Badner and Gershon (2001) presented a technique of meta-analysis of linkage data that could be applied to published genome scans. It combines the reported P-values of individual studies, after correcting each value for the size of the region containing a minimum P-value. Simulations demonstrated that the type I error rate was at least as low as that for a single genome scan and thus genome-wide significance criteria may be applied. Power to detect linkage was at least as high as the power of pooling the data from all the studies. We applied this method to all the published genome scans for bipolar disorder and schizophrenia. We found the strongest evidence for susceptibility loci on 13q (P < 6 x 10(-6)) and 22q (P < 1 x 10(-5)) for bipolar disorder, and on 8p (P < 2 x 10(-4)), 13q (P < 7 x 10(-5)), and 22q (P < 9 x 10(-5)) for schizophrenia.