University of South Carolina Scholar Commons

Faculty Publications Biological Sciences, Department of

11-2000

Effect of Harvest and Effective Population Size on in a Striped Bass Population

Marilyn Diaz University of South Carolina - Columbia

David S. Wethey University of South Carolina - Columbia, [email protected]

James Bulak University of South Carolina - Columbia

Bert Ely University of South Carolina - Columbia, [email protected]

Follow this and additional works at: https://scholarcommons.sc.edu/biol_facpub

Part of the Biology Commons

Publication Info Transactions of the American Fisheries Society, Volume 129, Issue 6, 2000, pages 1367-1372. © Transactions of the American Fisheries Society 2000, American Fisheries Society.

This Article is brought to you by the Biological Sciences, Department of at Scholar Commons. It has been accepted for inclusion in Faculty Publications by an authorized administrator of Scholar Commons. For more information, please contact [email protected]. Transactions of the American Fisheries Society 129:1367±1372, 2000 ᭧ Copyright by the American Fisheries Society 2000

Effect of Harvest and Effective Population Size on Genetic Diversity in a Striped Bass Population

MARILYN DIAZ1 AND DAVID WETHEY Department of Biological Sciences, University of South Carolina, Columbia, South Carolina 29208, USA

JAMES BULAK Department of Biological Sciences, University of South Carolina, Columbia, South Carolina 29208, USA and South Carolina Department of Natural Resources, Eastover, South Carolina 29044, USA

BERT ELY* Department of Biological Sciences, University of South Carolina, Columbia, South Carolina 29208, USA

Abstract.ÐA major factor that contributes to loss of population size, particularly if sustained over sev- genetic variation in natural populations is a small ef- eral generations (Wright 1938; Nei et al. 1975). A fective population size. In species with a complex life population with a low effective population size for history that involves overlapping generations and de- layed maturity, the impact of infrequent annual repro- several generations experiences an increase in ho- ductive bottlenecks is likely to be small because effec- mozygosity that can result in further reductions of tive population size is de®ned by the number of indi- effective population size when deleterious reces- viduals contributing to a generation and not to a single sive alleles are expressed in homozygous individ- year-class. The striped bass Morone saxatilis is a long- uals (Charlesworth and Charlesworth 1987; Hed- lived species with overlapping generations and age structure, whose recreational and commercial impor- rick and Miller 1992; Caughley 1994). Thus, a tance has made it a target of intense harvest. We analyzed major goal of conservation genetics is to promote allele frequency ¯uctuation among juvenile year-classes the preservation of genetic diversity in managed of the Santee±Cooper, South Carolina, population from populations by maintaining high effective popu- 1990 through 1994 with three independently segregating lation sizes. polymorphic nuclear DNA loci to examine The striped bass Morone saxatilis is an anad- and estimate the number of breeders each year. Signif- icant ¯uctuations in allele frequencies among juvenile romous species with overlapping generations, a year-classes were observed, and most of the variation long life span (up to 20 years), and delayed ma- was attributed to a small number of parents in 1992. The turity (Rago and Goodyear 1987; Winemiller and potential impact of this year-class is likely to be low Rose 1992). Its natural range includes the Gulf because 1992 was a poor recruitment year, and striped Coast of the United States and Atlantic estuaries bass have multiple opportunities to breed. However, high adult mortality due to ®shing may increase the impact and inshore coastal sites from the St. Johns River Downloaded By: [University of South Carolina] At: 19:52 31 May 2011 of the 1992 year-class by decreasing the number of adult in Florida to the Canadian maritime provinces (Ra- age-classes in this population. Thus, high exploitation ney 1952). The striped bass is exploited commer- in species with overlapping generations can reduce the cially in the northeastern United States and is a long-term effective population size by abrogating the popular sport ®sh throughout its range. Thus, it is possibility of multiple breeding opportunities. important to study the factors that affect genetic diversity in this species. Loss of genetic variation can effectively reduce Natural recruitment of juveniles has been shown the long-term prospects of survival of a natural to ¯uctuate considerably and can result in several population and ultimately a species. A major cause consecutive years of poor recruitment (Ulanowicz of loss of genetic diversity is a small effective and Polgar 1980). However, sporadic bursts of high recruitment can restore historic abundance * Corresponding author: [email protected] levels in striped bass populations (Richards and 1 Present address: Department of Immunology, Imm- Rago 1999). Bulak et al. (1997) showed that nat- 16, Scripps Research Institute, 10550 North Torrey Pines ural recruitment is highly variable in the Santee± Road, La Jolla, California 92037, USA. Cooper system, South Carolina. It is not clear, Received August 2, 1999; accepted June 6, 2000 however, whether variability among year-classes

