Considering the Value and Limits of Evolutionary Prediction*

vol. 193, no. 1 the american naturalist january 2019

American Society of Naturalists Address From the Past to the Future: Considering the Value and Limits of Evolutionary Prediction*

Ruth G. Shaw†

Department of Ecology, Evolution, and Behavior, University of Minnesota, Twin Cities, Saint Paul, Minnesota 55108 Submitted April 27, 2018; Accepted June 22, 2018; Electronically published October 30, 2018 abstract: The complex interplay of the multiple genetic processes to characterize in detail. In addition, environment, which of evolution and the ecological contexts in which they proceed frus- mediates much selection, varies over time and space, often trates detailed identification of many of the states of populations, both idiosyncratically. The dependence of a population’sevolu- past and future, that may be of interest. Prediction of rates of adapta- tionary change on its history (e.g., Travisano et al. 1995) tion, in the sense of change in mean fitness, into the future would, compounds the challenge. Together, these characteristics however, valuably inform expectations for persistence of populations, of evolution obviateaccurateprediction of apopulation’sde- especially in our era of rapid environmental change. Heavy invest- gree of adaptation—that is, its mean absolute fitness many ment in genomics and other molecular tools has fueled belief that — those approaches can effectively predict adaptation into the future. I generations into the future (Gerrish and Sniegowski 2012) contest this view. Genome scans display the genomic footprints of and, even over the short term, frustrate prediction of many the effects of natural selection and the other evolutionary processes attributes of interest. over past generations, but it remains problematic to predict future All of these reasons for doubting the possibility of de- change in mean fitness via genomic approaches. Here, I advocate for a tailed evolutionary prediction raise two key questions: What direct approach to prediction of rates of ongoing adaptation. Following evolutionary changes, if any, are currently feasible to pre- an overview of relevant quantitative genetic approaches, I outline the dict? And even more pertinent: Would it be worthwhile to promise of the fundamental theorem of natural selection for the study have these predictions? It is important to be clear about of the adaptive process. Empirical implementation of this concept can “ ” productively guide efforts both to deepen scientific insight into the pro- the meaning of the word prediction. In this essay, I am us- cess of adaptation and to inform measures for conserving the biota in the ing “predict” in the common language sense of foretelling at- face of rapid environmental change. tributes of a population sometime in the future, an objective that must draw not only on theory but also on data from the Keywords: adaptation, experimental evolution, fundamental theorem of natural selection, quantitative genetics. population(s) of interest. I am not considering retrodiction or hindcasting, that is, inference of past states. Nor am I re- ferring to qualitative prediction, the province of evolutionary Introduction theory developed as proof-of-concept models (Servedio et al. 2014), nor yet to prediction in the statistical sense of values Evolutionary biologists recognize immense impediments to fi fi tted from data in accordance with a statistical model. Each speci c, quantitative evolutionary prediction about popula- of these kinds of effort is valuable in its own right, but they tions in nature. The challenge stems from the size and com- are distinct from predicting states of populations in the fu- plexity of genomes, along with the multiple processes that ture. One kind of evolutionary prediction for which all of can change the genetic composition of populations: natural the necessary approaches are well established is prediction selection, gene flow, mutation, and genetic drift. In most fi fi fi of change in mean tness, that is, the rate of adaptation. I here speci c cases, each of these processes is profoundly dif cult propose that this research program merits extensive implementation. Beyond its intellectual interest, evaluating the rate of adaptation now looms in practical importance as environ- * Ruth Shaw received the 2017 Sewall Wright Award. The Sewall Wright mental change at extreme rates threatens the persistence of Award, established in 1991, is given annually and honors a senior but still ac- populations and species globally. tive investigator who is making fundamental contributions to the society’s goals, namely, promoting the conceptual unification of the biological sciences. † Email: [email protected]. Can Genomics Predict Change in Mean Fitness? Am. Nat. 2019. Vol. 193, pp. 1–10. q 2018 by The University of Chicago. 0003-0147/2019/19301-58432$15.00. All rights reserved. With current heavy investment in genomics and other mo- DOI: 10.1086/700565 lecular tools, some have opined that those approaches can

