UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Speciation in sympatric sticklebacks: hybridization, reproductive isolation and the maintenance of diversity Hatfield, Todd 1994

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1995-059731.pdf [ 2.08MB ]
JSON: 831-1.0088044.json
JSON-LD: 831-1.0088044-ld.json
RDF/XML (Pretty): 831-1.0088044-rdf.xml
RDF/JSON: 831-1.0088044-rdf.json
Turtle: 831-1.0088044-turtle.txt
N-Triples: 831-1.0088044-rdf-ntriples.txt
Original Record: 831-1.0088044-source.json
Full Text

Full Text

SPECIKHON IN SYMPATRIC STICKLEBACKS: HYBRIDIZATION,REPRoDucTIvE ISOLATION AND THE MAINTENANCE OF DIVERSITYbyTODD HATFIELDB.Sc., Daihousie University, 1986M.Sc., Daihousie University, 1989A THESIS SUBMITIED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FORTHE DEGREE OF DOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADuATE STuDIEs(Department of Zoology)We accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH CoLuMBIAJune 29, 1995© Todd Hatfield, 1995In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature)__________________________Department of 2514/The University of British ColumbiaVancouver, CanadaDate /9‘DE-6 (2/88)ABSTRACTThis research examined genetic, morphological, ecological and behavioral mechanismsmaintaining divergence between a pair of sympatric stickleback species (Gasterosteus aculeatuscomplex) from Paxton Lake, British Columbia, Canada.When raised in the lab, F1 hybrids showed no evidence of inferiority relative to parental speciesfor the following characters: fertilization success, hatch success, growth rate, fecundity, fluctuatingasymmetry, and parental care. Although these are common postmating isolation mechanisms betweenmany other species, if they operate in the wild in this system their effects are likely small and of littlebiological importance. When raised in the wild, in either of two main habitats, F1 hybrids suffered asignificant reduction in growth rate relative to parental species. Hybrid disadvantage resulted frommorphological intermediacy which affected resource exploitation efficiency. Hybrid disadvantage is thusa function of the ecological environment, rather than developmental inviability or physiologicalinferiority. This ecological disadvantage provides the strongest postmating barrier known between thesespecies. Though this is one of the first demonstrations that niche-based selection pressures are a dominantmechanism of postmating isolation, I argue that this may be common in the early stages of speciation.Since gene flow between species must occur through reproductively fit F1 hybrids, I askedwhether male F1 hybrids are sexually selected against by females of the parental species. I provideexperimental evidence that mate choice by females of the parental species must be coupled with habitatpreference for there to be sexual selection against hybrids.I measured the degree of genetic divergence between the species in four morphological charactersand one fitness component by rearing six lines (parentals and first and second generation hybrids; 109families total) in the lab. Joint-scaling tests on phenotype means and variances suggest that epistasiscontributes significantly to genetic divergence of parental lines for two of the characters, gill raker lengthUand growth rate. Dominance effects contribute significantly to divergence in plate number and pelvicspine length. A simple additive model was sufficient for only gill raker number. A biometrical approachto gene number estimation suggested that character differences are coded for by 1 to 60 genes, dependingon the trait. The tremendous discrepancy in these estimates may reflect real differences among charactersin genetic divergence, but the estimates are likely unreliable due to the rejection of several importantassumptions of the method. My findings are compared to other recent estimates of genetic divergence innatural populations, and discussed in the context of speciation in sticldebacks.111TABLE OF CONTENTSAbstract iiTable of Contents ivList of Tables VList of Figures VAcknowledgements viiGeneral Introduction 1Chapter One Ecological Speciation: Lab Measurements of Hybrid Fitness 8Chapter Two Ecological Speciation: Niche-Dependent Growth Rates of F1 Hybrids 40Chapter Three Mate Choice, Habitat Preferences and Sexual Selection Against Hybrids 67Chapter Four Genetic Divergence in Adaptive Characters 91General Discussion 120Literature Cited 128ivLIST OF TABLESTABLE 1.1 Fertilization success of crosses carried out in the laboratory. 25TABLE 1.2 Analyses of variance testing for the effect of cross type on asymmetry. 26TABLE 1.3 Combined fitness measures for lab-reared crosses. 27TABLE 2.1 Measurements of initial mass, and seven morphological traits from fishused in the field experiments. 57TABLE 2.2 Results from a two-way ANOVA testing for the effects of species andhabitat type on ln-transfonned growth rates in Paxton Lake. 58TABLE 2.3 Stomach contents of benthics and F1 hybrids from littoral zone enclosures;and F1 hybrids and limnetics from open water enclosures. 59TABLE 3.1 Combinations and sample sizes of mate choice trials. 83TABLE 3.2 Female responses to courting males. 84TABLE 3.3 Analyses of variance for female courtship responses. 85TABLE 3.4 Locations of nests built by lab-reared males placed in enclosures in the wild. 86TABLE 4.1 Character means and standard errors for the six lines. 111TABLE 4.2 Tests for character differences between sexes in first generation fish(benthics, limnetics and F1 hybrids); and maternal effects in F1 malesfrom reciprocal crosses. 112TABLE 4.3 Parameter estimates from joint-scaling tests for four characters. 113TABLE 4.4 Goodness-of-fit (chi-square) test statistics from joint-scaling tests onmeans of five characters. 114TABLE 4.5 Estimates of gene number for four morphological characters. 115VLIST OF FIGURESFIGURE 1.1 Design used to make crosses for this study. 29FIGURE 1.2 Boxplots of hatch success from laboratory crosses. 31FIGURE 1.3 Comparison of mean family growth rates of six cross types rearedin the lab on identical diets. 33FIGURE 1.4 Fecundity versus length relationship for lab-reared females. 35FIGURE 1.5 Asymmetry in gill raker length for eight groups of fish. 37FIGURE 1.6 Parental care provided by limnetic and F1 hybrid males. 39FIGURE 2.1 Mean growth rates of families of benthics, F1 hybrids and limneticsreared in the lab on identical diets. 62FIGuRE 2.2 Growth rates of fish raised in the littoral zone and in the open waterof Paxton Lake. 64FIGURE 2.3 Fitness set calculated from growth trials in 1991, 1992, and 1993. 66FIGURE 3.1 A series of hypothetical fitness sets showing how different levels ofhabitat selection can affect fitness isoclines. 88FIGURE 3.2 Fitness sets showing mean responses by limnetic and benthic femalesto the three male types (limnetic, F1 hybrid and benthic). 90FIGURE 4.1 Observed character means and standard errors for five traits measuredin each of the six lines. 117FIGURE 4.2 Observed character variances plotted against observed character meansfor four morphological traits in each of the six lines. 119viAcKNowLEDGEMENTsI thank Spencer Cotton, Kay Gopaul, Laurin Hummeibrunner, Laura Nagel and Sarah Wu forfirst-rate help with experiments; Max Blouw for patient instruction on methods of stickleback farming;Steve Heard, Arnon Lotem, Jeff McKinnon, and David Westcott for reading one or more chapters andproviding many helpful criticisms; and Harold Diggon and Otto Jespersen for site access and logisticalhelp on Texada Island. I gratefully acknowledge the use of unpublished chapters on genetic analysis fromMichael Lynch and Bruce Walsh, and the use of computer code for statistical analysis from Bill Rice andStephen Gaines. I thank my committee, Don McPhail, Bill Neill, and Jamie Smith for allowing me tomake mistakes, but providing supervision when it was needed, and for providing superb criticisms of mywritten work. My supervisor, Doiph Schiuter, had an enthusiasm that was entirely infectious, and heconstantly encouraged me to perform at my highest level. Finally, I thank my family and friends forexceptional support, and for the distractions needed to keep my feet on the ground. I especially thankElizabeth Vibert for extraordinary confidence, encouragement, and patience, and for making me feel likeI belong.This research was funded by grants from NSERC to Dolph Schluter. I also gratefullyacknowledge three years’ support from a UBC graduate fellowship.viiGENERAL INTRODUCTIONThe study of speciation is the oldest of the various fields comprising evolutionary biology. Thedesire to understand patterns of diversity in nature motivated virtually all of the early naturalists such asLinnaeus, Lamarck and Darwin. Their motivations were to describe biological diversity, to establishwhether there was an order to it, and then to examine the mechanisms promoting it. Nearly one hundredand forty years after Darwin’s Origin ofSpecies many would argue that the study of speciation is in itsinfancy—that speciation is still a “mystery of mysteries” (Coyne and Barton 1988; Coyne 1992).Speciation occurs when a single population gives rise to two or more reproductively isolatedpopulations. The critical event in speciation is thus the evolution of reproductive isolation. Reproductiveisolation is traditionally broken down into premating and postmating isolation. Premating isolation isbrought about by factors which prevent matings between populations or incipient species—for example,behaviors which cause individuals of two populations to be in different places when it comes time tomate; or, if they are in the same place, to prefer only their own kind. If individuals from the twopopulations do mate, then postmating isolation lowers the probability that hybrid offspring will surviveand/or reproduce.I carried out research on the speciation of two species of stickleback from Paxton Lake, BritishColumbia, Canada. Threespine sticidebacks (Gasterosteus aculeatus complex) are small fish widelydistributed in marine and freshwaters of the Northern Hemisphere (Bell and Foster 1994). In BritishColumbia anadromous sticklebacks colonized many low elevation coastal lakes and streams at the end ofthe Pleistocene (McPhail 1994). The great majority of these lakes contain solitary populations, but severallakes on islands in the Strait of Georgia contain sympatric species pairs (McPhail 1984, 1992, 1993,1994; Schluter and McPhail 1992). Much work has been done on the Paxton Lake and Enos Lake pairs(see McPhail 1994 for review). The species show a similar pattern of morphological and ecological12divergence (Schiuter and McPhail 1992), but each pair is thought to have been independently derived(McPhail 1993). The species have not been formally described, so we refer to them by their preferredforaging habitats. “Limnetics” are found in the open water where they forage for plankton, and “benthics”forage for benthos in the littoral zone. Limnetics and benthics are distinct morphologically and are wellisolated by premating mechanisms. Limnetics have a fusiform body, narrow gape and many, long gillrakers; benthics have a robust body form, wide gape and few, short gill rakers. Gill rakers are the feedingapparatus of fish—they are thought to directly filter food particles, or otherwise aid in particle capture(Sanderson et al. 1991).There are two exceptionally fascinating aspects of these species’ biology for the student ofspeciation. The first is that one can successfully hybridize limnetics and benthics in the lab, and thePaxton Lake pair, at least, hybridize to a small degree in the wild (McPhail 1992). The second is that thespecies are so young that it is reasonable to expect that the forces which drove speciation are the sameforces which maintain their present reproductive isolation. My approach has therefore been to use hybridsin experimental studies to describe the current forces maintaining the two species as distinct. Thisapproach assumes that reproductive isolation is still evolving in sticklebacks. I have concentrated onpostmating isolation, by measuring potential costs of hybridization at multiple levels—genetic,developmental, behavioral and ecological. By taldng this very broad view of reproductive isolation I havetried to weigh the relative strengths of diverse mechanisms and to assess speciation as an evolutionaryprocess rather than as an outcome of evolution.I avoid a longer introduction to stickleback biology here because each of the following chaptersis meant to be largely self-contained, in that each describes the natural history of the system relevant tothe question(s) posed in the chapter, and describes the methods used to answer those questions. There isinevitably some overlap in these descriptions, but I have attempted to minimize it by referring to earlierchapters where appropriate.3Issues in speciation research.—I will now briefly outline some of the major problems inspeciation so that the following chapters may be placed in the general context of speciation studies. Tostart, it is worth touching on the subject of species concepts, since studies of speciation and speciesconcepts are inescapably intertwined. What one emphasizes as interesting and important in speciationdepends on which species concept one explicitly or implicitly adopts (Endler 1989). Defining a species isdifficult simply because the idea of species has different meanings depending on context and observer:paleobiologists will inevitably think of species differently than population geneticists, and ecologists viewspecies differently than systematists. Biologists have devoted many pages to the inadequacies ofparticular concepts and/or the promotion of an entirely new species concept, but the efforts more oftendescribe differences among researchers than provide useful insights for research on speciation. Thebewildering variety of concepts and the pros and cons of each have been recently treated usefully by bothTempleton (1989) and Grant (1994). My inclination has been to accept the traditional “biological speciesconcept” of Mayr (1942) and Dobzhansky (1937). I do, however, offer two slight revisions.The first revision is that I relax Mayr’ s (1942) requirement of zero gene flow as the criterion ofspecies status. In reality many organisms hybridize in nature. And in practice there is often little argumentover whether many of these organisms are reproductively isolated enough to be called species (see e.g.,Hubbs 1955; Grant and Grant 1992). My second revision is really just a suggestion that a slight change ofemphasis to acknowledge ecology would not be amiss. A good test of whether two populations arespecies is whether they can coexist in sympatry. To coexist species must be distinct ecologically. Goodspecies can, of course, be ecologically similar and one species may simply exclude another from the samearea. However, if we are interested in the mechanisms promoting speciation, then this does not seem auseful way to regard species. The most impressive radiations in nature (e.g., Darwin’s finches, AfricanGreat Lake cichlids, Caribbean anoles) have a common signature: the traits distinguishing these speciesare those associated with resource use. The implication is that not only do species often have different4niches, but that these niche differences may have played a substantial role in promoting the diversity weare interested in explaining. This ecological context of a species definition pervades all research in thisthesis.The great success of the biological species concept has been to promote a research program forspeciation which focusses on reproductive isolation (Coyne 1994). Though the literature is teeming withstudies of reproductive isolation, there are many who feel that our progress in understanding speciationhas nevertheless been insubstantial (Coyne and Barton 1988). I prefer to think of the study of speciationmore optimistically by drawing parallels to studies of natural selection and adaptation. The study ofadaptation has undergone what can really be described as a revolution over the last decade or two. Thefield has progressed from descriptions of adaptations as an outcome of evolution to studies of the processof adaptation, of “evolution in action”. To be sure, we still know little about why natural selection occurs(Endler 1986, 1989; Wade and Kalisz 1990), but direct tests of hypotheses using populations in the wildare providing some of the most impressive results in evolutionary biology. This profound shift inemphasis came from the realization that evolution and natural selection can be studied in time framesfamiliar to us—days, months or years, instead of millennia—and that they can be studied in the wild(Endler 1986).Studies of speciation are currently undergoing a similar revolution. Studies of the mechanisms ofreproductive isolation are numerous and have described the many ways in which organisms remainisolated from one another. These studies have inspired much theoretical work and many tests ofhypotheses with laboratory populations. The natural progression is now to build on these studies bytesting theories in situ with wild populations. I believe it is this step which has rejuvenated the field andled to its resurgent popularity.I have tried to take this approach for my thesis research. I have been motivated more by the desireto assess the ultimate causes of reproductive isolation than to describe the proximate mechanisms of5premating and postmating isolation. A solid description of isolating mechanisms remains a necessaryfoundation though, and much of the work here is traditional in that sense; it simply elaborates on orcorroborates excellent earlier work on the system. However, I think progress is being made towarddescribing the ultimate causes of speciation. By concentrating on hybrid fitness I have approached someof the most tractable yet under-studied aspects of speciation (Coyne and Barton 1988). My workhighlights the importance of ecological processes (especially feeding efficiency and habitat selection) indetermining reproductive isolation. The work also highlights gaps in our knowledge, suggesting fruitfulareas for further research with sticklebacks and other species.The thesis.—The presentation of the thesis is as follows. Chapter one measures the cost ofhybridization by measuring the fitness of hybrids in the lab. Chapter two measures the fitness of F1hybrids in the wild. Chapter three measures the cost of hybridization by asking how F1 hybrids would beexpected to fare in the competition for mates. Chapter four describes genetic divergence between thespecies by applying methods of quantitative genetics to families reared in the lab.Chapters one and two adopt an explicitly ecological view and focus on the process I call“ecological speciation”. The term has been used before to apply to speciation which indirectly produceslineages with different niches. For example, Lande and Kirkpatrick (1988) used the term when theymodelled speciation by sexual selection which resulted in ecologically divergent species. However, I useit more specifically to refer to speciations which are a direct result of selection on traits affecting nicheuse. These traits may include those responsible for efficient resource use, or avoidance of competitors,predators and parasites. Chapters one and two set up and carry out a test of ecological speciation due todivergent selection on resource use efficiency. The results demonstrate that ecological trade-offs inresource use not only allow divergence but actively determine postmating isolation by selecting againstindividuals with intermediate morphology, for example hybrids.6Chapter three tackles the issue of behavior and reproductive isolation. A long-standing and rarelytested theory has been that divergent mating preferences result not just in premating isolation, but also inpostmating isolation by selecting against hybrids (Fisher 1930). I test this idea for the sticklebacks. Thefindings indicate substantial premating isolation between limnetics and benthics, but postmating isolationby sexual selection against hybrids is uncertain. Premating isolation exists in the absence of habitatpreferences, and divergent habitat choice would only magnify the premating isolation I describe.However, hybrid mating success appears to be context specific. I show that hybrid males are preferred toan intermediate extent by females of both parental species, and that habitat preferences must accompanymating preferences for there to be significant sexual selection against hybrids. Habitat choice exists inmales, but we do not yet have data on female habitat preferences. The question also remains as to howhabitat preferences have evolved.Finally, in chapter four I explore the genetics of speciation. In essence, the work continues theexamination of ecological speciation. Traits which determine the divergent ecologies of limnetics andbenthics are those affecting feeding efficiency in different habitats, and divergent strategies to predatordefense. How the traits are inherited is directiy linked to postmating isolation, because the pattern ofinheritance affects how genes from the two species interact when combined in hybrid offspring. Itherefore subjected two feeding traits, two armor traits, and one fitness component to genetic analysis.Genetic divergence may show up as additive, dominance and epistatic effects. The relative magnitude ofeach of these effects influences hybrid fitness and therefore postmating isolation.It will be clear in the following chapters that there is much yet to be learned about speciation insticklebacks. The challenge, as I see it, is to use sticklebacks as a model system and to weigh the relativeroles of adaptation and evolutionary caprice in the speciation of these fish. We must be careful whenextrapolating beyond this system, but I think there is much in common between Paxton Lake sticklebacks7and most other radiations. This challenge has so far been both interesting and fun. I have no doubt it willcontinue to be so.Chapter One EcoLoGIcAL SPECIATION: LAB MEASUREMENTS OP HYBRID FITNESS8INTRODUCTIONSpeciation occurs when premating and/or postmating reproductive isolation evolves betweenpopulations. The central challenge for those studying speciation is first to describe how organisms remainreproductively isolated from one another, but ultimately to evaluate the processes causing these isolatingmechanisms to evolve. The proximate mechanisms of reproductive isolation have been well-studied in agreat variety of organisms, but we remain some way from consensus on the ultimate causes of speciation(Endler 1989; Templeton 1989). In this chapter, and the next, I examine the hypothesis that selection onresource use efficiency has been the ultimate cause of speciation in two sympatric stickleback species inBritish Columbia, Canada.Van Valen (1976) suggested that species are discrete units because natural selection removesforms that are intermediate. This idea was embedded in his ecological species concept, which proved toounwieldy to be useful for taxonomic classification (Ridley 1993). However, the idea persists as an explicithypothesis of how species are formed, and how diversity is maintained over evolutionary time (Ridley1993). This view is an extension of ideas of adaptive zones (Simpson 1944, 1953), adaptive landscapes(Wright 1932), and the ecological niche (Hutchinson 1957). The ideas are linked by a common thread:organisms have certain attributes to exploit resources, or to avoid parasites and predators. Those that lacksuch attributes occupy valleys in the adaptive landscape and are selected against. That species occupydiscrete niches or peaks in an adaptive landscape is intuitively obvious for most ecologists. Less obviousis the idea that speciation could be driven by adaptation to different niches (Schluter 1995a).Ecological speciation occurs when reproductive isolation evolves as an outcome of selection foradaptation to different resources. I refer to the products of this process as “ecological species”. Thebiological species concept is still easily applied to these species (Schluter 1995a), but designating a group910of organisms as ecological species is useful because it highlights the importance of natural selection andits ecological context to the speciation process.How can we test whether speciation has been aided by niche-based selection pressures? Threepieces of evidence are necessary: (1) a demonstration of disruptive natural selection acting betweenspecies, (2) evidence that the mechanism by which selection operates is resource exploitation efficiency,(3) that premating isolation evolved in response to divergent niche-based natural selection. In this chapterand the next I examine the first two criteria (i.e., postmating isolation) by assessing fitness of hybrids inthe lab and in the wild. Future work with these sticklebacks will examine the role, of ecological processesin the evolution of premating isolation. A general discussion of ecological speciation and prematingisolation can be found in Schluter (1995a) and Schiuter and Nagel (in press).A demonstration of selection against hybrids provides the best evidence of disruptive selectionbetween species. Studies of mechanisms of selection against hybrids have concentrated almost entirely onhybrid sterility or hybrid developmental inviability. Ecological speciation requires that there be ecologicalmechanisms of selection against hybrids. Support for ecological speciation is measured by the relativestrength of selection by ecological mechanisms versus selection by other mechanisms. Because organismswhich have speciated by any mechanism will continue to accumulate differences between them, time willeventually obscure differences which had arisen earlier, thereby potentially masking the originalmechanisms of speciation. It is therefore most appropriate to experimentally study ecological speciationby identifying and working with populations at an early stage of divergence.Ecological Species Pair?I now review prior evidence which