Accordingly, reliance has been placed on natural experiments that, in one way or another, can separate genetic and environmental influences. These tend to be thought of in terms of twin and adoptee research strategies. However, both actually involve several different research designs that differ in their patterns of advantages and disadvantages (Rutter et al., 1990, 1999, 2001). Thus, for example, twin strategies include comparisons of separated and nonseparated twins and also studies of the offspring of twins. Similarly, there are several different varieties of adoptee design.
The traditional twin design is predicated on the fact that monozygotic (MZ) pairs share all their segregating genes, whereas dizygotic (DZ) pairs, on average, share just half. The contrast between within-pair correlations or concordances in MZ pairs and DZ pairs can be used to infer the genetic effects on population variance for any particular trait. However, various additional assumptions have to be made. The most important, by far, of these is the equal environments assumption (EEA). This specifies that the environmental variance in MZ pairs is closely similar to that in DZ pairs insofar as the environments are associated with the trait in question. It should be noted that the assumption does not require that the environmental variances in the two sorts of the pair are the same but only that, insofar as they are different, they do not relate to the trait being studied. Thus, MZ pairs are much more likely to be dressed alike than are DZ pairs, but this is of no consequence for the vast majority of traits. However, there are some features that, although not directly related to mating behavior or procreation, are likely to influence it. For example, this would be likely to apply to both physical attractiveness and age at puberty. Nevertheless, this is unlikely to violate the EEA because the features are very strongly genetically influenced and therefore would be unlikely to lead to within-pair differences in most circumstances for most traits.
On the other hand, such features may be very informative because they could indicate the route by which genetic effects are mediated. Thus, it is not implausible that part of the genetic effects on population variance in fertility could derive from genetic effects on either physical attractiveness or age at puberty (although it is most unlikely that that is the only mode of genetic mediation). Most reviews by behavior geneticists (e.g., Kendler et al., 1994) have concluded that there is a lack of evidence for violation of the EEA. However, recent evidence has shown that this conclusion is mistaken (Rutter et al., 1999, 2001). It is necessary to note why the conclusion is mistaken as well as its implications for inferences about genetic effects.