See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/305656790

Distortion of allele frequency distributions provides a test for recent population

bottlenecks

Article  in  Journal of Heredity · May 1998

DOI: 10.1093/jhered/89.3.238

CITATIONS READS
1,361 518

4 authors, including:

Gordon Luikart Fred Allendorf

University of Montana University of Montana
216 PUBLICATIONS   31,599 CITATIONS    262 PUBLICATIONS   31,309 CITATIONS   

SEE PROFILE SEE PROFILE

William Bruce Sherwin

UNSW Sydney
150 PUBLICATIONS   5,947 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Brucellosis View project

Population genetics study: Dispersal in and out of an urban dolphin population View project

All content following this page was uploaded by William Bruce Sherwin on 23 July 2019.

The user has requested enhancement of the downloaded file.

 Distortion of Allele Frequency Distributions
Provides a Test for Recent Population
Bottlenecks
G. Luikart, F. W. Allendorf, J.-M. Cornuet, and W. B. Sherwin

We use population genetics theory and computer simulations to demonstrate that

population bottlenecks cause a characteristic mode-shift distortion in the distri-
bution of allele frequencies at selectively neutral loci. Bottlenecks cause alleles at
low frequency (,0.1) to become less abundant than alleles in one or more inter-
mediate allele frequency class (e.g., 0.1–0.2). This distortion is transient and likely
to be detectable for only a few dozen generations. Consequently only recent bot-
tlenecks are likely to be detected by tests for distortions in distributions of allele
frequencies. We illustrate and evaluate a qualitative graphical method for detecting
a bottleneck-induced distortion of allele frequency distributions. The simple novel
method requires no information on historical population sizes or levels of genetic
variation; it requires only samples of 5 to 20 polymorphic loci and approximately
30 individuals. The graphical method often differentiates between empirical data-
sets from bottlenecked and nonbottlenecked natural populations. Computer simu-
lations show that the graphical method is likely (P . .80) to detect an allele fre-
quency distortion after a bottleneck of #20 breeding individuals when 8 to 10 poly-
morphic microsatellite loci are analyzed.

Identifying populations that have experi- Lande and Barrowclaugh 1987; Leberg
enced a severe reduction in size (i.e., a 1992; Nei et al. 1975). Therefore, if biolo-
bottleneck) is important because bottle- gists recognize that a population has been
necks can increase demographic stochas- recently bottlenecked, they may be able to
ticity, rates of inbreeding, loss of genetic minimize loss of heterozygosity and quan-
variation, and fixation of mildly deleteri- titative genetic variation.
ous alleles, thereby reducing evolutionary Identifying recently bottlenecked popu-
potential and increasing the probability of lations may also allow biologists to mini-
population extinction ( Brakefield and Sac- mize or reverse the reduction of fitness
cheri 1994; Frankel and Soule 1981; Frank- and the fixation of deleterious alleles that
ham 1995a,c; Hedrick and Miller 1992; Ji- often result from bottlenecks ( Backus et
menez et al. 1994; Lande 1988, 1994; Mills al. 1995; Hedrick 1995; Newman 1996;
and Smouse 1994; Newman 1996; Ralls et Spielman and Frankham 1992). Unfortu-
al. 1988; Vrijenhoek 1994; but see Bryant nately it is often difficult to identify re-
et al. 1986; Goodnight 1987). cently bottlenecked populations because
It is especially important to identify re- historical population sizes and levels of
cently bottlenecked populations (i.e., pop- genetic variation are seldom known.
From the Division of Biological Sciences, University of
Montana, Missoula, MT 59812 ( Luikart and Allendorf ),
ulations bottlenecked within the past few Our objective is to use empirical data
the Laboratoire de Modelisation et de Biologie Evolu- dozen generations), because such popu- and computer simulations to illustrate and
tive, I.N.R.A./U.R.L.B., Montpellier, France (Cornuet), lations may not yet have had time to evaluate a qualitative graphical method
and the School of Biological Science, University of New
South Wales, Sydney, Australia (Sherwin). G. Luikart is adapt to the problems often caused by for identifying populations that have re-
currently at the Laboratoire de Biologie des Popula- small population size and therefore may cently been bottlenecked. The graphical
tions d’Altitude, CNRS, Université Joseph Fourier, BP
53 F-38041, Genoble Cedex 9, France. We thank S.
have a high risk of extinction. The more method requires no data on historical
Forbes, D. Tallmon, and A. Taylor for comments on ear- recent a bottleneck, the greater the prob- population sizes or historical levels of ge-
lier versions of this manuscript. P. England provided ability that the deleterious effects of a bot- netic variation; it requires only measure-
helpful discussions and comments. Funding for analy-
ses of mountain sheep samples was provided by the tleneck can be avoided or minimized by ments of allele frequencies from 5 to 20
Boone and Crocked Club and the National Rifle Asso- mitigative management procedures, such polymorphic loci in a sample of approxi-
ciation. J. Hogg and S. Forbes kindly provided unpub- as habitat enhancement or introduction of mately 30 individuals. The method in-
lished data. Mountain sheep tissue samples were pro-
vided by J. McCarthy, J. Cross, and J. Firebaugh (Mon- immigrants. Recently bottlenecked popu- volves comparing the distribution of allele
tana Department of Fish Wildlife and Parks), O. Dyre, lations are likely to have lost rare alleles, frequencies observed in a population sus-
H. Langin, A. Peat, A. Wolterson, and B. Lincoln ( British
Columbia Ministry of Environment, Wildlife Branch),
but may still contain substantial hetero- pected to have been bottlenecked to the
Jon Jorgenson (Alberta Fish and Wildlife Service), M. zygosity and quantitative genetic variation distribution expected in a nonbottle-
Miller (Colorado Department of Fish and Wildlife), J. which are lost more slowly than allelic necked population.
Hogg, and Marco Festa-Bianchet. Address correspon-
dence to Gordon Luikart at the address above or e- variation, and which influence fitness in We define a ‘‘nonbottlenecked’’ popula-
mail: [email protected]. current environments more than allelic tion as one that is thought to not have
q 1998 The American Genetic Association 89:238–247 variation (Allendorf 1986; Denniston 1978; been recently bottlenecked and is there-