led to the hypothesis that sympatric species of sticklebacks areecological species.11Anadromous threespine sticklebacks (Gasterosteus aculeatus complex) are small fish common tocoastal areas throughout the northern hemisphere, that move into shallow marine areas, estuaries, orfreshwater rivers and streams to reproduce. Like several other anadromous marine fish taxa (e.g.,Lampetra (Beamish 1982), Osmerus (Taylor and l3entzen 1993), Oncorhyncus (Foote et al. 1989)) theyhave given rise to permanently resident populations in freshwater lakes and streams. Many low elevationcoastal lakes in British Columbia were formed following isostatic rebound at the end of the Pleistocene,dating the lakes to a maximum of approximately 10,000 to 13,000 years old (Mathews et al. 1970). Mostof these young lakes contain only a single species of stickleback, which presumably invaded the lakesfrom the ocean following deglaciation (McPhail 1994). However, several lakes on islands in the Strait ofGeorgia contain sympatric species pairs (McPhail 1984, 1992, 1993, 1994; Schluter and McPhail 1992).The pattern of morphological and ecological divergence is very similar for each of the pairs. Ineach case, one of the species (referred to as limnetic) primarily exploits plankton, and has morphologicaltraits such as a fusiform body, narrow gape and many, long gill rakers which match this lifestyle. Theother species (referred to as benthic) mainly exploits benthic invertebrates in the littoral zone, and has arobust body form, wide gape and few, short gill rakers (Schluter and McPhail 1992). Gill rakers arethought to directly filter food particles, or divert flow to specialized buccal grooves for particle capture(Sanderson et al. 1991). Long numerous gill rakers are prevalent in planktivorous fish in general (Schiuterand McPhail 1993). I carried Out experiments on the pair of species from Paxton Lake, Texada Island.However, the findings I present here may be generalizable to the other pairs.Although benthics and limnetics exploit different habitats when feeding in these lakes, they arereproductively sympatric. Males acquire territories in the littoral region in the spring. They build nests,mate (usually with several to many females), and raise their offspring. Gravid females have a highprobability of encountering males of both species when searching for mates. The reproductive seasons ofbenthics and limnetics overlap substantially, and there is ample opportunity for the two to hybridize in the12wild. In the lab, hybrids (F1s,F2s and backcrosses) between limnetics and benthics are easy to obtain byartificial fertilization. In Paxton Lake, F1 hybrids are thought to be produced at a consistent but low level:three estimates over a twenty year span found that hybrids made up 1-2% of all adults (McPhail 1992).McPhail (1992) carried out a breeding program to examine whether morphological differenceswere maintained in the lab, and to what extent the forms were interfertile. Offspring of pure speciescrosses resembled parents which suggested genetic inheritance of morphology. McPhail concluded thatthe two forms were indeed species, yet he did not detect a lower hatch success of F1 or F2 hybrids. Ingeneral, McPhail’s results showed that limnetics and benthics had remarkably compatible genomes.In this chapter I expand on McPhail’s (1992) study by analyzing a much larger number offamilies (109 versus 25) to see whether there are subtle effects of hybrid breakdown that were notdetected in his study. I also investigated several additional measures of postmating isolation. This is anecessary first step in testing the hypothesis of ecological speciation since it allows me to weigh therelative effects of genetic and ecological mechanisms of selection against hybrids. In chapter two Iexamine postmating reproductive isolation between these species by studying selection against hybrids inthe wild.MATERIALs AN]) METhoDsExperimental crosses and rearing conditionsI made full sib crosses by extracting eggs from gravid females using gentle abdominal pressure,and combining the clutch in a petri dish with a small volume of water and macerated testes from a male. Ithen immediately selected a random subset of 30 eggs from the clutch. Densities were set to 30 fish perfamily, in order to hold constant possible effects of social interactions.13I raised each clutch separately in one half of a 100 L aquarium held at constant temperature andlight regime (180 C; 1 6L:8D). Assignment of families to aquaria was random. I placed clutches in mesh-bottomed 250 ml containers attached to the side of the aquarium, and aerated them from below using anairstone. I examined eggs every other day under a dissecting microscope until day five, after which theywere examined daily until hatching (7..9 days). I removed eggs infected by fungus, and those which hadceased development or were developing improperly, using Swarup (1958) as a guide to development.Fungus attacked only eggs which had ceased developing. As fish hatched they were removed from thehatching container and placed in the tank. Hatchuings were fed for the first few days on infusoria cultures,then transferred to diets of live Artemia nauplii, and frozen crustaceans and insect larvae. Fish were fedtwice daily to satiation by apportioning roughly the same amount of food to each aquarium. After sixmonths of growth, fish were brought into reproductive condition by taking them through an approximatethree month period of low temperature and short day length (4°-10° C; 8L:16D) followed by a gradualreturn to the original temperature and light regime.The crosses I made were pure benthic, pure limnetic, F1 hybrid, F2 hybrid, limnetic backcross, andbenthic backcross, using the design and sample sizes shown in figure 1.1. No sibs were ever crossed, andfamilies were not sampled more than once for each type of cross. Hence little or no inbreeding occurred.The first generation was made in 1992 using wild caught fish from Paxton Lake. I used only lab-rearedindividuals to make the second generation in 1993.For the second generation I did not have sufficient resources to carry out all reciprocal crosses. Itherefore used only male hybrids when making backcrosses with pure species (fig. 1.1). An initialkaryotypic study (Chen and Reisman 1970) suggested that neither sex is heterogametic in G. aculeatus,but the resolution of the study was low. The same study suggested that males are heterogametic in G.wheatlandii, the closest relative of G. aculeatus. I therefore assumed that if either sex were heterogametic,it was more likely to be the males. If males are heterogametic then hybrid males are more likely than14females to show reduced fertility and viability (an effect known as Haldane’ s rule). If males are notheterogametic then my crossing design will not bias the results.Measuring components offitnessThe null hypothesis of all the lab experiments was that hybrids would not be inferior to limneticsand benthics when raised in the laboratory. I examined six components of fitness which could influencehybrid disadvantage.fertilization success.—Twelve to 24 hours after making a cross I checked for fertilization.Fertilization was scored under a dissecting microscope when the egg plasma membrane had separatedfrom the shell, and cell division was clearly visible. I scored fertilization success as the percent (of 30eggs) which had been fertilized. I used an arcsine square root transfonnation, and a one-factor analysis ofvariance to test for the effect of cross type on fertilization success.hatch success.—Hatch success is the fraction of fertilized eggs hatched. I used an arcsine squareroot transformation, and a one-factor ANOVA to test for the effect of cross type on hatch success.growth rate.—I consider growth as a fitness component because size has important fitnessconsequences in fish. Larger individuals usually have higher over winter survival (Shuter and Post 1990;Conover 1992), greater nesting success (Dufresne et al. 1990), greater mating success (Wiegmann et al.1992), higher fecundity (Baker 1994; Schiuter 1 995b; fig. 1.4), and earlier breeding (Miller and Storck1984; Reznick and Braun 1987; Schultz et al. 1991; Wiegmann et al. 1992; personal observation).Eighteen weeks after fertilization I weighed each family, and calculated a per individual mass.Since all crosses hatch at approximately the same very small size, mass at 18 weeks is an estimate of totalgrowth to that date.Mortality during embryo development, and shortly after hatch, meant density varied among tanks(i = 22.52, s.d. = 4.56) though approximately the same range of densities occurred for each cross type.15Growth rate showed a strong nonlinear relationship with density. I therefore corrected for density byusing residuals from a locally-weighted regression (lowess algorithm, S-plus [StatSci 1991]) of growthrate on density. I then used these residuals to test for the effect of cross type on growth rate.fecundiry.—I examined whether hybrid females might be at a disadvantage reproductively in theirability to carry and provision eggs. If they were at a disadvantage, this should show up in the relationshipbetween fecundity and body size. I tested the hypothesis that F1 hybrids are equivalent to pure species inrelationships of fecundity versus body size by measuring the length of lab-reared females and relating itto the number of eggs produced in a clutch. I measured only a female’s first or second clutch. I thencarried out an analysis of covariance on in-transformed egg number and body length to test whetherslopes and intercepts varied significantly among limnetics, F1 hybrids, and benthics.fluctuating asymmetry.—Asyrnrnetry results from differences in expression of characters on theleft and right side of a bilaterally symmetrical organism. The asymmetry is said to be fluctuating whenthese differences are normally distributed around a mean of zero. I measured fluctuating asymmetry inthree gill raker traits: length of longest gill raker on the first gill arch, gill raker number on the short armof the first gill arch, gill raker number on the long arm of the first gill arch, and one external meristic trait,lateral plate number. I chose the traits because of their functional significance for feeding and protectionfrom predators (Reimchen 1983, 1988; Sanderson et al. 1991; Schluter and McPhail 1992).Asymmetry was calculated as the value of the left side minus the right side. To test whether theasymmetry differed in directionality among crosses I used a one-factor ANOVA to test for the effect ofcross type on mean asymmetry. To test whether variance in asymmetry differed among crosses I usedLevene’s test, a one-factor ANOVA on the absolute value of the asymmetry measures. This procedure ishighly sensitive for detecting even small differences in asymmetry among groups (Palmer and Strobeck1986). Asymmetry values were not corrected for body size since I found no significant relationshipbetween asymmetry in gill raker length and body length (R2 = 0.01, F1240 = 2.583, p = 0.109).16parental care.—Male sticklebacks are the sole providers of parental care. In the wild they mustvigorously defend their nests against predators (mainly other sticidebacks), and ventilate the eggs withfrequent fanning by the pectoral fins. Eggs without proper ventilation fail to develop (van den Assem1967; Sargent and Gebler 1980).I tested the hypothesis that F1 hybrid males provide adequate parental care by examining theability of a small sample of males to care for fertilized eggs. I isolated males in halved 100 L aquaria heldat constant temperature and light (200 C; 16L:8D). If a male built a nest within 7 days I supplied him witha single gravid female. When the female had spawned, and the male had fertilized the eggs, I removedher. The male was fed daily, but otherwise left undisturbed for seven days. I then removed all eggs fromthe male’s nest and preserved the clutch in 2% formalin for measurement under a dissecting microscope. Icompared the development of embryos to that reported in Swarup (1958). I scored an embryo asimproperly or poorly developed if the optical cup was not fully formed in one or both eyes.Gravid benthic females spawned in a male’s nest only rarely in the lab, so I present results fromlimnetics and F1 hybrids. Sample sizes are small even for these due to the low availability of gravidfemales at the time this experiment was conducted.combinedfitness measure .—Fitness is often determined by several components. Therefore, inaddition to measuring the above individual fitness components I combined fertilization success, hatchsuccess and growth rate to give a single combined measure. To do this I considered only families forwhich I had data for all three components. Fertilization and hatch success were first arcsine square roottransformed. To give each measure equal weight I scaled all three measures to mean 0 and variance 1before summing. The family sums were then analyzed using a one-way ANOVA.17RESULTSfertilization success.—The data gathered on fertilization success provide no support for thehypothesis that F1 hybrids (males and females) are less fertile than parental species. Fertilization rateswere very high for all cross types, and were below 99% only for benthic backcrosses (table 1.1). A one-way ANOVA detected no significant difference among crosses (Fs,io = 1,72, p = 0.14).This method may overestimate fertility by providing many more sperm to fertilize a clutch ofeggs than would be deposited by a male under natural conditions. For example, if sperm from hybridmales are poor swimmers, or if eggs from hybrid females are not as receptive as those from pure species,then my method might not detect it. This experiment therefore provides an upper estimate of fertility.However, results from the parental care experiment (see below) suggest that hybrid males and femalesusually obtain high levels of fertility under more natural circumstances. I conclude that there are nodifferences in the fertility of limnetics, benthics and F1 hybrids.hatch success.—Hatch success was very high for all types of crosses except benthic backcrosses(fig. 1.2). The effect of cross type was highly significant in an ANOVA of hatch success (F5,101 = 4.10, p =0.002), but Tukey’ s HSD post-hoc tests suggested that the effect was entirely due to low hatch success ofbenthic backcrosses.All cross types had at least one clutch with less than 80% hatch and several of the backcrosseshad very high hatch rates. Backcrosses had the smallest sample size, and their observed inferiority maysimply be due to chance, though the ANOVA suggested otherwise. If F1 hybrid males mated to purebenthic females give rise to poorly developing embryos, then this effect would also be expected within F2hybrids. I did not observe such an effect. F2 hybrids were among the most successful of cross types andhad the largest sample size. Even a subtie effect of hybrid inferiority should have been detected in the F2s.Given the generally high hatch rates of other crosses then, I conclude that hatch success is not affected by18the type of laboratory cross, and does not follow Haldane’ s rule. This conclusion agrees with McPhail’ s(1992) results on hatch success, though he did not carry out the backcrosses.growth rate.—As a null model I assumed growth rate to be a quantitative trait with additivegenetic variance (and no dominance, epistasis or maternal effects). With this assumption F1 and F2 hybridsare expected to have mean growth rates intermediate between the parental species, and backcrosses tohave growth rates intermediate between F1 hybrids and the parental species. This null model is equivalentto a regression of growth rate on cross type, where cross is coded as -1 for limnetics, 1 for benthics, 0 forF1 and F2 hybrids, -½ for limnetic backcrosses, and ½ for benthic backcrosses. Figure 1.3 shows thisregression with a 95% confidence interval around it. The regression is statistically significant (F1,63 =5.78, p = 0.0 19) and the means of each cross type are well within the confidence limits of the regressionline indicating a reasonable fit to the data. However, there remained a tendency for F1s to grow faster andbackcrosses to grow slower than expected, so I carried out further analysis to test for hybrid breakdown orvigor.I carried out a regression of growth rate on the coded values of the pure species (i.e., limnetics =-1 and benthics = 1 respectively). Using this regression equation I then calculated the expected means forthe F1s, F2s and backcrosses and calculated the residuals of the observed from the expected growth rates. Ithen carried Out an ANOVA on the residuals to examine whether there was a significant differencebetween pure species as a group and each class of hybrids. The ANOVA was significant (F4,61 =2.91, p =0.028), but Tukey’ s FISD post-hoc tests suggested that the effect was entirely due to the difference ingrowth rate between F1s and benthic backcrosses (i.e., either hybrid vigor in the F1s, hybrid breakdown inthe benthic backcrosses, or both). No other pairwise comparisons were significant.fecundiry.—A strong overall relationship existed in females between clutch size and body length(fig. 1.4). Analysis of covariance suggested no significant difference in slopes among limnetics, benthics,19and F1 hybrids (F2,64 = 0.469, p = 0.63). Likewise there was no significant difference among intercepts(F2,64 = 0.667, p = 0.52). Thus, the data do not suggest a fecundity disadvantage in hybrids.fluctuating asymmetry.—For each of the four traits no difference among crosses was detected inmean asymmetry (table 1.2; fig. 1.5). Likewise, using Levene’ s test there was no difference amongcrosses in variance of asymmetry (table 1.2). A non-parametric Kruskal-Wallis test confirmed the resultfrom the Levene’ s test. (In table 1.2 I present only results from the Levene’ s test.) These results provideevidence that developmental instability in hybrids (F1s,F2s, and backcrosses) is not greater than that ofparental species, at least under lab conditions.Asymmetry is known to be expressed to a greater extent under environmental stress (Parsons1990) and might thus be greater in the wild than in the lab. For comparison I therefore included a sampleof wild limnetics and benthics in figure 1.5. (I did not include these fish in the initial statistical analysessince they were raised in the wild, and some were much larger and presumably older than the lab-rearedindividuals.) Asymmetry in gill raker length is greater in wild benthics than in lab-reared benthics, thoughwild and lab-reared limnetics had similar asymmetry (fig. 1.5). When wild fish were included in theanalysis, Levene’ s test indicated a significant difference among groups for both plate number (F7,310 =2.165, p = 0.037) and gill raker length (F7,310 = 14.674, p < 0.001). Rearing in the wild appears to affectasymmetry.parental care.—Sample sizes for the parental care experiment are small yet provide preliminaryevidence that F1 hybrid males are not inferior parents (fig. 1.6). Of the seventeen males I used in theexperiment only one had a clutch with fewer than 80% of embryos properly developed—a limnetic male.Was this due to the parental care provided by the male, or inviable gametes from the female hybrid?Hybrid females used for artificial lab crosses did not show a detectable reduction in gamete viabilityrelative to limnetic females, so it seems more likely that abnormalities in embryo development were dueto inept care of the eggs. All imperfect eggs appeared to have been fertilized, so poor development was20most likely due to insufficient ventilation of the eggs by the male. Taken together the results of theparental care experiment suggest that hybrid males are capable of stimulating females to spawn, andfertilizing and raising their offspring.combinedfitness measure.—A one-way ANOVA suggested an effect of cross type on thecombined fitness variable (F5,6 = 4.86, p = 0.001). As in the fitness components above this effect camelargely from benthic backcrosses. The only significant pairwise comparisons (Tukey’ s HSD test) were ofbenthic backcrosses versus benthics, F1s and F2s. These latter three were not significantly different fromone another or any of the other crosses. I interpret this as support for hybrid breakdown in the benthicbackcrosses, but no support for F1 hybrid vigor. However, this result should be accepted with caution dueto unequal variances among cross types (table 1.3).DISCUSSIONEvaluating the hypothesis of ecological speciation is a several stage process. Data presented herelay the foundation for further examination of the hypothesis by demonstrating that limnetic and benthicsticklebacks froni Paxton Lake can produce viable hybrid progeny that have a fitness similar to theparental forms in a laboratory setting. Support for ecological speciation requires that there be ecologicalmechanisms of selection against hybrids, and that the strength of this selection be strong relative toselection by other mechanisms.I evaluated six mechanisms which could place hybrids at a disadvantage relative to parentalspecies. Four of the six fitness components showed no difference between hybrids, limnetics andbenthics. F1 and F2 hybrids had a mean combined fitness greater than limnetics and benthics (table 1.3).Reduced fitness in hybrids was apparent only for growth rate and hatch success in the backcrosses. Yetthe magnitude of this effect suggests that it may be of little biological importance in any case. The data21provide at most weak evidence of hybrid breakdown and my general conclusion reiterates that of McPhail(1992): hybrids have a fitness remarkably similar to the parental species. I now briefly discuss theevidence for hybrid breakdown and implications of these findings.F1 vigor and backcross inviability.—There was a significant difference among crosses for hatchsuccess, growth rate and the combined fitness measure. Further analysis of hatch data suggested an effectfrom backcrosses having lower success than other crosses. And further analysis of growth rate providedsome support for either F1 hybrid vigor, hybrid breakdown in benthic backcrosses, or both. Benthicbackcrosses were the least fit of all cross types in terms of hatch success (fig. 1.2), and growth rate (fig.1.3). When three of the fitness measures were combined there continued to be support for lower fitness ofthe benthic backcrosses.The hatch data are perplexing because low hatch success was limited (statistically) to a singletype of cross. In addition, all cross types had clutches with both low success (< 80%) and high success(>90%). Benthic backcrosses should be further studied to see whether this effect is repeatable and whatthe mechanism leading to inferiority may be.Backcrosses also grew somewhat slower than other crosses. Although statistically significant inan analysis of variance I am somewhat skeptical that the effect of cross type is biologically real. First, ifgrowth rate is determined genetically and there has been substantial genetic divergence (andincompatibility) between limnetics and benthics, then growth rate should show greater variance in F2hybrids than in the F1s (i.e., some F2s should show breakdown as in the backcrosses, and some shouldshow vigor as in the F1s). Lesser rather than greater variance is the pattern in the data (variance in residualgrowth rate: F1 = 0.00434, F2 = 0.00261; fig. 1.3). Secondly, the results do not fit Haldane’s rifle,assuming that males are heterogametic. Although backcrosses have somewhat lower growth rates thanexpected, growth rates of F2s fit the null model shown in figure 3 very well, which would not be expectedif Haldane’s rule were operating (since low viability of F1 hybrid males would lower the average growth22rate). Thirdly, genetic incompatibility between the parental forms would be expected to result in greatervariance ofF2s than F1s for asymmetry and my combined fitness measure (Parsons 1992). This patternwas not observed for asymmetry (fig. 1.5), though was true for the combined fitness measure (table 1.3).Thus, there is only equivocal support for hybrid breakdown in growth rate.However, if further crosses show this effect to be real, then the mechanism leading to inferioritycould be quite novel. To my knowledge, there is no simple genetic model which could explain theobservation of backcross breakdown in the absence of F2 hybrid breakdown. If these effects are real thegenetics of hatch success and growth rate are not operating in an additive fashion. It is possible that thereare maternal effects which are only expressed in pure limnetic or benthic females, though this remains tobe studied. The genetics of this system are covered more fully in chapter four.fluctuating asymmetry.—Fluctuating asymmetry is not commonly considered a mechanism ofreproductive isolation between species, but there are good reasons for considering asymmetry in hybrids.Asymmetry has been shown to be important functionally in the performance of several taxa (Parsons1990; MØller 1991), and appears to be also significant in mate choice (e.g., MØller 1991; Thornliill 1992).If asymmetry reflects performance and attractiveness as a mate then there is also a potential role as anisolating mechanism during speciation.In some previous studies (Palmer and Strobeck 1986), greater asymmetry has been observed inhybrids than in parental species though this was not observed here. Asymmetry in hybrids would implygenetic incompatibility between limnetics and benthics. That fluctuating asymmetry in hybridsticklebacks was no greater than within pure species is further evidence that these two species have highlycompatible genomes. It is nevertheless possible that fluctuating asymmetry will be expressed to a muchgreater degree under the different stresses of the wild (fig. 1.5). For example, sticklebacks that weremoved to a water body where they had not previously lived showed increased asymmetry (Zakharov 198923cited in Parsons 1992). If hybrids raised in the wild have greater asymmetry than that expressed in the labthey may suffer lower fitness as a result. This remains to be tested.conclusions and implications.