Psychosomatic Medicine. 1989. V. 51. P.46-57.

The Cook-Medley Hostility Scale: Item Content and Ability to Predict Survival

J.C. Barefoot, K.A. Dodge, B.L. Peterson, W.G. Dahlstrom, R.B. Williams

Previous studies have identified the MMPI-based Cook and Medley hostility scale () as predictor of health outcomes. To achieve better understanding of the construct measured by this scale, items were classified on an priori basis. Six subsets were identified: Cynicism, Hostile Attributions, Hostile Affect, Aggressive Responding, Social Avoidance, and Other. Study 1 examined the correlations of these subsets with scales of the NEO Personality Inventory in two samples of undergraduates. Good convergent and discriminant validity were demonstrated, but there was some evidence that items in the Social Avoidance and other categories reflect constructs other than hostility. Study 2 examined the ability of the scale and the item subsets to predict the 1985 survival of 118 lawyers who had completed the I in 1956 and 1957. As in previous studies, those with high scores had poorer survival (c 2, = 0.012). Unlike previous studies, the relation between scores and survival was linear. Cynicism. Hostile Affect, and Aggressive Responding subsets were related to survival, whereas the other subsets were not. The sum of the three predictive subsets, with c 2 of 9.45 ( = 0.002), was better predictor than the full scale, suggesting that it may be possible to refine the scale and achieve an even more effective measure of those aspects of hostility that are deleterious to health.

INTRODUCION

Hostility is broad psychological domain encompassing various cognitive, emotional, and behavioral aspects of an individual's negative orientation toward interpersonal transactions. Traits in the hostility domain include cynicism, anger, mistrust, and aggression. Recent studies have linked the presence of chronic hostility to adverse health outcomes including hypertension, coronary heart disease (CHD), and all-cause mortality (1-4), These studies have employed variety of hostility definitions and have operationalized the construct in number of ways. Such multiple approaches can lead to confusion and misunderstanding, for it is tempting to think of different measures of hostility as equivalent when, in fact, they may be tapping very different aspects of the hostility domain. Therefore, it is important for researchers to make the nature of their measures as explicit as possible.

The MMPI-based Cook and Medley scale (5) is measure that has been used in number of studies of the health consequences of hostility (6-8). The present research is designed to define more clearly the nature of the psychological domain measured by the scale (Study 1) and to provide further documentation further health impact of that domain (Study 2).

STUDY 1

The 50-item scale was originally devised to identify teachers who had difficulty getting along with their students (5). Those MMPI items discriminating between high and low scorers on the Minnesota Teacher Attitude Inventory, test correlated with teacher-student rapport, were chosen for the initial item pool. Refinements in item selection were performed by group of clinical psychologists. Probably because the items were chosen on an empirical rather than theoretical basis, the item content is diverse and the underlying dimension is not well defined. Not surprisingly, this has led to disagreement about the nature of the scale (9, 10).

Previous attempts to describe the psychological domain measured By the scale have relied on empirical approaches. Costa et al. (11) factor analyzed data from 1002 cardiac patients and found two major factors that they called "cynical mistrust" and "paranoid alienation." Smith and Frohm (12) examined the correlates of the scale and suggested that the scale measured "cynical hostility." Blumenthal et al. (13) examined the correlates of in sample of cardiac patients. They suggested that high scores might reflect neuroticism, social difficulties, and ineffective coping as well as anger and hostility. The present study takes different approach, using rational analysis of the item content to suggest descriptors of the psychological dimensions measured By the scale. Items were then grouped into subsets on an priori basis. Correlates of the item subsets were examined to evaluate the results of the categorization procedure.

Relying on the face validity of the items and previous theorizing, we identified six subsets of items in the scale. We labeled these item subsets Hostile Attributions, Cynicism, Hostile Affect, Aggressive Responding, and Social Avoidance, plus group of miscellaneous items which we placed in an Other category. The principles we used to derive these categories were based on theories of aggression, attitudes, and information processing (4, 14-19).

