I have been describing new empirical results and theoretical departures that seem to me to be shifting the scientific ambience of ideas in the biology of longevity away from their long-standing emphasis on limits and tradeoffs. None of the new developments are so tightly connected with demographic processes in humans that they compel the adoption of different models or forecasts. But if further results and theoretical developments in biology in the next few years reinforce the recent developments described in this volume, then I think that demographers will find themselves far less at ease with models that posit strongly diminishing returns in the near future in progress against old-age mortality.

Many chapters in this book point directly toward the future, to kinds of research that should come into their own in the next few years. Chief among these is quantitative trait locus (QTL) analysis of the kind discussed in Chapter 7 by Thomas Johnson and David Shook. We should learn soon how easy it is, in an organism like the elegantly named nematode worm C. elegans, to find genes or portions of the genome that do have specific effects on old-age survival. It is tempting for demographers who model hazard rates as functions of age to imagine that genes ''know about" ages beyond development and reproduction. In caricature, it is tempting to imagine a set of genes influencing the probability of dying between 50 and 55 and another set influencing the probability of dying between 80 and 85. Results that bolster or undermine this picture will shape the next generation of demographic models.

Tests of predictions from the evolutionary theory of senescence are beginning to capitalize on the quantitative specificity permitted by QTL analysis. We should soon be learning, for some laboratory organisms, whether it is commonplace for portions of the genome that correlate positively with net reproduction to correlate negatively with older-age survival rates, as antagonistic pleiotropy posits. We should be learning systematically about the genetic variance in portions of the genome correlated with older-age hazard rates, and whether the variance increases with age in accordance with mutation-accumulation theory.

QTL analysis, like most of genetics, focuses on polymorphic genes, genes for which two or more forms or alleles remain present in the population despite the ages-long operation of natural selection. However, most genes responsible for structures that promote longevity are doubtless not polymorphic. They have been selected and become fixed in the population. This is less of a worry for geneticists, who are interested in the persistence of genetic variation in its own right, than for demographers. Inducing mutations in genes that have become fixed may sometimes allow QTL analysis to circumvent this limitation, and any such results will be particularly intriguing to demographers. Necessarily, our speculations about fixed genetic components of programmed aging will continue