This content downloaded from 128.101.134.127 on January 07, 2019 07:34:03 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c). 2 The American Naturalist effectively yield predictions of evolutionary adaptation in covariance between fitness and each of the traits; and b, the sense of change in mean fitness. I challenge this claim. the selection gradient, as a measure of selection directly on Genome scans yield the genomic historical record of the ef- each trait, is the partial regression of fitness on the traits. In fects of natural selection, along with all of the other evolu- the single-trait case, h2 is the trait heritability, the ratio of tionary processes, accumulated over many generations. the additive genetic variance of a trait to its phenotypic vari- Distinguishing from this record the loci that contributed ance. Reliable prediction is here favored by considering in to past adaptation remains problematic because of history aggregate the many polymorphic loci that collectively con- that generates population structure and linkage disequilib- tribute to the standing genetic variation in traits and to the rium (see, e.g., Tiffin and Ross-Ibarra 2014; Schrider et al. genetic covariance between fitness and traits. These equa- 2015; Hoban et al. 2016). Moreover, loci whose effects are tions, built on the regularities of Mendelian transmission, individually small but that can collectively account for much yield predictions for the change in trait mean from one of the adaptive change are unlikely to be detected (e.g., Lau- generation to the next, a modest timescale. rie et al. 2004). Thus, loci or genomic regions that can be Artificial selection experiments in controlled conditions discerned as possibly subject to selection generally repre- have demonstrated the utility of the quantitative predic- sent a small fraction of those that actually contribute to se- tions, even though realized responses to selection vary among lection response, and even these may often be mistakenly replicate populations and generations, reflecting random- identified (Schrider et al. 2015). More importantly, genomic ness of population sampling and genetic transmission. Sher- approaches do not generally elucidate rates of change in mean idan (1988), in his review of selection in livestock and lab- fitness except, at best, extremely indirectly (see Bay et al. oratory populations of Drosophila and Tribolium, deemed 2017, especially the numerous caveats therein). The tempo- selection responses to differ substantially from predictions. ral scale of the genomic signal of adaptation is that of many However, Hill and Caballero (1992) offer a critique of how the generations, and the spatial scale of most genomic studies comparisons were made (see also Walsh and Lynch 2018, far exceeds that occupied by an interbreeding population. p. 607) and a more encouraging view of agreement between In contrast, adaptation of a population in nature proceeds predicted and realized responses to selection. Realized re- on the timescale of generations and can continue over the sponses often show high repeatability and accord well with vast sweeps of time into the future only in populations that predictions (e.g., Enfield et al. 1966; Carey 1983; Conner et al. adapt to the vagaries of environment that confront them 2011). Moreover, even though h2 is expected to change with generation by generation and thus maintain an absolute fit- the changes in allele frequencies implied by the observed ness sufficient to persist. Adaptation over this timescale war- response to selection, in practice the rate of change in a trait rants direct study prospectively. I propose that quantitative or traits under artificial selection sometimes holds far beyond genetic approaches offer evolutionary predictions that would the initial generation—for 10–30 generations (Yoo 1980) or be of great value, both in deepening scientific understand- many more for large populations (Weber 1990; Weber and ing of evolutionary process and in informing measures to Diggins 1990). Within highly inbred lines, in which the stand- address issues of pressing societal concern. This line of ing genetic variation is nil (or very nearly so), the response to study is challenging, but I urge that the payoffs warrant selection is, as expected, not detectable initially (López and the effort. López-Fanjul 1993; Mackay et al. 1994; Keightley 1998). Thus, in accord with the theory for predicting selection response, empirical studies in the laboratory and the green- house have demonstrated the general reliability of predic- Prediction via Quantitative Genetics tions of trait change under artificial selection. But is it rea- Whereas anticipation of evolutionary change is largely lim- sonable to expect such reliability under natural selection? ited to qualitative prediction, quantitative genetic study One reason to believe so is that the equations for predicting stands as a key exception. Its approaches have long enabled response to selection apply equally for natural selection quantitative prediction of change in one or more trait mean when all of the traits under selection are known and taken phenotypes in response to selection via the breeder’sequa- into account (Lande 1979). However, artificial selection tion, R p h2S (Falconer and Mackay 1996), and its multi- contrasts with natural selection in a crucial way. In the for- variate form, DZ p GP21S p Gb (Lande 1979; Lande mer the experimenter chooses the trait or traits as the basis and Arnold 1983). Here, for a set of traits of interest, DZ for selection, whereas in the wild many traits may be sub- is the vector of the change in their means from one gener- ject to natural selection, and we do not know which ones ation to the next (R for a single trait); G is the matrix of a priori. To address this problem, Lande and Arnold (1983) additive genetic variances of and covariances between the presented theory and methods for inferring the direction traits; P is the matrix of their phenotypic variances and and magnitude of natural selection on each of a specified covariances; S, the selection differential, is the phenotypic set of correlated quantitative traits.

Challenges in Predicting Trait Change required to predict evolutionary genetic change from the due to Natural Selection multivariate breeder’s equation (or directly, as the additive fi genetic covariance between tness and traits, covA(w, z); Lande and Arnold’s (1983) influential paper inspired many e.g., Etterson and Shaw 2001). Observational studies of ped- field studies to assess the relationship between a measure of igreed populations of vertebrates have also yielded estimates individuals’ fitness and measures of phenotypic values of of the parameters required for prediction of response to nat- multiple traits (the approach is outlined in fig. 1A; studies ural selection on traits (fig. 1B; e.g., Grant and Grant 1995; reviewed by Kingsolver et al. 2001). Such studies yield Kruuk et al. 2002; Sheldon et al. 2003). In some of these estimates of the strength and direction of selection directly cases, the realized change in the traits aligned well with the on individual traits, b, the vector of selection gradients. predictions (e.g., Grant and Grant 1995). In their review of They may also indicate curvature, estimated as the matrix the vertebrate studies, however, Gienapp et al. (2008) found g, in the relationship between fitness and traits, providing that this is often not the case, noting the vexing issue of an estimate of the adaptive landscape in which the popula- distinguishing genetic response to selection from pheno- tion is evolving. As in the examples Lande and Arnold gave, typic plasticity in response to changing environment (see many such studies have based inference of selection on a also Pujol et al. 2018). Pemberton (2010) emphasizes other single component of fitness (e.g., survival over a particular impediments to accurate prediction from observational stud- interval). Moreover, phenotypic selection studies have gen- ies. Paramount among them is the general problem of con- erally considered a small number of traits because studying founding of environmental influences on organisms with selection on many traits poses at least two important chal- genetic influences, which stymies the effort to ascertain lenges. First, there are practical problems with obtaining the genetic contribution to fitness and traits. This is the clas- measures of many traits, along with fitness records, on each sic problem of discerning the effects of nature versus nurture of many individuals. Automation has made trait measure- (Lewontin 2000). Ironically, comparison of predicted to re- ment feasible on very large study populations for single alized responses to selection are far scarcer for the experi- (Weber 1990; Weber and Diggins 1990) and numerous (e.g., mental studies of annual plants, which are less subject to Houle et al. 2003) traits, enabling artificial selection on large these impediments. As one instance, however, Franks et al. scales. However, not all traits that may be targets of selection (2007) contemporaneously grew generations collected be- will be amenable to automated measurement, which is likely fore and after drought and thus conclusively demonstrated a to be largely restricted to laboratory studies. Second, the de- genetically based response to natural selection on flowering mand for data to ensure statistical power and precision in- time that accorded with prediction. creases dramatically with the number of correlated traits Whereas the challenges for evolutionary prediction from considered (Mitchell-Olds and Shaw 1987). Thus, for a given observational evidence are clear, the examples of experi- number of individuals the precision of estimates of selection mental studies tantalizingly intimate the predictive promise on each trait declines markedly as the number of traits in- of quantitative genetics that is more widely achievable. Im- creases. Moreover, a crucial limitation of the approach is the plementation of DZ p GP21S will, however, always entail a missing character problem. The particular traits chosen for fraught choice of the characters to consider, of which there study need not (and are unlikely to) account for all, or even are many more possible candidates as targets of selection most, of the variation in fitness. Omission of traits that are, than it would be feasible to measure. Moreover, the strength in fact, under selection and are correlated with traits under of selection on them and even which characters matter for consideration biases the estimates of selection on those traits. fitness seem likely to vary as environment changes over Furthermore, environmentally induced covariation between time, although limits on statistical precision may often make fitness and traits leads to spurious inference of selection that difficult to prove (Morrissey and Hadfield 2012). (Mitchell-Olds and Shaw 1987; Rausher 1992). In contrast to the body of phenotypic selection studies, Predicting Adaptation: Fisher’s Fundamental Theorem few studies of selection in nature have been designed to yield estimates of the genetic parameters, G,neededfor An alternative and potentially more robust way forward is quantitative evolutionary prediction. Ideally, such estima- direct prediction of the rate of adaptation via the funda- tion employs experimental designs that place progeny of mental theorem of natural selection (FTNS; Fisher 1930): D p = formal genetic crosses in the wild as the individuals whose W V A(W) W. Here, the evolutionary (i.e., genetically fitness and traits are recorded (fig. 1C; e.g., Rausher and based) change in mean fitness from the current generation Simms 1989; Campbell 1996; Tiffin and Rausher 1999; Et- to the next is predicted by the ratio of the population’s cur- terson and Shaw 2001). Thus, studying expression of fit- rent additive genetic variance for fitness to its mean fitness. ness and traits in the context of a quantitative genetic ex- Fitness, W, can be defined operationally as the number of periment conducted in nature yields all of the components offspring an individual contributes to the population, that