The first theoretical division was between behavioral and experimental items. Aggression theorists distinguish between instrumental aggressive behaviors (14) and those motivated By the experience of anger (15). The former behaviors involve acts of coercion designed to achieve an external goal, whereas the latter are motivated By the experience of anger. Dodge and Coie (16) have found that these two aspects of hostility are separable into two types of aggression called "proactive" and "reactive." Siegman et al. (17) have made similar distinction between "expressive" and "neurotic" hostility, while Dembroski and Costa (4) emphasize the difference between "antagonistic" and "neurotic" hostility. Therefore, the grouping of items on the basis of their behavioral vs. experiential emphasis is derived from an extensive tradition in hostility research.

Further distinctions within the experiential domain were based on theories of attitudes and social information processing (18,19). We separated items that were statements of belief from those that reflected the experience of affect. Furthermore, we identified items in the scale that express two types of beliefs.

One subset of belief items, Hostile Attributions, reflects tendency to interpret the behavior of others as intended to harm the respondent. These items are admissions of suspicion, paranoia, and fear of threat to the self. Hostile attribution biases have been found to relate to reactive, but not proactive, aggression in children (16). Two aspects of statement are necessary for an item to qualify for this subset: 1) mistrust of the intentions of another person and 2) the perception that the other's hostile intention is directed toward the respondent. This latter requirement differentiates Hostile Attribution items from those in the second belief subset, Cynicism. Cynicism. items reflect generally negative view of humankind, depicting others as unworthy, deceitful, and selfish. They are statements about the respondent's interpretation of others' behavior in general, with the target of their actions unspecified. Cynicism items convey more general world view than those in the Hostile Attribution subset.

Items in the Hostile Affect subset refer to the experience of negative emotions associated with social relationships. They are admissions of anger, impatience, and loathing when dealing with others. Unlike the Aggressive responding items, they do not imply that the respondent overtly acts on the basis of these affects.

Within the behavioral domain, the scale includes items describing overtly aggressive behaviors as well as those referring to indirectly negative behaviors. Aggressive Responding items indicate the respondent's tendency to use anger and aggression as instrumental responses to problems or to endorse these behaviors as reasonable and justified. Overt interpersonal behavior is indicated or implied.

Items describing indirectly negative behaviors were grouped in subset termed Social Avoidance. They indicate the respondent's tendency to avoid others, refrain from social interaction, or withdraw from interpersonal involvement. They lack the flavor of interpersonal confrontation that is common to the subsets discussed above.

Finally, there is subset of items that do not appear to have common underlying psychological content. We have simply assigned them to an Other category.

The initial assignments of items to these categories were performed By the authors on the basis of face validity without regard to external empirical criteria. As check on our procedure, 14 clinical psychologists and psychologists-in-training were given detailed descriptions of the six categories and asked to assign each item to the most appropriate subset. The classifications of five items were changed in light of these data. We remained convinced of the accuracy of the initial assignments for nine items (52, 59, 250, 271, 292, 383, 406, 485, and 504), even though there was evidence of disagreement among the judges. There was good agreement between the judges and the initial assignment for the remaining items. Correlates of these item subsets were examined in two data sets to see if the subsets were differentially correlated with other personality variables. This was intended as test of the adequacy of the classification procedure and as way of placing the domain measured By the scale within the context of normal personally theory. To control for the possibility of chance correlations, only the findings that were replicated in both samples were treated as the basis for firm conclusions.

Methods

Subjects were participants in the Computer Administered Panel Survey (CAPS) administered by the Institute for Research in Social Science at the University of North Carolina, Chapel Hill (20, 21). CAPS respondents were chosen from those who responded to an invitation that was sent to random sample of undergraduates. Comparisons with university records have shown that they do not differ from the undergraduate population on any available demographic or academic variable. They attended computerized data collection sessions each week of the school year where they answered variety of questionnaires and participated in experiments. They were paid four dollars for each session. The measures of interest in the present study were administered as part of the regular CAPS sessions. Data were obtained from two CAPS cohorts, 1985-1986 and 1986-1987. Complete data were available for 45 men and 44 women in the 1985-1986 sample; 44 men and 44 women were available in the 1986-1987 sample. They represented all undergraduate classes (mean age = 19.9 years), came from predominantly middle class backgrounds, and were predominately (85%) white. Tests were performed on the data of the two cohorts separately, allowing us to assess the replicability of our findings across samples.

