BACKGROUND: The International Type 2 Diabetes Linkage Analysis Consortium was formed to localize type 2 diabetes predisposing variants based on 23 autosomal linkage scans. METHODS: We carried out meta-analysis using the genome scan meta-analysis (GSMA) method which divides the genome into bins of approximately 30 cM, ranks the best linkage results in each bin for each sample, and then sums the ranks across samples. We repeated the meta-analysis using 2 cM bins, and/or replacing bin ranks with measures of linkage evidence: bin maximum LOD score or bin minimum p value for bins with p value <0.05 (truncated p value). We also carried out computer simulations to assess the empirical type I error rates of these meta-analysis methods. RESULTS: Our analyses provided modest evidence for type 2 diabetes-predisposing variants on chromosomes 4, 10, and 14 (using LOD scores or truncated p values), or chromosome 10 and 16 (using ranks). Our simulation results suggested that uneven marker density across studies results in substantial variation in empirical type I error rates for all meta-analysis methods, but that 2 cM bins and scores that make more explicit use of linkage evidence, especially the truncated p values, reduce this problem. CONCLUSION: We identified regions modestly linked with type 2 diabetes by summarizing results from 23 autosomal genome scans. Copyright (c) 2008 S. Karger AG, Basel.