This content downloaded from 128.101.134.127 on January 07, 2019 07:34:03 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c). All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c). This content downloadedfrom 128.101.134.127 onJanuary07, 201907:34:03AM

Figure 1: Schematic diagram of approaches to studying natural selection in the wild highlighting their similarities and differences. Observational approaches: (A) study of phenotypic selection on traits and (B) study of phenotypic selection on traits, joined with pedigree information permitting estimation of the matrix of genetic (co)variances of the traits, G, and thus enabling prediction of change in trait means. Experimental approaches: (C) formal crosses ensure that progeny are derived by random mating; progeny are distributed at random over the environment, ensuring independence of genetic and environmental effects. Study of phenotypic selection on traits; use of pedigree to estimate G and thus enabling prediction of change in trait means. D, Implementation of the fundamental theorem of natural selection (FTNS). As in C, formal crosses ensure that progeny are derived by random mating; progeny are distributed at random over the environment, ensuring independence of genetic and environmental effects. Fitness records as complete as practical obtained on individuals. Estimates of = fi VA(W), W,andV A(W) W to predict change in mean tness. Asterisks indicate that estimates are subject to concerns about confounding between genetic and environmental effects (A, B) and missing characters (A–C). Prime and double prime symbols indicate distinct cohorts from the same set of progeny. Predicting Adaptation 5 is, the per capita population growth rate. Thus, W p 1yields posed to “hard sweeps” from de novo mutation (Messer the expectation that the population will remain the same and Petrov 2013). size, and a magnitude of W greater or less than 1 indicates Thus, a strong reason for optimism that FTNS can yield the rate at which the population is expected to grow or de- informative predictions of adaptive rate is that the magni- = fi cline. The ratio V A(W) W predicts the change in the pop- tude of the change in mean tness depends on a popula- ulation’s growth rate resulting from genetically based dif- tion’s current overall standing genetic variance for fitness ferential contribution of offspring to the next generation. in its current environmental context. Genetic contributions Importantly, in cases with an estimate of W ! 1, the magni- to variation in fitness are likely to be at least as highly poly- = tude of V A(W) W offers a prediction of the rate of adap- genic as many more readily measurable traits, such as oil con- tation toward W p 1 and, in particular, whether a single tent of corn kernels (Laurie et al. 2004) and human height generation of natural selection can be expected to result in (Yang et al. 2010). Moreover, at any given time it is not nec- evolutionary rescue, stabilizing the population’s persistence. essary to consider mutations that have newly arisen. Those Although the interpretation and even the validity of the that could eventually contribute to selection response are not FTNS were unclear at first, the derivation has now been ver- likely to contribute importantly just after they originate, when ified as quite general, and the interpretation has been eluci- they are rare. dated (e.g., Ewens 1989, 2004; Frank 2012). A crucial point To return to the key questions above: What evolutionary is that the FTNS predicts the change in mean fitness re- changes, if any, are currently feasible to predict? FTNS of- sulting strictly from genetic changes due to selection. This fers a strong conceptual basis for direct, quantitative pre- is the rate of evolutionary adaptation. The total change in diction of the rate of adaptation, in the sense of change in mean fitness may deviate from this substantially because mean absolute fitness. The methodologies of quantitative the environment in which the offspring develop may differ genetics enable its implementation, and experimental ap- in ways that directly cause their mean fitness to differ from proaches of population biology can guide assessment of the that of their parents (i.e., even in the absence of any genetic extent to which the predictions are realized. Would it be change; phenotypic plasticity with respect to fitness). worthwhile to have these predictions? Quantitative predic- The utility of FTNS in empirical contexts has not yet tions of the rate of genetic adaptation warrant the effort to been demonstrated; no published empirical research, to obtain them for both fundamental and practical reasons. my knowledge, has applied it, although we have such stud- Direct study of this kind would strengthen our under- ies under way (Eule-Nashoba 2016; Sheth et al. 2018; R. G. standing of the process of adaptation, including feedbacks Shaw and S. Wagenius, unpublished data [see the overview between evolutionary change and population dynamics in “Experimentally Applying FTNS” belowaswellasShaw (community genetics; Antonovics 1992; also known as eco- and Etterson (2012) and Gomulkiewicz and Shaw (2013)]). evolutionary feedbacks). It would also inform efforts to pro- Fisher (1930) himself likely dampened enthusiasm for empir- mote population persistence and guide those efforts to where ical application of FTNS by suggesting that concomitant en- they are most needed. vironmental deterioration (e.g., from competitive pressure Among the specific empirical questions that follow are due to an increase in population density as adaptation pro- these: What is the magnitude of VA(W), in particular natu- ceeds) would generally cancel out any increase in the mean ral populations? What is the rate of adaptation that it absolute fitness of individuals. This has a logical appeal. It predicts? How closely do realized rates of adaptation accord is clear that population growth cannot exceed 1 indefinitely, with predictions? If they align closely, over how many gen- but how tight is such a governor? Others have suggested that erations? If they differ substantially, why? These questions

