Abstract: The purpose of this paper is to give a short survey of the present problems concerning the genetic improvement of horse breeds. The evolution of these populations in Europe, characterized by a deep change from production of draught horses towards that of leisure horses, is described and the influence of the demographical parameters on the selection of these horse populations, is discussed. The generation interval represents an important handicap only surmounted in the case of racing breeds where a high selection intensity can be practised since all animals are subjected to performance testing. In the other cases, the farmer usually does not use modern breeding techniques, but uses crosses instead, which lead more easily to visible results. The available selection criteria are also dealt with. A distinction is made between direct estimates evaluating the abilities of the animals in practice and the indirect estimates measuring a character in correlation to previous ones. For the former estimates, a distinction is made between those resulting from competitions (handicap, records or earnings) and those resulting from direct in-station measurements (saddle, jumping, dressage abilities, draught power). For the indirect estimates, often used especially for the selection of mares, the most important analysis is obviously that of the conformation. However, in the future early selection criteria according to more physiological data should be sought and developed. Estimation of the breeding value according to a given ability is thereafter pointed out. There are two situations: “the panmictic case” concerning sport and draught horses and “the non-panmictic case” corresponding to racing horses, which give rise to some problems. The setting up of breeding plans is discussed. Due to the different economic situations and various objectives of horse production, conclusions are drawn about the role played by geneticists in the present development of this sector.