In the search for candidate loci having a role in susceptibility to common diseases such as multiple sclerosis (MS), studies often look for an 'association' between one allele at a putative candidate susceptibility locus and the disease, e.g. MS. However, reproducibility of results from association studies has been difficult. In MS, despite numerous studies, association has been reproducibly confirmed only for the major histocompatibility complex, the MHC. The present paper is designed to review the allelic associations which have been reported in MS and suggest possible reasons for the difficulties in replicating these studies. We also outline some suggestions for improving the validity of future association studies in MS and in other complex traits.