238
 The expected distribution of allele fre-
quencies in a recently bottlenecked pop-
ulation has not been thoroughly studied,
although numerous authors have reported
that alleles at low frequency are expected
to be lost rapidly during a bottleneck (Al-
lendorf 1986; Denniston 1978; Nei et al.
1976; Maruyama and Fuerst 1985; Watter-
son 1984). Furthermore, no one has de-
veloped a method for identifying recently
bottlenecked populations based on a dis-
tortion of allele frequency distributions
( but see the quantitative method of Cor-
nuet and Luikart, 1996). We present a qual-
itative graphical method for identifying
bottlenecked populations from distribu-
tions of allele frequencies and evaluate the
performance of the method using comput-
er simulations and 90 datasets from natu-
ral populations.
Three questions we address here are as
follows: (1) How are bottlenecks expected
to distort the distribution of allele fre-
quencies at neutral loci? (2) Is the expect-
ed distortion usually apparent in empirical
datasets from bottlenecked natural popu-
lations? (3) How small a bottleneck is re-
quired to cause a distortion in allele fre-
quencies that is likely to be detectable
(power . 0.80) by the graphical method
when using approximately 10 microsatel-
lite loci? Questions 1 and 3 are addressed
using Monte Carlo computer simulations.
If the graphical method is to be useful
for detecting bottlenecks, both the simu-
Figure 1. (a) Distribution of allele frequencies expected for loci evolving under the infinite allele model of mu-
lations and empirical datasets from known
tation ( IAM) in a nonbottlenecked population at mutation-drift equilibrium ( Nei et al. 1976). Black bars represent bottlenecked natural populations should
the proportion of alleles expected in each of 10 allele frequency classes. The mean heterozygosity expected for a reveal a characteristic distortion in the
random sample of loci having the illustrated distribution is 0.40. (b) Distribution of allele frequencies expected
(for a sample of 50 loci) in a population bottlenecked to eight breeding individuals (Ne 5 8) for four generations. distribution of allele frequencies. Further-
Open bars represent the number of alleles expected in each of 10 allele frequency classes. Expected numbers of more, datasets from nonbottlenecked nat-
alleles were calculated as the mean of 500 replicate bottleneck simulations. ural populations should not reveal this
distortion, but rather should have a large
proportion of alleles at low frequency, as
fore likely to be near mutation-drift equi- cies will vary with the mutation rate and expected in populations near mutation-
librium. For selectively neutral loci, allele the model of mutation at a given locus. For drift equilibrium.
number and frequency distribution in a example, the distribution of allele frequen-
natural population results from a dynamic cies expected for loci evolving under the
Methods
equilibrium between mutation and genetic stepwise mutation model (SMM; Ohta and
drift. This ‘‘mutation-drift’’ equilibrium will Kimura 1973) may have a slightly lower The Graphical Method
be approximately reached if the effective proportion of alleles at low frequency than The graphical method consists of group-
population size (Ne) remains stationary for the distribution expected for the infinite ing alleles from a sample of many poly-
(4–10 multiplied by Ne) generations ( Nei allele model of mutation ( IAM; Kimura and morphic loci (at least five loci) into each
and Li 1976). Crow 1964). However, alleles at low fre- of 10 allele frequency classes and then
The expected distribution of allele fre- quency (,0.1) are always expected to be plotting a frequency histogram. The 10 al-
quencies for neutral loci in a nonbottle- more abundant than alleles at intermedi- lele frequency classes are 0.001–0.100,
necked population has been established. ate frequency, regardless of the mutation 0.101–0.200, 0.201–0.300, etc. For the fol-
Nonbottlenecked populations that are rate and model ( Nei et al. 1976). Low fre- lowing discussions we define low- and
near mutation-drift equilibrium for selec- quency alleles are typically far more abun- high-frequency allele classes as 0–0.100
tively neutral loci are expected to have a dant than alleles at intermediate frequen- and 0.901–1.00, respectively. We define in-
large proportion of alleles at low frequen- cy in allozyme datasets from nonbottle- termediate frequency classes as those
cy ( Figure 1a). The expected proportion of necked natural populations (Chakraborty eight classes between 0.101 and 0.900.
alleles at low and intermediate frequen- et al. 1980). This classification system is arbitrary but

