Hostility

Abstract

Hostility, which MeSH classifies under emotions, is the cognitive and attitudinal member of the anger-hostility-aggression triad: a disposition to view others with cynicism, mistrust and ill will, measured chiefly by the Cook-Medley scale, the Buss-Durkee inventory and the Hostility subscale of the Aggression Questionnaire. This article covers the measures and their factor structure, the hostility and coronary heart disease hypothesis from the 25-year physician follow-up through two meta-analyses, the mechanisms proposed to carry it, the overlap of hostility with depression and anxiety, the hostile attribution bias in children and adults and its meta-analytic link to aggression, a hostility-reduction intervention, and the interpersonal perspective on hostility and health. The central dispute is whether hostility is a specific risk factor or one face of a general negative affectivity. Three demos show the triad, the attribution bias and the heart-disease evidence.

Keywords: hostility, Cook-Medley scale, coronary heart disease, hostile attribution bias, aggression

Hostility sits between a feeling and an act. Where anger is the emotion and aggression the behaviour, hostility is the attitude: a cynical, mistrustful stance toward other people that outlasts any single provocation. Replicated factor analyses of a new questionnaire on aggression yielded four scales, Physical Aggression, Verbal Aggression, Anger and Hostility, and correlational analysis revealed that anger is the bridge between both physical and verbal aggression and hostility (Buss & Perry, 1992). The construct was first put on a scale for a different purpose, as proposed hostility and Pharisaic-virtue scales for the Minnesota Multiphasic Personality Inventory (Cook & Medley, 1954), and an inventory for assessing different kinds of hostility followed (Buss & Durkee, 1957). What made hostility a research programme rather than a questionnaire was the discovery, nearly three decades on, that scores predicted coronary disease and death.

What Hostility Is

Hostility is the member of the triad that lives in cognition. The scale that carries most of the health literature was assembled from Minnesota Multiphasic Personality Inventory items as a hostility scale alongside a Pharisaic-virtue scale (Cook & Medley, 1954), and an a priori classification of its items later identified six subsets, Cynicism, Hostile Attributions, Hostile Affect, Aggressive Responding, Social Avoidance and Other (Barefoot et al., 1989). The subsets name what hostility is made of: a cynical view of human nature, a readiness to read ill intent into others' behaviour, the affect that accompanies that reading, and the inclination to answer it. Anger is the bridge between both physical and verbal aggression and hostility, and the scales correlated differently with various personality traits, which suggests the need to assess not only overall aggression but also its individual components (Buss & Perry, 1992). Men scored slightly higher on Verbal Aggression and Hostility and much higher on Physical Aggression, with no sex difference for Anger (Buss & Perry, 1992).

Figure 1

The Anger-Hostility-Aggression Triad

Measuring Hostility

Three instruments and one interview carry the literature. The Cook-Medley hostility scale was identified as a predictor of health outcomes before anyone knew what it measured, and the item classification into six subsets showed good convergent and discriminant validity against the NEO Personality Inventory in two undergraduate samples, with some evidence that items in the Social Avoidance and Other categories reflect constructs other than hostility (Barefoot et al., 1989). In 118 lawyers who had completed the MMPI in 1956 and 1957, those with high scores had poorer survival to 1985, the relation between scores and survival was linear, the Cynicism, Hostile Affect and Aggressive Responding subsets were related to survival whereas the other subsets were not, and the sum of the three predictive subsets was a better predictor than the full scale (Barefoot et al., 1989). The Buss-Durkee inventory was built to assess different kinds of hostility (Buss & Durkee, 1957), and its successor's four scales showed internal consistency and stability over time and correlated with peer nominations of the various kinds of aggression (Buss & Perry, 1992). In the health meta-analysis, structured interview indicators of potential for hostility gave a weighted mean r of .18 with coronary heart disease, while the Cook-Medley scale and other cognitive-experiential measures were most predictive of all-cause mortality (Miller et al., 1996).

Table 1. Hostility measures and what the cited abstracts report about them.

Note. Each row lists only what the cited abstract reports, or the title where no abstract is indexed (Cook & Medley, 1954; Barefoot et al., 1989; Buss & Durkee, 1957; Buss & Perry, 1992; Miller et al., 1996). The instruments differ in format and population, so the rows are not a single comparison.

