Journal of Consulting Psychology Vol. 21. No. 4.

1957

An Inventory for Assessing Different Kinds of Hostility1

In their everyday functioning, clinical psychologists are alert to the ways in which hostility is expressed, and they are usually careful to distinguish various modes of expression. When the aggression is overt and direct, a distinction is usually made between verbal hostility and physical assault. Overt manifestations are clearly separated from covert manifestations of hostility, e.g., cursing and threatening behavior vs. gossiping and roundabout derogation. Since meaningful distinctions can be made between subclasses of hostility, a global evaluation of hostility would seem to contain considerable ambiguity. The statement "He is hostile" would apply equally well to a man who beats his wife and to a man who is spitefully late for appointments. Thus, it should be expected that attempts to assess hostility would include not only a global estimate of intensity but also estimates of the intensity of the various subhostilities. The writers know of no published hostility inventory that attempts more than a global estimate of hostility. Three of the more recently developed inventories, those of Cook and Medley (2), Moldawsky (1), and Siegel (9), all consist of items selected from the MMPI by clinical psychologists. None of these investigators attempted to group items into subscales representing various aspects of hostility. Thus a nonsuspicious, assaultive individual might receive the same score as a nonassaultive, suspicious individual. A score on one of these inventories would appear to be as ambiguous as the statement "He is hos1 The writers wish to acknowledge the considerable efforts of Dr. Herbert Gerjuoy in obtaining subjects and in facilitating statistical analyses.

tile." What is clearly needed is an inventory that attempts to assess the various aspects of hostility. This paper describes the development of such an inventory. Construction of the Inventory Varieties oj Hostilities The first task was to define the subclasses of hostility that are typically delineated in everyday clinical situations. Such a classification was made in an earlier study (1), and the present classification is an elaboration of the previous one.
Assaultphysical violence against others. This includes getting into fights with others but not destroying objects. Indirect Hostilityboth roundabout and undirected aggression. Roundabout behavior like malicious gossip or practical jokes is indirect in the sense that the hated person is not attacked directly but by devious means. Undirected aggression, such as temper tantrums and slamming doors, consists of a discharge of negative affect against no one in particular; it is a diffuse rage reaction that has no direction. Irritabilitya readiness to explode with negative affect at the slightest provocation. This includes quick temper, grouchiness, exasperation, and rudeness. Negativismoppositional behavior, usually directed against authority. This involves a refusal to cooperate that may vary from passive noncompliance to open rebellion against rules or conventions. Resentmentjealousy and hatred of others. This refers to a feeling of anger at the world over real or fantasied mistreatment. Suspicionprojection of hostility onto others. This varies from merely being distrustful and wary of people to beliefs that others are being derogatory or are planning harm. Verbal Hostilitynegative affect expressed in both the style and content of speech. Style includes arguing, shouting, and screaming; content includes threats, curses, and being overcritical.

343

344

Arnold H. Buss and Ann Durkee

Item-writing Techniques The writers constructed a pool of items and supplemented this pool with items borrowed from previous inventories. Most of the borrowed items underwent modification, and the following principles served as guides in writing and selecting items.
1. The item should refer to only one subclass of hostility, since an item that overlaps several categories would not help in distinguishing patterns of hostility. 2. The behaviors and attitudes involved should be specific, and the stimulus situations that arouse them should be near universal, e.g., "It makes my blood boil to have people make fun of me." "Makes my blood boil" is a fairly specific response, and being ridiculed is a common situation for most people. 3. The item should be worded so as to minimize defensiveness in responding. It has been established that social desirability accounts for much of the variance of normals' responses to inventories (4, 5). In attempting to facilitate respondents' admitting to socially undesirable behaviors, three item-writing techniques were employed: First, assume that the socially undesirable state already exists and ask how it is expressed, e.g., "When I really lose my temper I am capable of slapping someone," "When I get mad, I say nasty things." In these items the loss of temper is assumed, and the subject is asked only whether he expresses it physically. This procedure emphasizes a report of behavior and tends to minimize the value judgments associated with hostility. Second, provide justification for the occurrence of hostile behavior, e.g., "Whoever insults me or my family is asking for a fight," "People who continually pester you are asking for a punch in the nose," "Like most sensitive people, I am easily annoyed by the bad manners of others." When the item provides a rationale for the aggression, the subject's defensive and guilt reactions are reduced, and he does not necessarily answer in the direction of social desirability. Third, use idioms, e.g., "If somebody hits me first, I let him have it," "When I am mad at someone, I will give him the silent treatment." Idioms have a high frequency of usage in everyday life, and these phrases are typically used by subjects to describe their own behavior and feelings to others. It is anticipated that these phrases will merely echo what the subject has previously verbalized, and therefore when such phrases apply, they will be readily accepted and admitted. 4. Take into account the effects of response set by including both true and false items. If all the items were scored in the direction of hostility only when marked "True," a subject could get a low score simply by answering all the items "False." Ideally, this kind of response set is best controlled when the number of true items equals the number of false