Luikart et al • Testing for Bottlenecks 239

useful for the qualitative graphical assess- distribution of allele frequencies conform Data From Natural Populations
ment of allele frequency distributions. We well to the distribution expected for loci Both the allozyme and microsatellite da-
group alleles into only 10 allele frequency at mutation-drift equilibrium evolving un- tasets from recently bottlenecked natural
classes because if more than 10 classes der the stepwise mutation model ( Figure populations often revealed a characteris-
are used, a meaningful assessment of the 2e,g; Luikart G and Cornuet J-M, unpub- tic mode-shift distortion in the distribu-
distribution of allele frequencies would of- lished data). Second, random samples of tion of allele frequencies. That is, these da-
ten not be possible, because too few al- 30 individuals were generated from the tasets often had fewer alleles at low fre-
leles exist in most empirical datasets, es- last generation of each bottleneck repli- quency than in one or more intermediate
pecially datasets from bottlenecked pop- cate to simulate the process of estimating allele frequency class. For example, the
ulations. The graphical method concludes allele frequencies from a sample, as is bottlenecked population of Epping wom-
that a population has been recently bottle- done in empirical studies of natural pop- bats had only one allele at low frequency,
necked if fewer alleles are found in the low ulations. but nine alleles between the frequencies
frequency class than in one or more inter- 0.4 and 0.5 ( Figure 2c). In total, 13 of 19
mediate frequency classes (see below). Empirical Datasets datasets from bottlenecked natural popu-
We analyzed 9 microsatellite datasets and lations revealed a mode-shift distortion
Simulations 10 allozyme datasets from natural popu- (appendix 1; Figure 2a–l, open bars).
We conducted Monte Carlo computer sim- lations thought to have been recently bot- Allozyme datasets from nonbottle-
ulations to determine the expected distri- tlenecked and isolated based on evidence necked natural populations usually had a
bution of allele frequencies in a recently from demographic, biogeographic, and/or large proportion of alleles at low frequen-
bottlenecked population. To determine independent molecular data (appendix 1). cy, and thus had an allele frequency dis-
the expected distribution, we calculated Some datasets were selected as represen- tribution with a mode in the low frequency
and graphed the mean number of alleles tatives of numerous published datasets class ( Figure 2i–l, black bars; see also dis-
in each of 10 allele frequency classes after from a given species (e.g., common myna tributions from 138 populations in Chak-
500 bottleneck simulation replicates. In birds; Baker and Moeed 1987; Fleisher et
raborty et al. 1980). Only one of the 46
our simulation model, the genotypes of in- al. 1991).
allozyme datasets from nonbottlenecked
dividuals in the initial generation of each We analyzed 25 microsatellite datasets
populations revealed a mode-shift distor-
bottleneck replicate are generated by ran- and 46 allozyme datasets from natural
tion in which alleles at intermediate fre-
domly sampling from an allele frequency populations that are not known to have
quencies were more abundant than alleles
distribution of a nonbottlenecked popula- been recently bottlenecked (appendixes 2
at low frequency (Sun River population of
tion (e.g., the distribution in Figure 1a). and 3). These datasets were chosen be-
mountain sheep; appendix 3). Microsatel-
Genotypes of each subsequent generation cause they include many individuals (at
lite datasets from nonbottlenecked popu-
are generated by simulating Mendelian in- least 20, except for the Brookfield wombat
lations also had a large proportion of rare
heritance and random mating between population), many polymorphic loci (at
least 5), and populations from relatively alleles ( Figure 2a–g, black bars; see also
males and females (sex ratio 1:1). Loci are
assumed to be selectively neutral. The undisturbed habitats, for example, Pacific Allen et al. 1996; England et al., in press;
model does not incorporate new muta- salmon (Oncorhynchus sp.) from remote and Roy et al. 1995). Only one (M. domes-
tions; however, the number of new muta- areas without hatchery influences, bears tics-3) of the 25 microsatellite datasets
tions over a small number of generations (Ursus sp.), mountain sheep (Ovis cana- from nonbottlenecked populations re-
(i.e., 2 multiplied by Ne generations) will densis), and wolves (C. lupus) from Alaska vealed a mode-shifted distribution of allele
be negligible (Maruyama and Fuerst 1985). and Canada. However, it is difficult to be frequencies (appendix 2). Analyses of em-
certain that these populations, or any nat- pirical datasets suggest that the graphical
Power Analysis ural populations, have not experienced a method is not likely to incorrectly identify
Computer simulations were also used to recent bottleneck because information sel- a nonbottlenecked population as a recent-
determine the bottleneck size required to dom exists on a population’s historical ly bottlenecked population.
cause a distortion in allele frequencies size and Ne.
that is likely to be detectable (power .
Simulations and Power Analysis
0.80) when sampling 8–10 polymorphic mi-
Results To further determine if the qualitative
crosatellite loci and 30 individuals. These
simulations were conducted in the same Bottleneck-Induced Distortions graphical method is likely to detect mode-
way as those described above with two Our simulations showed that the defining shifted distributions in nonbottlenecked
exceptions meant to make the simulations characteristic of the distribution of allele populations, we used computer simula-
more realistic. First, the initial prebottle- frequencies expected in a recently bottle- tions to generate samples of 20 to 30 in-
neck allele frequencies were obtained necked population is a mode-shifted dis- dividuals from allele frequency data pub-
from actual data from 8 and 10 microsat- tribution, that is, a distribution with fewer lished from a nonbottlenecked population
ellite loci from nonbottlenecked popula- alleles in the low frequency class (,0.1) of brown bears (Western Brooks Range;
tions of brown bears and wolves, respec- than in one or more intermediate frequen- Craighead 1994). In ‘‘computer samples’’
tively (Western Brooks Range brown bear cy classes (e.g., 0.1–0.2). We characterize of 20, 25, and 30 individuals from the bear
population, Ursus arctos, Craighead 1994; this bottleneck-induced distortion of the dataset, the proportion of 500 samples re-
and Hinton wolf population, Canis lupus, distribution of allele frequencies as a vealing a mode-shifted distribution were
Forbes and Boyd 1996). These datasets mode shift, because bottlenecks shift the 0.05, 0.010, and 0.006, respectively.
were chosen because they are among the mode from low frequency to an interme- Our computer simulation power analy-
largest published (appendix 2), and their diate frequency (compare Figure 1a,b). sis suggested that bottlenecks of size

240 The Journal of Heredity 1998:89(3)

Figure 2. (a–l ) Allele frequency distributions from nonbottlenecked ( black bars) and bottlenecked natural populations (open bars). Figures h–l are from allozyme data; the
others are from microsatellites. Each figure (except g and h) shows a distribution from both a nonbottlenecked and a bottlenecked population from the same species or two
related species. The number in parentheses is the sample size of individuals.

ranging up to 20 individuals have a rea- these simulations, the prebottleneck allele er than those from the bear data were
sonably high probability of being detected frequencies for the eight loci were ob- achieved using data from 10 microsatellite
(0.78) using the qualitative graphical test tained from the Western Brooks Range loci from the Hinton population of wolves
for mode shift, and samples of 8 microsat- brown bears ( Figure 2g, black bars; Craig- ( Figure 2e, black bars; Forbes and Boyd
ellite loci and 30 individuals ( Figure 3). In head 1994). Power estimates slightly high- 1996).

Luikart et al • Testing for Bottlenecks 241

Discussion
Our simulations have shown that popula-
tion bottlenecks are expected to cause a
mode-shift distortion in the distribution of
allele frequencies at neutral loci such that
alleles in the low frequency class (,0.1)
become less abundant than alleles in one
or more intermediate frequency classes
(e.g., 0.1–0.2). We have illustrated a quali-
tative graphical method of analyzing allele
frequency data that can help detect mode-
shift distortions and thereby help identify
recently bottlenecked populations.
The main advantage of this method is
that no reference population or data on
historical levels of genetic variation is
needed to determine if a population has
been recently bottlenecked. For example,
based only on contemporary allele fre-
quency distributions and without compar-
ative data on levels of genetic variation in
nonbottlenecked reference populations,
one can conclude that the Epping Forest
wombats have recently suffered a genetic
bottleneck, loss of genetic variation (i.e.,
at least rare alleles), and possibly the fix-
ation of deleterious alleles ( Figure 2c,
open bars).
Even if reference populations are avail-
able, researchers testing for bottlenecks
should analyze distributions of allele fre-
quencies along with traditional indices of
genetic variation such as heterozygosity
(H), mean number of alleles per locus (A),
and proportion of polymorphic loci (P).
Traditional indices of genetic variation
can remain high after a bottleneck and
thus fail to detect a bottleneck. For ex-
ample, in the bottlenecked population of
brown bears from the western Carpathian
Mountains, the mean number of alleles per
locus is three (appendix 1, population 18;
Hartle and Hell 1994). This A is high in
comparison to allozyme data from other
large mammals, including nonbottle-
necked populations of brown bears from
North America (appendix 3; Figure 2l,
black bars). Although A is still high in the
Figure 3. Distributions of allele frequencies in a sample of 30 individuals and 8 microsatellite loci from ‘‘computer
western Carpathian bears, the distribution populations’’ bottlenecked to 4 individuals and 20 individuals for each of four different generation times. Open
of allele frequencies is mode-shift distort- bars represent the mean number of alleles in each allele frequency class, calculated from 500 bottleneck simulation
replicates. Error bars show the approximate 20–80 percentile range of the number of alleles found in each fre-
ed as expected in this recently bottle- quency class over 500 simulations. Power is the proportion of 500 bottleneck replicates that revealed a mode-shift
necked population ( Figure 2l, open bars). distribution of allele frequencies in the postbottleneck sample of 30 individuals.
Another advantage of the graphical
method is that it is most likely to detect
the type of bottlenecks that are most like- the more likely it will be detectable by mographic problems often caused by bot-
ly to be harmful, that is, recent small bot- tests for distortions of allele frequency tlenecks.
tlenecks. The smaller the bottleneck, the distributions ( Figure 3, bottleneck size 4 We define ‘‘recent’’ as within the past
more likely it will increase the frequency versus 20). Recent bottlenecks are impor- several dozen generations. This definition
of deleterious alleles and cause inbreeding tant to detect because recently bottle- is based on the time during which a dis-
depression and loss of genetic variation. necked populations are unlikely to have tortion of allele frequency distributions is
Fortunately the smaller the bottleneck, had time to adapt to the genetic and de- likely to be detectable. A bottleneck is

