Abstract
SUMMARY Records of cows each milked in two different herds have been used to estimate a component of variance due to cows that is free of variation arising from cow-by-herd interaction effects. By making use of a prior estimate of within-herd repeatability the variance component due to cow-by-herd interaction effects was estimated, and found to represent 35% of the total within-herd variance of butterfat yields and 9% of that of butterfat percentage records.