Existing breeds or closed lines of each livestock species are
essentially mildly inbred lines, whose relative straightbred performance
levels are determined by differences in both 1) mean gene frequencies and 2)
degrees of heterozygosity relative to a hypothetical population of all possible
breed crosses for a given species. These breed differences have developed over time from both 1) deliberate and
natural selection for adaptation to differing production-marketing
environments and 2) random drift in gene frequencies and in degree of
heterozygosity (inbreeding) from variable limits on the effective size of each
breed population. These differences include average, dominant and non-allelic
interaction effects of genes. Crosses of
breeds or lines produce changes in performance relative to the parental stocks
from complementary maternal/paternal effects, increased heterozygosity (reduced
inbreeding) for dominant alleles, and from changes in non-allelic interactions
as well. The challenge is to evaluate these genetic components of breeds
and their crosses accurately enough to predict the performance to be expected
from alternative choices of breeds and breeding systems. This task is
complicated by the multiple-trait nature of the breed differences and their interactions with production environments which
together determine the economic efficiency of livestock
production-marketing systems.
Alternative uses of genetic diversity among breeds or strains of
livestock include 1) expansion of the more efficient adapted breeds, 2) systematic
crossing of selected breeds and 3)
development of new composite breeds from selected combinations of several
preexisting breeds. All three alternatives permit more rapid short-term
genetic adaptation to changes in production-marketing
environments than can be expected from selection within a single breed population
(Smith and Banos, 1991). However, relative effectiveness of these alternative
breed uses depends upon both the nature of gene effects on performance traits
(average, dominant, epistatic) and the
reproductive rate of the species (Dickerson, 1969, 1973; Smith, 1989).
Expansion of a superior breed simply replaces other breeds that have
poorer gene-frequency and heterozygosity effects on performance. It can be
accomplished by outcrossing and then backcrossing to
the superior breed, and less efficiently, by reduced selection among female
progeny of the superior breed (Robertson and Asker, 1951). However, ultimate improvement from this method is limited to that
obtainable by continued selection within the better pure breeds, since
possible further improvement from crossbreeding is ignored.
Some system
of crossbreeding usually can make more effective commercial use of breed
differences than "grading up" to the best adapted breed, by
exploiting heterosis in individual, maternal or paternal performance, including
complementary breed differences in maternal vs transmitted individual
effects in performance (Winters et al., 1937; Lush et al., 1939; Smith, 1964;
Cartwright, 1970; Moav, 1966, 1973). The alternative two- three- or four-breed specific or rotation crossbreeding
systems differ not only in the proportion of the maximum heterosis used, but also in the proportion of purebred matings
required to provide replacements for
the industry crossbreeding system (Dickerson, 1973). 'Periodic' rotational
crossing, using sire breeds in unequal numbers of generations, but in strategic
sequence, has been shown to have less integenerational variation
(Bennett, 1987), and to potentially equal or exceed the efficiency of
conventional sire breed rotations.
Still
another alternative is formation of new "composite" breeds from a
combination of pre-existing breeds selected for superior adaptation to a
breed-role and production-marketing system. Such composite breeds use less of
the maximum Fl heterosis than in systematic crossbreeding, and alone
cannot utilize the "complementarity" of terminal crossing. However, a
new composite can be maintained by the much simpler straight breeding, and does not require continued
replacements from matings of parental purebreds. Desired selection may
be applied more directly and intensively than in separate parental breeds. New
composite breeds also can be selected to serve as specialized maternal or
paternal parents in specific two-breed crossbreeding systems, thus contributing
to some increased heterosis in maternal and/or paternal performance and to
complementary maternal and terminal sire breed effects, with a reduced
proportion of parental line matings (Cartwright, 1970; Cartwright et al., 1975). If desired, a composite breed
also can be propagated and possibly further improved by continuing use of only
F1 crossbred sires from the breeds of its origin.
The ultimate choice of optimum breeding system for any given
production-marketing environment depends upon the balance between the amount of
heterosis and breed of sire/dam complementarity in performance efficiency
achieved by the crossbreeding system and
the proportion of the total industry population represented by the crossbreds
(Dickerson, 1973). For this reason, systems such as rotational
crossbreeding which requires only sires of pure breeds, or new multi-breed
composites which require no parental purebred matings, are more efficient for
species with a low reproductive rate, such as cattle. In contrast, the higher
degree of heterosis and complementary paternal/maternal performance for
specific two-, three- or four-breed
crossing systems are likely to be more efficient for swine and especially for
poultry, where a higher reproductive rate requires a relatively small
proportion of purebred matings to provide replacement breeding stock for
crossbreeding.