242 The Journal of Heredity 1998:89(3)

likely to be detectable for only 40 to 80 frequency distributions can help detect Bottlenecked Populations
generations, assuming that the maximum these ‘‘cryptic’’ genetic bottlenecks in Simulations and empirical datasets both
bottleneck size likely to be detectable is populations thought to have a large Ne show that the qualitative graphical meth-
approximately Ne 5 20 ( Figure 3), and that based on demographic data. od often identifies recently bottlenecked
bottlenecks are detectable for only ap- populations. Moreover, the datasets with
proximately 2Ne to 4Ne generations until the largest proportion of alleles at inter-
Assumptions
genetic drift and new mutations begin to mediate frequency are from populations in
reestablish mutation-drift equilibrium When analyzing allele frequency distribu- which bottlenecks have been the most se-
(Cornuet and Luikart, 1996; Maruyama and tions to test for bottlenecks, it may be nec- vere, the most well-documented, and for
Fuerst 1985; Nei and Li 1976). essary to assume that the test population which the most polymorphic loci were an-
Houlden et al. (1996) and Huettle et al. is random mating, has no substructure, alyzed (see the Epping Forest population,
(1980) have reported ‘‘mode shifts’’ has no recent immigration, loci are neu- Figure 2c, open bars). We note that an un-
(called ‘‘bimodal’’ distributions by these tral, and that sampling is representative of published microsatellite dataset from the
authors) in allele frequency distributions the population. These assumptions and severely bottlenecked Bison Range popu-
from bottlenecked natural populations of the consequences of violating the assump- lation of mountain sheep also shows a
Mediterranean fruit flies (Ceratitis capita- tions have been discussed by Cornuet and strongly mode-shifted distribution of allele
ta) and Australian koalas (Phascolarctos ci- Luikart (1996). It is worth reiterating here frequencies similar to that of the Epping
nereus), respectively. These authors hy- that tests for Hardy–Weinberg proportions wombats (appendix 1; Hogg J and Forbes
pothesized that a mode shift may typically may detect violations of the above as- S, personal communication).
result from a bottleneck and that the sumptions. Loci not in Hardy–Weinberg The six bottlenecked populations that
mode should shift to an intermediate al- proportions should be excluded or used did not have a mode-shifted distribution
lele frequency between 0.4 and 0.6 or 0.3 only with caution. We tested for bottle- (appendix 1) may not have a distorted dis-
and 0.7. However, our simulations suggest necks with and without loci deviating from tribution for the following reasons: (1) the
that the mode at intermediate frequency Hardy–Weinberg proportions in the data- bottleneck was not recent or small enough
is usually expected to occur between the sets from which genotype frequency data to be detectable, (2) not enough polymor-
frequencies 0.1 and 0.2 ( Figure 1b and 3). was available, but it made no difference in phic loci and/or individuals were sampled
Furthermore, the mode at intermediate the test results. to have sufficient power for detecting the
frequency is only expected to occur be- Violating the assumption that loci are bottleneck, (3) the individuals sampled
tween the frequencies 0.2 and 0.5 when selectively neutral could cause nonbottle- were not representative of the bottle-
the bottleneck is both smaller than ap- necked populations to appear to have necked population, (4) a demographic
proximately eight breeding individuals been recently bottlenecked, if the type of bottleneck occurred but not a genetic bot-
and persists for several generations (data selection is heterozygote advantage or tleneck (i.e., Ne k N-census), and (5) the
not shown). balancing selection. Strong balancing se- bottlenecked population is not completely
The degree of mode shift is expected to lection could maintain alleles at interme- isolated and contains genes from immi-
increase with bottleneck severity. Thus diate frequencies and thereby reduce the grants that have obscured the genetic ef-
the degree of mode shift may be useful for proportion of alleles at low frequency and fects of the bottleneck.
inferring the approximate severity of a re- generate a mode-shift distortion. Re- The bottlenecked populations of Nepal
cent bottleneck. However, reliable infer- searchers have reported evidence of bal- rhinoceros and Sidney myna birds may
ences about the approximate severity of a ancing selection at allozyme loci in several not have a mode-shifted distribution be-
bottleneck may require many polymor- of the nonbottlenecked populations ana- cause the bottlenecks were not small (i.e.,
phic loci (.10) because there is substan- lyzed in this study [American oysters 60–80 individuals and 100 individuals, re-
tial variability among the different distri- (Crassostrea virginica), Karl and Avise spectively). The bottlenecked population
butions of allele frequencies that can ac- 1992; and Atlantic cod (Gadus morhua), of Illinois-WHIT tree sparrows may not
tually result from a bottleneck (see error Pogson et al. 1995]. However, the graphi- have a mode-shifted distribution because
bars in Figure 3). cal method detected no evidence of mode- the bottleneck occurred long ago (i.e.,
It is important to note that genetic bot- shift distortion in any of the oyster or cod more than 100 generations ago; appendix
tlenecks of less than 20 individuals are not populations (appendix 3). 1).
unrealistically small for many wild and If analyses of allele frequency distribu-
captive populations because the geneti- tions fail to detect a mode-shifted distri-
Performance of Method
cally effective size (Ne) of a population is bution of allele frequencies, one should
often only 10–20% of a population’s cen- It is difficult to evaluate the performance not conclude that a population has not
sus size, and occasionally a much smaller of methods for detecting bottlenecks us- been bottlenecked; one can only conclude
percentage ( Briscoe et al. 1992; Frankham ing data from natural populations because that a bottleneck is not likely to have oc-
1995b). Consequently even populations few datasets with many polymorphic loci curred in the recent past. Although a
with a large census size in national parks, exist from bottlenecked populations that mode-shifted distribution is likely to be
wildlife reserves, or fish hatcheries could are isolated and have a well-documented detectable (power 5 0.78) after a bottle-
experience a severe genetic bottleneck in history of bottleneck size and duration. neck of less than approximately 20 breed-
the absence of a demographic bottleneck. Nevertheless, it is important to evaluate ing individuals, it will not be detected ap-
For example, if only a few males mate with the behavior of genetic markers and the proximately 22% of the time when 8 to 10
all the females in a large population, a ge- performance of bottleneck tests in the few polymorphic microsatellite loci are
netic bottleneck can occur without a de- appropriate datasets available from natu- screened. Furthermore, it may take 5 to 10
mographic bottleneck. Analyses of allele ral populations. generations for bottlenecks of 20 breeders