1367 1368 DIAZ ET AL.

in the number of breeders reduces effective pop- from whole ®sh (excluding the digestive tract) by ulation in striped bass. This uncertainty is analo- a proteinase-K digestion followed by phenol-chlo- gous to Chesson's storage effect (Chesson 1983) roform extractions as described by Harrell et al. because overlapping generations and multiple (1993). breeding opportunities can reduce the genetic con- Assays.ÐThree independently segregating var- sequences of periods of poor recruitment and low iable nuclear DNA markers that are inherited in a numbers of contributing parents (Ellner and Hair- Mendelian fashion were analyzed by polymerase ston 1994). chain reaction ampli®cation±restriction fragment A primary objective of this study was to gen- length polymorphisms (i.e., restriction digests of erate annual estimates of the effective number of ampli®ed loci; Leclerc et al. 1996). These markers breeders in the Santee±Cooper population and to are single-copy regions of the striped bass nuclear ask whether ¯uctuation in the number of breeders genome that are polymorphic at one or more re-

has an impact in effective population size (Ne). striction enzyme recognition sites: SB14 contains Therefore, we analyzed the ¯uctuation of allele a variable A¯ II site, SB8-2 is variable at one of frequencies at three independently segregating, two Mnl I sites, and SB83 is polymorphic at the polymorphic loci among year-classes of juvenile recognition sites for Rsa I and Pst I, resulting in striped bass from the Santee±Cooper system. To a three allele system. The signi®cance of ¯uctu- examine the impact of adult mortality, we con- ations in allele frequencies was determined using structed a genetic submodel within an age-struc- the G-statistic test (Sokal and Rohlf 1981). tured population projection model that incorpo- Estimate of the effective population size.ÐAllele rated the life history and annual mortality of frequency variance (F) among young of the year striped bass in the Santee±Cooper system (Bulak was used to generate annual estimates of the ef-

et al. 1995). fective population size (Ne) using the methods of Waples (1989), Pollak (1983), and Jorde and Ry- Methods and Materials man (1995). The scaled allele frequency shift from Sample collection.ÐJuvenile striped bass (N ϭ time t to t ϩ 1 is (Pollak 1983) 55±119) were collected during the summers of 2(a X Ϫ X )2 1992 through 1994 with a 10.7 ϫ 1.8-m beach F ϭ ͸ i,t i,tϩ1 , seine with a 4.8-mm mesh. We sampled the historic a Ϫ 1 iϭ1 Xi,tϩ X i,tϩ1 nursery ground for Santee±Cooper striped bass at where a is the number of alleles at the locus and dusk (White and Lamprecht 1989). To ensure that X is the frequency of the ith allele in the sample samples were representative of the entire year- i,t from cohort t. The scaled allele frequency change class, collections began 50 d after the ®rst recorded adjusted to sample sizes is spawning event and continued until 65 d from the last spawning event. In 1992, when recruitment 11 was low, sampling was carried out twice weekly, FЈϭF ϪϪ , 2ntt2n ϩ1 with three 45-min sampling periods spaced by 15- where n is the number of individuals sampled at

Downloaded By: [University of South Carolina] At: 19:52 31 May 2011 min resting periods. In 1993 and 1994, when re- t cruitment was high, sampling was carried out once time t. per week. The daily collection limit was 100 ®sh. A weighted average FЈ was calculated using the Juvenile samples from 1990 and 1991 were ob- data from three loci (Waples 1989): tained from collections by the South Carolina De- (KjjϪ 1)FЈ partment of Natural Resources that used a similar FÅ Јϭ͸ , sampling protocol. To prevent oversampling of ͸ (Kj Ϫ 1) early spawners, ®sh larger than 60 mm were ex- where K is the number of alleles at locus j. cluded from the analysis. j Effective population size (N ) was estimated as Whole ®sh were placed in jars ®lled with 95% e (Jorde and Ryman 1995): ethanol and stored at room temperature. Because hatchery ®sh were stocked in the vicinity of the C N ϭ , collection site in all years except 1994, the otoliths e 2GFÅ Ј were examined for an oxytetracycline mark that identi®ed juveniles of hatchery origin (Bulak where C is a correction factor for overlapping gen- 1994), and ®sh showing the mark were excluded erations (7.81 for this population of striped bass from further analyses. Genomic DNA was isolated based on the correction factor of Jorde and Ryman NOTES 1369