VA(W) would be readily exhausted by incessant selection on have not been answered, not even for populations in the lab- fitness itself (e.g., Charlesworth 1987). However, recent theo- oratory—nor have they been answered for wild populations retical work by Zhang (2012; see also Ellner and Hairston in relatively undisturbed conditions, nor especially for popu- 1994; Shaw and Shaw 2014) has made clear that various re- lations whose fitnesses (hence, whose likelihood of persis- gimes of environmental change can maintain VA(W), such tence or increase) are compromised by rapid changes in en- that it would be available as a basis for response to selec- vironment. tion. Moreover, evidence now emerging from genomic studies indicates that past adaptation has generally proceeded through subtle changes in allele frequency at many simulta- Experimentally Applying FTNS neously segregating loci (Pritchard and Di Rienzo 2010; see also Boyle et al. 2017), in accordance with classical quantita- Implementation of FTNS requires estimation of the pop- tive genetic thought. Even when adaptation has resulted in ulation parameters VA(W)andW. Experiments to do so extremeallelefrequencies, this may have occurred primarily (fig. 1D) can draw on the general approaches of quantita- via “soft sweeps” from standing genetic variation as op- tive genetics for estimating the additive genetic variance,

This content downloaded from 128.101.134.127 on January 07, 2019 07:34:03 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c). 6 The American Naturalist as laid out in textbooks (Falconer and Mackay 1996, chap. 10; It is important to acknowledge challenges. For example, Lynch and Walsh 1998, chaps. 18, 20). In brief, estimation whereas individual fitness is, in principle, simple to define of additive genetic variance of any trait, including absolute as the number of offspring an individual contributes to the fitness, entails measures of the trait on a pedigreed set of in- population, the biology of fitness expression, which may span dividuals that is representative, both in the genetic sampling a single year or many, often complicates the measurement of and in the environment in which it is grown, of the popula- fitness in practice. Apart from the issue of life span, it is of- tion of interest. For the purpose at hand, it is appropriate to ten most convenient to assess fitness through maternity, consider a population in the sense of a deme (or, more spe- ignoring paternal fitness variation because of the difficulty cifically, a gamodeme; Gilmour and Gregor 1939), a local of attributing offspring to particular sires. It is now, however, group likely to be exchanging genes. Because of a deme’styp- feasible to assign paternity with considerable confidence (e.g., ically restricted spatial extent, most of its members may be Kulbaba and Worley 2012; Briscoe Runquist et al. 2017). With subject to many (if not all) of the same ongoing selective pro- such information, FTNS could be applied using estimates cesses. based on either maternal fitness or paternal fitness. Eventual interpretation of the experiment depends criti- cally on avoidance of confounding between genetic and en- Comparing Realized Adaptation to Prediction vironmental effects. To ensure this independence, sets of experimental individuals can be generated from random Predictions of the evolutionary change in mean fitness are choice of a large number of parental individuals by inter- warranted in their own right because they would directly crossing the parents in a formal, randomized mating scheme. characterize a population’s expected capacity for ongoing In a common approach, individual paternal parents are each adaptation to its current environment. They would be of mated to a distinct set of multiple maternal parents, as de- even greater value in comparison to the realized change in scribed by Falconer and Mackay (1996, chap. 10), to obtain mean fitness. To estimate the realized change in mean fit- progeny in paternal half-sib groups that include subgroups ness, a representative sample of the next generation must of full sibs. Any mating scheme that results in groups of pa- be grown and its mean fitness evaluated. As noted above, ternal half-sibs will allow estimation of additive genetic vari- the environments in which the parental and offspring co- ance unbiased by variance due to dominance or maternal ef- horts develop are likely to differ, and this could directly con- fects. The progeny are set out, randomly arrayed over the (one tribute to the difference in mean fitness between parental or possibly more) focal environment(s), to develop and ma- and offspring cohorts, in addition to the genetically based ture in conditions as similar as possible to the conditions un- change in mean fitness. To isolate the latter, the parental der which rate of adaptation is the focus. Raising the progeny and offspring cohorts must be grown in a common envi- at the site from which the parents were sampled offers the op- ronment—that is, a genetically representative sample of the portunity to predict the ongoing rate of adaptation of the original pedigreed cohort must also be grown and its fit- population under the conditions prevailing in situ. ness evaluated, along with the offspring cohort. This ap- The progeny are then followed, ideally throughout the life proach parallels that of mutation accumulation experiments, span, and measures of components of fitness are obtained with similar rationale (e.g., Schultz et al. 1999; Shaw et al. for each (see, e.g., Campbell 1997; Stanton-Geddes et al. 2012), 2000). to accumulate records of lifetime fitness. With these data At the outset, it is important to recognize that the realized fi in hand, it remains to estimate VA(W)andW. The frequency change in mean tness may often poorly match the predic- distribution of lifetime fitness does not lend itself to stan- tion, even when the direct effect of change in environment dard statistical analysis; as a compound distribution arising is addressed as described above. The overall change in mean from the serially expressed components of fitness, lifetime fitness also depends on genetic variation in sensitivity to fitness conforms to no standard probability distribution (Shaw such environmental change (i.e., the interaction between et al. 2008). To alleviate this problematic aspect of fitness genotype and temporally varying environment), which sig- data, Geyer et al. (2007) developed aster modeling, which nifies temporal change in genetic selection, whether this takes into account the dependence structure of arbitrarily change is erratic or due to more regularly fluctuating selec- many components of fitness to obtain, via maximum like- tion. Moreover, even in experimentally controlled labora- lihood, usefully precise inferences about fitness (see also tory environments, rates of change in mean fitness can vary Shaw et al. 2008). Its extension to accommodate random ef- substantially among replicate microbial populations subject fects, including random genetic effects (Geyer et al. 2013), to the same selection regime (e.g., Travisano et al. 1995; Ra- enables estimation of VA(W) and of VA(W)/W, the prediction miro et al. 2016), although it is important to note that these from the FTNS of the increase in mean fitness of the next gen- studies typically span 1,000 generations or many more. The eration over the current one due to the change in allele fre- match between realized and predicted change in mean fit- quencies resulting from selection in the current generation. ness for wild populations—and explanations for deviations

