higher incidence of sexually transmitted infections and consequent fecundity impairment. Part of the genetic variability in fertility might thus arise from variation in risk taking that affected many outcomes, all of which are related to reproduction.
One of the key features of behavioral genetics approaches to the analysis of human behavior is the increasing attention to the design of their studies (Plomin, 1994; Rutter et al., 2001). Moreover, there is increasing attention to interplays of nature and nurture, or gene-environment correlations and interactions. Teasing out these elements requires strong identifying and theoretical assumptions for the models used and is likely to demand longitudinal information too. By comparison with Reiss et al. (2000), for example, most behavioral genetics studies of fertility-related behavior seem fairly primitive, though it is of course early days.
Most of the studies on fertility-related behavior to date have relied on standard ACE or DCE models (Additive and Dominant; the latter include a dominant instead of an additive genetic component). Since many fertility-related studies rely on samples of twins (and occasionally siblings or other close relatives), we begin here. Essentially siblings and especially twins are defined as having a common environment, and monozygotic (MZ) twins are 100 percent genetically related, whereas dizygotic (DZ) twins and siblings share half their genes. The nature of the models and their fitting means that any gene-environment correlations are swept into the additive genetic component and that all unexplained residual variance is swept into the nonshared environment term (see, for example, Rutter’s chapter in this volume; Rutter and Silberg, 2002; Turkheimer and Waldron, 2000; Rutter et al., 2001; and for an accessible account of the basic models and ideas, Plomin et al., 1997 and Plomin, 1994). For a lively discussion of the use of such models in a demographic context, see the comments on Morgan and King (2001) by Capron and Vetta (2001) and Kohler (2001).
Since the majority of studies that have attempted partitions of variance in fertility using behavioral genetics models have been based on the information from the Danish Twin Registry (for an overview see Rodgers et al., 2001b), we shall begin by looking at these in some detail. These studies have all examined a variety of birth cohorts, including those of 1900-1923 (Kohler and Christensen, 2000), 1870-1910 and 1953-1964 (Kohler et al., 1999), 1953-59 (Rodgers et al., 2001a), 1953-1970 (this volume), and 1945-1968 (Kohler et al., 2001). No compelling rationale has been provided for these very different selections of birth cohorts, although data availability and selection issues undoubtedly play a part. The fertility-related outcomes considered have also varied, partly as result of censoring of