—The mechanisms of postmating isolation I examined in thischapter were straightforward, yet diverse. Many of these isolating mechanisms have been shown to existbetween other species (e.g., Mayr 1963; Dobzhansky 1970; Coyne and Orr 1989; Coyne 1992). Takentogether my data provide strong evidence that these postmating isolating mechanisms operate onlyweakly between Paxton Lake limnetics and benthics, and that the genomes of the two species are highlycompatible. The detectable frequency of adult F1 hybrids in the wild suggests potential for gene flowbetween the two species. If there is even a very small amount of hybridization, then the present resultssuggest that hybrids would have at most a slight disadvantage which should still allow substantialexchange of selectively neutral or mutually beneficial alleles (Barton and Bengtsson 1986).Are there other intrinsic factors which might place hybrids at a disadvantage and thereby preventgene flow? Although I have examined most of the major potential postmating isolating mechanisms thereare some factors which I have left untested. Yet, further sifting for a mechanism which would operate inthe lab would likely be fruftless: limnetics and benthics are not only interfertile, but capable of giving riseto hybrid offspring which are highly viable, show no developmental abnormalities, and are fertile.It might be suggested that the absence of strong isolating mechanisms in the laboratory isevidence that these are not true species. But the lab has conditions very different from the wild. Thehypothesis of ecological speciation predicts that limnetics and benthics would not exist in the prolongedabsence of ecological selection pressures, or under a substantially different set of ecologicalcircumstances. The absence of strong isolating mechanisms in the laboratory should strengthen a test ofecological speciation, since the strength of evidence hinges on the difference between hybrid fitnessmeasured in the lab and fitness measured in the wild. Whether these sticklebacks are ecological species isexamined further in chapter two in which I examine F1 hybrid fitness in the wild. I suggest that theabsence of these postmating isolation mechanisms is not unique to sticklebacks, and that the study ofother closely related species can provide further tests of ecological speciation.2425TABLE 1.1. Fertilization success of crosses carried out in the laboratory. Parental types are shown inparentheses.cross no. of i proportion s.e.families fertilizedlimnetic (LxL) 17 0.994 0.006limnetic backcross (LxF1) 10 0.997 0.003F1 hybrid (LxB) 24 1F2 hybrid (F1x) 35 0.998 0.001benthic backcross (BxF1) 8 0.988 0.009benthic(BxB) 16 126TABLE 1.2. Analyses of variance testing for the effect of cross type on asymmetry.mean asymmetry variance in asymmetrygill raker numberonthe long arm F5,6= 1.590 p=O.l64 F5,236 =0.818 p=0.538gill raker numberon the short arm F5,236 = 1.156 p = 0.332 F5,236 = 0.38 1 p = 0.86 1log(gill raker length) F5,6 = 2.009 p = 0.078 F5,6 = 0.763 p = 0.577plate number F5,236 = 0.782 p = 0.564 F5,236 = 1.600 p = 0.16127TABLE 1.3. Combined fitness measures for lab-reared crosses. Fertilization success and hatch successwere first arcsine square root transformed, then along with growth rate the three measures were scaled tomean 0 and variance 1 and summed. Note that the mean fitness values have not been standardized to fallbetween 0 and 1.n sdbenthic 6 0.680 1.700benthic backcross 7- 2.489 1.755F1 hybrid 5 0.997 1.153F2 hybrid 32 0.535 1.644limnetic backcross 7- 0.873 1.494limnetic 5 - 0.527 3.07728FIGURE 1.1. Design used to make crosses for this study. For the first generation ten families of each crosstype were made. n refers to the number of second generation families.29LfemaleB2nd generation (lab-reared parents)1St generation (wild parents)maleL BLLBLLBBBLLLBfemaleBLBBmaleLL LB BL BBn=6 n=4 n=4 n=2n=8 n=9n=8 n=8n=6 n=4 n=4 n=630FIGURE 1.2. Boxplots of hatch success from laboratory crosses. Wliiskers show the range of data, the boxshows the first and third quartiles, and the median is shown by the horizontal line within the box.0.8C.) 4-, 4-, C Cl) C.) L. U)0.60.5limneticlimneticFlF2benthicbenthicbackcrosshybridhybridbackcross32FIGURE 1.3. Comparison of mean family growth rates of six cross types reared in the lab on identicaldiets. Values are residual mass from a locally weighted regression. The dotted line shows the nullexpectations if growth rate is a quantitative character with strictly additive genetic variance. The linerepresents a regression of residual growth rates on cross type (where limnetics = -1, limnetic backcrosses= -½, F1 and F2 hybrids = 0, benthic backcrosses = ½, and benthics = 1). The solid lines are simultaneousconfidence intervals for the fitted y-values. Points are i ± s.e.0.2.2C 0 Cl) Cl) a) c,) a) I 0 0 E 0 Cl) Cl) a)0.1-0.0--0.1-0.2++.limneticlimneticbackcrossFlandF2hybridbenthicbackcrossbenthic34FIGURE 1.4. Fecundity versus length relationship for lab-reared female limnetics (1), benthics (b) and F1hybrids (h). The line represents a regression through all points.a) N Cl).C).- C) c,) 1.5. Asymmetry in gill raker length for eight groups of fish. Positive values occur when gillrakers on the left side are longer than on the right side. Boxplots show ranges, quartiles and medians ofthe data.0.6-c c,)C a) G)-Cu C ci) E E Cu0.40.2-0.0-0.2-IIIIIIIIIriIIIIll’I’llIIIII--wildlablimneticFlF2benthiclabwildlimneticlimneticbackcrosshybridhybridbackcrossbenthicbenthic(n=37)(n=33)(n=30)(n=90)(n=32)(n=29)(n=28)(n=39)38FIGURE 1.6. Parental care provided by limnetic and F1 hybrid males. Points show the percentage ofembryos from a single clutch which have properly developed (according to descriptions by Swamp[1958]) over a seven day period . Points have been jittered to avoid overlap.1.0-...........0.8-t3 ci) 0 ci)0.6> ci) 0 4-. C ci) C) ci) ci0.20.0-limneticmalelimneticmalehybridmalehybridmalewithwithwithwithlimneticfemalehybridfemalehybridfemalelimneticfemaleChapter Two EcoLoGIc&i. SPEcIATI0N: NIcHE-DEPENDENT GRowTh RATES OF F1 HYBRIDS40INTRODUCTIONOne of the most striking features of adaptive radiations is that closely related species often differgreatly in their use of resources. Yet the role of resources as a force driving speciation has not beenwidely accepted. In this chapter I provide experimental evidence for disruptive natural selection acting onresource exploitation efficiency, and propose that this has been a principal force in the speciation offreshwater threespine sticklebacks in British Columbia, Canada. I also suggest that such “ecologicalspeciation” may be much more common than is realized and discuss several examples where it seems tohave occurred elsewhere. I view ecological speciation as being potentially important whether a particularspeciation event began in allopatry or occurred entirely in sympatry.The hypothesis of ecological speciation—speciation driven by niche-based selectionpressures—was implicit in many earlier writings (e.g., Simpson 1944). The term itself came from adiscussion by Van Valen (1976) who applied the concept of “ecological species” to ecologicallydifferentiated taxa that hybridized yet maintained their distinctiveness over evolutionary time. Becausethey hybridized these taxa would normally be denied species status under Mayr’s (1942, 1963) biologicalspecies concept; Van Valen’s ecological species concept included these as species because theymaintained their distinct features despite hybridization, suggesting that hybrids experience an ecologicaldisadvantage. This view of species was atypical because it ignored gene flow and focussed instead onecological mechanisms of reproductive isolation. According to Van Valen, species are discrete unitsbecause natural selection removes forms that are intermediate.The ecological species concept itself had many disadvantages over the biological concept, and itproved too unwieldy to be useful for classification (Ridley 1993). However, it remains an explicithypothesis of speciation (Ridley 1993). This hypothesis is more than just a restatement of niche theory.The idea is that speciation could be driven by ecological adaptation to different niches (Schiuter 1995a).4142This view builds on the earlier ideas of adaptive zones (Simpson 1944, 1953), the adaptive landscape(Wright 1932), and the ecological niche (Hutchinson 1957). Taken together the ceniral theme is thatorganisms have certain attributes to exploit resources, or to avoid parasites and predators. Those that areintermediate for such attributes occupy valleys in the adaptive landscape, and are selected against. Thehypothesis of ecological speciation applies to any niche-based selection but I have restricted my study tothe ways in which food resources are used by organisms.Resources vary in quality and local abundance, and strategies to efficiently exploit one resourceoften preclude efficient exploitation of another. This trade-off in exploitation efficiency is the key tounderstanding the role of resources in speciation because it leads to selection between taxa for adaptationto different resources and selection against intermediate phenotypes. I refer to the products of this processas “ecological species”. The biological species concept is still easily applied to these species if one relaxesthe definition to allow some hybridization (Butlin 1995; Schluter 1995a).The traditional focus of speciation studies has been gene flow (Ridley 1993): selection may favordivergence, but the homogenizing effect of gene flow between populations prevents divergence. Rice andHostert (1993) recently reviewed forty years of published results from lab experiments on speciation, andshowed that there is substantial evidence for morphological and ecological divergence betweenpopulations connected by gene flow. The general conditions producing speciation (or substantialdivergence) in the lab were restricted (though not necessarily zero) gene flow, with strong, divergentselection on diverse characters. What we need now are studies in nature which examine how often theseconditions are met, and the mechanisms by which selection operates. This is well within the range ofexperimental investigation. In this report I do not argue the merits of alternative views concerning theimportance of reduced gene flow to speciation. Rather, I focus on the ecological mechanisms that causedivergence, and that allow coexistence of taxa in sympatry in the absence of perfect premating isolation.43Three pieces of evidence are necessary to test whether speciation has been due to resource-basedselection: (1) a demonstration of disruptive natural selection acting between species, (2) evidence that themechanism by which selection operates is resource exploitation efficiency, (3) evidence that prematingisolation evolved as a direct or correlated response to divergent natural selection. In chapter one Iexamined the first two criteria by assessing fitness of hybrids in the lab. In this chapter I extend myassessment of hybrid fitness by comparing fitness in the wild of F1 hybrids, limnetics and benthics. Futurework with these sticklebacks will examine the role of ecological processes in the evolution of prematingisolation, but a general discussion of ecological speciation and premating isolation can be found inSchluter (1995a) and Schluter and Nagel (in press).A demonstration of selection against hybrids provides the best evidence of disruptive selectionbetween species. The difference between fitness of hybrids in the wild and in the lab measures theecological contribution to hybrid inferiority. There are exceedingly few tests of this phenomenon, yetsuch tests are not out of reach. Many species in nature are interfertile (see for example, Hubbs 1955;Grant and Grant 1992; Coyne 1992), and viable hybrids can be raised by artificial fertilization in the lab.Hybrids can be used in experimental studies of natural selection to determine whether ecologicalmechanisms of selection operate.Test ofEcological Speciation in SticklebacksI now briefly outline my test of the hypothesis that sympatric species of sticldebacks areecological species. The natural history and prior experimental evidence which led to the hypothesis ofecological speciation in sticklebacks is reviewed in chapter one, but specific details are provided here tohelp set up my test.Several aspects of the biology of the species pair have led to the hypothesis of ecologicalspeciation. (1) Schluter and McPhail (1992) provided evidence that limnetics and benthics are character44displaced: trophic apparatus of limnetics and benthics are morphologically more different than would beexpected by chance. (2) There is experimental evidence that competition promotes morphologicaldivergence (Schluter 1994). (3) Schiuter (1993, 1995b) demonstrated a sharp trade-off in feedingefficiency and growth of species between the two habitats. (4) F1 hybrids have an intermediate phenotype(McPhail 1992; table 2.1). (5) Despite some evidence of breakdown, hybrids (F1s, F2s and backcrosses)raised in the lab have fitnesses remarkably similar to the parental species (McPhail 1992; chapter one).However, these results were not a test of ecological speciation due mainly to threemethodological shortcomings. First, both the feeding and growth rate experiments used F10 hybrids ratherthan F1 or F2s: Schiuter (1993, 1995b) did not know whether the performance ofF1 hybrids might bedifferent from F10 hybrids. Second, the sample size in the growth experiment was only n = 6 hybrids ineach habitat. The conclusion of hybrid disadvantage was therefore tentative due to the small sample size.Third, wild-caught fish were used that probably had different amounts of prior experience foraging in thetwo habitats. The hybrids he used were raised in a different lake from the limnetics and benthics, andinitial foraging experience was different for all three types. This difference in experience may haveaffected performance in alternative habitats; some morphological traits of these species are phenotypicallyplastic and respond in an adaptive direction to diet experience (Day et al. in press).The experiment I present here uses a large sample size of F1 hybrids, limnetics and benthics thathad very similar prior experience. All fish used in the experiment were raised in the lab on a commondiet.F1 hybrids are intermediate between the parental species in all morphological traits I havemeasured. For example, they possess an intermediate number of gill rakers, which are of intermediatelength (table 2.1). Because hybrids have an intermediate morphology I hypothesized that they would beselected against through resource use efficiency: they would do worse than limnetics when exploiting45plankton, and worse than benthics when exploiting benthos. Plankton and benthos are the two major foodresources available to sticklebacks.Hybrid fitness has been extensively studied in the lab (McPhail 1992; chapter one). In general,the results showed that limnetics and benthics have remarkably compatible genomes. Figure 2.1 presentsa subset of lab growth rate data from chapter one. Families of benthics, limnetics, and their F1 hybridswere reared in the lab and their growth rates measured. The null expectation ofF1s to grow at a rateintermediate between the parental species (fig. 2.1) was not rejected statistically (chapter one).In the study presented here I tested one aspect of the hypothesis of ecological speciation bycomparing growth rates of pure and F1 hybrid fish when raised in the wild. Since fish have indeterminategrowth, the rate at which they grow has important fitness consequences for both sexes. Fitness is oftenassociated more with size than with age: larger individuals usually have higher over-winter survival(Shuter and Post 1990; Conover 1992), greater nesting success (Dufresne et al. 1990), greater matingsuccess (Wiegmann et al. 1992), higher fecundity (Baker 1994; Schluter 1995b; chapter one), and earlierbreeding (Miller and Storck 1984; Reznick and Braun 1987; Schultz et al. 1991; Wiegmann et al. 1992;personal observation).MATERIALs AND METhoDsGrowth Rates in the FieldI tested the null hypothesis that growth rates of F1 hybrids are equal to pure species in the naturalecological setting of Paxton Lake. The field experiment was carried out in April and May 1993. Resultsfrom experiments done at this time of year suffer the fewest logistical constraints: enclosures in the openwater have the fewest benthic organisms settling on the mesh, and fish held in inshore enclosures have thehighest survival rates (Schluter 1995b).46I used 24 benthics taken from 10 families, 24 limnetics from 7 families, and 48 F1 hybrids from13 families (6 L x B d families, 7 B x L d’ families). All fish were approximately 10 months old, andcame from artificial crosses raised in the lab (see chapter one for rearing methods). All the fish used inthis experiment therefore had similar initial experience before being placed in the field.Growth in fish is indeterminate, but size tends to plateau as they get larger, especially as somaticdevelopment is diverted to reproduction. I therefore chose fish that were large enough to be retained inmesh enclosures, but small enough to have potential for substantial growth. None of the fish used showedsigns of becoming reproductive at the start, though some became reproductive by the end of theexperiment. The inclusion or deletion of gravid females did not alter patterns in the data. Hybrids hadmorphologies intermediate between those of the pure species (table 2.1). Nevertheless, characterdistributions for the three forms showed substantial overlap (table 2.1).Experimental HabitatsI confined fish to one of two habitats using enclosures. Enclosures for the open water habitat werecylindrical in shape, 1 m in diameter, 6 m in depth, and made of 6 mm knotless nylon mesh. They weredesigned to replicate a planktonic environment by allowing plankton to drift through the mesh, and allowdiel vertical migration of prey. I suspended 24 of these enclosures from two rafts anchored in the deepestpart of the lake. In addition, 24 enclosures placed in the littoral zone represented the benthic environment.The littoral zone enclosures had open bottoms, 1 m x 1 m, with sides made of the same nylon mesh. Iplaced these along the edge of the lake in undisturbed littoral substrate at a depth of 1-1.5 m, and wereemptied of wild fish prior to the experiment.Before I placed sticklebacks in the enclosures I weighed them and individually marked them byclipping one of the dorsal spines. The fish were kept in the enclosures for three weeks, removed andweighed again, then anaesthetized, preserved in formalin, stained with alizarin red, and stored in47isopropyl alcohol (Schluter and McPhail 1992) for later morphological and diet analysis. Enclosuredimensions, placement in habitats, and length of the experiment were the same as Schiuter (1995b).In the open water I assigned two hybrids to each of twelve enclosures, and two limnetics to theother 12 enclosures; in the littoral zone I assigned two hybrids to each of 12 enclosures, and two benthicsto 12 enclosures. Thus, the experiment compared growth of hybrids in each habitat to the single parentalspecies that predominantly forages there. Growth rate (mg day-1)was estimated simultaneously in thelittoral zone and open water. Since each enclosure contained two fish, their average growth rate was usedas a single independent data point.I deleted enclosures from the analysis if one or both the fish had died or was wounded. This gavea total of 20 littoral enclosures and 19 enclosures from the open water habitat.To confirm that enclosures were replicating the appropriate habitats I examined stomach contentsof fish from each habitat (all fish from the littoral habitat—i.e., 22 benthics and 18 hybrids; and due to themore numerous stomach contents of fish from the open water, a random subset of 12 limnetics and 12hybrids). I keyed prey items to family, and measured the width of each item. Width is the best preserveddimension of items in the gut. Differential passage of items through the gut may bias diet analyses, so Irestricted counts to stomach contents to minimize this bias (Schluter and McPhail 1992).My results were similar to those of Schluter’ s prior study (1995b) in which he used F10 hybrids. Itherefore combined my results with his to produce a single overall estimate of the fitnesses of limnetics,benthics and hybrids. I used a three-way analysis of variance (with species, habitat and year as effects) toobtain estimates of means and variances of growth rates corrected for differences between years. I thencombined these estimates into a single fitness set, which is a two-dimensional plot of an adaptivelandscape. A fitness set shows the extent of selection for resource specialization by plotting a measure offitness on one resource versus the same measure of fitness on an alternative resource (Levins 1968).48RESULTSGrowth Rates in the FieldThere was no evidence for F1 hybrid inferiority in the lab (fig. 2.1; chapter one). The picture wasquite different when hybrids were transplanted from the lab to the wild. Hybrids had a lower mean growthrate than the parental species when raised in either habitat in Paxton Lake (fig. 2.2). Hybrids grew at 73%the rate of benthics (on average, 8.57 mg/day more slowly) in the littoral zone, and at 76% the rate oflimnetics (on average, 3.42 mg/day more slowly) in the open water. Hybrids showed this disadvantagedespite conditions which were excellent for growth. For example, mean growth rate for benthics was3 1.53 mg/day which was 50% higher than mean benthic growth rates observed in two previous years(Schluter 1995b). Mean growth rate of limnetics was slightly higher than mean rates in these previousexperiments (Scliluter 1995b).The difference in hybrid growth rates between the littoral zone and open water habitats wassubstantial—hybrids grew at more than twice the rate inshore. These results agreed with the earlierexperiments in that the open water environment appears to afford lower growth rates, at least at this timeof year (Schluter 1995b). Wild hybrids would grow faster in the benthic foraging habitat than in the openwater or a mixture of habitats.I was interested in how well hybrids performed relative to both parental species, so for statisticalanalysis I compared hybrids from both habitats to limnetics and benthics as a group (“pure species”). Ithen conducted a two-way analysis of variance of group (hybrid vs. pure species) and habitat (littoral vs.open water) on ln-transformed growth rates. Hybrid disadvantage is supported by the results—the effectsof both species and habitat are highly significant (table 2.2). The lack of a significant interaction meansthat the magnitude of hybrid disadvantage is approximately equal in the two habitats on a logarthmicscale.49These findings were similar to those of Schluter (1995b) on F10 hybrids. This allowed me tocombine my data with his to produce an overall assessment of hybrid performance. Hybrid disadvantageis strongly supported by the fitness set constructed using the combined data (fig. 2.3). The dotted lineconnecting benthics and limnetics is a useful standard for comparing the performance of hybrids. The linedescribes a boundary along which hybrid individuals would have the same growth rate relative to the purespecies if all individuals were to select habitats at random. Because the mean for hybrids falls below thisline it implies selection against hybrids, even if the two species and hybrids were to choose habitatsrandomly. The fitness set therefore shows the minimum selection against hybrids (excluding the highlyunlikely scenario of each choosing the habitat where they grow most slowly).Analysis ofDietsTable 2.3 is a compiled list of stomach contents of fish from each habitat. The diets of hybridsand pure species were very similar in each habitat, as was the size of prey taken. Individuals from theopen water treatments selected predominantly planktonic prey; fish from the littoral zone enclosuresselected predominantly benthic prey. This suggests that the enclosures were replicating the appropriateenvironments. Since hybrids exploited the same resources as the pure species it implies that a nicheintermediate between the open water and littoral niches may not exist—hybrids would have to exploit oneor other of these niches.DISCUSSIONAre these Ecological Species?Ecological speciation occurs when reproductive isolation evolves as a consequence of divergentselection for efficient resource exploitation. Ecological species are the outcome of this process. My results50suggest that limnetics and benthics are ecological species, since postmating isolation appears to be mostlydetermined by the ecological setting. This result marks the first demonstration of a case where hybridshave a fitness similar to parental species when raised in the lab (McPhail 1992; chapter one) butsubstantially lower fitness when raised in the wild.Earlier studies demonstrated that Paxton Lake benthics and limnetics should be consideredbiological species despite some hybridization in the wild: they breed true in the lab, show strongassortative mating, and field observations show them to be reproductively isolated (McPhail 1992, 1993,1994; Nagel 1994). But lab studies have failed to detect any convincing evidence of substantial hybriddisadvantage (McPhail 1992; chapter one). These species are capable of exchanging genes, and may doso: since adult hybrids (presumptive F1s) have been found in the wild and F1 hybrids are reproductivelyviable in the lab, it seems likely that hybrids have at least some reproductive success. If this is true, thenfor these two morphologically divergent species to be maintained in sympatry, some form ofenvironmental selection must act against the hybrids.F10 hybrids are less efficient feeders on plankton than limnetics, and less efficient feeders onlittoral