.. .

THE GENETICS OF INBREEDING POPULATIONS

. .
R W Allard. S K Jain. f and P . L. Workman?
University of California. Davis. California

I. Introduction . . . . . . . . . . . . . . . . . . 55
I1. Theoretical Analysis of Genetic Models. . . . . . . . . 57
A. General Considerations . . . . . . . . . . . . . . 57
B. Single-Locus Models . . . . . . . . . . . . . . . 58
.
C Multilocus Population Models: Effects of Linkage and Epistasis
. . . . . . . . . . . . . . . .
under Inbreeding 68
D . Complex Models . . . . . . . . . . . . . . . . 81
.
E Conclusions from Theoretical Analyses . . . . . . . . . 85
111. Experimental Analyses of Polymorphisms . . . . . . . . . 86
A. Estimation of Parameters for Single-Locus Polymorphisms . . . . 86
.B Complex Polymorphisms . . . . . . . . . . . . . . 92
.
IV Genetic Variability in Quantitative Characters . . . . . . .
. 94
.A Geographical Variability . . . . . . . . . . . . . . 95
B. Variability within Populations . . . . . . . . . . . . 97
C . Variability within Families . . . . . . . . . . . . . 99
.
V Responses to Selection . . . . . . . . . . . . . . . 103
A . Competition in Mixtures of Pure Lines . . . . . . . . . 103
B. Responses to Natural Selection . . . . . . . . . . . . 105
C . Responses to Artificial Selection . . . . . . . . . . . 110
.
VI Effect of Altering the Mating System . . . . . . . . . . . 113
A . Enforced Self-Fertilization . . . . . . . . . . . . . 113
B. Increased Outbreeding by Introduction of Male Sterility . . . . 115
VII . Population Structure under Extreme Inbreeding : The Festuca
microstachya Complex . . . . . . . . . . . . . . . 117
VIII. Concluding Remarks . . . . . . . . . . . . . . . . 120
.
IX Summary . . . . . . . . . . . . . . . . . . . 123
References . . . . . . . . . . . . . . . . . . . 125
.
1 Introduction
The term inbreeding is applied when the individuals which mate
together are more closely related to each other than are random mem-
bers of an indefinitely large population . The notion of the closeness
of the relationship between two individuals is most easily visualized
in simple cases and much of inbreeding theory has been based on the
analyses of regular systems of inbreeding involving repeated sib, cousin,
* Departments of Agronomy and Genetics.
t Department of Agronomy.
55
56 R. W. ALLARD, S. I(. J A I N , AND P. L. WORKMAN

or other types of matings between close relatives. From a population

standpoint, however, these systems of inbreeding are interesting only
in a formal sense because, aside from special experimental situations,
matings in populations do not usually follow regular patterns.
I n natural populations there are two general causes of inbreeding:
restriction of population size and variations in the mating system. In
a population of bisexual organisms each individual has two parents
(uniparental, in case of selfing), four grandparents, and so on, so that
n generations back an individual has 2n ancestors. The number of
individuals required to provide separate ancestors for each present
individual obviously becomes very large within a few generations so
that any pair of individuals within a population must have common
ancestors in the not too remote past. The smaller the population the
less remote are the common ancestors and the higher the level of
inbreeding. Inbreeding due to restricted population size is not con-
fined to populations consisting of a few individuals isolated from others
of their kind. Many factors can lead to effective restriction of popula-
tion size, even though the total number of individuals in a population
may be large (Wright, 1943, 1946). Isolation by distance is a n ex-
ample: there is a tendency for close neighbors t o mate and proximity
in time or space increases the probability that these close neighbors
will be relatives, The amount of inbreeding that results from the
mating of close neighbors, even in a panmictic population, is largely
a function of the relative mobility of the organisms or their gametes
or propagules.
The other main source of inbreeding in natural populations derives
from various mechanisms in both plants and animals that affect the
mating system (reviews in Grant, 1956, 1958; Mayr, 1963). An example
in animals is the brother-sister mating observed in parasitic hymen-
opterans that mate within the host (Dreyfus and Breuer, 1944). I n
plants, modifications of floral structure which encourage self-fertiliza-
tion or assortative mating are widespread and in many instances they
are remarkably effective, reducing the amount of outcrossing to 1% or
less. Preferential mating between similar genotypes can also be an im-
portant source of inbreeding. It should be noted that these two main
causes of inbreeding, restriction of population size and control of mating
system, can occur simultaneously and reinforce one another in their
effects.
Inbreeding increases the probability that individuals which mate
carry alleles that are alike by virtue of descent from a common an-
cestor. The essential consequences, as shown theoretically by Wahlund
(1928) and Wright (1951), are a reduction in genetic variability within
 GENETICS OF INBREEDING POPULATIONS 57

families (kinships of related individuals) and an increase in genetic

variability between families, which become genetically differentiated
from one another under inbreeding. The magnitude of these effects
was well illustrated by Johannsen’s population of Princess beans which
was differentiated into a large number of strikingly different, true
breeding families (Johannsen, 1926).
There have been many deductions regarding the genetic structure of
inbreeding populations since Johannsen’s work. It has usually been
argued that inbreeding, by increasing homozygosity, also increases
genetic uniformity and hence that i t allows populations to achieve
“closer adaptation’’ (or adaptedness) to the immediate environment.
It is also often argued that the gain in adaptation (fitness in the cur-
rent environment) conferred by inbreeding is achieved largely through
loss in adaptability (flexibility or capacity for change in adaptation).
Thus, the mating system is commonly considered to be the chief fac-
tor determining the genetic structure and evolutionary potential of
populations and a sharp distinction is usually drawn between inbreed-
ers and outbreeders (see discussions by Darlington and Mather, 1949;
Stebbins, 1950). Such considerations have also been the basis for dis-
cussions of evolutionary changes in breeding systems (Mather, 1943;
Stebbins, 1957; Grant, 1958).
I n this review we shall consider, in a population context, the effects
of inbreeding in terms of genetic models of steadily increasing com-
plexity, We will then correlate theoretical effects with observations
and measurements which show that inbreeding populations contain
large amounts of genetic variability and that this variability is organ-
ized into highly integrated and flexible systems. The thesis that we
shall develop is that the observed structure of inbreeding populations
results from an appropriate integration of inbreeding into the con-
stellation of genetic and ecological factors which are involved in the
regulation of variability and maintenance of flexibility.

II. Theoretical Analysis of Genetic Models

A. GENERALCONSIDERATIONS
An understanding of the genetic structure of inbreeding species de-
rives from combining analyses of theoretical models with studies on
experimental and natural populations. By varying the restrictions on
the genetic parameters which define the theoretical models (e.g., those
describing population size, the mating system, the selective forces) i t
is possible to explore the relations between these various parameters.
58 R. W. ALLARD, S. K. J A I N , AND P. L. WORKMAN

Generally this entails a comparison of gene and genotypic frequencies,

both a t equilibrium and during the approach t o equilibrium, for models
with different specifications. One of the important discoveries of popula-
tion genetics is that all populations, including inbreeding popula-
tions, contain large stores of genetic variability. It is, therefore, of
particular interest to ascertain which combinations of values of the
genetic parameters result in a population structure in which genetic
variability is retained permanently.
We shall consider first models describing the distribution of geno-
types at a single diallelic locus in an indefinitely large diploid popula-
tion. An examination of the simple models of complete random mating
and complete selfing will be followed by a treatment of models per-
taining to partial inbreeding and partial inbreeding with selection.
Next we shall consider two locus models with particular attention di-
rected to the joint effects of inbreeding, linkage, and epistasis on
population structure. The aspect of epistasis that will be emphasized is
the manifestation of interactions between loci as they are expressed
on the fitness scale. Finally we shall discuss the view of population
structure provided by analyses of several more complex models such
as those including multiple loci, stochastic variation of the genetic
parameters, and restriction of population size.
All of the models to be considered assume that generations are dis-
crete (i,e., nonoverlapping). However, it can be shown that corre-
sponding continuous time models, that is, those assuming overlapping
generations, lead to the same conclusions as those derived from the
discrete models.

B. SINGLE-LOCUS
MODELS
I n the discussion of single-locus models we shall restrict the treatment
to a locus with two alleles, say A1 and A2. We shall denote the genotypic
proportions of (AlAI, AlA2, A2A2) in generation n by (fl(ll), fz("), fa("))
+
and the allelic proportions of (A1? A,) by ( p @ ) ,q(")),where ptn) q(")
= 1.
1. Complete Random Mating and Complete Inbreeding
The most elementary model is that of random mating in a n indefinitely
large population, described by the Hardy-Weinberg theorem. If the fre-
quencies of alleles Al and A%,at any generation, n, are given by p and
q = 1 - p , then the genotypic proportions in the next and all subsequent
generations are given by (&A1, A1A2, A2A2) = ( p 2 , 2pq, q z ) .For multi-
 GENETICS OF INBREEDlNG POPULATIONS 59
ple alleles the genotypic proportions are given by A d , = pi2, and A& =
2 p & p j ( i # j ) . For the diallelic model, the heterozygous genotype, A1A2,
has frequency fz = 2 p q and consequently the maximum possible heter-
ozygosity is determined from the solution of d(fz)/dq = 0, for which
q = $/2. Thus, under random mating without selection the maximum
heterozygosity possible is 50% and the range of possible values for f2 is
clearly 0 < f 2 5 1/2, depending only on p and q .
I n contrast to the case of random mating, consider population structure
under a system of complete inbreeding. For plant popul8tions this could
result from complete self-fertilization; for both plant and animal popula-
tions it might occur if the only matings permitted were those between
like genotypes (genotypic positive assortative mating) or like phenotypes
(phenotypic positive assortative mating).
Complete self-fertilization and genotypic assortative mating are
formally identical systems and by the technique of generation matrices
(Haldane and Waddington, 1931 ; Fisher, 1949; Kempthorne, 1957) it is
possible to derive n-step recurrence formulas for these mating systems,
T h e recurrences give the genotypic proportions a t any generation, n, in
terms of the initial genotypic distribution (fl@), f 2 @ ) , and n. For
complete selfing these can be shown t o be
A1A1: fi(") = fi'o) + ( 1 / 2 ) [ 1 - (1/2)n]f2(D)
AIAz: fiCn) = (1/2)" fz'"' (1)
A2Az: fa(") = fa(') + ( 1 / 2 ) [ 1 - (1/2)*]fi(O)
Equations (1) show that the amount of heterozygosity in a population is
halved every generation. I n the limit (limn+m f(@) the genotypic
+
distribution (fl", f z m l f a m ) is given by ( f ~ @ ) ( 1 / 2 ) * f z ( O ) , +
0, f a @ )
(1/2)f3(0)), or ( p @ ) ,0, q ( O ) ) . Thus, under this model, the limiting state
consists solely of homozygous genotypes in the same proportion as the
original gene frequencies and no heterozygosity is retained.
For phenotypic assortative mating, in the case of dominance, i t can be
shown (e.g., Hogben, 1946) that the limiting distribution is identical to
that for genotypic assortative mating. However, the rate of approach t o
equilibrium is different. The heterozygosity a t generation n is given by

and the rate of decrease in heterozygosity is much slower than in the case
.of genotypic assortative mating. For both cases, gene frequencies remain
constant over all generations (pi(")= p P for all n, m ) ; only the
genotypic proportions change.
These models of complete random mating and complete inbreeding
represent extremes that rarely, if ever, would occur in real populations.
60 R. W. ALLARD, S. K . JAIN, AND P. L. WORKMAN

A more realistic appraisal of population structure, therefore, requires

consideration of mating systems in which the actual amount of inbreed-
ing lies somewhere between the extreme values of 0 and 100%.
2. Partial Inbreeding
The mating system of a population may be one of partial inbreeding
for many different reasons. Dioecious plants (those producing both
male and female gametes) may mate according t o some mixed system
of random mating and self-fertilization. Genetic control of variation in
flowering time might result in assortative mating among those plants
flowering concurrently. Genetic variation in flower color or in the struc-
ture of the reproductive organs may lead to assortative mating in
insect-pollinated species when the pollinators have preferences for
particular colors or shapes. Positive sssortative mating arising from
mating preference between similar types has also been observed in
a number of animal populations (e.g., O'Donald, 1959; Cooch and Beard-
more, 1959). Pearson and Lee (1903) and others have noted its occur-
rence in human populations with respect to various physical charac-
teristics, Mating among members of the same family either as a regu-
lar system (e.g., sib mating in parasitic hymenopterans; review in
Mayr, 1963) or as some proportion of the total matings (uncle-niece
or cousin matings in man) is another common cause of partial in-
breeding.
The first systematic study of models of inbreeding systems was that of
Wright (1921), reported in a classic series of papers entitled "Systems
of Mating." Since then, numerous workers, employing a variety of
techniques, have considered the problem, both generally and in terms
of specific systems. We shall consider here only the most general models
of mixed random mating and selfing, or equivalently, partial positive
assortative mating, these two systems being formally identical. Refer-
ences to more specific models (e.g., regular systems such as those in-
volving uncle-niece, full sib, or parent-offspring mating) and descrip-
tions of the various techniques used for their analysis are found in
Kempthorne (1957).
Consider a population in which there is a constant probability, s,
for self-fertilisation and a corresponding constant probability, t = 1 -
s, for random outcrossing. Recurrence equations relating genotypic
proportions in successive generations are given by
s[fiCn) + (1/4) f2(")] +
t[fi(") + (1/2) f2'"'12
+ + + (1/2) f~'"']
fi(n+l' =
~[(1/2) fz'"'] 2t[fl'"' (1/2) fz'"'][f3'"' (3)
+
=
+ +
fz("+l)
f3("+') = ~[f3("' (1/4) fz'"'] t[f3("' (1/2) fz""Ia
 GENETICS OF INBREEDING POPULATIONS 61

It can be shown from these equations that the gene frequencies remain
constant in all generations, but the genotypic frequencies change until
an equilibrium state is attained. The general n-step formulas for this
system, given as Eqs. (4),can be obtained by matrix methods (Work-
man and Allard, 1962) or by the methods appropriate to the solution
of finite difference equations (Haldane, 1924) :
fi'") = S" fl'"' + (1/2) (S")[l - (1/2)"1 f2'"

f3(4 = s= .f3'") + (1/2)(9)"[1 - (1/2)"]fP

where p and q are

[fl'"' + (1/2) fi'"'1
and
If3'"' + (1/2) f2'")I

respertively. In the limit (as n + co ) the equilibrium state is given by

It is of particular interest to know how much heterozygosity can be

maintained indefinitely, given various assumptions about the initial
population and the amount of outcrossing. Table 1 gives the amount
of heterozygosity expected a t equilibrium for different values of p and
t. By the symmetry of the model, the tabular value of p is that of q
when p > 0.5.
An alternative derivation of the above results can be obtained by
the use of the coefficient of inbreeding, F (Wright, 1921), which was
first defined as "the correlation between homologous genes of uniting
gametes under a given mating pattern relative to the total array of
those in random derivations of the foundation stock" (Wright, 1965a).
Other interpretations of F in terms of probabilities of identity of unit-
ing gametes (MalBcot, 1948) or as a function of the relative amount of
heterozygosity are reviewed by Wright (1965a). The use of F statistics
has proven to be one of the most successful methods employed in the
study of inbreeding systems (see Kempthorne, 1957, for examples).
62 R. W. ALLARD, S. K. J A I N , AND P. L. WORKMAN

TABLE 1
Expected Equilibrium Proportions of Heteroeygotes under Mixed Random Mating
and Selfing (without Selection) for Various Assumptions about t and p (Initial Frequency
of Allele At)

t 0.05 0.10 0.20 0.30 0.40 0.50

-
0.1 0.0173 0.0327 0.0582 0.0764 0.0783 0.0909
0.2 0.0317 0.0600 0.1067 0.1400 0.1600 0.1667
0.3 0.0439 0.0831 0.1477 0,1939 0.2216 0.2308
0.4 0.0543 0.1029 0.1829 0.2610 0.2743 0.2857
0.5 0.0633 0.1200 0.2133 0.2800 0.3200 0.3333
0.6 0.0713 0.1350 0.2400 0.3150 0.3600 0.3750
0.7 0.0783 0.1482 0.2636 0.3459 0.3953 0.4118
0.8 0.0844 0.1600 0.2844 0.3733 0.4266 0.4444
0.9 0.0900 0.1705 0,3022 0.3969 0.4548 0.4737
1.0 0.0950 0.1800 0.3200 0.4200 0.4800 0.5oOo

At equilibrium, any inbreeding population can be written as (AIA1,

A I A z , A z A ~= + +
) ( p a pqP, 2 p q ( l - F ) , q2 p q F ) where F is the
equilibrium coefficient of inbreeding, and both p and F are constant (e.g.,
Wright, 1942). For the case of mixed random mating and selfing, an
n-step equation for F can be shown to be (Kempthorne, 1957)

where F(O) is the inbreeding coefficient in the original population. The

limiting formula ( n +=00) gives Frn) = s/(2 - s) and the equilibrium
hetero~ygosity,given by 2 p q ( l - F ) , is f 2 = 2pq{2t/ (2 - s) }, which is
equal to the expression ( 5 ) obtained by matrix or finite difference
methods.
It is possible that the probabilities for selfing versus outcrossing differ
for each of the genotypes. For this situation let a,/3, y be the probabilities
for selfing of AIA1, A1A2, A2A2, respectively. Then (1- a), (1 - p ) ,
(1- 7) are the corresponding probabilities for outcrossing of these three
genotypes. Note that when a = p = y = s , this model reduces to that
of mixed random mating and selfing already considered. The case when
ff, p, and y are different and operate on both sexes has been considered

by O’Donald (1960b), who obtained the recurrence equation for t h e

hetero~ygote,fz, and an exact equation for the equilibrium state. He also
considered the case of dominance (partial phenotypic assortative mating).
His results showed that when the tendency for assortative mating is
 GENETICS OF INBREEDING POPULATIONS 63

small, the equilibrium distribution is very close to that expected under

random mating.
A variant of the above is the model in which only the female parent
has different probabilities for random mating versus selfing (or assorta-
tive mating). This could apply to plant populations in which there is
an excess of pollen disseminated a t random, or to animal populations in
which mating is determined by female preference. The relevance of
models based on 0, 8, and y , as opposed to s alone, is indicated, for
example, by Harding and Tucker (1964), who showed that estimates of
the amount of outcrossing a t a locus varied with the particular genotype
used in the estimation.
Mating preference leading to assortat.ive mating might also be de-
termined in part by the genotype of one or both of the parents. The
effect of this imprinting on genotypic proportions has been con-
sidered by O’Donald (1960a) and Mainardi et al. (1965). Such a sys-
tem was suggested as the basis for the assortative mating in the Blue
Goose (Huxley, 1955; Cooch and Beardmore, 1959).
The models thus far considered vary only with respect to the mating
system and represent, as such, extremely idealized representations of
natural populations. In the following section we shall introduce vari-
ation due to differential selective forces in order to determine the
kinds of interactions between the mating system and the selective
forces which result in the maintenance of genetic heterogeneity.
3. Inbreeding and Selection
There are many different ways in which selective forces can act to
alter gene and genotype frequencies, in terms of both their possible
equilibrium distributions and the approach to equilibrium. Selection
pressure can be assumed to be constant over time or regularly cyclic
within generations or between generations [e.g., this has been suggested
as the explanation for a chromosomal polymorphism in the mantid,
Ameles heldreichi (Wahrman, 1965)l. The intensity of selection may
vary over an environment (clinal selection, habitat, or disruptive
selection), differ for different sexes or morphs, vary according to the
frequencies of the genotypes in the population (Clarke and O’Donald,
1964) , or vary with population density. Selection can act directly
on the gametes or on zygotes a t different stages in the life cycle
(differential viability, differential fecundity), Combinations of these
selective forces acting a t several different times may account for the
differential reproductive capacities of genotypes a t a given locus. I n
practice, however, it is usually necessary to consider only simple models
in which selection, operating a t a specified stage, is ascribable to a
64 R. W. ALLARD, S. K. JAIN, AND P. L. WORKMAN

single specified cause. We shall consider only two general models in

detail: selfing with zygotic selection, and mixed random mating and
selfing with zygotic selection. The conclusions drawn from analyses of
these models regarding the relation between inbreeding and selection
are applicable to a wide range of more specific models.
a. Selfing and Zygotic Selection. The most general solution is that
given by Hayman and Mather (1953, 1956)' who incorporated into their
model selection between lines, seIection within lines, and seed selection.
They obtained the n-step relations,

a wncfs(") - vfi(0)) +-1

2"
sf2(0)

where (z,1, y) are the relative survival rates of (fl, fi,

fa) in segregating
families, z and w are the relative survival rates of fl and f 3 in true-
breeding families, and u = x / [ 2 (1 - 22)] and v = y/[2 ( 1 - 2w)l are
compound survival parameters. When z = w = (%)x +
(%)y + 1/2 se-
lection occurs only within lines; when z = 'y = 1 selection occurs only
between lines; and when z = x and w = y only seed selection operates.
The equilibrium distributions are as follows. The population will be
homozygous f l (or fa) if z > 1/2, w (w > 1/2, 2 ) . For z = w > 1/2 there
will be a mixture of homozygotes fl and fa. In particular, if 1/2 > z, w,
heterosygosity persists and the equilibrium genotypic frequencies are

