The objective of this study was to investigate the feasibility of developing a system whereby measurements taken on bulk tank milk samples could be used to monitor the level of subclinical mastitis in dairy herds. The variables that were examined were the logarithmically transformed total somatic cell counts and percentages of cell volume in channel 8 (volumes from 89.2 to 178.3 microm(3)), the presence or absence of Streptococcus agalactiae and various husbandry/management factors including herdsize and the use of teat dips. Each of the use of actual monthly and rolling average bulk tank cell count determinations was investigated. It was found that the inclusion of all variables resulted in a correct classification of approximately 85% of herds and that no improvement was achieved by the use of rolling as opposed to actual monthly values. The inclusion of various husbandry/management practices improved the percentage correct classification to some extent over that achieved by the sole use of total somatic cell counts and percentages of cell volume in channel 8 when the herds were grouped on the basis of quarter infection rate (<10%, >10%) but not in the case of the cow infection rate categories (<20%, >20%). The use of both total cell counts and percentages of cell volume in channel 8 did not improve the overall predictive value over that achieved by the sole use of percentage of cell volume in channel 8 in the case of the quarter infection rate groupings but did to some extent in the case of the cow infection rate groupings. When the classification functions were applied prospectively and considering combinations of the two cell count determinations only, it was found that they were able to correctly classify, on the basis of the quarter infection rate groupings, approximately 75% of the study herds. It is concluded that the system described herein has limited application as a basis for selecting problem herds.