AbstractAttempts to identify bipolar disorder (BP) genes have only enjoyed limited success. One potential cause for this problem is that the traditional categorical BP phenotypes currently used in genetic linkage studies are not the most informative, efficient, or biologically relevant. An alternative to these strict categorical BP phenotypes is quantitative BP phenotypes. By isolating one aspect of a complex trait such as BP into a simple, intermediate, quantitative trait, genes that contribute to the larger complex trait can be more readily identified. Along these lines, we utilized a temperament‐based measure (cyclothymic temperament) as a quantitative, intermediate BP phenotype in linkage analyses and hypothesized that this measure might more efficiently detect loci for BP or temperamental traits that predispose to BP. A total of 158 individuals with temperament data from 28 BP families were used in the linkage analyses. All pedigrees had a proband diagnosed with BPI or BPII and at least two other family members with a mood disorder diagnosis. An 8 cM genome scan was performed and analyzed using MERLIN nonparametric multipoint regression linkage for a cyclothymic temperament trait. The highest overall LOD score was on chromosome 18 (LOD = 2.71, P = 0.0002). Other linkage peaks which may indicate potential regions of interest were found on chromosomes 3 and 7. The temperament‐based cyclothymic trait yielded a higher peak LOD score and a lower P‐value than analyses using traditional, categorical phenotypes in a separate analysis including these same families. © 2007 Wiley‐Liss, Inc.