Thus, for the case of full selfing, the heterozygotes must have a marked
superiority over the homozygotes (z, y 5 1/2) in order to maintain varia-
bility in the population. Hayman and Mather (1953) also treated par-
ticular cases of heterozygote advantage for several regular systems of
inbreeding (e.g., sib mating, parent-offspring mating) and made compari-
sons among these systems and the case of complete selfing.
b. Mixed Random Mating and Selfing with Zygotic Selection. Models
involving selection give rise to a series of nonlinear recurrence equations
for which, except for the case of full selfing, it has not been possible t o
obtain general n-step formulas. The analysis of such models involves the
examination of the distribution of genotypic frequencies at equilibrium
either directly, or in terms of gene frequency and Wright's F. At equilib-
rium ft("+l) = f i @ ) = f 6 and the solution is given by the simultaneous
 GENETICS OF INBREEDING POPULATIONS 65

solution of the equations in f4. The values of the relevant parameters

which result in stable nontrivial equilibria can be determined by testing
the stability by whichever of the available methods is most appropriate
(e.g., Owen, 1953; Lewontin, 1958; and general methods of Saaty and
Bram, 1964).
Suppose that the relative viabilities of (Il, f2, f 3 ) are in the ratio
(2, 1, y) and assume constant probabilities for selfing (s) and random
mating ( t = 1 - s) , If the genotypic frequencies are assumed to be scored
just before mating and after all of the zygotic selection has occurred, the
model is that described by Hayman (1953). The recurrence relations are

X( s(fi(") + (1/4) + + (1/2)

+ (1/2) j ~ ' ~ )(9))
+ 2t(jI(n) + (1/2) fz(n))(ja(n)
fi(n+l) a f2(n)) t(fi(") fi'n')2)
f2(n+l) a (1/2) sj2'")
f~(~+') a y(S(fs'"' (1/4) fz'"') t(f3(n) (1/2) ~Z(~)'I
The most direct approach to a solution of these proportionalities is that
described by Workman and Jain (1966) and Jain and Workman (1966),
If we define F = 1 - f n / 2 p q , when p = ( f l + ( 1 / 2 ) f 2 ) , then the geno-
typic distribution a t any generation, n, can be written as

If the only force causing genotypic frequencies to depart from Hardy-

Weinberg expectations is that of inbreeding, then F(n)is the coefficient of
inbreeding a t generation n. If selection is also involved, then F(*) denotes
the joint effects of inbreeding and selection and, in Wright's (1965s)
terminology, F(n)is the fixation index. At equilibrium both Ap = 0 and
AF= 0 (or dp/dt = dF/dt = 0 ) . Using this transform, the above equa-
tions (9) at equilibrium can be written as

+
pZ p q F +
x ( p 2 (1/2)s p q ( 1 F))+
+
fi: a
2Pdl - F ) (2PP - pq(1 F)I (11)
+ +
f2: a
s3: !I2 P q F a Y{q2 +(1/2)s P d l F)1
where F is the fixation index a t equilibrium and p is the equilibrium gene
frequency. From the relations (11) Workman and Jain (1966) showed
that both of the following equations must be satisfied a t equilibrium:
- (1/2)s(l - X)(1 + F )
(1 - y)
- y) [ l - (1/2)8(1 + F ) ]
= (2 -x
sF2(1 - X) + F ( 2 ( ~ y- 1) + ~ ( -
2 z - g)}
- 2(1 - ~ ) ( 1- y) + ~ ( -l ZY) = 0 (13)
66 R. W. ALLARD, S. K. J A I N , A N D P. L. WORKMAN

I n considering equilibrium populations it is useful to examine phase dia-

grams (Figs, l, 2) as described by Hayman (1953) and Workman and
Jain (1966). These diagrams show types of equilibria for different magni-
tudes of the relative viabilities (z, y) of the two homozygotes, given
some particular amount of selfing ( 5 ) . The areas marked A and B cor-

""I /

3
x
FIG.1. Phase dingrrtm for s = 1.00. See text for description of areas A, B, C, D.

0.4 1 \ I a

FIG.2. Phase diagram for s = 0.95. See text for description of areas A, B, C, D.

respond to populations homozygous for AIAl (fl) and AzAz (fs) , respec-
tively; in C,heterozygotes are present but in a frequency lower than in
a population mating a t random without selection ( F > 0) ; in D there
are more heterozygotes than in a random mating population without
selection ( F < 0). It can be seen that the amount of heterosygote ad-
 GENETICS OF INBREEDING POPULATIONS 67

vantage needed to maintain variability is a function of the amount of

selfing. As the amount of selfing increases, increasing heterozygote
advantage is required for the maintenance of a nontrivial equilibrium.
Two variations of this model have been considered by Workman and
Jain (1966). First, suppose that genotypic frequencies are scored just
after mating. The genotypic distribution a t equilibrium will be equal to
that obtained by imposing one generation of mixed random mating and
selfing upon the genotypic proportions obtained in the above model (11)
in which the genotype proportions are determined prior to mating. Con-
sequently, the value of p will be the same since gene frequencies remain
constant under mixed mating; but F, which depends on both genotypic
and gene frequencies, will be different. As can be shown by setting n = 1
in (6) above, the appropriate value of F , say F‘, will be equal to ( x ) s
+
(1 F ) , Thus, for this model, the equilibrium is described by

(1 - y) - F’(1 - z)
= (2 - z - y)(l - F’)
2F’2(1 -X)(1 - y > + F ’ ( ( z + y - 2 ) -t(z+y-2xy)}
+ (X + 2/ - 2zy)(l - t ) = 0 (13’)

The phase diagrams for this model, which differs only operationally from
the previous one [see Eqs. ( l l ) ] , are the same as given by Figs. 1 and 2.
If selection acts only on one sex, the model is one of differential
fecundity, rather than one of differential viability as above. If genotypes
are counted after mating, the equilibrium conditions are found to be
identical to those in the above model of differential viability (selection
in both sexes) when genotypes are scored after mating [Eqs. (12‘, 13’)].
Thus, for each of the above three models of selection and inbreeding
the equilibrium values of p and F are identical a t the same stage of the
life cycle. However, the approach to equilibrium differs among the three
models. Figures 3 and 4 show changes in p and F plotted against time
for x = 0.50, y = 0.75 and ( P o , Ro, Qo) = (0.25, 0.50, 0.25). Note that
p goes monotonically to its equilibrium value, but F can approach its
equilibrium value by complex paths.
The models so far discussed show that the permanent maintenance of
variability under inbreeding requires some net heterozygote advantage,
the degree of advantage required varying directly with the intensity of
inbreeding. It should be noted that net heterozygote advantage does not
necessarily imply that the heterozygote is adaptively superior, per se, to
both homozygotes since various mechanisms such as frequency-dependent
selection or disruptive selection can create a net heterozygote advantage
(see Dempster, 1955). Other kinds of interactions that result in the
68 11. W. ALLARD, S. K. J A I N , AND P. L. WORKMAN

5 10 15 20 25
Generation
FIG.3. Changes in p and F for three different models during the approach to
equilibrium for z = 0.50, y = 0.75; s = 0.0. p,FI, selection on both sexes, census prior
to mating, pI,FII, selection on both sexes, census after mating. pIIIFIII,selection on
one sex only, census after mating.
1.0 I

5
FIG.4. Changes in p and F for three different models during the approach to
equilibrium for z = 0.50, y = 0.75; s = 0.50. See Fig. 3 for key.
maintenance of genetic variability will be considered next in the discus-
sion of two-locus and more complex genetic models.

C. MULTILOCUS POPULATION MODELS : EFFECTS OF LINKAGE

AND EPISTASIS UNDER INBREEDING
The simplest model that permits the effects of linkage to be considered
is one of two linked loci with two alleles each. The notation that we shall
 GENETICS OF INBREEDING POPULATIONS 69

use in the discussion of two-locus models is given in the accompanying

tabulation.

Loci and allelic designation A a B b

Gene frequencies PI q1 = I - Pl pz qz = 1 - pz
Gametes AB Ab aB ab
Gametic frequencies 91 Ba 93 94 = 1 - 81 - Yz - 93

Genotypes
A
- B AB -
- Ab -
AB A
- B Ab - -aB_ aB_ ab
- Ab
AB Ab A6 aB ab aB ab aB ab ab
Genotypic frequencies fl f? f3 f4 f6 f6 f7 fS fB f10

I n order to discuss the effects of linkage and epistasis on genetic

changes a t two or more loci jointly, it is necessary to define the terms
“linkage equilibrium” and “linkage disequilibrium.” Linkage equilibrium
is defined as the state in which gametic frequencies correspond to the
products of the appropriate gene frequencies :
91 = PIP2 93 = QlPZ
92 = m q 2 g4 = QlQZ
It can be shown (Geiringer, 1944, 1945) that gametes in populations in
linkage disequilibrium have the frequencies
+
gi = P ~ P Z D 93 = pzqi -D
gz = piqz - D y4 = ~ 1 +D ~ 2

where D = glg4 - g2g3. These relationships also hold for inbreeding. The
parameter D is consequently a useful general measure of unbalance in
the proportion of coupling and repulsion gametes.
As early as 1909 Weinberg suggested that linkage equilibrium (D = 0)
is, in general, not reached in a single generation of random mating, even
without linkage. Jennings (1917) and Robbins (1918) provided explicit
expressions for the rate a t which any initial linkage disequilibrium in a
population, such as that in a population derived from the cross of two
homozygous parents, disappears under random mating. The recursion for
change in D is simply
D(n) = (1 - c)DCn-I)
or (14)
D(n) = (1 - c)nD(o)
where D(0) and D(”) measure the amount of gametic array unbalance in
the initial and nth generations and c is the recombination value between
the two loci. I n the absence of selection limn+mD ( n )= 0, so that after
70 R. W. ALLARD, S. K. J A I N , AND P. L. WORKMAN

a suf6cient number of generations in any random mating population the

loci will come into linkage equilibrium (D= 0).
When the mating system is one of complete selfing the recursion for
the gametic frequency of AB(gl) is (Bennett and Binet, 1956)

so that D does not become zero at genotypic frequency equilibrium. Thus

a correlated gene distribution, once created by any factor causing fti # j e t
persists to genotypic frequency equilibrium even in the absence of linkage.
As we shall see in the next section this result does not hold for partial
inbreeding. Consequently the case of complete selfing cannot be treated
simply as the limiting case of partial inbreeding systems.
I. Partial Inbreeding
Systems of partial inbreeding have been treated analytically by three
main methods. The generation matrix method was developed by Haldane
and Waddington (1931), who used it to analyze rates of approach t o
equilibrium for certain regular systems of inbreeding (e.g., parent-
offspring and full sib mating). Wright (1933) approached the problem
through the shorter method of path coefficients, using the correlation
coefficient ( r ) between gene distributions a t two linked loci, where r is
given by
D
r =
(PlqlPzqz)lIa
His analysis of a wide variety of mating systems led to the conclusion
that “different pairs of allelomorphs, even in the same chromosome, come
to be combined practically a t random in any freely interbreeding popula-
tion of long standing” and so also “within different subgroups of a popu-
lation unless these subgroups are very small or linkage is extremely close.”
Bennett and Binet (1956) and Binet et al. (1959) developed n-step recur-
sions for determining changes in the quantities D and v = (f6 - fe) for
the case of mixed selfing and random mating. Using methods of matrix
algebra they showed that for s < 1, both these quantities eventually
become zero in the absence of selection. Thus, mating systems involving
mixed selfing and random mating also lead t o linkage equilibrium.
Haldane (1949) used the probability method of Malecot (1948) to
investigate the joint effects of inbreeding on two linked loci. He estab-
lished that excesses or deficiencies of doubly homozygous and doubly
heterozygous gene combinations occur under partial inbreeding, even in
populations which are in linkage equilibrium (D= 0). Such zygotic as-
 GENETICS OF INBREEDING POPULATIONS 71

sociations become increasingly pronounced as linkage intensity increases.

It was shown later by Schnell (1961) and Kimura (1963) that rates of
approach to equilibrium are also affected by Haldane’s joint inbreeding
function, defined as a measure of identity of allelic pairs by descent.
The role of this “mating system effect” under mixed selfing and random
mating has been investigated by Bennett and Binet (1956). They showed
that the excess of double heterozygotes under free recombination ( c =
0.50) is given by

or generally, for any c,

w =
(4
1 6 ~ ( l- S )
- ~ ) (-
[2 - s -4 ~ ( 1- ~ ) ( l
2 s ) ~ 4 - 2s f 4 ~ ( l C) -- ‘’1 ~ 1 ~ 2 4 1 ~ (18)
2

where the parameter w provides a measure of zygotic associations inde-

pendent of the gametic unbalance parameter D. Changes that occur in D
and o as an arbitrary population progresses toward equilibrium are
plotted in Figs. 5 and 6. Figure 7 gives the magnitude of o measuring
the zygotic associations in equilibrium populations for various specifica-
tions regarding s and c. The effect of w when there is selection will be
considered later.
These results lead to two main conclusions. First, the description of
genotypic frequencies is complicated under partial inbreeding since
0.25 I I

-1 5 10 15 20 25
Genera tion
Fra. 5. Changes in D and w under random mating for c=0.10 and c =OM.The
genotypes AABB and aabb were equally frequent in the original populations 80
that D and LO had their maximum values of 0.25. Notice that the value of o de-
creases slowly when linkage is tight.
72 R. W. ALLARD, S. K. JAIN, AND P. L. WORKMAN

0.25p,
-...\.
---___
-----__ C =0.01 D

.\\
'\
0.20- \
\ %\<:\ ;0,
OD
\
0.15 '\
~

/ A \

Proportion of selfing, S

FIG.7. Equilibrium valuw of w under varying proportions of self-fertilization (8)

and different recombination values (c).

gametic associations ( D ),zygotic associations (a),and the joint inbreed-

ing effect must all be taken into account. Second, linkage has greater
effects on both gametic and zygotic associations under partial inbreeding
than under random mating.
Before turning to models in which selection is incorporated, note should
 GENETICS OF INBREEDING POPULATIONS 73
be taken of investigations of multilocus systems in terms of the distribu-
tion in succeeding generations of complete linkage blocks. Fisher (1948,
1949) Bennett (1954) and Hanson (1959), using the concept of average
) )

recombination rates, have shown that linkage has characteristically dif-

ferent effects under different mating systems and that these effects have
a bearing on population change under selection.
6. Random Mating: Selection and Linkage
It is tempting to conclude from studies of models of random mating
without selection that linkage effects are eventually dissipated and that
they have no significant influence on the structure of the equilibrium
population. Indeed, as emphasized by Lewontin (1964a)) most of the
formulations of multiple gene theory have been extensions of single gene
models that ignore linkage and epistasis. Wright, however, pointed out
in 1942 that selection “tends to bring about departures from random
combination among different series of alleles. The effects are unimportant
in most cases, especially if all relative selective differences are slight.’’
Later, in investigations of an optimum model, Wright (1952) demon-
strated that epistasis expressed on the fitness scale can lead to nonrandom
gene combinations (D# 0) in equilibrium populations. The role of
linkage in the quantitative theory of selection has also been discussed by
Griffing (1960) and Gibson and Thoday (1962). Information on condi-
tions which permit permanent linkage disequilibrium and on the magni-
tude of the resulting gametic and zygotic associations is therefore clearly
necessary to an understanding of population structure.
Consider again the two-locus model presented earlier in which the fls
are the ten genotypic frequencies. Further assume that the selective
values, wi,of the various genotypes represent relative probabilities that
a zygote in one generation will leave a zygote in the following genera-
tion. Numerical values can be assigned to the w;s so as to produce selec-
tion models without epistasis, or with any kind of epistasis. Given the

AR Ab aB ab Marginal means
AB w1 W2 w4 wb WAB
A6 WS Wa wa W7 WAb
aB w4 ws wa ?UQ WaB
ab W5 w7 WD WlO Wab

accompanying matrix of selective values, the average selective value of

gametic type AB in all its combinations under random mating is
w.48 = wlgl f w2gz 4- wag3 i-w&74
The average selective values of the other gametic types, W A b ) W,B, and
74 R. W. ALLARD, 8. K. J A I N , AND P. L. WORKMAN

wab, can be obtained similarly from appropriate marginal means. Since

the genotypic frequencies, ji, are obtained directly as the products g4g5, the
population is a t equilibrium when Agi = 0 (Lewontin and Kojima, 1960) :

+ + +
where W = glWAB g2wAb g3waR g4Wab denotes mean population
fitness.
The expressions (19) are cubic equations in the gi)s and they are con-
sequently difficult to solve for general wls. However, analytical solutions
are possible for w1 = w3 = w8 = wlo = u, w4 = w7 = v, w2 = w9 = w,
WE = we = x, where values of u, v, w, x can be taken which will produce
overdominance a t one or both loci, and epistasis results when ( u 2 - +
w - v) # 0. By virtue of the symmetry of the model, Eqs. (19) reduce
to a single cubic equation
&(ZJ f W - X - U)(4gi2 - 3gi -k 1/2) - W(g1 - 114) = 0 (20)
+
When there is no epistasis (u z - w - v = 0) the only nontrivial
solution of Eq. (20) is G1 = & = = J4 = 1/4. Hence a t equilibrium
+
D = 0. However, when the loci interact ( u x - v - w # 0) Eq. (20)
has three solutions :
= 114
cx
and 61 = 1/4 f

where GI= -
= 1/2 G2 = 1/2 - i3.Thus for recombination values c <
+
(u x - v -w) /4 there will be permanent gametic phase unbalance
(D # 0) a t gene frequency equilibrium and the linkage value required
to yield D # 0 a t gene frequency equilibrium depends on the type and
extent of epistasis. Lewontin and Kojima concluded from studies of an
heterotic model that “as a general rule joint effects of linkage and
epistasis do not produce serious changes in population structure except
under special circumstances. These circumstances are the simultaneous
existence of marked epistasis and tight linkage.”
 GENETICS OF INBREEDING POPULATIONS 75