The Cook and Medley scale was given as part of the full I to the 1986-1987 CAPS respondents. Only the 50 items were administered to the 1985-1986 respondents. The means were 20.8 (SE = 0.77) for the 1985-1986 sample and 22.3 (SE = 0.85) for the 1986-1987 sample. Separate scores were computed for each of the item subsets described above.

Correlations between scores on the subsets and scores from the NEO Personality Inventory (NEO-PI) (22) were calculated. The NEO-PI is 181-item test of normal personality based on Norman's (23) five-factor taxonomy of personality traits. It yields scores for the five major dimensions, Neuroticism, Extraversion. Openness, Agreeableness. and Conscientiousness, plus number of subscales reflecting facets of these dimensions. Two NEO-PI scales would be expected to reflect hostility: the dimension of Interpersonal Antagonism vs. Agreeableness and the Hostility (N2) subscale of the Neuroticism dimension. According to the manual of the NEO-PI, the Agreeableness vs. Antagonism domain assesses the quality of one's interpersonal orientation along continuum from compassion to antagonism in thoughts, feelings, and actions. High Antagonism (or low Agreeableness) scorers are perceived as cynical, rude, suspicious, uncooperative, vengeful, ruthless, irritable, and manipulative. The Hostility (N2) facet of the Neuroticism domain assesses more affective form of hostility, with high scores suggesting person who is frequently hot-tempered, angry, and easily frustrated.

Results

Intercorrelations of the item subsets are presented in Table 1. Data from the 1985-1986 cohort are presented above the diagonal, while the 1986-1987 data are below the diagonal. Cynicism, Hostile Attribution, Aggressive Responding, and Hostile Affect tend to be highly correlated, Correlations involving the Social Avoidance and Other subsets are not as consistently strong, These data call attention to the fact that these subsets are not independent. This is not surprising since they were derived in fashion that neither required not tried to achieve empirical orthogonality. They were conceived as aspects of the same set of predispositions.

The correlations of the full scale with the NEO-PI scores are presented in Table 2. The correlations of the item subsets with the NEO-PI scales are presented in Table 3. There is high level of consistency in the results of the two samples, and the patterns support both the convergent and "discriminant validity of the subsets**. The convergent validity of four subsets (Cynicism, Hostile Attribution, Aggressive Responding, and Hostile Affect) is supported by the fact that they were consistently correlated in the expected direction with Agreeableness and Hostility (N2). The discriminant validity of these four subsets is supported by the fact that none of them were consistently related to the dimensions of Extraversion, Openness, or Conscientiousness. The relations of these subsets to Neuroticism is more complex. The Cynicism and Hostile Affect subsets were consistently correlated with Neuroticism, and the Aggressive Responding and Hostile Attributions subsets were not. These correlations were not substantially changed when the N2 items were omitted from the calculation of the Neuroticism score.

The Other and Social Avoidance subsets differed from the other four in their patterns of correlations. Social Avoidance scores were correlated with Agreeableness, but not Hostility. They were also negatively correlated with Extraversion, an expected result given the nature of the items. The items in the Other subset consistently correlated only with Neuroticism.

Table 1. Intercorrelations of Item Subsets

 

Cynicism

Hostile

Attribution

Aggressive Responding

Hostile

Affect

Social

Avoidance

Other

Cynicism

Hostile Attribution

Aggressive Responding

Hostile Affect

Social Avoidance

Other

- a

0.64 b

0.48 b

0.41 b

0.38 b

0.37 b

0.54 b

- a

0.40 b

0.51 b

0.38 b

0.44 b

0.40 b

0.32 b

- a

0.21 c

0.11

0.19

0.43 b

0.32 b

0.40 b

- a

0.28 b

0.37 b

0.16

0.22 c

0.21 c

0.11

- a

0.25 c

0.35 b

0.29 b

0.13

0.18

-0.01

- a

Correlations above the diagonal are based on the data of the 1985-1986 cohort; those below are based on the data of the 1986-1987 cohort.

b < 0.01.

< 0.05.

