Association mapping is a method potentially useful for detection of marker-trait associations based on linkage disequilibrium, but little information is available on the application of this technique to plant breeding populations. In this study, the conditional expectation of a quantitative trait, given a marker allele, is defined as a function of the gene effect, the conditional probability of the gene allele given the marker allele, and the covariances caused by population structure. The conditional probability can be computed based on assumptions related to the history of the breeding population and is useful for predicting the efficiency of marker-assisted selection and the potential of different populations for association mapping. Valid association analysis can be done in structured populations if covariances among related lines are recognized and included in mixed-effects models. However, because breeding populations typically have a short history of recombination, the most significant markers are not necessarily the closest to the functional gene. Additionally, we compared the potentials and limitations of germplasm bank core collections, synthetic populations, and elite germplasm as alternative experimental materials for association mapping analysis integrated with plant breeding practice. Synthetic populations offer a favorable balance of power and resolution for association mapping and would allow increasingly fine mapping of quantitative trait loci over generations.