Hostility and the Heart

The hostility hypothesis grew out of research on the Type A behaviour pattern. In the prospective Multiple Risk Factor Intervention Trial, 192 cases of coronary death or nonfatal myocardial infarction were compared with 384 matched controls on the behavioural characteristics associated with the pattern; the global Type A pattern was not significantly related to coronary heart disease incidence, and after adjustment for age, serum cholesterol, blood pressure and cigarette smoking only Potential for Hostility, dichotomized into low and high, showed a significant relative risk of 1.5, while ordinal logistic regression revealed a nonsignificant effect (Dembroski et al., 1989). The hypothesis also acquired its own cohort. High levels of hostility as assessed by the Minnesota Multiphasic Personality Inventory scale had been found associated with increased levels of arteriographically documented coronary atherosclerosis, and in a 25-year follow-up of 255 medical students who completed the inventory while in medical school, high hostility scores were found to be predictive of both clinical coronary disease incidence and total mortality (Barefoot et al., 1983). A review a decade later found that the evidence available from prospective studies, although not entirely consistent, suggested that hostile persons may be at increased risk for subsequent coronary heart disease and other life-threatening illnesses, and that the most central remaining concerns were the assessment of individual differences in hostility and the role of social contexts in the psychosomatic process (Smith, 1992).

Two meta-analyses estimated the direction and size of the effect. A meta-analysis that included 15 studies used in previous reviews and 30 new independent studies found that overall, the results suggest hostility is an independent risk factor for coronary heart disease; after controlling for other risk factors, the widely used Cook-Medley Hostility Scale and other cognitive-experiential measures were most predictive of all-cause mortality (weighted mean r = .16) and, to a lesser extent, coronary heart disease (weighted mean r = .08), and the increased use of high-risk studies in recent years produced an increase in null findings (Miller et al., 1996). A review of prospective cohort studies, 25 studies of initially healthy populations and 19 studies of samples with existing coronary heart disease, found anger and hostility associated with increased coronary events in the healthy population studies (combined hazard ratio 1.19, 95% confidence interval 1.05 to 1.35) and with poor prognosis in the patient studies (hazard ratio 1.24, 95% confidence interval 1.08 to 1.42), with indications of publication bias, a harmful effect in the healthy populations that was greater in men than women, and an association in patients that persisted in studies that controlled fully for basal disease status and treatment (Chida & Steptoe, 2009). The authors concluded that this supports the use of psychological management focusing on anger and hostility in the prevention and treatment of coronary heart disease (Chida & Steptoe, 2009).

Mechanisms

How a questionnaire score reaches the coronary arteries is the harder question. Several plausible mechanisms possibly linking hostility and health have been articulated and subjected to initial evaluation: hostile individuals display heightened physiological reactivity in some situations, report greater degrees of interpersonal conflict and less social support, and may have more unhealthy daily habits (Smith, 1992). Inflammation is one candidate route. In 90 healthy, nonsmoking men, analyses of plasma interleukin-6 adjusting for age, body mass index, lipids and resting blood pressure revealed a significant interaction between depressive symptoms and Cook-Medley hostility: hostility was correlated with interleukin-6 (r = 0.59) only among men with Beck Depression Inventory scores of 10 and above, depressive symptoms were correlated with it (r = 0.61) only among men with hostility scores of 23 and higher, and men with high scores on both exhibited the highest median levels, suggestive of a subpopulation at increased risk for future cardiac events (Suarez, 2003).

That interaction points to the field's standing problem. Several reviews have identified three affective dispositions, depression, anxiety and anger-hostility, as putative risk factors for coronary heart disease, with mixed and negative results, and recognition of the construct and measurement overlap among the three negative affects necessitates more complex affect-disease models and leaves open the possibility that a general disposition toward negative affectivity may be more important for disease risk than any specific negative affect (Suls & Bunde, 2005). The interpersonal perspective answers in part: a review of research on the assessment of hostility, anger and aggressiveness, their association with coronary heart disease and longevity, and the mechanisms possibly underlying the association illustrated the value of the interpersonal tradition in personality psychology for research on the health consequences of these traits (Smith et al., 2004). A later integrative review holds that emotional characteristics and the presence and quality of personal relationships are robust predictors of major medical illness and premature mortality that have largely been studied separately, and calls for integration across their interconnected roles over development, the role of emotion regulation in relationship contexts, and dyadic and dynamic designs (Smith & Weihs, 2019).