This content downloaded from 128.101.134.127 on January 07, 2019 07:34:03 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c). Predicting Adaptation 7 between them—are empirical questions whose answers can aspects. Even so, measurement of traits of individuals, in con- come only from direct experimental study. junction with their fitness components, would illuminate selection on those traits and suggest their roles in the adaptive response. Inclusion of trait values could also bear on the role Potential Breadth of Application and Variants on It of trait plasticity in adaptation (see, e.g., Chevin et al. 2010). Unquestionably, experiments of this kind require consider- Similarly, measurements of environmental attributes near in- able effort, and, even more obviously, some kinds of organ- dividuals could suggest roles those features play in selection isms are better suited to such experiments than others. In and adaptation. Because environmental attributes are gener- particular, plants and other sessile organisms “stand still ally highly intercorrelated, conclusive evidence about their and wait to be counted” (Harper 1977, p. 515). Even studies roles as agents of selection would entail experimental manip- of these organisms alone would be highly informative, but ulation (e.g., increasing CO2 concentration; Lau et al. 2007). others may also be good candidates. Via’s (1991) confinement As noted at the outset, it is not selection alone that of aphids on host plants in the field to obtain full fitness eval- changes the genetic composition of populations. Genetic uations demonstrates that there is much greater potential for drift, gene flow, and mutation also play important roles. such studies. More vagile organisms would have to be tagged The stochasticity of drift implies that it undermines the pre- and reliably recoverable for fitness assessment, as has been dictability of change in mean fitness. The extent to which done with several experimentally manipulated vertebrate pop- it generates discrepancies between predicted and realized ulations (e.g., De Lisle and Rowe 2015; Bolnick and Stutz 2017). change in mean fitness could be directly studied by exper- Organisms whose lives span decades pose obvious chal- imentally varying effective population size of study cohorts lenges. Nevertheless, for a few perennial plants with life (e.g., Newman and Pilson 1997). Likewise, gene flow could spans up to a decade, investigators have conducted formal also be varied experimentally to assess its interplay with se- quantitative genetic experiments in the field, obtaining lection. It would be of interest to vary the amount of gene complete lifetime fitness measures (Campbell 1997; Camp- flow via both haploids (e.g., pollen) and diploids, for which bell et al. 2008; Shefferson and Roach 2012), and others the genetic and demographic consequences differ, impor- are in progress with plants of still greater life span (R. G. tantly affecting the evolutionary dynamics and outcomes Shaw and S. Wagenius, unpublished data). A priori, for re- (Aguilée et al. 2013). It would also be of value to vary the pop- liable prediction from FTNS it seems necessary to obtain ulation sources (e.g., their distances or the disparity of the fitness records over the full lifetime, particularly in cases habitats). Such studies can inform the pressing debate about with strong genetically based trade-offs between early- and the merits and risks of human-mediated gene flow (managed late-life components of fitness. However, it is an empirical relocation, assisted migration) for enhancing the persis- question whether useful predictions of the rate of adaptation tence of populations threatened by environmental change can be obtained from incomplete fitness records. An en- (Richardson et al. 2009; Aitken and Whitlock 2013). Newly abling condition would be that, after reproductive maturity, arising mutations are unlikely to substantially influence the the genetic variance in fitness and mean fitness change to- change in the mean fitness of the immediately subsequent gether through the life span such that the ratio remains generation due to their initial rarity. It would be of great in- roughly constant. This condition would promote the feasi- terest, however, to conduct a study as outlined here on sets bility of quantitative genetic studies to predict adaptation of of mutation-accumulation lines as a basis for augmenting much longer lived organisms (e.g., trees; Yeh and Heaman understanding of the contribution of newly arising mutations 1987; Aitken and Adams 1996; Warwell and Shaw 2017). to ongoing changes in mean fitness in the wild. The foregoing outline represents a general approach on To predict a population’s rate of adaptation to different which many variants can be envisioned to address ques- conditions, the same reference population can be grown tions related to the central ones. To evaluate how consistent at different sites (e.g., a chronosequence representing climate selection is over time, it would be worthwhile to grow change; Etterson and Shaw 2001). Even within a site, the se- subsets of individuals from the same pedigrees at successive lective environment (sensu Antonovics et al. 1988) may vary. times. The predicted rate of adaptation may differ, but even The approach outlined here could be further extended to clar- if it hardly differs, comparison of genetic effects on fitness ify the spatial scale of selective heterogeneity (although at no between cohorts could reveal interaction between genotype finer scale than could be addressed in an informatively repli- and temporally varying environment such that the genetic cated experiment). Fitness expression may depend on genetic composition favored by selection differs between them. The composition of neighboring conspecifics or their density (e.g., specifics underlying predicted and observed changes in mean Shaw 1986; Campbell et al. 2017). The nature of such “soft se- fitness (which traits, which alleles, and what aspects of envi- lection” and how it bears on adaptation could be investigated ronment played key roles?) will often be elusive, simply be- through experimental manipulation of neighbor density and cause of the enormously high dimensionality of each of these relatedness. To evaluate the role of a partner species in medi-

