Bipolar-related subphenotypes that cluster within families may help identify subsets of patients that are more genetically homogeneous. Environmental or assessment factors that segregate by family may influence estimates of familiality. We aimed to determine familiality of subphenotypes of bipolar disorder (BP), accounting for effects of age, sex, diagnosis, and site/wave of ascertainment. We studied 589 sibships with 1416 siblings affected with bipolar I (BPI), schizoaffective disorder, bipolar type (SAB), bipolar II (BPII), or recurrent unipolar depression (RUDD). Sibships were from families with ≥2 BPI cases collected by the NIMH Bipolar Genetics Initiative (NIMHBGI). Rapid cycling showed the strongest evidence for familiality [odds ratio (OR) (95%CI) = 2.02 (1.43, 2.85), P = 6.0 × 10−5] in a model including age, sex, diagnosis, and site/wave of ascertainment. Additional significantly familial traits were comorbid alcohol abuse/dependence (P = 2 × 10−4) and comorbid panic disorder (P = 8 × 10−3), as well as psychosis, suicidal thoughts, and rapid mood switching (P = 6 × 10−3 − 0.03). Omission of the effect of site/wave of ascertainment from the model inflated the significance level of the apparent familial association of almost all subphenotypes from one to four orders of magnitude. We have found evidence of familiality for subphenotypes of BP. In multicenter samples, familiality may be overestimated if variability in diagnosis of subphenotypes between site/wave of ascertainment is not considered. © 2007 Wiley-Liss, Inc.