In recent years enormous progress in determining the sequence of the human genome has led to a rapid development of research into polymorphisms in genes involved in complex diseases. It is clear, however, that there are important limitations in many of these association studies. Problems with reliable and reproducible phenotyping, the number of individuals studied, racial heterogeneity, population stratification (founder effect), functionality and multiple testing often mean that studies are not reproducible. In this review we describe a number of the limitations related to this type of research; from both our own experience with studies on polymorphisms in the glucocorticoid receptor gene, and shortcomings and solutions from the literature.