prey than benthics and this is best explained by their intermediate phenotype (Schluter 1993). Thisreduced ability to exploit resources is the most likely mechanism for selection against F1 hybrids in thewild. Other than a foraging disadvantage, no mechanism of selection against F1 hybrids has beenidentified. A foraging disadvantage did not operate in the laboratory because exploitation efficiency wasirrelevant to growth rate there—food was freely available and fish were fed to satiation.One alternative hypothesis is that limnetics and benthics are genetically incompatible to someextent, and that hybrids express a disadvantage only under stressful environmental conditions. However,conditions in my field experiments were probably more benign than in the lab—densities weresubstantially lower, and the high growth rates suggest that food was abundant.51The diets of hybrids are very similar to the parental species’ diets when in the parental species’habitat. This suggests it is unlikely that an intermediate habitat exists in which hybrids would performbest, though this is currently being investigated. On the whole my results suggest that resources in PaxtonLake are distributed such that limnetics and benthics each occupy an adaptive peak, separated by a“valley” of lower fitness.Ecological success of sticldeback hybrids should continue to be studied. I observed a 25%reduction in growth rate for F1 hybrids which to some readers may seem insubstantial. However, thefitness set shown in figure 2.3 was constructed from three years of data and shows that the advantage forlimnetic and benthic forms persists over this period. Hybrid disadvantage may be even greater if growthwas measured over a period longer than the three week duration of these experiments. It should also benoted that I have examined many possible mechanisms of selection against hybrids. This is by far thegreatest difference between hybrids and pure species I have found. Additionally, growth rate is only oneof many possible ecological measures of success. For example, hybrids may also be at a disadvantage inescaping predators or parasites. Additional ecological mechanisms of selection against hybrids, and atdifferent stages in their life history, await further study.To this point I have satisfied two of the criteria laid out above as an adequate test of ecologicalspeciation: I have demonstrated disruptive natural se1ection acting between these species, and I havedemonstrated that the likely mechanism for this selection is resource exploitation efficiency. My thirdcriterion is probably the most difficult to satisfy, but there is some evidence that premating isolationbetween limnetics and benthics has evolved as a result of selection for efficient use of different resources.Species pairs of sticklebacks exist in lakes that are nearby to Paxton Lake, but in separatewatersheds (McPhall 1984). Each pair is thought to have evolved independently (McPhail 1993, 1994).Nagel (1994) carried Out a series of mate choice studies and showed that limnetics will mate withlimnetics from a different lake, and benthics will mate with benthics from a different lake. However,52benthics and limnetics are reproductively isolated both within and between lakes. If the species pairs haveindeed evolved independently (work is currently underway on this question) then it appears that selectionfor resource specialization has caused premating isolation to evolve in parallel (Schiuter and Nagel inpress). This pattern of “parallel speciation” seems to not be unique to sticklebacks (Schiuter and Nagel inpress).Is Ecological Speciation Common?There are indications that natural selection on resource use efficiency has played a significant rolein many speciation events, but few studies fully test the hypothesis of ecological speciation. In thissection I review several cases which suggest that ecological speciation may be widespread, and highlightsome of the information still needed.criteria 1 and 2 satisfied.—Studies of the Galapagos finches have best satisfied the criteria whichI laid out in the introduction for testing the hypothesis of ecological speciation. Darwin’s finches arelikely ecological species since hybrids are often formed in the wild (Grant and Grant 1992), they areintermediate in morphology and their success depends on resource abundances (Grant and Grant 1993).The pure species are morphologically specialized and natural selection for efficient resource exploitationhas been repeatedly demonstrated (Grant 1986). Premating isolation exists between the species, but it isnot yet clear how assortative mating evolved.Benkman (1993) indirectly demonstrated disruptive selection between several species of birds(red crossbills; Loxia spp.) and had strong support for an ecological mechanism. He showed that eachcrossbill species had an optimal beak morphology for utilizing one of several major conifer seedresources, and he observed a morphologically induced trade-off between resource use efficiency ondifferent seeds. Morphological specialization for one seed type resulted in less efficient exploitation ofother seed types, rendering intermediate morphologies inefficient on either seed. Benkman hypothesized53that the crossbill “types” (subspecies or species) he tested likely hybridize in the wild, that hybrids shoulddo less well than parental forms, and that diversity of conifer seed crops was ultimately responsible forthe diversification of crossbills. However, studies of hybrids in the wild and of premating isolationbetween crossbill types remain to be done.criteria 1 and 2 partially satisfled.—Other likely cases of ecological speciation are insect hostraces (e.g., Bush 1969; Katakura et al. 1989; Tauber and Tauber 1989). Reproductive isolation betweenhost races is often accomplished through specialization on resources which mature at different times ofthe year. The ecological trait presumed to be under divergent selection is emergence time of the adultinsect from a larval stage. An intermediate emergence time would be selected against because resourcesare less abundant than at an earlier or later time, but how this life history trait is tied to resource useefficiency remains to be assessed directly. F1 hybrids between races of apple maggot flies (Rhagoletis)have been raised in the lab (Feder et al. 1988) and show no evidence of hybrid breakdown, but hybridfitness has yet to be studied in the wild. In the Japanese ladybird beetle (Epilachna) the resources used byeach of two host races are available in the wild simultaneously, and resource use efficiency has beenmeasured. Katakura et aL (1989) showed that survival to second instar was substantially greater whenraised on their own food type versus the other’s diet. In this case, hybrids were produced in the lab anddeclared viable, but the authors did not measure their relative ecological success, either in the lab or in thefield. Whether premating isolation has evolved as a result of resource specialization also remains aquestion. Theoretical models and lab experiments motivated by studies of host races (e.g., Rice 1987;Rice and Salt 1990) show this is possible, but experimental evidence is lacking for specific cases.In many lake and river systems in western North America anadromous sockeye salmon(Oncorhynchus nerica) have independently given rise to nonanadromous, freshwater species calledkokanee (Foote et al. 1989). Sockeye and kokanee appear to occupy different niches, but they spawnsympatrically and are known to occasionally interbreed in nature (Wood and Foote 1990). F1 hybrids are54completely viable in the lab (Wood and Foote 1990; Foote et al. 1992), but the relative perfonnance ofhybrids in the wild is not known. The hypothesized mechanisms of selection against hybrids have beenecological in nature.The near ubiquity of hybridization in plants was the original impetus for Van Valen’s (1976)concept of ecological species. The best example of ecological speciation in plants was provided by Briggs(1962). She studied fitness of buttercup species (Ranunculus spp.) and their hybrids in Australianhighland regions where wet and dry soil conditions form a network of small patches with almost noecologically intermediate habitat. Briggs documented several cases where one species, adapted to wet soilconditions, gave way to another, adapted to dry soil. Putative hybrids were morphologically intermediate,found in the wild, and in the lab showed no evidence of hybrid inferiority. Hybrids were rare (andpresumably selected against) in both wet and dry areas—in the wild they were largely restricted to theremarkably abrupt (sometimes <1 m wide) transitional zones from wet to dry conditions. Niche-basedselection pressures were clearly keeping the species separate. Further work seems warranted for thissystem: the ecological or physiological basis of adaptation to wet or dry conditions is not known, fulltransplant experiments need to be carried out with experimentally produced hybrids, and the pollinationsystem (i.e., premating isolation) needs to be described. Especially interesting would be to apply themethods employed by Arnold and co-workers in an Iris hybrid zone (Cruzan and Arnold 1993; Arnoldand Bennett 1993), to search for associations between genetic markers and ecological conditions.Implications ofEcological SpeciationIf viable hybrids can and are produced between a number of taxa, why is it that they remain rarerelative to parental species? One answer may be that the nature of selection against hybrids is oftenecological (Wade and Kalisz 1990), in which case natural selection would be playing a primary role in themaintenance of diversity in the early stages of speciation, by acting as a source of reproductive isolation.55A second implication of ecological speciation is that the genetics of speciation may be quitedifferent than has previously been suggested (Coyne 1992). For example, when Coyne and Orr (1989)reviewed Haldane’ s rule (sterility of the heterogametic sex in interspecific hybrids) in Drosophila, theircriterion for postmating isolation was hybrid sterility or complete developmental inviability. Fifteen of 81interspecies crosses included in the review showed no such isolation. Essentially the scenario theyexplored was one in which postmating reproductive isolation evolves by the accumulation in allopatry ofgenetic differences which produce some level of hybrid sterility or developmental inviability. Given myresults with sticklebacks, and similar results mentioned above, there seems no reason that such stringentrestrictions need exist for the level of differences which might accumulate in allopatry. If ecologicalspeciation is common, then postmating isolation instead depends on the nature of divergent selection forefficient resource use, and attention should focus instead on those genes which cause hybrids to have anintermediate phenotype. It is conceivable that the interspecific genetics reviewed by Coyne and Orr(1989) are then the study of random accumulation and expression of genetic differences after speciationhas occurred. Nevertheless such genetic differences would ultimately prevent ecological species fromhybridizing if selection pressures changed.The third implication of ecological species and speciation is that species may often be transient orephemeral (cf. Futuyma 1987). If current selection pressures change then a species or group of speciesmay collapse through hybridization. This is relevant to conservation. If species hybridize more frequentlyas they become rare (e.g., Hubbs 1955), the benefits of assortative mating may evaporate as selectionpressures change due to fluctuations in the abundance of resources, competitors and predators. (Thisimplies that ecological speciation might also be assessed with selection experiments.)A final implication is that the rate of speciation may be rapid relative to other modes of speciationbecause some reproductive isolation can evolve as a by-product of an evolutionary niche shift (Rice1987), which can itself be quite rapid. For example, if the niche shift alters the probability of56encountering a mate, then premating isolation evolves as a by-product of selection for resource useefficiency.I believe that ecological speciation is relevant to many radiations in nature. I have focussed onresource exploitation, but niche-based selection pressures also include predation and parasitism, andstudies might as easily document divergent selection for alternative adaptations to predators or parasites.Since interfertile species are not rare, evolutionary ecologists have the opportunity to directly assess theplausibility of ecological speciation. The explicit predictions and experimental tractability of the process,and its relevance to other fields of study should make research of this mode of speciation especiallyfruitful.57TABLE 2.1. Measurements (i ± s.d.) of initial mass, and seven morphological traits from fish used in thefield experiments.hybridsmass (g)standard length abody depth amouth width agill raker number b,cgill raker length a,dpelvic spine length aplate number1.069 ± 0.2073.930 ± 0.0242.342 ± 0.0251.434 ± 0.03723.879 ± 0.9930.412 ± 0.0761.838 ± 0.05812.848 ± 1.1491.234 ± 0.2 143.904 ± 0.0292.362 ± 0.0381.469 ± 0.04721.255 ± 1.3770.251 ±0.1331.596 ± 0.1298.891 ± 1.8331.386 ± 0.1943.872 ± 0.0292.379 ± 0.0291.516 ± 0.03717.968 ± 1.5380.048 ± 0.093-0.050 ± 0.1622.194 ± 2.286a size-adjusted logarithm in mm. Size correction was done using a method modified from Schiuter andMcPhail (1992). A general body size variable was calculated as PCi from a principal component analysisof ln-transformed body length, body depth, and mouth width. Size-corrected characters are residuals fromlinear regression of individual characters on PCi, standardized to mean PCi.b total number on first gill archno relationship with sizelength of longest raker on first gill archplate number includes any staining placode (regardless of size) from both sides of an individuallimnetics benthics58TABLE 2.2. Results from a two-way ANOVA testing for the effects of species and habitat type on lntransformed growth rates in Paxton Lake. The two categories of “group” are pure species (limnetic inopen water, benthic in littoral zone) and hybrids.df F pgroup 1 8.0809 0.007habitat 1 45.9152 <0.001group*habitat 1 0.0954 0.759residuals 3659TABLE 2.3. Stomach contents of 22 benthics and 18 hybrids from littoral zone enclosures; and 12 hybridsand 12 limnetics from open water enclosures.littoral zonebenthics hybridsn proportion n proportioninsect larvae 266 0.46 67 0.29ostracods 189 0.33 69 0.30amphipods 28 0.05 3 0.01molluscs 20 0.03 5 0.02copepods 11 0.02 5 0.02cladocera 9 0.02 7 0.03stickleback eggs 39 a 0.07 76 b 0.33other 12 0.02total 574 232mean preywidth in mm(± s.e.) 0.501 (0.020) 0.528 (0.043)60open waterlimnetics hybridsn proportion n proportioncopepods 1750 0.74 2018 0.60insect larvae 160 0.07 354 0.10cladocera 452 0.19 900 0.27amphipods 2 0.001 18 0.005ostracods 9 0.004 18 0.005other’ 2 0.001 66 0.02total 2375 3374mean preywidth in mm(± s.e.) 0.306 (0.005) 0.320 (0.005)a from stomachs of 2 individuals1,from stomachs of 3 individuals° mites and rotifersd mites, molluscs, rotifers, stickleback eggs or fry61FIGURE 2.1. Mean growth rates of families of benthics (n = 6), F1 hybrids (n = 9) and limnetics (n = 5)reared in the lab on identical diets. Total mass of each family was taken at 18 weeks afier fertilization,and divided by the number of individuals alive. Values are residuals from a locally weighted regression ofmass on density (see chapter one). Large dots with whiskers denote means of family means ± s.e.Adjacent smaller dots are the mean growth rates of each family. The dotted line shows that hybrids growfaster on average than would be expected if their growth rates were intermediate between benthics andlimnetics, but not significantly.residualmassperindividualafter18weeks(g)IIoo0000IIIwCDD•C)11-I••••••••-‘I3D•••••C•)1963FIGURE 2.2. Growth rates of fish raised in the littoral zone and in the open water. Large dots withwhiskers denote means ± s.e. Smaller dots are the enclosure means (two individuals per enclosure).littoralopenwater.40> E.ci)..2O.C ci) E.. .10•SS .0•FlHybridBenthicFlHybridLimnetic65FIGURE 2.3. Fitness set calculated from growth trials in 1991, 1992, and 1993. Analysis of variance wasused to correct for effects of variation among years. Dots with whiskers are means ± s.e. for limnetics (L),F1 hybrids (H), and benthics (B). Hybrids fall below the dotted line connecting benthics and limneticswhich implies selection against them even if all three types were to choose habitats randomly.18L> E16 14oBC) 0 c,)________C C) E1015202530meangrowthrateonbenthos(mg/day)Chapter Three MATE CHOICE, HABITAT PREFERENCES AND SEXUAL SELECTIoN AGAINsT HYBRIDS67INTRODUCTIONThe evolution of premating reproductive isolation is usually integral to the formation of speciesin sexually reproducing organisms. Premating isolation is brought about by factors which prevent matingsbetween populations or incipient species—for example, behaviors which cause individuals of twopopulations to be in different places when it comes time to mate (e.g., habitat preferences); or, if they arein the same place, to prefer only their own kind (e.g., mating preferences). If individuals from the twopopulations do mismate, then postmating isolation lowers the probability that hybrid offspring willsurvive and/or reproduce. Despite its importance in speciation we know virtually nothing about howpremating isolation evolves. There have been many excellent descriptions of premating isolation, but fewattempts to elucidate how it has evolved (Endler 1989).Divergent mating preferences and premating isolation are thought to arise in a number of ways.The main explanations focus on sexual selection (e.g., Lande 1981 a, 1982; Schiuter and Price 1993)reinforcement (e.g., Butlin 1989; Sanderson 1989) genetic drift (e.g., Kaneshiro 1980; Carson andTempleton 1984), or correlated responses to selection (e.g., Rice 1987; Rice and Salt 1990). But divergentmating preferences can also affect postmating isolation by selecting against hybrids. Sexual selectionagainst hybrids in the competition for mates was suggested by Fisher (1930), but few studies haveempirically addressed the issue. Most empirical studies of hybrids measure the cost of mismating byfocussing on hybrid viability and mechanisms of natural selection rather than aspects of hybridreproductive success. I set two tasks in this study of a pair of sympatric stickleback species. The first wasto describe premating isolation between the species. The second was to ask whether mating preferences ofthe parental species may set up selection against F1 hybrids during the competition for mates.There are important implications of postmating reproductive isolation by divergent matingpreferences. The first is that it may be an important “piece” of the total reproductive isolation between6869species which lack complete isolation by other mechanisms. Incomplete isolation may occur when hybridviability and fertility is high or when mating preferences have not diverged sufficiently to preventmismatings. The second implication is that sexual selection against hybrids may be an importantcomponent of reinforcement. Coyne and Off (1989) suggested that mating discrimination against hybridsmay 1) select for greater premating isolation, 2) select for greater postmating isolation through furthermating discrimination against hybrids, and/or 3) trigger divergent sexual selection. Although there areformidable theoretical obstacles to reinforcement (e.g., Spencer et al. 1986) there is some empirical andtheoretical support for the process (Coyne and Orr, 1989; Liou and Price, in press). It is therefore worthconsidering whether some of the preconditions for reinforcement exist in nature.Stickleback species pairsThreespine sticklebacks (Gasterosteus aculeatus complex) are small fish widely distributed inHolarctic marine and freshwaters (Bell and Foster 1994). In British Columbia anadromous sticldebackscolonized many low elevation coastal lakes and streams at the end of the Pleistocene (McPhail 1994). Thegreat majority of these lakes contain a single species of stickleback, but several lakes on islands in theStrait of Georgia contain sympatric species pairs (McPhail 1984, 1992, 1993, 1994; Schluter and McPhail1992). Each pair is thought to have been independently derived (McPhail 1993). The species pairs eachexhibit a similar pattern of morphological and ecological divergence (Schiuter and McPhail 1992). Thespecies have not been formally described, so we refer to them by their preferred foraging habitats.“Limnetics” primarily exploit plankton, and “benthics” exploit mainly benthic prey in the littoral zone.All research for this chapter has been on the pair from Paxton Lake, on Texada Island.Male sticklebacks are the sole providers of parental care. In the spring, they acquire territories inthe littoral region where they build nests, mate (sometimes with several to many females), and raise theiroffspring. Thus, benthics nest in their preferred foraging habitat, whereas limnetics do not. In Paxton70Lake, benthic males tend to build nests in areas of the freshwater algae Chara sp., while limnetic malestend to nest in the open (McPhail 1994). The two habitats form a mosaic, and it is not rare forheterospecific males to be nearest neighbors (McPhail 1994). Gravid females have a high probability ofobserving males of both species when searching for mates, so the species may be regarded asreproductively sympatric. Over a twenty year span, three separate estimates found that F1 hybrids makeup somewhere in the range of 1-2% of all adult sticklebacks in Paxton Lake (McPhail 1992). Prematingisolation in the wild is therefore incomplete.Hybrids (F1s, F2s and backcrosses) between limnetics and benthics of Paxton Lake can beobtained by artificial fertilization (McPhail 1992; chapter one). F1 hybrids are morphologicallyintermediate between the parental species in all traits measured so far (McPhail 1992; table 2.1). Hybridviability has been extensively studied in the lab, and there is no convincing evidence of F1 or F2 inferiority(McPhail 1992; chapter one), though there is some evidence of breakdown in backcrosses (chapter one).The only known postmating disadvantage for F1 hybrids is a foraging disadvantage in the wild, but thislikely produces less than complete postrnating isolation (McPhail 1992; chapter one and two). For thesetwo species to be maintained there must be substantial premating isolation and/or further postmatingisolation.This chapter reports measurements of mating success of male F1 hybrids relative to limnetic andbenthic males. Two main considerations led me to examine success of male rather than female hybrids.First, only males develop conspicuous secondary sexual characters, so I assumed that females are thelimiting (and therefore choosing) sex and that female preferences are stronger than male preferences.Second, I assumed that encounters between male and female hybrids in the wild would be rare relative toencounters between hybrids and the pure species, and therefore that F1 hybrids would have moreopportunity to backcross than to form F2 hybrids.71MAmRIALs AND METhoDsExperimental crosses and rearing conditionsAll fish used in these experiments were reared in the lab under identical conditions (see chapterone). I made full sib crosses by extracting eggs from gravid wild females, and combining the clutch in apetri dish with macerated testes from a wild male. I then immediately selected for rearing a random subsetof 30 eggs from each clutch. I raised each family separately in halved 100 L aquaria held at constanttemperature and light (18° C; 1 6L:8D). I brought juvenile fish into reproductive condition by taking themthrough an approximate three month period of low temperature and short day length (4°-10° C; 8L:16D)followed by a gradual return to the original temperature and light regime. The amplitude in daylightcycles approximated those of the wild. When conditions reached 16L:8D and 18° C they were maintainedto keep fish in breeding condition until the end of the experiments.I made crosses in April-May 1991 and 1992, and raised pure benthics (20 families), purelimnetics (17 families), and reciprocal F1 hybrids (15 L x BdN families, 19 B x Ld’ families). Fish inrearing tanks were not supplied with nesting material, and could not view the courtship displays of fish inexperimental tanks. All the fish I used in experiments were therefore naive.Preference testsI conducted experiments in June-July 1992 and 1993 when fish were about one year old. I usedindividuals as they came into reproductive condition with the stipulation that no within family matechoice trials be conducted.I conducted the experiments using a “no-choice” design for logistical reasons. Extensive pilotexperiments demonstrated that a choice protocol (e.g., Hill 1990; Milinski and Bakker 1990; Zuk et al.1990; Dugatldn and Godin 1992; McKinnon, in press) was not feasible for two reasons. First, when two72nesting males were not separated by a complete barrier, one male invariably took over the entire tank,regardless of tank size. This prevented females from interacting directly with two nesting males. Second,females were very reluctant to interact through glass when presented with a choice of males in separateaquaria. Females exhibited little or no interspecific preference with such a design. I therefore used no-choice tests. No-choice tests are commonly employed in mate choice studies, they are efficient, and cangive results which are very reliable (e.g., Milinski and Bakker 1990, 1992; Bakker and Milinski 1991).Single males were introduced at random to one side of a 100 L aquarium divided in two by anopaque barrier. Each side had a single corner filter, and was supplied with a tray of sand and plantmaterial (“Java moss”, Vesicularia dubyana) for nesting. The bottoms of the aquaria were covered withcoarse limestone gravel and the sides were covered by a neutral grey paper of a similar shade.