1995) and G is the generation time (5.25 for this speci®c fecundity. All age-classes were subdivided population; Bulak et al. 1995). into categories representing the number of indi- The 95% con®dence limits on FЈ were calculated viduals of each genotype. In each time step, we as in Waples (1989), using the chi-square distri- used a random number generator to decide on an bution: individual-by-individual basis who survived and who died, using the age-speci®c mortality rate to ÅÅ nFЈ nFЈ determine the individual probability of death. We 22,, XX0.025[n] 0.975[n] assumed no difference in mortality rate or fecun- dity among genotypes. The initial population at where n is the number of independent alleles (four time zero had equal proportions of all ®ve alleles, in this case). distributed among genotypes in Hardy±Weinberg Estimate of the effective number of parents con- ratios. The simulations ran until one allele was tributing to a year-class (Nb).ÐIf one considers lost, and the number of years elapsed was record- the young of the year to be a genetic sample of ed. Fifty replicate simulations were performed for their parents, then one can use the allele frequency each condition. difference between the young and the adults in a The rate of loss of alleles was calculated assum- population to estimate N . We calculated the b ing the process to be exponential (Wright 1938). weighted allele frequency difference and deter- Wright's model assumes that the drift of allele fre- mined Nb as quency A from A0 to At over t years is r: At ϭ Ϫrt 1 A0e . Taking natural logarithms of both sides of N ϭ . b 2FÅ Ј this equation,

Simulation model.ÐTo examine the impact of 1 At r ϭϪ loge . adult mortality on the distribution of age-classes tA΂΃0 and on the rate of loss of genetic variation, a one- We determined the rate r from our simulations by locus, ®ve-allele genetic model was incorporated determining the length of time in years to lose one within an age-structured population projection out of ®ve alleles, which makes A ϭ 4 and A ϭ model for the Santee±Cooper striped bass popu- t 0 5. In Wright's theory, if time, t, is measured in lation (Bulak et al. 1995). For simplicity, selection, generations, the rate, r, is 1/2N . gene ¯ow, and mutation were ignored. Fecundities e of all adult age-classes were taken from Bulak et Results and Discussion al. (1995). Yearly adult survivorship was 0.85 in populations not exposed to harvest and was 0.40 Fluctuation of Allele Frequencies among Year- when harvest was allowed (Bulak et al. 1995). To Classes from the Natural Reproduction determine the in¯uence of effective population Signi®cant (P Ͻ 0.05) ¯uctuation of allele fre- size on the loss of genetic variation, each year, a quencies among juvenile year-classes was ob- small number of parents was selected at random served at the SB83 and SB14 loci (Table 1). Most

Downloaded By: [University of South Carolina] At: 19:52 31 May 2011 and allowed to mate; separate simulations were run of the difference originated from the comparisons with 5, 10, 20, 50, and 100 mating pairs in each of the 1992 year-classes with all others, suggesting year. No other adults were allowed to mate. Mo- that a reproductive bottleneck occurred that year. nogamy and equal sex ratio were assumed to sim- Removal of the 1992 year-class from the SB83 plify the calculations. Recruitment to age-1 was analysis increased the probability of homogeneity obtained by randomly removing individual larvae above the signi®cance level (from P ϭ 0.0182 to until the numbers equaled 150,000 (the minimum P ϭ 0.1412 without 1992). In contrast, the ¯uc- number of age-1 recruits estimated for the Santee± tuation remained signi®cant when any of the other Cooper lakes; Bulak et al. 1995). Thus, the model years were removed from the analysis. This result represents a constant low recruitment scenario, indicates that the allele frequencies at this locus with low effective population size. The number of were relatively homogeneous in years 1990, 1991, individuals in each subsequent age-class was de- 1993, and 1994 but were distinct from those of termined by standard population projection meth- 1992. In the case of the SB14 locus, removal of ods (Caswell 1989; Bulak et al. 1995). either the 1992 or the 1990 year-classes resulted The numbers and genotypes of all individual in increases of the probability of homogeneity offspring were determined from the genotypes of above signi®cance levels (without 1992; P ϭ each mating pair of parents and the female's age- 0.1914; without 1990, P ϭ 0.2088). The overall 1370 DIAZ ET AL.