More recent,ly, i t has been possible t o extend studies of the effects of

linkage and epistasis to a wider variety of two-locus models, including
asymmetrical selection models, through simulation on digital computers
(Lewontin, 1964a,b; Kojima, 1965; Jain and Allard, 1966). Some of the
conclusions are
(a) When there is no epistasis, linkage affects the rate of progress
toward equilibrium but not the final equilibrium state of a population.
(b) Under certain conditions of epistasis permanent gametic phase
unbalance is maintained even for genes that are unlinked.
(0) The epistasis which results in gametic phase unbalance can be
generated by certain multiplicative processes and also those which char-
acterize optimising selection.
(d) The initial composition of a population respecting gametic phase
can affect the final gene frequency equilibrium which is attained. Thus
populations which have identical gene frequencies but differ respecting
gametic phase may reach different gene frequency equilibrium states.
(e) Gametic unbalance enhances mean populational fitness a t gene
frequency equilibrium.
These points are illustrated by the following two genetic models
(Table 2). The first model (Model 1) involves asymmetrical selection
T.4BLE 2
Selective Values for Models 1 and 2

Model 1 Model 2

AA AlZ aa AA Aa aa

BB 0.8 0.5 0.8 BB 0.64 1 .oo 0.96

Bb 0.6 1.0 0.5 Bb 1.00 1.05 1.00
bb 0.7 0.6 0.6 bb 0.94 1 .oo 0.64

with mixed under- and overdominance a t the two loci (Jain and Allard,
1966). Gene frequency equilibria and the trajectories of change in gene
frequency leading to these equilibria for four different linkage values
are superimposed on Wrightian adaptive topographies in Fig. 8. These
topographies were drawn under the assumption that D = 0, an assump-
tion that is not correct since genotypic frequencies are functions of both
D and gene frequencies and this model produces permanent gametic
phase unbalance a t gene frequency equilibrium even for c = 0.50. The
peak at A and the saddles a t S are expected equilibrium points if D is
ignored. Note that the equilibria actually reached do not correspond to
these points. For c = 0.50, D takes the value 0.0046 a t gene frequency
76 R. \V. ALLARD, S. K . J A W , A S D 1’. L . WORKMAN

FIG.8. Adaptive topographies for Model 1 with (a) c = 0.50, ( b ) c = 0.25, (c)
c = 0.10, and ((1) c = 0.001. Tlir isoiidi~~)ts
(tlottcd cwntoiir lines) shown in all four
topographies were computed assuming 1) = 0. Numbers along the trajectories give
the generation a t which the population point reached the position indicated. Ini-
tial grnc frcquency sets were: I, pl = p z = 0.6, D = 0.25; 11, pl = 0.6, p 2 = 0.4,
D = - 0 . 2 5 ; 111, pi=pn=0.5, D = O ; IV, p i = p n = 0 . 2 , D = O ; V, pi=p:!=OS,
D = 0 ; VI, p , = 0.9, p 2 = 0.1, D = 0 ; and VII, pl = 0.1, p~ = 0.9, D = 0. The peak
at A and saddlc~at S are given by the topography drawn assuming D = 0, whereas
the nctual peitks obtained by simiilation are nenr A in Fig. 8a and a t B-F in Figs.
Sb,c.tl. After Jain and Allard (1966).

equilibrium. In this case the dcpurture of D from zero is small and the
actual equilibrium lies near the peak ( A ) which is expected when D is
ignored. With c = 0.25, all trajectories lead to fixation of AABB. With
tight linkage ( c = O.OOl), however, there are two stable equilibrium
points, a t E and F, for which Zl = 0.2214 and D = -0.2393, respectively.
These points itre locatcd a t some distance from the single peak ( A )
 GENETICS OF INBREEDING POPULATIONS 77

expected when D is ignored, and they represent equilibria for which mean
population fitness is higher (w
= 0.866 for E and 0.879 for F) than for
the peak a t A (p= 0.712). An interesting feature revealed by these
figures is the crossing of trajectories of gene frequency change. This
indicates that populations which have identical gene frequencies may
have very different evolutionary futures owing to differences in their
evolutionary histories.
The second model (Model 2) is one of optimizing selection due to
Wright (1959). I n common with other models of optimiaing selection
this model leads to permanent gametic phase unbalance, with D < 0, a t
gene frequency equilibrium (Table 6). If D is ignored three equilibrium
points are indicated for c = 0.50: (a) 61 = $2 = 0.746; (b) = fj2 =
0.158; (c) GI = lj2 = 0.548. This corresponds to an adaptive topography
with stable equilibria with peaks a t (a) and (b) and an unstable equi-
librium a t the saddle point (c). When D is taken into account there is
found to be only a single equilibrium point a t = jj2 = 0.1764 with
D = -0.0041 (Jain and Allard, 1965). Thus D cannot be ignored, even
when it takes small values. Notice from Table 3 that gene frequencies

TABLE 3
Stable Equilibria under Model 2 for Random Mating (t = 1 .O)*

Recombination value (e)

Item? 0.50 0.25 0.10 0.01

= 0.1764 0.212Y 0.3995 0.4534

w
$1 (i2
0.9713 0.9725 0.9792 0.9962
D -0.0041 -0.0121 -0.0599 -0.1664

* After Jain
- and Allard (1965).
t f,, i2,r,and D represent equilibrium gene frequencies, mean population fitness,
and linkage disequilibrium values, respectively.

tend to be intermediate and that I D I and are larger with tighter

linkage.
Another feature of linkage which should be stressed is its general
effect of reducing the rate of approach to equilibrium, or fixation. This
can lead to the maintenance of variability over large numbers of
generations, producing quasi-stable equilibrium populations as discussed
by Lewontin (1964b) and Kimura (1965).
Wright (1965b) has shown that overdominance tends to reduce the
number of peaks (equilibrium points) in multigenic systems whereas
increasing linkage, epistasis, and asymmetry of selection tend to in-
78 R. W. ALLARD, S. K. JAIN, AND P. L. WORKMAN

crease the number of peaks. Numerical examples indicate that over-

dominance on marginal means a t individual loci is a necessary but
not a sufficient condition for stable equilibrium (Lewontin and Kojima,
1960; Jain and Allard, 1966).
The above analytical and numerical studies of multilocus models
make i t clear that linkage and epistasis can have significant effects on
population structure under random mating. I n addition, Kimura
(1956) and Bodmer and Parsons (1962) have discussed models which
illustrate the critical role of linkage in the initial stages of survival
of newly arising mutations and their maintenance in stable polymor-
phisms. An experiment on changes in the recombination system pro-
duced by selection was reported by Allard (1963). These studies sug-
gest that evolutionary adjustments in the recombination system are an
integral feature of the population structure. Next, we shall consider
the effects of linkage and epistasis on population changes under in-
breeding.

3. Selection and Linkage under Inbreeding

The case of complete selfing, with selection, for two-locus models
has been treated by Reeve (1955) and Shikata (1963) and the case
of sib mating, with selection, has been treated by Reeve and Gower
(1957). In general, however, partial inbreeding with selection leads to
systems of nonlinear equations that are difficult to handle analytically
and it has consequently been necessary to resort to numerical methods
to gain insight into the effects of linkage and epistasis under inbreeding.
This approach was used by Jain and Allard (1965, 1966) in studies of
a wide variety of selection models under mixed selfing and random
outcrossing.
The general pattern of the results indicated that increasing the
level of inbreeding has much the same effect on D as tightening
linkage, The amount of gametic phase unbalance ( D # 0) is relatively
greater with inbreeding than with random mating and it rises more or
less directly as the amount of selfing increases. This effect of inbreeding
is particularly marked with loose linkage. As a result, permanent gametic
unbalance is attained under a wider range of conditions with inbreeding
than with random mating. I n fact, under certain conditions of selection,
permanent gametic unbalance is possible a t gene frequency equilibrium
under intense inbreeding even without epistasis.
These results are illustrated by the models given in Table 4, which
represent four main types of selection (intermediate optimum, mixed
 GENETICS OF INBREEDING POPULATIONS 79

TABLE 4
Selective Values for Models 3 to 6

3. Intermediate Optimum 4. Mixed Optimum-Heterotic

0 . 6 0 . 8 1.0 0.50 0.70 0.85
0 . 8 1.0 0 . 8 0.70 1.0 0.70
1.0 0 . 8 0.6 0.85 0.70 0.50

5. Het,erotio (epistatic) 6. Mixed Under-, Overdominance

0.5 0.7 0.5 0.9 0.2 0.9
0 . 7 1.0 0 . 7 0 . 2 1.0 0 . 2
0.5 0.7 0.5 0.9 0.2 0.9

optimum-heterotic, heterotic with epistasis, and mixed under- and

overdominance). Values of F , D, and w,
a t gene frequency equi-
librium, for various specifications regarding mating system ( t ) and
linkage (c) are given in Table 5. It can be seen that values of F,
the fixation index (see Wright, 1965a; Jain and Workman, 1966),
were small and negative for t = 1 (random mating) and that they
increased steadily as t decreased. Thus, inbreeding had its usual
effect of increasing homozygosity. It can also be seen that inbreeding
led to larger values of D. The effect was small for c = 0.01 (tight
linkage) but it became increasingly pronounced as linkage became
looser. Mean population fitness is correlated with 1 D 1 and, with the
intermediate optimum model (Model 3), inbreeding led to small in-
creases in v. The relationships among the variables t, c, F , and D
are, however, not simple and their interactions with each other and
with selection as stipulated in the model influenced in complex ways.
I n addition to gametic unbalance (D) inbreeding also affects zy-
gotic associations as measured by the parameter o f a= ) [AaBb](")-
[Aa](")* [Bb](").As noted earlier, inbreeding leads to an excess of
multiple homozygotes or heterozygotes even without selection. Table
6 gives values of w for various values of c and t for the case of
no selection and for Models 3-6. It can be seen that w takes its
maximum value for different combinations of values of c and t for
the different selection models, indicating that the interrelationships
between linkage and selection are complex under inbreeding. More-
over, the effects are often large and take forms that can, in theory,
significantly influence the genetic organization of populations. Clearly,
gametic and zygotic associations as influenced by linkage and inbreed-
ing may be important factors in the maintenance of both stable and
quasi-stable gene frequency equilibria.
 TABLE 5 Fd
Values of D, w,and F at Gene Frequenry Equilibria (Models 3-6)* 9

t C F
Model 3

D
-
W F
Model 4

D i7
F
Model 5
AD F
Model 6
fD - 1
W
%-

*
m
1.0 0.01 0 -0.2375 0.9901 -0.0784 -0.2388 0.9151 -0,2702 0.1936 0.7400 -0.0524 0.2466 0.9400 .
0.10 -0.0114 -0.1359 0.9192 -0.0688 -0.1408 0.8398 -0.1724 0 0.7250 -0.0510 0.2141 0.8500
0.5750
0.50 -0.0460 -0.0274 0.9616 -0.0942 -0.0251 0.7782 -0,1724 0 0.7250 -0.0434 0
5
Q

0.90 0.01 0.0524 -0.2387 0.9910 -0.0102 -0.2396 0.9122 -0.2412 0.1970 0.7315 -0.0020 0.2468 0.9380
0.10 0.0440 -0.1469 0.9256 -0,0190 -0,1496 0.8423 -0.1366 0 0.7152 0.0048 0.2154 0.8528 ”
0.50 0.0094 -0.0339 0.8621 -0.04% -0.0307 0.7755 -0.1328 0 0.7146 0.0396 0 0.5888 *
0.30 0.01 0.5396 -0.2457 0.9966 0.4446 -0.2451 0.8830 0.0790 0.2143 0.6519 0.4822 0.2479 0.9191
0.10 0.5464 -0.2123 0.9710 0.4658 -0.2066 0.8516 0.2340 0 0.6274 0.5234 0.2268 0.8780 ‘d
0.50 0.5484 -0.1389 0.9210 0.4872 -0.1265 0.8081 0.2752 0 0.6212 0.6868 0 0.7594
0.10 0.01 0.8190 -0.2484 0.9988 0.7616 -0.2480 0.8639 0.3762 0.2201 0.5934 0.7862 0.2486 0.9078 8
0.10 0.8248 -0.2366 0.9895 0.7774 -0.2329 0.8516 0.5528 0 0.5674 0.8160 0.2331 0.8916 0
0.50 0.8326 -0.2094 0.9702 0.7978 -0.2014 0.8293 0.6106 0 0.5592 0.9022 0 0.8502 $
* Adapted from Jain and Allard (1966). E
 GENETICS O F INBREEDING POPULATIONS 81

TABLE 6
Equilibrium Values of w for the Case of No Selection and for Models 3 to 6

Selection model C 0.1 0. 3 0.9 1.o

No selection 0.01 0.0365 0.0614 0.0104 0

0.25 0.0265 0.0427 0.0079 0
0.50 0.0239 0.0376 0.0064 0

3 0.01 0.0809 0.1726 0.2321 0.2303

0.25 0.0571 0.1032 0.0468 0.0338
0.50 0.0501 0.0810 0.0284 0.0201

4 0.01 0.1031 0.1953 0.2347 0,2317

0.25 0.0675 0.1110 0.0508 0,0385
0.50 0.0581 0.0860 0.0329 0.0252

5 0.01 0.1940 0.2092 0.1492 0.1371

0.25 0.0480 0.0491 0.0075 0.0012
0.50 0.0369 0.0368 0.0054 0.0012

6 0.01 0.0951 0.1912 0.2486 0.2479

0.25 0.0629 0.1475* 0.2062* 0.2018*
0.50 0.0376 0.1038 0.1684 0,1626

* Metastable equilibria. See Jain and Allard (1966).

D. COMPLEX
MODELS
Deterministic one- or two-locus models can be extended by intro-
ducing stochastic genetic parameters or by considering populations of
finite size. Stochastic models involving the process of gene frequency
change in small random mating populations with nonoverlapping gen-
erations assume that the process can be described by a finite Markov
chain and solutions are obtained from the diffusion equation or by
simulation on digital computers (for reviews, see Kimura, 1964;
Wright, 1964). The effects of random drift arising from sampling er-
rors in small populations and the interactions of random drift with
various directed processes have been formulated in terms of effective
population size N and the distribution of gene frequencies (Wright,
1948). The effects on the distribution of gene frequencies of fluctua-
tions in the coefficients of the parameters describing the directed
processes have also been investigated. Both fluctuations due to
sampling in small populations and those due to varying selective
values affect the distribution of gene frequencies. While the former
82 R. W. ALLARD, S. K. JAIN, AND P. L. WORKMAN

can bring about fixation in small populations, the latter type of fluc-
tuation, theoretically, leads to gene frequencies in the neighborhood
of fixation but never a t fixation (quasi-fixation; Kimura, 1954).
The joint effect of small size and inbreeding on the relative rates
of gene dispersion is of interest in that variance in gene frequency
changes ( U q 2 ) can in theory be expressed in terms of the effective
population number (Kimura and Crow, 1963), or alternatively, in
terms of the fixation index, F, and its variance, oF2.Using a single
locus model, with size N = 10, 20, 100, and partial selfing of vary-
ing degrees, Allard and Hansche (1964) studied the gene dispersion
process due to drift with and without selection. When N is as small
as 10 the drift effects are so large that even heavy inbreeding only
slightly increases the rate of dispersion (Figs. 9 and 10). However,
as N increases, selection and mating system have an increasingly
greater effect on the rate of dispersion until, in populations of modest
size (say N = 100 to lOOO), drift effects tend to become small rela-
tive to the effects of the directed processes.
Similar analyses of rates of fixation in models involving six linked
loci, with and without selection, showed that rate of approach to
fixation is initially greater in small populations under inbreeding
( t < 1) than under random mating. This is particularly the case when
inbreeding is heavy and linkages are tight. However, when popula-

018-
5
0
2.

016-

Random mating (51% lost)

95% Selfing (65% lost)

3
Allele frequency

FIQ.9. Random dispersion of allelic frequencies in the Bth generation among

populations of size N = 10 (no selection, initial allelic frequencies= 0.1). Dispersion
among random mating families was determined using generation matrix techniques.
Diaperdon among inbreeding populations (s = 0.95) reflects the results of 300 Monte
Carlo simulations. After Allard and Hansche (1964).
 GENETICS OF INBREEDING POPULATIONS 83

-5 0.12
.-
%n a0
2 ace m mating (41% lost)
95 % Selfing (59 %lost
0.06
ao4
0.E

'0 0.1 0.2 0.3 0.4 0.50.6 0.7 0.8 0.9 0

Allele frequency
FIG.10. Random dispersion of allelic frequencies in the fifth generation among
population of size N = 10 (z= 1, y = 0.9; initial allelic frequencies = 0.1). Disper-
sion was determined as in Fig, 9. After Allard and Hansche (1964).
tion size is as small as 20 (Fig. l l ) , and there is no selection,
fixation is rapid whether mating occurs a t random or there is heavy
inbreeding. Perhaps the most interesting point revealed by these
studies is that tighter linkage tends t o reduce rates of gene disper-
sion, especially under inbreeding, when the genetic model is one
which leads to gametic phase unbalance (e.g., optimum models).

100

80
0
2
G.60
.-
0
-
-2 40
-

-
0

0
z
20 -- c
-c = 0.01
= 0.50

'0 20 40 60 80
Generation
FIG.11. Random drift for 6 linked loci in populations of size N = 20. All gene
frequencies were initially 0.5 and no selection was assumed. The percentage of loci
fixed was computed from 20 runs for each pair of values of s (percent self-fertilization)
and c (recombination value) (Jain, 1968a).
84 R. W. ALLARD, S. K. J A I N , AND P. L. WORKMAN

With increasingly larger population sizes, the pattern of the effects

of linkage and inbreeding for six-locus optimum, heterotic, or mixed
models becomes much the same as observed with two-locus deter-
ministic models. As population size increased, drift effects diminished
relative to the effects of selection and mating system until, for popu-
lations of size N = 1000, the distribution of gene frequencies ap-
proached those for deterministic cases (Jain, 1968).
An additional feature revealed by studies of six-locus models is a
cumulative effect of linkage along the chromosome such that genes
far apart on the chromosome can be held in gametic phase unbal-
ance by genes located between them, particularly under inbreeding.
This linkage effect was observed by Lewontin (1964a) in his study
of five-locus heterotic models, and as pointed out by him, such a
property of linked gene complexes could play a significant role in the
storage of variability under conditions leading t o permanent gametic
unbalance. Studies of six-locus models also reveal that factors such
as pleiotropic gene action and fluctuating positions of the optimum
provide for an extensive range of conditions favoring the maintenance
of unfixed loci in permanent gametic unbalance.
Stochastic processes treating the mating system ( t ) and the selection
parameters (x, y) as normal random variables have also been investi-
gated by simulation, As an example, E(E) = 0.10 and E ( o t ) = 0.05 or
0.09 were taken as expected means and variances over 100 generations,
with 20 replicate runs. The resulting variation among replicates is illus-
trated in Fig. 12 in terms of changes in the fixation index, F. Similar
1.0

0.8

0.6

0; 20 40 60 80 Ib
Generation
FIG.12. Effect of fluctuation in t (percent of outcrossing) on the fixation index, F.
Three replicates are plotted for t = 0.10, me = 0.0025 and for t = 0.10, utp = 0.0080.
No selection was assumed. Note that for this model F is relatively insensitive to
differences in the magnitude of the variance in t .
 GENETICS OF INBREEDING POPULATIONS 85

results were obtained by taking z, y as normal random variables with

known variances us2, ow2and plotting the resulting fluctuations in the
trajectories around the expected deterministic path of changes in allelic
frequency, p , and fixation index F (Jain, 1968). The average of all repli-
cates was found to approach the deterministic curve, and the variation in
F, or up2,was as expected found to be proportional to the input variances
of the parameters t, z, y. The important point is that fluctuations in the
parameters x: and y cause the bounds of the areas describing stable
equilibria (e.g., Figs. 1 and 2) to become larger such that a wider range
of conditions allowed stable polymorphisms under a model with stochastic
z and y (Jain and Marshall, 1968).