Luikart et al • Testing for Bottlenecks 243

to generate a mode-shifted distribution enly identifying a nonbottlenecked popu- netic variation. Even when reference pop-
( Figure 3e,f ). Consequently a mode-shift- lation as a recently bottlenecked popula- ulations are available for comparison, re-
ed distribution occasionally may not be tion. searchers testing for recent bottlenecks or
detected even though a population has loss of genetic variation should analyze
been recently bottlenecked. the distribution of allele frequencies in ad-
Conclusions
dition to traditional indices of genetic vari-
Nonbottlenecked Populations Empirical data and computer simulations ation. Analyses of allele frequency distri-
Both the simulations and empirical data- show that population bottlenecks cause a butions may identify bottlenecked popu-
sets suggested that the qualitative graph- characteristic mode-shift distortion in the lations when traditional indices of genetic
ical method is not likely to incorrectly distribution of allele frequencies at selec- variation do not.
identify a nonbottlenecked population as tively neutral loci. This distortion is often Analyses of allele frequency distribu-
a recently bottlenecked population. Only detectable by the qualitative graphical tions can help detect ‘‘cryptic’’ genetic
2 of the 71 datasets from ‘‘nonbottle- method. Eight to ten microsatellite loci bottlenecks in which Ne has been severely
necked’’ populations revealed a mode- provide approximately an 80% probability reduced in the absence of a reduction of
shifted distribution of allele frequencies (power) of detecting a recent historical a population’s census size. The graphical
that is characteristic of bottlenecked pop- bottleneck of fewer than 20 breeding in- method provides a useful tool for identi-
ulations. The simulations showed that dividuals. The graphical method is unlike- fying recently bottlenecked populations
sample sizes of 20 to 30 individuals are un- ly to mistakenly identify a nonbottle- that may be of concern for conservation
likely to suggest that a nonbottlenecked necked population as a bottlenecked pop- biology.
population has been recently bottle- ulation if at least 30 individuals are sam-
necked. Nonetheless, small samples are pled.
likely to miss alleles at low frequency (Sjo- The graphical method for detecting dis- References
gren and Wyoni 1994) and thus cause al- tortions in the distribution of allele fre- Allen PJW, Pomeroy PP, and Twiss SD, 1996. Microsa-
lele frequency distributions to resemble quencies can identify recently bottle- tellite variation in grey seals (Halichoerus grypus)
shows evidence of genetic differentiation between two
those from a bottlenecked population. necked populations without use of refer- British breeding colonies. Mol Ecol 4:653–662.
Consequently we recommend sampling ence populations or information about Allendorf FW, 1986. Genetic drift and the loss of alleles
more than 30 individuals to avoid mistak- historical population size or levels of ge- versus heterozygosity. Zoo Biol 5:181–190.
Backus VL, Bryant EH, Hughes CR, and Meffert L, 1995.
Effect of migration or inbreeding followed by selection
Appendix 1. Census sizes and sample sizes of loci and individuals in 19 datasets from bottlenecked on low-founder-number populations: implications for
populations captive breeding programs. Conserv Biol 9:1216–1224.
Baker AJ and Moeed A, 1987. Rapid genetic differenti-
Mean ation and founder effect in colonizing populations of
Number of number of common mynas (Acridotheres tristis). Evolution 41:525–
polymorphic individuals Historical Source of 538.
Species and loci/alleles sampled population genetic and
populations sampled per locus census size/date census data Bancroft DR, Pemberton JM, Albon SA, Robertson A,
Maccoll ADC, Smith JA, Stevenson IR, and Clutton-
Mountain sheep (Ovis canadensis) Brock TH, 1995. Molecular genetic variation and indi-
vidual survival during population crashes of an unman-
1. Wildhorse Island, 6/22 26 8 founders/1947 (trans- Microsatellites
aged ungulate population. Phil Trans R Soc Lond Biol
Montana planted from Sun River, Luikart G, u.d.; Matthews
Sci 347:235–358.
Montana, although 2 of the 1973
8 may be from a different Billington HL, 1991. Effect of population size on genetic
source), 90/1954, 130/ variation in a dioecious conifer. Conservation Biology
1964, 309/1971, 205/1972, 5:115–119.
200/1994 Brakefield PM and Saccheri IJ, 1994. Guidelines in con-
2. Tarryall, Colorado 6/25 24 900/pre-1952, 44/1953, 100/
Microsatellites servation genetics and their use of the population cage
1970, 150/1981, 200/1988 Luikart G, u.d.; Buchner experiments with butterflies to investigate the effects
1960; Bailey J, personal of genetic drift and inbreeding. In: Conservation Genet-
communication ics ( Loeschcke V, Tomiuk J, and Jain SK, eds). Berlin:
3. Bison Range, Mon- 7/20 23 12 founders/1921 (trans- Microsatellites Forbes S, Birkhauser; 165–179.
tana U.S.A. planted from Banff near Hogg J, u.d.; National Bi- Briscoe DA, Malpica JM, Robertson A, Smith GJ, Frank-
Sheep River, Canada), son Range census records ham R, Banks RG, Barker JSF, 1992. Rapid loss of ge-
90/1929, 8/1939, 12/1950, netic variation in large captive populations of Drosoph-
50/1984 ila flies: implications for management of captive pop-
Wombats ( Epping 5 Lasiorinyus krefftii) ulations. Conserv Biol 6:416–425.
4. Epping Forest, 9/31 43 20–30/1980, 70/1994 Microsatellites Bryant EH, McCommas SA, and Combs LM, 1986. The
Queensland, Aus- Taylor 1995 effect of an experimental bottleneck upon quantitative
tralia genetic variation in the housefly. Genetics 114:1191–
Red fox (Vulpes vulpes) 1211.
5. San Remo, Victo- 7/28 22 5 founders/1870, however, Microsatellites Buchner JK, 1960. The bighorn sheep in the United
ria, Australia other undocumented in- Lade et al. 1996 States—its past, present and future. Wildlife Mono-
troductions of V. vulpes graphs, no. 4:1191–1211.
may have occurred. The Buroker NE, 1983. Population genetics of the American
population has grown and oyster Crassostrea virginica along the Atlantic coast
spread across much of and the Gulf of Mexico. Mar Biol 75:99–112.
Australia
Chakraborty R, Fuerst PA, and Nei M, 1980. Statistical
6. Phillip Island, Vic- 7/17 23 Unknown number of foun- See 5
studies on protein polymorphism in natural popula-
toria, Australia ders from the Australian
tions. III. Distribution of allele frequencies and the num-
mainland population de-
ber of alleles per locus. Genetics 94:1039–1063.
scribed in 5 above
Craighead LF, 1994. Conservation genetics of grizzly

