For ex-ample, for physical damage to self, others, and objects,high-anger students were from two to five times more likelyto report a consequence than low-anger students, but forthose who reported a consequence, the severity of theseconsequences was not significantly different. Only for rela-tionship damage in one incident, self-esteem in the other,and overall costs in both situations were higher proportionsof high-anger participants found to experience more severeconsequences. Thus, this study supported the prediction thathigh-anger individuals suffer greater anger-related conse-quences, but with few exceptions, differences occurred be-cause high-anger individuals were more likely to suffer aparticular consequence, not because the consequence wasmore severe.

There were no significant gender interactions in thisstudy, suggesting that when the most severe anger-relatedincidents are considered, men and women are equally likelyto suffer similar consequences. When day-to-day conse-quences were considered in Study 4, there were genderinteractions with high-anger men being more likely to ex-perience adverse consequences, particularly those involvingphysical assault on others and objects. It appears that on aday-to-day basis, high-anger men are more likely to engagein behaviors that create these kinds of consequences, butthat when the most severe situations occur, high-angerwomen "catch up" with high-anger men and are equallylikely to suffer these kinds of consequences.

Frequent, less severe anger consequences and infrequent,more severe consequences should be studied separately asthey may contribute differentially to various aspects offunctioning. For example, frequent, day to day conse-

quences may have greater relevance to the quality of theindividual's life and to incremental disease processes, suchthat high-anger individuals generally, and in some caseshigh-anger men specifically, may be more affected by theseday-to-day consequences. On the other hand, less frequent,more severe consequences may have greater impact onvocational and legal aspects of the individual's world, andhigh-anger men and women may suffer equally from inthese more severe outcomes.

Study 6

Studies 1-5 provided data on convergent validity for theconstruct of trait anger, showing that high trait anger indi-viduals experience more frequent and intense anger, expresstheir anger less constructively, and experience more fre-quent and, in some cases, more severe consequences fromtheir anger. Construct validity, however, would be strength-ened by evidence of discriminant validity. If, as the theorysuggests, trait anger is an individual difference variablepredictive of the frequency and intensity of state anger, thennot only should trait anger correlate positively with thefrequency of state anger (convergent validity), but also thiscorrelation should be stronger than the correlation of traitanger with other frequent negative emotions and behaviors(discriminant validity). That is, trait anger should predict thefrequency of state anger and should predict it with greaterpower than it predicts the frequency of other nonangryemotions and behaviors. Study 6 evaluated the discrimina-tion hypothesis by correlating trait anger with the number oftimes the individual had been angry, anxious, depressed,and intoxicated in the last month.


Method

At the end of the first class, 880 students (536 women and 344men) in five introductory psychology classes completed a single-page questionnaire and placed it in a box as they exited. Thequestionnaire consisted of (a) the TAS; (b) ratings of the numberof times (0 to 5 or more) the student had been angry, anxious,depressed, and drunk in the last month; and (c) places for age andgender, and name and phone number if interested in research onanger.


In all analyses the category of 5 or more was treated as a5. A one-way (Gender) MANOVA across measures re-vealed a significant multivariate gender effect, F(5, 874) =14.48, p < .001. Significant univariate gender effects werefound on four variables, with men reporting higher traitanger and intoxication and lower anxiety and depression(Ms = 21.19, 2.33, 3.73, and 2.03, respectively) thanwomen (Ms = 20.25, 1.67, 3.96, and 2.56, respectively),Fs(l, 878) = 6.85, 24.19, 6.39, and 22.50, respectively,ps < .05. Eta square values, however, revealed that the largesample size had allowed small effect sizes to be significant,as gender effects actually accounted for less than 1% of thevariance of the TAS and anxiety frequency and less than 3%of variance in times depressed and drunk. Even thoughgender differences on these variables were small, separatecorrelation matrices for men and women (Table 10) wererun in order to assess possible gender differences in rela-tionships among variables. Tests for differences in depen-dent correlations showed that the correlation of the TASwith anger frequency was significantly stronger than thecorrelation of the TAS with anxiety, depression, or intoxi-cation frequencies for both men, fs(341) = 6.54, 4.93, and6.80, respectively, and for women, fc(533) = 6.64, 5.13,and 6.87, respectively, ps < .001. The correlation of theTAS with frequency of anger accounted for 31 % of variancefor men and 23% for women, whereas the correlation of theTAS with other variables only accounted for 1% to 8% ofvariance. Tests for differences in strengths of independentcorrelations showed that men and women did not differ inthe strengths of relationships among variables. In summary,the two primary predictions of state-trait theory were sup-ported. Specifically, trait anger was positively related to

frequency of state anger (elicitation hypothesis) and was abetter predictor of the frequency of state anger than it was ofthe frequencies of other common emotions such as anxietyand depression and behaviors such as becoming intoxicated(discrimination hypothesis).