TABLE 1.ÐAllele frequencies for individual year-classes. Ho/He is the observed heterozygosity divided by the het- erozygosity expected from the observed allele frequencies.

SB83a SB14b SB8-2c

Year N A1 A2 A3 HoHe N B1 B2 Ho/He N C1 C2 HoHe 1990 72 0.59 0.14 0.27 0.55/0.56 69 0.56 0.44 0.48/0.49 61 0.30 0.70 0.38/0.42 1991 68 0.63 0.12 0.25 0.46/0.53 67 0.48 0.52 0.55/0.50 49 0.38 0.62 0.39/0.47 1992 55 0.44 0.16 0.40 0.67/0.62 49 0.37 0.63 0.33/0.47 35 0.51 0.49 0.51/0.50 1993 88 0.51 0.13 0.36 0.61/0.59 119 0.48 0.52 0.48/0.49 93 0.31 0.69 0.53/0.43 1994 86 0.63 0.14 0.23 0.52/0.54 107 0.44 0.56 0.41/0.49 47 0.35 0.65 0.45/0.46 a Overall G ϭ 18.41, P ϭ 0.018. b Overall G ϭ 9.53, P ϭ 0.049. c Overall G ϭ 9.43, P ϭ 0.052.

probability of homogeneity among years for the the Ne values from Table 2, the ratio of Ne to the comparison at the SB8-2 locus was P ϭ 0.052 (Ta- estimated number of spawning females in the pop- ble 1), and again, the 1992 year-class was respon- ulation is between 5 ϫ 10±3 and 3 ϫ 10±4. Thus, sible for most of the variation. Thus, data from all these analyses indicate that only a small fraction three loci indicate that most of the variation in of the population makes a signi®cant contribution allele frequencies occurred in the 1992 year-class to the next generation. sample. The genotypes for all three loci in all year- Low effective population sizes should be re- classes were in Hardy±Weinberg equilibrium (chi- ¯ected in low numbers of parents contributing to square, P Ͼ 0.05). each year-class. Since allele frequency data were

The effective population size Ne was estimated available for adult samples collected on the spawn- to be approximately 20 in three of the ®ve com- ing grounds during the 1992±1994 spawning sea- parisons of consecutive year-classes (Table 2). In sons (Diaz et al. 1998), we were able to estimate

the other comparisons, the Ne estimates of 79 and that fewer than 100 parents contributed to the 268 were approximately 4-fold and 10-fold higher, 1992, 1993, and 1994 year-classes (Table 3). Thus, respectively, than the other three estimates. All although direct comparisons cannot be made be-

®ve Ne estimates are substantially lower than the tween Nb and Ne values, both numbers indicate that population size. Based on the equations, assump- only a small fraction of the adult population makes tions, and rates given in Bulak et al. (1995), we a genetic contribution to the next generation. calculated that it would take 27,727 female striped bass to produce 18.9 billion eggs, the average num- Impact of Adult Mortality on Age Structure and ber spawned over a 3-year period, as reported by on Genetic Variation Bulak et al. (1993). Assuming that 5% of age-3 In species with overlapping generations and and 50% of age-4 female striped bass participated multiple breeding years, the impact of a bottleneck in spawning, an equal sex ratio, and 100% fertility in a single year, such as the one we observed in

Downloaded By: [University of South Carolina] At: 19:52 31 May 2011 of age-3 or older males, this number of spawning 1992, will be minimal because the effective pop- females equates to approximately 100,000 sexu- ulation size is de®ned by the number of breeders ally mature striped bass of age 3 or greater. Using in a generation rather than in a single year (Hill 1972; Felsenstein 1985). However, the estimated rate of adult mortality resulting from the combined TABLE 2.ÐEstimated per generation effective popula- effects of ®shing and natural causes in the Santee± tion size (Ne, adjusted for identity by descent) based on comparisons of allele frequencies between consecutive year-classes. Other abbreviations are as follows: CI ϭ con- TABLE 3.ÐAnnual number of breeders (N ) estimated ®dence interval; Fk is the adjusted variance in allele fre- b quency between samples, S is the harmonic mean sample from comparisons of allele frequencies of adults sampled size. on the spawning grounds to those of the resulting progeny sampled in the nursery area; Fk and S are as in Table 2; Period Fk SNe (95% CI) CI ϭ con®dence interval.

