[en] Background information on the use of molecular genetic markers to detect and select for genes controlling genetic variation (quantitative trait loci, QTL) is provided by Williams and van der Werf (these proceedings). Use of QTL is predicted to be most beneficial for traits that have low heritability or are difficult, expensive or impossible to record during a breeding program. These conditions apply to disease resistance and adaptation traits in the low to medium inputs systems of the developing world and use of QTL could therefore be expected to be particularly beneficial here. The failure to use QTL information in the developing world reflects a lack of investment in QTL mapping. Such investment is needed not only to detect QTL that could be useful in genetic improvement programs, but also to design improvement programs utilising QTL information that would be sustainable under developing world conditions. A strategy for use of molecular markers that is hypothesis driven and has clear goals and routes to impact poor farmers is outlined here. The strategy is presented in greater detail by Gibson. Population genetics theory predicts that natural selection will fix different genetic solutions in populations that are isolated from each other. Selection acts stochastically on the variation available, and this variation will differ in nature and extent between populations. The more genetically distinct are any two populations, the greater the likelihood they will contain distinct genetic polymorphisms and the greater the chance that selection will lead to fixation of different genetic solutions to the same problem in the two populations. In order to identify the best possible genotype for each of a range of production environments, the ideal situation would be to test all breeds with potentially useful characteristics globally, along with all their crosses in each production environment. In practice such testing is not feasible, due to economic and logistical limitations. It would be feasible in many cases to undertake testing of just two breeds from two different countries (or regions). Many countries would see the advantage of a reciprocal exchange of germplasm with another country, which could overcome concerns related to benefit sharing in many cases. The critical question is which two breeds would maximise the probability of being able to develop a better genotype than currently exists? Obviously choice of breeds will involve careful examination of existing data on breed performance and the environments under which they evolved. Where it is desired that a particular trait be further improved one consideration would be the likelihood that two breeds have evolved different mechanisms of adaptation, such that a higher level of adaptation (and/or performance and/or resistance) could readily be developed from a cross between them. In this case one would seek breeds that have suitable phenotypes in the targeted environment, yet are as genetically distant from each other as possible. Genetic distances among existing breeds can be estimated using molecular genetic markers and a global survey of distances among all the breeds of each species need be completed only once. Having selected two breeds it will be important to test the hypothesis that they carry different genetic mechanisms controlling the desirable traits, before proceeding with a breeding program. A suitable method for testing that hypothesis is to perform a genome-wide interval mapping for QTL based on anonymous genetic markers in an F2 and/or backcrosses between the two breeds. Based on whether or not the hypothesis is confirmed, the size of the QTL detected and the performance of the pure breeds and the F2 or backcrosses, an informed decision can then be taken on a suitable genetic improvement program. The outcome might be to utilise one of the purebreds, or to develop a crossbreeding program, or to develop a new breed through selection from a crossbred or backcross population, or to introgress QTL from one breed to the other. In many cases the population used for testing the QTL hypothesis can also be used as the base population of a breeding program. An informed decision can be taken on whether or not the genetic improvement program should incorporate marker-based selection. This decision will depend not just on the potential value of the marker information, but also the cost and logistics of collecting and using the marker information in the genetic improvement program. The steps outlined above, from initial mapping of global livestock diversity through to hypothesis testing and possible use of markers in selection forms a coherent strategy for detecting useful genetic variation between different populations, with a clear route to utilizing such variation using molecular genetic information