On the basis of the foregoing considerations, a pool of hostility items was compiled. Next it was decided to add the variable of guilt because the relationship of guilt to the various subhostilities is of clinical interest. Accordingly, items were compiled for a Guilt scale, with guilt being denned as feelings of being bad, having done wrong, or suffering pangs of conscience. Item Analyses The first version of the inventory consisted of IDS items, with items from each scale randomly scattered throughout the inventory. It was administered in group fashion to 85 male and 74 female college students. In an attempt to reduce defensiveness, all protocols were anonymous. The various hostility scales and the Guilt scale were scored, and separate item analyses were performed for men and women. Two criteria were used in item selection: frequency and internal consistency. Frequency refers to the occurrence of the particular behavior in the population, as measured by the proportion of the sample answering in the direction of hostility (or guilt). If a given behavior is near-universal in the population or virtually absent, it obviously does not distinguish between individuals. A criterion of frequency is necessary to eliminate items that are answered in one direction by virtually everyone, and it was decided to accept only items answered in one direction by 15-85% of the sample. Internal consistency was measured by the correlation of an item with the score of the scale in which it belonged. Since the items are scored dichotomously, the biserial correlation coefficient was used. The criterion for item selection was a correlation of at least .40 for both the male and female samples. Only 60 of the original 105 items met the frequency and internal consistency criteria. The number of items in several of the scales was so low that unreliability (lack of testretest stability) seemed inevitable. Therefore, additional new items were written and old ones modified. Most of the modifications were

Assessing Different attempts to alter the frequency measure, i.e., decrease the popularity of items universally endorsed and increase the popularity of items rarely endorsed. The revised inventory contained 94 items. It was administered in group fashion to 62 male and 58 female college students, and separate item analyses were performed for each sex. Again the minimum item-scale correlation was set at .40, but this time the frequency criterion was modified. The first item analysis had revealed sex differences in the proportion of the sample answering in the direction of hostility (or guilt). For several items the proportion of male students was over 15%, but the proportion of female students was under 15%. Since the 15-85% frequency criterion might eliminate items that differentiated between men and women, a less stringent criterion was adopted: 15-85% for either men or women. In addition, an attempt was made to insure that each scale contained items whose frequencies varied over a wide range. The second item analysis yielded 75 items, 66 for hostility and 9 for guilt. It was found that more False items were discarded than True items, and the final form of the inventory contains 60 True items and 15 False items, a ratio ot tour to one. The items comprising the final form of the inventory are listed in Table 1. Each item is grouped with the other items in its scale, and the False items are marked "F." Social Desirability Responses to inventory items are at least in part determined by the respondent's desire to place himself in a favorable light. This tendency assumes great importance in a hostility inventory, which deals with behaviors that are generally regarded as socially unacceptable. The potency of the tendency to give socially desirable answers has been demonstrated by Edwards (4). He had college students assign each of 140 personality trait items to one of nine intervals of social desirability. Scale values for social desirability were obtained by the method of successive intervals. Then the 140 items were administered to different college students, with standard inventory instructions. The correlation between social desir-

Kinds of Hostility

345

ability and probability of endorsing the items was .87. Subsequent studies with other inventories have confirmed the fact that social desirability is an important uncontrolled variable in many present-day inventories (5, 8). In constructing the present inventory, an attempt was made to minimize the variable of social desirability. In order to test the success of this attempt, Edwards' procedure (4) was followed. The 66 hostility items of the final inventory were scaled for social desirability, using the method of successive intervals. The judges were 85 male and 35 female college students. The men's and women's judgments were quite similar, and they were pooled. Next, the inventories of 62 men and 58 women (who had previously taken the inventory and were different from the judges) were used to determine the probability of endorsement for each of the 66 hostility items. The productmoment r's were .27 for the men and .30 for the women. Both correlations are significantly above zero at the .05 level of confidence, which suggests that the influence of social desirability is having a small but significant effect on the direction of responding. However, these two correlations are considerably lower than the correlation of .87 reported by Edwards (4). In accounting for this discrepancy two differences between his study and the present one should be noted. First, the present items were designed to measure only the hostile components of personality, whereas Edwards' inventory taps a variety of personality components. Since hostile acts are generally regarded as being socially undesirable, the upper end of the social desirability continuum is not represented in the present inventory. The present inventory ranged from extremely undesirable to moderately desirable behaviors; the inventory used by Edwards includes not only extremely undesirable but also extremely desirable behaviors. The restriction of range can be clearly seen when social desirability scale values of the two inventories are compared. The scaling procedures were identical, but the present range of scale values was .23 to 2.38, while Edwards' range was .50 to 4.70.2 A curEdwards' scale values for social desirability and his probability of endorsement values were estimated from his Fig. 2 (4).
2