1990±1991 0.002777 65.1 268 (32±609) Period F SN(95% CI) 1991±1992 0.046987 51.5 16 (2±44) k b 1992±1993 0.026976 60.5 27 (3±76) 1992 0.01150 60 65 (8±179) 1993±1994 0.009465 80.3 79 (9±218) 1993 0.01347 117 55 (7±153) 1994±1995 0.037057 85.2 20 (2±56) 1994 0.02905 71 26 (3±71) NOTES 1371

TABLE 4.ÐMean number of years to lose one allele in a one-locus, ®ve-allele system in simulated populations.

Number of breeders per yeara Population 10 20 40 100 200 Harvested 81 181 338 939 1,784 Unharvested 204 563 1,121 2,111 5,064 a Males plus females.

(Figure 1). When adult survival was ®xed at 40% (the estimate for this population when subjected to harvesting), most of the adults in the population were age 4. In contrast, when the adult survival was set at 85% (the estimated rate without har- vesting), the age-structure of the population was far more heterogeneous. These results demonstrate that high adult mortality changes the age structure of the population by minimizing the number of age-classes that de®ne a generation. The high adult mortality of the harvested pop- ulation in the simulations signi®cantly increased the rate of loss of alleles regardless of the number of breeders reproducing annually (Table 4). Sim- ilarly, regardless of the number of breeders con- tributing to each year-class, effective population sizes were approximately threefold higher in the unharvested population than in the harvested pop- ulation (Table 5). As a result, under conditions of high adult mortality, the genetic impact of a single- year bottleneck, like the one observed in 1992, would be to accelerate the loss of alleles because individuals have a reduced chance of breeding more than once. Under conditions of high adult mortality, the inclusion of a variable recruitment rate among the contributing parents in the genetic simulation model would exacerbate the loss of al- leles during poor recruitment years. However,

Downloaded By: [University of South Carolina] At: 19:52 31 May 2011 there would be little compensation during high re- FIGURE 1.ÐEffect of adult mortality on the number of age-classes in the population. (A) shows age distri- cruitment years since high adult mortality tends to bution where survival is 0.85, corresponding to the un- eliminate multiple breeding opportunities. These harvested state. Age-classes for sector are shown with percentages in parentheses. (B) shows age distribution where survival is 0.4, corresponding to the current an- TABLE 5.ÐMean effective population size per genera- nual harvest. Simulations were carried out for 300 years. tion (Ne) in simulated populations. Effective population size Ne was calculated from the rate of loss of alleles r: Ne ϭ 1/(2 Gr), where G is the generation time in years (5.25 for this population). Each value is the mean of 50 Cooper striped bass population is 60±70% (Bulak replicates. Initial populations had an equal abundance of et al. 1995). Such a high adult mortality rate re- each allele and Hardy-Weinberg genotype frequencies. duces the number of age-classes in the population, a and as a result, reduces the probability that an adult Number of breeders per year can contribute to the gene pool. In our age-struc- Population 10 20 40 100 200 tured population model, the estimated mortality Harvested 35 77 144 400 761 rates for the Santee±Cooper population had a pro- Unharvested 87 240 478 1243 2161 found effect on the age-structure of the population a Males plus females. 1372 DIAZ ET AL.

