Conflicting Selection Pressures and Life History Trade-Offs

Conflicting selection pressures and life history trade-offs

DOLPH SCHLUTER1, TREVOR D. PRICE2 AND LOCKE ROWE1

1 Departmentof Zoology, University of British Columbia, Vancouver,British Columbia, Canada V6T 1Z4 2Biology Department 0-116, University of California, San Diego, La Jolla, California 92109, U.S.A.

SUMMARY

We review studies of natural selection in wild populations in which selection has acted in opposite directions at different stages of the life history. For example, the phenotype with highest probability of survival may have the lowest reproductive success. We discuss two important implications of these findings. First, measurements of opposing selection confirm that evolution of traits is governed by a balance of conflicting fitness advantages. Second, studies of opposing selection are informative about mechanisms underlying life history trade-offs. We outline difficulties in measuring opposing selection, particularly the problem that patterns of selection may be masked by the positive effects of nutrition on size of metric traits and fitness components. We discuss some solutions to these problems, and present a statistical technique to help disentangle direct selection from nutritional effects. Finally, we show how fluctuations in selection pressures lead to norms of reaction for life history traits in the absence of developmental plasticity.

fitness. We define the of fitness to be 1. INTRODUCTION components synonymous with life history traits; they include Natural selection is common in wild populations (see survival, fecundity, mating success and age at maturity. Endler 1986; Manly 1985). Here we draw attention to More formally, a life history trait (fitness component) studies in which selection has been discovered to act in is any character correlated with total fitness when all opposite directions on a trait at different life history other traits are held constant. For example, higher stages. For example, on Mandarte Island, British survival is always advantageous, all else being equal, Columbia, immature female song sparrows having and we therefore include survival in the set of life long beaks survived better over their first winter than history traits. Beak length, however, should not be shorter-beaked individuals, but short-beaked adult correlated with total fitness when all other traits, females had higher reproductive success (Schluter & including fecundity and survival, are held constant. Smith 1986). We refer to these other phenotypic traits, including We identify two broad implications of such dis- morphological, behavioural and physiological traits, as coveries. First, the presence of opposing selection 'metric traits', after Falconer (1989). pressures directly confirms the common perception Direct demonstrations of opposing selection pres- that phenotypic evolution is governed by a balance sures at different life history stages are summarized in between conflicting fitness advantages. Second, the table 1. The list is confined to observations from wild results show how the value of a metric trait which populations, and includes all examples that we found maximizes one fitness component (e.g. survival) may in a survey of the literature. Bird studies outnumber differ greatly from the value maximizing another (e.g. others, probably because of their determinate growth adult fecundity). Fitness components comprise the set and the relative ease with which they may be measured, of life history traits, and studies of opposing selection on marked and subsequently observed in the wild. The list metric traits are therefore informative about pheno- includes two studies of germination date in annual typic mechanisms underlying life history trade-offs. We plants. This is a metric trait rather than a fitness suggest that studies of selection in wild populations can component, under the definition given above, because complement traditional approaches to the study of life it does not refer to age at first breeding (which is a history evolution. fitness component) but to time within a season (during which all individuals breed). In this case, germination time no intrinsic fitness We also 2. CONFLICTING NATURAL SELECTION provides advantage. include two of selection on PRESSURES ON METRIC TRAITS examples enzyme morphs. Although these are arguably not phenotypic traits, we We define natural selection as non-random variation presume that their fitness consequences derive from among phenotypes in one or more components of their association with specific morphological, physio-

Proc.R. Soc.Lond. B (1991) 246, 11-17 11 Printedin GreatBritain 12 D. Schluter and others Conflictingselection pressures

