1. A statistical analysis has been made of lactation records from 500 dam-daughter pairs of the Ayrshire breed. Measurements were taken of milk yield, fat, S.N.F., crude protein, casein and lactose.

2. A correction was made for the age of the cow. This contributed a substantial fraction of the total variation for milk yield, S.N.F. and lactose contents. Month of calving had a trivial effect on the lactation average for the characters when compared with the individual variation between cows.

3. After correction for age, the variation between lactations was highest for fat content (standard deviation 0·35%) and slightly less for S.N.F. (0·26%).

4. The heritability was estimated by the dam-daughter correlation within herds. Values in the neighbourhood of 0·5 were obtained for S.N.F., crude protein, casein and ‘ash’ (determined by difference), about 0·35 for fat and lactose, 0·25 for milk yield and 0·13 for non-casein nitrogen.

5. The phenotypic correlations between yield and lactose were positive for both dam and daughter groups. The correlations between yield and the other constituents were more negative in the daughter group than in the dam group. The correlation between yield and S.N.F. was, in fact, positive in the latter. The correlation between fat and S.N.F. averaged 0·36, being derived mainly from the correlation between fat and casein (0·43), though that with lactose was also positive (0·13).

6. The variation in S.N.F. was mostly determined by that in casein (r=0·85). There was a positive correlation between lactose and casein (0·24), and a negative one between lactose and non-casein nitrogen (–0·41). Lactose was negatively correlated with ‘ash’ (–0·31). ‘Ash’ was positively correlated with non-casein nitrogen (0·23)

7. There were positive and significant genetic correlations between casein and both fat and lactose, and that between the last two was almost significant.

8. No significant genetic correlations were found between yield and the chemical constituents, although figures of practical importance may exist which did not reach significance in an experiment of this size.

9. The practical significance of these results is discussed. It is concluded that selection for yield alone is not likely to cause any great decline in S.N.F. content of the milk, and that this could probably be held in check by paying some attention to the fat content of the milk.