Mastitis history, California Mastitis Test scores, and filter-deoxyribonucleic acid cell counts from 232 cows were used to project the effectiveness of criteria which could be used to determine which cows to treat in a selective dry cow therapy program. Bacteriological analyses of quarter milk samples were used to identify infected cows. Effects of month, lactation parity, stage of lactation, and interaction of stage of lactation with lactation parity on somatic cell numbers in milk were determined. The various criteria singly or in combinations correctly identified from 50 to over 92% of the cows with one or more infected quarters at drying off. However, from 25 to 80% of the cows free of infection would have been treated as well. The most discriminating criteria were two monthly cell counts exceeding 500,000 cells per milliliter where 53% of the cows with infected quarters were identified correctly and only 25% of the uninfected cows would be treated, or the two monthly cell counts exceeding 500,000 cells per milliliter plus a California Mastitis Test score of +2 or +3 at drying off. Here 89% of the cows with infected quarters would be treated, and 56% of the uninfected cows would be treated.