244 The Journal of Heredity 1998:89(3)

 Appendix 1. Continued Appendix 2. Twenty-five microsatellite datasets
from populations not known to have been
Mean recently bottlenecked
Number of number of
polymorphic individuals Historical Source of Aver-
Species and loci/alleles sampled population genetic and Num- age
populations sampled per locus census size/date census data ber of number
poly- of indi-
Koalas (Phascolarctos cinereus) morphic viduals
7. French Island, 4/16 43 As few as 2–3/around 1900 Microsatellites loci sampled
Australia Houlden et al. 1996 Species and ana- per Refer-
population lyzed locus ence
Wolves (Canis lupus)
8. Mexican-certified 10/26 21 4 founders/1984 (from a Microsatellites Mountain sheep
Mexico remnant population of the Garcia-Moreno et al. 1996 Sun River 6 32 a
endangered Mexican wolf ) Vaseux Lake 6 25 a
Soay sheep (Ovis aries) Sheep River 8 50 b
9. Herta Island, Scot- 6/23 662 107 founders/1932 (intro- Microsatellites Wolves
land duced from Soay Island), Bancroft et al. 1995 Hinton 10 32 c
size fluctuates between N.W. Territory 10 21 d
600–1,500 every 3–5 years
Coyotes (Canis latrans)
10. Herta Island, Scot- 5/16 973 Same as in 9 Allozymes
land Bancroft et al. 1995 California 10 22 d
Common myna bird (Acridotheres tristis) Brown bears
11. Oahu, Hawaii 7/28 38 About 100 founders/1982 Allozymes W. Brooks Range 8 152 e
(introduced from India), Fleischer et al. 1991 Polar bears (Ursus maritimus)
now abundant on all Ha-
W. Hudson Bay 8 30 f
waiian islands
Davis Strait 8 26 f
12. Sidney, New South 9/32 42 About 100 founders/1862 See 11 N. Beaufort 8 30 f
Wales, Australia (introduced from India), S. Beaufort 8 22 f
now widely abundant in
Sydney and surrounding Wombats (Lasiorhinus krefftii)
cities Brookfield 14 16 g
Land snails (Thebia pisana) Koalas
13. City, Australia 5/23 28a Unknown number intro- Allozymes Gold Coast 6 27 h
duced in the 1890s Johnson 1988
Field mice (Mus musculus and M. domesticus)
14. Cott, Australia 6/25 28a Same as in 13 See 13
M. domesticus-3 6 24 k
Galaxid fish (Galaxias truttaceus) M. musculus-7.92 5 24 k
15. Isabella Lagoon, 5/28 40 Bottleneck inferred from Allozymes Chimpanzees (Pan troglodytes)
Tasmania mitochondrial DNA data, Ovenden and White 1990
Gombe Kasakela 8 36 q
population became land-
locked 3000–7000 years Humans (Homo sapiens)
ago Sardinia 10 46 j
Eurasian tree sparrow (Passer montanus) Egypt 10 46 j
Kachari 6 40 l
16. Illinois-WOOD 8/23 24 20/1870 (introduced from Allozymes
New Guinea 6 39 l
Europe) St. Louis and Barlow 1998
17. Illinois-WHIT 9/25 52 Same individuals as in 16 See 16 Gray seals (Halichoerus grypus)
Isle of May
Brown bear (Ursus arctos)
(adults) 8 35 m
18. Western Carpathi- 5/15 57 40/1932, 700/1995 Isolated Allozymes
North Rona
ans, Romania from other populations Hartle and Hell 1994
(adults) 8 176 m
since the late 1800s
Fruit flies (Drosophila melanogaster)
Rhinoceros (Rhinoceros unicornus)
Tyrell 8 68 h
19. Chitwan Valley, 10/20 22 .1,000/1950, 60–80/1962, Allozymes
Nepal .251/1988 Dinerstein and McCracken Bumble bees (Bombus terrestris)
1990 Corsica 7 21 i
Sardinia 7 22 i
All data sets had a mode-shifted distribution except 5, 9, 12, 14, 17, and 19. u.d. 5 unpublished data; n.r. 5 not reported.
a
Approximately 28, exact number not reported. Only one dataset revealed a mode-shifted distribution
(field mice, M. domesticus-3 population). a 5 Luikart G,
unpublished data, available upon request; b 5 Hogg J
bears (PhD dissertation). Bozeman, Montana: Montana diversity: implications for endangered species. In: En- and Forbes S, personal communication; c 5 Forbes and
State University. dangered birds: management techniques for preserving Boyd 1996; d 5 Roy et al. 1994; e 5 Craighead 1994; f
threatened species ( Temple SA, ed). Madison, Wiscon- 5 Paetkau et al. 1995; g 5 Taylor et al. 1994; h 5 En-
Cornuet J-M and Luikart G, 1996. Description and pow-
sin: University of Wisconsin Press; 281–289. gland et al., in press; i 5 Estoup et al. 1996; j 5 Di
er analysis of two tests for detecting recent population
Rienzo et al. 1994; k 5 Dallas et al. 1995; l 5 Deka et
bottlenecks from allele frequency data. Genetics 144: Dinerstein E and McCracken GF, 1990. Endangered one- al. 1991; m 5 Allen et al. 1996; n 5 Waits et al. 1996; q
2001–2014. horned Rhinoceros carry high levels of genetic varia- 5 Morin et al. 1994.
Dallas JF, 1992. Estimation of microsatellite mutation tion. Conserv Biol 4:417–422.
rates in recombinant inbred strains of mouse. Mammal Di Rienzo A, Peterson AC, Garza JC, Valdes AM, Slatkin
Genome 5:32–38. M, and Freimer NB, 1994. Mutational processes of sim-
Dallas JF, Dod B, Boursot P, Prager EM, and Bonhomme ple sequence repeat loci in human populations. Proc
F, 1995. Population subdivision and gene flow in Danish Natl Acad Sci USA 91:3166–3170.
house mice. Mol Ecol 4:311–320. England PR, Briscoe DA, and Frankham R, in press. Mi-
Deka R, Chakraborty R, and Ferrell RE, 1991. A popu- crosatellite polymorphisms in a wild population of Dro-
lation genetic study of six VNTR loci in three ethnically sophila melanogaster. Genet Res Camb.
defined populations. Genomics 11:83–92.
Estoup A, Solignac M, Cornuet JM, and Scholl A, 1996.
Denniston C, 1978. Small population size and genetic Genetic differentiation of continental and island popu-