Table 1. Opposing selectionpressures identified on metric traits in the wild (Life history traits refer to the fitness components whose performance is enhanced (1) and reduced (2) by an increase in the metric trait. Order of life history traits is arbitrary for polymorphic traits. Mating success refersto the rate at which individuals obtain a mate.) organism metric trait life history trait 1 life history trait 2 reference Darwin's finch" body size adult survival age at 1st reprod. Price (1984) Darwin's finch" body size adult survival juvenile survival Price & Grant (1984) song sparrowa beak length juvenile survival fecundity Schluter & Smith (1986) Darwin's fincha plumage colour fecundity survival Grant (1990) swallowa b tail-length mating success future fecundity Moller (1988, 1989a) herring gull body-size survival (non-breeding) survival (breeding) Monaghan & Metcalfe (1986) Arctic skuaa plumage colour age at 1st reprod. mating success O'Donald (1983) house sparrow" plumage badge size mating success adult survival Moller (1989b) red deera enzyme morph juvenile survival age at 1st reprod. Pemberton et al. (1991) dragonfly hindwing length mating success survival Koenig & Albano (1987) damselfly body size adult survival mating success Anholt (1991) barnaclec shell form survival fecundity Lively (1986a,b) snail shell size fecundity survival Bantock & Bayley (1973) wild oata enzyme morph survival fecundity Clegg & Allard (1973) blue-eyed Mary germination date winter survival spring survival Kalisz (1986) Leavenworthia germination date fecundity survival Baskin & Baskin (1972) jewelweed cotyledon area fecundity survival Stewart & Schoen (1987) a Traits known to be heritable. b Opposing selection pressures identified only by experiment. ? Known to show adaptive phenotypic plasticity of metric trait. logical or behavioural (i.e. metric) traits as yet wing length in the scorpionfly (Thornhill 1983); tarsus unmeasured. length in a flycatcher (Alatalo & Lundberg 1986); Although many studies have observed selection in horn length in a fungus beetle (Conner 1988), and leaf nature (Endler 1986), few have reported selection over length in a perennial grass (Scheiner 1989). Subsequent more than one life history stage. When directional analyses (Arnold & Wade 1984; Alatalo et al. 1990) selection is present on a trait at one life stage, three have suggested that masking effects of nutrition may patterns are possible at any other stage: no directional be important in some of these cases. Both opposing selection, selection in the opposite direction to that seen selection and complementary selection were present in at the first life stage (i.e. opposing selection), and two examples (Price 1984; Kalisz 1986; cf. table 1). selection in the same direction to that at the first stage Thus, as a first estimate, opposing selection appears to (complementary selection). Examples of all three be more frequent than complementary selection. patterns are known, occasionally from the same field Variability in selection intensities is the hallmark of study (e.g. Kalisz 1986). However, unknown biases in virtually all field studies of natural selection, and the reporting, sampling error and technical difficulties in studies in table 1 were no exception. Two types of measuring selection make it difficult to assess the variability were shown. First, traits subject to strong relative frequency of the three types. For example, selection in one year (or place) experienced weak or no biases may arise if some components of fitness (e.g. selection in another (Kalisz 1986; Schluter & Smith fecundity) are more difficult to measure than other 1986; Stewart & Schoen 1987; Anholt 1991). Second, components (e.g. mating success). Another major the direction of the association between a given trait problem is the tendency of opposing selection to be and a fitness component fluctuated between years or masked by nutritional effects. For example, good locations (Stewart & Schoen 1987). nutrition may enhance both survival and fecundity as The conflict between selection pressures on metric well as size of body parts, giving the false impression traits can be summarized in a simplified path diagram that size itself is under complementary directional (figure 1). We focus on survival and fecundity for selection. This problem is discussed more fully below purposes of illustration only; the same idea can be and in Appendix 1. Note also that if a trait is associated applied to other fitness components. fi and f2 are with several fitness components (three, for example), directional selection coefficients from separate linear then only a subset of the associations can be in opposing regressions of survival and fecundity on the phenotypic directions (two); the rest must be complementary. (metric) trait (Lande & Arnold 1983). When the two Despite these problems, opposing selection appears selection pressures on the trait conflict, these co- to be relatively common in nature (table 1). We efficients are of opposite sign (f1 x f2 < 0). An example located only five examples in which selection was of selection giving rise to this conflict is shown in the studied over two or more life stages, and comple- top half of figure 2. mentary selection, but no opposing selection, was We add a third path to figure 1 to incorporate the observed: body size in an Anolis lizard (Trivers 1976); possibility that each increment to fecundity causes a