This content downloaded from 128.101.134.127 on January 07, 2019 07:34:03 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c). 8 The American Naturalist ating ongoing adaptation, exposure to the partner could be Acknowledgments varied (e.g., microbial inoculation; Alexander et al. 2017; I am deeply grateful to the American Society of Naturalists Ossler and Heath 2018). Yet further, it is of interest to learn for honoring me with the Sewall Wright Award and for in- about how capacity for adaptation changes over time. The viting this contribution. I thank my mentors, especially Janis Project Baseline archive of seeds representing numerous spe- Antonovics and Joe Felsenstein, who introduced me to these cies in substantial seed collections from populations occupy- ideas and career-long colleagues who helped develop them. ing diverse habitats over the period 2012–2017 (Etterson et al. My thanks to G. May and M. Travisano for thoughtful dia- 2016) is an unparalleled resource for such studies. As experi- logue; H. M. Alexander, J. R. Etterson, E. R. Grosholz, M. ence with this approach advances, it will eventually be possi- Kulbaba, A. C. Love, S. N. Sheth, P. Tiffin, and A. Weis for ble to address the extent to which generalization about rates comments on the manuscript; and participants in the Biolog- of adaptation beyond individual populations or over longer ical Interest Group of the Minnesota Center for Philosophy of periods is warranted. Science for probing, constructive discussion of these ideas and their presentation. D. Bolnick, P. Nosil, and an anony- Conclusion mous reviewer provided comments that helped clarify key points. I gratefully acknowledge support from the National Among the profound aspects of evolution is its playing out Science Foundation and the Environment and Natural Re- over enormous sweeps of time, from more than 4 billion sources Trust Fund of Minnesota. years in the past onward into the indefinite future. Evolu- tionary biologists, through a multiplicity of approaches, have woven countless insights into a fabric of understanding about organisms in the remote past and their changes up to the present. These considerations of evolution over Literature Cited unimaginable spans of time have tended to dwarf interest Aguilée, R., F. H. Shaw, F. Rousset, R. G. Shaw, and O. Ronce. 2013. in evolutionary change on the timescale of organismal gen- How does pollen vs. seed dispersal affect niche evolution? Evolu- erations. Darwin’s (1859) view that adaptation would pro- tion 67:792–805. doi:10.1111/j.1558-5646.2012.01816.x. ceed via imperceptible changes accumulating exceedingly Aitken, S. N., and W. T. Adams. 1996. Genetics of fall cold hardiness fi slowly surely influenced this emphasis in the early period in coastal Douglas- r in Oregon. Canadian Journal of Forest Re- search 26:1828–1837. of evolutionarystudy(Antonovics1987). Now,however,nu- Aitken, S. N., and M. Whitlock. 2013. Assisted gene flow to facilitate merous retrospective studies have detected adaptation in local adaptation to climate change. Annual Review of Ecology, the wild over tens of generations or fewer (see above). Even Evolution, and Systematics 44:367–388. so, it is also clear, from field observations (Bradshaw 1991) Alexander, H. M., E. Bruns, H. Schebor, and C. M. Malmstrom. and from experiments in the field (e.g., Newman and Pilson 2017. Crop-associated virus infection in a native perennial grass: 1997) and laboratory (e.g., Bell and Gonzalez 2011; Lindsey reduction in plant fitness and dynamic patterns of virus detection. – et al. 2013), that adaptation under natural selection may not Journal of Ecology 105:1021 1031. Antonovics, J. 1987. The evolutionary dys-synthesis: which bottles suffice to maintain a population in a habitat where its fitness for which wine? American Naturalist 129:321–331. has been severely compromised. Can we predict evolutionary ———. 1992. Towards community genetics. Pages 426–449 in R. S. rescue and adaptive persistence versus failure to adapt, result- Fritz and E. L. Simms, eds. Ecology and evolution of plant resis- ing in ultimate extirpation for particular populations in the tance to herbivores and pathogens: ecology, evolution, and genet- wild (Gomulkiewicz and Shaw 2013)? This is an empirical ics. University of Chicago Press, Chicago. question that has importance both for advancing basic evolu- Antonovics, J., N. C. Ellstrand, and R. N. Brandon. 1988. Environ- tionary understanding and for informing measures intended mental variation and genetic variation: expectations and experiments. Pages 275–303 in L. D. Gottlieb and S. K. Jain, eds. Plant to conserve the biota as the environment changes so rapidly. evolutionary biology. Chapman & Hall, New York. Efforts using this direct approach to predicting and eval- Bay, R. A., N. Rose, R. Barrett, L. Bernatchez, C. K. Ghalambor, J. R. uating evolutionary adaptation require humility and mod- Lasky, R. B. Brem, S. R. Palumbi, and P. Ralph. 2017. Predicting re- est expectations. From the outset, it is important to ac- sponses to contemporary environmental change using evolutionary knowledge that predictions may often be imprecise even response architectures. American Naturalist 189:463–473. when they are accurate. Characterizing the magnitude of Bell, G., and A. Gonzalez. 2011. Adaptation and evolutionary rescue in uncertainty is itself worthwhile. Predictions of the geneti- metapopulations experiencing environmental deterioration. Sci- ence 332:1327–1330. cally based change in mean fitness will not likely be met Bolnick, D. I., and W. E. Stutz. 2017. Frequency dependence limits in all cases. Even when they are not, well-designed experi- divergent evolution by favouring rare immigrants over residents. ments that enable both these predictions and evaluation Nature 546:285–288. of the extent to which they are met will directly illuminate Boyle, E. A., Y. I. Li, and J. K. Pritchard. 2017. An expanded view of the adaptive process. complex traits: from polygenic to omnigenic. Cell 169:1177–1186.

