Data were collected for 11 yr in five institutional herds in Virginia to determine the response to selection for milk yield and to examine correlated response in other traits. A randombred control population allowed separation of genetic and environmental effects. The genetic base was the same for both control and selection groups, and little directional selection took place in the control population. The control line was generated from 12 sire lines initiated by 8 unproven sires sampled by artificial insemination and sons of 4 proven bulls. Four bulls high for Predicted Difference milk were used in the selection group each year. By three methods of analysis, best linear unbiased prediction, deviation of selection from control group means, and least squares differences between selection and control for milk, fat, and fat percentage were: 370, 374, 438 kg; 10, 8, 6 kg; and -.02, -.02, and -.10%. These agree closely with expectations from modified contemporary comparisons of the bulls. Differences in fat percentages were more variable although all analyses indicated a negative correlated response to selection for milk yield. Generation effects varied with method of analysis but were generally small and agreed with differences in sire Predicted Differences. Response differed across herds.