Proc. R. Soc. Lond.B (1991) Conflictingselection pressures D. Schluter and others 13

approach to measure selection on a specific trait is to 33 ~ -- - - - manipulate it experimentally, independently of the rest survival ---- -fecundity of the phenotype, as was done by Moller (1988, 1989 a) in his study of tail length in the swallow (table 1), although this is not always feasible. The task of of selection is 0 /3 identifying targets particularly difficult when size in metric traits is correlated with the overall condition or nutritional health of the individual, and this nutrition also elevates both survival and fecundity directly (Price et al. 1988; Price & Liou 1989; Alatalo et al. 1990). This mechanism could be in 1 the metrictrait represented figure by addition of a new trait ('nutrition') directly connected Figure 1. Path diagram summarizing the main causal links to all three nodes arrows with between and a metric trait. original by positive survival, fecundity phenotypic /, coefficients. In this case the correlation between each and are the linear selection intensities on the metric trait, f/2 fitness and the metric trait is a function of which are of opposite sign when the survival and fecundity component both the direct selection intensities and nutrition. advantages of the trait conflict (/J, x ,2 < 0). The dashed line fl Nutrition is a kind of correlated trait because it and &/ represent the possible reduction in survival stemming special directly from an increase in fecundity. is strongly affected by local environment, and although it may affect the size of other heritable metric traits, nutrition may not itself be heritable. reduction in fitness components at later life stages. For Most crucially, because it positively affects the example, brood enlargement in blue tits may reduce metric trait as well as the two fitness components, the future survival of parents caring for the young (Nur nutrition may mask opposing selection pressures. An 1984), and induced fruiting in an orchid reduces excellent example is provided by the swallow. Moller fecundity in subsequent years (Primack & Hall 1990). (1988, 1989 a, 1991) showed that tail length in the male These effects are mediated through physiological, swallow is positively correlated with both survival and morphological and behavioural changes that affect the mating success, suggesting that the tail is under fitness components in opposite ways (for example, persistent selection pressure to increase. However, this induced fruiting in orchids led to a reduction in growth pattern proved to be spurious when tail length was and hence size relative to control plants), but they are manipulated independently of nutritional state of the consequences of the increment to fecundity itself. The male: long tails enhanced mating success (Moller full correlation between survival and a metric trait 1988) but may have reduced survival and subsequent The true of selection (figure 1) depends on I, and /2 x f3 (and also on the fecundity (Moller 1989 a). pattern effects of correlated metric traits on survival and on tail length in unmanipulated birds was hidden by fecundity, whose paths of influence are not illustrated, the fact that males in better health or nutrition grow for simplicity). This third path is not always applicable; longer tails as well as survive and reproduce better for example IN= 0 when pre-reproductive survival is than males in poor nutrition. Nutrition-dependent of interest. expression of metric traits and fitness components is probably common (Price et al. 1988; Zeh & Zeh 1988; Price & Liou 1989; Alatalo et al. 1990; Rowe & The be 3. PROBLEMS WHEN METRIC TRAITS ARE Ludwig 1991). phenomenon may especially in with indeterminate CORRELATED prevalent organisms growth (for example, insects and plants), in which most of the Before discussing the implications of opposing selec- variation among individuals in size-related traits may tion, we first mention a difficulty encountered in be nutrition based (e.g. Roach & Wulff 1987). attempts to precisely determine patterns of selection on It is known that the effects of nutrition can mask traits. The problem is that the true target of selection trade-offs between life history traits (Lande 1982; may not be among the metric traits measured (Lande Reznick 1985; van Noordwijk & de Jong 1986; & Arnold 1983; Mitchell-Olds & Shaw 1987; Crespi Partridge & Harvey 1988; Stearns 1989). However, & Bookstein 1989). This problem is not unique to nutrition's equally confounding effects on measures of studies of opposing selection, but special difficulties selection in the wild are less widely appreciated. The worth mentioning arise in this case. For example, in swallow example shows that conflicting selection any given breeding season, male Arctic skuas with dark pressures are particularly prone to masking effects, and plumage obtain a mate earlier in the year than males lead us to suggest that the frequency of cases of with pale plumage (table 1; O'Donald 1983). How- opposing selection on metric traits in nature is ever, this apparent selection on plumage coloration underestimated by the number of studies recorded thus may stem indirectly from a tendency in females to mate far. Moller's (1988, 1989a) experiments reveal the with older males (Baker & Parker 1979). power of phenotypic manipulations to tease apart This problem of unmeasured, correlated characters selection on specific traits from the effect of nutrition is ubiquitous in observational studies of selection, and and other correlated characters. A second, indirect, careful choice of traits on the basis of their presumed method that may be used when experiments are not function is needed to minimize it. The most promising possible is discussed in Appendix 1.

