PRINCIPLES

OF

COMMERCIAL

POULTRY

BREEDING

¥1

I. MICHAEL LERNER

v I -

UNIVERSITY OF CALIFORNIA • COLLEGE OF AGRICULTURE

AGRICULTURAL EXPERIMENT STATION and EXTENSION SERVICE

THE LIBRARY

OF

THE UNIVERSITY

OF CALIFORNIA

DAVIS

I. MICHAEL LERNER

Principles of

COMMERCIAL

POULTRY

BREEDING

A semi-technical account of recent developments in
genetics applied to breeding for the improvement
of economic traits in poultry. For the breeder, the
hatcheryman, and the commercial poultryman.

UrtiVfcJKbl I Y OF CALIfURIMiA

LIBRARY

COLLEGE OF AGRICULTURE
DAVIS

LIBRARY

university of california
Davis

THE AUTHOR: I. Michael Lerner is Associate Professor of Poultry Husbandry
and Associate Poultry Husbandman in the Experiment Station, Berkeley.

Contents

THE GENETIC BACKGROUND 3

The Basis of Inheritance 3

Sex Linkage 3

Homozygosity and Heterozygosity 4

Dominance 5

Genetic and Phenotypic Variation 5

Heritability 6

Measuring Heritability 8

Non-Additive Genetic Variation 9

The C Effects 9

Heritability and the Fixation of Characters 10

The Breeder's Tools 11

SELECTION .12

Migration 12

Types of Selection 13

The Efficiency of Individual Selection 14

The Efficiency of Family Selection 15

Individual and Family Selection Compared 16

Combination Selection 18

The Weighting of Family Averages 18

C Effects and Family Averages 19

Selection Criteria 21

Part versus Full Production Records 22

Selection for Several Traits 23

Total Score Selection 25

Selection Procedures 26

Selection and Culling 28

MATING 31

Mating Systems 31

Inbreeding 32

Other Mating Systems 35

PRACTICAL APPLICATIONS 37

The Heritability of Economic Traits 37

Breeding from Pullets 40

Other Details of Breeding Plans 41

THE COMMERCIAL POULTRYMAN 43

Breeding Methods and the Commercial Poultryman ... 43

Evaluation of Advertising 44

Laying Tests and Official Improvement Schemes 45

Conclusion 46

Reading 47

Digitized by the Internet Archive

in 2012 with funding from

University of California, Davis Libraries

http://archive.org/details/principlesofcomm01lern

Zk

1>V HltiiiHiil outlines the current status
of the techniques of commercial poultry breeding in the light of recent de-
velopments in genetics and applied animal breeding. It is now half a century
since the basic principles of inheritance were recognized. Since then the
advances made in fundamental genetics have been reflected in the theory and
practice of breeding only to a limited extent. The general scheme of trans-
mission by inheritance of characters which can be classified as simply present
or absent in a bird (like "rose comb") has been fully understood and ex-
ploited by many practical breeders. But the problem of breeding for economic
traits like egg production, whose expression is usually continuous— that is,
not naturally falling into a few classes such as are formed by the more com-
mon comb types— has been a difficult one to solve. It is not that breeding
methods have not been successful with such characters. The difficulty has
been in making them efficient, and putting them on a quantitative basis.

Although the genetic principles of the inheritance of continuous traits
were formulated some years ago, they have only recently been applied to
economic characters. The method involved calls for a somewhat different
outlook than we have used in the past. This manual has been written to
explain it.

There is no denying that the more we learn of a biological subject the more
complex it becomes. Commercial poultry breeding is no exception. Under-
standing the reasons and implications of many statements to be made in the
course of our discussion may require an advanced statistical and biological
training. To make this manual useful to the commercial poultryman at large
as well as to the hatcheryman and specialized breeder, you will be asked to
accept some statements on faith. No matter what old-fashioned practitioners
had to say to the contrary, breeding is a mathematical process. It is impos-
sible to eliminate mathematics from a full discussion of the subject. But
formulas can be held at a minimum if you are willing to accept a few without

[1]

I PRINCIPLES OF COMMERCIAL

proof. The specialist who wants to verify the conclusions to be brought out
may consult the vast and intricate literature on the subject, which is scat-
tered in many of the technical journals. If others remain unconvinced, they
will have to seek remedy by becoming specialists themselves.*

* References to the many contributions to the subject of genetics in relation to animal
improvement have been omitted for the sake of readability. A few suggestions for further
reading will be found on page 47. 1 am glad to record here the names of three geneticists
on whose theoretical work much of the discussion to follow is based. They are Sewall
Wright of the University of Chicago, R. A. Fisher, now of the University of Cambridge,
and J. L. Lush of Iowa State College.

POULTRYBREEDING 3

Zhe Qenetic background

The Basis of Inheritance

The unit of inheritance in all forms of plant and animal life is the
gene. The nucleus of each sperm and each egg contains an assortment of
threadlike bodies known as chromosomes, which are composed of genes.
Fertilization involves the union of a sperm and an egg nucleus, so that the
newly arisen individual possesses two sets of chromosomes, and therefore a
pair of each of the different genes. Half of the inheritance thus comes from
the sire and the other half from the dam, and all the body cells normally con-
tain within their nuclei representatives of both paternal and maternal genes
in paired form. When the individual itself starts producing germ cells, a
regular process of reduction of chromosome number happens. Thus the
number of chromosomes entering each germ cell is half of that found in body
cells. The identity of the paternal and maternal set of chromosomes is not
preserved in this process, so that the chromosomes are re-sorted in each
generation.

There are 39 chromosome pairs in each cell of a chicken. One member of
each pair is derived from the sire and the other from the dam. The bird
itself will contribute 39 single chromosomes to its own offspring, but which
of them will be of paternal and which of maternal origin is determined by
chance.

Sex Linkage

One member of the set of 39 behaves differently. It is called the
sex chromosome. The male has a pair of these, the female a single unpaired
one. When the germ cells of the female are formed, half of them will each
contain one set of 39 chromosomes. The other half will each contain 38 of the
regular chromosomes but will lack a sex chromosome. If a germ cell of the
first type is fertilized, the new individual will be a male, since it will contain
a pair of sex chromosomes in addition to the regular set of 38 pairs. If an
egg of the second type (lacking a sex chromosome) unites with a sperm, the
new individual will be a female: it will once more contain 38 pairs and an
unpaired sex chromosome.

Genes carried on the sex chromosome are called sex-linked. They differ
from the others (autosomal genes) in that, as above, a female receives her
complement of sex-linked genes only from her sire, and in turn transmits
them only to her son. A male, on the other hand, receives one set of sex-
linked genes from each of his parents and similarly transmits a full set to
both his sons and daughters.

4 PRINCIPLES OF COMMERCIAL

Homozygosity and Heterozygosity

When any given gene pair contains identical members, that is to
say when the member of the pair contributed by the sire is the same as the
one contributed by the dam, the individual involved is said to be homozygous
for that gene. When the two genes are different, the individual is heterozygous.

Genes are commonly designated by letters. Different members of a given
gene pair are distinguished either by using capitals for one and lower-case
letters for the other (like R and r) , or by subscripts (like A1 and A2) . Thus
for a given gene pair A, it is possible that the bird received from its father the
gene At, while the mother contributed A2. The heterozygous bird will have
the constitution AXA2. A homozygous bird may be of the constitution A1A1
or A2A2. In the course of germ-cell formation, only one member of the pair
will enter each sperm or egg. Hence heterozygotes will pass Ax to one half of
their offspring, and A2 to the other half. Homozygous birds will, however,
produce only one kind of germ cell: if their constitution is A1A1 it will be A1 ;
if A2A2, it will be A2.

