To investigate whether milk of abnormal quality from some quarters of twelve cows on a Latin square type feeding trial would bias the results for chemical composition, the compositions of milk from normal and affected quarters were compared with each other and with that of the milk obtained from the whole udder by normal milking.

The average number of quarters producing abnormal milk in each of the four periods was eighteen, with only small variation between periods. Cell counts ranged from 50000 to 25 × 106/ml with 31% of them below 500000/ml; as cell count increased, lactose percentage and casein number fell. Pathogenic bacteria were detected in only 70% of the abnormal milks. The solids-not-fat (S.N.F.) percentage of milk from affected quarters was lower on average than that from healthy quarters by 0·43, the decrease resulting entirely from a lowered lactose percentage. The total nitrogen content was the same in both types of milk, but there was 4·5% less casein in milks from affected quarters.

Because in this particular experiment the distribution of the abnormal milks was almost uniform over the four feeding treatments, the results obtained for treatment effect on milk composition by the normal milking method were not biased by the occurrence of disease, but the general level of composition was depressed by 0·2% fat and 0·2% S.N.F. In other experiments the incidence of disease might happen by chance to differ between treatments, with the result that the findings about milk composition could be erroneous.