Study 7

Study 7 provided evidence of both convergent and dis-criminant validity relative to the frequency of anger. Traitanger, however, should predict not only the frequency ofstate anger and anger-related behaviors but also the intensityof such conditions. Additionally, trait anger should be morestrongly correlated with the intensity of anger-related con-ditions than with the intensity of other common emotionssuch as anxiety and depression, or psychological states suchas psychoticism and paranoid ideation. Study 7 extendedStudy 6 by correlating trait anger with the intensity ofanger-hostility in the last week as measured by the Symp-tom Checklist-90-Revised (SCL-90-R; Derogatis, 1983)and by comparing the strength of these correlations withthose between trait anger and other SCL-90-R symptomintensities.

Method

Participants

Participants were 178 volunteer students (64 men and 114women; 8 of the original 186 opted not to participate) from anupper division abnormal psychology class (Afdn age = 21). Forparticipation, students earned 6 points (in a system of more than400) toward a final grade.

Instruments

Trait anger was measured by the TAS, and intensity of thefollowing characteristics were measured by scales from the SCL-90-R: (a) anger-hostility by the 6-item Hostility scale reflectingthoughts, feelings, and actions relative to anger (e.g., temperoutbursts, frequent arguments, and urge to harm); (b) somatizationtendencies by the 12-item Somatization scale reflecting distressarising from perception of bodily dysfunction (e.g., headachesand dizziness); (c) obsessive-compulsiveness by the 10-itemObsessive-Compulsive scale reflecting irresistible thoughts (e.g.,

Table 10Correlations Among Trait Anger and Last Month's Frequency of Anger, Anxiety,Depression, and Intoxication for Men and Women in Study 6

Measure

1. Trait Anger Scale2. Angry3. Anxious4. Depressed5. Drunk

M

.56—

2

W

.48—

3

M

.23

.38—

W

.17

.24

4

M

.29

.41

.25—

W

.25

.39

.29—

5

M

.18

.22

.18

.17—

W

.08

.14

.07

.10—

Note. Except for the Trait Anger Scale, measures refer to the frequency of moods or behaviors inthe last month. Correlations greater than .09 are significant at p < .01. M = men; W = women.


unpleasant thoughts that will not leave one's mind) and compul-sive behaviors (e.g., checking and indecisiveness); (d) self-con-sciousness and interpersonal anxiety by the 9-item InterpersonalSensitivity scale reflecting interpersonal uneasiness and anxiety(e.g., easily hurt feelings, feeling shy with the opposite sex, andfeeling inferior); (e) depression by the 13-item Depression scaleassessing a broad range of dimensions reflecting depression (e.g.,feeling blue and hopeless about the future); (f) general anxiety bythe 10-item Anxiety scale tapping anxiety not related to specificsituations (e.g., periods of panic, nervousness, and suddenly scaredfor no reason); (g) phobic anxiety by the 7-item Phobic Anxietyscale reflecting persistent fear of and desires to avoid specificsituations (e.g., open spaces, crowds, and specific things, places, oractivities); (h) paranoid thinking and behavior by the 6-item Para-noid Ideation scale representing paranoid thinking and behavior(e.g., feeling others cannot be trusted and feelings of beingwatched or talked about by others); and (i) psychotic or thoughtdisordered tendencies by the 10-item Psychoticism scale reflectingsymptoms ranging from marked loneliness (e.g., feeling lonelyeven when with others) to delusions (e.g., thought control) andhallucinations (e.g., hearing voices). SCL-90-R items are rated ona 5-point (0 = not at all, 1 = a little bit, 2 = moderately, 3 = quitea bit, 4 = extremely) scale of intensity for each item over the past7 days. The SCL-90-R yields an intensity measure for each scale(i.e., the sum of the intensity ratings within that scale) and a TotalIntensity index (i.e., the sum of all the symptom intensities). Alphareliabilities for SCL-90-R scales range from .77 to .90, and 1-weektest-retest reliabilities from .78 to .90 (Derogatis, 1983). Substan-tial validity evidence is summarized in Derogatis (1983).