Luikart et al • Testing for Bottlenecks 245

Appendix 3. Forty-six allozyme datasets from populations thought not to have been recently bottlenecked lations of Bombus terrestris ( Hymenoptera, Apidae) in
Europe. Mol Ecol 5:19–31.
Mean number Fleischer RC, Williams RN, and Baker AJ, 1991. Genetic
Number of of individuals variation within and among populations of the com-
Species and polymorphic analyzed mon myna in Hawaii. J Hered 82:205–208.
population loci analyzed per locus Reference
Forbes SH and Boyd DK, 1996. Genetic variation of nat-
Mountain sheep urally colonizing wolves in the central Rocky Moun-
Sun River 5 29 aa tains. Conserv Biol 10:1082–1090.
Wolves Frankel OH and Soule ME, 1981. Conservation and evo-
Tuktoyaktuk 5 93 bb lution. Cambridge: Cambridge University Press.
Brown bear Frankham R, 1995a. Conservation genetics. Annu Rev
Western Brooks Range 5 42 cc Genet 29:305–327.
Common myna birds Frankham R, 1995b. Effective population size/adult
Bophal 15 40 dd population size ratios in wildlife: a review. Genet Res
Loknow 16 40 dd Camb 66:95–106.
Bhubaneswar 15 36 dd
Frankham R, 1995c. Inbreeding depression: a threshold
Eurasian tree sparrows effect. Conserv Biol 9:792–799.
Germany 12 30 ee
Sweden 25 ee Garcia-Moreno J, Matoco MC, Roy MS, Geffen E, and
Wayne RK, 1996. Relationships and genetic purity of
Minke whales (Balenoptera acutorostrata) the endangered Mexican wolf based on analysis of mi-
MKC 9 45 ff crosatellite loci. Conserv Biol 10:376–389.
MBC 12 190 ff
Goodnight CJ, 1987. On the effect of founder events on
Bryd whales (Balenoptera edeni)
epistatic genetic variance. Evolution 41:80–91.
BMA 6 100 ff
BJA 6 118 ff Hartle GB and Hell P, 1994. Maintenance of high levels
Galaxid fish of allelic variation in spite of a severe bottleneck in
Allens Creek 12 40 gg population size: the brown bear in western Carpathi-
Fortesue Lagoon 10 42 gg ans. Biodiv Conserv 3:546–554.
Pink salmon (Oncorhynchus gorbuscha) Hedrick PW, 1995. Gene flow and genetic restoration:
Ivashka 21 75 hh the Florida panther as a case study. Conserv Biol 5:
Kik-chik 16 78 hh 996–1007.
Pymta 23 79 hh
Arman 21 79 hh Hedrick PW and Miller PS, 1992. Conservation genetics:
techniques and fundamentals. Ecol Applic 2:30–46.
Chum salmon (O. keta)
Anadyr 26 100 ii Houlden BA, England PR, Taylor A, Greville WD, and
Ola 29 80 ii Sherwin WB, 1996. Low genetic variability of the koala
Kamchatka-b 21 39 ii (Phascolarctos cinereus) in southeastern Australia fol-
lowing a severe population bottleneck. Mol Ecol 5:269–
Sockeye salmon (O. nerka)
281.
Skilak 13 50 jj
Yenta 7 50 jj Huettle MD, Fuerst PA, Maruyama T, and Chakraborty
Dalnee 89-90 7 250 kk R, 1980. Genetic effects of multiple population bottle-
Atlantic cod (Gadus morhua) necks in the Mediterranean fruit fly (abstract) Genetics
Cod-E 8 95 ll 94:s47–s48.
Cod-F 7 96 ll Irawan B, Kijima A, and Fujio Y, 1993. Genetic diver-
Cod-I 5 98 ll
Cod-G 7 96 ll gence among the three species of estuarine crab, Helice
tridens, H. japonica and Chiromantes dehaani. Tohoku J
Crabs (Halice tridans and Chiromantes dehaani) Agric Res 43:101–109.
H. tridans-2 6 39 mm
H. tridans-3 7 40 mm Jiminez JA, Hughes DA, Alaks G, Graham G, and Lacy
C. dehaani-2 8 39 mm RC, 1994. An experimental study of inbreeding depres-
C. dehaani-1 5 23 mm sion in a natural habitat. Science 266:271–273.
Land snails Johnson MS, 1988. Founder effects and geographic vari-
Vale, France 11 28 nn ation in the land snail Thebia pisana. Heredity 61:133–
CNRS, France 11 28 nn 142.
American oysters (Crassostrea virginica) Karl SA and Avise JC, 1992. Balancing selection at allo-
Cape Cod, Massachusetts 5 90 oo zyme loci in oysters: implications from nuclear RFLPs.
Charleston, South Carolina 5 100 oo Science 256:100–102.
Bay Grabe, Louisiana 5 88 oo
Brownsville, Texas 5 97 oo Kimura M and Crow JF, 1964. The number of alleles that
Milk fish (Chanos chanos) can be maintained in a finite population. Genetics 49:
Oahu 7 60 pp 725–738.
Tarawa 10 38 pp Kennedy PK, Kennedy ML, Clarkson PL, and Liepins IS,
Christmas Island 6 47 pp 1991. Genetic variability in natural populations of the
New Zealand conifers (Halocarpus bidwillii) gray wolf. Can J Zool 69:1183–1188.
Pop-14 7 38 qq Lade JA, Murray ND, Marks CA, and Robinson NA, 1996.
Pop-16 5 76 qq
Pop-2 5 40 qq Microsatellite differentiation between Phillip Island and
Pop-6 5 40 qq mainland Australian populations of the red fox Vulpes
vulpes. Mol Ecol 5:81–87.
Scots pine (Pinus sylvestris)
Yllastunturi 12 44 rr Lande R, 1988. Genetics and demography in biological
conservation. Science 241:1455–1459.
Only one dataset revealed a mode-shifted distribution (Mountain sheep, Sun River population). aa 5 Knudsen K and Lande R, 1994. Risk of population extinction from fixa-
Allendorf FW, personal communication; bb 5 Kennedy et al. 1991; cc 5 Knudsen K and Allendorf FW, personal com- tion of new deleterious mutations. Evolution 48:1460–
munication; dd 5 Baker and Moeed 1987; ee 5 St. Louis and Barlow 1988; ff 5 Wada and Numachi 1991; gg 5 Ovenden 1469.
and White 1989; hh 5 Shaklee and Varnavaskya 1994; ii 5 Winans et al. 1994; jj 5 Allendorf FW and Knudsen K, personal
communication; kk 5 Varnavaskya et al. 1994; ll 5 Mork et al. 1985 (these loci were suspected to be under balancing Lande R and Barrowclaugh GF, 1987. Effective popula-
selection when they were compared to nuclear DNA RFLP loci, Pogson et al. 1995); mm 5 Irawan et al. 1993; nn 5 tion size, genetic variation, and their use in population
Johnson 1988; oo 5 Buroker 1983—we used only the five loci used by Karl and Avise (1992) and suspected by them to management. In: Variable populations for conservation
be under balancing selection; pp 5 Winans 1980; qq 5 Billington 1991; rr 5 Savolainen and Hedrick 1995. (Soule ME ed). Cambridge: Cambridge University
Press; 87–123.