The relative
efficiency of alternative breeding systems for use of genetic diversity among
breeds is determined mainly by 1) average transmitted breed deviations in
individual (gI), maternal (gM) and paternal (gP)
effects on progeny performance; 2) magnitude of crossbred heterosis for individual (hI), maternal (hM)
and paternal (hP) effects; 3) change in non-allelic gene
interaction effects from non-parental recombination in crossbred progeny and
parents (rI, rM and rP); and 4) the
reproductive rate of each species, and of breeds within a species, , which controls the proportion of purebred vs crossbred
matings required in each industry breeding system.
The
expectations for dominance effects in systems of mating were first defined by
Wright (1921, 1922). He also recognized that deviations from linear association
with changes in heterozygosity among
parental, F1, F2 and backcross generations provide
evidence for effects of non-allelic gene interaction (Wright, 1977).
Expectations for dominance effects in rotational crossbreeding, using sires of
n breeds, were given long ago by Cannon et al. (1956). Breed average
transmitted direct (gIi), and indirect maternal (gMi)
or paternal (e.g., in conception rate, gPi) effects in
breed crossing can be measured in some type of diallel mating design
(Table 1). Here, heterosis can be estimated for the mean of all
crosses included (hI..),
for those crosses having a common breed of sire 
or dam 
or both 
for each pair of reciprocal crosses 
and for possible specific reciprocal sex-linked
or cytoplasmic effects 
Eisen et al., 1983).
Differences between reciprocal crosses are assumed caused by breed indirect
maternal effects 
although differences
in average breed paternal 
or in specific 
effects also
can possibly be involved. Deviations of breed average or specific
cross heterosis (hij) from the
mean for all crosses (h..) indicate differences either in degree of
change in heterozygosity (correction of inbreeding effects) or in
non-allelic interaction effects, or both.
Heterosis for indirect maternal (hM) or paternal (hP)
effects requires mating designs (Table 2) using crossbred females
and/or males as parents in experimental comparisons with purebred parents (i.e., 
These parameters are very
useful in choosing breeds for specific crossbreeding (e.g., two-, three, or
four-breed) but not to evaluate the possible role of a new composite relative
to rotational or specific crossbreeding systems. The latter also requires
information about epistatic deviation from
linear association with changes in heterozygosity that can be obtained only
from designs comparing parental, F1, F2, F3
and backcross generations.
The
formulation used here (Table 3) expresses expectations for alternative crossbreeding systems in terms of deviations from
the weighted mean of the n parental pure breeds 
where qi; =
fraction of each parental breed in progeny or parents of a given mating. The
expectation for heterosis (h) includes effects of increased heterozygosity on
expression of both dominance (d) and non-allelic interactions of average (gg),
average × dominant (gd) and dominant
(dd) gene effects in the deviation of various crosses from the mean of the
parental breeds 
. Thus the expected effects of non-allelic interaction on
differences among various types of crosses can be expressed as deviations from
the proportion of such effects included in the average F1, heterosis
of crossbreds (h = d + 1/2 gg). In this approach (Dickerson 1969, 1973; Koch et
al., 1985), the r parameter measures
epistatic deviations of observed heterosis from linear association with
expected change in degree of heterozygosity from the mean for the parent
breeds. This partition of epistatic gene effects (gg, gd, dd) agrees with that
proposed by Hill (1982), except that expectations are expressed as deviations
from the combined dominance and epistatic effects in the F1 (h = d +
1/2 gg), as well as from the weighted mean of parental breeds, rather than from
the F2 generation. Also scaling was reduced by one-half. Hill's (1982)
formulation, in turn, was derived from
earlier work of Cockerham (1954) and Kempthorne (1957). Alternative parameters
for additive, heterotic and non-allelic gene interaction effects have
been developed by Harvey (1960), Eberhart and Gardner (1966), Kinghorn (1980);
Sheridan (1981); Willham and Pollak (1984), and compared by Eisen (1989); and
Jacobec et al. (1991).
The effects
of changes in heterozygosity among alternative types of matings on expression of dominance (d) are partially
confounded with those of possible non-allelic gene interaction (gg and dd).