Procedure

In the fifth week of class, interested students picked up the TASand SCL-90-R, which were completed at home and returned overthe next three class periods. TAS and SCL-90-R data were anon-ymous, except for age and gender of student, as name and studentnumber for extra credit were turned in separately.


Because tests for differences in independent correlationsrevealed no gender differences among correlations, datawere collapsed across gender (Table 11). Trait anger corre-lated significantly with intensity of angry, hostile feelings,and hostile behaviors in the past week, accounting for 30%of the variance. Correlations of trait anger with other symp-tom intensities were significant, but smaller, accounting for3% to 17% of the variance and for 18% of the variance ofTotal Intensity. A series of tests for differences betweendependent correlations (Table 11), which partially correctfor the dependence among SCL-90-R scales, revealed thatthe correlation of the TAS with anger-hostility intensityand frequency was significantly stronger than the correla-tion of the TAS with all other respective symptom intensi-ties. Although trait anger correlated with nearly all psycho-logical SCL-90-R states, the correlations were higher withstate anger-hostility intensity than with the intensities ofother emotional and psychological states as well as theindex of general symptom intensity, again providing sup-port for state-trait anger theory and evidence of both con-

Table 11Correlations of Trait Anger Scale (TAS) With SCL-90-RSymptom Intensity in Study 7

Symptom

HostilitySomatizationObsessive-compulsiveInterpersonal sensitivityDepressionAnxietyPhobic anxietyParanoid ideationPsychoticismTotal symptom intensity

r

.55

.17

.38

.40

.30

.32

.29

.41

.30

.42

f(175)a

5.94*3.60*3.04*4.26*4.14*4.41*2.99*5.26*3.16*

Note. Correlations greater than .14 are significant at p < .05.a Test for difference in strengths of dependent correlations com-paring the strength of the TAS and Hostility intensity correlationwith the correlation of the TAS and the Symptom Checklist-90-Revised (SCL-90-R) symptom intensity.* p < .05 (alpha level has been adjusted for experimentwise error).

vergent and discriminant validity for the TAS as a measureof trait anger.

Study 8

Study 8 extended the logic of Studies 6 and 7 to angerexpression. The discrimination hypothesis also suggests thattrait anger should correlate with ways of expressing angerand that correlations with anger expression should begreater than the correlation of anger expression with othergeneral emotions such as anxiety and depression. In earlierstudies (e.g., Spielberger, 1988; Spielberger et al., 1985),trait anger was found to correlate positively with angersuppression (AX-I) and with outward negative expression(AX-0) and negatively with a controlled anger expressionstyle (AX-C). Also, AX-I was unrelated to AX-0 andAX-C, which were inversely related to each other. More-over, trait anger was more strongly related to AX-0 andAX-C than was trait anxiety, but trait anxiety was morestrongly related to AX-I than was trait anger (Deffenbacher,1992). The present study furthers understanding of theserelationships by simultaneously assessing trait anger, angerexpression styles, trait anxiety, and depression. Inclusion oftrait anxiety and depression allowed for a more sensitiveexploration of state-trait theory in regard to anger expres-sion, as theory logically extends to predict that correlationsbetween trait anger and anger expression styles will bestronger than those between trait anxiety and anger expres-sion and between depression and anger expression.

Method

Participants

Participants were 233 introductory psychology students (107men and 126 women) who received one research credit.


Instruments

Trait anger, anger expression styles, trait anxiety, and depres-sion were assessed by the TAS, AX, TAI, and the Beck DepressionInventory (BDI; Beck, Rush, Shaw, & Emery, 1979), respectively(see Study 1 for descriptions of the TAS and TAI). The AX(Spielberger, 1988) includes 8-item measures of anger suppressionor anger held in (AX-I), outward negative expression of anger(AX-0), and calm, controlled anger expression (AX-C). AX scaleshave adequate alpha reliability and evidence of discriminant andconvergent validity, which is reflected in different patterns ofcorrelations with emotional, physiological, and other variables(Deffenbacher, 1992; Spielberger, 1988). The BDI is a 21-item,self-report measure of severity of depression over the past week;symptoms are rated on scales of 0-3 resulting in scores rangingfrom 0 to 63, with higher scores reflective of greater depression.The BDI has alpha reliability of .86 (Beck & Steer, 1984) and isone of the most widely validated measures of depression.

