1. A statistical study of the causes of variation in milk yields has been made on the basis of 5000 lactation records from twelve leading herds of Ayrshire cattle in south-west Scotland for the period 1930–9.
2. The unit of measurement of milk production employed was the yield during the first 180 days of the lactation period. This eliminated the effect of variations in length of current calving interval on milk yield.
3. The effect of month of calving on milk yield varied significantly between herds, and it was shown that correction factors for month of calving should be calculated on a within-herd basis. The average difference in 180-day yield between the summer and winter calvers of all herds was about 10% in favour of winter calvers.
4. The milk yield of a cow was found to be influenced both by the number of her previous lactations and also by her age at calving. The types of corrections for age employed by previous investigators were discussed, and it was shown that percentage corrections are the most satisfactory.
5. There was a positive correlation between milk yield and length of preceding calving interval. From an economic point of view, the optimum length of calving interval was about 400 days for the first lactation, and 1 year for subsequent ones. Corrections for preceding calving interval, like those for age, were most satisfactory when they were proportionate and not additive.
6. No significant differences were found among the first three records of a cow in their ability to indicate her actual production capacity. The probable performance of a cow in any lactation was predicted as accurately from the lactation immediately preceding it as from the average of a number of previous lactations.
7. The average repeatability of milk yield was 0.46, and heritability was of the order of 0.25–0.30.
8. The probable effect on herd improvement of selecting breeding females was found to be very small, extremely little genetic progress being attained by this method in the twelve herds.