E. CONCLUSIONS
FROM THEORETICAL
ANALYSES
In this section the theory of inbreeding and selection was intro-
duced in terms of simple one-locus, deterministic models and then
developed by reference to genetic models of steadily increasing com-
plexity. One-gene models of mixed random mating and positive as-
sortative mating (e.g., selfing) in which selection confers net heterozy-
gote advantage provide a simple theoretical basis for explaining the
maintenance of balanced polymorphisms in inbreeding populations. I n
such models the net superiority of heterozygotes need not result from
superiority of heterozygotes per se but might be associated with
frequency-dependent selection, cyclic selection, or other causes. Varia-
tions in the mating system, such as the occurrence of some proportion
of negative assortative mating, can also provide a basis for main-
tenance of stable polymorphisms under otherwise predominant in-
breeding. Thus, the a priori assumption that inbreeding leads to
homozygosity and hence to genetic uniformity does not find support
in theoretical models which take both selection and mating system
into account.
Investigations of two-locus models in which the effects of linkage
and epistasis are incorporated show for many models that one of the
effects of inbreeding is to reinforce the effect of linkage in the de-
velopment and maintenance of nonrandom gametic or zygotic associa-
tions. Heterozygote advantage on marginal means is a necessary
condition for stable nontrivial equilibrium in multilocus situations
and the maintenance of genetic variability in general involves the net
selective advantage of heterozygous segments in conjunction with
interlocus interactions. There are, however, a wide range of biologically
reasonable values for the relevant parameters which allow for the
maintenance of unfixed loci under inbreeding. Investigations of mul-
86 R. W. ALLARD, S. K. J A I N , AND P. L. WORKMAN

tilocus models involving stochastic variation in the genetic parameters

and restricted population size lead to essentially the same conclu-
sions as the simple models, that is, dispersive processes reinforce the
tendency toward reduction in the within-family genetic variability and
concomitant increase in between-family genetic variability without
much reduction in the total populational variability. Migration be-
tween subdivisions of a population is a factor that also promotes
maintenance of genetic variability in the total population. Theoretical
studies suggest that the many factors which affect population struc-
ture are interrelated and that there are a multiplicity of pathways
by which inbreeding populations might evolve a coadaptive gene pool
featuring a stable high level of genetic variability.

111. Experimental Analyses of Polyrnorphisrns

A. ESTIMATION
OF PARAMETERS POLYMORPHISMS
FOR SINGLE-LOCUS

A body of evidence is now in existence to indicate that polymor-

phisms are a commonplace in both natural and agricultural species
of inbreeding plants (review in Allard, 1965). The genetic models
discussed in the previous section identify the parameters that are be-
lieved to be relevant to the development of stable polyrnorphisms
but it is obvious that the models can be little more than abstrac-
tions in the absence of quantitative estimates of the parameters ob-
tained from actual populations. In particular, it is important t o
determine whether the parameters describing the mating system and
the relative viabilities take numerical values that will explain the
polymorphisms which are observed in natural and experimental popu-
lations. In this section we shall consider methods for estimating the
genetic parameters and some of the problems involved in estimation
and interpretation of the resulting statistics. The discussion will be
restricted to plant populations in which there is predominant self-
fertilization because adequate data do not appear to be available
in other species.
1. Analyses of the Mating System
Census data on genotypic frequencies in two successive generations,
(fl("), f ~ ( ~fg("))
) , and ( f l ( n + l ) , f 2 ( # + l )f3(nf1))
, provide two degrees of
freedom and allow only two parameters to be estimated in the recurrence
equations which relate genotypic proportions in successive generations.
Consequently, before relative viabilities can be estimated it is necessary
to obtain an independent estimate of the amount of outcrossing ( t ) .
 GENETICS OF INBREEDING POPULATIONS 87

Methods of estimating t , assuming no selection and equilibrium gene

frequencies, have been given by Fyfe and Bailey (1951) and Nei and
Syaktdo (1958). I n general, however, these assumptions will not be
satisfied. Allard and Workman (1963) and Harding and Tucker (1964)
have given methods for estimating t when selection is present and
(or) equilibrium cannot be assumed. These methods are based on
estimates of gene frequency in the population and estimates of the
frequency of dominant individuals appearing in the progeny of re-
cessive individuals taken a t random from the population.
It has been assumed in the formulation of the above models that
outcrossing occurs a t random. A number of studies suggest that this
assumption may not be justified (Guitierrez and Sprague, 1959;
Barnes and Cleveland, 1963; Harding and Tucker, 1964). If the out-
crossing that occurs tends to be between unlike individuals, that is,
the outcrossing involves negative assortative mating, then, as shown
by Workman (1964), lower levels of heteroxygote advantage suffice
to maintain polymorphisms than is the case with random outcrossing.
Conversely, if part of the outcrossing involves positive assortative
mating, greater heteroxygote advantage is required. Some of the diffi-
culties pertaining to a correct assessment of the mating systems are
made less critical when an “effective” amount of random outcrossing
rather than the actual amount of outcrossing is estimated (Allard
and Workman, 1963).
The estimate of t used in the estimation of the relative viabili-
ties should, if possible, be the average of a series of estimates. Studies
in a wide variety of organisms (e.g., the French bean, Bateman, 1952;
safflower, Claassen, 1950; flax, Dillman, 1938) show the actual fre-
quency of outcrossing is often quite variable. It can differ between
years or locations within a year for a given marker locus, and i t can
differ from locus to locus. Some of the results of Harding and Tucker
(1964) showing variation in the estimate of t in experimental lima
bean populations are summarized in Table 7. The amount of out-
crossing can also vary with respect to sex. Smeltzer analyzed out-
crossing in a population made up of five varieties of sorghum. Each
line was considered in terms of its contribution to the outcrossing
both as a pollen source and an ovule source. Significant differences
between lines with respect to sex were noted as shown in Table 8,
2.. Estimation of Relative Viabilities
As noted earlier, selection models vary according to the time and
mode of action of the selection and the stage of the life cycle a t which
the genotypic frequencies are determined. In estimating relative via-
88 R. W. ALLARD, S. K. JAIN, AND P. L. WORKMAN

TABLE 7
Estimates of Outorossing in Three Experimental Lima Bean Populations*
Year

Populationf Locus 1957 1958 1959 1960

I D/d - - 0.0892 0.0730

Wlw - - 0.1792 0.2979

I1 Did - 0.0254 - 0.0479

Vlv - 0.0478 0.0180 0.0293
c/c 0.0806 0.0632 0.0722 0.1710
Srlsr - - 0.19291 0.0924
ss - - - 0.1173
ss - - 0.5247$ 0.1667

Location and year

(Davis) (Irvine) (Davis) (Irvine)

Population Locus 1961 1961 1962 1962

I11 CIC 0.0539 0.0615 0.0712 0.0053

85 - - 0.1718 0.0295
88 - - 0.0834 0.0292

* Adapted from Hardmg and Tucker (1964).

t Populations I and I1 grown at Davis, California;population I11 grown at Davis and
Irvine, California.
$ Standard error of estimates > O . 05.

bilities it is therefore necessary to choose the model which most

closely fits the biological situation. Estimators of 2 and y, and their
variances, can then be obtained from the recurrence equations by
standard methods (e.g., by the method of maximum likelihood, or
least-squares analysis).
These estimates of 5 and y are assumed t o represent so-called
net Darwinian fitnesses defined in terms of the relative number of
progeny left by different genotypes, These estimates will be biased
if the selection has not been completed by the time of the genotypic
census, as shown by Prout (1965). For the case of predominantly
self-fertilixing populations, with which we are concerned here, Work-
man and Jain (1966) have shown that the errors which attend the
estimation of selective values at a partially selected stage are too
small to be of any practical significance.
Examples of estimators for differing modes of selection and time
of genotypic census are give11 by Allard and Workman (1963) and
 GENETICS OF INBREEDING POPULATIONS 89
TABLE 8
Estimates of Outcrossing in Mixtures of Grain Sorghum Varietiea at Davis, California*

Outcrossing as estimated Proportional contribution to

from female parentst outcrossing aa male parentst

Variety 1959 1963 1959 1963

DD38 0.1321 0.0721 0.0362 0.0857

CK60 0.0279 0.0258 0.1335 0.2200
DWD 0.0508 0.0447 0.1014 0.0859
DW39 0.0506 0.0246 0.1733 0.1667
EH 0.0721 0.0355 0.1563 0.4417
BDS - - 0.3993 -
~ ~~ ~~

* Smelteer (1965).
?All estimates have been adjusted to compensate for differences in the relative pro-
portions of the linea in the mixture.
$ Accurate clas~ificationwas found to be difficult in the 1959 experiment. This geno-
type waa therefore not included in 1963.

Workman and Jain (1966). Table 9 gives numerical values of the

selection parameters obtained for various marker loci in some experi-
mental populations of liina beans and barley and in some natural
populations of wild oats, These data indicate that heterozygotes
nearly always produce more viable progeny than homozygotes and
reference to the phase diagrams for t = 0.05 (Fig. 2) shows that the
observed level of heterozygote advantage is sufficient to account for
permanent maintenance of the polymorphisms. The data of Table 9
can be considered in another way. Inbreeding species have usually
been described in terms of the amount or intensity of inbreeding
without consideration of the effects of selection. Allard and Workman
(1963) defined a parameter, t*, the effective amount of outcrossing,
which permits one to equate the observed population with parameters
x, y, and t, t o a population in which x = y = 1 and there is out-
crossing in the proportion t*. Although the actual t was of the order
of 3 to 5% for the lima bean populations, t" was approximately 13
to 14%. Thus, heterozygote advantage had the effect of substantially
reducing the amount of homozygosity that would have resulted if in-
breeding had been the major factor in determining variability.
Some other aspects of the problems involved in the estimation of
the genetic parameters should be mentioned. Allard and Workman
(1963) studied fluctuations in estimates of x and y over several gen-
erations in experimental lima bean populations (Fig. 13). Their re-
sults show that year-to-year fluctuations in environment lead to sharp
90 R. W. ALLARD, S. K. JAIN, AND P. L. WORKMAN

TABLE 9
Estimated Relative Viabdities for the Two Homozygotea and the Heterozygotea
for Chromosome Segments Marked by Major Genes
Mean Relative viabilities of
percentage of
Locus or population outorwing AIAI AiAt A 4 2

Lima bean'
s/s 4 0.53 1 .oo 0.35
Dld 5 0.58 1 .oo 0.47
SIs 3 0.66 1 .oo 0.66
Barleyt
Blb 1.06 1.00 1.31
Sb 0.81 1 .oo 0.96
Wlg 1.04 1.00 0.82
Ele 0.47 1.00 0.59
Bllhl 0.61 1 .oo 0.54
Rlr 0.82 1 .oo 0.68
Bt/bt 0.96 1 .oo 1.06
Sh./sh 0.71 1.00 0.63
Wild oatst (lemma color locus)
Population 1 (1960) 0.50 1 .oo 0.30
Population 1 (1960) 0.64 1.00 0.56
Population 5 (1960) 0.41 1.00 0.44
Population 7 (1961) 0.57 1 .oo 0.54

*After Allard and Hansche (1964).

t After Jain and Allard (1960).
2 After Imam and Allard (1965).

shifts in the relative viabilities of different genotypes suggesting that

caution must be exercised in interpreting results based on data from
a limited sample of generations. Sampling errors, particularly if one
of the genotypes is in low frequency, can also be an important source
of fluctuations in the estimates of the viabilities (Workman and
Allard, 1964).
Frequency-dependent selection may be involved but go unrecog-
nired. Harding et al. (1966) submitted one of the lima bean popu-
lations studied by Allard and Workman (1963) to a more detailed
experimental analysis and established that the fitness of heterozy-
gotes increased strikingly as their frequency decreased in the popula-
tion (see Fig. 14). Consequently, it seems likely that some of the
observed fluctuations in x and 9 reported by Allard and Workman
may be ascribed to the effect of frequency-dependent selection.
Such selection can be a potent force in retaining variability in popu-
 GENETICS O F INBREEDING POPULATIONS 91

x, y 0.8
0.4

.,as[
,:-:\' ' Pdpuiknk ' Li
---- Y
.

0.4

0.0 k ,ik2iAiMizi
Year
7imi m / !
k

FIG.13. Year-to-year fluctuations in estimates of selective values for the S/s locus
in three experimental lima bean populations. After Allard and Workman (1963).

All populations
0 26
c% 4.0 - 03
m
=.
N 016

f
E 3.0-
c
r

-F
v)

2.0-
._
r
200

s
._
L

QCT 1.0- Y e = 3.38 - 16.77):

s b = 2.74

FIG.14. Relationship between frequency of heterozygotes and their fitness relative

to homozygotes for the S/s locus in lima bean populations. Homozygotes were as-
signed a relative fitness of 1.0. After Harding et al. (1966).

lations when the selective value of any genotype increases as its fre-
quency decreases.
Workman and Jain (1966) have shown that restriction of selection
to one sex affects rate of approach to equilibrium but that the ulti-
mate equilibrium reached is the same, for given 2, y, and t, as when
92 R. W. ALLARD, S. K. J A I N , AND P. L. WORKMAN

selection affects both sexes equally. Thus, it is possible that differ-

ent models may appear to fit a given set of experimental data, par-
ticularly for populations near equilibrium. This can lead to serious
problems of interpretation, particularly if the data available span
only a limited number of generations.
The genetic basis for the selfing can vary from control by a
single locus to control by a large number of loci (e.g., East, 1919;
Clarke, 1935; Holden and Bond, 1960). When a single locus or very
few loci control selfing there will be little or no outcrossing a t all
in many families and the distribution of outcrossing among families
may be satisfied approximately by a Poisson or negative binomial
distribution. Control by a large number of loci is expected to lead
to an approximately normal distribution of amount of outcrossing
in different families. Differences in the genetic control of selfing lead
to considerable differences in the observed variation in the estimate
of t , especially in experiments in which only a few families a r e
studied.
Linkage is another factor which affects estimates of the relevant
parameters. Suppose that a marker locus under study (say A ) is
linked to a locus ( B ) a t which variability is retained due to hetero-
zygote advantage. The estimates of x and y for locus A then de-
pend not only on the relative viabilities of the genotypes a t the A
locus but also on the intensity of linkage, the viabilities a t the locus
B, and the presence or absence of linkage disequilibrium between
loci A and B. This problem has been considered indirectly by Kimura
(1956) and by Parsons (1957) for the case of full selfing.
Despite the caution that must be observed in analyzing estimates
of genetic parameters, certain conclusions can be drawn with con-
fidence. The general pattern is that heterozygotes usually show con-
siderable selective advantage over the corresponding homozygotes for
marker loci and the net heterozygote advantage observed is often at
levels which permit permanent retention of polymorphisms, even in
very heavily inbreeding populations.

B. COMPLEX
POLYMORPHISMS
The models discussed in Sections 11, C and 11, D indicate some
of the possibilities regarding equilibrium states and progress toward
equilibrium in populations t.hat are polymorphic for more than a single
Mendelian unit. However, a number of difficulties are encountered
when attempts are made to obtain numerical estimates of the rele-
vant parameters in actual populations. Not the least of these diffi-
 GENETICS OF INBREEDING POPULATIONS 93
culties is the sampling problem associated with the necessity of es-
timating the frequencies of nine genotypes when two polymorphisms
are studied simultaneously, 27 genotypes with three-unit polymor-
phisms and discouragingly large numbers of genotypes with more com-
plex cases. It is not surprising that experiment has lagged behind
the development of theory.
Despite the difficulties, there have been several studies of complex
polymorphisms in experimental populations and also a few studies
in natural populations. Examples are wing characters of the grass-
hopper Pazatettix by Fisher (1939),shell color of Cepaea nemoralis
studied by Lamotte (1951) and Gain and Sheppard (1952),the in-
version polymorphisms of Drosophilia robusta reported by Levitan
(1955, 1958), the complex mimicry patterns in certain butterflies
(Sheppard, 1959), mutant markers in D . melunogaster studied by Can-
non (1963) and further analyzed by Lewontin (1964a),marked chro-
mosomes of D . pseudoobscura studied by Spassky et al. (1965), and
lethals in D. uri2listoni studied by Magahaes et al. (1964). The gen-
eral pattern of results with these presumably random mating species
indicates that different Mendelian units are often not independent in
their fates in populations. Instead interactions between different units
frequently affect selective values and hence the organization of the
population genotype.
Experimental data on complex polymorphisms are also scanty in
inbreeding populations. One of the most thoroughly analyzed cases oc-
curs in the Australian grasshopper, Moraba scurra. White (1957) and
Lewontin and White (1960) presented evidence that pericentric in-
versions carried on two different chromosome pairs in this species are
not combined a t random in adult male individuals. Their analysis of
the deviations from the expected random combinations provided evi-
dence for genetic interaction between the two systems of cytological
polymorphism as they affect viability. However, these wingless grass-
hoppers have limited mobility and if it is assumed that there is a
tendency for close neighbors to mate and that close neighbors are
likely to be relatives, i t is expected that some equilibrium level of
inbreeding will develop in populations of these sedentary insects. As
discussed by Allard and Wehrhahn (1964) the assumption of a low
level of inbreeding (0.10< F < 0.15) gave a good fit t o the data.
Thus epistatic interactions can be a feature of an apparently stable non-
trivial gene frequency equilibrium in populations in which inbreeding
arises as a result of isolation by distance.
The polymorphisms in experimental populations of lima beans dis-
cussed previously are favorable materials for the study of interactions
94 R. W. ALLARD, S. K. JAIN, AND P. L. WORKMAN

between different Mendelian units under heavy inbreeding and they

have been used for that purpose by Harding and Allard (in preparation).
In the population chosen for detailed study it was possible to score
seven marker loci individually and 21 possible two-locus combina-
tions. In 16 of the 21 two-locus cases studied, certain genotypes were
in significant excess or deficiency. Two of the loci are located on the
same chromosome and the type of gametic unbalance observed for
these two loci ( D > 0, loci in coupling in the initial population)
suggested that the unbalance represented undissipated coupling as-
sociations due to the linkage. However, the remaining 15 cases of
zygotic association cannot be attributed to linkage because the loci
involved are known to be located on different chromosomes. The
basis of these associations is therefore presumably either the mating
system effect of Bennett and Binet (see Section 11), or epistatic in-
teractions between the marked linkage blocks.
More extensive and more precise numerical estimates are clearly
necessary before the relative importance of the parameters identified
in the two-locus models of the previous section can be assessed. Re-
sults obtained thus far point, however, to the possibility, indeed the
probability, that the destiny of genes in populations depends not
only on the effects of the genes themselves but also on their inter-
actions with all of the factors that affect the organization of the gene
pool. The mating system appears t o be only one among a number
of interdependent factors that affect the organization of populations,
and its effects, no less than those of the other factors, are integrated
in the whole. As an example, the effects of inbreeding due to re-
stricted population size in Cepaea nemoralis appear to be mitigated
by multiple matings and sperm storage mechanisms (Murray, 1964).
The discovery that complex interactions are a feature of two-locus
systems thus provides an experimental basis for extrapolation to the
more complicated systems of interaction which are presumably in-
volved in the control of continuously varying characters and which
in total lead presumably to the “coadapted” population genotype.