(Background color is variable for nesting sites in the wild, but limestone gravel is a common substrate.)The aquaria were illuminated by “cool white” fluorescent bulbs, with the addition of a 60 W incandescentbulb during mating trials. All nesting males were visually isolated from one another.I “primed” males by introducing a gravid female of the same type for 10 minutes each day untilthe male was used in a trial. Males would generally fail to build or maintain adequate nests if they werenot exposed to gravid females periodically. I used males in a mating trial only if they had built nests witha well defined opening within seven days, and had responded with a vigorous courtship display to afemale during priming. Before using females in trials they were isolated for 12 to 24 hours; the femalesused in the trials had not been used for priming. All males developed nuptial coloration, and males ofeach cross type had similar color expression.Females exhibit stereotypical behaviors in response to male courtship. The sequence of femalebehaviors is (1) to orient her body in a head up posture, (2) to follow the male, (3) to examine the nest bysticking her snout in the nest opening, and finally (4) to enter the nest to spawn. Both in the lab and in thewild females may break off a courtship at any of these points (personal observation).73I placed a gravid female in a tank with a nesting male and videotaped the courtship for two hours.From the videotapes I later extracted elapsed time to each of the four female responses. My protocollikely measured female preference rather than male choice because rejection was almost always initiatedby females while males continued to court. Subtle male choice may nevertheless have occurred in thetests.All males courted the females, but only females that responded with at least a head up wereincluded in analyses. I interpreted a lack of response subsequent to a head up by a female as a measure ofpreference, and therefore included females which did not continue with a follow, examine or enterbehavior in all calculations of mean times and variances. This is still a conservative criterion since evenfemales presented with a hybrid or heterospecific male had to give at least some response to be included.In only four of 100 trials did females fall to give a head up response. Final sample sizes are shown intable 3.1.I analyzed mate choice data using analyses of variance on the ln-transformed times. I usedordered expectation one-way ANOVAs (Gaines and Rice 1990; Rice and Gaines 1994) to test whetherconspecifics were preferred over hybrids or heterospecifics. Ordered expectation ANOVAs are used whenthere are prior expectations of patterns in the data. For example, hybrids are intermediate in morphologyso I expected that female responses to male hybrids would be intermediate also. The test is analogous tousing one-tailed probabilities for a t-test—it substantially improves power to detect an effect. I usedstandard two-way ANOVAs to test for interactions between female response and male type and toexamine whether there was a difference in male mating success when averaged across females. Two-wayANOVAs could be performed only on the first three responses since benthic females did not spawn withany of the males.74Estimates ofhybrid reproductive successTo ascertain whether divergent mating preferences result in selection against hybrids Iconstructed fitness sets from my data. A fitness set plots a measure of fitness (e.g., mating success) on oneresource along the y-axis versus fitness on an alternative resource along the x-axis (Levins 1968). Ienvisioned limnetic and benthic females as the resources which male sticklebacks must exploit. Such afitness set shows whether hybrid males would be selected against during competition for mates.Standard fitness sets implicitly assume that females of each species encounter males of the threetypes in proportion to their relative abundance. But encounter rates may be profoundly affected bybehaviors such as habitat preference. If the two species have divergent habitat preferences, thenencounters between males and females will be in proportion to their relative abundance in those habitats,rather than their relative abundance averaged over all habitats. Note that habitat preference should beexpressed in both males and females for it to affect encounter rates. Figure 3.1 displays the effect ofhabitat choice on a fitness set. The shaded region of the figure is an area of isoclines, four examples ofwhich are shown, ranging from complete segregation of the parental species by habitat (1), to no habitatchoice (4). Lines 2 and 3 show isoclines for intermediate degrees of habitat segregation.At one extreme, if all males encounter pure type females at random then hybrid reproductivesuccess must fail below isocline 4 to demonstrate a mating disadvantage. At the other extreme, iflimnetics and benthics are completely segregated by habitat then hybrid mating success should fall belowisocline 1 in order to conclude sexual selection against hybrids. The distance between the mean for malehybrids and the relevant isocline describes the mean strength of sexual selection against hybrids.Habitat preferences in malesMy lab experiments minimized differences among males to force females to choosemorphological and behavioral characteristics of males rather than extrinsic characters such as habitat. Yet,75the extent of habitat preference determines the relevant isocline for comparing hybrid mating success, so Iexamined habitat choice by placing lab-reared males in the field. As noted above, I should know thehabitat preferences of both males and females. Determining habitat preferences only of males does notsolve the issue of which is the relevant isocline, but it is a necessary step. Future work will determinefemale habitat preferences. I incorporate the data on male habitat preferences when I discuss thepossibility of sexual selection against hybrids.I placed lab-reared reproductively mature limnetic, benthic and F1 hybrid males in enclosures inthe littoral zone of Paxton Lake. Each enclosure was 1 m x 1 m and contained a single fish. I spread algae(Chara sp.) over half of the bottom of each enclosure; the other half had no cover. A single naive malewas placed in each enclosure and primed for 5-10 minutes per day by presenting him with a gravid femaleof his own type in a 1 L jar. After four to seven days the location of the male’s nest was noted using threeclasses of cover: no cover (nest in the open, 5 cm or more from cover), partial cover (nest within 5 cm ofChara), or complete cover (nest under a canopy of Chara). The partial cover category is somewhatarbitrary, but my intention was to ascertain whether hybrids are intermediate in nesting sites. Thesecategories thus reveal whether hybrids tended to nest at the transition from covered to open habitat.RESULTSBoth limnetic and benthic females readily distinguish among conspecific, hybrid, andheterospecitIc males (table 3.2). Mean response times indicate that females responded much faster to theirown type than they did to hybrid or heterospecific males, though the stage of courtship at whichpreference was strongest was different for the two species. Limnetic females showed strong preferenceslate in the courtship sequence whereas preferences of benthic females were strongest early in courtship.Weak preferences were generally observed at the other stages.76Of females which responded to males with a head up, all but one (a benthic) also responded witha follow (table 3.2). But proportions of females responding fell off beyond this point in courtship. Ingeneral, proportions corroborate mean response time data: females more readily court their own speciesthan hybrids or heterospecifics. In some cases though, hybrids seem as successful as conspecifics.Two of the three two-way ANOVAs found statistically significant interactions between males andfemales (table 3.3) supporting the view that limnetic and benthic females rank males differently. (Thethird was close to significant at a = 0.05). Limnetic and benthic females also differed significantly in theiroverall response to males because mean responses by limnetic females were faster than those of benthicfemales (see table 3.2). The lack of significance of the male terms indicate that males were statisticallyequivalent when averaged across females, suggesting no disadvantage for hybrids.In constructing fitness sets, ideally I would use the same stage of courtship response by femalesso that axes are equivalent. Two observations prevented this approach. First, I knew that the two femalegroups responded over different time frames (tables 3.2 and 3.3). Second, limnetic and benthic femalesexhibited their strongest preference at different ends of the courtship sequence. Therefore, a fitness setusing the same stage of courtship would have different units on each axis, and show an essentially flatresponse on one axis. I therefore used the two behaviors for limnetic and benthic females whichdemonstrated the clearest interspecific preferences. For limnetic females these were nest examination andnest entry; for benthic females these were head up and follow. I combined the responses to give fourfitness sets (fig. 3.2).The fitness sets show that limnetics and benthics are well-isolated by premating mechanisms.Means of pure species males are separated by more than two standard errors on each axis. The graphs alsoshow that limnetic and benthic mating preferences do not necessarily result in sexual selection againsthybrids. Hybrid attractiveness is intermediate for both limnetic and benthic females, but responsesnevertheless add up to be equivalent for the three groups of males. Thus, the mean for hybrid males falls77roughly on the line drawn between limnetic and benthic males. The four graphs are in general agreementthat my data do not support a hybrid mating disadvantage assuming no habitat preferences.In enclosures, lab-reared limnetic and benthic males chose to build nests in different habitats(table 3.4). Benthic males built only under cover, whereas limnetic females built primarily in the open.Hybrid males were not intermediate in their choice of nesting sites—they were virtually identical tolimnetic males. Differences among males in nest locations is highly significant by a Pearson chi-squaretest (x2 = 31,3, df= 4 , p <0.001).DISCUSSIONPaxton Lake limnetic and benthic females showed measurable premating isolation in the lab.They responded more often, and faster to their own type than to hybrid and heterospecific males (table3.2). For behavioral stages at which preference was most pronounced, the differences between meanresponse to conspecifics versus heterospecifics were more than two standard errors. My results parallel astudy by Ridgway and McPhail (1984) who studied premating isolation between a different species pairof sticklebacks—the Enos Lake limnetics and benthics. They found that the Enos Lake pair were wellisolated by behavioral mechanisms. They also encountered difficulties working with benthics. Reasonsfor benthics’ reluctance to spawn in the lab are not readily apparent. Benthics’ recalcitrance is in starkcontrast to the compliant behavior of limnetics and sticklebacks from other populations (e.g., Milinskiand Bakker 1990) which tolerate lab conditions and mating manipulations much more readily.Differences among populations’ or species’ amenability to lab study make assessing mating behaviordifferences more difficult. Possible repercussions of this difficulty are discussed in the next section.Setting aside differences in the species’ amenability to laboratory manipulations, my resultsdescribe the potential success of hybrid males in the competition for mates. Limnetic and benthic females78ranked males in the following order, from most to least preferred: conspecific, hybrid, heterospecific.Two-way ANOVAs and fitness sets suggested that despite their intermediate attractiveness, hybrid malesmay have reproductive success in the wild similar to that of pure type males. However, such aninterpretation depends critically upon the pattern of encounters between males and females of each crosstype. The potential for nonrandom encounters was demonstrated by divergent nest site preferences of purespecies males (table 3.4). These patterns were similar to those observed by McPhail (1994) for nestingmales observed in the wild. The finding that male hybrids chose sites similar to male limnetics wassurprising though given the morphological and behavioral intermediacy found in previous studies ofhybrids (e.g., McPhail 1992; Schluter 1993; table 2.1). It means that in the wild hybrid males willgenerally only be interacting with limnetic males.Habitat choice by females remains to be studied. But, if pure type females have habitatpreferences similar to pure type males this would inflict a substantial disadvantage on hybrids because allmales could interact with females of only one species. Divergent female habitat preferences wouldgenerate sexual selection against hybrid males, because hybrids could only court female limnetics.Size differences among my experimental fish were intentionally very small, though in the wildbenthics are usually larger than limnetics (McPhail 1984, 1992). In sticklebacks size differences oftenplay a role in mate acquisition (Rowland 1989; Whoriskey and FitzGerald 1994). Nagel (1994) found thatbody size may also play a role in premating isolation between Paxton Lake limnetics and benthics.Limnetic females responded more to small than to large benthics; and benthic females responded more tovery large limnetic males than to small limnetic males. I chose to minimize size differences in order tofocus on morphological and behavioral differences. During data analysis I found no relationships betweenthe difference in length between males and females, and the responses given by females to the male in atrial. However, I acknowledge that body size may play a role in mate selection when variation in bodysize is greater.79Strength ofpremating isolation and sexual selection against hybridsExperimental tanks contained a single habitat with little structure, which is likely closer to thetype of habitat preferred by limnetics (see table 3.4). This may be one explanation for benthics’ weakpreference at the nest examination stage of courtship and their unwillingness to spawn. Matingpreferences are best measured directly with unequivocal responses such as time to spawning, rather thanwith a surrogate measure, unless the surrogate measure is highly correlated with probability of mating. Adirect measure was pOssible for limnetic females, but not benthics. Although benthic females courted withnesting males and showed significant preferences early in courtship (table 3.2), I do not know whetherthis preference is correlated with spawning success.Benthics’ refusal to spawn within the allotted time and their generally slower responses (tables3.2 and 3.3; fig. 3. 2) suggest that caution should thus be used when extrapolating from these results tosituations in the wild. I therefore present two alternative scenarios of habitat preference and assess theramifications of weak benthic responses for the conclusion of premating isolation and sexual selectionagainst hybrids.Scenario one: females have no habitat preference. If the preferences of benthics early in courtshipand limnetics late in courtship are indicative of female mating preferences in the wild then there issubstantial premating isolation between limnetics and benthics. Absence of habitat preferences in femaleswould not lessen this premating isolation.The two-way ANOVAs (table 3.3), and the fitness sets (fig. 3.2), suggested no sexual selectionagainst hybrids under scenario one. If females exhibit no habitat preferences they will encounter males inproportion to their overall abundance in the lake. The strong habitat preferences of males (table 3.4;McPhail 1994) cannot produce any hybrid disadvantage in the absence of female habitat preferences.However, since mean hybrid mating success is at the margin (fig. 3.2), any habitat preference by femalesshould generate sexual selection against hybrids.80The greatest potential problem with this conclusion is that I do not know if benthic females’preferences early in courtship correlate with spawning probability. A head-up response in females isstrongly correlated with probability of spawning in some populations (e.g., McLennan and McPhail 1990)and has been used as a measure of female preference in many other studies (e.g., Milinski and Bakker,1990, 1992; Bakker and Milinski, 1991). I believe the behavior also correlates with mating preferences inPaxton Lake benthics, but the strength of the conclusion of equal mating success for hybrid males hingeson a strong response (i.e., reliable data) from both limnetic and benthic females to males in experimentalconditions. Weak responses by one or both types of females lessen the strength of such a conclusion.Scenario two: females prefer the same habitats as males of their species. This scenario predictsgreater premating isolation than that observed in our experiments. It also predicts sexual selection againstmale hybrids because they must compete for limnetic females with the preferred limnetic males. Thestrength of selection against hybrids would increase with increasing degrees of habitat preference.The qualitative prediction from this scenario should not be affected by weak benthic responsesbecause limnetic females behaved clearly and consistently, and spawned more often with limnetic malesthan with hybrids. Given that male habitat choice is strong (table 3.4), and limnetic females have apreference for limnetic males over hybrids (table 3.2; fig. 3.2), the strength of sexual selection againsthybrids in the wild will depend on two aspects of the mating system: (1) the strength of habitat preferenceby females (both limnetic and benthic), and (2) the strength of mating preference by limnetic females forlimnetic males versus hybrid males. If both habitat and mating preferences are strong, then there ispotential for strong sexual selection against hybrids. Mean response times (table 3.2) showed that limneticfemales have substantial preference for conspecifics over hybrids: mean time to spawning withconspecifics was two-thirds of the mean time with hybrids; the difference in mean times to spawn was9.35 minutes. But in a no-choice situation many limnetic females eventually spawned with hybrid males:13 out of 20 (65%) limnetic females spawned with limnetic males, and 10 out of 17 (59%) spawned with81hybrid males (table 3.2). This statistic indicates only weak preference, but choosiness is expected todecline as females approach the point at which they can no longer retain their eggs (Milinski and Bakker1992). Thus, preferences may be stronger when females are given a choice of males.Habitat preferences and speciationDivergent mate choice and habitat preferences can independently lead to reproductive isolation. Ihave tried to show how interaction between the two may produce even greater reproductive isolation.However, ft is not clear how or why habitat preferences would evolve when mating takes place in habitatsdifferent from those preferred for foraging. (Limnetics forage in the open water, but nest inshore.) Modelsexist to explain the evolution of premating isolation as a “by-product” or correlated response to selectionfor foraging in different habitats (e.g., Rice 1987; Diehl and Bush 1989). However, in these modelsmating takes place in the preferred foraging habitats.Two hypotheses might explain divergent habitat preferences in sticklebacks. The first is that therehas been or continues to be competition for nest sites. This competition could cause divergence in nestsite locations (e.g., larger benthics could exclude limnetics from vegetated areas). In Paxton Lake it seemsunlikely that nesting habitat is limiting (personal observation), though some sites may be better thanothers. Also, nest sites chosen by males in enclosures (table 3.4) and in the wild (McPhail 1994) suggestthat divergent habitat preferences exist both in the presence and absence of potential competition for nestsites. Previous competition may nevertheless have caused this divergence.An alternative to competition driven divergence in habitat preference is that nesting sitepreferences may have been indirectly shaped by selection on the visual system for increased foragingefficiency (Endler 1989). The hypothesis is supported by empirical studies which have shown markeddifferences in visual systems between species which forage on plankton versus species which forage onbenthos (e.g., Collin and Au 1994). Adaptation of the visual system to foraging needs may result in82sensory biases and lead to divergent mating preferences and nesting site preferences (Endler 1992). Thesealternative hypotheses await testing.In this study I described premating isolation between Paxton Lake limnetics and benthics, andassessed whether their divergent mating preferences lead to a mating disadvantage for F1 hybrids.Premating isolation between the species exists in the absence of habitat preference, but its presence wouldenhance premating isolation beyond that shown here. On the other hand, sexual selection against hybridsappears to rest critically on the extent of female habitat preference. Habitat preferences may have played asubstantial role in the speciation of these fish by providing both premating and postmating isolation. But,whether females have habitat preferences needs to be addressed before we know if divergent matingpreferences set up reinforcing selection in these species.83TABLE 3.1. Combinations and sample sizes of mate choice trials (n = 96).malehybridlimneticfemale_____________________________________benthiclimnetic benthic20 17 1514 16 1484TABLE 3.2. Female responses to courting males. Values are in-transformed time in seconds for females torespond to courting males. p-values are from ordered expectations one-way ANOVAs on mean times (seetext).limnetic femaleswith: limnetic males hybrid males benthic malesi s.e. proportion s.e. proportion i s.e. proportionhead up 2.24 0.26 20/20 2.44 0.31 17/17 2.52 0.52 15/15 p = 0.424follow 3.25 0.28 20/20 3.37 0.27 17/17 3.22 0.48 15/15 p=0.6l8examine 5.40 0.23 20/20 6.40 0.44 14/17 7.12 0.50 9/15 p=0.003enter 6.98 0.42 13/20 7.40 0.42 10/17 8.33 0.32 3/15 p = 0.021benthic femaleswith: iimnetic males hybrid males benthic males5 s.e. proportion i s.e. proportion 5 s.e. proportionheadup 4.56 0.53 14/14 3.58 0.48 16/16 2.58 0.50 14/14 p=0.Ollfollow 4.97 0.59 13/14 4.17 0.47 16/16 2.96 0.49 14/14 p = 0.011examine 7.65 0.43 6/14 7.42 0.36 10/16 7.22 0.52 8/14 p = 0.39085TABLE 3.3. Two-way analyses of variance for female courtship responses. Each behavior represents astage in the courtship sequence. Data are in-transformed times in seconds to exhibit the indicatedbehaviors by females. The ANOVAs demonstrate three main results: 1) females differ significantly inresponse times, 2) males are equivalent when averaged across females, and 3) significant interactionsindicate a different order of preference by females.head up follow examine nestdf F p F p F pfemale type 1 11.22 0.001 4.66 0.034 11.33 0.001male type 2 1.90 0.155 2.89 0.061 1.24 0.296female*male 2 3.42 0.037 2.63 0.077 3.40 0.038residuals 90TABLE 3.4. Locations of nests built by lab-reared males placed in enclosures in the wild. Pure specieshave divergent nest site preferences, and F1 hybrids resemble limnetics.limnetic hybrid benthicno cover 5 12 0partiaL cover 2 3 0full cover 0 1 128687FIGURE 3.1. A series of fitness sets showing how different levels of habitat selection in limnetics andbenthics can affect mating success isodines. The open triangle depicts a hypothetical mean for benthicmales, the open square depicts limnetic males, and the filled diamonds depict equally fit hybrid malesunder different levels of habitat selection. If limnetics are found only in one habitat and benthics only in adifferent habitat, then isocline 1 is appropriate for comparing mating success of hybrids. If the mean formale hybrids falls below this line then on average there is sexual selection against male hybrids. If allmales encounter limnetic and benthic females in proportion to their total abundance in all habitats (i.e., nohabitat segregation) then isocline 4 is the relevant line for comparing hybrids to pure species.Intermediate levels of habitat segregation are shown by isoclines 2 and 3.fastC)LCl) 0 Cl) ci) ci) E ci) 9- C) ci) C EslowslowfastbenthicfemaleresponseB000089FIGURE 3.2. Fitness sets showing mean responses by limnetic and benthic females to the three male types.Bars indicate standard errors and letters denote male type: L = limnetic, H = hybrid, B = benthic. Thedotted line shows the mating success isocline if males encounter females of the two parental species inproportion to their overall abundance in the lake. The four graphs were constructed using femalebehaviors (two for each species) which showed statistically significant differences in preference for thethree male types. For benthic females the behaviors were early in courtship; for limnetics the behaviorswere later in the courtship sequence. The graphs are in agreement that hybrids do not suffer a matingdisadvantage in the absence of habitat selection.01CD0DCD:-C)0CD’CD01-,01CDCl)-Do01—oC)01—OCDCD01responsebylimneticfemaleIn(seconds)toenternestp—101001001—-‘CD0CDCDOc-0oc,•C2.z-0CD’-CD001-‘CDO10-•TCD 0Do01—C)01-OCDCD01responsebylimneticfemaleln(seconds)toexaminenest-0)0)0101001001I0I-.Iresponsebylimneticfemaleln(seconds)toenternest0)0)-0101010responsebylimneticfemaleln(seconds)toexaminenest-J-.10)0)C)101001001I06Chapter Four GENETIC DIVERGENCE IN ADAPTIVE CHARACTERs91INTRODUCTIONOur understanding of mechanisms of adaptation and speciation depends, in part, on anunderstanding of genetic divergence between species. Genetic divergence between populations andspecies