Developmental Instability as a Bet-Hedging Strategy

Abstract

In temporally varying environments selection will often act, at the expense of expected fitness in any given generation, so as to maximize geometric mean fitness across a number of generations (Gillespie 1977). Such (Slatkin 1974) may occur in two ways (Seger and Brockmann 1987). In conservative bet hedging a genotype minimizes the risk of a very low fitness bout by producing a narrowly unimodal distribution of safe trait values. In contrast, a genotype practising diversification bet hedging spreads the risk by producing traits with increased variance (see Philippi and Seger 1989 for a full discussion). A classical example of such diversification both in theoretical and empirical studies is that of seed germination behaviour (Cohen 1966, Janzen 1977, Venable and Lawlor 1980, Cooper and Kaplan 1982, Bulmer 1984, Ellner 1985, Le6n 1985, Venable 1985, Kalisz 1986, Bull 1987, Venable and Brown 1988, Biere 1991), but diversification or multiple strategies (Lloyd 1984) could be equally relevant to other processes such as insect diapause (Tuljapurkar and Istock 1993). Little is known, however, about how such within-genotype diversification might be produced. Here we propose that diversification bet hedging may be attained through a mechanism traditionally considered to be exclusively detrimental: developmental instability. We discuss the plausibility of this perspective, present testable hypotheses that emerge directly if bet hedging is achieved through instability of development, and suggest appropriate tests of these hypotheses. In this paper we discuss the mechanisms underlying diversification bet hedging although these mechanisms should apply to other, nonbet-hedging situations in which within-genotype trait variance is shown to be adaptive (see Geritz 1995 for density-dependent selection of seeds under spatial heterogeneity). There exists considerable confusion surrounding the use of terms related to the degree of variability inherent to developmental processes (Zakharov 1992). The relative fidelity of development to a program is known variously as developmental stability (Mather 1953, Waddington 1957, Parsons 1992, Moller 1995), homeorhesis (Waddington 1957), homeostasis (Lerner 1954, Orzack 1985, Parsons 1992, Moller 1995), buffering capacity (Van Valen 1962), and developmental canalization (Mather 1953, Thoday 1958, Waddington 1960, Levin 1988). The mechanisms which these terms describe all have the effect of decreasing trait variance. Developmental instability, environmental sensitivity (Jinks and Pooni 1988), and phenotypic plasticity (Bradshaw 1965, Via and Lande 1985, Schlichting 1986) are all means by which trait variance is increased. There is considerable overlap in the definitions of some of these terms, and their usage may vary by author (Zakharov 1992). Phenotypic variance expressed among individuals of a genotype is usually attributed to two general sources: plasticity and noise. Plasticity is measured as the phenotypic expression of genotypes across an environmental gradient, whereas developmental noise is assumed to result from random errors of development, but in fact is variance resulting from all sources unknown. Developmental noise, then, may include plasticity in response to environmental variables that have not been identified (Bradshaw 1965). There is ample evidence for the existence of genetic variation for plasticity (Bradshaw 1965, Perkins and Jinks 1971, Schlichting and Levin 1986, Scheiner et al. 1991, Oyama 1994). Although phenotypic plasticity is the variable expression of a genotype under differing environmental circumstances, it is quite possible to have well-canalized reaction norms, and the degree of plasticity is a genotype-level property. There is no reason to suppose that, like plasticity, developmental noise could not be maintained at some optimal level by stabilizing selection. The genetic basis of developmental stability is not yet well established (see Clarke 1993 for a discussion of competing hypotheses). Developmental instability may be assessed through the occurrence of phenodeviants or, in (normally) bilaterally symmetrical organisms, as the fluctuating asymmetry (FA) of paired characters