Also, the number of potential genetic parameters (Table 3) is large, especially for traits of progeny that are
influenced indirectly by maternal (gM, hM, rM)
or even paternal (gp, hp, rp) genotype
(e.g., seasonal date of birth, fertility of matings). Thus experimental
estimation of effects from each genetic source can be difficult. The major
objective is to determine the magnitude of additive breed differences,
heterosis in crossbreeding, and of deviations from expectations for purely
additive and dominant gene effects (Tables 1, 2 and 4). These effects for
individual components need to be combined into
the total effects on production efficiency, using the relevant economic weights
(partial regressions) for component traits (Dickerson, 1982).
Efficiencies
of alternative breeds and breeding systems can then be compared (Table 3),
including effects of reproductive rate on proportion of purebred vs crossbred
matings in each system. If epistatic
deviations from linear association with heterozygosity are important and
negative, specific crossbreeding systems will tend to be more efficient than
new composites in using crossbreeding heterosis, especially if large breed
differences in maternal vs individual performance can be exploited and if a
high species reproductive rate minimizes the proportion of industry
purebred matings required to produce the crossbred replacements. If deviations
from dominance expectations are minor, heterosis retention in a composite can
be high and choice of this alternative could approach maximum industry
efficiency for species with low reproductive rate, when the composite is used
as a straightbred only or as a specialized maternal stock in terminal-sire
crosses (Dickerson, 1973). Systematic crossbreeding is probably impractical in
much of world livestock production because it requires progeny identification
by breed composition and two or more separate breeding herds or flocks. Thus
use of composites often may be the most feasible approach for using breed and heterosis effects to improve production
efficiency, when the costs of maintaining separate breeding herds for
crossbreeding systems are considered. Continued use of crossbred sires of the parent breeds to maintain or to improve an
adapted composite also can minimize parental breed costs.
If the
performance ranking of alternative breeds or crosses differs significantly
between predictable environments (e.g., between temperate vs tropical
climatic zones), evaluations of alternative genotypes obviously should be made
within the same environment in which they
are to be used (Hammond, 1947). This general principle applies in choosing
genotypes for use in any predictable production-marketing system. However,
selection of genotypes for use across a variety of important but random
and unpredictable environments is most effective when based on average
evaluations across a sample of those environments (Dickerson, 1962). In
choosing genotypes for use in environments which seriously limit expression of
genetic potential (e.g., for survival, reproduction, growth or lactation),
economically feasible improvements in the production environment should be
considered before choosing the environment in which the alternative genotypes
will be evaluated. A serious mistake to be avoided is choosing breeds for use
in one environment based on evaluations under another environment in which the
ranking of breeds is seriously different.
Some
environmental factors have obvious interactions with genetic potential. For example, differences in exposure to pathogens or
parasites definitely affect the expression of differences in genetic
resistance. Here, unless an alternative of immunization or eradication is
possible, genetic evaluations must be done under exposure. Less extreme
differences in such environmental factors as temperature, humidity, nutrient
availability, reproductive management or market preferences also can change
performance ranking of alternative genotypes. Alternative genotypes include
both average breed transmitted effects and those from various levels of
crossbreeding heterosis. In some cases, the increased average breed effect in the backcross to the higher (milk)
producing breed may offset the reduction of one-half in heterosis,
compared with the F1 crossbred (e.g., Syrstad, 1989).
Finally, to be most useful, evaluations
need to include effects of all important traits on the lifetime efficiency of production
(e.g., mortality, culling, fertility, body size, and replacement costs rather
than only first parity or survivor
lactation milk records). This requires assessment of the relative economic
importance of component traits, as discussed under Performance Objectives.
The
appropriate experimental measure of genetic by environmental interaction is the
product moment correlation (rG) between
performances for the same breed genotype (G) in the contrasting environments
(Falconer, 1952; Robertson, 1959; Dickerson, 1962; Yamada, 1962; Wilson, 1974).
Differences between environments in only the scaling of genetic effects do not
reduce the genetic correlation. However, differences between environments in
only the scaling of genetic effects do reduce the intra-class estimate of
genetic correlation (r'G)
obtained from the ANOVA genetic (VG) and interaction (VGF.)
variance components of variance
because they inflate the 
. component 
. This underestimation of true rG can be avoided by computing the product
moment correlation separately for each possible pairing of environments, or by
adjusting the interaction variance 
for scaling effects to 
before calculating
true 
(Robertson, 1959). The second alternative
requires separate estimation of the total genetic 
within each
environment. Both are more labourious than avoiding the problem by standardizing
phenotypic variation within each environment before doing the ANOVA (i.e., dividing observations within each environment by the standard deviation
in that environment, Dickerson, 1962).