Table 2. Correlations of Full Scale with NEO Scales

 

1985-1986 Cohort

1986-1987 Cohort

Neuroticism

Extraversion

Openness

Conscientiousness

AgreeabIeness

N2 (Hostility)

0.26

-0.06

-0.27 b

-0.01

-0.49 b

0.48 b

0.52 b

-0.04

-0.11

0.12

-0.49 b

0.48 b

a < 0.05.

b < 0.01.

Table 3. Correlations of Item Subsets with NEO Scales

 

Cynicism

Hostile Attribution

Aggressive Responding

 

1985-1986

1986-1987

1985-1986

1986-1987

1985-1986

1986-1987

Neuroticism

Extraversion

Openness

Conscientiousness

AgreeabIeness

N2 (Hostility)

0.21 a

-0.11

-0.27 b

-0.01

-0.40 b

0.37 b

0.34 b

-0.06

-0.18

0.16

-0.40 b

0.25 b

0.17

-0.05

-0.27 b

0.12

-0.38 b

0.23 a

0.53 b

0.01

-0.06

0.19

-0.39 b

0.39 b

0.05

0.17

-0.15

0.01

-0.44 b

0.46 b

0.06

0.11

-0.26 a

0.30 b

-0.42 b

0.39 b

 

Hostile Affect

Social Avoidance

Other

1985-1986

1986-1987

1985-1986

1986-1987

1985-1986

1986-1987

Neuroticism

Extraversion

Openness

Conscientiousness

AgreeabIeness

N2 (Hostility)

0.32 b

-0.03

0.01

-0.10

-0.21 a

0.50 b

0.53 b

-0.12

0.11

-0.05

-0.33 b

0.42 b

0.00

-0.37 b

-0.09

0.05

-0.31 b

0.09

0.31 b

-0.34 b

0.05

-0.12

-0.39 b

0.30 b

0.29 b

0.04

-0.09

-0.20

-0.07

0.14

0.49 b

0.10

0.01

-0.19

-0.10

0.25 a

a < 0.05.

b < 0.01.

 

Discussion

It appears that the item subsets were related to Norman's (23) five-factor model of personality in an expected fashion. Cynicism, Hostile Attribution, Hostile Affect, and Aggressive Responding were strongly correlated with the NEO-PI measures of Hostility and Agreeableness in both samples. These two NEO-PI scales were themselves highly correlated (-0.53 and -0.52 in the two samples). On the other hand, Social Avoidance was negatively related to Extraversion and Agreeableness but was not consistently related to Hostility. Social Avoidance items probably reflect different kind of disagreeableness than that conveyed by the other subsets, one based on withdrawal rather than confrontation. Items in the Other category appear to be associated with aspects of Neuroticism other than Hostility.

Items in the Other and Social Avoidance categories differ from the other items in that their face valid content is not as obviously reflection of the hostility. Their patterns of correlations support this impression. The presence of items apparently unrelated to hostility in the scale is easily understood if one remembers the origin of the scale. The criterion used to choose items was ability of teachers to establish good interpersonal rapport with their students. It is not surprising, therefore, that items related to domains other than hostility (such as neuroticism and tendencies toward social withdrawal) could also be selected using this criterion.

When the items were assigned to subsets there was some ambiguity about the placement of nine items because of disagreements among the authors and judges. However, most of these disagreements centered around which hostility subset (Cynicism, Hostile Attribution, Hostile Affect, and Aggressive Responding) was most appropriate for the item. Since we see these subsets as highly related measures of the same psychological construct, we feel that these disagreements are of minor importance. Only two items (485 and 504) received ratings that reflected substantial disagreement with the authors about whether they belong in the hostility subsets or in the remaining two categories.

In summary, the scale appears to measure number of different aspects of interpersonal hostility: beliefs about the trustworthiness of others, negative emotions associated with social relationships, and aggressive behavior toward others. The scale also contains items reflecting tendency to avoid social contacts and group of items that appear to reflect general neuroticism. It is possible that only some components of the scale account for its relation to health. We would expect that those subsets that appear to be the best reflections of hostility (Cynicism, Hostile Attributions, Hostile Affect, and Aggressive Responding) should also be the best predictors of health outcomes. preliminary test of this hypothesis was performed in Study 2.