results demonstrate that in a long-lived species Harrell, R. M., X. L. Xu, and B. Ely. 1993. Evidence with overlapping generations and multiple breed- of introgressive hybridization in Chesapeake Bay ing seasons, adult mortality must be considered in Morone. Molecular Marine Biology and Biotech- nology 2:291±299. the implementation of a management program de- Hedrick, P. W., and P. S. Miller. 1992. Conservation signed to minimize loss of genetic diversity. genetics: techniques and fundamentals. Ecological Applications 2:30±46. Acknowledgments Hill, W. G. 1972. Effective size of populations with This manuscript is part of a doctoral dissertation overlapping generations. Theoretical Population Bi- by M. Diaz, who received funding support from ology 3:278±289. Jorde, P. E., and N. Ryman. 1995. Temporal allele fre- the Electric Power Research Institute through a quency change and estimation of effective popu- doctoral fellowship. Additional funding was pro- lation size in populations with overlapping gener- vided by a grant from the Cooperative Institute for ations. Genetics 139:1077±1090. Fisheries Molecular Biology (FISHTEC) (NOAA- Leclerc, G. M., M. Diaz, and B. Ely. 1996. Use of PCR- NMFS (RT/F-1)). This paper is FISHTEC contri- RFLP assays to detect genetic variation at single- bution FT00-03. copy nuclear loci in striped bass (Morone saxatilis). Molecular Marine Biology and Biotechnology 5: References 138±144. Nei, M., T. Maruyama, and R. Chakraborty. 1975. The Bulak, J. S. 1994. Factors affecting recruitment of bottleneck effect and genetic variability in natural striped bass (Morone saxatilis), in the Santee±Coo- populations. 29:1±10. per system, South Carolina. Doctoral dissertation. Pollak, E. 1983. A new method for estimating the ef- University of South Carolina, Columbia. fective population size from allele frequency chang- Bulak, J. S., J. S. Crane, D. H. Secor, and J. M. Dean. es. Genetics 104:531±548. 1997. Recruitment dynamics of striped bass in the Rago, P. J., and C. P. Goodyear. 1987. Recruitment Santee±Cooper system, South Carolina. Transac- mechanisms of striped bass and Atlantic salmon: tions of the American Fisheries Society 126:133± comparative liabilities of alternative life histories. 143. Pages 402±416 in M. J. Dadswell, R. J. Klauda, C. Bulak, J. S., N. M. Hurley, Jr., and J. S. Crane. 1993. M. Mof®tt, R. L. Saunders, R. A. Rulifson, and J. Production, mortality, and transport of striped bass E. Cooper, editors. Common strategies of anadro- eggs in Congaree and Wateree rivers, South Caro- mous and catadromous ®shes. American Fisheries lina. American Fisheries Society Symposium 14: Society, Symposium 1, Bethesda, Maryland. 29±37. Raney, E. C. 1952. The life history of striped bass, Bulak, J. S., D. S. Wethey, and M. G. White. 1995. Roccus saxatilis (Walbaum). Bulletin of the Bing- Evaluation of management options for a reproduc- ham Oceanography Collection, Yale University ing striped bass population in the Santee±Cooper 14(1):5±97. system, South Carolina. North American Journal of Richards, R. A., and P. J. Rago. 1999. A case history Fisheries Management 15:84±94. of effective ®shery management: Chesapeake Bay Caswell, H. 1989. Matrix population models: construc- striped bass. North American Journal of Fisheries tion, analysis and interpretation. Sinauer, Sunder- Management 19:356±375. land, Massachusetts. Sokal, R. R., and J. F. Rohlf. 1981. Biometry. Freeman, Caughley, G. 1994. Directions in conservation biology. New York. Journal of Animal Evolution 63:215±244.

Downloaded By: [University of South Carolina] At: 19:52 31 May 2011 Ulanowicz, R. E., and T. T. Polgar. 1980. In¯uences of Charlesworth, D., and B. Charlesworth. 1987. Inbreed- anadromous spawning behavior and optimal envi- ing depression and its evolutionary consequences. ronmental conditions upon striped bass year-class Annual Review of Ecology and Systematics 18: success. Canadian Journal of Fisheries and Aquatic 237±268. Sciences 37:143±154. Chesson, P. 1983. Coexistence of competitors in a sto- Waples, R. S. 1989. A generalized method for estimating chastic environment: the storage effect. Lecture effective population size from temporal changes in Notes in Biomathematics 52:188±198. allele frequency. Genetics 121:379±391. Diaz, M., J. Macpherson, B. Ely, and J. S. Bulak. 1998. White, M. G. I., and S. Lamprecht. 1989. Fisheries in- Striped bass population subdivision within the San- vestigations in lakes and streams, District V. South tee±Cooper system, South Carolina. Molecular Ma- Carolina Wildlife and Marine Resources Depart- rine Biology and Biotechnology 7:191±196. ment, Columbia. Ellner, S., and N. G. Hairston, Jr. 1994. Role of over- Winemiller, K. O., and K. A. Rose. 1992. Patterns of lapping generations in maintaining genetic variation life-history diversi®cation in North American ®sh- in a ¯uctuating environment. American Naturalist eries: implications for population regulation. Ca- 143:403±417. nadian Journal of Fisheries and Aquatic Sciences Felsenstein, J. 1985. Con®dence limits on phylogenies: 49:2196±2218. an approach using the bootstrap. Evolution 39:783± Wright, S. 1938. Size of population and breeding struc- 791. ture in relation to evolution. Science 87:430±431.