The broad principles of animal breeding are outlined and the ‘normal-probability curve’ is used to illustrate how progress should be made in improving the genetic merit of dairy cattle. The importance of genetically correlated responses is discussed, together with the need to test progeny. The way in which genetic improvement accumulates in dairy cattle is described and the need to use high-production sires with reliable proofs is stressed. Present and proposed milk-quality payment schemes require that selection should be based on the yield of recorded solids as the best indicator of total solids. As cattle breeding involves a long time-scale, it is not always possible to take into account short-term changes in the value of milk constituents. The effect of changing values upon the ranking of bulls is small and, for the foreseeable future, the use of the weight of recorded solids, i.e. the weight of fat plus the weight of protein, as the selection factor, is likely to maximize progress. The relationship between functional type and herd life is examined. It is concluded that there is no obvious relationship between type and herd life; yield is the dominant factor in determining the length of time an animal remains in the herd. Any breeding programme should therefore place maximum emphasis on production.