STUDY 2

Interest in the trait measured by the scale was initially sparked by the finding that its scores were associated with angiographically documented coronary artery disease in clinical population (6). Following this lead, Barefoot et al. (7) assessed the prospective relationship of scores to CHD incidence in study of nonclinical sample. Medical students (n = 255) at the University of North Carolina (UNC) who had completed the I in the late 1950s were surveyed about their health in 1980. As expected, those who scored high on the scale were more likely to report the occurrence of CHD events during the follow-up period. There was also an important, unanticipated finding in this study: scores were predictive of mortality from all causes. Those with scores. above the median had more than six times the mortality rate of those with scores at or below the median.

The health impact of scores was further documented in prospective study of 1877 middle-aged male Western Electric. workers conducted by Shekelle et al. (8). The ten-year incidence of major CHD events was related to scores in this sample even after statistical adjustment for the effects of traditional risk factors. As in the study by Barefoot et al. (7), scores also significantly predicted all-cause mortality during the 20-year follow-up period.

In contrast to these positive studies, three recent studies (25-27) have failed to find an association between scores and later CHD or mortality outcomes. While there may be plausible explanations for these findings (3, 26), it is not possible to identify conclusively the reasons for the failure of these studies to replicate previous results. The conflicting findings suggest the need for further evaluations of the ability of scores to predict health outcomes. Study 2 is designed to provide such test by assessing the relationship between scores and mortality rate over 29-year follow-up period.

As demonstrated in Study 1, the item content of the scale is somewhat heterogeneous. It is possible that only some of the items account for its predictive ability and that certain subsets of items may bear more consistent and theoretically coherent relation to health outcomes. Therefore, the item subsets described in Study 1 and the empirically derived subscales described by Costa et al. (11) were tested individually against the survival criterion.

Methods

The I was administered to 128 law students ! the University of North Carolina during the 1956 and 1957 school years. The test was administered in order to compare the profiles of law students to those of students in other professions. The testing took place during class time and was presented to the subjects as research project. Ages of the subjects ranged from 20 to 45 years at the time of testing, with mean age of 24.85 years. Data on traditional risk factors were not available.

mortality follow-up of the students took place in 1985. Alumni records and legal directories were the primary sources of follow-up information. Ten subjects were lost to follow-up because they had no alumni records or listings in legal directories. Their scores on the scale did not differ significantly from those remaining in the study. Thirteen individuals were identified as deceased. Causes of death were determined from death certificates (10 subjects) and obituaries (3 subjects).

Results

The mean score in this sample was 14.65 (SE = 0.65) ***, figure slightly higher than that found in the UNC medical students and well within the normal range of scores for adult samples (28). This suggests that the conditions of test administration were appropriate, since it has been suggested that evaluative testing conditions can result in unusually low scores that may be invalid (3). Furthermore, all scores on the L scale were in the normal range, adding support to the conclusion that social desirability biases did not play major role in the test administration.

Six of the 13 deaths were attributed to probable CHD and 1 death was due to congestive heart failure. Cancer accounted for four deaths and two were due to thrombosis associated with diabetes mellitus. The number of deaths was too small to perform separate analyses by cause of death.

The association between scores and the probability of being deceased is presented in Figure 1. The categories of scores in the figure are based on the quartiles of the Barefoot et al. study (7). If categories are based on quartiles of the present sample, the proportions of deceased are 0.04, 0.10, 0.10, and 0.20 as one goes from the lowest to the highest category.

Fig. 1. Relationships between Ho scores and mortality incidence over the follow-up period.

model proportional hazards survival analysis (29) found that scores were linearly related to survivorship, c 2 (1 df) = 6.37, p = 0.012. Tests for deviations from linearity were nonsignificant. This analysis controlled for age, which was also significant, c 2 (1 df) = 7.3, p = 0.007.