Procedure

At a table that described various experiments available, intro-ductory psychology students read a description of a questionnaire-type study and signed up to participate. Groups of approximately60 were assessed in a large university classroom. Following in-formed consent procedures, students completed, in order, the TAI,BDI, TAS, and AX. The measures of anxiety and depression werecompleted prior to measures of anger so that responses to thesemeasures would not be contaminated by any participant hypothe-ses about the nature of the study involving the relationship of angerto these variables. When questionnaires were complete, partici-pants were debriefed and given experimental credits.


Because no multivariate gender effects were found in aone-way (Gender) MANOVA, and because tests for differ-ences between independent correlations revealed no genderdifferences in correlation matrices, data were collapsedacross gender (Table 12). Before exploring the primaryhypotheses, intercorrelations of measures were assessed. Asfound in prior research (Deffenbacher, 1992; Spielberger,1988), AX-I did not correlate with either AX-0 or AX-C,which correlated negatively with each other. Also, anger,anxiety, and depression correlated positively with eachother, although tests for differences in dependent correla-tions showed that anxiety and depression were more highlyrelated to each other than either anxiety or depression were

Table 12Correlations Among Trait Anger, Trait Anxiety,Depression, and Anger Expression Styles in Study 8

Measure 2 3 4 5 (

1. Trait Anger Scale2. Trait Anxiety Inventory3. Beck Depression Inventory4. Anger Expression-In5. Anger Expression-Out6. Anger Expression-Control

.42 .33 .28 .70 - .58— .75 .50 .34 - .41

— .37 .29 - .31— .07 - .07

— - .60

Note. Correlations greater than |.17| are significant at p < .01.

to anger, tt(230) = 6.80 and 8.95, respectively, ps < .001.With regard to the primary hypotheses, the TAS correlatedsignificantly with all anger expression variables; however,so did the TAI and BDI. Tests for differences in dependentcorrelations were run to test the predictions that the TASwould form stronger relationships with anger expressionstyles than would the TAI or BDI. In these analyses, abso-lute values of correlations were employed in order to re-move measurement artifacts that could lead to artificialdifferences. The TAS formed significantly stronger rela-tionships with AX-0 and AX-C than did the TAI, fs(230) =7.12 and 3.01, respectively, ps < .001, or the BDI,?s(230) = 6.18 and 4.39, respectively, ps < .001. This pat-tern, however, did not hold for the AX-I. In fact, the TAIcorrelated significantly stronger with the AX-I than did theTAS, f(230) = -3.59, p < .001, and there was no signifi-cant difference in the strengths of correlation of the TASand BDI with AX-I, rs(230) = 1.29. Moreover, the TASwas more strongly correlated with AX-0 and AX-C thanwith AX-I, fs(230) = 6.91 and 4.15, respectively, ps <.001.

In summary, state-trait anger theory received further sup-port, as trait anger was related to all anger expression styles(convergent validity), and the correlations of trait anger withtwo of the expression styles were stronger than their corre-lations with trait anxiety and depression (discriminant va-lidity). However, the evidence for discriminant validitylinking trait anger with anger suppression is weak, as angersuppression had more to do with anxiety than anger per se.The overall pattern of results suggests that trait anger, asassessed by the TAS, is characterized more by expression ofanger in outward, negative, and less controlled ways and toa much lesser extent by anger suppression.

Integrative Discussion

State-trait anger theory (Spielberger, 1988; Spielberger etal., 1983, 1988) proposes trait anger as a fundamental indi-vidual difference in the propensity to become angry. Thisleads to the two general predictions that high trait angerindividuals tend to become angry more frequently and moreintensely than low trait anger individuals (elicitation andintensity hypotheses) and to three corollary predictions thathigh-anger individuals express their anger in more maladap-tive ways and experience more negative consequences fromtheir anger (negative expression and consequence hypothe-ses) and that as a unique personal characteristic or person-ality factor, trait anger should predict anger phenomenamore powerfully than other emotional and behavioral phe-nomena (discrimination hypothesis).

Findings from the current research supported all five ofthese hypotheses. High-anger individuals reported that awider range of potential provocations angered them, as wellas reporting more frequent anger on a weekly and monthlybasis, all suggesting that more things were angering them.They reported more intense anger across a wide range ofsituations, in their most angering, ongoing situations, in dayto day activities, in angry and hostile feelings and behaviors


over the last week, and in the level of reported anger-relatedphysiological arousal. Moreover, anger intensity findingsdid not appear confounded by different types of provoca-tions, but reflected higher intensity levels to similar provo-cations. High-anger individuals also reported more intensestate anger in response to experimental provocations. Thus,the two primary predictions were supported as it was con-sistently shown that high-anger individuals were more fre-quently and intensely angered.