Proc. R. Soc. Lond. B (1991) 14 D. Schluter and others Conflictingselection pressures

evolution. We doubt that there is a for 4. SELECTION AND THE EVOLUTION OF gene directly survival other life individuals METRIC TRAITS (or history trait); rather, in a population differ in survival either because of The view that morphological and other phenotypic chance, or because they differ in one or more aspects of traits are under conflicting selection pressures is a morphology, physiology and behaviour, metric traits central concept in evolutionary ecology (Krebs & that may have a partly genetic basis. The measurement Davies 1984; Pianka 1988; Ricklefs 1990). Direct of selection on underlying metric traits is therefore a measurements of selection provide powerful evidence useful way to identify trade-offs between life history of this principle (table 1). They show that the traits: the demonstration that even one metric trait phenotypic values at which survival and fecundity (or cannot be optimized simultaneously for survival and other fitness components) are maximized often lie to fecundity (table 1) is equivalent to a demonstration either side of the population mean. Traits are pulled in that survival and fecundity themselves cannot be alternate directions at different stages of the life history, simultaneously maximized. and the point at which these conflicts are balanced The model of opposing selection (figure 2) is also a may lie in the vicinity of the observed population useful basis for evaluating and comparing other mean. methods to identify life history trade-offs. These include Phenotypic selection pressures are relevant to the the measurement of genetic correlations among life evolutionary dynamics of natural populations when history traits, and the manipulation of fecundity variation in the metric trait has a genetic component. (Reznick 1985; Partridge & Harvey 1988). Below, we Traits under opposing selection pressures are known to briefly note the link between the latter two methods be heritable in about half of the cases (table 1); a and patterns of opposing selection on underlying metric genetic component remains to be verified in the other traits. half. Populations may thus appear to be evolutionarily The method of genetic correlations considers that life stable over the long term, but the finer measurements history traits have an additive genetic component, and summarized in table 1 reveal that the mean values of asks whether individuals with a high genetic value for traits lurch and wobble continuously over time. Studies survival also have a low genetic value for fecundity. by Gibbs & Grant (1987) on Galapagos finches show This information is preferable to a measure of the that selection pressures on heritable morphological phenotypic correlation between survival and fecundity, traits oscillate between generations. The present survey as a negative genetic correlation can be masked by the of studies shows that similar oscillations occur between positive effects of nutrition (see above). Our approach life stages within a generation. is closely related to this method because, as has been argued elsewhere (Price & Schluter 1991), underlying 5. CONFLICTING SELECTION PRESSURES heritable metric traits impart the genetic component to AND LIFE HISTORY TRADE-OFFS life history traits. Hence, selection on metric traits, and the heritability and genetic correlations between these Fitness components are with life synonymous history traits, are the source of genetic correlations between traits, and selection pressures on heritable metric traits fitness components (Price & Schluter 1991). therefore identify mechanisms underlying life history All else being equal, opposing survival and fecundity selection on a heritable metric trait such as beak length environment1 (figures 1 and 2) yields a negative genetic correlation survival between survival and fecundity (Schluter & Smith 1986; Price & Schluter 1991). However, the actual genetic correlation will be determined by the selection pressures accumulated over all heritable metric traits, with less predictable results. For example, if parasite resistance is heritable and has a strong positive effect on both survival and fecundity, then the net genetic correlation between the fitness components may be positive despite the fact that beak length is under selection environment2 conflicting pressures. Hence, although many survival metric traits in the organism may be subject to opposing selection, with the result that neither survival nor fecundity is maximized, the genetic correlation will not necessarily reveal it. This problem has been modelled by van Noordwijk & de Jong (1986) and Houle (1991). Experimental manipulation of fecundity, such as metrictrait adding eggs or young to the nest of individual birds and noting survival of the for Figure 2. Hypothetical example of selection on a metric trait subsequent parents (see, Nur in two environments or time periods. The shaded curve example, 1984; Gustafsson & Sutherland 1988; shows the distribution of the trait in the population, assumed Pettifor et al. 1988), is another common approach to to be the same in the two environments. The top curves are investigating life history trade-offs. This approach fitness functions for survival and fecundity. illuminates only part of the trade-off between survival