may be the result of divergent natural selection, genetic drift, mutation and/or gene flow. Theoriesof speciation emphasize these processes to different extents.Several questions relevant to the study of speciation may be answered from empirical studies ofgenetic divergence among species. The most common of these involves the evolution of postmatingisolation: how do genes from separate populations work when combined in hybrid individuals? Tounderstand how postmating isolation mechanisms evolve we need to understand the genetic architectureof traits which directly or indirectly provide reproductive isolation. For example, we should know howmany loci are involved, whether some loci are dominant, whether there are interactions between loci, andwhether the distribution of effects of different loci is uniform. The answers will help us understand howselection, drift and recombination affect traits conferring reproductive isolation.Most models of the evolution of quantitative characters assume a particular genetic architecture:that traits are polygenic and variation is entirely additive. Currently, our knowledge of whether thisassumption holds is remarkably scant (Hard et al. 1992, 1993). Compounding this problem is that thetraits we know most about at a genetic level we generally know little about at an ecological level (howthey affect the fitness of organisms in the wild). Linldng genetic analysis with evolutionary ecologyprovides promise that we can understand how an organism’s fitness in the wild is mediated at the geneticlevel.This chapter has the primary objective of describing genetic differences between two interfertilefish taxa. Rearing six lines (parentals and first and second generation hybrids) in the lab allowed me toexplore how the two species have diverged in two foraging characters (gill raker number and gill raker9293length), two armor characters (plate number and pelvic spine length), and one direct fitness component(growth rate). There is good ecological information for the four morphological traits. I used regressiontechniques to examine the contribution to phenotypic divergence of additive, dominance and epistaticgenetic effects. I also used a biometrical technique to estimate the minimum number of loci contributingto differences in mean expression of the five traits.The study speciesThreespine sticklebacks (Gasterosteus aculeatus complex) are small fish common to coastal areasthroughout the northern hemisphere (Bell and Foster 1994). Marine sticklebacks have given rise topopulations in many coastal lakes and streams by colonization after deglaciation at the end of thePleistocene (McPhail 1994). Several lakes on islands in the Strait of Georgia, British Columbia, Canadaare exceptional because they contain sympatric species pairs of sticldebacks (McPhail 1984, 1992, 1993,1994; Schluter and McPhail 1992). Each pair is thought to have been independently derived (McPhail1993). In this chapter I deal exclusively with the species pair from Paxton Lake, on Texada Island.Each species pair exhibits similar divergence in morphological and ecological characters(McPhail 1984, 1992, 1993, 1994; Schluter and McPhail 1992). The species have not been formallydescribed, so they are referred to by their preferred foraging habitats. “Limnetics” primarily exploitplankton, and “benthics” exploit mainly benthic prey in the littoral zone. Limnetics have a fusiform body,narrow gape and many, long gill rakers; benthics have a robust body form, wide gape and few, short gillrakers (McPhail 1984, 1992, 1994; Schluter and McPhall 1992). Gill rakers are the feeding apparatus offish—they are thought to directly filter food particles, or otherwise aid in particle capture (Sanderson etal. 199 1)—and gill raker number and length is heritable in sticklebacks (Hagen 1973; D. Schiuterunpublished data). Long numerous gill rakers are prevalent in planktivorous fish in general (Schiuter andMcPhafl 1993). Limnetics and benthics have also diverged substantially in armor characteristics (McPhail941984, 1992, 1994; table 2.1): benthics exhibit substantial annor reduction sometimes to the point of acomplete lack of lateral plates and pelvic and dorsal spines. Divergent armor characteristics may reflectdivergent predation pressures. Armor characteristics are also heritable in Gasterosteus (Hagen 1973).MATERIALS AND METHoDsExperimental crossesThe crosses I made were pure limnetic (16 families), pure benthic (15 families), F1 (21 families)and F2 (31 families) hybrid, and limnetic and benthic backcrosses (8 families of each). Each cross type isreferred to as a “Line”, and will be referred to respectively as P1. P2,F1,F2, B1 and B2. The crossing designand rationale are discussed in more detail in chapter one. Hybrid lines may be made from more than onereciprocal cross. I carried out all reciprocal crosses for F1 and F2 hybrids, but concentrated on certainbackcross reciprocals due to rearing limitations (see chapter one). All crosses were obtained by artificialfertilization (McPhail 1992; chapter one), and viability of hybrids and pure types in the lab is high(McPhail 1992; chapter one). All fish were reared under identical conditions (see chapter one for fulldescription). No sibs were ever crossed, and each replicate line was established from parents from adifferent pair of families. Hence little or no close inbreeding occurred. The first generation was made inApril-May 1992 using as parents wild caught fish from Paxton Lake. I used only lab-reared individuals tomake the second generation in April-May 1993. Second generation fish were not reproductively mature atthe end of the experiment, so could not be sexed.Fish were anaesthetized in MS222, fixed in 4% formaldehyde for at least one week, stained withalizarin red, and preserved in 37% isopropyl alcohol. I measured morphology using a dissectingmicroscope fitted with an ocular micrometer. Gill raker number is the total on the first gill arch; gill rakerlength is length in mm of the longest raker on the first gill arch; plate number includes any staining95placode regardless of size from left and right sides; pelvic spine length is total length in mm from tip tohinge. Gill rakers and pelvic spines were measured on the left side.Gill raker length and pelvic spine length were corrected for body size differences amongindividuals using a method modified from Schiuter and McPhail (1992). A general body size variable wascalculated from the full sample of measured individuals as PCi from a principal component analysis ofln-transformed body length, body depth, and mouth width. Size-corrected characters are residuals fromlinear regression of ln-transformed spine or gill raker length on PCi, added to the overall mean body size(i.e., mean PCi of all individuals measured). The measurements are therefore on a logarithmic scale.Growth is a fitness component because fish have indeterminate growth, and size has importantfitness consequences. Larger individuals usually have higher over winter survival (Shuter and Post 1990;Conover 1992), greater nesting success (Dufresne et al. 1990), greater mating success (Wiegmann et al.1992), higher fecundity (Baker 1994; chapter one), and earlier breeding (Miller and Storck 1984; Reznickand Braun 1987; Schultz et al. 1991; Wiegmann et al. 1992; personal observation).Eighteen weeks after fertilization I weighed each family, and calculated a per individual mass.Since all crosses hatch at approximately the same very small size (eggs are < 2 mm diameter and < 350ig wet mass [Wooton 1973]), mass at 18 weeks ( = 0.364 g, s.d. = 0.080, water displacement mass) is anestimate of total growth to that date. The number of families per line for growth rate measurements differfrom those for the morphological analysis and are as follows, limnetic (n = 5), benthic (n = 6), F1 hybrid(n = 9), F2 hybrid (n = 32), limnetic backcross (n = 7), benthic backcross (n = 7). Growth rate wasmeasured on second generation families only.Mortality during embryo development, and shortly after hatch, meant density varied among tanks(1 = 22.52, s.d. = 4.56) though approximately the same range of densities occurred in each cross type.Growth rate had a strong nonlinear relationship with density. I corrected for density by using residuals96from a locally-weighted regression (lowess algorithm, S-plus [StatSci 1991]) of growth rate on density. Iused these residuals for all analyses of growth rate.Analysis of line meansA karyotype study by Chen and Reisman (1970) found no evidence of sex chromosomes inGasterosteus aculeatus. Thus, traits are not expected to be sex-linked in this species complex. Maternaleffects and sex differences may nevertheless occur (though sex differences would not be the result of sex-linkage). I tested for sex differences in expression of morphological characters infirst generation fish(benthics, limnetics and F1 hybrids) using two-tailed t-tests on character means. I tested for maternaleffects in expression of morphological characters by comparing character means of F1 males from the tworeciprocal crosses using two-tailed t-tests. I did not assess sex differences in growth rate because themeasure was based on entire families not individuals; sample size for F1 hybrid growth rates wasinadequate to assess maternal effects. I maintained statistical independence of data in all analyses byusing family means rather than pooling all measured individuals.I tested genetical models using a regression technique called joint-scaling, developed by Cavalli(1952) and Hayman (1958, 1960a, 1960b). I used the joint-scaling method to derive estimates ofcomposite additive, dominance, and epistasis effects for differences between limnetics and benthics in thefive Iralts. Additive effects cause a cross between different phenotypes to produce offspring withphenotypes half way between the parental phenotypes. Dominance effects cause offspring to be more likeone parental type. Epistasis causes offspring phenotypes to differ consistently (i.e., not simply throughrandom measurement error) from expectations of additivity or dominance. The three effects may occursimultaneously.The joint-scaling tests for phenotype means and variances are summarized in Mather and Jinks(1982) and Lynch and Walsh (in preparation). Although originally developed for genetic analysis of97divergent inbred lines, the method can be used for the analysis of interfertile wild populations (Matherand links 1982). The lines need not be homozygous (i.e., inbred) provided that close relatives are notmated.Briefly, the technique involves fitting the following multiple regression model.z=t+M2a+3ö e (1)where z4 is the trait of interest in the ith line, i is the overall mean of all lines, a is the additive geneticeffect, ö is the effect of dominance, and is the measurement error associated with the ith line. M is amatrix of coefficients which is similar to a design matrix in multiple regression. The coefficients in Mcome from equations for the predicted line means (see Mather and links 1982); column one is all is,coefficients for additive effects are in column two, coefficients for dominance effects are in column three.The regression model is fit using the following formulae.a = (MTV-1M)-1MW (2)(3)where âis the vector of mean, additive, and dominance parameters, t x and M is the matrix ofcoefficients; V is a diagonal weighting matrix of sampling variances of observed line means (squaredstandard errors); is the vector of observed line means; and 2 is the vector of predicted line means. (Datafrom the lines are therefore in and V.) Superscripts T and -1 indicate transpose and inverse respectively.Standard errors of parameter estimates are square roots of the diagonal elements of (MTV-1M)-1 (Lynchand Walsh in preparation).98I tested the fit of regressions by comparing observed and predicted line means, using thefollowing goodness-of-fit test statistic (Mather and Jinks 1982; Lynch and Walsh in preparation),(.)2Var( (4)where the degrees of freedom are the number of lines minus the number of estimated parameters.I tested genetical models by sequential model fitting (Lynch and Walsh in preparation) beginningwith the additive-dominance model. Rejection of this model by the above goodness-of-fit test indicatesthat epistasis and/or linkage are contributing to genetic divergence of the lines. If the additive-dominancemodel is not rejected, then testing of the simple additive model is appropriate. The above regressions arerepeated after dropping the column of dominance coefficients from M (equations 2 and 3). Rejection ofthe additive model indicates that dominance effects are contributing to the genetic divergence of the lines.Failure to reject the additive model indicates that genetic divergence of the lines has been primarily at lociwith additive effects.Analysis of line variancesVariances for these analyses are based on pooled measures from all individuals in each line—formost families I measured more than one individual. I ignored family effects on variance becausesegregation variance is a within family phenomenon. That is, the increase in phenotypic variance insecond generation hybrids occurs entirely within each family. Any attempt to remove family effects byaveraging variances across families effectively removes the variation one is trying to estimate. Astatistical ideal would measure a single individual from many families, but this is generally not feasible.Phenotypic variances generally have little or no measurable heritability (e.g., Leamy and Atcbley 1985;99Coyne 1987; Tuinstra et al. 1990; MØller 1994) so measuring several individuals from the same familyshould not cause spurious conclusions.Because I measured growth rates for families and not individuals, some analyses could not beperformed. I did not use joint-scaling tests on variances because variance in growth rate could not bemeasured adequately. Since variances could not be estimated adequately, I also did not estimate genenumber for growth rate differences among lines (see section on gene number estimation below).Joint-scaling methods can be used to interpret variances from the line crosses, though scalingtests based on variances are much less powerful than those based on means (Lynch and Walsh inpreparation). For this reason I did not use the method to test the fit of variances to genetical models. Themethod is useful however for obtaining a direct estimate of segregation variance (a measure of increasedphenotypic variance in second generation hybrids versus first generation hybrids), and for a visualcomparison between observed and expected variances under the additive model.The method for estimating variance components is similar to that described above for means, anduses the equations,= (MTY-1M)-1MIY-1v (5)(6)where is the vector of variance components; M is the matrix of coefficients for these parameters; v is thevector of observed line variances and Y is the sampling variance-covariance matrix for the line variances.However, this equation cannot be solved unless one is willing to assume that the parental lines arecompletely homozygous. This assumption is likely inappropriate for wild species of sticklebacks (seetable 4.1). Hayman (1960b) developed a maximum likelihood procedure to circumvent this problem andobtain predicted line variances under the additive model. Y is reduced to a diagonal matrix with variance100elements, 2v?I n; the variance component vector is reduced to Var(z1),Var(z), and Var(s), thesegregation variance; and the coefficients in M are changed to those from equations predicting linevariances from parental and segregation variance (Hayman 1960b). Equations 5 and 6 are then iterated(substituting the elements of ‘ into Y) until the elements of stabilize.Estimates ofgene numberI used a biometrical approach to estimate the minimum number of genes contributing todifferences between limnetics and benthics in the four morphological characters. The original method ofCastle (1921) and Wright (1968) was intended for use with divergent inbred lines, but was extended byLande (1981b) for use with wild populations. The method! employed incorporates modificationssuggested by Cockerham (1986) to correct for sampling variances in the estimates of parentalpopulations. The formulae are,-- Var(1)- Var() (7)lie8Var(s)Var[Var(s)] (8)Var(lie)= h - +[(EN) - ()]2 [Var(s)]2The method assumes the following: additivity of mean phenotypes ofF1, F2 and parental lines;line variances plotted against their means have a specific geometric relationship (i.e., they form a triangle,see fig. 4.2); normality of data in each line; and measured individuals should be unrelated and from largepopulations. The resulting he rarely exceeds the recombination index, which in higher plants and animalsis usually one to several times the haploid chromosome number (Lande 198 ib). If the assumptions aremet the method provides “reasonably accurate” predictions with the following sample sizes: n = 20-30 forP1,P2,F1; n 100 for second generation hybrids (Lande 198 ib). Failure to meet assumptions of this101method generally bias the estimate downward, sometimes substantially (Zeng et al. 1990; Hill andCaballero 1992). The estimate is therefore ofien referred to as the minimum number of loci contributingto divergence of the measured characters.If gene action is additive and loci are unlinked then an upper bound may be placed on the locus ofgreatest effect (Lande 198 lb). Additionally, if one parental line is fixed for “positive” loci and the otherparental line is fixed for all “negative” loci, then a lower bound may also be placed (Lander and Botstein1989).bounds on magnitude of locus of greatest effect 1 I 1 % 1/ ‘i (9)Due to the small number of replicates in backcross lines, I calculated 1 using three estimates ofVar(s) and Var[Var(s)]. Estimate 1 uses Var(s) and Var[Var(s)j from the maximum likelihood approachdescribed in the previous subsection (analysis of line variances); estimates 2 and 3 use equations fromLande (1981),Var(s) = Var(F2)- Var(F1) (10)Var[Var(s)j 2[Var(F)]I N + 2[Var(F1)12 / N1 (11)Var(s) = Var(F2)-{ [2Var(F1)+ Var(P1)+ Var(P2)]I 4} (12)Var[Var(s)] 2[Var(F)]I N1 + ½[Var(F1)]2I N1 +1/a[Var(P)}2I N1 + h/a[Var(P2)]2/ Np2 (13)102RESULTSCharacter values for the four traits are provided in table 4.1. Hybrid lines were intermediatebetween limnetics and benthics for each trait. I tested normality of the data for the four trails usingquantile-quantile plots. Both gill raker traits fit normal expectations well, however the armor characters fitpoorly. Non-normality of the data appeared to result from the truncated nature of the distribution of thesedata. Armor phenotypes are restricted to be greater than or equal to 0. Zero values were frequentlyobserved, especially in benthics. When quantile-quantile plots were constructed from observations greaterthan 0, the fit to normal expectations was good.Tests of sex differences and maternal effects in the four traits are given in table 4.2. The testsindicate that differences in sex are not statistically significant for any of the traits (though sex differencesapproach statistical significance for gill raker length). I therefore pooled data from both sexes forsubsequent analyses.Tests of maternal effects in F1 hybrid males were significant for pelvic spine length andapproached statistical significance for gill raker number. When mean values from the two reciprocalcrosses are transformed back into their original units, the difference in spine length is approximately 0.5mm. Despite evidence of maternal effects I pooled reciprocal crosses for all characters for subsequentanalyses. Reciprocal crosses are weighted equally since the number of families measured is the same foreach cross. However, maternal effects inflate the variance of the F1 line thereby deflating estimates ofsegregation variance. Additionally, if the maternal effect is large it may lead to spurious conclusions ofdominance in the joint-scaling tests.Character means with standard errors are shown in figure 4.1. The dotted lines join observedparental means. If limnetics and benthics have diverged primarily in genes with additive effects then onewould expect character means for each hybrid cross to fall along this line. If the species have also103diverged in genes with dominance effects then one would expect character means for hybrid crosses to bedisplaced from the line—the means for F2 hybrids and backcrosses should be displaced (perpendicularly)from the line half as much as F1 hybrids. The effect of dominance is measured by. the displacement fromthe line.Observed line means for gill raker number fit the expectations best of the five traits. The nonlinear relationships for the other characters do not conform to expectations of an additive model.Joint-scaling tests of mean phenotypes provided parameter estimates t ix and (table 4.3).Statistical tests (table 4.4) suggested that epistasis (in addition to additive and dominance effects)contributes to genetic divergence of limnetics and benthics for two of the characters, gill raker length andgrowth rate. Rejection of the simple additive model indicated dominance effects (in addition to additiveeffects) are important for plate number and pelvic spine length. A simple additive model was accepted foronly gill raker number.Observed trait means and variances (± standard errors) of each cross are shown in figure 4.2. Forcomparison the predicted means and variances from the additive model form the depicted triangles.Observed means and variances for the four traits are often located well away from their expectedlocations on the triangles indicating a poor fit to the additive model. However, the often substantialstandard errors suggest that additive models should not be rejected outright.Estimates of gene number (table 4.5) varied tremendously among the four characters, but therewas a general consistency among the three estimates of ne within each trait. However, since only gillraker number fit a simple additive model, the other estimates may be unreliable. The very large standarderrors for some estimates (especially gill raker number) compound the unreliability. Gene numberestimates for pelvic spine length were consistently low (0.2 to 1.3). The general finding for the othercharacters is that they involve more than simply one or two genes, though the possibility of genes of large104effect should not be ruled out (especially for pelvic spine length), as indicated by the broad range inmagnitude of locus of greatest effect.DISCUSSIONThe current view of speciation of stickleback species pairs is that two colonizations, separated intime by an unknown number of years, gave rise to first the benthic and then the limnetic species (McPhail1993, 1994). How did the ancestral marine sticklebacks respond to the new selective enviromnent oflakes, which is thought to differ in a number of ways from that of the marine environment? Since thetraits I studied are of ecological significance, they would have been shaped by the new selectionpressures. Exploring the genetic divergence between limnetics and benthics helps to answer how thesetraits may have responded. Genetic differences of limnetics and benthics suggest that evolutionaryresponses to the new selection pressures differed among characters.Joint-scaling tests (table 4.3) suggested that epistasis contributes significantly to geneticdivergence between limnetics and benthics for gill raker length and growth rate, but does not contributesignificantly to genetic divergence in the other three traits. The simple additive model was rejected forplate number and pelvic spine length, suggesting that dominance effects play a role in the geneticdivergence of this trait. Maternal effects show up as “pseudo-dominance”, so the maternal effect found forpelvic spine length may have affected the conclusion of significant dominance in this trait. However, itseems likely to have augmented an already existing effect rather than to have produced the entireeffect—the other armor character showed no maternal effects yet had significant dominance effects. Gillraker number was the only one of the five traits I examined for which a simple additive model of geneticvariance was not rejected. Taken together the joint-scaling tests suggest that adaptive divergence betweenlimnetics and benthics has taken place through a variety of genetic mechanisms specific to different traits.105Any genetic models of stickleback speciation should acknowledge that genetic divergence incorporatesmore than additive components of genetic variance.Despite the inability of a simple additive model to explain phenotypic differences among lines,figures 4.1 and 4.2 suggest that additive models may often do a reasonable job—the observed means andvariances are generally within a single standard error of their expected values, and almost always withintwo standard errors. The truncated nature of data for armor characters may lead to the poor fit for benthicsand benthic backcrosses in figure 4.2c and d. Additionally, although dominance effects are obvious forthe armor characters in figure 4. ic and d, it is equally apparent in all of the figures that there is a muchgreater additive component to phenotypic differences, save for spine length.There have been three recent studies similar to mine. Hard et al. (1992, 1993) examined adaptivedivergence between northern and southern populations of the mosquito, Wyeomyia smithii. Northernpopulations undergo diapause through a lengthy winter, while southern populations have a short diapause.Hard et al. made first and second generation lines between one northern and two southern populations.Genetic analysis suggested that divergence in diapause included additive effects, dominance, andsignificant epistasis. Estimates of gene number affecting divergence in diapause were 14 for one set ofcrosses, and 19 for the other set. However, the authors stressed the unreliability of these estimates giventheir evidence of significant nonadditive genetic effects for diapause.The other two studies investigated differences