State-trait theory was also extended successfully tomodes of expressing anger and to the outcomes or conse-quences of anger expression (negative expression and con-sequence hypotheses). High-anger persons experiencedmore dysfunctional coping when provoked, reported usingboth outwardly negative and inwardly suppressive expres-sion styles, and reported more frequent and, in some cases,severe anger consequences.

Finally, the discrimination hypothesis received supportfrom the patterns and strengths of correlations. Trait angerwas more strongly associated with the frequency and inten-sity of anger and anger-related thoughts and behaviors thanit was with general indexes of psychological distress, withother specific emotions such as anxiety and depression,with psychological states such as paranoid thinking, or withbehavioral acts such as becoming intoxicated. Moreover,trait anger was highly related to expression of anger in anoutwardly negative way and had a strong negative relation-ship to controlled styles of expression, whereas anxiety anddepression showed much weaker relationships with thesemodes of anger expression. Although there was some evi-dence for general negative affectivity (Watson & Clark,1984) because trait anger was correlated with other emo-tional, cognitive, and behavioral conditions, trait anger wassignificantly more strongly related to angry feelings, angry-hostile symptoms, and outward negative expression of an-ger than to other emotions, behaviors, symptoms, or angersuppression. An underlying construct of anger pronenessthus appears more explanatory than the construct of generalnegative affectivity.

Overall, there was considerable convergent and discrimi-nant validity for the concept of trait anger, and the TASappears to be a good measure of this tendency. Data fromseveral different measures and methodologies converged tosupport trait anger as a relatively unique emotional dimen-sion of personality. In fact, across studies there were onlytwo nonconfirmatory findings: (a) high- and low-anger par-ticipants did not differ on heart rate immediately after theexperimental provocation in Studies 1 and 2, and (b) traitanxiety was a better predictor of anger suppression than traitanger. Perhaps, heart rate did not change because it takesgreater provocation than the studies provided (i.e., thatproduced by naturalistic provocations and interpersonalchallenges) to lead to reliable physiological differences, assuggested by Smith (1992) in his review of studies ofhostility and physiological arousal. A portion of the expla-nation may also lie in the use of a single physiologicalresponse. Even though pulse-related responses were a fre-quently mentioned anger symptom (Table 5), they werelisted by less than 20% of participants. Heart rate may not

be the most sensitive physiological anger response for ev-eryone, and assessment of multiple response systems maybe required to identify the individual's most responsivesystem and confirm the theoretically predicted physiologi-cal changes for high-anger participants, at least in situationsnot eliciting strong physiological arousal.

The fact that anger suppression was more strongly linkedto trait anxiety than trait anger suggests that anger suppres-sion might be better conceptualized in frameworks involv-ing other emotional states and traits, especially those in-volving anxiety. Throughout these studies, direct andindirect evidence suggested that trait anger was related moreto outward negative, less controlled expression styles. Forexample, although trait anger was correlated with anger-in,it formed larger correlations with anger-out and anger-control styles. Also, in response to experimental provoca-tion, high-anger individuals reported more verbal and phys-ical antagonism and lower constructive coping, none ofwhich are reflective of suppression or inhibition of expres-sion. More frequent and severe consequences experiencedby high-anger individuals also suggest less controlled, morenegative expression. Moreover, trait anger was morestrongly correlated with the anger-hostility scale on theSCL-90-R, which includes hostile thoughts and actionitems, than with general or phobic anxiety scales morereflective of anxiety and inhibition. Thus, trait anger ap-pears to be more strongly associated with less controlled,outwardly expressive styles than with anger suppression.Research in areas such as health psychology (e.g., Diamond,1982) and depression (e.g., Biaggio & Goodwin, 1987;Riley, Treiber, & Woods, 1989), which have noted theimportance of anger suppression and anxiety, should takenote of the greater relationship of anger suppression toanxiety than to other anger variables and include both angerexpression and anxiety measures so that results are notattributed to conceptually different processes when, in fact,they are empirically linked.

