The regression of kcal./kg. milk on fat percentage in some sets of observations departs significantly, in a probability sense, from linear. In some cases the regression line is concave upward; in others, concave downward. The quantitative difference between the linear and curvilinear regression lines for the same set of observations is small. Where only the wts. of milk and its fat content are known it seems best to estimate milk energy by a linear equation. For this purpose there seems no need to modify the widely used formula, FCM = 0.4[image] + 15 F, where FCM is milk energy in terms of 4% milk, [image] is milk, F is milk fat, all in the same unit of wt.