IV. Genetic Variability in Quantitative Characters

Populations of inbreeding species have commonly been supposed to
be highly uniform respecting measurement characters. However, it
has been known since ancient times that selection is effective in im-
proving performance and modifying various other characteristics in
populations of wheat, barley, oats, and other predominantly self-
pollinated crop plants (Roberts, 1929). The first recorded accounts
showing that selection is effective in changing inbreeding populations
 GENETICS OF INBREEDING POPULATIONS 95
appeared in the eighteenth century as a result of methodical at-
tempts at plant breeding. Van Mons in Belgium, Knight in England,
and Cooper in America demonstrated that selection can lead to worth-
while improvement in agricultural performance. More fully docu-
mented evidence followed in the early nineteenth century. John Le
Couteur, a farmer of the Isle of Jersey, described the diversity of
plant types in his wheat field and established that there were differ-
ences in the agricultural value of various selections. B y the late
nineteenth century selection in “land varieties” had become an es-
tablished method of plant breeding. Land varieties still exist in some
of the less advanced agricultural areas and it is easily demonstrated
by progeny tests that such populations contain many different geno-
types. Harlan (1957), in discussing variability in barley, states that
in each local area “there has evolved a type peculiarly fitted for
conditions as they exist there. Slight changes in altitude are accom-
panied by corresponding changes in the barleys, The barleys from
each tiny area are made up of large numbers of strains that look
much alike, but that may differ greatly in ways useful to the plant
breeder. Even out on the plains where the superficial appearance of
the crop may be the same over a large area, these constituents are
present in endless variety and shift as one goes to drier or colder
sections.” Thus, it has been recognized since the earliest days of cul-
tivation that populations of predominantly self-pollinating agricul-
tural plants are repositories of great amounts of genetic variability.
It is only recently that detailed quantitative studies of variability
in measurement characters have been undertaken in inbreeding popu-
lations and that adequate estimates have been made of the com-
ponents of this variability under population conditions. I n these
studies the basic procedure has been to collect seeds from random
plants in a number of populations. These seeds were then sown to
establish families of 20 or more plants in replicated experiments in a
common nursery environment and each individual in each family was
measured for various morphological and physiological attributes.
Data from such studies permit quantification of three aspects of
genetic variability, that associated with broad geographical regions,
that associated with specific sites within regions , and variability oc-
curring within sites. The general pattern of variability that exists in
inbreeding species can be established by considering a few examples.

VARIABILITY
A. GEOGRAPHICAL
Knowles (1943) made collections of Soft Chess (Bromus mollis),
a highly self-pollinated annual grass, along a transect from the cold,
96 R. W. ALLARD, 6 . K. JAIN, AND P. L. WORKMAN

humid coastal regions of northern California t o the warmer semi-

arid Sacramento Valley to the interior. Data from progenies grown
in a common environment (Yolo County) revealed a clinal pattern
of variation (Table 10). Size of plant, as measured by height, and
time required to reach maturity, as measured by days to flowering,
decreased progressively but tillering capacity increased with increas-
ing aridity. Similar clines have been found in all other inbreeding
species for which adequate measurement data are available. Thus,
for example, seed size in Bur Clover (Medicago hispidu) decreases
progressively with increasing altitude, petal length of Erodium &u-
turiurn increases with increasing rainfall, leaf length in Foxtail barley
(Hordeurn nodosum) decreases from north to south in California,
TABLE 10
Genetic Variability between and within Populations of Bromus mollis'
- ~ ~~

Location Number Daya t o heading Plant height Tiller number

of of
populationt progenies Mean Range Mean Range Mean Range

Del Norte 16 186 174-201 88 5&102 31 22-45

Humboldt 23 174 163-183 77 49-103 27 17-43
Contra Costa 18 164 155-169 80 72-89 29 2043
Solano 44 153 151-164 67 55-92 33 23-62
Yolo 9 154 151-160 71 57-86 40 32-49
Sutter 6 151 151 60 58-63 35 30-40

* After Knowles 11943).

t Listed in increaeing order of aridity.
and wild oats (Avena fatua) from deep fertile soils are more robust
and tiller more profusely than wild oats from infertile soils (Allard,
1956).
Superimposed on this clinal variability is another and much more
striking pattern of variation. This is a patchwork or mosaic pattern
which is reflected in sharp differentiations between different sites
(populations) within regions. Such variation is illustrated by a sample
of data on wild oats (Avena jatua) from the studies of Imam and
Allard (1965) and Jain and Marshall (1967). Comparisons of collec-
tions taken from a number of sites within one of the regions studied
(the Coast Range of California) indicated sharp differentiations be-
tween sites (Table ll). For example, flowering was earlier for site 4
than for the other sites. The habitat in this case was a west-facing
arid hillside which might be expected to be unfavorable for late ma-
turing genotypes in a Mediterranean climate.
 GENETICS OF INBREEDING POPULATIONS 97

TABLE 11
Estimated Means for Different Natural Populations of Wild Oats
within a Geographical Region

Spikelet Panicle
Flowering time Height number length
Population* (date in April) (cm) Populationt per panicle (cm)

4 11.5 112 El 7.7 9.9

5 16.5 103 E2 6.3 8.2
6 25.4 107 FI 14.0 15.2

* After Imam and Allard (1965).

t After Jain and Marshall (1966).

Differentiations are often striking over very short distances. Within one
site heading time changed more than 15 days in a distance of 5
meters, apparently in association with local topography. Progeny of
planta taken from a flat area immediately above a short steep slope
flowered on April 18 on the average, those from the steep slope
flowered on April 11, and those from the well-watered area a t the
base of the slope flowered on April 26.
Results for many different species follow a pattern that can be sum-
marized as follows, Inbreeding species of plants are differentiated geo-
graphically and the differentiations observed appear to be those which
provide each population with the adaptive properties needed to meet
the requirements of the local environment. The predominant pattern
is one of striking local differentiations, often involving areas of a
few square meters or less, but clinal gradients also occur in association
with progressive changes in rainfall, temperature, and other factors of
the physical or biotic environment.

WITHIN POPULATIONS
B. VARIABILITY
Perhaps the first quantitative study of genetic variability arising
from differences among families whose seed parents were single plants
taken from the same population was Johannsen’s (see Johannsen,
1926) classic experiment with the Princess variety of garden bean
(Phaseolus vulgaris). Johannsen observed that a great deal of varia-
tion occurred in the unselected population with respect to seed size
and other characters and, by dividing the seed into classes according
to weight, he was able to show that smaller seeds in general produced
small-seeded and large seeds produced large-seeded progeny. B u t he
also observed that seeds in a given class produced progeny with
98 It. W. ALLARD, 6. K. JAIN, AND P. L. WORKMAN

rather widely different weights and this observation led him to compare
mean seed weights of families. Johannsen established families by grow-
ing separately the progenies of 19 plants from the original population
and he found that each family had a characteristic seed weight form-
ing a continuous series from 35.1 centigrams for the smallest family
to 64.2 centigrams for the largest family. It is not clear whether
these 19 families represented a random or selected sample from the
original population. However, the nearly twofold range in seed size
from smallest to largest family established that the original population
was highly variable respecting this character.
Knowles’ (1943) study of Soft Chess demonstrated that similar vari-
ability occurs in natural populations of inbreeding species. It can be
seen from Table 10 that families from the same population varied
over very wide ranges respecting flowering time, height, and number of
tillers. The difference between earliest and latest family was as much as
25 days in flowering time, the range from shortest to tallest was more
than 50 inches, and there was more than twofold difference in tillering
capacity. Knowles concluded from his measurements and observations
on several characters that each family is genetically distinct from each
other family, implying that few if any individuals in the original
populations had exactly the same genotype.
Knowles’ conclusion regarding the genetic distinctiveness of each in-
dividual has been placed on a firmer basis by subsequent studies in other
species. In wild oats, for example, very wide differences occur among
progenies from the same population with respect to flowering time,
height, germination percentage, number of tillers, length of panicle,
seed size, and other characters. When the means of families are com-
pared for any single character, using a Duncan multiple range test,
it is found that the families from a single population fall into a t least
two but usually more groups which differ significantly from one an-
<
other ( P 0.05). Families in different groups clearly have different
genotypes. When two characters are considered simultaneously i t is
found that families which fall in the same group for one character
(say flowering time) often fall into different groups with respect t o the
second character (say height); such families also do not have the
same genotype. Extension of this sort of analysis over additional
characters ultimately separates the families into as many groups as
there are families. This measure of intrapopulation variability there-
fore leads to the conclusion that each population includes plants of
many different genotypes respecting measurement characters. No single
genotype is represented by more than a few individuals and perhaps
 GENETICS O F INBREEDING POPULATIONS 99

every individual in t,he population differs genotypically from every

other individual.

C. VARIABILITY
WITHIN FAMILIES

Again, the first precise quantitative study appears to have been

Johannsen’s study of the Princess variety of the garden bean. This
study is too well known to require detailed description. It is sufficient
to say that six generations of selection within families failed to estab-
lish a significant difference between lines selected for light seeds and
lines selected for heavy seeds. This result led t o the formulation of the
pure-line theory and it is also the basis for many of the deductions that
have been made concerning the genetic structure of inbreeding popula-
tions.
More recently, evidence has been accumulating that Johannsen’s
population of garden beans represented a special and apparently very
rare situation, that of a population in which selective pressures have
been low and reproduction has been exclusively by self-fertilization for
many generations. There have been many experiments to indicate that
continued propagation of selfed families under conditions of low com-
petition leads to high levels of homozygosity and these experiments
provide a basis for the result observed by Johannsen. In the popula-
tion studied by Johannsen a history of propagation under garden condi-
tions may have provided for high survival and the absence of out-
crossing might have been due t o lack of appropriate insect pollinators
in Northern Europe. However, the garden bean is typically not com-
pletely self- fertilized in the more southerly areas where this species
is grown on a large scale. Instead, a low rate of outcrossing (less
than 1%)is usual and in some years there may be as much as 5% of
outcrossing between certain genotypes. Furthqrmore, plantings are
usually dense, which suggests that competition may be intense. The
effect of this low order of outcrossing and presumed competition on
within-family variability was determined by Allard and Golden
(1954), who conducted an experiment analogous to that of Johannsen.
A lot of foundation seed of the Red Kidney bean ( P . vulgaris) provided
by the California Crop Improvement Association was sown and seed
weight index (grams/100 seeds) was determined for each of 100 plants.
The range in seed weight index varied from 49.1 to 54.3. This repre-
sents a smaller range of variability than Johannsen found in the
Princess variety and i t no doubt reflects the stringent selection for
conformity to “type” that characterizes present day seed practices.
100 R. W. ALLARD, S. K. JAIN, AND P. L. WORKMAN

Twenty of these plants were then chosen by random methods and seeds
from each plant were sown to establish 20 families each containing
approximately 25 individuals. Seed-weight indexes were determined
a t maturity and seeds from the plant with the smallest seed-weight
index and from the plant with the largest seed-weight index were sown
the next year to establish minus and plus selection lines. This process
was repeated for four generations. The results of selection in the
families are presented in Table 12. I n contrast to Johannsen’s result
a significant response to selection occurred in each of the 20 families.

TABLE 12
Effect of Four Generations of Selection for Seed Weight within Families 7 and 16
of the Red Kidney Bean*

Family 7, Family 16,

seed weight index seed weight index
(grams/lOO seeds) (grams/100 seeds)

Generation Small Large Difference Small Large Difference

0 52.2 52.2 0 51.1 51.1 0

1 50.1 52.3 2.2t 50.0 53.1 3.1$
2 50.8 52.9 2.lt 51.2 55.5 4.3f
3 50.3 52.6 2.3$ 49.0 54.1 5.1$
4 49.2 52.9 3.7$ 48.1 56.1 6.0$

* Among 20 families studied, response to selection was smallest in family 7 and greatest
in family 16. From Allard and Golden (1954).
t Difference significant P 5 0.05.
3 Difference signifirant P 5 0.01.

The greatest progress occurred in family 16. I n that family the final
difference between the small and large selection lines was 6.0 g/100
seeds, which represents a spread of nearly 12% in terms of the base
weight of line 16 (51.1 g/100 seeds). The smallest response was in
family 7 but even in that family the final difference between small
and large was 3.7 g/100 seeds, or about 7% of the base weight of the
family (52.2 g/100 seeds).
Similar responses to plus and minus selection within families were
obtained in a companion experiment with the Henderson variety of the
lima bean (Phaseolus lunatus). In this case the base population was
a highly selected elite seed lot obtained from a commercial seedsman.
The seed weight index of the base population was 43.8 and the mean
observed seed weights for progenies of plants derived from 100 random
seeds taken from the base population varied from 39.8 to 46.1 g/100
 GENETICS OF INBREEDING POPULATIONS 101

seeds. Selection for low and high seed-weight index in 20 families led
t o statistically significant responses in each case. The smallest response
observed after four generations of selection was 4.1 and the largest re-
sponse was 8.4 g/lOO seeds (Table 13).
It can be inferred from studies of intervarietal rates of outcrossing
that a very high proportion, probably more than 99%) of the fertiliza-
tions result from selfing in both the Red Kidney and Henderson varie-
ties. However, the precise rate of outcrossing which occurs within these
varieties under population conditions cannot be determined because

TABLE 13
Data Showing Response to within Family Selection in the Henderson Variety
of the Lima Bean (Pha-seoluslunatus)*

Family 11, Family 8,

seed weight. index seed weight index
(gramsj100 seeds) (grarrm/100 seeds)

Generation Small Large Difference Small Large Difference

43.6 43.6 0 41.5 41.5 0

41.8 43.8 2.0t 41.3 42.6 1.3
41.9 43.9 2.0t 40.8 42.6 1.8t
41.5 45.1 3.6$ 39.1 43.4 4.3t
41.7 45.1 3.4$ 37.9 43.3 5.4t
41.6 45.7 4.1$ 37.1 45.5 8.4$

* Family 11 showed the smallest response and family 8 the largest response among 20
families studied. From Allard and Golden (1954).
t Difference significant, p 5 0.05.
1Difference significant, p 5 0.01.

both varieties are monomorphic. It is therefore of particular interest t o

examine response to selection within families in populations in which
precise quantitative data on outcrossing rates have been obtained from
studies of polymorphisms.
I n their study of variability in wild oats, Imam and Allard (1965)
noted that, in addition to the differences between families that were
considered in the previous section, there was also distinct variability
within the great majority of families. A selection experiment was con-
ducted to determine whether at least part of the variability for meas-
urement characters could be ascribed t o heterozygosity of the plants
taken from the natural population. The selection experiment was based
on 10 plants chosen by random methods from a natural population in
which the outcrossing rate varied from 1.49 to 6.68%) depending on
102 R. W. ALLARD, S . K. JAIN, AND P. L. WORKMAN

the year and the marker gene used to determine rate of outcrossing.
A progeny of 20 individuals was grown from the seeds harvested from
each of these 10 plants and within each progeny the individual with
the lowest score and the individual with the highest score were selected
respecting the following three characters : prostrate vs. erect growth
habit; early vs. late flowering time; short vs. tall. The next season
paired plots were grown, one plot of each pair containing 10 plants
derived from the extreme low individual and the other containing 10
plants derived from the extreme high individual for each character.
The 30 paired plots for each character were arranged in a randomized
design, replicated twice. Each replication also included a plot derived
from reserve seed of each of the ten original plants. The results of
measurements taken on an individual plant basis in these 180 plots
are summarized in Table 14.
TABLE 14
Effects of a Single Generation of High and Low Selection on Three Quantitative
Characters in Lines Derived from 10 Wild Oat Plants Taken
from a Natural Population*
Unselected lines Low selected lines High selected lines

Character Range Mean Range Mean Range Mean

Growth habitt 2.70-3.71 3.26 1.89-2.46 2.2418 3.34-4.11 3.68

Days to 10.63-16.99 15.62 4 . 6 5 1 5 . 5 1 11.1918 16.16-19.18 17.72
flowering
Height 93.1-115.2 105.6 82.5-106.3 89.3$0 107.3-139.4 116.2

* After Imam and Allard (1965).

t Scored on an arbritrary scale; 1 = prostrate; 5 = erect.
1Significantly different (P< 0.05) from the mean of the unselected lines.
Q Significantly different (P < 0.01) from the mean of the high selectians.

Two aspects of the results can be considered. First the data given in
Table 14 show that the mean of the 10 low selected lines was sig-
nificantly lower ( P < 0.05) than the mean of the 10 high selected lines
for all three characters, Also the means of both the low and high
selected lines were significantly different from the means of the un-
selected lines for each character. The second aspect of the results is
the divergence within individual families produced by the selection.
For growth habit and flowering time the difference between the high
and low selections was significant in each of the 10 families and the
difference between the selected and unselected lines was also significant
in all comparisons. The response to selection for height was only slightly
 GENETICS OF INBREEDING POPULATIONS 103
less clear-cut. For this character one generation of selection failed
to produce a significant difference between high and low selection lines
in only two families and the difference between unselected and selected
lines failed to reach significance in only 8 of the 20 comparisons that
could be made.
The results of various experiments involving selection within families
in several different heavily inbreeding species can be summarized as
follows. Individuals within a population are often heterozygous for
many genes governing quantitative characters. The variability that is
commonly observed within individual families is therefore not ex-
clusively environmental or developmental but much of it can be
ascribed to segregation. It is clear that the genetic organization of a
population cannot be deduced solely from the amount of outcrossing
that occurs in the population.

V. Responses to Selection
The responses of inbreeding populations to mass selection, both
artificial and natural, have been studied in many different species.
I n this section we shall consider some representative experiments which
illustrate the main patterns of response.

A. COMPETITION OF PURELINES
IN MIXTURES

There have been a number of studies of natural selection in popula-

tions synthesized by mechanical mixing of two or more pure lines
(e.g., Harlan and Martini, 1938; Suneson, 1949; Allard e t al., 1966).
The results of these experiments have been remarkably consistent; al-
most without exception one pure line rapidly became predominant and
all other pure lines were drastically reduced in frequency. A typical
result is given in Fig. 15. I n this experiment the four competing geno-
types were fully vigorous and well adapted commercial varieties.
Nevertheless, relative selective values (Fig. 15) show that the poorest
competitor produced 23% fewer progeny on the average than the best
competitor (Workman and Allard, 1964). I n some other experiments
selective differences were much greater. In certain cases genotypes that
were not obviously defective left only of the order of 5% as many
progeny as the best competitor. It is obvious that selection of very
great intensity occurs under population conditions.
Results such as these lead to the expectation that local populations
of highly inbreeding populations should soon be reduced t o a single
highly competitive genotype. However, we have already seen that both
104 R. W. ALLARD, S. K. JAIN, AND P. L. WORKMAN

natural and agricultural populations of inbreeders are variable and

that, in fact, no single genotype makes up any substantial proportion
of the population. One line of evidence that provides a partial answer
to this seeming paradox is a tendency noted in several populations
for the selective values of certain genotypes t o increase as these geno-
types become increasingly rare in the population. For example, Work-
man and Allard (1964) found evidence to suggest that the selective
values of Club Mariout, Vaughn, and Hero were higher when they
were in low than when they were in high frequency. This type of
frequency-dependent selection could be a powerful force in maintain-
ing many genotypes in a population, each in low frequency.

100,
Selected Values

. ._

._______-
--*'.
220
10
0
\--.
-.
I 2 3 4 5 6 7 8 9 10 I I f2 13 14 15 16
Generation
ha. 15. Effect of natural selection on genot,ypic frequencies in a mixture of four
pure lines of barley. Data of Suneson (1949).