When a character depends for its expression on a single gene pair, whose
effect is not greatly modified by environment, we can often distinguish the
two types of homozygotes from each other and sometimes from the hetero-
zygote. Take plumage color in the Blue Andalusian fowl. The two homozygote
forms are respectively black (A^A^) and blue-splashed white {A2A2), while
the heterozygote (A1A2) is blue. We can readily see that homozygous birds
mated within their own type will breed true. Two homozygotes of contrast-
ing types mated to each other will always produce heterozygous offspring.
The blue birds, being heterozygous, can never be true-breeding. Their eggs
will be of two kinds, At and A2. Each of these has an equal chance of being
fertilized by its own or by a different kind of sperm. Thus on the average,
one quarter of the offspring will be black, one quarter white, and one half
blue, as the following diagram shows.

Kind of
sperm

Kind of eggs

A,

A2

A,

A,At

AXA2

A2

AXA2

A2A2

Among every four offspring the average ratio will be one A1A1: two AtA2 :

one A 2A2.

POULTRY BREEDING

Dominance

There are other cases where one kind of gene may be dominant
to the other of the pair. Here the heterozygote will not be distinguishable
from the dominant homozygote.

Thus we may designate a pure-breeding rose-combed bird RR, indicating
that it has a pair of identical genes. A single-combed bird will be of the con-
stitution rr. It possesses a pair of recessive (r) genes. The crosses between
these types will be heterozygotes (Rr) but will exhibit the dominant trait. In
other words, an Rr bird cannot be distinguished by its appearance from RR
birds, since it will also be rose-combed. But the breeding behavior of the
two types will be different. When an RR bird is mated to an rr bird, all of the
first generation offspring will be rose-combed as indicated. But among the
first generation progeny of an Rr bird mated to single-combed birds (rr)
one half will be rose-combed and one half single-combed:

Kind of germ cells
produced by rr
birds

Kind of germ cells
produced by Rr birds

R

r

r

Rr

rr

On the average, half of the offspring will be Rr and half rr.

This example demonstrates that the appearance of a bird (its phenotype)
may be different from its breeding potentialities, that is to say from its ac-
tual genetic makeup, or genotype. Here the difference is caused by the phe-
nomenon of dominance. There are other and more usual reasons why the
correspondence between phenotype and genotype is not perfect.

Genetic and Phenotypic Variation

Suppose we are dealing with a quantitative character such as
body weight. The number of gene pairs contributing to it is probably very
large. Ordinarily, when more than two or three gene pairs are involved in
the inheritance of a character, it is unprofitable and almost impossible to
isolate and measure the effects of each. We have no way of writing out the
genetic constitution of any given individual for body weight in the way we
have for comb shape. We can weigh a bird and determine that it has a pheno-
type of three pounds, five pounds or whatever it may be for any specified age.

6 PRINCIPLES OF COMMERCIAL

The phenotype is the product not of the bird's genetic constitution acquired
from its parents, but of the interaction between its genes with various non-
genetic or environmental forces. The more the actions of the genes involved
are susceptible to environmental modification, the less accurate will be our
judgment as to the bird's actual genotype.

Statistical methods have been devised to measure approximately what per-
centage of the variation of a given character is due to genetic forces and what
to environmental influences. In body weight, the percentage of genetic varia-
tion is about 40 per cent, and the environmental percentage about 60 per
cent. The statistic expressing the percentage of genetic variation is known
as the degree of heritability . Heritability for body weight is thus about 40
per cent, usually written .40.

Heritability

The degree of heritability is an exceedingly important figure for
us. It determines the amount of gain which breeding selection can accom-
plish. More important, its magnitude governs the choice of an efficient selec-
tion method. So it is vital to understand what heritability (we shall call it h2)
actually means.

Heritability does not mean that 40 per cent of a given animal's body
weight is due to heredity and 60 per cent to environment. What the state-
ment that for body weight h2 approximates .40 means is this : that in a popu-
lation of birds there will be differences between individuals, partly caused
by the fact that each bird has a somewhat different genetic constitution, and
partly by the fact that each bird has been under an environment to some extent
peculiar to itself; the degree of heritability is the fraction of the total of such
individual variation which is traceable to genetic differences. For body
weight this fraction is roughly 40 per cent.

Standardizing Heritability. If the birds in a population are deliberately
subjected to different environments, the total variation among them will be
increased. But the absolute amount of genetic variation will remain un-
changed. In other words, heritability will drop. If we call the genetic varia-
tion G, and the environmental variation E, the degree of heritability will be

r

h2 =

E + G

Suppose by manipulation of management the amount of E is doubled. The
degree of heritability in this new population,

h-2ETG'
will now be obviously less than it was in the first place.

POULTRYBREEDING 7

In order to have a uniform standard for heritability, it therefore seems best
to use the term to mean the relative amount of genetic variation when the en-
vironment is uniform or random for all the birds in a population. This means
that the E fraction of the total variation will be due to uncontrollable environ-
mental differences. Hence, when we say that body weight has a heritability
of .40, we mean that when all the birds in a population are raised and kept
in the same houses, are given the same diet, and in general are subjected to
uniform treatment, 60 per cent of the total variation is still not traceable to
genetic differences between them. This fraction then must be due to de-
velopmental accidents and deviations, to noninherent peculiarities of in-
dividuals, and to other sources of this kind.

The conditions of genetic variation must also be standardized. We have
shown that it is possible to double the E fraction by manipulation of en-
vironment. It is also possible to modify (either increase or decrease) the
absolute amount contributed by the G fraction to the total variation.

Suppose the population we deal with consists entirely of full sisters or
brothers. The amount of genetic variation in such a population will be much
less than in an ordinary flock of chickens which contains full sisters, half
sisters, cousins, less closely related individuals, and some with no common
ancestors for many generations. This should be clear from the fact that full
sisters have more genes in common with each other than unrelated individu-
als. Some genes will be held in common by all birds in a flock. Genetic rela-
tionships are measured from a base representing the average proportion of
genes common to all members of a population and taken as zero. If we say,
then, that two non-related individuals have a genetic relation of zero, and
that animals with exactly the same genetic constitution (such as identical
twins in mammals) bear a genetic relation to each other of 1, full sisters are
related to each other to the extent of .5.

In a group of full sisters, then, the amount contributed to the total vari-

Q

ation will be — as compared to G in a random-bred population. The degree
of heritability in this restricted population will be

E + G/2

The same principles apply to any population in which the amount of genetic
variation is reduced by such means as inbreeding. Conversely, the G frac-
tion may be increased in a population created by a mixture of completely
unrelated inbred lines of the same breed, or by a mixture of different breeds.
We therefore adopt another convention to standardize the meaning of the
term heritability. By the degree of heritability we mean the proportion of
the variation which is genetic in a random-bred flock, that is to say, a flock

8 PRINCIPLES OF COMMERCIAL

in which males and females are mated together without regard to their geno-
typic or phenotypic resemblance to each other. Heritability determinations
made on a different basis can be statistically reduced to such a common form.
The heritability estimates we shall make later for various traits have been
so reduced.