Proc. R. Soc. Lond. B (1991) Conflictingselection pressures D. Schluter and others 15 and fecundity, as we show in figure 1: the experiment highly dependent on the environment in which they can determine whether or not f/3 is negative, but the are measured (Stearns 1989; Stearns et al. 1991). Such manipulation bypasses consideration of the changes in genotype-environment interactions in life history traits organism design that would otherwise be necessary to are generally thought to result from developmental produce that extra nestling, and the possibly negative plasticity in the organism's underlying phenotype. consequences of that design to survival (P/l in figure 1). Consideration of fluctuating selection on heritable Reznick (1985) has criticized the fecundity manipu- metric traits reveals an alternative mechanism, where- lation as 'phenotypic', arguing that only genetic by genetic correlations between life history traits may trade-offs are relevant to life history evolution. We even change sign in the absence of developmental demur, and consider the fecundity manipulation as plasticity. analogous to a selection experiment, much like Moller's We illustrate this mechanism by comparing the (1988, 1989a) on tail length of the swallow, the chief selection pressures acting on a heritable trait in utility of which lies in its ability to identify the fitness alternative environments (figure 2; e.g. different consequences of variation in specific traits. Individual locations or years of a field study). Survival and survival depends on underlying phenotype by all paths fecundity are presumed to depend on a metric trait in in figure 1 (/, and /2 x/3, plus paths through other different ways. For example, a large beak in a finch metric traits not illustrated); the genetic correlation may be advantageous to survival in the non-breeding between survival and fecundity also depends on these season because it helps crack large seeds, but a small same paths. Only by manipulating fecundity inde- beak may be best in the breeding season when parents pendently of the underlying metric trait are we able forage for insects to feed to their young. Fluctuations in to measure /p3 alone. However, the manipulative the sizes of seeds and insects may lead to different approach measures only one part of the survival- patterns of selection in different years (figure 2). For fecundity trade-off. Perhaps most of the burden of simplicity, we assume that the metric trait does not reproduction lies instead in the traits that organisms itself change in response to these environments. The must carry around with them in order to reach the mean survival and fecundity of genotypes of alternate point of reproducing successfully. If so, then a beak size can then be calculated from the fitness conceivably large portion of the survival cost associated functions in figure 2, producing the results shown in with reproduction (/,) may be paid well before figure 3. In this example, the genetic correlation breeding begins. changes sign simply as a consequence of an altered pattern of selection. Moreover, all the complexities of interpreting the shifting pattern of genetic correlations 6. SELECTION, vanish when selection pressures on the underlying trait GENOTYPE-ENVIRONMENT are understood. INTERACTION AND REACTION NORMS The few studies of selection on metric traits that have been across environments or The alternative phenotypes that a given genotype replicated years (Kalisz Schluter & Smith Gibbs & Grant produces across a range of environments are labelled its 1986; 1986; 1987; Scheiner Anholt that 'norm of reaction' (Stearns 1989). Reaction norms are 1989; 1991) suggest fluctuating selection are the rule in nature. it ubiquitous in life history traits, and as a result genetic pressures Hence, may variances and correlations between these traits are be worthwhile to partition the causes of genotype- environment interaction in life history traits into components, one attributable to developmental plasticity in the organism and a second to variation in selection pressures. In birds there may be restricted in in which case environment 2 plasticity morphology, fluctuating selection on these traits may be the main cause of environment 1 variation in life history trade-offs. In contrast, phenotypic plasticity probably dominates variation in behav- a ioural traits, in which case studies of selection alone will ci \ _ not the causes of variation in measures of life U2 identify P()sitive genetic " \ history. An outstanding example of a joint approach, correlation N in which variable selection and developmental plas- negative genetic ticity were both measured and found to be large, is that correlation of shell form in the acorn barnacle (Lively 1986a, b). Levels of genetic variation in this trait are not yet known. survival Figure 3. Genotype-environment interaction in life history traits arising from fluctuating selection on a metric trait 7. CONCLUSION (figure 2). Solid lines give mean survival and fecundity of It is commonly recognized that heritable phenotypic different genotypes for the metric trait in two environments. traits mediate life evolution & The slope of the line shows the sign of the genetic correlation history (Partridge Stearns It is less between survival and fecundity. A dashed line connects the Harvey 1988; 1989). commonly same genotype in the two environments, and describes its recognized that this mediation is equivalent to natural norm of reaction. selection on morphology, physiology and behaviour,

