The etiological role of genotype and environment in recalled age at menarche was examined using an unselected sample of 1,177 MZ and 711 DZ twin pairs aged 18 years and older. The correlation for onset of menarche between MZ twins was .65 +/- .03, and that for DZ pairs was .18 +/- .04, although these differed somewhat between four birth cohorts. Environmental factors were more important in the older cohorts (perhaps because of less reliable recall). Total genotypic variance (additive plus nonadditive) ranged from 61% in the oldest cohort to 68% in the youngest cohort. In the oldest birth cohort (born before 1939), there was evidence of greater influence of environmental factors on age at menarche in the second-born twin, although there was no other evidence in the data that birth trauma affected timing. The greater part of the genetic variance was nonadditive (dominance or epistasis), and this is typical of a fitness trait. It appears that genetic nonadditivity is in the decreasing direction, and this is consistent with selection for early menarche during human evolution. Breakdown of inbreeding depression as a possible explanation for the secular decline in age at menarche is discussed.