It should be clear now that what the degree of heritability measures is the
accuracy of identification of the genotype from the phenotype, or the correla-
tion between them. The higher the heritability, the greater is this correlation.
If the heritability of a character is 100 per cent (or 1 in our terms) it means
that the genotypic value of a bird coincides with its phenotypic value. At
the other extreme, a heritability of zero would mean that the genotype and
phenotype are not at all correlated : that all of the variation in the phenotype
is due to non-genetic or environmental sources.

Measuring Heritability

It may be easier to get the idea of heritability if we give a simpli-
fied example of one method of computing it. Suppose in a random-bred
flock we have several groups or families of full sisters as well as a number
of half-sister families (birds from different dams but sired by the same
male). We have already noted that the genetic relationship between full
sisters is equal to .5. The similar figure for half sisters is .25. If we have body
weights, for example, for all of the birds in the flock, we will note that those
of full sisters will resemble each other more than those of half sisters do. As
a matter of fact, the genetic resemblance between the former will be twice
that between the latter, because as indicated by the respective genetic rela-
tionships, full sisters have twice as many genes in common as half sisters
(that is, twice as many of the genes which are not common to the whole
flock).

The phenotypic resemblance within any group of birds can be computed
from measurements on them by means of what is known as the coefficient of
correlation. Now if we find that the coefficient of correlation for body weight
between full sisters is .2 and between half sisters .1, we can attribute the
difference (.1) to the fact that the genetic relationship in the first case was
.5 and in the second .25. Thus, for every .25 difference in the genetic rela-
tionship, there is a difference of .1 in the phenotypic correlation coefficient.
The figure .1 corresponds to one quarter of those differences between pheno-
types of completely unrelated birds which are due to differences in their
heredity. All of the genetic difference between them is then four times this
fraction. In other words, the heritability of body weight is 4 x .1 or .4.

So far we have considered the total phenotypic variation in terms of its
two components, genetic and environmental. Each of these can be subdivided

POULTRYBREEDING 9

further. The full complexities of the situation would take us far afield, but
two points must now be made because we shall need to refer to them in our
later discussion. One refers to genetic variability and the other to environ-
mental variability.

Non-Additive Genetic Variation

We have assumed that the action of genes is additive. This means
that if gene A has a given effect on the trait in question, a bird having the
genetic constitution AA will show twice the effect. To say it in another way,
bird AA will be as different from bird Aa as the latter will be from bird aa.
Additiveness also means that for two or more gene pairs, the combined effects
will equal the sum of the individual effects. Take two pairs of genes, A -a and
B-b, for which the differences between the effects of the capital-letter and
small-letter genes are equal. The bird possessing the genetic constitution
A Abb will be superior to one of the genotype aabb by a certain amount. The
amount will equal the difference between birds aaBB and aabb, since we
assumed that the differences between A and a, and between B and b are equal
in effect. The additive idea means that genotype AABB will be superior
to A Abb, AaBb or aaBB to the same extent that each of them is superior to
aabb.

Dominance and interaction between different gene pairs may interfere with
additive gene action, but it so happens that the idea of additiveness seems
reasonably well justified for many quantitative traits in random-bred flocks,
at least in such flocks at lower levels of performance. However, when inten-
sive inbreeding is practiced, or after a long period of selection, nonadditive
gene effects apparently become significant.

The difference in the two situations means in practice that whereas in aver-
age random-bred flocks, knowledge of the genotypes of two prospective
parents permits the prediction within certain limits of the performance of
their offspring, in the case of matings between inbred birds, or birds at
upper levels of improvement, such prediction is usually not possible.

The C Effects

We have shown that one of the ways of determining the degree
of heritability depends on the closeness of resemblance between full sisters,
as compared to that between half sisters. It was assumed that the difference
between the two is due to the fact that full sisters hold more genes in common
than half sisters. But they may also resemble each other more because, being
full-sister embryos, their pre-hatching environment was supplied by the same
dam, while half sisters were provided with an embryonic environment by
different hens. If such additional differences exercise significant effects, we

10 PRINCIPLES OF COMMERCIAL

may have included in what we assumed to be net genetic differences also
some environmental differences.

We designate environmental differences of this type as environmental ef-
fects common to members of the same family, and assign to them the letter C,
in the same way as we assigned G to genetic differences and E to uncontrolled
environmental effects. Besides strictly maternal influences, the C fraction
may include other effects when, for instance, members of each family are
housed separately, and thus subjected to non-random environment.

So far few C effects have been found for chickens when all birds irrespec-
tive of family origin are housed together. For instance, with respect to body
weight, although the C factor is great at hatching time (because of the egg
size characteristic for each given dam, which exercises such powerful effects
on the weight of day-old chicks), by the time the birds reach maturity it is
only about 5 per cent. More significant C effects however are apparent in egg
traits, such as shell thickness, and possibly in percentage of firm white. We
shall later come back to the problem of the C fraction.

Heritability and the Fixation of Characters

The heritability concept has an important application to the pos-
sibility of fixing desirable characters in a flock. In the early days of Men-
delian genetics, it was assumed (and many still believe this) that if all the
genes controlling a given trait were identified, it would be possible to produce
by selection a completely uniform flock with respect to their performance
for this trait. It became known somewhat later that this objective cannot be
accomplished by selection alone and that intensive inbreeding must be
brought into play. But we can now see that even if inbreeding is carried to
the point where every bird in the flock is of exactly identical genetic consti-
tution, uniformity of performance will not follow.

The genotype determines only the hereditary part of the variation and not
the environmental part. If, for body weight, we were to remove all the genetic
sources of variation, 60 per cent of the original variation in phenotype would
still be exhibited in the genetically uniform group of birds (h2 being .40).
Many other productive traits have an even lower heritability than body
weight. Only a minor fraction of variability in them can be removed by
inbreeding. There are other aspects of inbreeding which we shall mention
later. But we may say at once that not only the possibility but even the de-
sirability of complete uniformity is doubtful. A flock of birds all possessing
identical genotypes cannot be improved further by genetic means. Of course,
if their genetic constitution is the best possible, they need no further im-
provement. But the difficulty is that there is no such thing as the best possible
genotype for all environments.

POULTRY BREEDING 11

Suppose a slight change in the environment occurs, such as the onset of
a hitherto unexperienced disease: the genotype which was superior in the
previous environment will no longer be superior in the new. Attainment of
complete uniformity thus destroys the flexibility of a population, making it
rigid and unadaptable to any unpredicted changes of environment. When
uncontrolled environment plays such a dominating role in determining pheno-
typic levels of performance, as it does in the case of economic characters, the
objective of complete genetic uniformity cannot be a wise one. Eventually,
perhaps, methods will be discovered to overcome this difficulty (for example,
a complete control of environment) . But today a genetically flexible flock is
both the desirable one to have and the only practicable kind.

The Breeder's Tools

There are three and only three ways in which a breeder can con-
tribute to the genetic improvement of his flock. First, he can decide which
birds of those available will become the parents of the next generation. This
is selection. Second, it is in his power to decide which particular male will be
mated to any given female. This involves a choice of a mating system. Third,
he can within certain limits determine what proportion of the next generation
will originate from each of the birds selected to be parents. This is his power
to control reproductive rates.

For our purposes, reproductive rates may be included as part of selection.
Although theoretically the breeder can decide to have a variable number of
offspring from each mating in the next generation of his flock, in practice it
is simpler to permit free and unrestricted reproduction of the chosen parents
throughout the length of the normal hatching season. Only occasionally is
it worth while to produce an extra hatch or two from part of the selected
group of birds. Such a procedure will lead to confusion if followed too often,
because the environmental effects of out-of-season hatching may make com-
parisons difficult between the performances of birds widely differing in
hatching date. Late-hatched birds from superior families may, of course, be
profitably used in a hatchery flock for production of commercial breeding
stock. Measurements on such birds however are of restricted value.