Proc. R. Soc. Lond. B (1991) 16 D. Schluter and others Conflictingselection pressures

and that the growing number of demonstrations of measurement z'). Let H refer to the slope of this conflicting selection pressures (table 1) provide ex- regression. Second, compute the deviation d of each amples of mechanisms in nature for life history trade- individual measurement z from the regression line: offs. not all mechanisms of life Clearly, history d- z-Hz'. (3) evolution will be discovered by field studies of opposing out a of (o-1 on z selection: many phenotypic traits under selection are Finally, carry multiple regression and d. It can be shown that the extremely plastic, and others are difficult to observe partial regression coefficient for z is the of directional and measure. Nevertheless, if life history trade-offs are /f, intensity selection. The coefficient for d is common then opposing selection pressures on metric partial regression Thus the terms and 8 of the selection model traits are also common, and there should be much to 8o'/ol'a. f are and their to fitness is now gain from studying them. We suggest that the link separated, significance evaluated. between selection studies and the study of life histories The above result seem counterintuitive because will grow as further field studies of selection are may conducted. d is not nutrition itself, but like z is a sum of genetic, environmental and nutrition components. However, We thank P. R. Grant, L. Gustafsson, D. Reznick and an when z is held constant, residual covariation between d reviewer for their comments. The authors anonymous helpful and fitness can arise only through variation in the were supported by the NSERC (Canada), and by the nutrition part, and the rate at which fitness increases University of California Education Abroad Program. with d depends on S and the fraction of the variance in d attributable to nutrition. Hence, the partial re- APPENDIX 1. DISENTANGLING DIRECT gression coefficient for d evaluates the magnitude of 6 SELECTION FROM EFFECTS OF (scaled by o/of-) disentangled from selection on z. NUTRITION Similarly, when d is statistically held constant, the rate at which fitness changes with variation in z depends The problem of nutrition masking patterns of only on the magnitude of /, the selection intensity. natural selection is probably widespread. Experimental The above partial regression coefficients were de- manipulation is one way in which effects of nutrition termined P-1s, where P is the matrix of can be controlled for, but this will not be by calculating always variances and covariances between z and and s is a Here we show how one d, possible. may statistically vector whose two entries are Cov and Cov natural selection on a trait z from (w, z) (w, d). distinguish directly We made use of the substitutions: the effect of nutrition n, when nutrition cannot be following measured. The gist of the method was suggested and Coy (w, z) = 80-2 + 6a2z ni (4) Alatalo et al. in = applied by (1990) their illuminating CovCouv(z,d) (), d) =6'v %.2?+6'O.2n. d and (5) 'bubble plot' (their figure 1), and requires measure- Cov (z, d) = 82. (6) ments on relatives. It has also been used with some This derivation makes a number of and success on song sparrows (W. M. Hochachka, un- assumptions, hence results from its should be published results). Consider the following regression application interpreted model of natural selection: cautiously. In particular, we assume that selection is linear (equation 1); this may often be achieved with a o - I = f/z + 6n + random error, (1) suitable transformation of the data. We also assume that of variation are additive and un- where o is relative fitness, wo= Wl/w; W is the fitness components correlated that nutrition is component (e.g. survival, measured as a binary (0,1) (equation 2), non-heritable, and that environmental and maternal sources of variable, or reproductive success, measured as number resemblance relatives are absent that the of offspring), and #w is the mean of W. ,/ is the intensity among (i.e. resemblance is The method of directional selection on the trait z, and 6 is the effect entirely genetically based). of nutrition on relative fitness. The trait z is assumed to will be most successful when the heritability of the trait is moderate or consist of three independent components: high. = z x++e; (2) REFERENCES an additive term x, a n attributable genetic component Alatalo, R. V. & Lundberg, A. 1986 Heritability and to and a random error term e. z nutrition, Therefore, selection on tarsus length in the pied flycatcher (Ficedula and n are correlated with Cov (z, n)= ', where cr2 hypoleuca).Evolution 40, 574-583. refers to variance (Price et al. 1988). Alatalo, R. V., Gustafsson, L. & Lundberg, A. 1990 A univariate regression of ( on z would have the Phenotypic selection on heritable size traits: environmental slope 8+?6'/o2 (cf. equation 1). Hence, if z has a variance and genetic response. Am. Nat. 135, 464-471. nutrition component, its covariance with fitness may Anholt, B. R. 1991 Measurement of selection on a population of damselfies with a represent the effect of nutrition (8) rather than direct manipulated phenotype. Evolution 1091-1106. selection (/,) on the trait z. If n is measurable, its effect 45, & 1984 On the be from natural selection on z with a Arnold, S.J. Wade, M.J. measurement of may disentangled natural and sexual selection: Evolution of o on z and n & Arnold applications. 38, multiple regression (Lande 720-734. The be taken when n is not 1983). following steps may Baker, R. R. & Parker, G. A. 1979 The evolution of bird measurable, provided that z is heritable and n is not. coloration. Phil. Trans.R. Soc. Lond.B 287, 63-130. First, carry out a regression ofz on the same trait z' in Bantock, C. R. & Bayley, J. A. 1973 Visual selection for a relative (e.g. offspring measurement z on midparent shell size in Cepaea(Held.). J. Anim.Ecol. 42, 247-261.

