THE PARALLEL EVOLUTION OF REPRODUCTIVE ISOLATION IN THREESPINESTICKLEBACKSbyLAURA MARIA NAGELB.Sc., The University of Toronto, 1989A THESIS SUBMITTED IN PARTIAL FULFILMENT OFTHE REQUIREMENTS FOR THEDEGREE OFMASTER OF SCIENCEinTHE FACULTY OF GRADUATE STUDIES(Department of Zoology)We accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIAApril 1994©Laura Maria Nage1 1994In 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.Department of ‘/cThe University of British ColumLIVancouver, CanadaDate 7 O/ 1DE-6 (2188)11ABSTRACTReproductive isolation within and between sympatric populations of threespinestickleback was investigated to find evidence for parallel speciation, which is the independentevolution of the same reproductive isolating mechanisms. One of the requirements of parallelspeciation is reproductive isolation between descendant populations of sympatric species.Mate choice tests between benthic and limnetic sticklebacks in two lakes were thereforeconducted to examine reproductive isolation between them. Strong reproductive isolation wasfound. The probability of hybridization between limnetic females and benthic males wasaffected by the size of both the male and the female. I suggest that reproductive isolationbased on body size may have evolved as a by-product of natural selection on body size.Mate choice tests within benthic and limnetic forms were conducted to determine ifthere was reproductive isolation among benthic and limnetic forms from different lakes. Ihypothesized that because each form evolved under similar selective regimes, they would notbe reproductively isolated. The results supported this hypothesis, although there was a trendin limnetics to prefer mates from their own population. There was no reproductive isolationbetween benthics from different lakes. Similar reproductive isolating mechanisms thereforearose independently in several lineages, probably as a by-product of natural selection.TABLE OF CONTENTS111ABSTRACT.TABLE OF CONTENTSLIST OF TABLESLIST OF FIGURESACKNOWLEDGEMENTSDEDICATIONINTRODUCTIONChapter 1: REPRODUCTIVE ISOLATION IN BENTHIC ANDSTICKLEBACKSIntroductionParallel evolution and speciation in sticklebacksStudy populationsMethodsCollection and maintenance of fishMate choice testsTest procedureAnalysisResultsDiscussionLIMNETIC11111ivVvivii1THREESPINE3356991011121314Chapter 2: A TEST OF PARALLEL SPECIATIONIntroductionStudy populationsMethodsMate choice tests1. No-choice tests (limnetics and2. Choice tests (limnetics only)AnalysisResultsDiscussion .CONCLUSIONSLITERATURE CITEDAPPENDIX IAPPENDIX IIbenthics)2727293132323334343649515659ivLIST OF TABLESTable one. Results of chi-square tests for assortative mating between benthics and limneticsfrom Paxton Lake 19Table two. Results of chi-square tests for assortative mating between benthics and limnetics fromPriest Lake 19Table three. Results of univariate Maximum-Likelihood estimates carried out individually on theeffect of date of test, female and male standard length and male colour score on the probabilityof interspecific hybridization in limnetic female and benthic male pairs (both lakescombined) 20Table four. Results of univariate Maximum-Likelihood estimates carried out individually on theeffect of date of test and female and male standard length on the probability of interspecifichybridization in benthic female and limnetic male pairs (both lakes combined) 20Table five. Results of choice tests with limnetic females from Paxton and Priest Lakes given achoice between spawning with a limnetic male from Priest or Paxton Lake 41Table six. Results of within-form tests with limnetics from two lakes 42Table seven. Results of within-form tests with benthics from three lakes 43Table eight. Results of Maximum-Likelihood estimates on the effect of date of test and femaleand male standard length on the probability of spawning in limnetic pairs (both lakescombined) 44Table nine. Results of univariate Maximum-Likelihood estimates carried out individually on theeffect of date of test and female and male standard length on the probability of interspecifichybridization in benthic pairs (males from three lakes paired with females from twolake) 44VLIST OF FIGURESFigure 1. Results of no-choice spawning tests for assortative mating with limnetic females22Figure 2. Results of no-choice spawning tests for assortative mating with benthic females23Figure 3. Estimates of nuptial colour score for males from four populations used in no-choicetests 24Figure 4. Standard lengths of limnetic females and benthic males used in no-choice tests25Figure 5. Standard lengths of benthic females and limnetic males used in no choice tests26Figure 6. Schematic of parallel evolution in sympatric threespine sticklebacks 46Figure 7. No-choice spawning tests with limnetic females from each lake tested with limneticmales from both lakes 47Figure 8. No-choice spawning tests with benthic females from each lake tested with benthicmales from three lakes 48viACKNOWLEDGEMENTSI thank my supervisor, D. Schiuter, for his boundless enthusiasm throughout this project,and for showing me how not to season a curry. J.D. McPhail, among other helpful advice,suggested that I use wild-caught females, thus enabling me to escape from the city each summer.The encouragement of J.N.M. Smith in the early stages of the work is appreciated. L. Gass wasan excellent replacement for him near the end.0. Jesperson provided logistical help in the barn, and M. Linden helped me to spend myfree time on Texada in innovative activities. I thank S. Heard for help with statistical analysisand his invaluable comments on this manuscript.My parents have always been an inspiration for me and their continual support and theoccasional ticket to Whitehorse are greatly appreciated.I thank my fellow biologists and friends, the alpine birds, for the sarcasm and witticismsthat brightened many a rainy day.T. Day has been patiently explaining my project to me from the beginning. I thank himfor all his help with it and for helping me to stay sane for the last two years.Finally, I want to thank all the sticklebacks who contributed to this study and withoutwhom it would not have been possible.vi’This thesis is dedicated to Troy Day, because his interest is real.INTRODUCTIONMost experimental studies of the process of speciation have been conducted withlaboratory populations of Drosophila. As a result, we know little about how new species formin nature. Natural selection is usually invoked as the process which causes divergence betweenincipient species, but there is little evidence of this actually occurring in natural populations(Endler 1989).Natural selection can cause speciation in an adaptive or non-adaptive way. Reproductiveisolation between groups can evolve as a by-product of natural selection acting on other traits.The mechanism may be evolution as a result of genetic correlation (indirect) or by directselection on mate preferences as a result of changes in other traits caused by natural selection.The repeated independent evolution of the same reproductive isolating mechanisms has beentermed parallel speciation (Schluter and Nagel, ms). Three criteria are required to demonstrateparallel speciation:1. Each lineage sharing a trait must have an independent origin.2. Reproductive isolation must evolve between the ancestral population and the descendantpopulations. In cases where descendant populations are sympatric, reproductive isolationmust evolve between these populations.3. Descendant populations which have evolved under similar selective regimes in differentlocalities must not be reproductively isolated from each other.To test for parallel speciation in a natural population, I examined assortative mating in sympatricthreespine sticklebacks from three lakes. These populations of threespine sticklebacks arevaluable organisms with which to test speciation theory because they are very young (<15 000yrs.) and because they exhibit extensive parallel evolution. Each lake contains a benthic and alimnetic stickleback which probably evolved independently from a common marine ancestor2(McPhail 1993). The speciation process may therefore have been replicated many times.A complete phylogeny of these populations is not yet available, but evidence to date isconsistent with the hypothesis that each population has had an independent origin.Evidence supporting the first criterion of parallel speciation is thus inconclusive.The second criterion is addressed in chapter one. I describe mate choice tests betweenbenthics and limnetics in two lakes to examine reproductive isolation between them. If there ispositive assortative mating, and if it is based on a trait thought to have been under direct orindirect natural selection (like body size) then I suggest that reproductive isolation may haveevolved as a by-product of natural selection on body size. Since there is already some evidenceof assortative mating between the benthics and limnetics in one lake (Ridgway and McPhail,1984), these tests also allowed me to assess the accuracy of my experimental design so that Icould be confident in the results of the mate choice tests described in chapter two.In chapter two I address the third criterion of parallel speciation. I describe tests ofreproductive isolation among limnetics and benthics from different lakes. Limnetics