An estimate of effect size was calculated from the beta coefficient of the proportional hazards analysis. The estimated risk of dying for persons who scored 1 SD (SD = 7.06) above the mean on the scale was 4.19 times that for persons scoring 1 SD below the mean.

second goal of Study 2 was an evaluation of the predictive validity of the subscales empirically derived by Costa et al. (11) and the item subsets described in Study 1. This was done by performing separate model analyses (controlling for age) for each subscale and subset. Results are presented in Table 4. The empirically derived subscales were significant but did not differ from each other or from the full scale in their ability to predict survival. However, the chi-squares for the rationally derived item subsets did differ great deal. Cynicism, Hostile Affect, and Aggressive Responding were significant predictors of survival, whereas the Hostile Attribution, Social Avoidance, and Other subsets were not. Moreover, the sum of the three predictive subsets yielded chi-square nearly 50% larger than that produced by the full scale. This increased ability to predict survival was due to both an increase in effect size and decrease in the standard error of the beta coefficient. The estimated risk of dying for person who scored 1 SD above the mean (SD = 4.53) was 5.54 times that of person who scored 1 SD below the mean.

 

Table 4. Survival Analysis Chi-Squares for Item Subsets

Item set

c 2

p

Full scale

Costa t al. (22) factors

Cynicism

Paranoid Alienation

Study 1 Subsets

Cynicism

Hostile Attribution

Aggressive Responding

Hostile Affect

Social Avoidance

Other

Sum of Cynicism, Hostile Affect, and Aggressive Responding

6.37

4.95

5.37

4.63

1.83

7.30

5.77

0.80

0.36

9.45

0.012

0.026

0.021

0.031

0.177

0.007

0.016

0.371

0.550

0.002

Discussion

scores were significant predictors of early mortality despite the fact that this study was performed on relatively small sample. When this finding is viewed in combination with the two previous positive prospective studies, it presents strong case for the health impact of the trait measured by the scale. However, the existence of previous negative studies (25-27) suggests that high hostility, at least as measured by the scale, is not invariably harmful. Perhaps some aspect of the environment is needed to potentiate the adverse health consequences of the trait measured by the scale. Future research should concentrate on identifying those aspects of the environment that act in conjunction with high hostility to produce adverse health outcomes.

In the present study, the combination of the three item subsets, Cynicism, Hostile Affect, and Aggressive Responding, comprising only about half of the items, had larger chi-square and nearly one-third greater relative risk than did the full scale. The Other and Social Avoidance item subsets, those identified in Study 1 as less clearly related to hostility, were not significant predictors of survival. The Hostile Attribution subset, whose items may indicate paranoid tendencies, was also not significant.

In addition to the performance of these subsets against the purely empirical crit8rion of survival, the combination of Cynicism, Hostile Affect, and Aggressive Responding forms theoretically coherent group. That is, Cynicism items are statements of belief. Hostile Affect items reflect emotional experiences, and Aggressive Responding items refer to behavior. In combination, these three subsets may define those items that come closest to the comprehensive measurement of the hostility construct. They appear to correspond to the functions of knowing, feeling, and acting, distinction frequently made by psychologists and philosophers in their analyses of human experience (19). It is important to emphasize that the subsets do not measure independent entities but appear to be components of the same construct, hostility.

The data of this study also call our attention to the fact that hostility is not simple construct but has multiple components. We should be able to improve our operationalization of hostility and our ability to predict health outcomes by using measures that assess combination of beliefs, affects, and behaviors.

It may be tempting to conclude that we have found the items that account for the health impact of scores, but we wish to avoid premature closure on this matter. The small sample size of the present study argues for such caution. Therefore, we prefer to encourage replication so that future research can determine which subsets best relate to health outcomes and can enable us to arrive at proper term to describe those aspects of the hostility domain that are truly health-damaging (or health-promoting). Future studies should examine the predictive validity of these items subsets in addition to that of the full scale. This research strategy has been successfully used to isolate hostility as the most important component of the Type behavior pattern for the prediction of CHD outcomes (3). Such tests are now under way on the data from other samples, and results will be communicated in subsequent papers.

The size of the Ho-survival association among the lawyers in this study was roughly comparable to that found by Barefoot et al. (7). Shekelle et al. (8) reported much smaller effect, 1.42 increase in risk of dying associated with an increase of 23 points. Since the Western Electric study differs from the other two in number of ways, its smaller effect size has several possible explanations. The estimate of the effect size in the Western Electric study was adjusted for traditional risk factors (blood pressure, cholesterol, smoking, and ethanol intake) that were not assessed in other studies. The Shekelle et al. study (8) also differed from the others in the occupational status of the participants and their age at intake into the study.