346

Arnold H, Buss ana

Assault:)' 4. When people are bossy, I take my time just to show them. (36) 4, Cmce in a while I cannot control my urge to harm 5. Occasionally when I am mad at someone I will others. (9) -.give him the "silent treatment." (28) 2F. I can think of no good reason for ever hitting __^_ anyone. (17) / 3. If somebody hits me first, I let him have it. (23) Resentment,'' 4. Whoever insults me or my family is asking for a. 1. ,t-dbn't seem to get what's coming to me. (5) fight. (33) ^2. Other people always seem to get the breaks. (13) 5. People who continually pester you are asking for 3. When I look back on what's happened to me, I a punch in the nose. (41) can't help feeling mildly resentful. (29) 6F. I seldom strike back, even If someone hits me 4. Almost every week I see someone I dislike. (37) first. (1) 5. Although I don't show it, I am sometimes eaten 7. When I really lose my temper, I am capable of up with jealousy. (45) slapping someone. (49) 6F. I don't know any people that I downright hate. 8. I get into fights about as often as the next person. (21) (57) 7. If I let people see the way I feel, I'd be consid9. If I have to resort to physical violence to defend ered a hard person to get along with. (53) my rights, I will. (65) ., ..8 .At times I feel I get a raw deal out of life. (61) JO. I have known people who pushed me so far thai ~"w came to blows. (70) Suspicion:, Indirect: --"' IT" "f know that people tend to talk about me behind my back. (6) 1. I sometimes spread gossip about people I don't 2. I tend to be on my guard with people who are like. (2) somewhat more friendly than I expected. (14) 2F. I never get mad enough to throw things. (10) 3. There are a number of people who seem to dislike 3. When I am mad, I sometimes slam doors. (26) me very much. (22) 4F. I never play practical jokes. (34) 4. There are a number of people who seem to be 5. When I am angry, I sometimes sulk, (18) jealous of me. (30) 6. I sometimes pout when I don't get my own way. 5. I sometimes have the feeling that others are (42) laughing at me. (38) 7F. Since the age of ten, I have never had a temper 6. My motto is "Never trust strangers." (46) tantrum. (50) 7. I commonly wonder what hidden reason another 8. I can remember being so angry that I picked up person may have for doing something nice for the nearest thing and broke it. (58) me. (54) 9. I sometimes show my anger by banging on the 8. I used to think that most people told the truth """table. (75) but now I know otherwise. (62) Irritability: 9F. I have no enemies who really wish to harm me. (67) 10F. I seldom feel that people are trying to anger or ^ f. r lose my temper easily but get over it quickly. (4) insult me. (72) 2F. I am always patient with others. (27) 3. I am irritated a great deal more than people are , aware of. (20) { Verbal: ; 4. It makes my blood boil to have somebody make\ ten I disapprove of my friends' behavior, I let fun of me. (35) them know it. (7) 5F. If someone doesn't treat me right, I don't let it 2. I often find myself disagreeing with people. (15) annoy me. (66) 3. I can't help getting into arguments when people 6. Sometimes people bother me just by being around. disagree with me. (23) (12) 4. I demand that people respect my rights. (31) 7. I often feel like a powder keg ready to explode. (44) 5F. Even when my anger is aroused, I don't use 8. I sometimes carry a chip on my shoulder. (52) "strong language." (39) 9. I can't help being a little rude to people I don't 6. If somebody annoys me, I am apt to tell him like. (60) what I think of him. (43) 10F. I don't let a lot of unimportant things irritate 7. When people yell at me, I yell back. (47) me. (71) 8. When I get mad, I say nasty things. (51) ~lli-_.Lately, I have been kind of grouchy. (73) 9F. I could not put someone in his place, even if he needed it. (55) Negativism) 10. I often make threats I don't really mean to carry out. (59) T; Unless somebody asks me in a nice way, I won't 11. When arguing, I tend to raise my voice. (68) do what they want. (3) 12F. I generally cover up my poor opinion of others. 2. When someone makes a rule I don't like I am tempted to break it. (12) (63) 13F. I would rather concede a point than get into an 3. When someone is bossy, I do the opposite of what he asks. (19) argument about it. (74)
* The numbers In parentheses indicate the sequence of Items In the mimeographed form of the Inventory.