Gender did not appear to mediate anger relationships, asthere were relatively few gender differences in the eightstudies. Even where gender differences were found, theywere small and did not tend to replicate. In general, resultssuggest that, within the limits of methodologies employed,men and women are angered by similar things and to similardegrees, express themselves in similar ways, and suffersimilar consequences. This conclusion is not meant to implythat men and women do not differ, for they may. Forexample, there was suggestive evidence that high-angermen may somewhat more frequently suffer negative conse-quences resulting from physically assaultive behavior.However, even here, results were not consistent, for whenthe most serious anger incidents were sampled, women wereas likely as men to suffer negative consequences. Thus, thestudies in this article and other recent reviews (e.g., Sharkin,1993) do not suggest that there are large gender differencesconsistent with sex role stereotypes of anger as a maleemotion. One potential explanation for these findings is thatgender role may be a better predictor of anger (Kopper,1993; Kopper & Epperson, 1991) than gender per se. How-ever, should this prove to be the case, the variance ac-


counted for may still be relatively small because gender andgender role are significantly correlated (Bern, 1981).

Findings from these studies have several implications forcounseling practice and research. First, the TAS appears tohave good clinical utility as a measure of general anger.When combined with a statement of personal problems withanger and a desire for help with those problems, the TASdiscriminated a meaningful group of angry individuals.They report elevated anger on a large number of indexesand in considerable portions of their lives, and they handledtheir anger less well and suffered significant consequencesbecause of their anger. Moreover, although the TAS corre-lated with other negative psychological conditions, therewas good evidence that it was not seriously confounded bythese issues and that it primarily assessed anger. Thus, theTAS may prove useful as a brief screening device forgeneral anger and be used as an outcome measure for theindividual case or treatment studies involving general angerreduction.

Second, high trait anger individuals experience both ele-vated angry emotions and difficulties in positive, prosocialexpression of that anger. That is, they appear marked byboth heightened emotionality and skill deficits in socialproblem-solving, assertion, and communication skills.Heightened emotionality might be targeted directly withrelaxation interventions (e.g., Deffenbacher, Demm, &Brandon, 1986; Deffenbacher & Stark, 1992; Hazaleus &Deffenbacher, 1986) or cognitive interventions (e.g., Def-fenbacher et al., 1988; Hazaleus & Deffenbacher, 1986;Moon & Eisler, 1983; Novaco, 1975) to reduce anger-engendering cognitions or a combination of cognitive andrelaxation interventions (e.g., Deffenbacher, Story, Stark,Hogg, & Brandon, 1987; Deffenbacher et al., 1988, 1994;Deffenbacher & Stark, 1992; Novaco, 1975). Skill deficitsin interpersonal communication and conflict resolutionmight be addressed by social problem-solving interventions(e.g., Moon & Eisler, 1983) or by social skills training (e.g.,Deffenbacher et al., 1987, 1994; Moon & Eisler, 1983),which enhances the individual's capacity to generate andexecute calm, problem-focused negotiation, assertion, andexpression.

Third, even though there was a pattern of convergingcharacteristics of high-trait anger individuals, there was alsoconsiderable variability in the situations that triggered theiranger, the patterns of expression, and the consequencessuffered. This suggests that when dealing with an individualclient, the counselor and client should carefully map thesedimensions to identify areas of greatest vulnerability and tomake sure that therapeutic insights and interventions areapplied successfully to these areas. That is, effective angerreduction interventions will still benefit from a careful as-sessment and focusing of intervention on the specific trig-gers and vulnerabilities of the individual. A related inter-vention issue is that of personal strengths. Even thoughdeficits and difficulties were strongly suggested by thepresent research, high-anger individuals also reported con-trolled expression on some occasions. Control strategies andpositive examples of dealing with anger should also beexplored to identify personal strengths and assets that may

be adapted to other elements of the individual's angerproblems.

Fourth, although anger appeared to stand alone as anidentifiable set of concerns, it too was associated with othernegative emotions and issues such as anxiety and depres-sion. This suggests that these issues might benefit fromattention, too. Therapeutic strategies developed for angermight be adapted to other issues, especially in the latterstages of therapy when anger issues have been addressed toa great extent. For example, relaxation interventions mightbe adapted to other arousal states such as anxiety, guilt, andshame, or cognitive interventions might be focused on cog-nitive themes and schema contributing to anxiety or depres-sion. Social problem-solving and social skill training inter-ventions could be adapted to issues of social skill deficitsinvolved in interpersonal anxiety and depression.

Finally, the relative lack of gender differences suggeststhat men and women are similar enough that, if a groupintervention is employed, men and women can be combinedefficiently in a group. In fact, there may be some addedbenefits such as balanced discussion of anger in partnered orwork relationships, the renorming of expectations of angerin the opposite sex, and the availability of both genders forrole plays and other interventions.

References

Averill, J. R. (1983). Studies on anger and aggression: Implica-tions for theories of emotion. American Psychologist, 38, 1145-1160.