and benthicshave probably experienced the same selective regimes in each lake. If there is no reproductiveisolation among the forms from different lakes, then the same selective regimes may have causedthe same reproductive isolating mechanisms to evolve. This may have occurred as an adaptiveor non-adaptive by-product of natural selection. This would suggest that natural selection cancause speciation to unfold in a remarkably similar fashion.3chapter 1. REPRODUCTIVE ISOLATION IN BENTHIC AND LIMNETICSTICKLEBACKSINTRODUCTIONModels of speciation fall into two groups: those in which gene flow between populationsis present and those in which populations are entirely allopatric (Mayr 1963). Models with geneflow include reinforcement (Dobzhansky 1937), sympatric and parapatric speciation (reviewedin Maynard Smith 1966; Endler 1977; Felsenstein 1981) and bottleneck speciation (Mayr 1970;Carson and Templeton 1984). In a recent review, Rice and Hostert (1994) summarize 40 yearsof laboratory experiments which have tested these major speciation models. Despite intensiveeffort, few of these studies have actually achieved speciation (defined by Mayr (1963) asreproductive isolation between groups of natural populations). The authors concluded that thereis evidence from laboratory experiments for most of these models, except for bottleneckspeciation and reinforcement.Only in the reinforcement model of speciation is natural selection thought to act directlyon mate preferences. In most other models, the evolution of prezygotic reproductive isolationis thought to occur as a by-product of natural selection or genetic drift. Although genetic driftmay be an important mechanism promoting speciation, the role of natural selection will befocused on here.Mate preferences evolve as a by-product of natural selection in two ways. In the nonadaptive scenario, mate preferences evolve via a genetic mechanism. This may be eitherpleiotropy (one gene having several phenotypic effects) or linkage disequilibrium between allelesaffecting the character(s) under selection and alleles affecting positive assortative mating (Slatkin1982; Rice 1984, 1987; Rice and Hostert 1994). This is distinct from an adaptive scenario inwhich natural selection acts directly on some trait, resulting in secondary selection on mate4preferences to accommodate the new properties of available mates. Mate preferences will thenchange along with changes in other traits as a ‘chain of selection’ (Schluter and Nagel, ms).Although it may appear that natural selection is acting directly on mate preferences, the selectionis actually a by-product of selection on other trait(s). I therefore refer to this scenario as a byproduct mechanism. Sexual selection may generate or amplify premating isolation which hasoccurred as a by-product if there is heritable variation for female choice and for male traits (e.g.Endler 1989; Liou and Price 1994).By-product speciation has been shown experimentally to be a feasible mechanism instudies with flies (Sans et al. 1974; Hurd and Eisenberg 1975; Kilias et al. 1980; Dodd 1989;Rice 1985; Rice and Salt 1988, 1990). In some of these tests, the effects of genetic drift areindistinguishable from those of selection, but two studies (Kilias et al. 1980; Dodd 1989)explicitly demonstrated that divergent selection is the factor which promoted the evolution ofreproductive isolation. However, it is unclear whether this occurred in an adaptive or nonadaptive way. Rice and Hostert (1994) favour the interpretation that a genetic mechanism isresponsible and suggest that pleiotropy is a more tenable mechanism than linkage disequilibrium.This is because linkage disequilibrium requires two steps for the evolution of reproductiveisolation (see Felsenstein 1981), while pleiotropy requires only one: disruptive selection on a trait(such as body size) produces positive assortative mating as a by-product. Pleiotropy also appearsto be widespread in natural populations (Price and Langen 1992).The fact that the by-product mechanism is the only mechanism of speciation with strongsupport from laboratory tests suggests that it is very important in nature as well. Some labstudies which have tested the by-product mechanism have resulted in the evolution of only weakreproductive isolation. It is important to note that these studies involved selection on only asingle factor (e.g. bristle number, temperature tolerance). Those which resulted in stronger5reproductive isolation involved selective regimes in which several traits were selected forsimultaneously. This is a more likely scenario for natural selection in the wild. Presumably,selecting for several traits increases the probability that one of them will have an association withpositive assortative mating.Studying speciation is inherently difficult in natural populations because of the historicalnature of the process. This is why most studies of speciation have been conducted withlaboratory populations of Drosophila. In this chapter, I will describe an indirect test of thepossible importance of the by-product mechanism in the evolution of reproductive isolation. Acrucial difference in my study is that the populations concerned are natural populations that havealready achieved the level of biological species: my study is therefore one of the first to test thismechanism in a wild population. I examined reproductive isolation in populations of sympatricthreespine sticklebacks (Gasterosteus aculeatus) from two lakes test one of the criterion ofparallel speciation: reproductive isolation between sympatric descendant species (seeintroduction). I was also interested in determining if reproductive isolation was based on bodysize, a trait thought to be under direct or indirect natural selection in these populations. If thereis evidence for reproductive isolation based on body size, then reproductive isolation may havearisen as a by-product of natural selection.PARALLEL EVOLUTION AND SPECIATION IN STICKLEBACKSOne of the most important reasons for conducting studies of speciation in laboratories isthat the experiments can be replicated, but there are also replicated “experiments11 in nature.Parallel evolution occurs when similar selective regimes act on closely related populations andin response, the same trait evolves independently in two or more lineages (Futuyma 1986).Parallel evolution has occurred in populations of freshwater threespine sticklebacksthroughout their holarctic range. When marine sticklebacks establish freshwater populations,6there is a consistent pattern of shifts in morphology to accommodate the different requirementsof a freshwater environment. For example, the large, slender anadromous form invariably givesrise to a small, stocky stream resident form (McPhail 1993). Since it is very unlikely that geneticdrift alone could cause independent, replicate shifts in the same direction, natural selection isstrongly implied (Endler 1986; Clarke 1975). If these shifts are independent events caused bynatural selection, then they are essentially a replicated “experiment” which can be used to testevolutionary theory.In southwestern British Columbia, most morphological variation in threespine sticklebackpopulations is associated with local selective regimes (eg. Moodie 1972; Lavin and McPhail1985, 1986; Schluter and McPhail 1992). Body size is highly variable among sticklebackpopulations (Bell 1976; Schiuter and McPhail 1992), and is thought to have been under director indirect natural selection. Body size has also been implicated as a factor maintainingreproductive isolation between some sympatric populations of sticklebacks (Borland 1986;McPhail 1993).Study populationsPopulations of sympatric benthic and limnetic threespine stickleback are found in arestricted geographical area in southwestern British Columbia, Canada. These populations arelocated in four watersheds on three islands in the Straight of Georgia. Each lake contains aspecies pair. One species is a small, limnetic stickleback and the other is a much larger, benthicstickleback. The limnetic stickleback feeds largely on plankton while the benthic form forageson benthic invertebrates.McPhail (1993) suggested that the species pairs resulted from a double invasion of marinesticklebacks after the last (Fraser) glaciation during the Pleistocene. Marine sticklebacks invadedmany coastal watersheds as the glaciers receded approximately 15 000 years ago. These fish7evolved to exploit the lake habitat, including benthos. Approximately 2000 years later, the sealevel rose or the land was depressed a second time, and a second colonization of marinesticklebacks took place. Ecological character displacement occurred and these secondary invaderscontinued to exploit planktonic resources while the first colonizers become more benthic in theirresource exploitation (Schluter and McPhail 1992). There is geological evidence which supportsthis double-invasion scenario (Mathews et al. 1970; Clague 1981; Clague et al. 1982). Thebenthic and limnetic pairs in each lake represent distinct gene pools: for example, in the EnosLake population, one locus is fixed in the benthic but limnetics show an 18% frequency of avariant allele (McPhail 1984). Hybrids of the two species appear to be very rare (McPhail 1984,1993). There is some evidence for postmating isolating