Genetic association studies of SLC6A4 (SERT) and obsessive-compulsive disorder (OCD) have been equivocal. We genotyped 1241 individuals in 278 pedigrees from the OCD Collaborative Genetics Study for 13 single-nucleotide polymorphisms, for the linked polymorphic region (LPR) indel with molecular haplotypes at rs25531, for VNTR polymorphisms in introns 2 and 7 and for a 381-bp deletion 3' to the LPR. We analyzed using the Family-Based Association Test (FBAT) under additive, dominant, recessive and genotypic models, using both OCD and sex-stratified OCD as phenotypes. Two-point FBAT analysis detected association between Int2 (P = 0.0089) and Int7 (P = 0.0187) (genotypic model). Sex-stratified two-point analysis showed strong association in females with Int2 (P<0.0002), significant after correction for linkage disequilibrium, and multiple marker and model testing (P(Adj) = 0.0069). The SLC6A4 gene is composed of two haplotype blocks (our data and the HapMap); FBAT whole-marker analysis conducted using this structure was not significant. Several noteworthy nonsignificant results have emerged. Unlike Hu et al., we found no evidence for overtransmission of the LPR L(A) allele (genotype relative risk = 1.11, 95% confidence interval: 0.77-1.60); however, rare individual haplotypes containing L(A) with P<0.05 were observed. Similarly, three individuals (two with OCD/OCPD) carried the rare I425V SLC6A4 variant, but none of them passed it on to their six OCD-affected offspring, suggesting that it is unlikely to be solely responsible for the 'OCD plus syndrome', as reported by Ozaki et al. In conclusion, we found evidence of genetic association at the SLC6A4 locus with OCD. A noteworthy lack of association at the LPR, LPR-rs25531 and rare 425V variants suggests that hypotheses about OCD risk need revision to accommodate these new findings, including a possible gender effect.