Genome-wide association studies using recently developed large scale single nucleotide polymorphism platforms are beginning to be performed, and results reported. Initial indications are that these studies are capable of discovering loci associated with relative risks too modest to have been detectable through family-based linkage studies. However, as these studies initially test 500,000 or more polymorphisms in a first series of cases and controls, the need for robust replication in one, or preferably, several independent studies is paramount to winnow out the true positive results from the large number of expected false positives. We discuss the need for the formation of consortia to conduct these multi-stage studies, and stress the importance of full disclosure of allele frequencies in cases and controls from these studies in order to facilitate joint analyses across datasets to speed discovery of reproducible associations, and to explore more complex associations such as gene-gene interactions. Desirable characteristics of studies in which genome-wide association studies will be most informative are discussed. The validation of genetic variants that alter risk of specific cancers may be relevant to screening, the identification of high risk persons for risk-reducing interventions, and the discovery of new biological mechanisms that may provide insight into cancer causes and preventive strategies.