mechanisms. Although hybrids arefertile and viable, their intermediate phenotypes put them at a selective disadvantage relative tothe parental forms in certain environments (Schluter 1993, 1994).A complete phylogeny of interpopulation relationships is not yet available, but molecularanalyses to date are consistent with the hypothesis that the species pair in each lake aroseindependently from the marine ancestor. For example, fish from Paxton Lake have one mtDNAgenotype not present in the Priest or Enos Lake populations (E.B. Taylor, pers. comm.).Benthic sticklebacks are larger than limnetics. Sexual dimorphism is similar in all speciespairs: breeding benthic females are larger than benthic males, and limnetic females are smallerthan limnetic males. The reason for the differences in size within and between species is notknown. Body size may not have been under direct natural selection, but may have beencorrelated with traits which were. However, size is one of the most important determinants offeeding efficiency. For example, Schiuter (1993) found that feeding efficiency of benthics onplankton decreases with increasing size. Gape width (which is correlated with body size) maybe important for foraging: a small mouth may enhance capture success of plankton, particularly8where suction is used (eg. Werner 1977). Limnetics may have been selected for small body sizefor efficiency in foraging on plankton. The large gape width of the benthic may have evolvedfor foraging on larger benthic prey items. Natural selection may therefore have favoured anincrease in size for the benthics and a decrease in body size for limnetics as a result ofspecialization on different prey types.Alternatively, there is some evidence that gape-limited predators can exert considerableselection pressure on sticklebacks (McPhail 1969; Reimchen 1991). All of the lakes with speciespairs have endemic populations of cutthroat trout (Onchorhnychus clarkii) (J.D. McPhail, pers.comm.) so intense predation pressure may have selected for larger body size in the first coloniststo the lakes (benthics). Benthics are usually solitary and cryptic while limnetics school in largegroups.I conducted mate choice tests using benthics and limnetics from two lakes. If body sizeis the basis of assortative mating, limnetics should choose smaller mates and benthics shouldprefer larger mates. Any breakdown in positive assortative mating should be correlated withbody size (e.g. hybridizations with limnetic females should only occur when benthic males aresmall, and therefore within the normal size range of limnetic males). If body size is shown tobe the basis of reproductive isolation, and if we accept that body size has been under naturalselection, then this is indirect evidence that reproductive isolation may have evolved as a byproduct of natural selection in these populations.9METHODSCollection and maintenance of fishSticklebacks were collected from two lakes (Paxton and Priest) on TexadaIsland, British Columbia, Canada (49° 40’N, 124° 30’W). Females were collected shortly beforeneeded with baited minnow traps set for 12-14 hours. Fish captured in the traps were examinedfor reproductive condition, and gravid animals were immediately brought into the lab and housedin communal tanks of 102 L or 180 L. Females were fed a mixture of frozen bloodworms andlive Artemia sp. On the second day after capture females judged ready to spawn by theirdistended abdomens were used in a single mate choice test.Males were collected with baited minnow traps on February 12 and March 7 in 1992 andon February 11 in 1993. Benthics and limnetics were held separately in 102 L tanks in anenvironment chamber in Vancouver. The tanks were lit by rows of “cool white” fluorescentlamps. The photoperiod was gradually increased from 1OL:14D to 16L:8D over a two monthperiod in order to bring the males into breeding condition. The temperature during this periodwas increased from 7 to 10 degrees Celsius. All fish were fed to satiation once daily with frozenbrine shrimp (Artemia sp.) and bloodworms (chironomid larvae).All males had developed some degree of nuptial colouration and were ready to breed byearly March. Benthic males from Priest Lake differed considerably from other populations inacclimating to captivity. They were very cryptic and easily frightened, even after months incaptivity. They also never fully developed the nuptial colouration typical of wild males from thislake (pers. ohs.). The reason for this is not known. The Priest Lake populations probablyexperience more predation since trout are less abundant in Paxton Lake. This may explain thenervous behaviour of Priest Lake benthic males in the aquaria, which are relatively bright andin which there is little cover. Limnetics and female benthics from this population did not seem10to be affected to the same extent as the benthic males were.In April, males were returned to Texada Island and housed under similar conditions. Nomortality or aberrant behaviour was associated with this move. The light regime was maintainedat 1 6L: 8D, and water temperatures fluctuated between 15 and 18 degrees Celsius. The holdingtanks were lit with fluorescent lights supplemented with 60W incandescent bulbs. Males wereremoved haphazardly from the communal tanks for use in mate choice tests. Each male wasplaced in a 55 L aquarium or in one half of a divided 102 L aquarium. The aquaria werecovered on three sides by heavy brown paper so that males were visually isolated. Each row oftanks was illuminated by two ‘cool white” fluorescent lights. Each aquarium contained a bubblerfilter and a plastic tray filled with sand to serve as a nesting substrate. Males were discouragedfrom building nests outside of the sand tray by covering the bottom of the aquarium with gravel.Each male was provided with nesting materials obtained from the lakes and allowed five daysin which to build a nest. Males that did not build a nest within five days were replaced (about8% of males).Mate choice testsInitially, three different protocols were used to assess mate choice. Orientation tests inwhich females chose between two males in separate tanks were tried first. In these tests, afemale in a small tank could see each of two males, but the males could not see each other.Although this worked reasonably well with limnetic females, benthic females would not respondto the visual stimulus of courting males. These tests were therefore not included in this analysis.Since physical contact appeared to be important to benthic females, I attempted a secondprotocol in which females were allowed to spawn with one of two nesting males who wereconfined to either side of a 180 L aquaria. No combinations of benthic-limnetic male pairs weresuccessful: the benthic always destroyed the limnetic male’s nest and took over his territory.11The protocol finally used was one in which a single gravid female was placed in anaquarium with a reproductively active male, and whether or not the pair spawned in 30 minuteswas recorded. Although this was essentially a no-choice test, it appeared to be a reasonableestimate of a pair’s willingness to spawn. Spawning did not appear to be coerced: only 33% ofthe pairs tested spawned. Preliminary results showed that pairs which did not spawn within 30minutes would not do so even if left for much longer. In addition, orientation tests withlimnetics corroborated results from this protocol.Test ProcedureA gravid female was removed from the communal tank and placed in an inverted 500 mlclear plastic container in a corner of a male’s tank. Males usually began to court femalesconfined to the jar within two minutes of its placement in the tank. On rare occasions, a maledid not approach the jar within three minutes, in which case the test was cancelled. Within about30 seconds of the male’s approach, the female in the jar swam up from the bottom and beganto respond to the male’s courtship by swimming at the sides of the jar. After three minutes thejar was gently raised above the female using a piece of clear monofilament line attached to theit’s bottom. The line was secured such that the jar was positioned just below the surface of thewater to minimize disturbance and to serve as a retreat for females. Courtship in sticklebackscan be very aggressive, and in natural situations females who are unwilling to court with a maleare able to leave the area. Having a safe and familiar area for retreat appeared to be especiallyimportant in tests involving limnetic females and benthic males because benthics were often veryaggressive toward the smaller limnetics.A test began when the male first directed courtship activity to a freed female (usuallywithin one minute). I made observations from approximately two meters from the front of thetank with an event recorder. The test animals did not appear to respond to a motionless observer12at this distance. A qualitative colour score was assigned to each male in order to determine ifbright male colour was a factor in female preference for spawning partners. The score rangedfrom one to five, with five being the highest score for overall brightness and intensity ofcolouration. Since males usually become more colourful as the time engaged in courtshipincreases, this score was recorded only once: when the male directed the first courtship