246 The Journal of Heredity 1998:89(3)

Leberg PL, 1992. Effects of a population bottleneck on Ovenden JR and White RWG, 1990. Mitochondrial and applications to conservation and management (PhD
genetic variation as measured by allozyme electropho- allozyme genetics of incipient speciation in a land- dissertation). Sydney, Australia: University of New
resis. Evolution 46:477–494. locked population of Galaxias truttaceus (Pisces: Gal- South Wales.
axiidae). Genetics 124:701–716.
Maruyama T and Fuerst PA, 1985. Population bottle- Taylor AC, Sherwin WB, and Wayne RK, 1994. Genetic
necks and non-equilibrium models in population ge- Paetkau D, Calvert W, Stirling I, and Strobeck C, 1995. variation of microsatellite loci in a bottlenecked spe-
netics. II. Number of alleles in a small population that Microsatellite analysis of population structure in Ca- cies: the hairy-nosed wombat (Lasiorhinus krefftii). Mol
was formed by a recent bottleneck. Genetics 111:675– nadian polar bears. Mol Ecol 4:347–354. Ecol 3:277–290.
689.
Pogson GH, Mesa KA, and Boutilier RG, 1995. Genetic
Varnavskaya NV, Wood CC, and Everett RJ, 1994. Ge-
Matthews JW, 1973. Ecology of bighorn sheep on Wild- population structure and gene flow in the Atlantic cod:
netic variation in sockeye salmon (Oncorhynchus ner-
horse Island, Flathead Lake, Montana (Masters thesis). a comparison of allozyme and nuclear RFLP loci. Ge-
ka) populations of Asia and North America. Can J Fish
Missal, Montana: University of Montana. netics 139:375–385.
Aquat Sci 51(suppl 1):132–145.
Mills SL and Smouse PE, 1994. Demographic conse- Ralls K, Ballou JD, and Templeton A, 1988. Estimates of
quences of inbreeding in remnant populations. Am Nat lethal equivalents and the cost of inbreeding in mam- Vrijenhoek RC, 1994. Genetic diversity and fitness in
144:412–431. mals. Conserv Biol 2:185–193. small populations. In: Conservation genetics ( Loesch-
cke V, Tomiuk J, Jain SK, eds). Berlin: Birkhauser; 37–
Morin PA, Wallis J, Moore JJ, and Woodruff S, 1994. Pa- Roy MS, Geffen E, Smith D, Ostrander EA, and Wayne 53.
ternity exclusion in a community of wild chimpanzees RK, 1994. Patterns of differentiation and hybridization
using hypervariable simple sequence repeats. Mol Ecol in North American wolflike canids, revealed by analysis Wada S and Numachi K, 1991. Allozyme analyses of ge-
3:469–478. of microsatellite loci. Mol Biol Evol 11:553–570. netic differentiation among the populations and spe-
cies of the Balaenoptera. Report of the International
Mork J, Ryman N, Stahl G, Utter F, and Sundnes G, 1985. Savolainen O and Hedrick P, 1995. Heterozygosity and
Whaling Commission (special issue 13):126–154.
Genetic variation in Atlantic cod throughout its range. fitness: no association in scots pine. Genetics 140:755–
Can J Fish Aquat Sci 42:1580–1587. 766. Waits LP, Paetkau D, Strobeck C, and Ward RH, 1996. A
Nei M, Chakraborty R, and Fuerst PA, 1976. Infinite al- Shaklee JB and Varnavaskaya NV, 1994. Electrophoretic comparison of genetic diversity in brown bears from
lele model with varying mutation rate. Proc Natl Acad characterization of odd-year pink salmon populations Alaska, Canada and the lower 48 states. Int Conf Bear
Sci USA 73:4164–4168. from the Pacific coast of Russia, and comparison with Res Manage 10.
selected North American populations. Can J Fish Aquat Watterson GA, 1984. Allele frequencies after a bottle-
Nei M and Li W-H, 1976. The transient distribution of
Sci 51(suppl 1):158–171. neck. Theor Popul Biol 26:387–407.
allele frequencies under mutation pressure. Genet Res
Camb 28:205–214. Sjogren P and Wyoni PI, 1994. Conservation genetics
and detection of rare alleles in finite populations. Con- Winans GA, 1980. Geographic variation in the milkfish
Nei M, Maruyama T, and Chakraborty R, 1975. The bot- Chanos chanos. I. Biochemical evidence. Evolution 34:
serv Biol 8:267–270.
tleneck effect and genetic variability in populations. 558–574.
Evolution 29:1–10. Spielman D and Frankham R, 1992. Modeling problems
in conservation: improvement in reproductive fitness Winans GA, Aebersold PB, Urawa S, and Varnavskaya
Newman D, 1996. Importance of genetics on the surviv-
due to immigration of one individual into small par- N, 1994. Determining continent of origin of chum salm-
al of small populations: genetic drift, inbreeding and
tially inbred populations. Zoo Biol 11:343–351. on using genetic stock identification techniques: status
migration (PhD dissertation), Missal, Montana: Univer-
of allozyme baseline in Asia. Can J Fish Aquat Sci
sity of Montana. St. Louis VL and Barlow JC, 1988. Genetic differentia-
51(suppl 1):95–113.
tion among ancestral and introduced populations of
Ohta T and Kimura M, 1973. A model of mutation ap-
the Eurasian tree sparrow. Evolution 42:266–276. Received November 20, 1996
propriate to estimate the number of electrophoretically
Accepted July 8, 1997
detectable alleles in a finite population. Genet Res Taylor AC, 1995. Molecular ecology of the endangered
Camb 22:201–204. northern hairy-nosed wombat Lasiorhinus krefftii, and Corresponding Editor: Masatoshi Nei

Luikart et al • Testing for Bottlenecks 247

View publication stats