Some of the difficulties in the interpretation of the group-feeding method are removed when the animals are fed individually, and when for each animal a record of the amount of feed consumed is obtained. The statistical treatment of the data may then be extended to the economy of gains as well as to the rate of gains. But still certain difficulties remain. We are still handicapped in the application of statistical methods to these results by the fact, pointed out by Dr. Lush, that the animals in a group receiving the same ration have not been selected at random, and hence that the gains secured and the feed costs of gains may not themselves possess the characteristic of randomness essential to the application of the ordinary probability methods. The force of this objection relates particularly to experiments in swine feeding, since for these animals it may be shown clearly that initial body weight, the dominant factor in the judgment of the animal as a feeder, is rather highly correlated with gains secured in a subsequent feeding period.