behaviourat a female.After each trial, females who did not spawn were examined to see if their receptivity priorto the test had been improperly assigned. When gentle pressure is placed on the abdomen of afemale with mature eggs, the eggs slide easily into the lower oviduct and can be seen throughthe body cavity. If a female had been misjudged, her trial was excluded from the analysis. Thisoccurred only twice in over 200 tests. Males were sometimes tested a second time with a femalefrom another population (18% of all males used in this analysis). When this was done, maleswere always left for 24 hours after a test.After a test, females were anaesthetized with MS222 and standard length was measured.Males were anaesthetized, measured, weighed, and preserved in 10% formalin. Standard lengthsof males and females were measured on all but 25 pairs. The range and median standard lengthsof all fish used in no-choice tests are in appendix II.AnalysisI used Yates’ corrected chi-square tests (SYSTAT version 5.01, SYSTAT Inc., 1989) totest for assortative mating between species in each lake. I used logistic regression to examineeffects of male and female standard length on spawning probability in interspecific pairs (SASVersion 6.03, CATMOD procedure; SAS Institute Inc., 1988). All tests were two-tailed exceptfor the effect of male colour on the probability of spawning.13RESULTSStrong positive assortative mating was evident in all but one population (figures 1 and 2).In tests with Paxton Lake fish, limnetics strongly preferred to spawn with limnetics and benthicspreferred to spawn with benthics (table 1). Results from tests with Priest Lake fish (table 2) werecomplicated by problems with benthic males. Female benthics rejected benthic males nearly asoften as they did linmetic males. This resulted in non-significant assortative mating. This resultwas probably due to the unusual behaviour and dull colour (figure 3) of Priest Lake males.Interspecific pairings resulted in spawning in eight of 87 tests conducted. Male benthicssometimes responded very aggressively towards limnetic females, especially when the femaleswere very small. Male limnetics also sometimes would not court benthic females, and wereespecially aggressive towards very large females.When the data on the effect of variables of interest on interspecific pairings in each lakewere analyzed separately, the magnitude and sign of all variables was similar, so the data werecombined for further analysis. Interspecific spawnings involving limnetic females (6 of 49 tests)occurred only when the benthic males were smaller (mean=5 1mm) than most benthic males usedin the tests (mean=58.6mm). The limnetic females they spawned with were among the largestfemales tested (mean=49. 1mm) (figure 4). Although pairing of males and females was meantto be random, a problem arose in analysis because limnetic females and benthic males wereinadvertently size-matched. Small females tended to be paired with large males, resulting in anegative correlation (-0.60) between male and female standard length. Because of the correlationbetween male and female length, it was not possible to determine which of these variables is thecause of the variation in the probability of hybridization (or whether there is a significantinteraction between them). When analyzed separately, the effects of female length, male length,and male colour on the probability of hybridization are significant (table 3). The effect of trial14date was not significant. The probability of hybridization was therefore probably not aconsequence of males or females becoming less choosy as the breeding season progressed.Only two hybridizations with benthic females and limnetic males occurred. The femaleswho spawned were both small (mean=52.5mrn), and the limnetic males they spawned with werelarge (mean=5 1.7mm) compared with the mean length of the entire sample (49.5mm, seeAppendix II) (figure 5), but because of the small number of spawning events, there was nosignificant effect of male or female size (table 2). The effect of male colour score could not beevaluated because one of the colour scores was missing. There was no significant effect of thedate of the test, although it may be biologically significant that both hybridizations occurred verylate in the late in the breeding season (June 16 and 17, 1993). Limnetic males and/or benthicfemales may have become less choosy as the breeding season progressed.DISCUSSIONBenthics and limnetics within lakes are reproductively isolated. The second criterionof parallel speciation (reproductive isolation between descendant populations) is thereforefulfilled. It is not surprising that there is strong assortative mating in these populations. Hybridsbetween the two species appear to be very rare in nature (McPhail 1993), and Ridgway andMcPhail (1986) have already documented positive assortative mating in a similar species pair inEnos Lake. Reproductive isolation between sympatric species has also been demonstrated forother populations of threespine stickleback, although they too have not been formally describedby systematists as distinct species (eg. Hagen 1967; Moodie 1972; Hay and McPhail 1975;Borland 1986; Blouw and Hagen 1990). However, most of these studies do not identify whichtraits are important in mate choice, and hence may be the basis of reproductive isolation.The case for body size being the basis of reproductive isolation between species is notterribly strong. Because of the limited number of spawnings with benthic females and limnetic15males, little can be said about the causes of hybridization. The effect of body size in maintainingreproductive isolation between limnetic females and benthic males is more apparent. Limneticfemales spawned with benthic males who were smaller than most benthic males. The meanstandard length of these males was very close to the mean length of limnetic males used in thesetests. This suggests that limnetic females use body size to evaluate potential mates: large maleswere rejected by them, but they spawned with small males. The same pattern was apparent ifmale choice is considered to be a determining factor in mate preference: benthic males preferredto spawn with large females, even if these females were of the wrong species.In chapter two, I examine the effect of standard length on the probability of spawningwithin each form. In limnetics, there is a significant effect of male standard length on theprobability of spawning but not of female length. This suggests that limnetic females prefer tospawn with larger males, but not with male who are much larger than the average limnetic male.It appears that positive assortative mating between limnetics and benthics may be largelydetermined by body size. This may have arisen as a by-product of natural selection on body size.Body size may be one of the most important visual cues to mate recognition in the species pairsbecause it is very obvious and easily assessed. However, the statistical significance of this traiton the probability of hybridization has not been strongly demonstrated. I have also not testedfor the relative importance of other factors and so cannot discount them. Female mate choicewithin populations of threespine sticklebacks has been shown to be influenced by malecolouration (eg. McLennan and McPhail 1989; Milinski and Bakker 1990), territory size(Goldschmidt and Bakker 1990), territory quality (Sargent 1982), and presence of eggs in themales’ nest (eg. Jamieson and Colgan 1989). Male courtship behaviour may also be an importantfactor (eg. McPhail and Hay 1983; Ridgway and McPhail 1986; Jamieson and Colgan 1989).The species pairs differ in many other morphological traits as well as behaviourally and16ecologically. These and other factors probably also contribute to reproductive isolation.There are also alternative hypothesis for the evolution of reproductive isolation in thesepopulations that cannot yet be ruled out.First, body size may have diverged as a result of reproductive character displacement.This hypothesis requires that complete postmating reproductive isolation evolved in allopatry.When the two species contacted each other at the time of the second invasion, an exaggerationof differences in traits used in premating isolation may have resulted. There have been fewdemonstrations of reproductive character displacement (but see Littlejohn (1965) and Bell(1976)). One of the requirements to demonstrate reproductive character displacement is that atrait important in mate selection is more divergent in sympatry than in allopatry (Brown andWilson 1956). Obviously, this would be not be possible to test directly because limnetics andbenthics do not occur in allopatry. However, this hypothesis could be supported if it were shownthat the sympatric populations of benthics and limnetics have a stronger preference for mates witha certain body size than the closely related solitary or marine populations.Second, reinforcement (Dobzhansky 1937) may have occurred. In reinforcement, naturalselection acts directly on mate preferences to avoid the production of less fit hybrids. Thishypothesis suggests that both premating and postmating reproductive isolation between thespecies pairs was incomplete at the time of secondary contact. Premating reproductive isolationthen evolved