increase substantially at later ages, subject to comparability in the number of loci affecting each age class, as well as showing roughly uniform mutation rates over age classes and similar magnitude effects of the individual alleles over age classes. Of course, if these ceteris paribus conditions are violated, the additive genetic variance may not increase with age. But that is a typical caveat for population genetics theory.
Essentially all other possible population mechanisms fall under the pleiotropic heading. When pleiotropic effects are in the same direction, such that one allele is beneficial for all affected life-history characters, then evolution is somewhat analogous to that in the case with nonpleiotropy. The balance between selection and mutation will be determined primarily by genetic effects at early ages, and the form of the equation determining allele frequency will be much more complicated, but qualitatively the results will be similar. One complication, however, might be a reduction in the genetic variance at later ages, if early, stronger, selective forces are determining allele-frequency equilibria. If allelic effects are opposed among life-history characters, then antagonistic pleiotropy arises. Under these conditions, selection will often be the dominating factor in the evolutionary outcome, favoring alleles that are beneficial at early ages but deleterious later (Williams, 1957; Charlesworth, 1980; Rose, 1985).
Overall, both mutation-selection balance and pleiotropy can act so that aging evolves. They can also maintain genetic variability for demographic characters. However, these two mechanisms have different implications for the nature of that genetic variability. With mutation-selection balance, genetic variance should increase with age (Charlesworth, 1990, 1994), all other things being equal, while with pleiotropy it may or may not. Antagonistic-pleiotropy models also require that there be antagonistic genetic effects between some life-history characters, which may (Rose, 1985) or may not (Houle, 1991) be manifest as negative genetic correlations between some life-history characters. Fortunately, these corollaries are testable in experimental systems.
The first thing to be said about the experimental population genetics of life history is that additive genetic variation is maintained for a variety of demographic characters in a variety of species (Rose, 1991; Roff, 1992). This is a corollary of both of the theoretical population genetic mechanisms described above. Beyond this point most of our experimental information comes from Drosophila. In that genus, at least, there are some cases in which genetic variances increase with age (e.g., Hughes and Charlesworth, 1994), and other cases where it does not (Rose and Charlesworth, 1981; Promislow et al., 1996). There are some cases where negative genetic correlations between life-history characters have been inferred (Rose and Charlesworth, 1981: Rose, 1984a: Luckinbill et