The etiology of type 2 diabetes (T2D) is complex and remains poorly understood. Differences in individual susceptibility to this condition reflect the action of multiple variants, each of which confers a modest effect, and their interactions with a variety of environmental exposures. Several complementary approaches to the identification of the etiological variants have been adopted, though, for all, association analyses provide the final common pathway. The genes and/or chromosomal regions studied have been selected on the basis of their presumed biological relevance to diabetes, known involvement in monogenic forms, or animal models of the condition and/or signals arising from whole-genome linkage scans. These association studies have featured a wide variety of designs and analytical approaches, but reliable biological insights have been few, largely because of difficulties in obtaining reproducible findings. However, in recent years, several examples of robustly replicated associations have emerged, largely as a result of an emphasis on the need for improved power and more appropriate analysis and interpretation. New strategies for the large-scale identification of T2D susceptibility variants are now becoming possible, including the prospect of genuine genome-wide association scans, but caution in their design, analysis, and interpretation remains essential.