The Hostile Attribution Bias

The cognitive side of hostility has its own literature, built in children. Aggressive and nonaggressive boys from grades 2, 4 and 6 were exposed to a frustrating negative outcome instigated by an unknown peer who had acted with either a hostile intent, a benign intent or an ambiguous intent; all groups responded with more aggression in the hostile condition than in the benign condition, and aggressive and nonaggressive subjects differed only in the ambiguous condition, where aggressive subjects responded as if the peer had acted with a hostile intent and nonaggressive subjects as if the peer had acted with a benign intent (Dodge, 1980). The hypothesis that aggressive children respond to ambiguous-intention-negative-consequence situations with aggression because they infer a hostile intention was supported by a follow-up study, and the total pattern suggested a picture of the aggressive child caught up in a spiraling cycle of reputation and behaviour (Dodge, 1980). In Dodge and Coie's studies the bias tracked one kind of aggression: across four studies, attributional biases and intention-cue detection deficits were related to reactive aggression but not to proactive aggression, with only the two reactive-aggressive groups displaying biases and deficits in interpreting videorecorded provocations (Dodge & Coie, 1987). The reformulated social information-processing model placed hostile attributional biases among the processing steps likely to be causal of behaviours that lead to social status (Crick & Dodge, 1994).

The meta-analyses agree on the association and both report effect sizes that vary widely between studies. Forty-one studies with 6,017 participants showed a robust significant association between hostile attribution of intent and aggressive behaviour, with effect sizes that differed considerably between studies; larger effects were associated with more severe aggressive behaviour, rejection by peers as a selection criterion, inclusion of 8- to 12-year-old participants and absence of control for intelligence, and staging of actual social interactions was associated with the largest effects (Orobio de Castro et al., 2002). A multilevel meta-analysis of 111 studies with 219 effect sizes and 29,272 participants again found a positive association with effect sizes that varied widely, and suggested that hostile intent attribution is a general disposition guiding behaviour across a broad variety of contexts, whereas the strength of its relation to aggression depends on the level of emotional engagement; the relation was stronger for more reliable measures but not stronger for reactive aggression or comorbid attention-deficit hyperactivity disorder than for aggression in general (Verhoef et al., 2019). Across cultures, when children attribute hostile intent to peers they are more likely to predict they would react aggressively, and differences in this tendency statistically account for differences in future chronic aggressive behaviour problems across children and across ecological-context groups (Dodge et al., 2015). Systematic reviews have since asked how the bias varies by aggression subtype and gender in children and adolescents (Martinelli et al., 2018) and how it relates to aggression in adults (Klein Tuente et al., 2019), and the mediating role of anger rumination between the bias and aggression has been examined (Quan et al., 2019).

Reducing Hostility

If hostility raises coronary risk, lowering it should be tested. A preliminary study examined the effects of a hostility-reduction intervention on patients with coronary heart disease: 22 high-hostile men were matched on age and hostility and randomly assigned to a hostility intervention (10) or an information-control group (12), and were reassessed immediately and 2 months after treatment on hostility, by self-report and structured interview, and resting blood pressure (Gidron et al., 1999). The intervention's overall effect size was moderately strong (d = .62), intervention patients were reported and observed to be less hostile than controls, at follow-up they had significantly lower diastolic blood pressure, reductions in hostility were significantly and positively correlated with reductions in diastolic pressure, and replication with a larger sample and coronary outcomes was recommended (Gidron et al., 1999). The meta-analytic conclusion points the same way: the evidence supports the use of psychological management focusing on anger and hostility in the prevention and treatment of coronary heart disease (Chida & Steptoe, 2009). The cited intervention did not measure cardiac events.