Proc. R. Soc. Lond.B (1991) Conflictingselection pressures D. Schluter and others 17

Baskin,J. M. & Baskin, C. C. 1972 Influence of germination Nur, N. 1984 The consequences of brood size for breeding date on survival and seed production in a natural blue tits. I. Adult survival, weight change and the cost of population of Leavenworthiastylosa. Am. Midl. Nat. 88, reproduction. J. Anim.Ecol. 53, 479-496. 318-323. O'Donald, P. 1983 The ArcticSkua. Cambridge University Clegg, M. T. & Allard, R. W. 1973 Viability versus Press. fecundity selection in the slender wild oat, Avenabarbata L. Partridge, L. & Harvey, P. 1988 The ecological context of Science,Wash. 181, 667-668. life-history evolution. Science,Wash. 241, 1449-1455. Conner, J. 1988 Field measurements of natural and sexual Pemberton, J. M., Albon, S. D., Guinness, F. E. & Clutton- selection in the fungus beetle, Bolitotheruscornutus. Evolution Brock, T. H. 1991 Countervailing selection in different 42, 736-749. fitness components in female red deer. Evolution45, 93-103. Crespi, B.J. & Bookstein, F. L. 1989 A path-analytic Pettifor, R. A., Perrins, C. M. & McLeery, R. H. 1988 approach for the measurement of selection on morphology. Individual optimization of clutch size in great tits. Nature, Evolution43, 18-28. Lond.336, 160-162. Endler, J. A. 1986 Natural selectionin the wild. Princeton Pianka, E. R. 1988 Evolutionary ecology, 4th edn. New York: University Press. Harper & Row. Falconer, D. 1989 Introductionto quantitativegenetics, 3rd edn. Price, T. D. 1984 The evolution of sexual size dimorphism New York: Longman. in a population of Darwin's Finches. Am. Nat. 123, Gibbs, H. L. & Grant, P. R. 1987 Oscillating selection on 500-518. Darwin's finches. Nature,Lond. 327, 511-513. Price, T. D. & Grant, P. R. 1984 Life history traits and Grant, B. R. 1990 The significance of subadult plumage in natural selection for small body size in a population of Darwin's finches, Geospizafortis.Behav. Ecol. 1, 161-170. Darwin's finches. Evolution38, 483-494. Gustafsson, L. & Sutherland, W.J. 1988 The costs of Price, T. D., Kirkpatrick, M. & Arnold, S.J. 1988 Directional selection and the evolution of date in reproduction in the collared flycatcher Ficedulaalbicollis. breeding Nature,Lond. 335, 813-815. birds. Science,Wash. 240, 798-799. T. D. L. 1989 Selection on in Houle, D. 1991 Genetic covariance of fitness correlates: Price, & Liou, clutch size birds. Am. Nat. 950-959. what genetic correlations are made of and why it matters. 134, T. D. D. On the Evolution45, 630-648. Price, & Schluter, 1991 low heritability of life traits. Evolution 833-861. Kalisz, S. 1986 Variable selection on the timing of history 45, R. B. & P. 1990 Costs of in the germination in Collinsiaverna (Scrophulariaceae). Evolution Primack, Hall, reproduction 40, 479-491. pink lady's slipper orchid: a four year experimental study. Am. Nat. 131, 348-359. Koenig, W. D. & Albano, S. S. 1987 Lifetime reproductive D. N. 1985 Costs of an evaluation of success, selection and the opportunity for selection in the Reznick, reproduction: the evidence. Oikos 257-267. white