to prevent the production of less fit hybrids. Although there is no experimentalevidence to support speciation by reinforcement and there are strong theoretical objections to themodel (for a review, see Rice and Hostert, 1994) it cannot be entirely discounted here.Both of these alternative scenarios would require hybrids to have zero or very low fitness,which is unlikely in this case. Genetic drift is probably not the cause of reproductive isolationbecause it is a random process, and so would not likely cause parallel shifts morphology or17behaviour in the same direction in each population. The most plausible scenario for speciationin these species pairs seems to be the evolution of reproductive isolation as a by-product ofnatural selection acting directly or indirectly on traits conferring reproductive isolation. Underthe ‘chain of selection’ scenario (Schiuter and Nagel ms), the modification of phenotypes (e.g.body size) due to natural selection may have placed secondary selection directly on matepreferences to accommodate the new phenotypes of available mates. Another way in whichselection could act directly on mate preferences as a by-product of natural selection is that newsensory modes to detect mates may be favoured in new environments (Endler 1992; Marchetti1993; Schluter and Price 1993). It is not possible in this example to distinguish between the ideathat selection acted in an adaptive way on mate preferences or that mate preferences weregenetically correlated with other traits under selection.The potential for invoking the by-product mechanism is present in many species in whicha trait which has been under natural selection is also important in reproductive isolation. Otherexamples of assortative mating between species that is based on morphological traits which havebeen under disruptive selection include finches (Ratcliffe and Grant 1983), anadromous andfreshwater-resident threespine sticidebacks (Borland 1986), and salmon (Foote and Larkin 1988).By-product speciation may have had important implications for adaptive radiation inthreespine sticklebacks and in general. Because divergence in sticklebacks is driven by naturalselection (McPhail 1993), the possibility for the evolution of reproductive isolation via the byproduct mechanism seems very high. This suggests that speciose lineages may be those who,for ecological reasons, adapt quickly to different environments. Marine threespine sticklebackswhich invade freshwater respond quickly to the local selective regime, and, if the by-productmechanism occurs, reproductive isolation evolves as well. This could be especially importantfor our understanding of adaptive radiations in general if certain lineages have a greater incidence18of correlated responses for genes for assortative mating and for traits under selection. By-productspeciation is an important alternative to some of the more untenable speciation models whichhave dominated much of the discussion on the origins of reproductive isolation, and for whichthere is minimal experimental support.19Table one. Results of chi-square contingency tests for assortative mating between benthics andlimnetics from Paxton Lake.Pair N #Spawning Chi-square p-valuelimnetic X 12 7limnetic o’4.38 0.04limnetic X 23 4benthic d’benthic X 28 17benthic a”13.85 0.0002benthic X 21 1limnetic a”Table two. Results of chi-square contingency tests for assortative mating between benthics andlimnetics from Priest Lake.Pair N #Spawning Chi-square p-valuelimnetic X 23 11limnetic a” 8.13 0.004limnetic X 26 2benthic a’benthic X 26 4benthic a” 0.21 0.64benthic X 17 1limnetic a”20Table three. Results of univariate Maximum-Likelihood estimates carried out individually bylogistic regression on the effect of date of test, female and male standard length and male colourscore on the probability of interspecific hybridization in limnetic female and benthic male pairs(both lakes combined). Asterisks indicate a significant effect on the probability of hybridizationusing chi-square analysis. (N=42).Variables Estimate (SE) df Likelihood ratiodfdate 0.066 (0.04) 1 24female length* 0.707 (0.29) 1 34male length* -0.733 (0.31) 1 36male colour* 1.000 (0.53) 1 6*p < 0.05Table four. Results of univariate Maximum-Likelihood estimates carried out individually bylogistic regression on the effect of date of test and female and male standard length on theprobability of interspecific hybridization in benthic female and limnetic male pairs (both lakescombined). All variables are not significant using chi-square analysis. (Nt34).Variables Estimate (SE) df Likelihood ratiodfdate 0.349 (0.36) 1 10female length -0.301 (0.21) 1 28male length 0.275 (0.31) 1 2321Figure LegendsFigure 1. Results of no-choice spawning tests for assortative mating with limnetic females.Points are mean values of the proportion of pairs which spawned for all pairs tested + one-waySE bars. a) Priest Lake (N=49) b) Paxton Lake (N=35)Figure 2. Results of no-choice spawning tests for assortative mating with benthic females. Pointsare mean values of the proportion of pairs which spawned for all pairs tested + one-way SE bars.a) Priest Lake (N=43) b) Paxton Lake (N=49)Figure 3. Estimates of nuptial colour score for males from four populations used in no-choicetests. Note the low score for Priest Lake benthic males. Points are mean values ± SE (1 =dullest,5=brightest). Paxton benthic: mean=3.5, N64, SE= 0.13; Paxton limnetic: mean=4.1, N=39,SE=0.15; Priest benthic: mean=2.0, N=64, SE=0.09; Priest limnetic: mean=3.0, N=46, SE= 0.22).Figure 4. Standard lengths of limnetic females and benthic males used in no-choice tests.Circles are pairs from Paxton Lake. Triangles are pairs from Priest Lake. A filled symbolindicates that the pair spawned. In the six pairs that spawned, the mean male standard lengthwas 51.0mm (SD=0.998) and the mean female standard length was 49.1 (SD=1.287).(N=42).Figure 5. Standard lengths of benthic females and limnetic males used in no-choice tests.Circles are pairs from Paxton Lake. Triangles are pairs from Priest Lake. A filled symbolindicates that the pair spawned. In the two pairs that spawned, the mean male standard lengthwas 51.7mm (SD=1.56) and the mean female standard length was 52.5 (SD=2.40).(N=42).______________________________________________________1.0-oci)ci)C0.u,0.80.8C)0.6•?i3EE•0C)0.4-0,4C-0.2-IL.00aci000.0I0.0IIlimneticbenthiclimneticbenthica)(Priest)mateformb)(Paxton)maleformII.01•00.8-ci0.8-c0°-CC’)cI(tci)CI).4—9-04-04-.0‘S____________________10.0IIbenthiclimneticbenthiclimnetica)(Priest)maleformb)(Paxton)maleformcJTh:ha) L. 0 0 C’) 0 0 C) ci) (‘I E5 4 3 2 1 0\eSpopulationN-climneticfemalestandardlength(mm)GDGD0101001001001-(iiIICD-CD0100:3rm•-\-,..L501:’Qt010010t>0ci 0)-0C-C)00:3CC0:3-030)030105z70II 065-V0(Th-0 C0Cl)00(11AE0004045505560limneticmalestandardlength(mm)27chapter 2. A TEST OF PARALLEL SPECIATIONINTRODUCTIONParallel evolution is a form of homoplasy in which similar evolutionary changes occurindependently in closely related lineages (Futuyma 1986), and there are many possible examples.Some recent examples include ecological and morphological diversification of Anolis lizardsinhabiting Caribbean islands (Losos 1992); reduced eye size and increased antennae length incave-dwelling amphipods in eastern North America (Jones et al. 1992; Culver et al. 1994); anda small, freshwater form of sockeye salmon (kokanee) in many lakes in western North America(Foote et al. 1989). In other fish species, including threespine sticklebacks, repeated paralleldivergence into ‘benthic’ and ‘limnetic’ forms has been identified (e.g. Schiuter and McPhail1993).Parallel evolution occurs when selective regimes acting on closely related populationsare very similar, and result in the same trait evolving independently in each lineage. Naturalselection is presumably the cause of these repeated evolutionary events because genetic driftalone would not produce independent, replicate shifts in the same direction (Endler 1986; Clarke1975). When similar selective regimes cause evolutionary transitions to occur independently inmultiple populations, then each population is a replicate of the same process and can be used totest hypotheses of evolutionary theory. The evolutionary process that I will examine isspeciation. I use Mayr’s (1963) definition of a species: a group of organisms reproductivelyisolated from other such groups. This is known as the biological species concept.One of the ways in which reproductive isolation between groups is thought to evolve isas a by-product of natural selection. Natural selection may act in an adaptive or non-adaptivefashion on mate preferences which cause reproductive isolation between groups. Matepreferences may simply be correlated via linkage disequilibrium or pleiotropy with traits which28are under direct natural selection (for a review, see Rice and Hostert (1994). Natural selectioncan also act on mate preferences as a consequence of selection on other traits in a population.Although it may appear that selection is acting directly on mate preferences (as in reinforcement),this is actually secondary selection. Mate preferences will then change along with changes inthese traits as a ‘chain of selection’ (Schluter and Nagel ms).In chapter one, I described tests which showed that body size may be an importantcomponent