among wild species of the plant genus Mirnulus.Macnair and Cumbes (1989) examined a wide array of floral and morphological differences; and Fensteret al. (1995) examined differences in floral development. Using line crosses, Macnair and Cumbes (1989)found evidence for both significant epistasis and dominance (in addition to additivity) in a number offloral characters. Gene number estimates varied from three to seven for some characters but suggestedcontrol by major genes for other characters. Fenster et al. (1995) used different species but also foundevidence for epistasis and concluded that developmental traits in Mimulus are primarily polygenic.106The pattern of genetic divergence found between populations of Wyeomyia, and between speciesof Mimulus is similar to that between species of Gasterosteus. It would be interesting to know whethergenetic differences between recently diverged species are generally similar to divergence betweenpopulations which have not speciated. Such comparisons would tell us how often substantial geneticreorganization is associated with speciation.Implications ofepistasis and dominance in morphological traitsThe finding of significant epistasis in one of the four morphological traits measured here hasimplication for reproductive isolation between limnetics and benthics. The morphological traits Iexamined have a known ecological hinction, and therefore a direct tie to fitness. If there is significantepistasis in the expression of an ecologically important trait we should therefore expect to see epistasis infitness also. I predict that since gill raker length determines (at least in part) an individual’s feedingefficiency, that there should be epistasis for feeding efficiency and fitness in the wild. The extent ofepistasis in fitness depends though on the ability of individuals (hybrids especially) to mitigate poorfeeding efficiency by choosing the most profitable foraging habitat for their phenotype.The implications of dominance in morphological characters are similar to those for epistasis. Aswith epistasis, fitness effects from dominance may also depend on the ability of individuals to choose thebest habitat for their phenotype, though this may not always be possible. For example, dominance inarmor characteristics would tend to give “excessive” armor to otherwise benthic-like hybrids (e.g., benthicbackcrosses). Reist (1980) and Zyuganov (1989) showed that well-armored individuals from populationsof brook sticldeback (Culea inconstans) and ninespine stickleback (Pungitius pungitius) suffered thegreatest mortality from insect predation. Insect predators tend to hunt in the covered littoral habitat that ismost efficiently exploited by fish with benthic feeding morphology. Thus, patterns of inheritance infeeding and armor traits may lower hybrid fitness below that expected if all traits were additive.107Figure 4. le shows mean growth rate in the lab for the six lines. The deviations of mean hybridgrowth rates around the line in the figure suggest epistasis for this trait—an interpretation supportedstatistically by the joint-scaling tests (table 4.4). Growth in the lab is essentially a baseline growthrate—conditions should not have favored one line over another since differences in feeding efficiencywere negated by ad libftum feeding (chapter one). Growth differences among lines in the field may bequite different than those in the lab because foraging efficiency is determined in part by the morphologyof the fish (Schiuter 1993; chapter two). For example, fish with long gill rakers are more efficientplanktivores. Any epistasis in feeding efficiency would be in addition to epistasis already observed forgrowth rate in the lab. It would be fascinating to measure growth rates in the wild (e.g., in enclosures) forindividuals from the six lines and attempt to partition the effects of rearing in the wild from rearing in thelab.Gene number estimatesThe Castle-Wright-Lande method of gene number estimation gave a very broad range ofestimates. Estimates suggested that differences between limnetics and benthics in three of the traits (gillraker number, gill raker length and plate number) are determined by the action of more than three loci.The estimates suggested that differences in pelvic spine length are determined by a single locus.I now turn to the reliability of these estimates, since there are a variety of assumptions andshortcomings in the method (Zeng et al. 1990). The estimates are sensitive to hybrid breakdown andlimited by the recombination index of the organism. There is no evidence in Paxton Lake sticidebacks ofbreakdown in F1 or F2 hybrids, though there is some evidence of breakdown in the backcrosses (chapterone). Heterosis is therefore unlikely to have affected the estimates significantly. Nevertheless, as aprecaution backcrosses were not used to estimate Var(s) in the latter two of the three estimation methods(see equations 10 to 13) due to the evidence of breakdown in these lines. The recombination index of G.108aculeatus is not known, but the haploid chromosome number is 21 (Chen and Reisman 1970), Therecombination index is therefore expected to be high, and is not likely to be a severe restriction onestimates. If the loci affecting the traits measured here are not spread throughout the genome, however,linkage may cause substantial underestimation (Zeng et al. 1990).There are other significant shortcomings. The method assumes that divergence in meanphenotypes can be fit to an additive model, at least on some scale. Significant epistasis and dominanceeffects for four of the five characters suggest violation of this assumption and standard transformationsdid not render additive the differences in mean phenotypes. The additional assumption of normality ofcharacter distributions could not be met for either of the armor characters. The assumptions were mostclosely met for gill raker number, however the very large standard errors restrict the reliability ofestimates even for this trait. These reservations suggest that all the estimates of gene number (table 4.5)should be taken with some caution.Despite the reservations outlined above, my results are consistent with the view of adaptation bymore than just one or two major genes for most of these traits. If we accept that the estimates of genenumber and joint-scaling test results can be taken as at least rough approximations of a minimum locinumber, they suggest that selection on three of the four morphological traits would be spread over severalto many loci. (The finding of epistasis for growth rate suggests that this trait is also polygenic.) However,due to the finding of epistasis, the response to selection on these characters may not be straightforward.My study can be added to a growing body of evidence (Macnair and Cumbes 1989; Hard et al. 1992,1993; Cabot et al. 1994; Davis et al. 1994; Lai et al. 1994) that says that all characters do not fit a simpleadditive model. Epistatic interactions between loci may generally play a significant role in the geneticdivergence between populations and species.109Implicationsfor the genetics of adaptationIn a recent paper, Orr and Coyne (1992) reviewed studies of the genetics of adaptive characters.Their findings suggested that there was, in fact, little empirical support for the neo-Darwinian view thatadaptive traits are coded for by many genes of small effect. The authors were able to produce only eightstudies which had looked at adaptive differences between natural populations. (Many others have lookedat traits of unknown significance using lab populations and inbred lines.) The conclusion from thesestudies was surprising: in six of the eight studies the divergent characters were coded for by only onegene. Their basic conclusion was that “while some adaptations are clearly based on many genes of smalleffect, major genes are also sometimes involved”. The challenge remains to find out how often adaptationinvolves major genes.There is room for substantial skepticism toward Orr and Coyne’s conclusion. First, the characterswe study affect our conclusions. Consider a character that is only present in one of two closely relatedspecies. Many genes may be required to build a complex character, but only one to delete it. Therefore, asingle point mutation may be the only difference between two species, even if the initial adaptation(present in ancestors) requires the action of many genes to produce the functioning character. Thiscriticism applies to a broad range of characters.Further criticisms of the Off and Coyne (1992) study revolve around methodologicalshortcomings of the individual studies they reviewed. 1) There is little evidence that the character stateslisted in their table 1 are in fact “adaptive differences”. 2) Some of the characters might be expected to beunder the control of major genes simply because the trait itself is dimorphic. For example, the expressionof diapause in the Drosophila and Papilio species investigated is either present or absent. It seems likely apriori that a dimorphic trait will be under the control of one gene. Hard et al. (1992, 1993) examinedgenetic divergence in the length of diapause between northern and southern populations of Wyeomyiapopulations and found the differences to be coded for by several to many loci. 3) The most common110approach to gene number estimation in the tabulated studies was the biometrical method (Lande 1981)which has several assumptions and drawbacks outlined above. The lack of genetic analysis to test theseassumptions is a serious deficiency in all the tabulated studies. Violations of assumptions made by thismethod generally bias estimates of loci number downward (Hill and Caballero 1990). 4) Finally, theextent of hybrid breakdown in most crosses was not reported. More sophisticated methods (e.g.,association of characters with known genetic markers) should be used when species cannot be crossedfreely.The conclusion by Off and Coyne (1992) that we know little about the genetic basis of adaptivedifferences among populations will remain until there have been many studies of genetic differencesamong wild taxa. The biometrical approach to gene number estimation is valuable for estimating theminimum number of loci, but the method has many inherent weaknesses, and little power to distinguishbetween many genes of equal effect and one or two major genes with many modifiers. The most valuabledata will likely come from finer grained genetic studies (e.g., Cabot et aL 1994; Davis et aL 1994; Lal etal. 1994).My study offers a view that is nevertheless quite similar to that of Off and Coyne (1992). Myresults are consistent with the view that some adaptive differences between species of stickleback arepolygenic while others may be explained by very few genes of large effect. A more completeunderstanding of relative magnitude of effects of genes coding for these traits must wait for a finergrained genetic analysis.TABLE4.1.Charactermeansandstandarderrorsforthesixlines.Valuesaremeansoffamilymeansandtheirstandarderrors.no.ofno.ofgillrakergillrakerplatepelvicspinegrowthspeciesindividualsfamiliesnumberlengthnumberlengthratelimnetic331624.156(0.257)0.409(0.023)13.052(0.316)1.744(0.015)-0.0306(0.0189)benthic301518.239(0.269)0.064(0.016)1.250(0.322)0.087(0.091)0.0299(0.0271)fl882121.029(0.215)0.259(0.017)9.256(0.198)1.487(0.043)0.0300(0.0219)f21223121.177(0.154)0.260(0.015)8.613(0.261)1.191(0.078)0.0034(0.0090)limneticbackcross33823.244(0.395)0.296(0.026)11.338(0.681)1.557(0.080)-0.0424(0.0104)benthicbackcross30819.706(0.340)0.111(0.018)5.388(0.890)0.990(0.166)-0.0322(0.0302)112TABLE 4.2. t-tests for (A) character differences between sexes in first generation fish (benthics, limneticsand F1 hybrids); (B) maternal effects in F1 males from reciprocal crossesmale(n=24)21.10 (0.42)0.30 (0.03)9.16 (0.65)1.30(0.12)p-value (df= 46)p=O.6l9p0.068p=O.657p=0.994B.traitgill raker numbergill raker lengthplate numberpelvic spine lengthlimnetic x benthic d’(n=10)21.47 (0.36)0.26 (0.03)9.47 (0.28)1.42 (0.02)benthic x limnetic a’(n=8)20.49 (0.29)0.32 (0.03)9.33 (0.36)1.53 (0.03)p-value (df = 16)p=O.O57p=0.l9Op=0.755p=0.OlOA.trait female(n=24)gill raker number 21.43 (0.50)gill raker length 0.23 (0.03)plate number 8.72 (0.72)pelvic spine length 1.30 (0.15)TABLE4.3.Parameterestimatesfromjoint-scalingtestsonmeansofthefivetraits.The joint-scalingtestswerecarriedoutfortwogeneticalmodels,theadditive-dominancemodelandthesimpleadditivemodel.additive-dominancemodelgillrakergillrakerplatepelvicspinegrowthnumber(s.e.)length(s.e.)number(s.e.)length(s.e.)rate(s.e.)çt21.240(0.070)0.248(0.005)8.394(0.108)1.212(0.021)-0.0047(0.006)-3.000(0.139)-0.173(0.010)-5.551(0.214)-0.827(0.033)0.0400(0.013)6-0.086(0.102)0.021(0.008)1.006(0.124)0.287(0.023)0.0063(0.013)additivemodelgillrakernumber(s.e.)t21.235(0.070)-2.996(0.139)114TABLE 4.4. Goodness-of-fit (cu-square) test statistics from joint-scaling tests on means of the fivecharacters.trait additive-dominance model (df = 3) additive model (df =4)gill raker number 2.31; p> 0.50 a 2.58; p> 0.50 agill raker length 9.67; p < 0.05plate number 1.86; p > 0.50 a 49.97; p < 0.001pelvic spine length 1.96; p > 0.50 a 82.93; p < 0.001growthrate 9.00; p <0.05 ba do not reject the modelreject the modelTABLE4.5.Estimatesofgenenumberandstandarderrorsforthefourmorphologicalcharacters.Alsoestimatedistherangeinmagnitudeofthelocusofgreatesteffect,usingf1.s.e.magnitude(proportionoftotal)iraitgillrakernumbergillrakerlengthaplatenumberapelvicspinelengthahel29. 4.1. Observed character means and standard errors for the five traits measured in each of the sixlines. An additive model would predict that all means lie along the dotted lines drawn between parentalmeans.growthrate(residualsfromlocalregression)platenumbergillrakernumberpppLLLr’)r’)-0L£00£0.OC0i).CD0IIIIIIIIIIII4.0:00-.c.0oB0-S..-.-__________________pelvicspinelengthgill rakerlengthooLI)o000ooiooi0-IIIIIIII-.--a-.S.F-..S.0.h.S.0(D—.-.-—-pS.’....pN...118FIGURE 4.2. Observed character variances (± s.e.) plotted against observed character means (± s.e.) forfour morphological traits in each of the six lines. The triangles depict the maximum likeithood predictedvalues for a simple additive model.ci) C) > ci) C.) C Cu 1 cu >0510meanplatenumber15ci) 0 C Cu I Cu > benf2.Nmfibbkci) C) C Cu Cu >bbbkfi[be+T+ 14 12 10 8 6 4 20.0140.0120.0100.0080.0060.0040. 0.018202224meangillrakernumberCbbkbe0. DISCUSSIONIn the general introduction I noted that my motivation has been fuelled more by the desire tounderstand the ultimate causes of speciation than to describe proximate mechanisms of premating andpostmating isolation. I chose to evaluate mechanisms of reproductive isolation at multiplelevels—genetic, developmental, behavioral and ecological—because assessing these different levelspresents a clearer and fuller picture of postmating isolation in Paxton Lake sticklebacks than examinationin finer detail at only one of these levels. My results shed light on the relative magnitude of differentmechanisms of postmating isolation, and they highlight where we should concentrate our efforts next. Inthis discussion I will briefly review the results from this study, assess standard explanations ofreproductive isolation with results from my experiments with these fish, and then discuss areas ofprofitable future research.The thesis.—Chapters one and two set up and carried out a test of “ecological speciation”—speciation due to divergent niche-based selection. The results demonstrated that ecological trade-oils inresource use actively determine postmating isolation by selecting against hybrids and other individualswith intermediate morphology. Data in chapter one provide strong evidence that selection against hybridsis not due to genetic incompatibility between limnetics and benthics. Hybrid offspring were generally asviable as the parental types, although there was some evidence that backcrosses were less viable. Hybridshatched in the same proportions as parentals, grew at the same rate, were developmentally normal, andcould provide adequate parental care.Chapter two offered a rather different view of hybrid fitness. I demonstrated that hybrids, whenplaced in the wild, were not as viable as parentals, and suggested that their lower growth rates were mostlikely due to morphological intermediacy which renders them inefficient foragers on the two main food120121resources. The conclusion of very high hybrid fitness in the lab made for a strong test of ecologicalspeciation, since it is the difference between fitness in the lab and fitness in the wild which measures thecontribution of niche-based selection to reproductive isolation. The concept of trade-offs is a central tenetin much of evolutionary theory and this was certainly not the first demonstration of a trade-off. However,this was one of the first demonstrations that trade-offs can be a dominant mechanism of posimatingisolation. I hypothesized that limnetics and benthics are not somehow unique, that trade-offs in resourceuse may commonly provide reproductive isolation, especially at the early stages of speciation.Chapter three addressed the issue of behavior and reproductive isolation. Premating isolation isinvariably a behavioral phenomenon since avoiding heterospecifics requires preferring conspecifics overheterospecifics and/or being in separate locations. Theory suggests that divergent mating preferences mayresult not just in premating isolation, but also in postmating isolation by selecting against hybrids (Fisher1930). The idea that sexual selection against hybrids may be an important component of reproductiveisolation has rarely been examined before. I tested the idea with sticklebacks. My findings confirmedsubstantial premating isolation between limnetics and benthics, even in the absence of habitat preferences.However, sexual selection against hybrids remains uncertain pending further work. I showed that hybridmales are preferred to an intermediate extent by females of both parental species, but that habitatpreferences must accompany female mating preferences for there to be sexual selection against hybrids.Habitat choice exists in males, but we do not yet have data on female habitat preferences.In chapter four I explored the genetics of ecological speciation. I used quantitative geneticanalysis on two feeding traits, two armor traits, and one fitness component. These are some of the sametraits which determine the divergent ecologies of limnetics and benthics; they affect feeding efficiency indifferent habitats, and divergent strategies to predator defense. Therefore, how the traits are inheriteddirectly influences postmating isolation, because inheritance patterns affect how genes from the twospecies interact when combined in hybrid offspring. Genetic divergence between limnetics and benthics122included additive, dominance, and epistatic effects. The magnitude and statistical significance of theeffects depended on the particular trait examined, but since the traits are of ecological importance therelative magnitude of each of these effects influences hybrid fitness and therefore postmating isolation.Epistasis (and/or dominance) in any of these traits should result in epistasis (and/or dominance) in hybridfitness. The next step in genetic analysis of sticklebacks should be to examine these potential effects byraising hybrid lines in the wild. Surprisingly few studies have examined how closely-related species differgenetically, and none has tried to extend analysis by raising hybrid lines in the wild.Traditional explanations of speciation.—The traditional approach when discussing speciation isto divide explanations, models and data into those which may account for speciation either with orwithout geographic isolation—that is, a division into sympatric versus allopatric speciation. The tensionbetween researchers with these differing world views is legendary (see e.g., Mayr 1942, 1963; White1978; Bush 1975, 1994; Rice and Hostert 1994). This tension has led to numerous excellent studies whichhave focussed our attention on particular processes, for example the idea that recombination canoverwhelm diversifying selection (Felsenstein 1981). Yet despite the ancient discord there are very fewunequivocal examples of sympatric speciation in animals, and the allopatric explanation remains the nullmodel. Many believe we will never know the relative rates of different modes of speciation (Coyne1994), though some have attempted to quantify this (e.g., Lynch 1989; Chesser and Zink 1994).My experience suggests that there is not always a straightforward division between sympatric andallopatric speciation. Diversification in sticklebacks, for example, seems to have occurred both inallopatry and sympatry. It therefore seems unnecessary and misguided to try to pigeonhole speciationevents in this way. I believe a more profitable division concentrates on modes of process rather thangeography. There are three major classes of process in species formation: 1) genetic drift, 2) adaptiveprocesses, and 3) by-products of adaptation. These three classes are not mutually exclusive, and123adherence to only one could be misleading. I will now consider their explanatory power for sticklebacksin the light of my results. Along the way I will also point to some outstanding questions.Speciation by genetic drtfL —My data on hybrid fitness offer little evidence for completespeciation by genetic drift. The very high genetic compatibility between limnetics and benthics (chapterone) refutes any suggestion of incompatibility arising through chromosomal speciation. Substantialreductions in hybrid fitness would be seen in the lab if isolating mechanisms such as polyploidy orchromosomal translocations had arisen.The extent of genetic drift is more difficult to assess because the nature of sticklebackcolonizations is probably unknowable. There may have been founder effects if Paxton Lake wascolonized by a small number of individuals from the sea, and population numbers had remained low forsome time after colonization (Barton and Charlesworth 1984). If McPhail’ s (1993) double colonizationscenario is correct then some differences between limnetics and benthics may be attributable to geneticdrift, especially during the period when the initial colonizers and marine fish were separated. However, Ibelieve overall that evidence strongly supports the view that the effects of drift have been minor incomparison to explanations of species differences based on selection. Chapters one and two stronglysupport the notion that morphological and ecological differences between limnetics and benthics areadaptive, and not simply random. It is possible that premating isolation has been molded somewhat bygenetic drift, and this would best be tested by lab experiments like those of Meffert and Bryant (1991)where populations are subjected to repeated bottlenecks. However, I am skeptical that there has beenmuch isolation in sticklebacks by genetic drift, since experimental results in other species do not supportthe idea of substantial premating isolation evolving by genetic drift (Rice and Hostert 1994).124Speciation as an adaptation.—Data in chapter two demonstrated disruptive selection andtherefore support the view that a discontinuous or bimodal distribution of morphologies and ecologicalstrategies is favored in Paxton Lake. However, the extreme geographical restriction of the sticidebackspecies pairs suggests that speciation has not been by purely adaptive mechanisms: what would bepreventing such speciation in the many hundreds of similar lakes with solitary populations? Adaptivemechanisms probably provide at most only a partial explanation for stickleback speciation. One potentialrole is through reinforcement—the evolution of premating isolation in direct response to hybridinviability or sterility. Substantial theoretical and experimental results suggest that the process is unlikely(Spencer et al. 1986; Butlin 1989; Rice and Hostert 1994), but there is also some empirical and theoreticalevidence in its favor (e.g., Coyne and Off 1989; Howard 1993; Liou and Price in press). Whilereinforcement is itself an adaptive process, the origin of hybrid inviability need not be.There is still much we do not know about reinforcing selection imposed by resource distributionsin these lakes. The experiment in chapter two was one of the first to attack the problem. We don’t know,for example, whether an intermediate habitat exists, or whether solitary populations are under disruptiveselection. It is possible that disruptive selection only appears as the morphological distribution of theconsumers broadens (Rosenzweig 1978; Wilson 1989). It would be easier to evaluate a potential role forreinforcement if we knew how the strength of disruptive selection changed over the course of divergence.Regardless, the observed selection against hybrids (chapter two) does favor greater premating isolation.The role of reinforcing selection in the speciation of sticklebacks will be difficult to assess, but is a topicworthy of further investigation.Speciation as a by-product of adaptation.