A recent study suggested that the cadherin gene FAT exerts an influence on susceptibility to bipolar affective disorder (BPAD). We aimed to replicate this finding in a German sample (425 BPAD I and 419 controls). In addition, we performed a comprehensive linkage disequilibrium mapping of the whole genomic region of FAT and the neighboring circadian gene MTNR1A (48 single nucleotide polymorphisms (SNPs) covering 191 kb). No significant association was observed for SNPs located in the MTNR1A gene. In FAT, however, nine SNPs showed association, eight of them being located in the same haplotype block found to be associated with BPAD by Blair et al. The smallest P-value of 0.00028 (OR 1.71) was seen for non-synonymous SNP rs2637777. A combination of five markers including this marker showed a haplotype distribution with a nominal P-value of 1.8 x 10(-5) that withstands correction for multiple testing. While the control allele frequencies between our sample and the samples of the original study are comparable, tendencies of risk allele frequencies are opposite. Possible explanations for this include potential differences in linkage disequilibrium structure between the German, Australian, UK, and Bulgarian populations sampling variation, multilocus effects and/or the occurrence of independent mutational events. We conclude that our results support an involvement of variation at the FAT gene in the etiology of BPAD, but that further work is needed both to clarify possible reasons for the observed risk allele differences and to ultimately identify the functionally relevant variant(s).