Hip dysplasia (malformation of the coxofemoral joint) in dogs is a major health problem for which the British Veterinary Association (BVA) had set up a control scheme in 1965. Based on scoring nine components of the radiographs of both the left and right joints, the degree of hip joint malformation is now quantified by an overall hip score (a measure of the condition of the hip joint). The hip scores of 1258 flat-coated retrievers and 1566 Newfoundlands (registered with The Kennel Club in the United Kingdom) were analysed after merging with Kennel Club pedigree data for 19036 flat-coated retrievers registered by 1995 and 14336 Newfoundlands registered by 1997. The merged data included the animal's identity, date of birth, sex and hip scores and also similar records for the dog's relatives including the hip score if the relative had been tested. In recent years, breeding has been increasingly from tested parents. There has been some reduction in offspring hip scores - presumably because breeders avoided breeding from males with very high scores. However, a much greater reduction in offspring hip score would be achieved by stricter science-based selection of potential sires and dams. Regression modelling quantified the positive relationship between offspring and parental hip scores. The genetic heritability of hip scores was large and significant in both breeds (particularly from dams). The breeders in UK tended to use healthy sires for breeding but they have taken less care in selecting dams. Our regression models emphasise the need for both sires and dams, particularly dams, to be healthy with very small hip scores.