Assessing Different Table 1Continued

Kinds of Hostility

347

4. It depresses me that I did not do more for my parents. (32) 5. I am concerned about being forgiven for my sins, (40) 6. I do many things that make me feel remorseful afterward. (48) 7. Failure gives me a feeling of remorse. (56) 8. When I do wrong, my conscience punishes me severely. (64) 9. I often feel that I have not lived the right kind of life. (69)

tailed distribution decreases the magnitude of a correlation coefficient, but it is possible to adjust for a difference in standard deviations (7, pp. 149-150). When Edwards' correlation of .87 between social desirability and probability of endorsement is adjusted to the present range of values, it becomes .74. There is still a large disparity between Edwards' corrected correlation of .74 and the present ones of .27 and .30, and the curtailment of the range of social desirability evidently accounts for only a small part of the discrepancy. The second difference between the studies lies in the construction of the present inventory. The writers were aware that social desirability might influence inventory responses, and attempted to minimize its effect by: (a) assuming that anger was present and inquiring only how it is expressed; (b) providing justification for admitting aggressive acts; and (c) including cliches and idioms that

would find ready acceptance. On the other hand, Edwards used a list of unelaborated personality trait names, and there was no attempt to manipulate the wording of the items. Thus, the present low correlations between social desirability and probability of endorsement would seem to reflect the success of the item construction techniques used in the present study. Previous attempts at controlling social desirability have taken two forms. The first is to develop suppressor variables like the "validity" scales of the MMPI (6). The second approach is to scale items for social desirability and then use a paired comparisons type of inventory, in which each item is paired with another item of matched social desirability (3). The present study suggests a third approach, that of focusing on the process of item construction. Perhaps the influence of social desirability can be substantially reduced or eliminated at the source, i.e., in the actual wording of the item. Factor Analyses The final form of the inventory was administered in group fashion to 85 male and 88 female college students. The eight scales were scored, and product-moment correlations were computed for men and women separately. The correlation matrices are presented in Tables 2 and 3. None of the women's correlations, and only two of the men's correlations, are above .50, which suggests that the various scales are tapping at least partially independent behaviors. Thurstone's centroid method (10) was used to extract two factors from each intercorrelation matrix. The axes

Table 2 Table of Intel-correlations for Men (N 85) Indirect Hostility Verbal Hostility

Variable Indirect Ho Irritability Negativism Resentment Suspicion Verbal Ho Guilt

Assault

Irritability

Negativism Resentment

Suspicion

.28 .32 .30 .16 .11 .40 -.03

.44 .27 .33 .27 .40 .28

.20 .44 .26 .66 .24

.31 .38 .25 .08

.58 .37 .27

.21 .25

.16

348

Arnold H. Buss and Ann Durkee

Table 3
Table of Intercorrelations for Women (N 88) Indirect Hostility Verbal Hostility

Variable Indirect Ho Irritability Negativism Resentment Suspicion Verbal Ho Guilt

Assault

Irritability

Negativism Resentment

Suspicion

.38 .30 .27 .14 .11 .37 -.07

.31 .34 .23 .19 .19 .05

.29 .30 .30 .44 .16

.23 .15 .30 .01

.45 .22 .33

.21 .27

.10

for men and women were rotated to the same simple structure so that the factor loadings of the two sexes would be comparable. These factor loadings are presented in Table 4. If only factor loadings of .40 and over are considered meaningful, the first factor is defined by Resentment and Suspicion for men, and by Resentment, Suspicion, and Guilt for women. The second factor is defined by Assault, Indirect Hostility, Irritability, and Verbal Hostility for both sexes, with the addition of Negativism for women. However, both Guilt and Negativism had positive loadings on their respective factors for the men, also, and the sex differences just noted are slight. In fact, the men's and women's factor loadings are generally similar, differences being small and random. Since the same axes were used for men and women, this similarity of factor loadings suggests that the factor structure is stable. The two factors extracted from the intercorrelation matrix divide hostility into an
Table 4 Rotated Factor Loadings for Men and Women

Men
Variable Assault Indirect Hostility Irritability Negativism Resentment Suspicion Verbal Hostility Guilt

