On page 1095 of this issue, Reznick et al.1 describe how they have investigated one of the main factors that influence the evolution of an organism's lifespan. That factor is the risk of dying that a population faces as a result of environmental conditions (such as, in this case, predation). The study subjects are guppies, small tropical fish that are widely used in evolutionary studies, and the authors provide the first experimental support for the prediction that a higher environmental risk of mortality can select for inherently longer-lived organisms.

Guppies from the lower reaches of several rivers in Trinidad are subject to much higher rates of predation than those in the upper parts of the same rivers, where waterfalls block access by larger fish. In predator-free lab experiments, Reznick et al. found that guppies from the high-predation segments of two of the rivers lived up to 35% longer than those from low-predation segments of the same watercourse. In addition, the guppies from high-predation sites had a 40% longer reproductive span and reproduced at a higher rate. So a background of higher mortality under natural conditions has apparently led to the evolution of both a longer lifespan and a longer reproductive span. Their longest-lived fish, a female, is pictured in Figure 1.

Figure 1: Star survivor — the longest lived of the guppies studied by Reznick et al.1.
figure 1


The photo was taken shortly before her death at the age of 1,464 days.

A bit of history is required to see why this observation is surprising. Environmentally caused (‘extrinsic’) mortality has long been recognized as a key factor determining how natural selection moulds ‘intrinsic’ mortality — the death rate that a population would have under some standardized, generally benign, set of environmental conditions. Although evolution should favour lower intrinsic mortality (and a longer intrinsic lifespan) when all else is equal, many organisms face a trade-off between higher levels of reproduction or lower levels of intrinsic mortality. One of the main reasons that senescence occurs is because repair is costly: resources that are devoted to maintaining an organism are not available for reproduction. In the 1950s, Peter Medawar2 and George Williams3 pointed out that high extrinsic mortality could favour shorter intrinsic lifespan. Why, they reasoned, should an organism invest in costly repair that will probably only ensure that it is in prime physical condition when its life ends? Higher extrinsic mortality should favour low investment in repair, and thus a high intrinsic mortality and a short intrinsic lifespan.

But this reasoning didn't take account of two further factors. One is that higher extrinsic mortality also slows the rate of population growth, and more slowly growing populations are expected to evolve to have lower rates of intrinsic mortality and a longer lifespan4,5. The other is the interaction between extrinsic mortality factors and physiological repair or maintenance5,6. If predators can be evaded by fast, but not by slow prey, greater predation risk should select for greater maintenance of the body systems essential for fast movement. This higher level of repair would then prolong intrinsic lifespan.

Higher extrinsic mortality (more predators) could also have indirect effects that Medawar and Williams did not consider. For example, it reduces population size, which in turn increases the abundance of food or other resources. These changes may have their own effects on both population growth and the level of intrinsic mortality favoured by selection. Other complications arise if the mortality factor has a greater effect on some ages than on others5,6 — if, for example, predators prefer to capture larger, older prey. As a result of these complicating features, many types of mortality are expected to reduce intrinsic death rates at some ages while increasing them at others6. In any event, theory suggests that higher extrinsic mortality will produce evolutionary conditions that can either extend or shorten the intrinsic lifespan.

Given these complexities, the curious feature of previous observational7 and experimental work8 has been its support for the Medawar–Williams prediction. There have been exceptions9, if only suggestively so. But the almost unanimous evidence that high extrinsic mortality is associated with shorter lifespan is puzzling because there is no reason to believe that the conditions that produce the opposite outcome are rare in nature.

So it is reassuring that Reznick et al.1 found longer intrinsic lifespans in guppies from populations characterized by higher predation rates. The authors also looked at whether these evolutionary changes might be an indirect consequence of predator-caused deaths, such as the availability, in natural settings, of more food for the remaining guppies. Reznick and colleagues' study is unique in examining this effect. They found that food alone could not account for the difference in intrinsic mortalities seen in their experiments, but that having more food enhanced the lifespan-lengthening effect of a high-predation background.

There is no doubt that the guppies from high-predation sites have both longer intrinsic lifespans and longer reproductive spans. But is it valid to conclude that they have slower senescence? This is a more difficult question. A hypothetical population with no senescence (that is, no age-related decline in survival or reproduction) could still have a short lifespan if it had a high mortality rate that was independent of age. If guppies from high-predation sites begin their adult life with a lower rate of intrinsic mortality than those from low-predation environments, they could have the same rate of increase with age in their mortality rate, but would still have a longer lifespan. One could then argue that the two populations had identical rates of senescence. Some measures of the rate of change of intrinsic mortality with age suggest that senescence is delayed in guppies from high-predation sites.

However, senescence encompasses relationships between many different components of fitness and age, none of which can be adequately summarized by a single number: there are many potential measures of the rate of senescence, and conclusions about this rate depend on the measure chosen. It might be possible to make a case that guppies from high-predation environments are more robust, but age at a rate equal to or higher than that of low-predation guppies. Regardless of the mathematical measure used to quantify the rate of senescence, the work of Reznick et al. clearly shows that rates of senescence differ among the different components of fitness examined: survival, reproduction or swimming performance. The reasons for these differences are not yet understood.

It would be surprising if guppies were the only species for which an added risk of mortality lengthens intrinsic lifespan. Similar studies on other species will help us understand the underlying reasons why Medawar and Williams' predictions hold for some species and not for others. Such studies should follow Reznick and colleagues' lead in quantifying declines in several fitness components and studying the indirect ecological consequences of higher extrinsic mortality.