We tested the hypothesis that hostility is associated with increased relative risk (RR) for coronary death and nonfatal myocardial infarction among participants in the prospective Multiple Risk Factor Intervention Trial (MRFIT). Cases (N = 192) were compared with matched controls (N = 384) on a variety of behavioral characteristics associated with the Type A behavior pattern (TABP), including three different but interrelated components of hostility. Logistic regression analyses revealed that only two of the eight TABP attributes analyzed on the overall sample were significant. Only total Potential for Hostility, when dichotomized into 'low' and 'high' categories, and the antagonistic interpersonal component of hostility (Stylistic Hostility) had positive unadjusted associations with coronary heart disease (CHD) incidence (RR = 1.7, P = 0.003; and RR = 1.+5, p = 0.016, respectively). The global TABP= and related paralinguistic attributes were not significantly related to CHD incidence. After adjustment for the traditional risk factors of age, serum cholesterol, blood pressure, and cigarette smoking, only dichotomous Potential for Hostility showed a significant relative risk (RR = 1.5, p = 0.032). Ordinal logistic regression revealed a nonsignificant effect. When the total sample was stratified into younger (≥47 years) and older (>47 years) subjects, dichotomous Potential for Hostility scores in unadjusted analyses were positively associated with CHD incidence in the younger group (RR = 2.4, p ≤ 0.001), and were significantly associated after adjustment for traditional risk factors (RR = 2.1, p = 0.011); the corresponding unadjusted and adjusted values for the stylistic component of hostility were RR = 0.2.3, p ≤ 0.01 and RR = 1.9, p = 0.016, respectively. In addition, ordinal logistic regression revealed that only Stylistic Hostility showed a significant adjusted association with incident CHD. Neither Potential for Hostility nor the Stylistic Hostility component was significantly related to CHD incidence in the older group. These findings are consistent with a growing body of evidence suggesting that hostility - especially an antagonistic interactional style - but not global TABP is an independent risk factor for CHD.