Another way of controlling reproductive rates, besides extending the
hatching season, is to repeat the same matings in successive years, so that
the repeated pairs of parents will contribute more descendants to following
generations that unrepeated ones. The decisions to be made about repeating
birds previously used as parents are really a matter of selection. Perhaps the
basis of selection in such cases is different from that used for birds previously
not mated. But it is still a question of selection, and to some extent of mating
system (page 31).

12 PRINCIPLES OF COMMERCIAL

Selection

In practicing selection, the breeder may choose the parents of his
next year's flock either from his own breeding stock or from someone else's.
Our discussion will limit itself to the first of these methods. A word about the
second— migration— will explain why we will not be concerned with its details.

Migration

The use of breeding stock produced outside the breeder's own
flock involves what is known in genetics as migration. Each closed flock (a
flock reproduced entirely from its own members) contains within it an assort-
ment of certain genes, desirable, undesirable, and indifferent, so far as their
effect on the character selected is concerned. The task of the breeder working
with such a flock or isolate, as it is called, is to increase the concentration of
the desirable genes at the expense of the undesirable ones. When stock from
another isolate Js introduced, different genes from those already present in
the home flock may be brought in, or more likely the proportions of the same
types of genes may change rapidly. In either case, instead of a continuous
increase in frequency of desirable genes or wanted genetic combinations, a
certain discontinuity results. Since there is no way to identify the new genes,
products of the introduced germ plasm must be subjected to the same per-
formance tests applied to the original birds. In this way selection is made
operative on the migrants before their contributions are fully incorporated
in the fund of genes already present.

In general, it can be shown that the most efficient way to improve the total
population genetically is to maintain a large number of non-interbreeding
flocks. Application of proper selection procedures and mating systems within
each flock will lead to a rise in average genetic quality. Then, occasionally,
when rate of improvement within a given isolate slows down or stops because
of the exhaustion of potentialities for improvement, an introduction from
another isolate can be made, and a similar process of upgrading once more
undertaken. Migration should thus be an exception rather than a rule, and
the selection principles which are appropriate to the breeder's own stock
should be similarly applied to the introduced migrants.

Systematic crossbreeding may be a good method of producing com-
mercial stock for specific purposes (e.g. broiler production), but it is not a
technique for genetic improvement. A crossbred generation may be superior
to its parents, but the superiority achieved is a dead-end one: it cannot be
utilized for further improvement. Naturally this refers to the scheme of
crossing breeds, strains and lines afresh every year, and not to methods which

POULTRYBREEDING 13

involve crossbreeding for foundation and continued selection from the closed
group for the purposes of synthesizing a new breed, variety, or strain.

To return to selection, the general problem presents two aspects: 1, the
choice of animals upon whose performance judgment with respect to any
given individual will be based (see below) ; 2, the choice of measurements
to be made on these animals (page 21).

Types of Selection

So far as the choice of animals is concerned, there are three bases
of selection : mass, pedigree, and family. The first is also known as individual
selection. It refers to the simple evaluation of the breeding worth of birds
from their own phenotypic performance on the trait actually selected for, or
with respect to some other character. Thus in attempts to improve egg pro-
duction by breeding, mass selection for females can be conducted on the basis
of their trapnest records. The mass selection of males (and of course, if de-
sired, of females) can be carried out on the basis of breed type, health, body
weight or characters other than the egg record itself.

The method of pedigree selection makes use of information on the ances-
tors of the animals to be chosen. Whereas mass selection can be carried out
without knowing the identity of a bird's parents, pedigree selection (as well
as family selection) requires that the ancestry of the birds in the flock should
be known. The simplest form of pedigree selection considers the bird's par-
ents. More elaborate schemes may call upon information about grandparents
or more remote ancestors. We need not consider this method in too great
detail, because by comparison with the others it is not very efficient.

It is true that in the past (and even today in many animals) pedigree selec-
tion has been extensively used. The idea of blood lines, so much beloved by
the breeders of racehorses and to some extent of cattle, is founded on this
type of selection. Even in poultry one can often find breeders wasting their
time in constructing pedigrees carried back 20 or more generations, in the
belief that knowledge of remote ancestry is an aid to judgment of the breed-
ing worth of an individual. In sexual reproduction, however, the contribution
that each given ancestor makes to an individual is on the average cut by half
in every generation intervening between them. Thus an ancestor five genera-
tions removed can be expected to have contributed on the average approxi-
mately only S1/^ per cent of the bird's genes; an ancestor removed ten
generations will have contributed less than one-tenth of 1 per cent of the
total hereditary makeup of the individual (unless it appears in the pedigree
many times) . Surely such a small fraction of the genetic constitution can add
little to the accuracy of judgment of the bird's genetic merit.

14 PRINCIPLES OF COMMERCIAL

It may be argued that each of the parents contributes half of the genes
comprising the genotype of the bird, and so should be of great value in as-
sessing the individual's worth. But it must be remembered that knowledge of
the phenotypes of the parents is a different thing from knowledge of their
genotypes. The phenotype is determined partly by heredity and is useful to
the extent that it is; but if we are to judge by the phenotype, the individual
itself is more useful than even its immediate ancestors (except in such cases
as egg number in males, for which there is no phenotype). If we decide to
bring into consideration the genotypes of parents rather than their pheno-
types, we are not using simple pedigree selection, but family selection in one
or another form, because more accurate information on genotypes can usually
be obtained by consideration of the performance of the bird's relatives.

Family selection is of two types. We can evaluate the respective genotypic
merits of a series of birds, first by the performance of their collateral rela-
tives, sisters, brothers, half sisters or individuals of different degrees of
relationship, or second, by the performance of descendants. Of the first kind
of family selection only full sister or brother and half sister or brother per-
formances offer worthwhile opportunities. The comparatively low degree of
genetic similarity between more remote relatives introduces too much inac-
curacy to be very useful. The second kind of family selection will be recog-
nized as the progeny test. Before considering full- and half-sib testing (sib is
a word meaning sister and/or brother) separately from progeny testing, we
will deal first with the general merits of family selection (or family testing) .

Our special interest is in the comparative advantages of individual selec-
tion and family selection. They can be used separately or in combination. By
considering the properties of each we should be able to decide under what
circumstances does one or the other or a combination of both provide for
the most efficiency.

The Efficiency of Individual Selection

We have noted that the degree of heritability indicates the ac-
curacy with which the genotype can be identified from the phenotype. Now it
must be realized that genetic improvement depends on the breeder's skill in
choosing for reproduction the superior genotypes present in his population.
Only that part of the phenotypic excellence which is due to genetic causes is
transmitted from generation to generation. It then seems fairly obvious that
the rate of improvement obtained by individual selection is a direct function
of the degree of heritability.

Suppose that out of a population averaging 2000 grams in body weight (we
may forget the difference in size between the two sexes for now) , the group
selected as parents of the next generation had a mean weight of 2400 grams.
The phenotypic superiority of the chosen birds over the average for their

POULTRY BREEDING 15

generation, or the so-called selection differential, is then 400 grams. This does
not mean that their offspring will exceed the previous generation by that
amount. Since the heritability of body weight is .4, only this fraction of the
selection differential will be added to the previous mean. The balance can be
considered to have been due to non-genetic (or non-additively genetic)
sources. Hence we may expect gains from selection to be 400 x .4 or 160
grams. The average body weight of the offspring of the selected parents will
be 2000 + 160 or 2160 grams.

