1. A statistical analysis has been made of the production records in all lactations, up to the third, of 10,967 British Friesian cows. The work was done in two sections, for those animals by home-bred sires and for those by sires imported from Holland and Canada. Within each section, three parallel sets of calculations were made, based on all animals for lactation 1, on those with at least two completed lactations for lactations 1 and 2, and on those with three completed lactations. The analysis was done in a hierarchical manner, taking variables in the order sire, herd, year, month and age at calving. The production measurements were 305-day yield, fat content and lactation length, the latter having an upper limit of 305 days.
2. It was found that in all three characters the variance between animals identical for all environmental variables was approximately two-thirds of that between animals commencing the same lactation in the same herd in the same year. This difference could not be attributed to any particular variable and appeared to be an effect of exact contemporaneity.
3. Culling produced appreciable selection differentials in yield and lactation length but not in fat content. The regression coefficient, calculated within sire groups, of selection differential for heifer yield within sires on the heifer progeny test was 0·11 for culling at the end of the first lactation and —0·07 for culling at the end of the second. In the analysis of British sires for yield, selection in the first lactation reduced both sire and within sire components by about 20%, so that the heritability was little changed. In the Imported sires, the sire variance was reduced by nearly 40 %, so that the heritability fell from 0·26 to 0·22. Because of the truncated distribution of lactation length, selection reduced the within sire variance for the Imported sires by about 60% and the heritability from 0·13 to 0·09.
4. The heritability of yield was lower in second and third lactations than in first. This seemed to be due to an increase of about 50 % in the within-sire component whereas that between sires remained constant. Fat content showed a similar trend and the heritability was slightly less in later lactations.
5. The repeatability of the characters was in general agreement with the results of other workers. The genetic correlation of first yield with later ones averaged 0·80 (significantly different from unity), but that between second and third yields averaged 0·91. The genetic correlations between fat content in different lactations were all close to unity, as were those between length in different lactations.
6. In agreement with other results, the negative genetic and phenotypic correlations between yield and fat content appeared to approach zero as the animals aged. This was due to an above average increase of fat with age in the daughters of bulls high for yield.
7. As would be expected from the genetic correlations between yield in first and later lactations, the increase of yield with age had a heritability significantly different from zero. First lactation yield was negatively correlated (—0·27) with increase from first to second but positively (0·03) with increase from second to third. The heritability of the average yield and fat content increased little with age.
8. The consequences of these results for the optimum combination of different records in selection decisions are discussed.