Recent genome-wide association (GWA) studies have localized genetic variants involved in medical conditions such as Type I and Type II diabetes, and breast cancer. GWA-studies for complex psychological traits, such as depression and schizophrenia, have however been less successful. Yet, establishing the genetic variants involved in complex traits would increase our knowledge of the genetic basis of behaviour, and may eventually inform treatment. Complex traits are thought to be influenced by multiple genetic variants, each explaining a small percentage of the variance. In addition, complex traits are often multidimensional and relatively difficult to measure reliably. Whereas statistical and technical challenges of GWA-studies (e.g., multiple testing, power, and population stratification) have received ample attention, psychometric problems (e.g., misclassification, measurement (in)variance, and multidimensionality) have largely been neglected. Yet, psychometric problems may hamper the search for genetic variants involved in complex traits, complicate comparisons between studies, and impede replication of results. The aim of this proposal is to increase the chances of localizing genetic variants involved in complex traits through two related psychometric projects. First, I will determine, through simulation studies, how phenotypic multidimensionality should be accommodated in GWA studies. In this context, I will develop a new statistical approach, which integrates phenotypic and genetic information, to uncover phenotypic dimensions that actually map onto genetic pathways. Second, I will focus on how violations of measurement invariance over subpopulations affect GWA results. Feasibility of this study is ensured by the current public availability of multiple GWA data sets.