of the mate recognition system in sympatric populations of threespine sticklebacks.I suggested that reproductive isolation evolved as a by-product of natural selection on body size.In this chapter, I describe tests conducted among these sympatric populations to determine ifreproductive isolation has evolved along parallel lines in these populations as a result of theparallel evolution of body size and shape. Traits which evolve by parallel evolution may formpart of the specific mate recognition system (Patterson 1985) of a population. If a trait used inmate recognition (such as body size) evolved by parallel evolution (see figure 6), thenreproductive isolation between the ancestor and the descendent species could occur in a parallelfashion as well if mate preferences evolve as a by-product. Speciation will thus occur in asimilar way in each lineage such that each form is reproductively isolated from the ancestralpopulation but not from one another. The independent evolution of the same isolatingmechanism is termed parallel speciation (Schiuter and Nagel, ms).Three criteria should be satisfied before it can be said that parallel speciation hasoccurred:1. Each lineage sharing a trait must have an independent origin.2. Reproductive isolation must evolve between the ancestral population and the descendantpopulations. In cases where descendent populations are sympatric, reproductive isolationmust evolve between these populations.293. Descendant populations which have evolved under similar selective regimes indifferent localities must not be reproductively isolated from each other.To determine if natural selection is responsible for parallel evolution an adaptive mechanismshould be identified and tested.Some of these criteria have been met in laboratory experiments with Drosophila, but therelevance of parallel speciation in nature has not been addressed. I conducted a test of parallelspeciation with populations of sympatric threespine sticklebacks which evolved from a commonancestor and have reached the level of biological species. A complete phylogeny ofinterpopulation relationships is not yet available, but biochemical analysis to date is consistentwith the hypothesis that each population has an independent origin. For example, fish fromPaxton Lake have one allele not present in the Priest or Enos Lake populations (E.B. Taylor,pers. comm.). Evidence for the first criterion is therefore lacking.In chapter one, I described tests which met the requirement of reproductive isolationbetween sympatric descendant populations in the same lake (requirement #2). Note thatreproductive isolation between the marine ancestor and the descendant populations was notaddressed. In this chapter, I address requirement #3 by conducting mate choice tests amongpopulations which have experienced the similar selective regimes in different lakes. If thedifferent populations are not reproductively isolated, then I suggest that the same reproductiveisolating mechanisms arose independently in each lineage.STUDY POPULATIONSThe origin of these sympatric populations is discussed in detail in chapter one, but keypoints are summarized here. Located in a restricted geographical area in southwestern BritishColumbia, Canada, they occur in four watersheds on three islands in the Straight of Georgia.Each lake contains two kinds of stickleback. One form is a small, limnetic stickleback and the30other is a much larger benthic stickleback. The limnetic stickleback feeds largely on planktonwhile the benthic forages on benthic invertebrates. All limnetic sticklebacks are morphologicallyand ecologically similar, so we refer to a limnetic form. Similarly, all benthics belong to thebenthic form.McPhail (1993) suggests that the origin of these populations is due to an unusualglaciation event which occurred in this region. As the glaciers receded approximately 15 000years ago, two temporally separate invasions of marine sticklebacks occurred. Ecologicalcharacter displacement occurred between the two colonists, resulting in the very distinct formsinhabiting each lake today. There is geological evidence which supports this double-invasionscenario (Mathews Ct al. 1970; Clague 1981; Clague et al. 1982).The benthic and limnetic pairs in each lake represent distinct gene pools. For example,in one population from Vancouver Island, one locus is fixed in the benthic but limnetics showan 18% frequency of a variant allele (McPhail 1984). The morphological differences areheritable, and the two species are reproductively isolated within each lake (McPhail 1993).Hybrids of the two species appear to be very rare (less than 1% based on morphologicalmeasures (McPhail 1993)).Limnetics were collected from two lakes (Paxton and Priest) on Texada Island, BritishColumbia, Canada (49° 40’N 124° 30’W). Benthics were collected from these two lakes, andfrom Enos Lake on Vancouver Island (49° 17’N 124° 09’W).Body length and depth are remarkably similar among forms from different lakes but thereare some morphological traits which differ slightly (e.g. gill raker number, number of lateralplates, extent of pelvic girdle development) (Bentzen and McPhail 1984; McPhail 1991; Schiuterand McPhail 1992). These traits are unlikely to be important in mate recognition. Sexualdimorphism in body size is consistent among populations: breeding benthic males are usually31smaller than breeding females and in limnetics, breeding males are larger than females.Male threespine sticklebacks develop bright nuptial colouration at the beginning of thebreeding season. The usual pattern is a reddish throat region and blue-green body colouration.This visual cue is used in both male-male aggression and mate attraction. Males build nests outof plant fibres, court females, and raise fry. The nuptial colouration of male limnetics is similarin each lake and is typical of most threespine sticklebacks. However, benthics from one lake area striking exception to the usual pattern. Males from Enos Lake are black over their entire body,with no reddish throat, while males from Priest and Paxton Lakes develop the typical blue-greenbody coloration.METHODSFemales were collected with baited minnow traps left out for 8-13 hours several daysbefore being used in tests. Fish captured in the traps were examined for reproductive conditionand gravid animals were immediately brought into the lab and housed in 102 L or 180 L tanks.They were fed a mixture of frozen bloodworms and live Artemia sp. After 48 hours, they wereclosely examined for reproductive condition. Females who are ready to spawn have a verydistended body cavity, and these animals were then used in mate choice tests.Males were collected with baited minnow traps from Texada Island on February 12 andMarch 7 in 1992 and on February 11 in 1993. They were collected from Enos Lake onVancouver Island on March 15, 1993. Benthic and limnetic species were held separately in 102L tanks in an environment chamber at The University of British Columbia in Vancouver. Thetanks were lit by rows of “cool white” fluorescent lamps. The photoperiod was graduallyincreased from 1OL: 14D to 16L:8D over a two month period in order to bring the males intobreeding condition at approximately the same time. The temperature during this period wasincreased from 7 degrees to 10 degrees Celsius.32All males had developed nuptial colouration by early March. In April, the males weremoved to Texada Island and housed under similar conditions. No mortality or aberrant behaviourwas associated with this move. The light regime was maintained at 16L:8D, and temperaturesfluctuated between 15 and 18 degrees Celsius throughout the season. The holding tanks were litwith fluorescent lights supplemented with 60W incandescent bulbs. The fish were fed to satiationonce daily with frozen brine shrimp (Artemia sp.),and bloodworms (chironomid larvae).Benthic males from Priest Lake differed considerably from other populations in adaptingto captivity in both years. They were very cryptic and easily frightened, even after four monthsin captivity. They also never fully developed the nuptial colouration of wild males from this lake(pers. ohs.). The Priest Lake populations probably experience considerably more predation thanfish in Paxton or Enos Lake, since trout are rare in those lakes. This may explain the nervousbehaviour of these fish in the aquaria, which are relatively bright and in which there is littlecover. However, limnetic males and wild-caught females from this lake did not seem to beaffected to the same extent.Mate choice testsInitially, I planned to conduct choice tests in which one female has access to two nestingmales. However, benthic males were very aggressive, and two males could not be kept in thesame aquaria. These tests were therefore only conducted with limnetics. No-choice tests wereconducted with limnetics and benthics.1. NO-CHOICE TESTS (limnetics and benthics’)The procedure for these tests was as previously described (chapter 1). A gravid femalewas put in a 55 L aquarium which contained a nesting male. Whether or not the pair spawnedin 30 minutes was recorded. Tests in which females were unresponsive in the first five minutesof the test were excluded from analysis. This was done to exclude females who were not33receptive, which occurred infrequently (about 10% of all