Worked Example

Three numbers from the heart-disease literature are worth translating. The hazard ratios of 1.19 for healthy populations and 1.24 for patients (Chida & Steptoe, 2009) mean that anger and hostility were associated with about 19% and 24% higher rates of coronary events and poor prognosis respectively, so the patient effect is about five percentage points larger (24 - 19 = 5). The weighted mean correlations of .16 with all-cause mortality and .08 with coronary heart disease for cognitive-experiential measures (Miller et al., 1996) square to about 2.6% (0.16 x 0.16 = 0.0256) and 0.6% (0.08 x 0.08 = 0.0064) of variance, so hostility explains about four times as much variance in mortality as in coronary disease (0.0256 / 0.0064 = 4.0). The attribution meta-analyses grew from 41 studies and 6,017 participants (Orobio de Castro et al., 2002) to 111 studies and 29,272 participants (Verhoef et al., 2019), a 2.7-fold increase in studies (111 / 41 = 2.71) and a 4.9-fold increase in participants (29,272 / 6,017 = 4.86), so the average study grew from about 147 participants (6,017 / 41 = 146.8) to about 264 (29,272 / 111 = 263.7). The intervention that showed d = .62 had 22 patients (Gidron et al., 1999), about one for every 1,331 participants in the later attribution meta-analysis (29,272 / 22 = 1,330.5). Small effects on large samples and a moderate effect on a small sample are the two halves of this literature.

Discussion

First, hostility earned its place in health psychology by prediction, and the prediction has survived two meta-analyses. The 255 physicians (Barefoot et al., 1983) gave the hypothesis its own cohort; the 45-study meta-analysis suggested hostility is an independent risk factor with weighted mean correlations of .16 for mortality and .08 for coronary disease, and noted that high-risk studies produced an increase in null findings (Miller et al., 1996); the prospective review found hazard ratios of 1.19 and 1.24 with indications of publication bias (Chida & Steptoe, 2009). The effect is modest and consistent in direction across both meta-analyses.

Second, the specificity question is open. Depression, anxiety and anger-hostility overlap in construct and measurement, and a general disposition toward negative affectivity may be more important for disease risk than any specific negative affect (Suls & Bunde, 2005); the interleukin-6 data, in which hostility predicted inflammation only among men with elevated depressive symptoms (Suarez, 2003), can be read either as hostility's specific contribution or as the overlap at work. The item analysis offers a partial resolution: the Cynicism, Hostile Affect and Aggressive Responding subsets predicted survival and the other subsets did not (Barefoot et al., 1989), so the active ingredient is narrower than the full scale.

Third, the cognitive and the cardiac literatures describe the same disposition from two ends. The hostile attribution bias is a general disposition guiding behaviour across a broad variety of contexts (Verhoef et al., 2019), it relates to reactive rather than proactive aggression (Dodge & Coie, 1987), and hostile individuals report greater interpersonal conflict and less social support (Smith, 1992). An integrative perspective, on which emotion and relationships are interconnected influences on health (Smith & Weihs, 2019), is the account that joins the two ends; a hostility-reduction trial with cardiac events as the outcome is the study that would test it.

Current Directions

The research front runs through integration and mechanism. The agenda set out after a conference co-sponsored by the American Psychosomatic Society and the Society for Affective Science includes common biological foundations of affective and relationship processes, the cultural embeddedness of affective and relationship processes, the potential contribution of affective-relational processes to health disparities, and implications for intervention research (Smith & Weihs, 2019). In the attribution literature, the importance of understanding specific moderators of effect size for theory development is the stated next step (Verhoef et al., 2019), identifying the hostile-intent mechanism could lead to novel interventions, education and policies that reduce or prevent aggressive behaviour (Dodge et al., 2015), and the adult literature has been reviewed systematically (Klein Tuente et al., 2019). The question Suls and Bunde posed, whether a general negative affectivity rather than hostility in particular drives disease risk (Suls & Bunde, 2005), remains the one a reader should carry to any new hostility study, and the recommendation to replicate the hostility-reduction intervention with a larger sample and coronary outcomes (Gidron et al., 1999) remains that study's own stated next step.

Common Misconceptions

“Hostility, anger and aggression are the same thing.”

Replicated factor analyses yielded four scales, Physical Aggression, Verbal Aggression, Anger and Hostility, and anger is the bridge between both kinds of aggression and hostility (Buss & Perry, 1992); attributional biases related to reactive but not proactive aggression (Dodge & Coie, 1987).

“Hostile people just have a hair-trigger temper.”

The items that predicted survival were the Cynicism, Hostile Affect and Aggressive Responding subsets of the Cook-Medley scale (Barefoot et al., 1989), and aggressive boys differed from nonaggressive boys only when a peer's intent was ambiguous, responding as if the peer had acted with hostile intent (Dodge, 1980); the disposition is a way of reading others, not only a speed of reacting.