The Efficiency of Family Selection

What will family selection accomplish under the same circum-
stances? The reason that the family is brought into the discussion is that it
provides additional information regarding an individual's genotype beyond
what is obtainable from its own phenotype. We said that full sisters bear to
each other a genetic relationship of .5. Half of the genes which are hetero-
zygous in the population are then held by them in common form. This is the
same as saying that the determination of a bird's genotype by the phenotype
of her sister is one half of the degree of heritability. Thus if we were to
assess the genotype of a female from the phenotypic performance of one of
her sisters, we would be only half as accurate in our judgment as we would
have been had we used her own phenotype for this purpose.

But when we have more than one sister as a base for our attempted evalua-
tion, our accuracy is increased. We may expect that the environmental forces
affecting the genotype of each individual sister will operate at random: they
may reduce the phenotypic expression of one sister below its true genotypic
merit and they may increase it above the genotypic worth in the case of
another sister. The more sisters there are in the family, the better is the chance
that the environmental influences will cancel each other out. If we had an
infinite number of sisters their phenotypic average would correspond exactly
to their genotypic average. This would, of course, give us perfect accuracy so
far as a family average is concerned, but it would still be less than perfect for
evaluation of the genotype of any one of the sisters in the family.

We can now show that the use of the family average increases the accuracy
of selection in accordance with a formula which looks complicated but is
really simple. If we call the heritability of a trait h2, the genetic relationship
between the members of the family r (equal to .5 in the case of non-inbred
full sisters) , and the number of individuals in a family n, then the accuracy
of identifying genotypes from family averages is

nrh3+ (l-r)h2
l+(n-l)rh2
times the accuracy obtained from individual records.

16 PRINCIPLES OF COMMERCIAL

Let us take the case of full sisters, and substitute for r the value .5. The
formula now becomes

.5nh2+.5h2

or, more simply,

l + .5(n-l)h2'

(n + l)h2
2+(n-l)h2'

Now, suppose that each of the families contains five sisters and substitute
the number 5 for n. We get

6h2
2 + 4h2'

All we need to know now is the heritability of the character in question to
determine whether full-sister family averages are more accurate than indi-
vidual records. For body weight, where h2 is .4, the comparative accuracy of
individual to family records (with five sisters per family) will be as .4 is to

6x.4

9

2 + 4 x .4

that is, as .4 is to .67, or approximately as 1 to 1.7. If the average number
of sisters in a family is increased to 10, the comparable ratio will be about
lto2.

In the same way we can compute the relative accuracy of half-sister aver-
ages (r = .25) and of progeny tests (r between parent and offspring = .5 ) .

Individual and Family Selection Compared

The figures obtained so far do not tell the whole story. They do
indicate the relative accuracy of genotype identification achieved by the use
of different selection methods, but they do not answer the question, which of
the two methods is actually more efficient. They do not because the genetic
gains possible depend not only on the accuracy of genotypic identification but
also on the intensity of selection used. Thus, in the example for body weight,
we found that the genetic gain from individual selection will be 160 grams
when the selection differential is 400 grams. Obviously if the selection differ-
ential were only 200 grams, the gain obtained would have been only 80 grams.
The selection differentials possible under individual selection are greater
than those under family selection. Consider a flock of birds averaging 2000
grams and ranging from 1400 to 2600 grams in body weight. When selection
is based on the individual phenotypes, it should be possible to choose enough
birds to produce next year's flock from those weighing 2300 grams or over.

POULTRYBREEDING 17

The average weight of the birds chosen will be, say, 2400 grams, and so the
selection differential will equal 400 grams. But if we consider family averages,
the range of variation will not be the same 1400 to 2600 grams. Since environ-
mentally induced deviations from genotype will tend to cancel each other out
to some extent, it is more likely that the best family will have an average of
only 2350 grams and the poorest 1650 grams (the actual range will depend on
the number of sibs in a family) . To choose on the family basis the same per-
centage of the flock which was needed for reproduction in the case of indi-
vidual selection, it may be necessary to use families averaging as low as 2100
grams, as against the lowest individual weight of 2300 grams. The average
of the selected families will thus be lower than the average of individuals
selected by the first method. Consequently, the selection differential now, in-
stead of being 400 grams, may be reduced to something like 250 grams.

The use of family averages instead of individual records thus has two
effects on selection : it increases the accuracy of choosing superior genotypes,
but it reduces the selection differential. There is a formula for computing
the joint effects of these two factors which we can use to make the comparison
between the overall efficiency of individual and of family selection. Accord-
ing to this formula, for full-sister families the gains obtained from individual
selection will be to those obtained from family selection as 1 is to

n + 1

V2n[2+ (n-l)h2]
With families of five each, the formula becomes

V20(l + 2h2)
Similarly, with families where n equals 10, we have

11

V40(l + 4.5h2)

In the case of body weight, where h2 equals .4, the two respective family
sizes lead to ratios of 1 to 1, and 1 to 1.04. It would seem that when the de-
gree of heritability is as high as .4, family selection does not offer any greater
efficiency than what can be gained from individual selection.

But when we consider characters with lower heritabilities, the results are
somewhat different. Suppose the trait in question has an h2 of .1. Family selec-
tion based on five sisters will be 1.23 times as efficient as individual selection.
Selection based on ten-sister families will be 1.45 times as efficient. The
reader might find it interesting to figure the effect of family size and heri-

18 PRINCIPLES OF COMMERCIAL

tability on the comparative efficiency of the two methods. He will find, in
general, that progeny testing and full-sister family selection, in cases where
between five and 10 sisters appear in a family, will be more efficient than
individual selection when the heritability is lower than it is for body weight.
(Larger full-sister families may lead to complications because they will in-
variably require a prolongation of the hatching season, which increases the
E fraction of variation and so lowers heritability.) He will also discover that
when heritability is higher than .4, family selection may be less efficient than
individual selection and cannot be recommended in place of it.

Combination Selection

It is easy to see that by combining information from an individual
and from its immediate relatives, we could get greater efficiency of selection.
Here is a formula for full sisters which gives the ratio of gains to be ex-
pected from the best combination of family and individual records. The
combination gain will be

V

(n-1) (1-h2)2
±+(2-h2)[2+(n-l)h2]

times as great as the gains from individual selection alone. This general for-
mula assumes no C effects. Note that the ratio will be at least 1 to 1.

Again we suggest that the reader investigate the efficiency of combined
selection as against the other types, by substituting in the above formula dif-
ferent values of n and h2.

To conclude: The choice of a selection method hinges on the degree of
heritability of the trait selected for. When it is high (around .4 or more),
individual selection is a satisfactory method, and becomes more efficient than
family selection as h2 rises. When it is low, combined selection is indicated
for traits where combined selection is possible, and family selection for traits
where it is not (egg production in males, and to some extent viability) .

There still remains the question of how combined selection is to be carried
out. The formula just given refers to the best combination of family and
individual records. How is this combination arrived at? In other words,
what is the relative amount of attention a breeder should pay to individual
and to family averages?

