At intake into a multiple coronary heart disease (CHD) risk factor intervention trial, 3110 individuals were interviewed to assess Type A behavior. After an average of 7 years follow-up, the 193 individuals who manifested their first CHD event were matched with 384 CHD-free individuals. To assess self-involvement, auditors counted all verbal self-references (I, me, my) and clauses spoken in the audiotaped baseline interviews. Self-references were entered into multiple logistic regression analyses that controlled for age, diastolic blood pressure, cholesterol, cigarette smoking, and Type A behavior. Relative to matched controls, those who incurred CHD spoke more self-references at baseline [p = 0.017; relative risk (RR) = 1.20], but did not self-reference more densely. Relative to matched controls, those who died from CHD spoke more self-references (p = 0.008; RR = 1.62) and self-referenced more densely (p = 0.027; RR = 1.54). Neither total self-references nor self-reference density was predictive of angina pectoris or nonfatal myocardial infarction (MI). However, among those who incurred MI, self-reference frequency was the strongest predictor of mortality among all the measured risk factors (p = 0.01, RR = 2.0). The results suggest that self-involement is related to CHD incidence.