“Hostility is a proven, large cause of heart attacks.”

The 1996 meta-analysis suggested that hostility is an independent risk factor for coronary heart disease, but the weighted mean correlations were .16 for all-cause mortality and .08 for coronary disease (Miller et al., 1996), and the prospective hazard ratios were 1.19 and 1.24 with indications of publication bias (Chida & Steptoe, 2009).

“Hostility is a risk factor on its own, separate from depression and anxiety.”

The construct and measurement overlap among depression, anxiety and anger-hostility leaves open the possibility that a general disposition toward negative affectivity may be more important for disease risk than any specific negative affect (Suls & Bunde, 2005), and hostility predicted interleukin-6 only among men with elevated depressive symptoms (Suarez, 2003).

Glossary

Aggression Questionnaire.

The Buss-Perry instrument whose replicated factor analyses yielded four scales, Physical Aggression, Verbal Aggression, Anger and Hostility.

Anger-hostility-aggression triad.

The related traits of hostility, anger and aggressiveness, long suggested as risk factors for coronary heart disease (Smith et al., 2004); anger is the bridge between hostility and aggression (Buss & Perry, 1992).

Buss-Durkee Hostility Inventory.

The 1957 inventory for assessing different kinds of hostility, predecessor of the Aggression Questionnaire.

Cook-Medley hostility scale.

The MMPI-based scale proposed in 1954 alongside a Pharisaic-virtue scale; the measure that predicted coronary disease and mortality in the physician cohort and survival in the lawyer cohort.

Cynicism.

One of six a priori item subsets of the Cook-Medley scale, and one of the three, with Hostile Affect and Aggressive Responding, that predicted survival.

Emotional engagement.

In the 2019 meta-analysis, the level of emotional engagement was indexed by the type of stimulus presentation, from hypothetical stories read by the participant to real-time interactions with a real peer, and the relation between hostile intent attribution and aggression was stronger in social situations that elicit sufficient emotional engagement (Verhoef et al., 2019).

Hazard ratio.

The ratio of event rates between groups over time; 1.19 for anger and hostility in healthy populations and 1.24 in coronary patients in the prospective meta-analysis.

Hostile attribution bias.

The tendency to attribute hostile intent to others whose intentions are ambiguous; a general disposition guiding behaviour that is robustly associated with aggressive behaviour in children.

Interleukin-6.

An inflammatory cytokine implicated in cardiovascular disease; in healthy men it rose with hostility only among those with elevated depressive symptoms.

Interpersonal perspective.

The tradition in personality psychology that Smith and colleagues used to study the health consequences of hostility, anger and aggressiveness and, more generally, the effects of emotion and personality on physical health.

Negative affectivity.

A general disposition toward negative affect that, given the overlap among depression, anxiety and anger-hostility, may matter more for disease risk than any specific negative affect.

Potential for hostility.

The structured-interview rating of hostility whose weighted mean correlation with coronary heart disease was .18 in the 1996 meta-analysis.

Psychosomatic hypothesis.

The proposal that hostile persons are at increased risk for coronary heart disease and other life-threatening illnesses through physiological reactivity, interpersonal conflict and unhealthy habits.

Reactive and proactive aggression.

Two forms of childhood aggression; in Dodge and Coie's studies, attributional biases and intention-cue detection deficits related to the reactive form but not the proactive one.

Social information-processing model.

The reformulated model of the steps between a social cue and a behavioural response, in which hostile attributional biases are likely to be causal of behaviours that lead to social status.