Women
If I .19 .00 .14 -.03 II

II

.17 <f.54>> .27 .19 740 .37 .11 C7S7'' .60 .23 .22 .25

L66,)-.02 1 .60 (V63 ..64 .14 .29 3

v -12 .55

f.ST^
154/

.04 .50

,61 .38 .48 .38 .47 .44 .48 .34 .04 .45 .02 .45 >A9,' .44 .28 .33

"emotional" or attitudinal component ("People are no damn good") and a "motor" component that involves various aggressive behaviors. However, it should be noted that the factor loadings are not high. The average communality of the eight variables was .43 for men and .40 for women, leaving considerably more than half of the test variance unexplained. Some of this specific variance may be attributed to unreliability of the scales (especially since they are short), but there seems to be much variance that is stable and unique. The presence of unique variance is not surprising, since it seems likely that there are more than two components of hostility. For example, the second factor includes both Assault and Verbal Hostility, yet there are obviously many verbally hostile individuals who are not assaultive. Similarly, with respect to the first factor, resentment may be seen in the absence of distrust and suspicion. The presence of unique variance would seem to reflect the presence of these patterns within each factor. The population used in deriving the two factors was normal, but the factors appear to have relevance for clinical populations. For example, the characteristics associated with paranoid personalities suggest that such individuals would score high on Resentment and Suspicion (Factor I) and low on the other scales. On the other hand, hysterical personalities should score low on Resentment and Suspicion and high on Irritability, Negativism, and Verbal Hostility. In both instances, no prediction can be made concerning Assault, since this variable is thought to be related to

Assessing Different Table 5

Kinds of Hostility

349

Variable

Mean SD

N,, Items
10 9 11 5 8 10 13 9 66

Women Mean SD
3.27 5.17 6.14 2.30 1.78 2.26 6.82 4.41 27.74 2.31 1.96 2.78 1.20 1.62 1.81 2.59 2.31 8.75

Assault 5.07 2.48 Indirect Hostility 4.47 2.23 S.94 2.65 Irritability Negativism 2.19 1.34 Resentment 2.26 1.89 Suspicion 3.33 2.07 Verbal Hostility 7.61 2.74 Guilt 5.34 1.88 Total Hostility 30.87 10.24

The hostility items were scaled for social desirability, and social desirability was correlated with probability of endorsement. The r's of .27 and .30 for college men and women, respectively, were considerably smaller than those of previous studies. The reduction in the effects of social desirability was attributed to item-writing techniques. Factor analyses of college men's and women's inventories revealed two factors: an attitudinal component of hostility (Resentment and Suspicion) and a "motor" component (Assault, Indirect Hostility, Irritability, and Verbal Hostility). The relevance of these factors to the study of abnormal as well as normal personalities was illustrated.
Received October 29, 1956.

the variables of sex, socioeconomic status, psychopathology, etc. Norms The collection of normative data for a new instrument is a long-time endeavor. In the present instance the process has just begun. Norms are being collected for clinical populations, and the construct validity of the inventory is being investigated. At present, the only norms available are for the 85 college men and 88 college women who were administered the final form of the inventory. The means and standard deviations of these two groups are presented in Table 5. Since these samples are small and not representative, the norms must be regarded as highly tentative. Summary This paper described the construction of an inventory consisting of the following scales: Assault, Indirect Hostility, Irritability, Negativism, Resentment, Suspicion, Verbal Hostility, and Guilt. The first and second versions of the scale were item analyzed, and the final revision consists of 75 items.

References
1. Buss, A. H., Durkee, Ann, & Baer, M. B, The measurement of hostility in clinical situations. /. abnorm. soc. Psychol., 1956, 52, 84-86. 2. Cook, W. W., & Medley, D. M. Proposed hostility and pharisaic-virtue scales for the MMPI. J. appl. Psyckol., 1954, 38, 414-418. 3. Edwards, A. L. Manual for the Edwards Personal Preference Schedule. New York: Psychological Corp., 1954. 4. Edwards, A. L. The relationship between the judged desirability of a trait and the probability that the trait will be endorsed. /. appl. Psychol., 1953, 37, 90-93. 5. Fordyce, W. E. Social desirability in the MMPI. /. consult. Psychol., 1956, 20, 171-175. 6. Hathaway, S. R., & McKinley, J. C. Minnesota Multiphasic Personality Inventory: Manual. (Rev. ed.) New York; Psychological Corp., 1951. 7. McNemar, Q. Psychological statistics. (2nd ed.) New York: Wiley, 1955. 8. Rosen, E. Self-appraisal, personal desirability, and perceived social desirability of personality traits. /. abnorm. soc. Psychol., 1956, 52, 151158. 9. Siegel, S. M. The relationship of hostility to authoritarianism. /. abnorm. soc. Psychol., 1956, 52, 368-372. 10. Thurstone, L. L. Multiple-factor analysis. Chicago: Univer. of Chicago Press, 1947.