While association studies for the mapping of genetic variants that influence complex traits have primarily focused on populations of European descent, more recent studies involve populations with admixed ancestry, such as African Americans and Hispanics. Genetic association studies in ancestrally admixed populations offer exciting opportunities for the identification of novel variants that underlie trait diversity. At the same time, the heterogeneous genetic background and dependencies among sample individuals from admixed populations, including both ancestry differences and relatedness among sample individuals, pose special challenges for genetic association testing. In these circumstances, it is necessary to devise statistical methods for association mapping that account for the diverse genomes of the sample individuals and are robust in the presence of a variety of complex sample structures. We will present a linear mixed-model score test for genetic association with quantitative traits in admixed populations. We will demonstrate that the method can provide an improvement over existing mixed-model methods in terms of both power and type 1 error in samples from admixed populations. We will also demonstrate the utility of the method with an application to a sample of 3,500 self-identified Hispanics from the Women's Health Initiative study.