Key Researchers

Arnold H. Buss

. Professor Emeritus of Psychology at the University of Texas at Austin (1924-2021); built the Buss-Durkee Hostility Inventory and, with Perry, the Aggression Questionnaire that separates hostility from anger and aggression. Wikipedia

Bram Orobio de Castro

. Professor of Child and Youth Care Sciences at the University of Amsterdam; led the 2002 meta-analysis of hostile attribution of intent and aggression in children and co-authored its 2019 successor. Faculty Page

Nicki R. Crick

. Distinguished McKnight University Professor at the University of Minnesota's Institute of Child Development (1958-2012); co-authored with Dodge the reformulated social information-processing model in which hostile attribution is one step toward aggression. Wikipedia

Kenneth A. Dodge

. Professor at Duke University's Sanford School of Public Policy; identified the hostile attributional bias in aggressive children and showed its link to aggression across cultures. ORCID - Faculty Page - Wikipedia - Google Scholar

Timothy W. Smith

. Distinguished Professor of Psychology at the University of Utah; framed hostility and health as a psychosomatic hypothesis, co-authored the 1996 meta-analysis and set out the interpersonal perspective on hostility and coronary heart disease. Faculty Page - Wikipedia - Google Scholar

Andrew Steptoe

. Professor of Psychology and Epidemiology at University College London; co-authored the meta-analysis of prospective evidence linking anger and hostility to future coronary heart disease. ORCID - Faculty Page - Wikipedia - Google Scholar

Jerry Suls

. Professor Emeritus of Psychological and Brain Sciences at the University of Iowa; co-authored the review showing that depression, anxiety and anger-hostility overlap as cardiovascular risk factors. Faculty Page

Redford B. Williams

. Professor Emeritus of Psychiatry and Behavioral Sciences at Duke University; co-authored the 25-year physician follow-up that tied Cook-Medley hostility to coronary disease and mortality, and the item analysis of the scale. Faculty Page

Frequently Asked Questions

What is hostility in psychology?

Hostility is the attitudinal member of the anger-hostility-aggression triad, measured as a distinct scale of the Aggression Questionnaire alongside Physical Aggression, Verbal Aggression and Anger, with anger as the bridge between hostility and aggression (Buss & Perry, 1992); its Cook-Medley items divide into Cynicism, Hostile Attributions, Hostile Affect, Aggressive Responding, Social Avoidance and Other (Barefoot et al., 1989).

How is hostility different from anger?

In the Aggression Questionnaire they load on separate scales, anger is the bridge between hostility and both kinds of aggression, men scored slightly higher on Hostility while there was no sex difference for Anger, and the scales correlated differently with personality traits (Buss & Perry, 1992).

Does hostility cause heart disease?

The 1996 meta-analysis suggested that hostility is an independent risk factor for coronary heart disease, with weighted mean correlations of .16 for all-cause mortality and .08 for coronary disease (Miller et al., 1996); prospectively, anger and hostility carried hazard ratios of 1.19 in healthy populations and 1.24 in coronary patients, with indications of publication bias (Chida & Steptoe, 2009).

How is hostility measured?

By the Cook-Medley scale drawn from MMPI items (Cook & Medley, 1954), whose Cynicism, Hostile Affect and Aggressive Responding subsets carry the survival prediction (Barefoot et al., 1989); by the Buss-Durkee inventory (Buss & Durkee, 1957) and the Hostility scale of the Aggression Questionnaire (Buss & Perry, 1992); and by structured-interview ratings of potential for hostility (Miller et al., 1996).

What is the hostile attribution bias?

Aggressive boys differed from nonaggressive boys only when a peer's intent was ambiguous, responding as if the peer had acted with hostile intent (Dodge, 1980); across 111 studies and 29,272 participants, hostile intent attribution is a general disposition guiding behaviour whose link to aggression depends on emotional engagement (Verhoef et al., 2019).

Why do hostile people get sick?

Hostile individuals display heightened physiological reactivity in some situations, report greater interpersonal conflict and less social support, and may have more unhealthy daily habits (Smith, 1992); in healthy men, hostility predicted plasma interleukin-6 only among those with elevated depressive symptoms (Suarez, 2003).

Can hostility be reduced?

In a preliminary study, 22 high-hostile coronary patients were randomly assigned to a hostility intervention or an information control; the intervention's effect size was moderately strong (d = .62), intervention patients were less hostile and had lower diastolic blood pressure at follow-up, and replication with a larger sample and coronary outcomes was recommended (Gidron et al., 1999).

Is hostility really distinct from depression and anxiety as a risk factor?

The three negative affects overlap in construct and measurement, which leaves open the possibility that a general disposition toward negative affectivity may be more important for disease risk than any specific negative affect (Suls & Bunde, 2005); an integrative perspective proposes studying emotion and relationships together as interconnected influences on health (Smith & Weihs, 2019).

References