Bradshaw, A. D. 1991. The Croonian lecture 1991: genostasis and the ation. Proceedings of the National Academy of Sciences of the limits to evolution. Philosophical Transactions of the Royal Society USA 104:1278–1282. B 333:289–305. Gerrish, P. J., and P. D. Sniegowski. 2012. Real time forecasting of near- Briscoe Runquist, R. D., M. A. Geber, M. Pickett-Leonard, and D. A. future evolution. Journal of the Royal Society Interface 9:2268–2278. Moeller. 2017. Mating system evolution under strong pollen lim- Geyer, C. J., C. E. Ridley, R. G. Latta, J. R. Etterson, and R. G. Shaw. itation: evidence of disruptive selection through male and female 2013. Local adaptation and genetic effects on fitness: calculations fitness in Clarkia xantiana. American Naturalist 189:549–563. for exponential family models with random effects. Annals of Ap- Campbell, D. R. 1996. Evolution of floral traits in a hermaphroditic plied Statistics 7:1778–1795. plant: field measurements of heritabilities and genetic correlations. Geyer, C. J., S. Wagenius, and R. G. Shaw. 2007. Aster models for life Evolution 50:1442–1453. history analysis. Biometrika 94:415–426. ———. 1997. Genetic and environmental variation in life-history Gienapp, P., C. Teplitsky, J. S. Alho, J. A. Mills, and J. Merilä. 2008. traits of a monocarpic perennial: a decade-long field experiment. Climate change and evolution: disentangling environmental and Evolution 51:373–382. genetic responses. Molecular Ecology 17:167–178. Campbell, D. R., A. K. Brody, M. V. Price, N. M. Waser, and G. Al- Gilmour, J. S. L., and J. W. Gregor. 1939. Demes: a suggested new dridge. 2017. Is plant fitness proportional to seed set? an experi- terminology. Nature 144:333. ment and a spatial model. American Naturalist 190:818–827. Gomulkiewicz, R., and R. G. Shaw. 2013. Evolutionary rescue be- Campbell, D. R., N. M. Waser, G. Aldridge, and C. A. Wu. 2008. yond the models. Philosophical Transactions of the Royal Society Lifetime fitness in two generations of Ipomopsis hybrids. Evolu- B 368:20120093. tion 62:2616–2627. Grant, P. R., and B. R. Grant. 1995. Predicting microevolutionary re- Carey, K. 1983. Breeding system, genetic variability, and response to sponses to directional selection on heritable variation. Evolution selection in Plectritis (Valerianaceae). Evolution 37:947–956. 49:241–251. Charlesworth, B. 1987. The heritability of fitness. Pages 21–40 in J. W. Harper, J. L. 1977. Population biology of plants. Academic Press, Lon- Bradbury and M. B. Andersson, eds. Sexual selection: testing the don. alternatives. Wiley, Chichester. Hill, W. G., and A. Caballero. 1992. Artificial selection experiments. Chevin, L.-M., R. Lande, and G. M. Mace. 2010. Adaptation, plastic- Annual Review of Ecology and Systematics 23:287–310. ity, and extinction in a changing environment: towards a predic- Hoban, S., J. L. Kelley, K. E. Lotterhos, M. F. Antolin, G. Bradburd, tive theory. PLoS Biology 8:e1000357. D. B. Lowry, M. L. Poss, L. K. Reed, A. Storfer, and M. C. Whitlock. Conner, J. K., K. Karoly, C. Stewart, V. A. Koelling, H. F. Sahli, and F. H. 2016. Finding the genomic basis of local adaptation: pitfalls, prac- Shaw. 2011. Rapid independent trait evolution despite a strong pleio- tical solutions and future directions. American Naturalist 188:379– tropic genetic correlation. American Naturalist 178:429–441. 397. Darwin, C. 1859. The origin of species. J. Murray, London. Houle, D., J. Mezey, P. Galpern, and A. Carter. 2003. Automated mea- De Lisle, S. P., and L. Rowe. 2015. Ecological character displacement surement of Drosophila wings. BMC Evolutionary Biology 3:25. between the sexes. American Naturalist 186:693–707. Keightley, P. D. 1998. Genetic basis of response to 50 generations of Ellner, S., and N. G. Hairston Jr. 1994. Role of overlapping generations selection on body weight in inbred mice. Genetics 148:1931–1939. in maintaining genetic variation in a fluctuating environment. Kingsolver, J. G., H. E. Hoekstra, J. M. Hoekstra, D. Berrigan, S. N. American Naturalist 143:403–417. Vignieri, C. E. Hill, A. Hoang, P. Gibert, and P. Beerli. 2001. The Enfield, F. D., R. E. Comstock, and O. Braskerud. 1966. Selection for strength of phenotypic selection in natural populations. American pupa weight in Tribolium castaneum. I. Parameters in base popu- Naturalist 157:245–261. lations. Genetics 54:523–533. Kruuk, L. E. B., J. Slate, J. M. Pemberton, S. Brotherstone, F. Guinness, Etterson, J. R., S. J. Franks, S. J. Mazer, R. G. Shaw, N. L. Soper Gorden, and T. H. Clutton-Brock. 2002. Antler size in red deer: heritability H. E. Schneider, J. J. Weber, K. J. Winkler, and A. E. Weis. 2016. and selection, but no evolution. Evolution 56:1683–1695. Project baseline: an unprecedented resource to study plant evolution Kulbaba, M. W., and A. C. Worley. 2012. Selection on floral design across space and time. American Journal of Botany 103:164–173. in Polemonium brandegeei (Polemoniaceae): female and male fit- Etterson, J. R., and R. G. Shaw. 2001. Constraint to adaptive evolu- ness under hawkmoth pollination. Evolution 66:1344–1359. tion in response to global warming. Science 294:151–154. Lande, R. 1979. Quantitative genetic analysis of multivariate evolu- Eule-Nashoba, A. 2016. Fitness and adaptive capacity in a Minnesota tion, applied to brain∶body size allometry. Evolution 33:402–416. prairie. PhD diss. University of Minnesota. Lande, R., and S. J. Arnold. 1983. The measurement of selection on Ewens, W. J. 1989. An interpretation and proof of the fundamental the- correlated characters. Evolution 37:1210–1236. orem of natural selection. Theoretical Population Biology 36:167–180. Lau, J. A., R. G. Shaw, P. B. Reich, F. H. Shaw, and P. Tiffin. 2007. ——— . 2004. Mathematical population genetics. I. Theoretical in- Strong ecological but weak evolutionary effects of elevated CO2 troduction. 2nd ed. Springer, New York. on a recombinant inbred population of Arabidopsis thaliana. Falconer, D. S., and T. F. C. Mackay. 1996. Introduction to quanti- New Phytologist 175:351–362. tative genetics. 4th ed. Longman, Essex. Laurie, C. C., S. D. Chasalow, J. R. LeDeaux, R. McCarroll, D. Bush, Fisher, R. A. 1930. The genetical theory of natural selection. 2nd ed. B. Hauge, C. Lai, D. Clark, T. R. Rocheford, and J. W. Dudley. 2004. Dover, New York. The genetic architecture of response to long-term artificial selection Frank, S. A. 2012. Wright’s adaptive landscape versus Fisher’s funda- for oil concentration in the maize kernel. Genetics 168:2141–2155. mental theorem. Pages 41–57 in E. Svensson and R. Calsbeek, eds. Lewontin, R. C. 2000. The triple helix: gene, organism, and environ- The adaptive landscape in evolutionary biology. Oxford University ment. Harvard University Press, Cambridge, MA. Press, Oxford. Lindsey, H., J. Gallie, S. Taylor, and B. Kerr. 2013. Evolutionary res- Franks, S. J., S. Sim, and A. E. Weis. 2007. Rapid evolution of cue from extinction is contingent on a lower rate of environmen- flowering time by an annual plant in response to a climate fluctu- tal change. Nature 494:463–466.