The Weighting of Family Averages

The theory underlying this question is complex. We will simply
give another formula as a guide. This must be presented in two forms, a sim-
plified form for the more usual cases where no C effects are present, and an
example illustrating situations with C effects. In the first instance, the weight

POULTRYBREEDING 19

of the family average as compared to that of an individual record is expressed

by the ratio
y nr(l-h2)

[l+(n-l)rh2](l-r)

Thus for a character with a heritability of .05, the average of full sisters
(where r equals .5) should be given the weight of

n x .5 x .95

[l+(n-l) x.5x.05]x.5'
or, more simply,

38n

39 + n'

To illustrate the use of family average weighting: Suppose the character
considered is the hen-housed average (the production index), which indeed
has an h2 of .05. We wish to decide which of two hens is to be preferred for
breeding purposes : hen A with a record of 300 eggs belonging to a family of
six full sisters (including the bird herself) with a production index of 200
eggs, or hen B with a record of 250 eggs belonging to a family of four full
sisters with an average hen-housed production of 240 eggs. The comparative
breeding value of each of these birds can be expressed by an index which
consists of the family average, given the weighting suggested above, plus her
own record. Further, the records must be expressed in relation to the flock
average. Thus, if the flock average in our example equals 180 eggs, the index
for bird A (substituting 6 for n in the formula) is

I—1! x (200 - 180) + (300 - 180) = 5 x 20 + 120 = 220.
For bird B, the index value is

(240 - 180) + (250 - 180) = 280.

39 + 4

Apparently bird B is to be preferred to bird A. Had the record of bird B been
only 190 eggs (still assuming a family production index of 240 eggs), the
figure 280 would have been reduced to 220, so that the difference between
representatives of the two families compared would be negligible.

C Effects and Family Averages

The problem of characters which may show C effects involves a
more complicated expression which right now is not particularly practical
to use in poultry breeding. We will give an example because it points out
certain pitfalls in breeding systems advocated by those who favor family
selection without individual pedigreeing.

20 PRINCIPLES OF COMMERCIAL

Suppose a breeder who wants to improve the production index decides to
use a family selection scheme, where the offspring of each male is toe-punched
and housed separately so as to avoid individual trapnesting. In other words,
he will have no individual records and will rely solely on family averages.
Each family will consist of a mixture of full and half sisters in which the
average genetic relationship (r) will be about .25. If there are 50 such birds
in a pen, the general formula for the family weighting factor which takes
into account the possible C effects is, for the production index, approximately

10 -45c2
1 + 33c2

where c stands for a fractional value comparable to h2, in that it measures
the proportion of the total variation attributable to C effects, just as h2 meas-
ures the proportion traceable to genetic differences or G. When there are no
significant C effects, c equals zero, which means that the family should re-
ceive about 10 times the attention paid to the individual.* Of course, with
the system of selection proposed, the individual will receive no attention at
all except from the standpoint of whether it lived or died, since this is the
only type of information which will be available for any given bird. The
point is that a positive value for the family weighting factor confirms what
common sense would suggest: that it is best to breed from the families with
highest production indexes.

Now suppose that because of housing conditions, environmental influences
have become more common to members of one family than to members of
different families. This could readily happen if, for instance, a severe attack
of coccidiosis or another disease affected some houses while others escaped.
If this type of variation accounts for nearly a quarter of the total variation
(c = .22) , we may substitute the value .22 for c in the above formula. f The
numerator will become zero, and so the whole expression will be zero. This,
of course, means that under the specified circumstances the phenotypic fam-
ily average is no guide whatsoever to evaluating the family genotype.

A more extreme case where c2 is higher than .22, say .5, will result in nega-
tive values for the family average weighting factor. In other words, when
environment is deliberately made more uniform for members of a family
than it is for the flock as a whole, it might mean that members of the poorer
families are to be preferred to members of better families as parents of the
next generation!

* This is the same result that we get when n = 50, r- .25, and h2 = .05 in the previous
weighting formula.

t Actually under these circumstances h2 itself will be reduced below the given value
of .05. But this a rough example and the difference does not matter.

POULTRY BREEDING 21

This sounds completely unreasonable. Yet we can give an example in
actual field practice. Suppose a purchaser of chicks has a choice of buying
them from one of two flocks. One of them has not been exposed to lympho-
matosis and the incidence of the disease in it is therefore zero. Another flock
has had a history of lymphomatosis so that some selection for resistance to
the disease has been practiced. It may still have an incidence of 10 per cent,
and thus phenotypically show poorer performance than the first flock. Few
breeders who have considered the matter will deny that the prospective pur-
chaser should be advised to buy his chicks from the second flock, especially
if the purchaser himself knows that his premises carry the infective agent.
Yet this advice represents the same sort of paradox that we have just de-
scribed, since each of the two flocks may be viewed as representing a family
of birds.

How often such a situation may actually come up in practice is not known,
but obviously the breeder must be on guard against selection methods which
may lead to such difficulties.

Selection Criteria

The second problem of selection procedure is the type of measure-
ment to be applied to the birds under selection. As a rule the poultry breeder
tries to use as a criterion of selection the very trait which he is anxious to
improve. If he is interested in raising the average body size of his birds at
the age of 12 weeks, he will use as his selection criterion body weight ob-
tained at that age. But sometimes he may find it more economical to use,
instead of a direct measurement, a related one.

For instance, a breeder attempting to raise the average annual production
may, instead of trapnesting his birds every day in the year (this is the direct
measurement of the trait under selection), trapnest them only five days a
week. His objective is to improve the seven-day- a- week production, yet the
indirect measure he uses is so closely correlated with the direct one that he
can afford to sacrifice some accuracy of measurement in order to cut down
his expenses. Indirect selection for egg size is even more striking. It can be
shown that selection on the basis of egg weight in the first November of life
is as efficient as selection on the basis of spring egg weight, for which im-
provement is sought.

The extent to which an indirect measure is useful depends on how close is
its correlation to the trait under selection. It is reasonably safe to say that
five-day-a-week records are closely enough correlated with full annual rec-
ords to make their use worth while. But can the five days a week be cut down
to four? Three? Perhaps one? On these points we have no complete answer
as yet, but it is quite clear that the hatcheryman or breeder who carries the

22 PRINCIPLES OF COMMERCIAL

process of reduction to the end by eliminating trapnesting altogether is on
very shaky ground. The indirect criteria which he can use, like breed type,
conformation, head points and so forth are not correlated strongly enough,
if at all, with production records to enable him to carry out much improve-
ment in the production record.

This point needs no emphasis for the specialized breeder. The multiplier
of improved stock or the hatcheryman, however, often labors under the idea
that physical selection for improved egg production can be effective. The
fact is that so far no genetic correlations between any specific body measure-
ment and production records have been discovered. It is true that periodic
examination of birds can lead to an estimate of their productive capacity.
The identification of a laying state in a bird at an early age can be interpreted
as evidence for early sexual maturity ; lack of neck molt in the winter months
as a reasonable sign of the lack of tendency to long winter pausing ; and so on.
Yet even careful periodic examination of birds will not lead to any great
accuracy of discrimination of various degrees of genetic merit, and in males
it is virtually of no value. At best, breeding policies based on such extremely
indirect criteria of selection may maintain stock quality, but have no powers
of improving it beyond the minimum economic standards of farm production.

