Fluctuating selection: the perpetual renewal of adaptation in variable environments

Abstract

Darwin insisted that evolutionary change occurs very slowly over long periods of time, and this gradualist view was accepted by his supporters and incorporated into the infinitesimal model of quantitative genetics developed by R. A. Fisher and others. It dominated the first century of evolutionary biology, but has been challenged in more recent years both by field surveys demonstrating strong selection in natural populations and by quantitative trait loci and genomic studies, indicating that adaptation is often attributable to mutations in a few genes. The prevalence of strong selection seems inconsistent, however, with the high heritability often observed in natural populations, and with the claim that the amount of morphological change in contemporary and fossil lineages is independent of elapsed time. I argue that these discrepancies are resolved by realistic accounts of environmental and evolutionary changes. First, the physical and biotic environment varies on all time-scales, leading to an indefinite increase in environmental variance over time. Secondly, the intensity and direction of natural selection are also likely to fluctuate over time, leading to an indefinite increase in phenotypic variance in any given evolving lineage. Finally, detailed long-term studies of selection in natural populations demonstrate that selection often changes in direction. I conclude that the traditional gradualist scheme of weak selection acting on polygenic variation should be supplemented by the view that adaptation is often based on oligogenic variation exposed to commonplace, strong, fluctuating natural selection.

Figure 1.

Simulated evolution in a multi-scale environment. Character state is governed by alleles with additive effects at 10 loci. Mutation creates a new allele with an exponentially distributed random effect with mean 1: the mutation rate is 0.001 per locus per generation. Environmental variation adds a random normal deviate to the character state of each zygote with mean 0 and variance σ²_E = 50. There is random gamete fusion and a single crossover with probability 0.5 at each meiosis. The population consists of 500 diploid individuals, of whom 100 are selected at random in each generation, each with probability proportional to its fitness. The width of the fitness function is ω² = 25. Over the 10⁵ generations shown in the figure, the realized population variables were: average distance of population mean from optimum = 1.13 σ_P; coefficient of variation = 0.13; average heritability = 0.63; z_E = 0.25; z_P = 0.46. The three panels show the population mean tracking the optimum over different time-scales, plotted at 25-generation intervals. (a) 1000 generations, (b) 10 000 generations, and (c) 100 000 generations. Thick line, optimum; thin line, mean.

Figure 2.

Fluctuating selection in Cepaea. Estimates of survival rates for brown (black squares) and non-brown (white squares) individuals from mark-recapture surveys. The bars are 1 s.e. (adapted from Cain et al. 1990). (a) Survival rates and (b) fluctuation of the selection coefficient.

Figure 3.

Fluctuating selection in natural populations. The standard deviation of the selection coefficient over time σ_s is plotted against its mean absolute value |s|. The solid line is σ_s = |s|.