It is desirable to develop criteria with which to judge whether given persons or groups of persons are likely to have coronary heart disease later in life. The technique of discriminant function analysis of variance has been applied to data on 100 normotensive patients under the age of 40 with coronary heart disease and to 146 putatively normal subjects. The presence or absence of coronary heart disease was correlated with the logarithms of height and of total serum cholesterol, serum phospholipids, and serum uric acid levels and with four anthropometric indexes and familial incidence. The highest correlation was between coronary heart disease and total serum cholesterol level. There was a negative correlation with stature, the taller subjects being less likely to develop coronary heart disease. The correlations found have been combined in a multiple regression equation which has a measurable predictive value in groups of persons and promises to be useful as a research tool.