Beck, A. T., Rush, A. J., Shaw, B. F., & Emery, G. (1979). Cog-nitive therapy of depression. New York: Guilford.

Beck, A. T., & Steer, R. A. (1984). Internal consistencies of theoriginal and revised Beck Depression Inventory. Journal ofClinical Psychology, 40, 1365-1367.

Bern, S. L. (1981). Gender schema theory: A cognitive account ofsex typing. Psychological Review, 88, 354-364.

Biaggio, M. K., & Goodwin, W. H. (1987). Relation of depressionto anger and hostility constructs. Psychological Reports, 61,87-90.

Dean, R. B., & Winters, C. (1991). TRYSYS: User's manual.Chicago: d2 Associates.

Deffenbacher, J. L. (1992). Trait anger: Theory, findings, andimplications. In C. D. Spielberger & J. N. Butcher (Eds.), Ad-vances in personality assessment. (Vol. 9, pp. 177-201). Hills-dale, NJ: Erlbaum.

Deffenbacher, J. L., Demm, P. M., & Brandon, A. D. (1986). Highgeneral anger: Correlates and treatment. Behaviour Researchand Therapy, 24, 480-489.

Deffenbacher, J. L., Eiswerth, L. E., & Stark, R. S. (1986, April).Situations and physiological symptoms in general anger. Paperpresented at the 56th Annual Convention of the Rocky MountainPsychological Association, Denver, Colorado.

Deffenbacher, J. L., & Sabadell, P. M. (1992). Leicht argerliche(high trait anger) personen und nur schwer zu argernde (lowanger) personen: Ein vergleich. [Comparing high trait anger withlow trait anger individuals]. In M. Muller (Ed.), Psychophysi-ologische risikofaktoren bei herz-kreislauferkrakungen: Grund-lagen und therapie (pp. 153-169). Gottingen, Germany: Verlag.

Deffenbacher, J. L., & Stark, R. S. (1992). Relaxation and cogni-tive-relaxation treatments of general anger. Journal of Counsel-ing Psychology, 39, 158-167.


Deffenbacher, J. L., Story, D. A., Brandon, A. D., Hogg, J. A., &Hazaleus, S. L. (1988). Cognitive and cognitive-relaxation treat-ments of anger. Cognitive Therapy and Research, 12, 167-184.

Deffenbacher, J. L., Story, D. A., Stark, R. S., Hogg, J. A., &Brandon, A. D. (1987). Cognitive-relaxation and social skillsinterventions in the treatment of general anger. Journal of Coun-seling Psychology, 34, 171-176.

Deffenbacher, J. L., Thwaites, G. A., Wallace, T. L., & Oetting,E. R. (1994). Social skills and cognitive-relaxation approachesto general anger reduction. Journal of Counseling Psychology,41, 386-396.

Derogatis, L. R. (1983). SCL-90-R administration, scoring & pro-cedures manual—//. Towson, MD: Clinical PsychometricResearch.

Desnoes, M., & Deffenbacher, J. L. (1995, April). Evaluation ofstate-trait anger theory. Paper presented at the 65th AnnualConvention of the Rocky Mountain Psychological Association,Boulder, Colorado.

Diamond, E. (1982). The role of anger and hostility in essentialhypertension and CHD. Psychological Bulletin, 92, 410-433.

Hazaleus, S. L., & Deffenbacher, J. L. (1986). Relaxation andcognitive treatments of anger. Journal of Consulting and Clin-ical Psychology, 54, 222-226.

Kassinove, H., & Sukhodolsky, D. (1995, August). Psychometricand constructivist anger measurement in Russia and America.Paper presented at the 103rd Annual Convention of the Amer-ican Psychological Association, New York, NY.

Kopper, B. A. (1993). Role of gender, sex role identity, and TypeA behavior in anger expression and mental health functioning.Journal of Counseling Psychology, 40, 232-237.

Kopper, B. A., & Epperson, D. L. (1991). Women and anger: Sexand sex role comparisons in the expression of anger. Psychologyof Women Quarterly, 15, 7-14.

Lazarus, R. S. (1991). Emotion and adaptation. New York: OxfordUniversity Press.

Leibsohn, M.T., Oetting, E. R., & Deffenbacher, J. L. (1994).Effects of trait anger on alcohol consumption and consequences.Journal of Child and Adolescent Substance Abuse, 3, 17-32.

Leonard, K. E., & Blane, J. T. (1992). Alcohol and marital aggres-sion in a national sample of young men. Journal of InterpersonalViolence, 7, 19-30.