Part versus Full Production Records

One kind of indirect selection which can be of utmost value in
breeding for improved egg production is the use of part-year records as a
measure of the full annual production index. The part-production index from
beginning of lay (in spring-hatched birds) to January 1 bears a high genetic
correlation with the full-production index. It is true that its accuracy in de-
termining the genotype for the production index is only three quarters of the
direct measurement itself. But there are many compensations. One of these
is the tremendous saving of labor costs which can be made by suspending
trapnesting of all but the birds selected for breeding after January 1. Another
is the fact that the indirect measurement becomes available nearly a year
earlier than the direct one. Thus birds selected on the basis of their produc-
tion index to January 1 may be used in breeding when they are one year old,
whereas birds selected on the basis of their full record cannot be placed in
the breeding pen until a year later. If breeding progress is measured in terms
of years rather than in terms of generations, this procedure would double the
rate of improvement were accuracy of selection preserved. But even when
accuracy is cut by a quarter, the net advantage is obviously on the side of the
indirect or part measurement.

The above example illustrates some of the problems of choosing between
sib and progeny testing. It so happens that the progeny test in the limit (when

POULTRYBREEDING 23

the number of offspring is infinite) is a more efficient measure of the geno-
type of a bird under test than the sib test. But the progeny test takes a longer
time to complete, while information on contemporaneous sibs becomes avail-
able as early as information on the bird itself. The data on the progeny test
cannot be used until the bird has been selected first on some other basis,
mated, and the offspring subjected to test. This is why the efficiency of gains
per year is greater with the sib test than it is with the progeny test.

This does not mean that progeny testing has no uses. The most efficient
system of improving the production index has been found to be one in which
a small amount of progeny testing is used. Thus, when the progeny test is
itself based on part records, it may be profitable to include about 10-15 per
cent of progeny-tested birds in the breeding flock. The precise proportion
matters little. In can be zero or 25 per cent if the occasion calls for it. But
higher figures will usually tend to reduce the rate of annual improvement in
most traits in which a poultryman is interested. The breeder's best judgment
will determine the value of the progeny test to him. It is not essential where
individual and sib selection are combined. It can be taken advantage of when
genotypes of exceptional merit are identified. The key to the success of a
selection scheme is in flexibility of this sort.

Selection for Several Traits

So far, we have tried to find a selection criterion for the breeder
who wants to improve a single trait, or an aggregate of traits which can be
expressed by a single measurement. As every poultryman knows, few if any
breeders can in practice limit their objective to a small number of solitary
characters. A breeder of birds for egg production must consider not only the
number of eggs, but also viability ("livability") , egg size, breed type (if his
customers demand it) , and a variety of other factors. A breeder of birds for
meat production has to include rate of growth, conformation, rate of feather-
ing and other traits in his overall improvement goal. Furthermore, both
breeders must consider characters which may be of no interest to their cus-
tomers, but which determine the efficiency of their own operations, like
fertility and hatchability.

These many objectives pose a problem as to the best means of combining
them in a breeding program. Let us say at once that the greater the number
of traits selected for, the lower the intensity of selection can be for each trait.
Suppose that in order to maintain his flock at constant size, the breeder must
use 10 per cent of the females in his flock as parents of the next generation.
This is the same as saying that he expects to produce an average of ten females
from each dam. It does not matter whether this figure is an over- or under-
estimation. On the basis of whatever tests the breeder desires to use (in-

24 PRINCIPLES OF COMMERCIAL

dividual, sib, or progeny) he can select the top 10 per cent of his flock for
the character under selection, let us say egg number. However, if adequate
egg size also forms one of his breeding objectives, he will of course find that
the top birds for egg number will not be the top birds for egg size. Even if
we make the conservative assumption that egg size and egg number are inde-
pendent of each other, we can still expect only one out of each ten high-
record females selected to be in the top 10 per cent of the flock for egg size.
If the breeder wants to combine selection for both traits he will have to reduce
his selection standards for each.

In our example the best 10 per cent of the birds in a flock may average (on
the basis of survivors' records) 70 eggs above the flock mean. If, however,
the breeder also pays attention to egg size, he may find that the birds found
adequate for both traits will exceed the average of the flock by only 35 eggs
in the egg record. This means that the selection differential instead of being
70 eggs is now only half that. If the heritability of survivors' production is
taken at .3, the expected gain in the flock average in the first instance would
be 70 x .3 or 21 eggs. In the second, the gain would be reduced to 35 x .3 or
about 10% eggs. Every time another objective is added to the selection
program, it means further reductions in gain.

The amount of the reduction will depend among other things on the close-
ness of correlation between the desired traits. In our example we assumed
that egg number and egg size are independent characters. Actually, in some
flocks there is a negative correlation between them which means that the
reduction in gain would be even greater than the amount given.

This situation leads us to a fundamental rule for breeders: the smaller the
number of characters concerned in a selection program, the greater the ex-
pected gains in each. The importance of this rule cannot be overemphasized.
It is this which should make the breeder think twice before he includes char-
acters with no economic importance in his breeding program. Everyone will
agree that a flock of Single Comb White Leghorns, each member having five
well-defined comb points, no more and no less, looks more attractive than a
flock widely varying in this respect. But every bit of attention paid to such
characters will in the long run lower the rate and efficiency of improvement
in the economically valuable traits. We assume that most breeders are pri-
marily interested in these.

Less obviously irrelevant traits than comb points are other characters en-
tering breed standards and culling guides, such as plumage color, breed type,
and standard defects and disqualifications. So far as we know, they bear
little or no relationship to the productive qualities of birds. Yet the com-
mercial poultryman who is the breeder's direct customer often evaluates his
stock on their basis. Poultrymen with this viewpoint are the only reason why
breeders might be justified in paying any attention to extraneous non-

POULTRY BREEDING 25

productive characters. As soon as the commercial poultryman gets con-
vinced that it is the viability, egg or meat yield and quality and not breed
points which determine whether his balance sheet shows profit or loss, the
breeder will no longer be obliged to include uneconomic traits in his selec-
tion procedures. The example of the American Dairy Cattle Club, which
judges the merit of breeding stock entirely on the basis of economic traits,
might be worth looking into in the field of poultry.

Total Score Selection

In the meantime, even when considering economic characters
only, the breeder has a difficult task. Three different methods of proceeding
with selection for many traits can be used. They are: (1) selection for one
character at a time, (2) simultaneous selection for several characters on the
basis of separate standards for each, and (3) so-called total score selection
on the basis of a single criterion combining all the desired traits.

It can be shown that the third method is more efficient than the first two,
at least in theory. We will limit our discussion to it.

An ideal total score is a selection index incorporating information on the
individual and family performance on every trait to be considered. In such
an index, each measurement would be weighted in accordance with three
factors : ( 1 ) the economic contribution each character makes to the overall
worth of a bird, (2) the heritability of each character, and (3) the extent to
which each of the desired traits is correlated with the other traits under
selection.

It is clear why the first two factors must be taken into account. The more
important a character is economically, the more attention it should receive
in a selection index. The greater the heritability of a trait, the faster are the
gains which can be obtained from selection. The third factor is less simple
and calls for an illustration.

Suppose we are interested in improving the conformation and rate of
growth of a flock devoted to broiler production. Body weight at 12 weeks of
age, breast width and keel length may be our three desirable traits. Now it so
happens that a positive genetic correlation exists between body weight and
keel length. This means that if we selected for body weight alone, keel length
would also increase as a result of the common genetic control of both. In one
flock of New Hampshires it was found, for instance, that in selecting for body
weight every additional unit increase in body weight would automatically
lead to an increase of .8 units in keel length.* Similarly, in selecting on the

* The units referred to here are not pounds and inches, but rather standard deviations
(a statistical constant) . In the New Hampshire males in the flock discussed, one standard
deviation in 12-week body weight equals about eight ounces, while one standard deviation
in keel length is about one quarter of an inch.

