The lack of a formal definition of human longevity continues to generate confusion about its genetic and nongenetic determinants. In order to characterize how differences in birth year cohorts and percentiles of survival are associated with familial contribution to variation in survival, we estimated sibling relative risk of living to increasingly rare percentiles of survival based on a dataset of 1,917 validated sibships each containing at least one individual living to age 90 years. About 1,042 of the sibships included at least one individual who survived to age 100 and 511 included at least one individual who survived to age 105 and older. We show that sibling relative risk increases with older ages, sex, and earlier birth year cohorts of the proband and siblings of male 90-year-olds (5th percentile of survival) have 1.73 (95% CI: 1.5; 2.0) times the chance of living to age 90, while siblings of both male and female probands who survived to age 105 years (~0.01 percentile of survival) have 35.6 (95%CI: 15.1; 67.7) times the chance of living to age 105 compared with population controls. These results emphasize the importance of consistently defining the longevity phenotype in terms of rarity of survival for appropriate comparisons across studies.