Another type of experiment provides additional information which is

helpful in understanding the large number of genotypes which occur in
natural populations. In the experiments that have just been con-
sidered, the integrity of the small number of competing pure lines
was protected by removing interline hybrids. If, however, the occa-
sional hybrids that occur were to be left in the population, the situ-
ation would correspond more closely to that in natural populations,
or in unselected agricultural populations. An experiment with a popula-
tion of lima beans which was synthesized by mechanical mixing of
two pure lines illustrates the type of result which is obtained. The
two homozygous lines that were mixed were differentiated by five
major genes affecting conspicuous morphological characters so that the
great majority of individuals derived from hybrids between the original
homozygotes could be identifled easily. Consequently it was possible t o
classify the population, generation by generation, into parental and
 GENETICS OF INBRREDING POPULATIONS 105
nonparental types, as shown in Fig. 16. Within a few generations the
original pure lines had been swamped by a hybrid swarm, even though
the proportion of outcrossing in the population averaged only about
5 % . Moreover, the population had become enormously variable for
quantitative characters by the fifth or sixth generation. Observations
and measurements on progenies derived from single random plants
taken from the population in the ninth and tenth generations indi-
cated that each progeny was characterized by a constellation of char-
acteristics, such as time to maturity, leaflet color (yellow green to
dark green), leaflet size, seed size, and other more subtle differences,

.--

c
0
m- I
I
560- I
I

Parental line L 121

" O I 2 3 4 5 6 7 8 9 1 0 1 1
Generation

FIG.16. Proportions of parental and nonparental types in a lima bean population

which was synthesized by the mechanical mixing of two pure lines (L21 and L121).
The average amount of outcrossing in this population was 5%. After Jans and
Allard (in preparation).

which distinguished it from all other progenies. It is clear that com-

petition among homozygotes rarely if ever occurs in pure form in
natural populations because even low levels of outcrossing set the
stage for conversion of simple mixtures of limited numbers of homozy-
gotes into a complex and dynamic populational system within a few
generations.

B. RESPONSES SELECTION
TO NATURAL

There have been a number of studies of both short- and long-term

responses to natural selection in populations that were developed from
intercrosses between different genotypes. An early experiment which
provided quantitative data on two aspects of response to natural
106 R. W. ALLARD, S. K. JAIN, AND P. L. WORKMAN

selection was reported by Adair and Jones (1946). These investigators

blended F2 seed of 20 different rice hybrids and grew the resulting
population without conscious selection at Biggs, California, Stuttgart,
Arkansas, and Beaumont, Texas, for eight generations. A random
sample of 1000 seeds was then drawn from the population a t each
location, grown a t Stuttgart, Arkansas, and the plants in each sample
measured or classified for various characteristics.
The average number of days from seeding to heading was shortest for
the California grown and longest for the Texas grown materials, as
shown in Table 15. This is in accord with differences in the length
of growing season a t the three locations and it indicates that rather
intense directional selection occurred for this character. Intense di-
rectional selection also occurred for grain type, awn length, and
pigmentation as evidenced by the differing proportions of plants with
varying expression of these characters a t the three locations. However,
mean height was about the same a t the three locations and it was also
about equal to the mean of the original parents. Apparently, there-
fore, the height of the original hybrid population was close to that
which is most desirable for each of the environments sampled. These
results make i t clear that one of the major patterns of change in
response to natural selection is rather rapid directional adjustment of
measurement characters toward optimum values and that the optimum
is often not the same for different environments.
Table 15 also reports the frequency with which plants with various
phenotypes survived a t the three locations. Plants with a wide range
of heading dates and with a wide range of height survived at all loca-
tions. The three populations also remained highly variable with respect
to grain type, degree of awning, and color. Among the surviving plants
these characters occurred in all combinations. The second major effect
of natural selection is therefore the preservation of many different geno-
types and not reduction to uniformity as has commonly been supposed.
The experiments of Akemine and Kikuchi (1958) and Allard and
Jain (1962) reveal that stabilising selection, as well as directional
selection, occura for measurement characters. Akemine and Kikuchi
grew hybrid populations of rice a t 20 experiment stations located from
31" to 43" north latitude in Japan. Each generation, random samples
of seed were drawn from the population a t each location, grown in
a common environment in central Japan, and measured for various
characters. The data showed that strong directional selection occurred
for earliness and for lateness a t northern and southern locations,
respectively, whereas there was little change in mean heading time a t
intermediate latitudes. The progressive elimination of late or early
 TABLE 15
Effect of Natural Selection in Three Environments on the Composition of a Population of Rice*
i
B
Days from Awn classes (percent) Color of hull apex g
seeding to Height. Grain types (percent) (percent)

Source of
heading
Mean R.ange Med- Awn- Tip
Pa-
tially Pur-
2
seed Mean Range (in.) (in.) Short ium Long less awDs awned Awned ple Red Green
LJ
California 102 83-127 38.8 20-48 46 48 6 53 24 17 6 2 16 82 cd

35
Arkansas 110 94-127 39.9 20-55 24 41 35 77 16 6 1 46 27 28 0
Texas 120 105-151 40.1 20-55 50 44 6 52 22 21 5 11 9 80
*Data of Adair and Jones (1946).
108 R. W. ALLARD, S. K. JAIN, AND P. L. WORKMAN

heading genotypes which occurred a t the extreme latitudes led to

skewed distributions in early generations and to some reduction in
variability, whereas a t central locations selection led t o less drastic
elimination of individuals from both tails of the distribution curve.
Nevertheless, the amount of variation remained large at all locations
showing that stabilizing as well as directional selection occurred for
heading time.
Similar concurrent directional and stabilizing selection responses
have also been observed for various metrical characters in experimental
populations of barley, wheat, and lima beans. For example, in a study
of a barley population (Composite Cross V) synthesized by mixing in-
tercrosses among 31 varieties of barley, Allard and Jain (1962) ob-
served that mean heading date of the population shifted rapidly in the
direction of earliness in the first 4 or 5 generations after synthesis and
then more slowly during the next 15 generations. Frequency distribu-
tions for various generations showed that there was steady elimination
of individuals from both tails of the curve, indicating that stabilizing
selection was also occurring for this character. The effect of this
combined directional and stabilizing selection was a small change in
the mean heading time of the population and a moderate decrease in
variance over 20 generations.
To determine the basis of these changes in mean and variance,
random samples of plants were taken from Composite Cross V in var-
ious generations and their progeny grown in replicated experiments.
Measurements on heading date and other metrical characters allowed
estimates to be made of the variance of family means and also of
within-family variances. The between-family variance was high in early
generations and it decreased steadily generation by generation. The dis-
tribution of family means indicated that genotypes with intermediate
heading dates were favored and that selection operated against both
early and late heading genotypes. However, between-family variance
remained high in the Flg generation, which was the most advanced
generation available for study. Measurements on the Fan generation of
a similar barley population (Composite Cross 11) indicated that 20
additionaI generations of stabilizing selection led to little if any
further decrease in between-family variance. Apparently, by the Fl9
generation, a balance had been struck between factors that increase
and factors that decrease variability.
Within-family variances for heading date and other metrical char-
acters were high in the Fr generation of Composite Cross V, as ex-
pected in materials recently derived from hybrids. During the next
15 generations the mean within-family variance decreased steadily.
 GENETICS OF INBREEDING POPULATIONS 109

However, in the FIQgeneration it remained larger than the average

within-family variance of the original homozygous parents of the
population. One possible explanation for this excess of variability over
that of the parents is that selection favored genotypes which, owing
to poor buffering, were more variable than equally homozygous but
well-buffered genotypes. However, in a study of seed size, tiller height,
and similar characters for which it is possible to make multiple meas-
urements on single plants, Kikuchi et al. (1967) found that within-
plant variances were higher in early than in late generations. Selection
thus appears to favor well-buffered individuals and this reinforces
the argument that the high within-family variability observed in the
Fle generation is due to the segregation of genes governing metrical
characters.
Studies of the FSQgeneration of barley Composite Cross I1 indi-
cate that little if any further decrease in within-family variability OC-
curred after the FIQgeneration. This result provides further evidence
that the individuals of which inbreeding populations are comprised
are not fully homozygous but that they provide new variability each
generation through segregation and recombination of genes governing
quantitative characters. This result also suggests that populations
derived from hybrids rather rapidly reach an equilibrium in which
loss of variability due to the combination of directional and stabilizing
selection is balanced by steady release of new variability resulting
from intercrosses between individuals within the population.
Studies of experimental populations derived from hybrids have also
provided information about the effect of natural selection on the mean
fitness of populations, as measured by seed yield. Changes in seed yield
of several representative populations are given in Fig. 17. I n the F z
to Fa generations yields were conspicuously inferior t o those of stand-
ard locally adapted varieties. This inferiority is not unexpected be-
cause many of the genotypes included as parents were unadapted
types. I n the Fs to F16 generations there was rapid improvement
in seed yield until the level of performance of each population ap-
proached or even exceeded that of the standard commercial variety
with which it was compared. Jain and Allard (1960) studied the num-
ber of seeds produced per individual in various generations. There was
a high proportion of individuals of very low fecundity in early genera-
tions. The proportion of these inferior genotypes declined sharply in
the generations when seed yield was improving rapidly and there was
a corresponding rapid increase in the proportion of individuals of
mediocre fecundity, I n late generations there was some increase in
the proportion of individuals which produced large numbers of seeds
110 R. W. ALLARD, S. K . J A I N , AND P. L. WORKMAN

120 I
-
I10 - ,I
I -.--.
7-.
I
L
- I

?'
100 I

501 ' I I I 1 I I
F4 F, 6, 56 40 64 58
Generation
FIG.17. Changes in seed yield in four representative populations synthesized from
hybrids between homozygous varieties. The yield of each population is given as a
percentage of the yield of a standard commercial variety. After Allard and Hansche
(1965).

and interestingly, a few individuals with very high seed-producing

capacity appeared. These results indicate that the mean fitness of
population derived recently from crosses between a set of arbitrary
parents is low owing to the production, by segregation and re-
combination, of a swarm of unbalanced genotypes. The initial stage of
the evolution of such populations is characterized by elimination of
these inferior genotypes, leading to rapid improvement in mean popula-
tion fitness. Continued evolution of the system appears to involve
the slow sorting out and incorporation into the populational system
of the occasional superior genotype produced by continued segregation
and recombination.

TO ABTIFICIAL
C . RESPONSES SELECTION
The question which must now be considered is whether the genetic
variability that occurs in inbreeding populations is organized into inte-
grated systems. A number of experiments involving directional selec-
tion for various single characters help to answer this question.
 GENETICS OF INBREEDING POPULATIONS 111
I n one such experiment selection for intense green seed coat color
has been practiced in an experimental population of lima beans. In
each of 15 successive generations the planting seed was chosen from
among the seeds which fell in the most green 10% of the harvest
(Allard and Sanchez, 1965). There was steady response to selection but
the increase in greenness of seed coat color was accompanied by a
series of correlated responses in other characters, such as maturity
date and seed size. Most important there was severe deterioration in
seed yield and general vigor of the population. Upon relaxation of
selection, seed coat color reverted within two or three generations to neat-
that of the unselected population and there was an accompanying
improvement in general thriftiness and productivity. Parallel results
have been obtained in several other experimental populations of lima
beans (Allard, 1965a) and in populations of barley (Suneson, 1965)
which were selected for large or for small seed size.
A selection experiment with the Yellow Double Dwarf 38 variety
of grain sorghum (Sorghum vulgare) is interesting because the base
population was a highly purified seed lot obtained from the California
Crop Improvement Association. Six mass selection populations were
established in which upward or downward selection was practiced for
seed size, height, or flowering date (Sanchez et al., 1963). I n each
generation 500 individuals were measured for these characters and
the planting seed for the next generation was taken from the 50 most
extreme individuals. As shown in Table 16 all three characters showed

TABLE 16
Effect of Six Successive Generations of Mass Selection on Three Measurement
Characters in a Highly Purified Seed Lot of the Yellow Double Dwarf 38
Variety of Grain Sorghum*

Seed size Flowering date

(gr-/lW Height (days after
kernels) (inches) seeding)

Generation Small Large Short Tall Early Late

B&W 2.46 2.46 53.3 53.3 71.7 71.7

1 2.35 2.53 49.7 55.0 71.7 72.3
2 2.34 2.82 50.7 54.0 71.0 71.3
3 2.25 3.07 - - 71.0 72.0
4 2.00 3.06 44.3 63.7 70.7 75.3
5 1.96 3.32 48.7 66.7 69.0 75.0
6 1.83 3.38 47.3 72.3 68.7 -
* Benjasil (1968).
 TABLE 17
Mean Seed Yield of Selected Populations of Grain Sorghum Expressed as a Percentage of Mean of the Urnelected Base Population*
Character under dection
~~

rA
Eeight Anthesis date
w
Short Tdl Early Late

Location I 11 I11 I I1 111 I I1 I11 I I1 I11 i

4
0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0
I 96.5 98.8 99.4 99.0 95.3 108.4 97.5 96.5 92.2 89.9 108.0 97.7 90.1 97.3 100.4 96.8 92.5 100.0
I1 79.3 105.4 106.1 103.5 100.6 97.9 83.8 97.6 91.9 92.8 101.4 92.9 86.4 93.8 99.1 101.1 98.6 96.8
I11 68.9 88.4 85.8 105.7 94.7 Q7.9 _ - - - - - 87.3 101.3 100.9 104.5 108.0 95.4
IV 63.9 71.7 81.1 99.1 98.4 96.2 87.0 94.6 88.2 104.6 107.0 90.5 81.7 104.8 93.2 106.7 102.4 97.5
V 51.0 62.2 71.2 102.4 90.4 101.6 82.9 88.1 97.9 116.3 100.9 98.6 93.5 99.3 99.3 104.9 106.2 93.3
- - -
3
VL 65.0 70.5 75.6 115.4 90.6 95.4 78.2 80.7 93.7 127.1 109.8 100.8 99.8 104.1 100.4

R2
 GENETICS OF INBREEDING POPULATIONS 113

steady response to selection in both plus and minus directions over

six successive generations. This response was accompanied by cor-
related responses in characters other than those under selection. In the
sixth generation the performance of each selected population was
compared with that of the base population in a replicated yield trial.
The results, which are shown in Table 17, show that the general
effect of directional selection for single characters was to reduce per-
formance, as measured by seed yield.
In general, directional selection for any one character changes the
mean of the character in the direction of the selection. At the same
time, changes occur in various other characters that are seemingly un-
related to the character under selection, and the totality of the popu-
lation phenotype is also altered, usually adversely; upon relaxation of
selection the population reverts toward its unselected state. Such re-
sponses are not consistent with the idea that the population is a simple
mixture of highly adapted genotypes. They indicate instead that every
component of the system depends on every other component and hence
that the population genotype is an integrated and cohesive system
(Allard, 1966).

VI. Effect of Altering the Mating System

SELF-FERTILIZATION
A. ENFORCED
Estimates of selective values given in Section 111 suggest that much
of the variability observed in inbreeding populations may be associ-
ated with loci held heterozygous as a result of some net advantage
of heterozygotes over homozygotes. If this hypothesis is correct, forced
self-fertilization should lead to a reduction in reproductive capacity.
However, it is virtually impossible t o design experiments which give
meaningful measures of reproductive capacity under natural condi-
tions and it has consequently been necessary to use characters such as
plant height, leaf number, survival after emergence, and flowering
time, measured under nursery conditions, as criteria of the fitness of
inbred materials.
I n an inbreeding experiment with wild oats, Imam and Allard
(1965) bagged panicles from 60 randomly chosen plants. A t maturity
seeds were harvested from these selfed panicles and also from com-
parable open-pollinated panicles on the same plants. The next season
these seeds were used to establish paired plots in a replicated experi-
ment conducted under nursery conditions. The data of Table 18 indi-
cate that plants from bagged and open-pollinated panicles were equal
114 R. W. ALLARD, S. K . JAIN, AND P . L. WORKMAN

in height but that plants from bagged panicles were significantly later
in flowering, tillered less profusely, and were less likely to survive
from seedling stage to maturity. Wild oats outcross at a rate of ap-
proximately 5%. Consequently 95% of the plants in plots from open-
pollinated panicles are expected to be selfs, that is, only 5% fewer than
in plots from bagged panicles. The 5% of outcrossed plants in plots
from open-pollinated panicles must have been substantially more
vigorous than the selfed plants since plots containing these few out-
crosses were significantly different from plots containing only selfs.
There was, however, some indication that the kernels produced under
TABLE 18
Effect of a Single Generation of Enforced Self-Fertilization
OD Four QuantitativeCharacters in Wild Oats*

Character

Heading time
Height Tiller Percentage survival (day8 after
(cm) number after emergence April 1)

Selfed 100.3 40.6t 90. lt 17.lt

Open-pollinated 101.2 45.8 98.1 13.3

* From Imam and Allard (1965).

t Significantly different (P < 0.05) from progenies derived from open-pollinated seed.
bags were smaller than kernels produced on unbagged panicles on
the same plant and i t is possible that the effects seen in Table 18
trace to adverse effects of bagging, rather than to inbreeding. Hence
this experiment did not fully resolve the issue of inbreeding depression
in wild oats.
I n an inbreeding experiment with grain sorghums, Sanchez et al.
(1963) bagged 100 random plants in a foundation seed lot of the
variety Yellow Double Dwarf 38. For the next 5 generations each
family was perpetuated from a single selfed plant chosen a t random.
These families gradually diverged from each other in height, flowering
time, seed size, and various other characters but there was no observa-
ble deterioration in general vitality or robustness. I n the Sa genera-
tion the seed yields of the 100 families and the base population were
determined in a replicated field experiment. The majority of the inbred
families were lower in productivity than the base population but differ-
ences were generally small and nonsignificant. Thus increasing the level
of inbreeding does not appear t o have very drastic effects on general
vigor in this already heavily inbred population of sorghums.
 GENETICS OF INBREEDING POPULATIONS 115

BY INTRODUCTION
B. INCREASED OUTBREEDING OF MALESTERILITY

Jain and Suneson (1964, 1966) have made use of a recessive male-
sterility allele (ms) to study the effect of increasing the level of out-
breeding on variability and fitness in barley. Their experimental mate-
rial was Composite Cross XIV, which was developed by intercrossing
eight barley varieties of nearly common origin by means of the male-
sterility gene.
Two levels of outcrossing were obtained by manipulation of the male
sterility gene. One level was obtained by propagating the population in
mass for 21 successive generations, ignoring the male-sterility gene. I n
this population the proportion of male-sterile plants decreased from 26%
in F2 to 5% in Flo, 2% in F14, and to 1% in F22.I n a second series the
population was propagated exclusively from seeds produced on male-
sterile plants. After 7 and 15 generations of enforced outcrossing samples
were taken from the population and propagated thereafter by the natural
mating system of self-fertilization, The generations studied in detail in
the selfing series were the F,, FI5, and FZ1and in the outcrossing series
the 07F4 (outcross 7, F4), O7Fl2,and O15F12 generations were studied
in detail. Progeny tests established that the proportion of heterozygotes
a t various marker loci was higher in the 07F4,07F12, and OlaF12 gen-
erations than in the selfing series. However, in both series i t was neces-
sary to invoke heterozygote advantage to explain the observed level of
heterozygosity a t marker loci.
Variability in quantitative characters was determined from measure-
ments made on the progeny of 64 randomly chosen individuals from each
of the six generations indicated above. These progenies were grown in a
replicated experiment in which the eight original parents of the popula-
tion were also induced. The Fl, F15, F21, 0,F4, O15F4,and O.rF12 families
grown along with parents were scored for plant height (in centimeters)
heading date (in days after March 31), and seed size on a single plant
basis. Yields were taken on a plot (row) basis and seed number per
family was determined by the yield to seed size ratio.
The range in height and seed size among the parents was rather small
but they differed from one another rather substantially in heading date,
seed number per plot, and yield per plot (Table 19). Table 19 also gives
generation means, the range in family means, and estimates of between-
and within-family variances (sa2, sW2) for various generations of the
selfing and outcrossing series. General features that should be noted are
(1) Large amounts of variability are present in all generations and
in general the range of types in the population exceeds the range
exhibited by the parents.
 TAl3LE 19
Estimates of Mean, Between-, and Within-Family Variances in Composite Cross XIV Populations*