The age difference between the samples is particularly promising explanation for the effect size differences. Williams et al. (30) found that Type behavior was related to the severity of coronary artery disease in younger, but not older, angiography patients. They hypothesized that, like cholesterol and cigarette smoking, psychosocial factors such as Type behavior and hostility have their major impact on health in younger age groups because those individuals at risk who survive to older ages may be biologically hardier. Higher mortality among hostile individuals at young ages might serve to eliminate those who are less fit from the population. This could result in population of older hostile persons that is biologically hardier than the older nonhostile population, which has not been subject to such selection. This would lessen the predictive value of hostility in studies of older people. similar argument was made by Siegman et al. (17), who found relationship between "expressive hostility" and angiographically documented coronary artery disease in patients less than 60 years old, but not in older patients. The observation that scores have stronger predictive power in studies of young men than in study of middle-aged men is consistent with this line of reasoning.

The form of the Ho-mortality relation is also noteworthy. In previous studies, there was tendency for scores to be associated with health outcomes in nonlinear, threshold fashion, so that the strongest relation was found when scores were treated as dichotomous variable. The data of the present study revealed linear dose-response relation between and survival. Obviously, the form of the relation between scores and health outcomes is an open question that can only be resolved with further research.

Portions of this study were presented at the Annual Meeting ! the American Psychosomatic Society, Philadelphia, , 1987.

This work was supported in part by grant HL-36587 from the National Institutes of Health, grant KO4 HDoo806 from the National Institute of Child Health and Human Development, grant -70482 from the National Institute of Mental Health, and the John D. and Catherine . MacArthur Foundation.

We thank Dr. Paul Costa, Dr. Ilene Siegler, and Dr. Larry Scherwitz for their mments jn an earlier draft of the r. The assistance f Dora Burton, Thomas n, Dr. Beverly Wiggins, and Dr. Bibb Latane is also appreciated.

 

* Items in the subsets have MMPI booklet numbers as follows: Cynicism 59, 71, 89, 93,117,124, 252, 265, 280, 319, 406, 436, and 558; Hostile Attribution 110, 136, 157, 237 (Reversed), 244, 278, 284, 348, 458, 469, 507, and 551; Hostile Affect 148, 226, 383, 399 (Reversed), and 438; Aggressive Responding 28, 250, 253 (Reversed), 271, 410, 426, 447, 504, and 520; Social Avoidance 52, 292, 368, and 455; Other 19, 183, 386, 394, 411, 485, and 531. summary of the judges' assignments is available upon request from the authors.

** Dr. Paul Costa (personal communication) has exarnined the correlations of the subsets to the NEO-PI scales in sarnple of 274 older adults nrolled in the Baltirnore Longitudinal Study of Aging (24). reports substantial agreement with the pattern of correlations presented in Table . These data available upon request to the authors of this r.

*** The scores of 30 participants who failed to answer ll 50 items were prorated to adjust for missing data. Results were not changed when analyses were performed without this adjustment.

 