tailed skimmer Plathemis lydia (Odonata: Libel- empirical 219, R. E. 1990 3rd edn. New York: Freeman. Evolution41, 22-36. Ricklefs, Ecology, lulidae). D. A. & R. D. 1986 Maternal effects in Krebs, R. & Davies, N. B. 1984 Behaviouralecology, Roach, Wulff, J. (ed.) A. Rev. Ecol. 2nd edn. Oxford: Blackwell Scientific. plants. Syst. 18, 209-235. L. & D. 1991 Size and of Lande, R. 1982 A of life Rowe, Ludwig, timing quantitative genetic theory history in life time constraints and evolution. Ecology63, 607-615. metamorphosis complex cycles: variation. Ecology72, 413-427. Lande, R. & Arnold, S.J. 1983 The measurement of Scheiner, S. M. 1989 Variable selection a successional selection on correlated characters. Evolution37, 1210-1226. along Evolution 548-562. C. M. 1986a Predator-induced shell in gradient. 43, Lively, dimorphism D. & N. M. 1986 Natural selection on the acorn barnacle Chthamalus Evolution Schluter, Smith, J. anisopoma. 40, beak and size in the Evolution 323-342. body song sparrow. 40, 221-231. C. M. 1986b life Lively, Competition, comparative histories, S. C. 1989 Trade-offs in evolution. and the maintenance of shell in a barnacle. Stearns, life-history dimorphism Funct.Ecol. 259-268. 858-864. 3, Ecology67, S. C., G. & B. 1991 The effects B. F. 1985 Thestatistics naturalselection on animal Stearns, deJong, Newman, Manly, J. of of on correlations. Trends New York: & Hall. phenotypic plasticity genetic populations. Chapman Ecol. Evol. 6, 122-126. Mitchell-Olds, T. & Shaw, R. G. 1987 Regression analysis Stewart, S. C. & Schoen, D. J. 1987 Patterns of phenotypic of natural selection: statistical inference and biological viability and fecundity selection in a natural population of Evolution41, 1149-1161. interpretation. Impatienspallida. Evolution41, 1290-1301. Moller, A. P. 1988 Female choice selects for male sexual tail Thornhill, R. 1983 Cryptic female choice and its implica- ornaments in the swallow. Nature,Lond. 332, monogamous tions in the scorpionfly Harpobittacusnigriceps. Am. Nat. 122, 640-642. 765-788. A. P. 1989a costs of male tail ornaments in Moller, Viability Trivers, R. L. 1976 Sexual selection and resource-accruing a swallow. Lond. 132-135. Nature, 339, abilities in Anolisgarmani. Evolution 30, 253-269. A. P. 1989b Natural and sexual selection on a Moller, van Noordwijk, A.J. & de Jong, G. 1986 Acquisition and plumage signal of status and on morphology in house allocation of resources: their influence on variation in life- Passerdomesticus. J. evol. Biol. 125-140. sparrows, 2, history tactics. Am. Nat. 128, 137-142. A. P. 1991 is related to a Moller, Viability positively degree Zeh, D.W. & Zeh, J.A. 1988 Condition-dependent sex of ornamentation in male swallows. Proc. R. Soc. Lond.B ornaments and field tests of sexual-selection theory. Am. 243, 145-148. Nat. 132, 454-459. Monaghan, P. & Metcalfe, N. B. 1986 On being the right size: natural selection and size in the body herring gull. Submittedby P. R. Grant;received 17 May1991; revised 19 July 1991; Evolution40, 1096-1099. accepted29 July 1991

Proc. R. Soc. Lond.B (1991)

2 Vol. 246. B (22 October 1991)