The utility of ergonovine testing for coronary artery spasm was assessed in 3,447 patients with angiographically insignificant (less than 50% diameter stenosis) or no coronary artery disease. No patients clinically had Prinzmetal's variant angina. Overall, 4% had a positive ergonovine test result, defined by spasm causing greater than or equal to 75% focal stenosis. Complications related to ergonovine use occurred in 11 patients (0.03%). In a training sample of 1,136 patients (studied between 1980 and 1984), two independent predictors of spasm were found by using multivariate analysis: the amount of visible coronary artery disease on the coronary angiogram (p less than 0.0001) and a smoking history (p = 0.001). A model to predict spasm based on these variables was validated in a test group of 2,311 patients who received ergonovine from 1985 to 1989. This model allowed the identification of a subset of 400 patients in the validation sample who had a 10% positive test rate compared with a 2% positive test rate in the remaining patients. These results should permit clinicians who use provocative testing in the catheterization laboratory to reserve testing for the subset of this group of patients most likely to have abnormal findings.