tests). After each test, the male andfemale were measured for standard body length.2. CHOICE TESTS (limnetics only)Twenty-four pairs of males were used with fourteen females from Priest Lake and tenfrom Paxton Lake. Each pair was used only once. Each limnetic female was allowed to choosebetween a limnetic male from her own lake or a limnetic male from the other lake who werenesting in the same aquarium. The criterion for choice was spawning.A 180 L tank was fitted with two plexiglass dividers: one opaque and one clear. Pairsof size-matched (+1- 2mm) males from the two lakes were put randomly into each side of thetank with both dividers in place. When both males had built nests (2-4 days), the opaque dividerwas removed. This allowed the males to interact visually, and in effect, created an artificialterritorial boundary. Males responded vigorously to each other when the divider was firstremoved, but gradually seemed to become accustomed to one another. Approximately 24 hoursafter the opaque divider was removed, the clear divider was removed and males were allowedto interact physically for 15 minutes. Although interactions sometimes became very aggressive,males tended to stay on their own side of the artificial boundary.When the clear divider was removed, a female was removed from the communal tank andplaced into an open 500 mL glass jar. During this acclimation period, the jar was placed nearthe test tanks but was visually isolated from all males. After fifteen minutes, the jar (which wasopen at the top) was placed in the centre of the test tank with the top about 20 cm below waterlevel. In most tests both males immediately began to court the female.A test began when the female swam out of the jar (usually within three minutes). Irecorded the time until the female entered one of the male’s nests. Spawning occurred within30 minutes in 83% of the tests conducted (see Appendix I). However, in four tests, spawning34took close to 60 minutes. These tests were included in the analysis because male-malecompetition was thought to be the main reason for the lengthy time interval. Aggressiveinteractions between the males seemed to be a major factor determining when spawning occurredin most of the tests. Females in these tests were actively engaging in courtship from early on inthe tests. Females always deposited eggs in the nests they entered.AnalysisI used Yates’ corrected chi-square tests (SYSTAT version 5.01, SYSTAT Inc., 1989) totest for assortative mating between forms from different lakes. I used logistic regression toexamine effects of male and female standard length on probability of spawning (SAS Version6.03, CATMOD procedure; SAS Institute Inc., 1988).RESULTSIn choice tests with limnetics, there was no statistically significant reproductive isolationbetween forms from different lakes (table 5). However, females tended to prefer to spawn withmales from their own lake. The data were combined for a single chi-square goodness-of-fit testof females choosing mates from their ‘own lake’ vs. ‘other lake’. This result was also notsignificant at p=O.O5 (Yates’ corrected chi-square=2.042, df=1).The results of no-choice tests with linmetics were similar in both populations and areillustrated in figure 7. The results of chi-square contingency tests are listed in table 6. Althoughthere is a trend towards an increased probability of spawning in pairs from the same lake, thisis not statistically significant. To determine if both samples could be combined for a strongertest of ‘own lake’ versus ‘other lake’, the data were then tested for heterogeneity (Zar 1984).The heterogeneity chi-square value was 0.66, which allowed both samples to be pooled for asingle test of ‘own lake’ versus ‘other lake’. The results for the pooled data are also listed intable 6. This produced a marginally significant result (p=O.05), which nevertheless shows that35there is some evidence for weak reproductive isolation between limnetics from different lakes.The trend towards an increased probability of spawning when pairs were from the samelake was not evident in no-choice tests with benthics (figure 8). The results for chi-squarecontingency tests are listed in table 7. All tests involving Priest Lake males resulted insignificant differences between male types in the proportion of pairs that spawned becausefemales from both populations rarely spawned with Priest Lake males (only 25% of pairs tested).However, females from both populations spawned equally frequently with males from Enos Lakeand from Paxton Lake (66% of pairs tested). This result was surprising because nuptialcolouration in Enos males is very different from both Priest and Paxton males. Enos males areblack, while males from the other population have the typical blue/red colouration of most malethreespine sticklebacks. The fact that Priest Lake females preferred males from two differentpopulations over males from their own is strong evidence that the Priest Lake population is notreproductively isolated from the other two populations. Females from Paxton Lake preferredPaxton Lake males over Priest Lake males (ptO.Ol), but did not discriminate against Enos Lakemales (p=l.OO). This too indicates that Priest benthics are not reproductively isolated from theother populations.In chapter one, I suggested that body size appeared to be an important trait used to assessmates in these populations. To gain further insight into the effect of body size on the probabilityof spawning, I analyzed the data for possible effects of male and female standard length. I wasalso interested in the possible effects of trial date because I suspected that fish of both sexesmight become less choosy later in the breeding season. Data on the effect of size and date onthe probability of spawning within each form in each lake were analyzed separately. Themagnitude and sign of all variables was similar in each lake, so the data were combined forfurther analysis. In limnetics, there was a significant positive effect of season and male length36on the probability of spawning (table 8). Female standard length was not significant. The effectof season on the probability of spawning was also significant in benthics, as were male andfemale standard length (table 9).DISCUSSIONThere is no strong evidence of reproductive isolation among sticklebacks that haveevolved under the same selective regimes, whether in the same or in different lakes. Limneticsticklebacks from two lakes spawned often with each other in two different protocols. Benthicsticklebacks from three lakes also spawned often with mates from the other two lakes.However, there is a trend in limnetics from both lakes to prefer mates from their ownpopulation. This trend is evident in both the choice and no-choice tests, but is marginallystatistically significant only when the data from no-choice tests from both lakes is combined.Nevertheless, this suggests that there are subtle differences between populations that are importantin mate choice. Molecular evidence (E.B. Taylor, pers. comm.) suggests that limneticsticklebacks evolved independently in each lake, probably from a common marine ancestor. Thisis in accordance with the double invasion hypothesis (McPhail 1993). The trend in preferencefor mates from their own lake in limnetics suggests that stronger reproductive isolation mayeventually develop due to genetic drift, or slightly different selective regimes in the two lakes.If this is the case, it suggests that parallel speciation may be detectable only for a short time inthe evolution of parallel lineages.No similar trend was evident in benthics. The benthics are also thought to have evolvedindependently from a common ancestor (McPhail 1993). Results from tests with benthics werecomplicated by a strong aversion by females of both lakes for males from Priest Lake. As inchapter one, Priest Lake benthic females rarely spawned with Priest Lake benthic males. Isuggest that this was probably an artifact of the inability of males from this population to adapt37to captivity. This appears to have resulted in erratic behaviour and dull nuptial colouration (seefigure 3) and a concomitant lack of female preference for them. Male colour score was notevaluated in Enos Lake males because they do not possess typical nuptial colouration; howevermale colour may have had some effect on female preference. McLennan and McPhail (1990)and Milinski and Bakker (1990) showed that female threespine sticklebacks prefer brighter males.This may explain why females from both populations discriminated against the pale Priest Lakemales (see figure 3). As mentioned above, this aversion may also be due to aberrant behaviouraltraits in these males. It is intriguing that females from both populations were strongly attractedto Enos Lake males, who do not have blue/red nuptial colour. It is possible that black nuptialcolouration was attractive because it is unusual, but it is also possible that males from thispopulation were attractive because of differences in courtship behaviour. Enos Lake males seemto be more aggressive courters compared to males from the other two lakes (L. Nagel, pers. ohs.;J.D. McPhail, pers. comm.).In the previous chapter, I showed that there is strong reproductive isolation betweenspecies within the same lake, and I suggested that this isolation is determined