Lopez, F. G., & Thurman, C. W. (1986). A cognitive-behavioralinvestigation of anger among college students. Cognitive Ther-apy and Research, 10, 245-256.

Maiuro, R. D., Cahn, T. S., Vitaliano, P. P., Wagner, B. C , &Zegree, J. B. (1988). Anger, hostility, and depression in domes-tically violent versus generally assaultive men and nonviolentcontrol subjects. Journal of Consulting and Clinical Psychology,56, 17-23.

Moon, J. R., & Eisler, R. M. (1983). Anger control: An experi-mental comparison of three behavioral treatments. BehaviorTherapy, 7, 493-505.

Novaco, R. (1975). Anger control: The development and evalua-tion of an experimental treatment. Lexington, MA: D. C. Heath.

Pan, H. S., Neidig, P. H., & O'Leary, K. H. (1994). Predictingmild and severe husband-to-wife physical aggression. Journal ofConsulting and Clinical Psychology, 62, 975-981.

Riley, W. T., Treiber, F. A., & Woods, M. G. (1989). Anger andhostility in depression. Journal of Nervous and Mental Disease,177, 668-674.

Schwartz, G., Weinberger, D., & Singer, J. (1981). Cardiovasculardifferentiation of happiness, sadness, anger, and fear followingimagery and exercise. Psychosomatic Medicine, 43, 343-364.

Sharkin, B. S. (1993). Anger and gender: Theory, research, andimplications. Journal of Counseling and Development, 71, 386-389.

Siegman, A. W., & Smith, T. W. (1994). Anger, hostility, and theheart. Hillsdale, NJ: Erlbaum.

Smith, T. W. (1992). Hostility and health: Current status of apsychosomatic hypothesis. Health Psychology, 11, 139-150.

Southard, L. D., & Katahn, M. (1967). The correlation betweenself-reported and mechanically-recorded pulse rates. Psy-chonomic Science, 8, 343-344.

Spielberger, C. D. (1966). Theory and research on anxiety. InC. D. Spielberger (Ed.), Anxiety and behavior (pp. 3-20). NewYork: Academic Press.

Spielberger, C. D. (1972). Anxiety as an emotional state. In C. D.Spielberger (Ed.), Anxiety: Current trends in theory and re-search. (Vol. 1, pp. 24-49). New York: Academic Press.

Spielberger, C. D. (1988). State-Trait Anger Expression Inven-tory. Orlando, FL: Psychological Assessment Resources.

Spielberger, C. D., Gorsuch, R., & Lushene, R. (1970). Manual forthe State-Trait Anxiety Inventory (Self Evaluation Question-naire). Palo Alto, CA: Consulting Psychologists Press.

Spielberger, C D . , Jacobs, G., Russell, S., & Crane, R. (1983).Assessment of anger: The State-Trait Anger scale. In J. N.Butcher & C. D. Spielberger (Eds.), Advances in personalityassessment. (Vol. 3, pp. 112-134). Hillsdale, NJ: Erlbaum.

Spielberger, C. D., Johnson, E., Russell, S., Crane, R., Jacobs, G.,& Worden, T. (1985). The experience and expression of anger:Construction and validation of an anger expression scale. InM. A. Chesney & R. H. Rosenman (Eds.), Anger and hostility incardiovascular and behavioral disorders (pp. 5-30). New York:McGraw-Hill.

Spielberger, C. D., Krasner, S., & Solomon, E. (1988). The expe-rience, expression, and control of anger. In M. P. Janisse (Ed.),Health psychology: Individual differences and stress (pp. 89-108). New York: Springer Verlag.

Spielberger, C. D., Reheiser, E. C , & Sydeman, S. J. (1995). Mea-suring the experience, expression, and control of anger. In H.Kassinove (Ed.), Anger disorders: Definition, diagnosis, andtreatment (pp. 49-67). Washington, DC: Taylor & Francis.

Tryon, R. C , & Bailey, D. E. (1966). The BC TRY computersystem of cluster and factor analysis. Multivariate BehavioralResearch, 1, 95-111.

Watson, D., & Clark, L. A. (1984). Negative affectivity: Thedisposition to experience aversive emotional states. Psycholog-ical Bulletin, 96, 465-490.

Received June 5, 1995Revision received September 18, 1995

Accepted September 18, 1995 •

state-trait anger theory and the utility of the trait ... · such as the buss-durkee hostility...

Documents