Twelve Friesland cows, in a Latin Square experimental design with 4 blocks and 42-day treatment periods, were subjected from the end of the second month of lactation, to 3 dietary treatments viz. a conventional system of roughage fed according to body size and concentrates according to production, an ad lib. offering of a complete diet and a free choice ad lib. offering of a concentrate mixture and hay, ground through screens of 2 different sizes. There were no statistically significant differences in the mean daily production of 4% fat corrected milk (FCM) on the 3 treatments, but cows on the third treatment showed a remarkably constant preference for the different dietary components while they produced significantly more milk but with a lower percentage of butterfat.