Character Statistic F4 Fis Fzi O7F4 olsF4 OBii Parents

Plant height (cm) Mean 98.7 100.5 102.14 102.4 100.7 105.8 99.3
Sb' 96.94 56.38 45.08 77.23 74.68 71.28 14.62
Range 75.8-138.8 83.3-116.4 88.4-118.8 82.4-122.4 77.6-118.6 88.8-125.3 92.7-104.5
s WZ 89.19 73.86 64.47 101.13 126.95 101.02 34.72

Heading date Mean 19.1 19.4 18.2 18.4 18.7 19.4 19.8 F
(days after March 31) 83 9.21 10.80 5.72 9.86 6.50 5.87 15.95
Range 15.4-27.5 11.8-29.4 13.2-24.9 12.7-27.3 12.9-24.4 12.4-23.6 14.9-2.82 4
sw= 9.04 6.99 7.03 9.18 10.11 8.48 7.81 2
Z
Seed size Mean 33.16 33.76 35.13 33.51 33.56 35.93 32.32 P
(decigrams) Sb' 18.50 13.48 10.71 27.40 18.15 16.32 9.98 3
Range 19.1-40.6 32.0-41.6 24.2-42.7 21.0-46.4 24.2-44.5 24.4-46.3 28.2-37.0

Yield/plot (grams) Mean 228.0 264.7 290.9 222.1 240.9 204.9 195.8
Sb' 4567.78 3274.67 3306.37 3528.72 3354.72 2566.53 2769.78
Range 84418 95-389 159-454 127-380 105-411 97-339 125313

Seed no. per family Mean 7088 7539 8329 6729 7282 5635 6035
3838306 4062509 3694965 3456909 3379905 2864554 2374271
- - -
8'b
Range - - - -
*After Jain and Suneson (1966).
 GENETICS OF INBREEDING POPULATIONS 117

(2) The within-family variance was large in all generations, includ-

ing the FS1, which suggests that many loci affecting quantitative
characters remain heterozygous, even under intense inbreeding.
(3) The generation means are in the intermediate range and they
remain virtually unchanged from early to late generations. This is
expected because the parents were largely adapted types with heading
dates appropriate to the local environment.
(4) Between-family variance decreased in later generations (except
for height in the outcrossing series), which suggests that optimizing
selection occurred in the population, that is, that genotypes producing
extreme phenotypes are at a disadvantage. These results are parallel
to and corroborate those discussed in Section IV.
The main point of this experiment, however, is the comparison be-
tween the selfing and outcrossing series. Variability, as estimated by
the between- and within-family components, tends to be a little larger
in the outcrossing series as would be expected from greater heterozy-
gosis and higher recombination potential than in the selfing series.
Yet, the mean population fitness (w) measured by average number
of seeds produced per family and the yielding ability of total popula-
tion were higher in the selfing than in the higher outbreeding series.
The rates of change, A V , computed from the yield trial results over
the period 1954-1965, are compared in Table 20. It can be seen that
the selfing series of two composite crosses, CC V, and CC XIV,
and the outcrossing series of CC XIV improved a t nearly the same
rate. Thus, it is apparent that the low levels of outcrossing (1-2%)
that occur normally in these barley populations are integrated into a
system which released sufficient variability so that increasing the re-
combination potential by altering the mating system did not increase
the rate of change in fitness.

VII. Population Structure under Extreme Inbreeding:

The Festuca microstachys Complex

The Festuca microstachys complex provides particularly favorable

material for the study of population structure under extreme inbreed-
ing because the cleistogamous habit of these slender annual grasses
virtually ensures self-fertiliaation (Kannenberg and Allard, 1966).
The stamens and stigma are tightly enclosed by the lemma and
palea and pollination occurs before the panicle is exserted from
the “boot.” Moreover, it is usual for only one of the three anthers
in a flower to develop to anthesis and this anther, which is tightly
 TABLE 20
(AF)
Average Yields (lb/Acre) and Yearly Rate of Change in Composite Cross V and XIV Populations*

CC V, Selfing series CC XIV, Selfing series CC XIV, Outcrossing seriea

Period Generation Average yield Generation Average yield Generation Average yield 4
2
z
1954-1957 Fu-Fi, 2415 Fii-Fu 2572 O?Fz*rFa 2390
1958-1961 FlrFzo 3474 F16-Fli 3860 OEsOBe 3675
1962-1965 Fu-Fu 3740 FirFa 3972 07Fl&7Fla- 3808
A w = 31.4 AF = 31.6 AW = 32.2

* Afhr Jein and Suneson (1966).

 GENETICS OF INBREEDING POPULATIONS 119
enclosed in the stigmatic hairs, is usually very small and contains only
a half dozen or so pollen grains, Under favorable environmental condi-
tions an occasional floret is chasmogamous, thus providing opportunity
for outcrossing. However, when Kannenberg and Allard examined
more than 20,000 progeny of plants taken from nature they found
no recognizable hybrids. They concluded that the rate of outcrossing is
not likely to be as high as one per 1000 fertilizations and it is prob-
ably lower than one per 10,000 fertilizations.
An extensive quantitative study of genetic variability between and
within populations revealed that these grasses exhibit an intricate
and complex pattern of variation in the hundreds of habitats they
occupy. Distinct clinal trends were observed but differentiations be-
tween local populations were even more striking. Measurements made
on the progeny of single plants revealed that any single microniche
contains individuals which are genotypically very different from one
another with respect to height, maturity date, and other continuously
varying characters. For any single measurement character the geno-
types within a local population represent a graded series such that
there is a normal distribution between wide extremes. Consideration
of several measurement characters simultaneously showed that each
natural population consists of a very large number of genotypes, no
one of which makes up more than a small fraction of the total. How-
ever, each population has a characteristic pattern of variability for
measurement characters which distinguishes it from other local popula-
tions. The population phenotype appears to result from interactions
among many different genotypes, giving the population a coordinated
and cohesive structure. Because of the low level of intrapopulational
gene flow, these interactions are presumably between individual plants
and result from factors such as frequency and density dependence.
The following inferences were made concerning breeding behavior.
The progeny of seeds collected from single plants are quite uniform,
giving the impression that the great majority of individuals in the
population are rather highly homozygous. Although it could not be
demonstrated it was presumed that outcrossing occasionally occurs
between different families and that numerous novel genotypes segre-
gate from such hybrids. If heterozygotes are at an advantage, or ad-
vantage attends rarity in the population, as observed in other inbreed-
ing species, a burst of segregation initiated by even a single outcross
could have considerable persistence and result in the release of large
amounts of genetic variability.
The population into which these new genotypes are infused is the
product of many generations of natural selection. It is prosperous in
120 R. W. ALLARD, S. K. J A I N , AND P. L. WORKMAN

the habitat it occupies owing to harmonious balances among the many

genotypes of which it is composed and it has cohesive properties due
to interactions and interdependencies among the constituent genotypes.
When new genotypes are introduced into the population as a result
of hybridization or migration, the ones that interact disharmoniously
with the other genotypes in the population are eliminated and the
ones which contribute to a better adapted population genotype are in-
corporated into the system. Although most individuals in the popula-
tion are presumably rather highly homozygous, and hence capable
of producing progeny only like themselves, each such potentially uni-
form family appears to be represented in the population by only a
single individual, or a t most a few individuals. The population owes
its variability to the genetic uniqueness of each of many different
relatively homozygous individuals of which it is composed.
Kannenberg and Allard compared measures of within-population
variability of the fescues with measures on other annual grasses ob-
tained in similar experiments. The results indicated that the fescues
were no less variable respecting comparable quantitative characters
than either wild oats (see Section I V ) , a species in which the extent
of outcrossing is many times higher, or Lolium nvultiflorum (Schulke,
1963), an outcrossing species. Measurement data therefore do not
support the idea that there is any general or consistent relationship
between mating system and total populational variability.

VIII. Concluding Remarks

If recent experiments conducted to quantify the extent of genetic

variability in inbreeding species point to any single conclusion, it is
that there is remarkable genetic diversity within natural and domestic
populations of such species. The demonstration that inbreeding species
are highly variable genetically calls for explanations of the persistence
of genetic variability under inbreeding and it also points t o the de-
sirability of a reassessment of deductions concerning the evolutionary
role of inbreeding which have been made on the assumption that
inbreeding popuIations are uniform genetically.
Quantitative studies of inbreeding species reveal that any given
population contains individuals of many different genotypes. Individ-
uals within populations often differ from one another with respect
to single- and multiple-unit polymorphisms and also for continuously
varying characters. The extent of variation for continuously varying
characters is striking. For example, the range in height from the
 GENETICS OF INBREEDING POPULATIONS 121
genetically shortest to the genetically tallest individual within a popu-
lation is often twofold or greater and between-family variation for
other measurement characters is usually equally large.
Another feature of variability in inbreeding populations is that the
individuals within a population are frequently heteroeygous a t many
loci, that is, there is a substantial component of within-family vari-
ability. The extent of the response to selection within families is
often large, which suggests that segregational variability is an im-
portant feature of the variability system under inbreeding.
Still another feature of variability in inbreeding species is genetic
differentiation between different populations. Clinal variation is fre-
quently observed in association with progressive changes in rainfall,
temperature, and other factors of the physical environment. Super-
imposed on such broad geographical variations is a patchwork or mo-
saic pattern of variation which reflects adaptations to the local en-
vironment. The pattern is often a very fine one and since the numbers
of niches included in areas as small as a few square meters may
be considerable, such local differentiation appears to provide for mas-
sive storage of genetic variability.
It is apparent that population structure in inbreeding species is
much more complicated than has been commonly supposed and that it
probably does not take the same form in all inbreeding species or
even in different populations of the same species. Population structure
in inbreeding species cannot be explained by focusing attention on
any single factor among the complex of interacting genetic and eco-
logical factors that are involved. It is clearly inadequate, for example,
to make predictions concerning population structure or evolutionary
potential on the basis of simple genetic models which take only
within-family variability into account. Such models predict that heavy
inbreeding leads to homozygosity, a prediction which is a t variance
with the observed facts. The incorporation of selection (heterozygote
advantage) into simple single-locus models helps to explain the ex-
istence of stable polymorphisms under heavy inbreeding. There is,
however, experimental evideiice to indicate that different polymor-
phisms interact with one another and theory shows that the effects of
such epistatic interactions on population structure vary with the type
of epistasis and the intensity of inbreeding. One of the discoveries from
theoretical investigations of multilocus models is that inbreeding can
have the same effect as linkage and that with some types of epistasis
inbreeding can reinforce or substitute for linkage in the maintenance
of balanced gene complexes. It seems clear that simple genetic models
are inadequate and that analyses of complex multilocus models are
122 R. W. ALLARD, S. K. JAIN, AND P. L. WORKMAN

prerequisite to an understanding of the genetic aspects of the struc-

ture of inbreeding populations.
The situation becomes even more complex when various factors
which are more ecological than genetic are taken into account. One
such factor is frequency-dependent selection in which the pattern of
interactions between individuals is a function of the frequency of vari-
ous genotypes in the population. In one case which has been analyzed
in detail, the S/s locus in the lima bean, heterozygotes and homozy-
gotes are equal in fitness when all three genotypes are frequent in the
population. However, when heterozygotes are rare in the population
their fitness increases relative to the homozygotes. In this case it is
clear that maintenance of a stable nontrivial polymorphism depends on
a complex set of interactions between genetic factors, mating system,
and ecological factors. The maintenance of numerous relatively homo-
zygous genotypes in populations of the heavily inbreeding grasses of
the Festuca microslachys complex may depend in part on a similar
complex pattern of interactions between individuals in the population.
Other items of a similar nature which are believed to affect the ex-
tent of genetic variability include niche arrangement, local subdivision,
mobility in animals, pollen and seed dispersal mechanisms in plants,
and interactions of these factors with each other and with the genetic
determinants of population structure.
Many aspects of genetic systems and of ecology have been studied
individually in different species but rarely if ever have several factors
been studied simultaneously in a species as components of a single
integrated system for the regulation of variability. The discovery that
remarkable genetic variability exists within populations of inbreeding
species suggests that high variability is essential to the survival of
populations. It also suggests that change of any component of the
variability system which leads t o undue restriction of variability must
be mediated by compensatory changes in other components if the
population is to survive. For example, should adoption of self-fertiliza-
tion as a device to ensure fertilization in unfavorable weather have
the correlated effect of reducing genetic variability in the population
to critical levels, survival of the population would presumably depend
on compensating adjustments in other components of population struc-
ture which affect the variability. Viewed in this way the mating system
is seen t o be only one component of the variability system and what is
needed are comprehensive studies to determine how it is integrated
into the totality of factors which determine population structure and
net evolutionary potential. Broad generalizations about the evolution-
 GENETICS OF INBREEDING POPULATIONS 123

ary implications of mating systems do not appear t o be warranted

until this type of information becomes available.

IX. Summary
Inbreeding can arise in populations either as a result of various
mechanisms which affect the mating system, or from restrictions in
actual or effective population size. In this review the effects of
inbreeding have been considered, in a populational context, with re-
spect to theory and also in terms of studies conducted with natural
and experimental populations.
The theoretical effects of inbreeding were introduced in terms of
single-locus population models in which population size was assumed
sufficiently large to avoid sampling effects and in which i t was assumed
that selective values, mating system parameters, and other population
parameters were constant in all environments. Examination of equi-
librium values and rate of approach to equilibrium in terms of simpli-
fied population models show that net heterozygote advantage is a
necessary condition for stable nontrivial equilibrium when the mating
system imposes some degree of inbreeding. The extent of heterozygote
advantage required to avoid fixation is least when homozygotes have
equal selective values and there is little inbreeding; i t increases with
increasing asymmetry of selection and increasing levels of inbreeding.
Under the most intense inbreeding possible with diploid organisms
(complete assortative mating) the selective value of the heterozygote
must be a t least double the selective values of the homozygotes to
maintain stable nontrivial equilibrium.
The effects on population structure of linkage and epistatic inter-
actions between different polymorphisms were then considered in terms
of multilocus genetic models involving deviations from a fixed optimum,
heterotic models, and mixed optimum-heterotic models. The basis for
this choice is the general experience that the types of selection im-
plied by these models are widespread in actual populations. The
results were expressed in terms of the extent of linkage disequilibrium
(gametic phase unbalance), mean population fitness, and the number
and nature of gene frequency equilibria. Complex interactions occur
among linkage, epistasis, and inbreeding, and different combinations
of these factors can lead to retention of variability in populations.
Finally, the models were extended to include stochastic variation,
which appears to have the effect of producing quasi-stable equilibria
a t critical values of the relevant parameters.
124 R. W. ALLARD, S. K. JAIN, AND P. L. WORKMAN

Estimates of selective values for the simple polymorphic characters

commonly found in natural and experimental populations of pre-
dominantly self-pollinated plants show that linkage blocks marked b y
major genes are often subject to very strong selection. I n a high pro-
portion of cases the estimated selective values show that such linkage
block heterozygotes have striking net selective advantage over homo-
zygotes. The selective values of heterozygotes often fall within ranges
which permit stable equilibrium a t intermediate gene frequencies. In-
breeding populations also frequently carry several t o many polymor-
phisms whose distributions in the population are interdependent. Such,
interactions extended over the totality of the gene pool may be the
basis of the cohesive properties and “coadapted” nature of the popula-
tion genotype.
There is a substantial body of measurement data which shows that
natural populations of inbreeding species contain great stores of
genetic variability respecting continuously varying characters. Part of
this variability takes the form of geographical differentiations, includ-
ing a mosaic pattern of differentiations associated with adaptation to
microniches. There is also extensive intrapopulational variability both
between and within families. Between-family variability is shown by
significant differences among the means of the progenies of individuals
taken from nature, Within-family variability is demonstrated by re-
sponses to selection within single kinships, indicating that the in-
dividuals within natural populations are heterozygous for numerous
genes affecting quantitative characters.
Theory and experimental studies lead to the following conception
of population structure under inbreeding. The natural population con-
tains a very large number of different genotypes none of which makes
up more than s small proportion of the total. For any single measure-
ment character the genotypes in a population represent a graded
series such that there is a more or less normal distribution from
one extreme to the other with all degrees of intermediacy represented.
When many characters are considered simultaneously, each local popu-
lation is found t o have a characteristic pattern of variability which
distinguishes it from other local populations. The mating system is
only one of many factors which affect the amount and kinds of vari-
ability in populations. The observed structure of inbreeding popula-
tions is seen to result from an integration of inbreeding into the
totality of genetic and ecological factors affecting population struc-
ture. The structure of inbreeding species provides both for storage of
large amounts of genetic variability and for retention of sufficient
flexibility t o meet the requirements imposed by heterogeneity of the
 GENETICS OF INBREEDING POPULATIONS 125

local environment, including season-to-season or year-to-year fluctu-

ations. Species which meet these requirements appear t o possess ample
stores of variability to satisfy the challenges imposed by long-term
‘changes in environment.

ACKNOWLEDGMENTS
This study was supported in part by grants from the National Science Foundation
(GB-3246) and the National Institutes of Health (GM 10476).