at least in part bydifferences in body size. I suggested that reproductive isolation evolved as a by-product ofnatural selection, and this process was repeated in each lineage. This satisfied the secondcriterion of parallel speciation: reproductive isolation between sympatric descendant populations.In this chapter I provide evidence which satisfies the third criterion of parallel speciation:descendant populations inhabiting the same environment are not reproductively isolated. Theabsence of reproductive isolation suggests that the traits involved in mate choice are similar ineach population.The adaptive mechanism responsible for the evolution of reproductive isolation isprobably natural selection on body size. Body size is highly variable among stickleback38populations (Bell 1976; Schiuter and McPhail 1992). Body size may have simply been correlatedwith traits which under direct natural selection. However, size is one of the most importantdeterminants of feeding efficiency, so it may have been under direct selection. For example, infeeding trials, Schluter (1993) found that feeding efficiency of benthics on plankton decreaseswith increasing size. Limnetics may have been selected for small body size for efficiency whenforaging on plankton. The large gape width of the benthic may have evolved for foraging onlarger benthic prey items. Natural selection may therefore have favoured an increase in body sizefor the benthics and a decrease in body size for limnetics as a result of specialization on differentprey types.Evidence to confirm the first criterion (independent origins) is still lacking. We do nothave a complete phylogeny of these populations, although molecular evidence to date isconsistent with the hypothesis that the sticklebacks in different lakes have independent origins.This phylogeny is forthcoming (E.B. Taylor, pers. comm.).These preliminary results indicate that parallel speciation may have occurred in thesepopulations of sticklebacks. Parallel speciation has probably been overlooked in many taxa.Wherever an ancestral population gives rise independently to descendant populations via parallelevolution, the potential for parallel speciation exists. The only requirement is that the selectiveregimes in each new environment must be similar so that the evolution of reproductive isolationoccurs in a similar direction in each lineage.The best evidence for parallel speciation so far may be from laboratory experiments withDrosophila. Kilias et. al. (1980) raised lines of D. melanogaster in two different selectiveregimes: a cold, dry, dark environment and a wet, damp, light environment. After five years,mate choice tests showed that flies from the different environments had evolved some degree ofreproductive isolation from one another. However, flies held separately under the same selective39regimes were not reproductively isolated from each other. In a similar experiment, Dodd (1989)raised lines of D. pseudoobscura on either a high-starch or a high-maltose diet. After one year,flies from the different environments had evolved a significant amount of reproductive isolationbut flies experiencing the same selective regimes did not. Both of these researchers concludedthat reproductive isolation had evolved as a genetically correlated by-product of selectionpressures involved in adapting to the different environments. This mechanism suggests that theevolution of mate preferences occurs as a non-adaptive, incidental result of pleiotropy or linkagedisequilibrium. However, mate preferences may also evolve as a result of secondary naturalselection on mate preference to accommodate new phenotypes in available mates which evolveby direct natural selection (Schluter and Nagel, ms). In a review, Rice and Hostert (1994) pointout that if genetic drift had indirectly caused the evolution of reproductive isolation, then it wouldbe expected to cause isolation between the populations who had experienced similar selectiveregimes. Since this did not occur in any of these examples, natural selection must be the causeof the divergence in mate preferences.Other potential examples of parallel speciation include sockeye salmon (Onchorhynchusnerka). Foote and Larkin (1988) describe size-based assortative mating between sympatricpopulations of large, anadromous sockeye and the small, freshwater kokanee. The anadromousform independently gave rise to these nonmigratory populations of kokanee in many drainagesystems in western North America (Foote et al. 1989). Thus, two of three criteria for thedemonstration of parallel speciation have been satisfied for this system. What remains to beshown is that different populations of kokanee are not reproductively isolated.Another possible example are cave amphipods which have independently evolved fromspring populations in several drainage basins (Culver et a!. 1994). Cave populations ofGammarus minus are larger and have reduced eyes and larger antennae than the ancestral spring40amphipods. Tests of mating success indicate that reproductive isolation may be size based.Small spring males are physically unable to amplex with large cave females. In addition, largecave males sometimes view small spring females as food rather than a potential mate.As well as providing possible evidence for the by-product mechanism of speciation,parallel speciation has other interesting implications. The existence of parallel speciationsuggests that biological species need not be monophyletic. Parallel evolution not only causes theevolution of the same reproductive isolating mechanisms, but according to the biological speciesconcept, the same species evolves several times (all limnetic sticklebacks are the same speciesbecause they are not reproductively isolated, as are all benthics). Yet evidence to date stronglysuggests that the sticklebacks in each lake have independent origins, and that they have beenisolated for at least 11 000 years.It has been suggested that evolution may be a largely random process and that, given thechance to repeat itself, would follow a different course each time a tape of it were ‘replayed’(Eldridge and Gould 1972). This view suggests that chance and historical accident are extremelystrong influences on the course that evolution follows. Schluter and McPhail (1993) provide arecent summary of examples of adaptive radiation which suggest that some patterns ofevolutionary change are repeatable. They give examples of organisms in which similarevolutionary changes have occurred independently multiple times under similar circumstances.In sticklebacks, evolutionary events have not been random, but show repeatable patterns ofevolutionary change. Similar selective regimes may cause not only similar morphologies andbehaviours to evolve, but can also cause speciation to unfold in a predictable sequence as well.That microevolutionary events are not random is well known- there are many examples ofadaptation. That these events are precisely repeatable is interesting, but that speciation eventsare repeatable is remarkable.41Table five. Results of choice tests with limnetic females from Paxton and Priest Lakes given achoice between spawning with a limnetic male from Priest or Paxton Lake nesting in the sameaquarium. Twenty-four different pairs of males were used with 24 females (chi-square=l .543;df=l; p=O.2l).Paxton Lake Priest LakePaxton Lake o’ 7 5Priest Lake d 3 942Table six. Results of cu-square contingency tests with limnetics from two lakes. N=the numberof trials of each pair conducted. “#spawning” is the number of these trials which resulted inspawning.Pair N #spawning chi-square p-valuePaxtonX 9 7Paxton c?’2.025 0.16PaxtonX 9 3 -Priest a”Priest X 20 11Priest a”I 1.326 0.25PriestX 19 6Paxton a”combined data 57 27 3.987 0.0543Table seven. Results of within-form tests with benthics from three lakes. N=the number of trialsof each pair conducted. ‘#spawning’ is the number of these trails which resulted in spawning.Pair N #spawning chi-square p-valuePaxtonX 21 6Priest d’ 6.404 0.01IPaxton X 24 17Paxton d’ 10.000 1.00PaxtonX 13 9 -Enos dPriest X 21 12 1Paxton d’ 4.48 1 0.03IPriest X 20 4Priest d’ 1I 4.096 0.04PriestX 9 6Enos a”44Table eight. Results of Maximum-Likelihood estimates analysed by logistic regression on theeffect of date of test and female and male standard length on the probability of spawning inlimnetic pairs (both lakes combined). Asterisks indicate a significant effect on the probabilityof spawning using chi-square analysis. Likelihood ratio degrees of freedom are 52 for date andlengths. (N=1 10).Variables Estimate (SE) dfintercept 115.9 (37.47) 1date** 0.6397 (0.21) 1female length 0.0776 (0.07) 1male length* 2.2664 (0.74) 1* P < 0.05= 0.001Table nine. Results of univariate Maximum-Likelihood estimates carried out individually bylogistic regression on the effect of date of test and female and male standard length on theprobability of spawning in benthic pairs (males from three lakes paired with females from twolakes). Asterisks indicate a significant effect on the probability of spawning using chi-squarecontingency analysis. (N=108).Variables Estimate (SE) df Likelihoodratio dfintercept 4.8259 (1.60) 1date**-0.0345 (0.01) 1 43intercept -6.7129 (2.37) 1female length** 0.1109 (0.04) 1 75intercept -5.8278 (2.46) 1male length* 0.099 1 (0.04) 1 59P <0.05**0.00l