REFERENCES

  1. Chesney , Rosenman R (eds): Anger and Hostility in Cardiovascular and Behavioral Disorders. New York, Hemisphere/McGraw Hill, 1985
  2. Diamond : The role of anger and hostility in essential hypertension and coronary heart disease. Psychol Bull 92:410-433,1982
  3. Williams RB. Barefoot JC: Coronary-prone behavior: The emerging role of the hostility complex. In Houston , Snyder CR (eds), Behavior Pattern: Research, Theory, and Intervention. New York, Wiley, 1988, 189-211
  4. Dembroski . Costa : Coronary-prone behavior: Components of the pattern and hostility. J Pers 55:211-235, 1987
  5. Cook W, Medley D: Proposed hostility and pharasaic-virtue scales for the MMPI. J Appl Psychol 38:414-418, 1954
  6. Williarns RB, n TL, Lee , Kong , Blurnenthal J, Whalen R: behavior, hostility, and ron heart disease. Psychosorn Med 42:539-549, 1980
  7. Barefoot JC, Dahlstrorn WG, Willians RB: Hostility, CHD incidence, and total rnortality. 25- follow-up study of 255 physicians. Psychosom Med 45:59-63, 1983
  8. Shekelle R, Gale , Ostfeld , Paul O: Hostility, risk of heart disease, and mortality. Psychosorn Med 45:109-114,1983
  9. Sallis JF, Johnson , Trevorrow TR, lan RM, Hovell MF: The relationship between cynical hostility and blood pressure reactivity. JPsychosorn Res 31:111-116, 1987
  10. Megargee : The dynarnics of aggression and their application to cardiovascular disoders. In Chesney , Rosenrnan R (eds): Anger and Hostility in Cardiovascular and Behavioral Disoders. New York, Hernisphere/McGraw ill, 1985, 31-57
  11. Costa , Zonderman , r RR, Williams RB: Cynicism and paranoid alienation in the Cook and Medley scale. Psychosorn Med 48: 283-285,1986
  12. Smith TW, Frohrn KD: What's so unhealthy about hostility? Construct validity and psychosocial correlates of the Cook and Med!ey scale. Health Psychol 4:503-520, 1985
  13. lumenthal JA, Barefoot JC, Burg , Williarns RB: Psychological correlates of hostility among patients undergoing ron angiography. r J Med Psychol 60: 349-355, 1987
  14. Bandua : Aggression: Social Learning Analysis. Englewood Cliffs, NJ, Prentice-Hall, 1973
  15. Berkowitz L: Aggression: Social Psychological Analysis. New Yok, McGraw Hill, 1963
  16. Dodge , Coie JD: Social-information-processing factors in reactive and proactive aggression in children's r groups. J Pers Soc Psychol 53:1146-1158, 1987
  17. Siegman AW, Dembroski , Ringel N: Conponents of hostility and the severity of coronary artery disease. Psychosom Med 49:127-135, 1987
  18. Dodge , Petit GS, McClasky CL, Brown : Social Competance in Children. Monogr Soc Res Child Dev 51:2,1986
  19. McGuire WJ: Attitudes and attitude change. In Lindzey G, Aronson (eds): Handbook of Social Psychology, vol 2. New York, Random House, 1985, 235-346
  20. Beza , Latane : multipurpose data collection facility at the Institute for Research in Social Science. Soc Sci Newsletter 68:27-31, 1983
  21. Wiggins : The UNC CAPS panel: new data facility for social science. Soc Sci Newsletter 69:100-104, 1984
  22. Costa , r RR: The NEO Personality Inventory Manual. Odessa FL, Psychological Assessment Resources, 1985
  23. Norman WT: Toward an adequate taxonomy of personality attributes: Replicated factor structure in r nomination personality ratings. J Abnorm Soc Psychol 66: 574-583, 1963
  24. Shock NW, Greulich RC, Andres R, Arenberg D, Costa , Lakatta EG, Tobin JD: Normal Human Aging: The Baltimore Longitudinal Study of Aging (NI Publication No. 84-2450). Bethesda, MD, National Institutes of Health, 1984
  25. McCranie , Watkins L, Brandsma J, Sisson : Hostility, CHD incidence, and total mortality: Lack of association in 25- follow-up study of 478 physicians. J Behav Med 9:119-125, 1986
  26. Leon GR, Finn SE, Bailey JM: The inability to predict cardiovascular disease from I special sls related to patterns. Psychosom Med 49:205, 1987
  27. rn MD, Murray DM, Luepker RV: Hostility, coronary heart disease, and total mortality: 33- follow-up study of university students. J Behav Med, in press
  28. Swenson WM, Pearson JS, Osbourne D: An I Source Book: Basic Item, Scale, and Pattern Data on 50,000 Medical Patients. Minneapolis, University of Minnesota Press, 1973
  29. Kalbfleisch J, Prentice R: The Statistical Analysis of Failue Time Data. New York, Wiley, 1980
  30. Williams RB, Barefoot JC, n TL, rrll FE, lumenthal JA, rr DB, Peterson BL: behavior and angiographically documented coronary atherosclerosis in sample of 2289 patients. Psychosom Med 50:139-152, 1988