REFERENCES
Adair, C. R., and Jones, J. W. 1946. Effect of environment on the characteristics
of plants surviving in a bulk hybrid population of rice. J . Am. Soe. Agron. 38,
708-716.
Akemine, H., and Kikuchi, F. 1958. Genetic variability among hybrid populations
of rice plants grown under various environments. (In Japanese.) In “Studies
on the Bulk Method of Plant Breeding” (K. Sakai, R. Takahashi and H.
Akemine, eds.), pp. 89-105.
Allard, R. W. 1956. Unpublished results.
Allard, R. W. 1963. Evidence for genetic restriction of recombination in the lima
bean. Genetics 48,1389-1395.
Allard, R. W. 1965. Genetic systems associated with colonizing ability in predomi-
nantly self-pollinated species. In “The Genetics of Colonizing Species” (H. G.
Baker and G. L. Stebbins, eds.), pp. 50-76. Academic Press, New York.
Allard, R. W. 1965. Unpublished results.
Allard, R. W. 1966. Population structure and performance in crop plants. Ciencia
Cult. Sao Paulo 19,145-150.
Allard, R. W., and Golden, W. G. 1954. Unpublished results.
Allard, R. W., and Hrtnsche, P. E. 1964. Some parameters of population variability
and their implications in plant breeding. Advan. Agron. 16, 281-325.
Allard, R. W., and Hansche, P. E. 1965. Population and biometrical genetics in
plant breeding. Proc. 11th Intern. Congr. Genet., The Hague, 1963 Vol. 3,
pp. 665-679. Pergamon Press,Oxford.
Allard, R. W., and Jain, S. K . 1962. Population studies in predominantly self-
pollinated species. 11. Analysis of quantitative changes in a bulk hybrid popu-
lation of barley. Evolution 16, 90-101.
.Allard, R. W., and Sanchez, R. L. 1965. Unpublished results.
Allard, R. W., and Wehrhahn, C. 1964. A theory which predicts stable equilibrium
for inversion polymorphisms in the grasshopper, Moraba scurra. Evolution 18,
129-130.
Allard, R. W., and Workman, P. L. 1963. Population studies in predominantly
self-pollinated species. IV. Seasonal fluctuations in estimated values of genetic
parameters in lima bean populations. Evolution 18, 470-480.
Allard, R. W., Harding, J., and Wehrhahn, C. 1966. The estimation of selective
values and their use in predicting population change. Heredity 21, 547-564.
Barnes, D. I<., m d Cleveland, R. W. 1963. Pollen tube growth o f diploid alfalfa
in vitro. Crop Sci. 3, 291-297.
126 R. W. ALLARD, S. K. JAIN, AND P. L. WORKMAN

Bartlett, M. S., and Haldane, J. B. S. 1936. The theory of inbreeding with forced
heterozygosis. J. Genet. 31, 327-340.
Bateman, A. J. 1962. Variation within French bean varieties. Ann. Appl. Biol. 39,
129-138.
Benjasil, V . 1968. IMecta of ma= selection for quantitative traits in sorghum. Ph.D.
Dissertation. Univ. California, Davis, California.
Bennett, J. H. 1964. The distribution of heterogeneity upon inbreeding. J . Roy.
Statist. SOC. B16, 88-99.
Bennett, J. H.,and Binet, F. E. 1956. Association between Mendelian factors with
mixed selfing and random mating. Heredity 10, 51-55.
Binet, F. E.,Clark, A. M., and Clifford, H. T. 1969. Correlation due t o linkage in
certain wild plants. Genetics 44,613.
Bodmer, W.F.,and Parsons, P. A. 1962. Linkage and recombination in evolution.
Advan. Genet. 11,147.
Cain, A. S.,and Sheppard, P. M. 1952. The effects of natural selection on body
color in the land snail Cepaea nemoralis. Heredity 6,217-231.
Cannon, G. B. 1963.The effects of natural selection on linkage disequilibrium and
relative fitness in experimental populations of Drosophila melanogaster. Ge-
netics 48,1201-1216.
Claassen, C. E. 1950. Natural and controlled crossing in safflower, Carthamus
tknctorius L. Agron. J . 42,381484.
Clarke, A. E. 1935. Inheritance of annual habit and mode of pollination in an
annual White Sweet Clover. Agron. J. 27, 492-496.
Clarke, B., and O’DonaId, P. 1964. Frequency-dependent selection. Heredity 19,
201-206.
Cooch, F. G.,and Beardmore, J. A. 1959. Assortative mating and reciprocal differ-
ence in the Blue-Snow goose complex. Nature 183, 1833-1834.
Darlington, C. D., and Mather, I(. 1949. “Elements of Genetics.” Allen & Unwin,
London.
Dempster, E. R. 1956. Maintenance of genetic heterogeneity. Cold Spring Harbor
Symp. Quant. Biol. 20, 25-32.
Dillman, A. C. 1938.Natural crossing in flax. Agron. J. 30, 279-286.
Dreyfus, A., and Breuer, M. E. 1944. Chromosome and sex determination in the
parasitic hymenopteron Telenomus furiai (Lima). Genetics 29, 75-82.
East, E. M. 1919. Studies on self-sterility. 111. The relation between self-fertile and
self-sterile plants. Genetics 4, 341-346.
Fisher, R. A. 1939. Selective forces in wild populations of Paratettix tezunus. Ann
Eugen. (London) 9,109-122.
Fisher, R . A. 1948. A quantitative theory of genetic recombination and chiasma
formation. Biometldcs 4,l-13.
Fisher, R. A. 1949. “The Theory of Inbreeding.” Oliver & Boyd, Edinburgh and
London.
Fyfe, J. L., and Bailey, N. T. J. 1961. Plant breeding studies in leguminous forage
crops. I. Natural crossing in winter beans. J . Agr. Sci. 41, 371-378.
Geiringer, H.1944. On the probability theory of linkage in Mendelian heredity.
Ann. Math. Statist. 15, 26-57.
Geiringer, H.1945.Further remarks on linkage theory in Mendelian heredity. Ann.
Math. Statist. 15,26-67.
Gibson, J. B., and Thoday, J. M. 1962. Effects of disruptive selection. VI. A second
chromosome polymorphism. Heredity 17,l-26.
 GENETICS OF INBREEDING POPULATIONS 127
Grant, V. 1956. The influence of breeding habit on the outcome of natural hy-
bridization in plants. Am. Naturalist 90, 319-322.
Grant, V. 1958. The regulation of recombination in plants. Cold Spring Harbor
Symp. Quant. Biol.23, 337-363.
Griffing, B. 1960. Accommodation of linkage in mass selection theory. Australian
J. Biol.Sci. 13, 501-526.
Guitierrez, M. G., and Sprague, G. F. 1959. Randomness of mating in isolated
polycross phntings of maize. Genetics 44, 1075-1082.
Haldane, J. B. S. 1924. A mathematical theory of natural and artificial selection.
Part 11. Proc. Cambridge Phil. SOC.1, 168-163.
Haldane, J. B. S. 1949. The association of characters as a result of inbreeding and
linkage. Ann. Eugen. (London) 15,16-23.
Haldane, J. B. S., and Waddington, C. H. 1931. Inbreeding and linkage. Genetics
16,357-374.
Hanson, W. D. 1959. The breakup of initial linkage blocks under selected mating
systems. Genetics 44, 857-868.
Harding, J., and Allard, R. W. Population studies in predominantly self-pollinated
species. XI1 Zygotic ‘and gametic interactions in a population of Phaseolus
Zunatus (in preparation).
Harding, J., and Tucker, C. L. 1964. Quantitative studies on mating systems. I.
Evidence for the non-randomness of outcrossing in Phaseolus lunatus. Heredity
19,369-381.
Harding, J., Allard, R. W., and Smeltzer, D. G. 1966. Population studies in pre-
dominantly self-pollinated species. IX. Frequency dependent selection in
Phaseolus lunatus. Proc. Natl. Acad. Sci. U.S. 56, 99-104.
Harlan, H. V. 1957. “One Man’s Life with Barley.” Exposition Press, New York.
Harlan, H. V., and Martini, M. L. 1938. The effect of natural selection in a
mixture of barley varieties. J. Agr. Res. 67, 189-199.
Hayman, B. I. 1953. Mixed selfing and random mating when homozygotes are a t
a disadvantage. Heredity 7, 185-192.
Hayman, B. I., and Mather, K. 1953.The progress of inbreeding when homozygotes
are at a disadvantage. Heredity 7, 166-183.
Hayman, B. I., and Mather, K. 1956. Inbreeding when homozygotes are a t a
disadvantage : A reply. Heredity 10, 271-274.
Hogben, L. 1946. “An Introduction to Mathematical Genetics.” Horton, New York.
Holden, J. H. W., and Bond, D. A. 1960. Studies on the breeding system of the
field bean, Vicia labia (L.). Heredity 15, 175-192.
Huxley, J. S. 1965. Morphism in birds. Acta 11th Congr. Intern. Omithol. Basel,
1954 pp, 309-328.
Imam, A. G., and Allard, R. W. 1965. Population studies in predominantly self-
pollinated species. VI. Genetic variability between and within natural popula-
tions of wild oats from differing habitats in California. Genetics 61, 49-62.
Jain, S.K.1968.Unpublished results.
J a b , S. K. 1968. Simulation of models involving mixed selfing and linkage in finite
populations. Theoret. Appl. Genet. 38 (in press).
Jain, S. K., and Allard, R. W. 1960. Population studies in predominantly self-
pollinated species. I. Evidence for heterozygote advantage in a closed popula-
tion of barley. Proc. Natl. Acad. Sci. U B . 46, 1371-1377.
Jain, S. K.,and Allard, R. W. 1966. The nature and stability of equilibria under
optimizing selection. Proc. Natl. Acad. Sci. U.S. 54, 1436-1443.
128 R. W. ALLARD, S. K. JAIN, AND P. L. WORKMAN

Jain, S. K., and Allard, R. W. 1966. The effects of linkage, epistasis, and inbreeding
on population changes under selection. Genetics 53, 633-659.
Jain, S. K.,and Marshall, D. R. 1967. Population studies in predominantly self-
pollinated species. X. Variation in natural populations of Awena f a t m and A .
barbata. Am. Naturalist 101,19-33.
Jain, 8. K.,and Marshall, D. G.1968.Simulation of models involving mixed selfing
and random mating. I. Stochastic variation in outcrossing and selection pa-
rameters. Heredity (in press).
Jain, S. K., and Suneson, C. A. 1964. Population studies in predominantly self-
pollinated species. VII. Survival of a male-sterility gene in relation to heterozy-
gosis in barley populations. Genetics 50, 905-913.
Jain, S. K., and Suneson, C. A. 1966. Increased recombination and selection in
barley populations carrying a male sterility factor. I. Quantitative variability.
Genetics 54,1215-1224.
Jain, S . K.,and Workman, P. L. 1966. The use of generalized F-statistic in t h e
theory of inbreeding and selection, Nature 214, 674-678.
Jana, S., and Allard, R. W. Population studies in predominantly self-pollinating
species. XIII. Genetic variability in populations of lirna beans composed origi-
nally of a mixture of two pure lines (in preparation).
Jennings, H. S. 1917. The numerical results of diverse systems of breeding, with
respeot to two pairs of characters, linked or independent, with special relation
to the effects of linkage. Genetics 2, 97-154.
Johannsen, W. 1926. “Elemmte der exakten Erblichkeitslehre,” 3rd ed. Fischer,
Jena.
Kannenberg, L. W., and Allard, R. W. 1967. Population studies in predominantly
self-pollinated species. VIII. Genetic variability in the Festuca microstachys
complex. Ewolution 21, 227-240.
Kempthorne, 0. 1957. “An Introduction to Genetic Statistics,” Wiley, New York.
Kikuchi, F., Allard, R. W., and Chapman, S. R. The effect of natural selection on
phenotypic plasticity and genetic polymorphism in a barley population. In
“Food for the Man of Tomorrow” Symp. Am. Univ. Beirut, 1967 (in press).
Kimura, M. 1954. Process leading to quasi-fixation of genes in natural populations
due to random fluctuation of selection intensities. Genetics 39, 280-295.
Kimura, M. 1956. A model of a genetic system which leads to closer linkage b y
natural selection. Evolution 10, 278-287.
Kimura, M. 1963. A probability method for treating inbreeding systems, especially
with linked genes. Biometries 19, 1-17.
Kimura, M. 1964. “Diffusion Models in Population Genetics,]’ p. 57. Methuen,
London.
Kimura, M. 1966. Attainment of quasi-linkage equilibrium when gene frequencies
are changing by natural selection. Genetics 52, 876-890.
Kimura, M., and Crow, J. F. 1963. The measurement of effective population number.
Evolution 17, 279-288.
Knowles, P.F. 1943. Improving an annual bromegrass, Bromus moUk L., for range
purposes. J. Am. SOC.Agron. 35, 584-694.
Kojima, K. 1959. Role of epistasis and overdominance in stability of equilibria
with selection. Proc. Natl. Acad. Xci. U.S. 45, 984-989.
Kojima, K. 1966. The evolutionary dynamics of two-gene systems. In “Computers
in Biomedical Research” (R. W. Stacy and B. D. Waxman, eds.), Vol. 1,
Chapter 9,pp. 197-220. Academic Press, New York.
 GENETICS OF INBREEDING POPULATIONS 129
Lamotte, M. 1961. Recherches sur la structure genetique des populations naturelles
de Cepaea nemoralis L. Bull. Biol. France Belg. Suppl. 35, 1-239.
Levitan, M. 1965. Studies of linkage in populations. I. Association of second
chromosome linkages in Drosophila robusta. Evolution 9, 2741.
Levitan, M. 1958. Nonrandom association of inversions. Cold Spring Harbor Symp.
Quant. Biol. 23,251-268.
Lewontin, R. C. 1958. A general method for investigating the equilibrium of gene
frequency in a population. Genetics 43,419434.
Lewontin, R. C. 1964a. The interaction of selection and linkage. I. General con-
siderations: Heterotic models. Genetics 49, 49-67.
Lewontin, R. C. 1964b. The interaction of selection and linkage. 11. Optimum
models. Genetics 50, 757-782.
Lewontin, R. C., and Hojima, K. 1960. The evolutionary dynamics of complex
polymorphisms. Evolution 14, 458-472.
Lewontin, R. C., and White, M. J. D. 1960. Interaction between inversion poly-
morphisms of two chromosome pairs in the grasshopper Moraba scurra. Evolu-
tion 14, 116129.
Magahaes, L. E.,Brito Da Cunha, A., De Toledo, J. S., Toledo, S. A., De Soura,
H. L., Targa, H. J., Setzer, V., and Pavan, A. 1964. On lethals and their
suppressors in experimental populations of Drosophila willistoni. Mutation Res.
5 45-54.
Mainardi, D., Scudo, F. M., and Barbieri, D. 1965. Assortative mating based on
early learning: population genetics. Ateneo Parmense 36, 583-606.
MalBcot, G. 1948. “Les math6matiques de l’hCr6ditk.” Masson, Paris.
Mather, K. 1943.Polygenic inheritance and natural selection. Biol. Rev. 18, 32-64.
Mayr, E. 1963. “Animal Species and Evolution.” Harvard Univ. Press, Cambridge,
Massachusetts.
Murray, J. 1964.Multiple mating and effective population size in Cepaea nemoralis.
Evolution 18, 283-291.
Nei, M., and Syaktdo, K. 1958.The estimation of outcrossing in natural populations.
Japan. J . Genet. 33,46-51.
O’Donald, P.1959.Possibility of assortatire mating in the Arctic Skua. Nature 183,
1210-1211.
O’Donald, P. 1960a. Inbreeding as a result of imprinting. Heredity 15, 79-85.
O’Donald, P. 1960b. Assortative mating in a population in which two alleles are
segregating. Heredity 15, 389-396.
Owen, A. R. G.1953. A genetical system admitting of two distinct stable equilibria
under natural selection. Heredity 7 , 97-102.
Parsons, P. A. 1957. Selfing under conditions favoring heterozygosity. Heredity 11,
411-421.
Pearson, K.,and Lee, A. 1903. On the laws of inheritance in man. I. Inheritanre
of physical characters. Biometrilia 2, 357-462.
Prout, T. 1965. The estimation of fitness from genotypic frequencies. Evolution
19, 546551.
Reeve, E. C. R. 1956. Inbreeding with the homozygotes a t a disadvantage. Ann.
Human Genet. 19, 332-346.
Reeve, E. C . R., and Gower, J. C. 1957. Inbreeding with selection and linkage. 11.
Sib mating. Ann. Human Genet. 23, 36-49.
Robbins, R. B. 1918.Applications of mathematics to breeding problems. 11. Genetics
3, 73-92.
130 R. W. ALLARD, S. K. JAIN, AND P. L. WORKMAN

Roberts, H.F. 1929. “Plant Hybridization Before Mendel.” Princeton Univ. Press,
Princeton, New Jersey.
Saaty, T. L., and Bram, J. 1964. “Nonlinear Mathematics.” McGraw-Hill, New York.
Sanchez, R. L. Smeltzer, D. G., and Allard, R. W. 1963. Unpublished data.
Schnell, F. W. 1961. Some general formulations of linkage effects in inbreeding.
Genetics 46, 947-967.
Schulke, J. D. 1963. Genetic variability in natural populations of Lolium multi-
fioruna. Ph.D. Dissertation, University of California, Davis, California.
Sheppard, P. M.1959.The evolution of mimicry; a problem in ecology and genetics.
Cold Spring Harbor Symp. Quant. Bwl. 24, 131-140.
Shikata, M. 1963. Representation and calculation of selfed population by group
ring. J. Theoret. Biol. 5, 142-160.
Smeltzer, D. G. 1965.Unpublished data.
Spassky, B., Dobzhansky, T., and Anderson, W. W. 1965. The genetica of natural
populations. XXXVI. Epistatic interactions of genetic load in Drosophila pseu-
doobscura. Genetics 52, 653-664.
Stebbins, G. L. 1950. “Variation and Evolution in Plants.” Columbia Univ. Press,
New York.
Stebbins, G. L. 1967. Self-fertilization and population variability in the higher
plants. Am. Naturalist 41, 337-353.
Suneson, C. A. 1949. Survival of four barley varieties in a mixture. Agron. J. 41,
459461.
Suneson, C.A. 1966.Personal communication.
Tucker, C. L., and Harding, J. 1965. Quantitative studies on mating systems.
11. Estimation of fitness parameters in a population of PhaseoZus Zunatus.
Heredity 20, 393-402.
Wahlund, S. 1928. Zusammensetzung von Populationen und Korrelationserscheinun-
gen vom Standpunkt der Vererbungslehre aus betrachtet. Hereditas 11, 65-106.
Wahrman, J. 1965.Permnal communication.
Weinberg, W. 1909. Ober Vererbungsgesetae beim Menschen. 2. Znduktive
Abstammungs- Vererbungslehre 1, 277-330.
White, M. J. D. 1957. Cytogenetics of the grasshopper, Moraba scurra. 11. Heterotic
systems and their interactions. Australian J . 2001.5, 305-337.
Workman, P. L. 1964. The maintenance of heterozygosity by partial negative
assortative mating. Genetics 50, 1369-1382.
Workman, P. L., and Allard, R. W. 1962. Population studies in predominantly
self-pollinated species. 111. A matrix model for mixed selfing and random
outcrossing. Proc. Natl. Acad. Sci. US. 48, 1318-1325.
Workman, P. L., and Allard, R. W. 1964. Population studies in predominantly
self-pollinated species. V. Analysis of differential and random viabilities in
mixtures of competing pure lines. Heredity 19, 181-189.
Workman, P. L., and Jain, S. K. 1966. Zygotic selection under mixed random
mating and self-fertilization : Theory and problems of estimation. Genetics
54,159-171.
Wright, 8. 1921. Systems of mating, I-V. Genetics 6, 111-1’78.
Wright, 6. 1933. Inbreeding and recombination. Proc. Natl. Acad. Sci. U S . 19,
4201433.
Wright, S. 1940. Breeding structure of populations in relation to speciation. Am.
Naturalist 74, 232-248.
Wright, S. 1942.Statistical genetics and evolution. Bull. Am. Math. SOC.48, 223-246.
 GENETICS OF INBREEDING POPULATIONS 131
Wright, S. 1943. Isolation by distance. Genetics 28, 114-138.
Wright, S. 1946. Isolation by distance under diverse systems of mating. Genetics
31, 39-59.
Wright, S. 1948. On the roles of directed and random changes in gene frequency
in the genetics of populations. Evolution 2, 279-294.
Wright, S. 1951. The genetic structure of populations. Ann. Eugen. (London) 15,
323-354.
Wright, S. 1952. Quantitative inheritance. In “The Genetics of Quantitative Vari-
ability” (E. C. R. Reeve and C. H. Waddington, eds.), pp. S-41. Agr. Res.
Council, London.
Wright, S. 1959. Physiological genetics, ecology of populations, and natural selection.
Perspectives Biol. Med. 3, 107-161.
Wright, S. 1964. Stochastic processes in evolution. In “Stochastic Models in Medicine
and Biology” (J. Gurland, ed.), pp. 199-244. Univ. of Wisconsin Press, Madison,
Wisconsin.
Wright, S. 1965a. The interpretation of population structure by F-statistics with
special regard to systems of mating. Evolution 19, 395-420.
Wright, S. 1965b. Factor interaction and linkage in evolution. Proc. Roy. SOC.
B162, 80-104.