—The final class of speciation processes viewsreproductive isolation evolving as an incidental by-product of adaptation. This view of speciation isprobably the most widely held and the best supported by evidence. The classical view is that of Fisher125(1930) and Mayr (1942, 1963), wherein two isolated populations diverge through local adaptation. Thetwo populations may be molded by similar selection pressures but diverge anyway due to adaptationalong different pathways, and/or the two may follow different pathways because selection is significantlydifferent for the two populations. Given enough time the populations accumulate sufficient differences tomake cross fertilization difficult or impossible and hybrid offspring inviable or sterile. I believereproductive isolation in Paxton Lake sticklebacks has evolved primarily as a by-product of adaptation.McPhail’ s double colonization scenario (1993) offers the view that an initial colonization fromthe sea gave rise to benthics, and a subsequent colonization gave rise to limnetics. Such a view issupported by geological evidence for sea level fluctuations in the Strait of Georgia (Mathews et al. 1970).Following colonization the now landlocked marine type needed to adapt to the lake environment. For theinitial colonizers these adaptations probably included a shift to foraging for benthic organisms (ratherthan a purely planktivorous habit) and seeking cover to avoid predation (rather than relying on armor).Data in chapter two suggest that this niche shift would affect postmating isolation. I showed thatadaptations allowing increased foraging efficiency on benthos could lead to postmating isolation betweenbenthivores and planktivores because offspring between the two would be intermediate and have lowerfeeding efficiency than either of the parental types. Secondly, in chapter four I showed that modificationof traits responsible for the ecological strategies of limnetics and benthics has led to dominance andepistasis in some traits. Genetic divergence from adaptation to different resources likely increasespostmating isolation through dominance and epistasis in hybrid fitness.Future work.—The evolution of premating isolation and habitat preferences remain the biggestmysteries of stickleback speciation. Virtually nothing is known about how premating isolation evolves inany system (Endler 1989). At this point each of the three processes outlined above remain potentialexplanations for the evolution of premating isolation in sticklebacks. But, since postmating isolation126appears to have been molded primarily by ecological selection pressures, it seems a reasonable place tostart future investigations of premating isolation. In particular, I think the following question should beaddressed: Has adaptation within the stickleback visual system for efficient foraging shaped femalemating preferences, and thereby driven the evolution of premating isolation? Whether adaptation todifferent resources affected the evolution of premating isolation in sticklebacks is one of the most excitingprospects for future research, and has implications for many speciation studies.The visual system of most organisms has evolved almost entirely in the context of finding foodand to avoid being found as food. If there are trade-offs in the ability of the visual system to carry out itsfunctions there may be inescapable biases which lead to predictable mating preferences (Endler 1992).There are indications that trade-offs exist: visual acuity may come at the expense of absolute sensitivity(e.g., to movement), and discrimination at one wavelength may be traded off for discrimination at another(King et al. 1993; McDonald and Hawryshyn in press).Characteristics of the visual system may vary in a way similar to that of morphology. Thephysical characteristics of prey items utilized by sticklebacks vary enormously between plankton andbenthos. Plankton are generally very small relative to benthic prey and appear against a very differentbackground; planktonic prey (e.g., copepods) move fast for their body size whereas most benthicorganisms move slowly. Studies of the visual system in other fish species have shown marked differencesin cellular structure in the retina between species which forage on planktonic versus benthic prey (Collinand Mi 1994). It is therefore reasonable to hypothesize differences in the visual system of sticklebackswhich correlate with the types of resources exploited. Sensory biases resulting from visual adaptation maylead to female preferences for particular male mating traits or to habitat preferences, and both could leadto premating isolation and sexual selection against hybrids.If habitat preferences and premating isolation have evolved in this way there would be evenstronger evidence of speciation as a by-product of adaption to divergent niches. Although there still127remains much to do, an examination of the ultimate causes of speciation in sticklebacks is well underway.I am confident that studies of sticklebacks, combined with the many superb studies currently underway inother systems, will help to maintain the vigorous discipline of speciation studies.LITERATURE CITEDArnold, M.L. and B.D, Bennett. 1993. Natural hybridization in Louisiana irises: genetic variation andecological determinants, in R.G. Harrison (ed.) Hybrid zones and the evolutionary process. OxfordUniversity Press, Oxford, UKBaker, J.A. 1994. Life history variation in female threespine stickleback. in M.A. Bell and S.A. Foster (eds.),The evolutionary biology of the threespine sticideback. Oxford University Press, Oxford, UKBakker, T.C.M. and M. Milinski. 1991. Sequential female choice and the previous male effect in sticklebacks.Behavioral Ecology and Sociobiology 29: 205-210Barton, N.H. and B.O. Bengtsson. 1986. The barrier to genetic exchange between hybridizing populations.Heredity 56: 357-376Barton, N.H. and B. Charlesworth. 1984. Genetic revolutions, founder effects, and speciation. Annual Reviewof Ecology and Systematics 15: 133-164Beaniish, R.J. 1982. Lampetra macrostoma, a new species of freshwater parasitic lamprey from the west coastof Canada. Canadian Journal of Fisheries and Aquatic Sciences 39: 736-747Bell, M.A. and S.A. Foster. 1994. The evolutionary biology of the threespine stickleback. Oxford UniversityPress, Oxford, UKBenkman, C.W. 1993. Adaptation to single resources and the evolution of crossbill (Loxia) diversity.Ecological Monographs 63: 305-325Briggs, B.G. 1962. Interspecific hybridization in the Ranunculus lappaceus group. Evolution 16: 372-390Bush, G.L. 1969. Sympatric host race formation and speciation in frugivorous flies of the genus Rhagoletis(Diptera, Tephritidae). Evolution 23: 237-251Bush, G.L. 1975. Modes of animal speciation. Annual Review of Ecology and Systematics 6: 339-364Bush, G.L. 1995. Sympalric speciation in animals: new wine in old bottles. Trends in Ecology and Evolution9: 285-288Butlin, R. 1989. Reinforcement of premating isolation. in D. Otte and J.A. Endler (eds.) Speciation and itsconsequences. Sinauer Associates, Sunderland, USA.Butlin, R. 1995. Introduction. in R. Butlin (ed.) Species and speciation. Cambridge University Press,Cambridge, UKCabot, E.L., A.W. Davis, N.A. Johnson and C.-I. Wu. 1994. Genetics of reproductive isolation in theDrosophila simulans Clade: complex epistasis underlying hybrid male sterility. Genetics 137: 175-189Carson, H.L. and A.R. Templeton. 1984. Genetic revolutions in relation to speciation phenomena: thefounding of new populations. Annual Review of Ecology and Systematics 15: 97-131128129Castle, W.E. 1921. An improved method of estimating the number of genetic factors concerned in cases ofblending inheritance. Science 54: 223Cavalli, L.L. 1952. An analysis of linkage in quantitative inheritance, in Reeve, E.C.R. and C.H. Waddington(eds.), Quantitative Inheritance. Her Majesties Stationery Office, London, UKChen, T.-R. and H.M. Reisman. 1970. A comparative chromosome study of the North American species ofsticklebacks (Teleostei: Gasterosteidae). Cytogenetics 9: 321-332Chesser, R.T. and R.M. Zink. 1994. Modes of speciation in birds: a test of Lynch’s method. Evolution 48:490-497Cockerham, C.C. 1986. Modifications in estimating the number of genes for a quantitative character. Genetics114: 659-664Collin, S.P. and M.A. Au. 1994. Multiple areas of acute vision in two freshwater teleosts, the creek chub,Semiotilus atromaculatus (Mitchill) and the cutlips minnow, Exoglossum maxillingua (Lesueur).Canadian Journal of Zoology 72: 721-730Conover, D.O. 1992. Seasonality and the scheduling of life history at different latitudes. Journal of FishBiology 41B:161-178Coyne, J.A. 1987. Lack of response to selection for directional asymmetry in Drosophila melanogaster.Journal of Heredity 78: 119Coyne, J.A. 1992. Genetics and speciation. Nature 355: 511-515Coyne, J.A. 1994. Ernst Mayr and the origin of species. Evolution 48:19-30Coyne, J.A. and N.H. Barton. 1988. What do we know about speciation? Nature 331: 485-486Coyne, l.A and H.A. Off. 1989. Patterns of speciation in Drosophila. Evolution 43: 362-38 1Cruzan, M.B. and M.L. Arnold. 1993. Ecological and genetic associations in an Iris hybrid zone. Evolution47: 1432-1445Davis, A.W., E.G. Noonburg and C.-I. Wu. 1994. Evidence for complex genic interactions betweenconspecific chromosomes underlying female sterility in the Drosophila simulans Clade. Genetics137: 191-199Day, T., J. Pritchard and D. Schiuter. in press. Ecology and genetics of phenotypic plasticity: a comparison oftwo sticklebacks. Evolution.Diehl, S.R. and G.L. Bush. 1989. The role of habitat preference in adaptation and speciation. in D. Otte andJ.A. Endler (eds.) Speciation and its consequences. Sinauer Associates, Sunderland, USADobzhansky, T. 1937. Genetics and the origin of species. Columbia University Press, New York, USA130Dobzhansky, T. 1970. Genetics and the evolutionary process. Columbia University Press, New York, USADufresne, F., G.J. FitzGerald and S. Lachance. 1990. Age and size-related differences in reproductive successand reproductive costs in threespine sticklebacks (Gasterosteus aculeatus). Behavioral Ecology 1:140-147Dugatkin, L.A. and J. -G. Godin. 1992. Reversal of female mate choice by copying in the guppy (Poeciliareticulata). Proceedings of the Royal Society of London, series B 249: 179-184Endler, J.A. 1986. Natural selection in the wild. Princeton University Press, Princeton, USAEndler, J.A. 1989. Conceptual and other problems in speciation. in D. Otte and J.A. Endler (eds.) Speciationand its consequences. Sinauer Associates, Sunderland, USAEndler, J.A. 1992. Signals, signal conditions and the direction of evolution. American Naturalist 139: S 125-S153Feder, J.L., C.A. Chilcote and G.L. Bush. 1988. Genetic differentiation between sympatric host races of theapple maggot fly Rhagoletis pomonella. Nature 336: 61-64Felsenstein, J. 1981. Skepticism toward Santa Rosalia, or why are there so few kinds of animals. Evolution35: 124-138Fenster, C.B., P.K. Diggle, S.C.H. Barrett and K. Ritland. 1995. The genetics of floral developmentdifferentiating two species of Mimulus (Scrophulariaceae). Heredity 74: 258-266Fisher, R.A. 1930. The genetical theory of natural selection. Clarendon Press, Oxford, UKFoote, C.J., C.C. Wood, and R.E. Withler 1989. Biochemical genetic comparison of sockeye salmon andkokanee, the anadromous and nonanadromous forms of Oncorhyncus nerka. Canadian Journal ofFisheries and Aquatic Sciences 46: 149-158Foote, C.J., C.C. Wood, W.C. Clarke and J. Blackburn. 1992. Circannual cycle of seawater adaptability inOncorhyncus nerkt: genetic differences between sympatric sockeye salmon and kokanee. CanadianJournal of Fisheries and Aquatic Sciences 49: 99-109Futuyma, D.J. 1987. On the role of species in anagenesis. American Naturalist 130: 465-473Gaines, S.D. and W.R. Rice. 1990. Analysis of biological data when there are ordered expectations. AmericanNaturalist 135: 310-317Grant, P.R. 1986. Ecology and evolution of Darwin’s finches. Princeton University Press, Princeton, USAGrant, P.R. and B.R. Grant. 1992. Hybridization of bird species. Science 256: 193-197Grant, B.R. and P.R. Grant. 1993. Evolution of Darwin’s finches caused by a rare climatic event. Proceedingsof the Royal Society of London, series B 251: 111-117131Grant, V. 1994. Evolution of the species concept. Biologisches Zentralblatt 113: 401-415Hagen, D.W. 1973. Inheritance of numbers of lateral plates and gill rakers in Gasterosteus aculeatus.Heredity 30: 303-3 12Hard, J.J., W.E. Bradshaw and C.M. Holzapfel. 1992. Epistasis and the genetic divergence of photoperiodismbetween populations of the pitcher-plant mosquito, Wyeomyia smithii. Genetics 131: 389-396Hard, J.J., W.E. Bradshaw and C.M. Holzapfel. 1993. The genetic basis of photoperiodism and itsevolutionary divergence among populations of the pitcher-plant mosquito, Wyeomyia smithii.American Naturalist 142: 457-473Hayman, B.I. 1958. The separation of epistatic from additive and dominance variation in generation means.Heredity 12: 371-390Hayman, B.I. 1960a. The separation of epistatic from additive and dominance variation in generation meansII. Genetica 31: 133-146Hayman, B.I. 1960b. Maximum likelihood estimation of genetic components of variation. Biometrics 16:369-381Hill, G.E. 1990. Female house finches prefer colourful males: sexual selection for a condition-dependent trait.Animal Behaviour 38: 563-72Hill, W.G. and A. Caballero. 1992. Artificial selection experiments. Annual Review of Ecology andSystematics 23: 287-310Howard, D.J. 1993. Reinforcement: origin, dynamics, and fate of an evolutionary hypothesis. in R.G..Harrison (ed.) Hybrid zones and the evolutionary process. Oxford University Press, Oxford, UKHubbs, C.L. 1955. Hybridization between fish species in nature. Systematic Zoology 4: 1-20Hutchinson, G.E. 1957. Concluding remarks. Cold Spring Harbor Symposia on Quantitative Biology 22: 415-427Kaneshiro, K.Y. 1980. Sexual isolation, speciation and the direction of evolution. Evolution 34: 437-444Katakura, H., M. Shioi and Y. Kira. 1989. Reproductive isolation by host specificity in a pair ofphytophagous ladybird beetles. Evolution 43: 1045-1053King, R.B., J. K. Douglass, J.B. Phillips and C.L. Baube. 1993. Scotopic spectral sensitivity of the optomotorresponse in the green treefrog Hyla cinera. Journal of Experimental Zoology 267: 40-46Lai, C., R.F. Lyman, A.D. Long, C.H. Langley and T.F.C. Mackay. 1994. Naturally occurring variation inbristie number and DNA polymorphisms at the scabrous locus of Drosophila melanogaster. Science266: 1697-1702132Lande, R. 1981 a. Models of speciation by sexual selection on polygenic traits. Proceedings of the NationalAcademy of Sciences, U.S.A. 78: 372 1-3725Lande, R. 198 lb. The minimum number of genes contributing to quantitative in populations. Genetics 99:541-553Lande, R. 1982. Rapid origin of sexual isolation and character divergence in a dine. Evolution 36: 2 13-223Lande, R. and M. Kirkpatrick. 1988. Ecological speciation by sexual selection. Journal of Theoretical Biology133: 85-98Lander, E.S. and D. Botstein. 1989. Mapping Mendelian factors underlying quantitative traits using RFLPlinkage maps. Genetics 121: 185-199Leamy, L. and W. Atchley. 1985. Directional selection and developmental instability: evidence fromfluctuating asymmetry of morphometric characters in rats. Growth 49: 8-18Levins, R. 1968. Evolution in changing environments. Princeton University Press, Princeton, USALiou, L.W. and T.D. Price. in press. Speciation by reinforcement of pre-mating isolation. Evolution.Lynch, J.D. 1989. The gauge of speciation: on the frequencies of modes of speciation. in D. Otte and J.A.Endler (eds.) Speciation and its consequences. Sinauer Associates, Sunderland, USALynch, M. and J.B. Walsh. in preparation. Fundamentals of Quantitative Genetics. Volume One: Biology andestimation. Sinauer Associates, Sunderland, USAMacnair, M.R. and Q.J. Cumbes. 1989. The genetic architecture of interspecific variation in Mimulus.Genetics 122: 211-222Mather, K. and J.L. Jinks. 1982. Biometrical genetics: the study of continuous variation. 3rd ed. Chapman andHall, New York, USAMathews, W.H., J.G. Fyles and H.W. Nasmith. 1970. Postglacial crustal movements in southwestern BritishColumbia and adjacent Washington state. Canadian Journal of Earth Sciences 7: 690-702Mayr, E. 1942. Systematics and the origin of species. Columbia University Press, New York, USAMayr, E. 1963. Animal species and evolution. Harvard University Press, Cambridge, USAMcDonald, C.G. and C.W. Hawryshyn, in press. Intraspecific variation of spectral sensitivity in threespinestickleback (Gasterosteus aculeatus) from different photic regimes. Journal of ComparativePhysiology AMcKinnon, J.S. in press. Video mate preferences of female threespine sticklebacks from populations withdivergent male coloration. Animal Behaviour133McLennan, D.A. and J.D. McPhafl. 1990. Experimental investigations of the evolutionary significance ofsexually dimorphic nuptial colouration in Gasterosteus aculeatus (L.), the relationship between malecolour and female behaviour. Canadian Journal of Zoology 68: 482-492McPhail, J.D. 1984. Ecology and evolution of sympatric sticklebacks (Gasterosteus): morphological andgenetic evidence for a species pair in Enos Lake, British Columbia. Canadian Journal of Zoology 62:1402-1408McPhail, J.D. 1992. Ecology and evolution of sympatric sticklebacks (Gasterosteus): evidence for a species-pair in Paxton Lake, Texada Island, British Columbia . Canadian Journal of Zoology 70: 361-369McPhail, J.D. 1993. Ecology and evolution of sympatric sticklebacks (Gasterosteus): origin of the speciespairs. Canadian Journal of Zoology 71: 515-523McPhail, J.D. 1994. Speciation and the evolution of reproductive isolation in the sticklebacks (Gasterosteus)of southwestern British Columbia. in M.A. Bell and S.A. Foster (eds.), The evolutionary biology ofthe threespine stickleback. Oxford University Press, Oxford, UKMeffert, L.M. and E.H. Bryant. 1991. Mating propensity and courtship behavior in serially bottlenecked linesof the housefly. Evolution 45: 293-306Milinsld, M. and T.C.M. Bakker. 1990. Female sticklebacks use male coloration in mate choice and henceavoid parasitized males. Nature 344: 330-333Milinski, M. and T.C.M. Bakker. 1992. Costs influence sequential mate choice in sticklebacks, Gasterosteusaculeatus. Proceedings of the Royal Society of London, series B 250: 229-23 3Miller, S.J. and T. Storck. 1984. Temporal spawning distribution of largemouth bass and young-of-yeargrowth, determined from daily otolith rings. Transactions of the American Fisheries Society 113:571-578MØller, A.P. 1991, Sexual ornament size and the cost of fluctuating asymmetry. Proceedings of the RoyalSociety of London, series B: 243: 59-62MØller, A.P. 1994. Sexual selection in the barn swallow (Hirundo rustica). IV. Patterns of fluctuatingasymmetry and selection against asymmetry. Evolution 48: 658-670Nagel, L. 1994. The parallel evolution of reproductive isolation in threespine sticklebacks. M.Sc. thesis.University of British Columbia, Vancouver, CanadaOff, H.A. and J.A. Coyne. 1992. The genetics of adaptation: a reassessment. American Naturalist 140: 725-742Palmer, A.R. and C. Strobeck. 1986. Fluctuating asymmetry: measurement, analysis, patterns. Annual Reviewof Ecology and Systematics 17: 391-421Parsons, P.A. 1990. Fluctuating asymmetry: an epigenetic measure of stress. Biological Reviews 65: 131-145134Parsons, P.A. 1992. Fluctuating asymmetry: a biological monitor of environmental and genomic stress.Heredity 68: 361-364Reimchen, T.E. 1983. Structural relationships between spines and lateral plates in threespine stickleback(Gasterosteus aculeatus). Evolution 37: 93 1-946Reimchen, T.E. 1988. Inefficient predators and prey injuries in a population of giant sticideback. CanadianJournal of Zoology 66: 2036-2044Reist, J.D. 1980. Predation upon pelvic phenotypes of brook stickleback, Culea inconstans, by selectedinvertebrates. Canadian Journal of Zoology 58: 1253-1258Reznick, D.N. and B. Braun. 1987. Fat cycling in the mosquitofish (Gambusia affinis): fat storage as areproductive adaptation. Oecologia (Berlin): 73: 401-413Rice, W.R. 1987. Speciation via habitat specialization: the evolution of reproductive isolation as a correlatedcharacter. Evolutionary Ecology 1: 301-3 14Rice, W.R. and S.D. Gaines. 1994. “Heads I win, tails you lose”: testing directional alternative hypotheses inecological and evolutionary research. Trends in Ecology and Evolution 9: 235-237Rice, W.R. and E.E. Hostert. 1993. Laboratory experiments on speciation: what have we learned in fortyyears? Evolution 47: 1637-1653Rice, W.R. and G.W. Salt. 1990. The evolution of reproductive isolation as a correlated character undersympatric conditions: experimental evidence. Evolution 44: 1140-1152Ridgway, M.S. and J.D. McPhail. 1984. Ecology and evolution of sympatric sticklebacks (Gasterosteus):mate choice and reproductive isolation in the Enos Lake species pair. Canadian Journal of Zoology62: 1813-1818Ridley, M. 1993. Evolution. Blackwell Scientific, Boston, USARosenzwieg, M.L. 1978. Competitive speciation. Biological Journal of the Linnaen Society 10: 275-289Rowland, W.J. 1989. The effects of body size, aggression and nuptial coloration on competition for territoriesin male threespine sticklebacks, Gasterosteus aculeatus. Animal Behaviour 37: 282-289Sanderson, N. 1989. Can gene flow prevent reinforcement? Evolution 43: 1223-1235Sanderson, S.L., J.J. Cech, and M.R. Patterson. 1991. Fluid dynamics in suspension-feeding blackfish.Science 251: 1346-1348Sargent, R.C. and J.B. Gebler. 1980. Effects of nest site concealment on hatching success, reproductivesuccess, and paternal behavior of the threespine stickleback, Gasterosteus aculeatus. BehavioralEcology and Sociobiology 7: 137-142135Schluter, D. 1993. Adaptive radiation in sticklebacks: size, shape and habitat use efficiency. Ecology 74:699-709Schiuter, D. 1994. Experimental evidence that competition promotes divergence in adaptive radiation.Science 266: 798-801Schluter, D. 1995a. Ecological species and speciation. in R. Butlin (ed.) Species and speciation. CambridgeUniversity Press, Cambridge, UKScliluter, D. 1995b. Adaptive radiation in sticklebacks: trade-offs in feeding perfonnance and growth.Ecology 76: 82-90Schiuter, D. and J.D. McPhail. 1992. Ecological character displacement and speciation in sticklebacks.American Naturalist 140: 85-108Schluter, D. and J.D. McPhail. 1993. Character displacement and replicate adaptive radiation. Trends inEcology and Evolution 6: 197-200Schiuter, D. and L.M. Nagel. in press. Parallel speciation by natural selection. American NaturalistSchiuter, D. and T. Price. 1993. Honesty, perception and population divergence in sexually selected traits.Proceedings of the Royal Society of London, series B: 253: 117-122Schultz, E.T., M. Clifton, and R.R. Warner. 1991. Energetic constraints and size-based tactics: the adaptivesignificance of breeding-schedule variation in a marine fish (Embiotocidae: Micrometrus minilnus).American Naturalist 138: 1408-1430Shuter, B.J. and J.R. Post. 1990. Climate, population variability, and the zoogeography of temperate fishes.Transactions of the American Fisheries Society 119: 314-336Simpson, G.G. 1944. Tempo and mode in evolution. Columbia University Press, New York, USASimpson, G.G. 1953. The major features of evolution. Columbia University Press, New York, USASpencer, H.G., B.H. McArdle and D.M. Lambert. 1986. A theoretical investigation of speciation byreinforcement. American Naturalist 128: 241-262S-plus. Statistical Sciences, Inc. 1991. Seattle, USASwamp, H. 1958. Stages in the development of the stickleback Gasterosteus aculeatus (L.). Journal ofEmbryology and Experimental Morphology 6: 373-383Tauber, C.A. and M.J. Tauber. 1989. Sympatric speciation in insects: perception and perspective, in D. Otteand J.A. Endler (eds.), Speciation and its consequences. Sinauer Associates, Sunderland, USATaylor, E.B. and P. Bentzen. 1993. Evidence for multiple origins and sympatric divergence of trophicecotypes of smelt (Osmerus) in northeastern North America. Evolution 47: 813-832136Templeton, A.R. 1989. The meaning of species and speciation: a genetic perspective, in D. Otte and J.A.Endler (eds.), Speciation and its consequences. Sinauer Associates, Sunderland, USAThornhill, R. 1992. Fluctuating asymmetry and the mating system of the Japanese scorpionfly, Panopajaponica. Animal Behaviour 44: 867-879Tuinstra, E.J. G. de Jong and W. Scharloo. 1990. Lack of response to family selection for directionalasymmetry in Drosophila melanogaster: Left and right sides are not distinguished in development.Proceedings of the Royal Society of London, series B 241: 146-152van den Assem, J. 1967. Territory in the three-spined stickleback Gasterosteus aculeatus (L.). BehaviourSupplement 16: 1-164Van Valen, L. 1976. Ecological species, multispecies, and oaks. Taxon 25: 233-239Wade, M.J. and S. Kalisz. 1990. The causes of natural selection. Evolution 44: 1947-1955White, M.J.D. 1978. Modes of speciation. W.H. Freeman, San Francisco, USAWhoriskey, F.G and G.J. FitzGerald. 1994. Ecology of the threespine stickleback on the breeding grounds. inM.A. Bell and S.A. Foster (eds.), The evolutionary biology of the threespine stickleback. OxfordUniversity Press, Oxford, UK.Wiegmann, D.D., J.R. Baylis and M.H. Hoff. 1992. Sexual selection and fitness variation in a population ofsmallmouth bass, Micropterus dolomieui (Pisces: Centrarchidae). Evolution 46: 1740-1753Wilson, D.S. 1989. The diversification of single gene pools by density- and frequency dependent selection. inD. Otte and J.A. Endler (eds.) Speciation and its consequences. Sinauer Associates, Sunderland, USAWood, C.C. and C.J. Foote. 1990. Genetic differences in the early development and growth of sympatricsockeye salmon and kokanee (Oncorhyncus nerka) and their hybrids. Canadian Journal of Fisheriesand Aquatic Sciences 47: 2250-2260Wooton, R.J. 1973. The effect of size of food ration on egg production in the female three-spined stickleback,Gasterosteus aculeatus L. Journal of Fish Biology 5: 89-96Wright, S. 1932. The roles of mutation, inbreeding, crossbreeding, and selection in evolution. Proceedings VIInternational Congress of Genetics 1: 356-366Wright, S. 1968. Evolution and the genetics of populations. Volume 1. Genetic and biometric foundations.University of Chicago Press, Chicago, USAZeng, Z.-B., D. Houle and C.C. Cockerham. 1990. How informative is Wright’s estimator of the number ofgenes affecting a quantitative character? Genetics 126: 235-247Zuk, M., K. Johnson, R. Thornhill. and J.D. Ligon. 1990. Mechanisms of female choice in red jungle fowl.Evolution 44: 477-485137Zyuganov, V.V. 1989. The discovery of reproductively isolated parapatric forms of the ninespine stickleback,Pungitius pungitius, with and without a pelvic girdle. Soviet Journal of Fish Biology 1989: 134-142


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items