López, M. A., and C. López-Fanjul. 1993. Spontaneous mutation for Shaw, R. G., and J. R. Etterson. 2012. Rapid climate change and the a quantitative trait in Drosophila melanogaster. I. Response to ar- rate of adaptation: insight from experimental quantitative genet- tificial selection. Genetics Research 61:107–116. ics. New Phytologist 195:752–765. Lynch, M., and B. Walsh. 1998. Genetics and analysis of quantitative Shaw, R. G., C. J. Geyer, S. Wagenius, H. H. Hangelbroek, and J. R. traits. Sinauer, Sunderland, MA. Etterson. 2008. Unifying life-history analyses for inference of fit- Mackay, T. F. C., J. D. Fry, R. F. Lyman, and S. V. Nuzhdin. 1994. ness and population growth. American Naturalist 172:E35–E47. Polygenic mutation in Drosophila melanogaster:estimatesfrom Shaw, R. G., and F. H. Shaw. 2014. Quantitative genetic study of the response to selection of inbred strains. Genetics 136:937–951. adaptive process. Heredity 112:13–20. Messer, P. W., and D. A. Petrov. 2013. Population genomics of rapid Shefferson, R. P., and D. A. Roach. 2012. The triple helix of Plantago adaptation by soft selective sweeps. Trends in Ecology and Evolu- lanceolata: genetics and the environment interact to determine tion 28:659–669. population dynamics. Ecology 93:793–802. Mitchell-Olds, T., and R. G. Shaw. 1987. Regression analysis of nat- Sheldon, B. C., L. E. B. Kruuk, and J. Merilä. 2003. Natural selection ural selection: statistical inference and biological interpretation. and inheritance of breeding time and clutch size in the collared Evolution 41:1149–1161. flycatcher. Evolution 57:406–420. Morrissey, M. B., and J. D. Hadfield. 2012. Directional selection in Sheridan, A. K. 1988. Agreement between estimated and realised ge- temporally replicated studies is remarkably consistent. Evolution netic parameters. Animal Breeding Abstracts 11:877–889. 66:435–442. Sheth, S. N., M. W. Kulbaba, R. E. Pain, and R. G. Shaw. 2018. Ex- Newman, D., and D. Pilson. 1997. Increased probability of extinction due pression of additive genetic variance for fitness in a population of to decreased genetic effective population size: experimental popula- partridge pea in two field sites. Evolution (forthcoming). tions of Clarkia pulchella. Evolution 51:354–362. doi:10.2307/2411107. Stanton-Geddes, J., P. Tiffin, and R. G. Shaw. 2012. Role of climate Ossler, J. N., and K. D. Heath. 2018. Shared genes but not shared ge- and competitors in limiting fitness across range edges of an annual netic variation: legume colonization by two belowground sym- plant. Ecology 93:1604–1613. bionts. American Naturalist 191:395–406. Tiffin, P., and M. D. Rausher. 1999. Genetic constraints and selection Pemberton, J. M. 2010. Evolution of quantitative traits in the wild: acting on tolerance to herbivory in the common morning glory mind the ecology. Philosophical Transactions of the Royal Society Ipomoea purpurea. American Naturalist 154:700–716. B 365:2431–2438. Tiffin, P., and J. Ross-Ibarra. 2014. Advances and limits of using Pritchard, J. K., and A. Di Rienzo. 2010. Adaptation—not by sweeps population genetics to understand local adaptation. Trends in Ecol- alone. Nature Reviews Genetics 11:665–667. ogy and Evolution 29:673–680. Pujol, B., S. Blanchet, A. Charmantier, E. Danchin, B. Facon, P. Travisano, M., J. A. Mongold, A. F. Bennett, and R. E. Lenski. 1995. Marrot, F. Roux, et al. 2018. The missing response to natural se- Experimental tests of the roles of adaptation, chance, and history lection in the wild. Trends in Ecology and Evolution 33:337–346. in evolution. Science 267:87–90. Ramiro, R. S., H. Costa, and I. Gordo. 2016. Macrophage adaptation Via, S. 1991. The genetic structure of host plant adaptation in a spa- leads to parallel evolution of genetically diverse Escherichia coli tial patchwork: demographic variability among reciprocally trans- small-colony variants with increased fitness in vivo and antibiotic planted pea aphid clones. Evolution 45:827–852. collateral sensitivity. Evolutionary Applications 9:994–1004. Walsh, B., and M. Lynch. 2018. Evolution and selection of quantita- Rausher, M. D. 1992. The measurement of selection on quantitative tive traits. Vol. 2. Oxford University Press, Oxford. traits: biases due to environmental covariances between traits and Warwell, M. V., and R. G. Shaw. 2017. Climate-related genetic var- fitness. Evolution 46:616–626. iation in a threatened tree species, Pinus albicaulis.American Rausher, M. D., and E. L. Simms. 1989. The evolution of resistance Journal of Botany 104:1205–1218. to herbivory in Ipomoea purpurea. I. Attempts to detect selection. Weber, K. E. 1990. Increased selection response in larger popula- Evolution 43:563–572. tions. I. Selection for wing-tip height in Drosophila melanogaster Richardson, D. M., J. J. Hellmann, J. S. McLachlan, D. F. Sax, M. W. at three populations sizes. Genetics 125:579–584. Schwartz, P. Gonzalez, E. J. Brennan, et al. 2009. Multidimensional Weber, K. E., and L. T. Diggins. 1990. Increased selection response in evaluation of managed relocation. Proceedings of the National Acad- larger populations. II. Selection for ethanol vapor resistance in Dro- emyofSciencesoftheUSA106:9721–9724. sophila melanogaster at two populations sizes. Genetics 125:585– Schrider, D. R., F. K. Mendes, M. W. Hahn, and A. D. Kern. 2015. 597. Soft shoulders ahead: spurious signatures of soft and partial selec- Yang, J., B. Benyamin, B. P. McEvoy, S. Gordon, A. K. Henders, tive sweeps result from linked hard sweeps. Genetics 200:267–284. D. R. Nyholt, P. A. Madden, et al. 2010. Common SNPs explain Schultz, S., M. Lynch, and J. Willis. 1999. Spontaneous deleterious a large proportion of the heritability for human height. Nature mutation in Arabidopsis thaliana. Proceedings of the National Genetics 42:565–569. Academy of Sciences of the USA 96:11393–11398. Yeh, F. C., and J. C. Heaman. 1987. Estimating genetic parameters of Servedio, M. R., Y. Brandvain, S. Dhole, C. L. Fitzpatrick, E. E. height growth in 7-year old coastal Douglas-fir from disconnected Goldberg, C. A. Stern, J. Van Cleve, and D. J. Yeh. 2014. Not just diallels. Forest Science 33:946–957. a theory—the utility of mathematical models in evolutionary biol- Yoo, B. H. 1980. Long-term selection for a quantitative character in ogy. PLoS Biology 12:e1002017. large replicate populations of Drosophila melanogaster.I.Re- Shaw, R. G. 1986. Response to density in a wild population of the sponse to selection. Genetics Research 35:1–17. perennial herb Salvia lyrata: variation among families. Evolution Zhang, X.-S. 2012. Fisher’s geometrical model of fitness landscape 40:492–505. and variance in fitness within a changing environment. Evolution Shaw, R. G., D. L. Byers, and E. Darmo. 2000. Spontaneous muta- 66:2350–2368. tional effects on reproductive traits of Arabidopsis thaliana.Ge- netics 155:369–378. Editor: Daniel I. Bolnick

This content downloaded from 128.101.134.127 on January 07, 2019 07:34:03 AM All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).