26 PRINCIPLES OF COMMERCIAL

basis of keel length without attention to body weight, a gain in one unit
would result in an increase of .8 units in body weight. The correlation be-
tween the two traits is so high that if attention is paid to one, the other need
not be weighted too heavily in the selection index, because a good share of
the possible gain in it would be made automatically.

Much less convenient for the breeder are desirable traits between which
negative correlations exist. This is the case for keel length and breast
width. A negative correlation means that when selection for increase is prac-
ticed for one of them, the flock average for the other will decrease. No selec-
tion procedure can entirely overcome the conflict of negative genetic
correlations. The purpose of a selection index is to find the most profitable
balance between the pulls exercised in opposite directions by negatively
correlated characters under selection: so that while the gains in respect to
each will not be the greatest possible, the gain in total economic worth will be.

Selection Procedures

Making selection indexes is a complex job, and it is out of the
question for each individual breeder to undertake a special index for the
special conditions of his own flock. Even in experimental flocks the work is
so great that no complete index for any comprehensive breeding project has
as yet been made. However, certain clues and guides are beginning to be
available for several types of breeding projects.

As one instance, it was found that if the main object of selection was breast
width in the flock of New Hampshires mentioned, the fastest gains could be
expected when an index incorporating body weight (W), shank length (S) ,
keel length (K) and breast width {B) took these proportions:*

Index of breeding worth = W + 2/35 - 1/55 - l/5£.

This, however, represents a relatively simple situation where improvement in
breast width is wanted, perhaps at the expense of other valuable traits not
considered in the index (for instance fertility, which may or may not be
affected by changes in breast width) . Besides, this index is intended for use
under individual and not combination selection, though the heritability of
breast width is in the range where some attention to family averages may be
profitably given.

In general, it is premature at this time to recommend any particular selec-
tion index for commercial use. But the principle involved can definitely be
put to good use, particularly by breeders for egg production. This is because
the two main characters such breeders are vitally interested in are egg num-

* The measurements are in kilograms and centimeters. Other units, such as pounds
and inches, would call for different fractions in the index.

POULTRYBREEDING 27

ber and viability. The breeder can select for each trait separately or he can
base his selection procedure on the combination of both. The hen-housed
production average, or as we have called it, the production index, is not nec-
essarily the ideal combination where precise weighting is applied to each
component, since it is not weighted for the heritabilities of the two com-
ponents ; but it does represent an approach to the total score. As such, the
use of the production index is a more efficient plan of selecting birds than
separately considering viability and egg records of survivors.

This fact can stand emphasis. Some breeders have objected that selection
on the basis of the production index would lead to increased mortality and
loss of "stamina." This argument does not carry much weight when we realize
that it is the balance sheet which is important to the producer. The production
index gives the best approximation to it (except where the meat value of culls
contributes an important share of the poultryman's income) . It provides for
the most profitable balance between viability and egg number. Neither of the
traits will be permitted to drop below an economically sound level. This is
quite likely to happen when undue attention is paid to one or the other. The
mortality rise of the 1920s and 1930s (when individual selection for high
egg number was in vogue) may well be an example of such a case.

All of this, of course, refers to selection on the family or combination basis,
since the production index as a selection criterion can have but little meaning
when applied to individuals alone.

If we consider viability and egg number further, we see that each of them
is in itself an aggregate of still further characters. Thus, the mortality level
of a flock may represent the net combination of resistances and susceptibili-
ties to various diseases. The survivors' egg record is the result of different
potentialities, such as rate of sexual maturity, the tendency to pause in the
winter, the extent of the broody instinct, the ability to maintain a laying
state into the period of the normal annual molt, and the rate of laying. The
production index has the virtue of combining all these diverse factors into
one figure, to which selection can be applied.

One adjustment in the use of the production index may be suggested. It
so happens that when appropriate weighing for heritability and economic
contribution is made, early production seems to be more important than late
production. The average performance to January 1 can then be profitably
given extra attention beyond what is given to the total annual record. This
fact further supports the use of part records in selection, and incidentally
illustrates the wisdom of several earlier generations of poultry breeders who
based their selection procedures on winter records in preference to spring
ones.

As breeding objectives change, selection indexes are bound to change also.
At the moment, the only practicable application of the total score idea is to

28 PRINCIPLES OF COMMERCIAL

use the production index rather than survivors' records and viability sep-
arately, and, all other things being equal, to give preference to families and
birds with higher production indexes in the first rather than in the last half
of the laying year.

This recommendation must be qualified too. For instance, a survey of re-
sults from laying tests indicates that the stock of some breeders may perform
adequately at the beginning of a test but cannot maintain the pace, and go to
pieces in the summer. Possibly these birds are deficient in genes responsible
for high persistency of production. Under these circumstances more attention
may be given to the performance in the summer and second fall of laying
than we have suggested. In general, the type of total score, or approximation
to it, which a breeder uses as a criterion of selection must be dictated by his
specific needs. No single index will apply for all flocks.

Selection and Culling

There is one more point about selection: the culling of flocks
under test. The no-culling provision of many official improvement schemes
has always been a thorn to breeders. In the United States and Canada com-
promise solutions have at times been reached, when breeders were permitted
to cull birds or families of birds in the early stages of the laying year. In
Great Britain, the no-culling regulation of a recently inaugurated improve-
ment scheme raised a storm of protest from breeders who did not want
unthrifty birds in their flocks, thinking that maintaining them in the laying
houses would make the scheme economically unworkable.

The no-culling provision will indeed raise the cost of the breeding pro-
gram. But there is good indication that the increased accuracy of identifying
the desirable genotypes from unculled flocks may more than offset the added
expense. A precise answer to this problem is not yet available. Perhaps we
may eventually reach a compromise between the two extreme viewpoints.
Available evidence in the meantime favors the no-culling scheme.

Let us take a possible example. Suppose we have two families of five birds
each with unculled production records as follows (a d indicating that that
bird died in the course of the laying year) :

Family A

Family B

250

300

240

250

230

200

180d

70

100

40d

Average: 200 eggs 172 eggs

POULTRYBREEDING 29

On the basis of an unculled population, the average of family A is clearly
superior to family B. Now suppose that the bottom two birds in each of the
families could be identified beforehand by an expert culler as inferior, if,
say, they showed unthrifty appearance due to the onset of disease. When such
birds are culled and their production records eliminated, the family averages
of the two groups will read in the breeder's summary-
Family A : Average 240 eggs 2 culls
Family B : Average 250 eggs 2 culls

—or, if zero production is assigned to the culls-
Family A : Average 144 eggs
Family B: Average 150 eggs.

To look at such a summary, family B is superior to family A. But this is
not the case. It is, of course, to some extent a speculative matter which of
these two particular families actually has the better genotype for hen-housed
production. But on the average, among a large number of A and B families,
greater genetic improvement can be expected in breeding from the A families,
not from the B families. In this way we can see that culling procedures may
obscure true genetic worth. Culling procedures also add an extra character
(percentage culled) to the production index, and this reduces the efficiency
of selection for the index.

There are technical difficulties in applying a no-culling rule. Some date
or age must be selected as the base point from which the production index is
computed. From the breeder's standpoint the date of hatch may not be the
best choice, since much pre-laying mortality is accidental in nature (that is,
the heritability of chick viability in some flocks seems to be low) . Use of the
original number of pullets hatched as a base for the family average is there-
fore not recommended. The choice of any other starting point in the life cycle
of a flock is arbitrary, and only a compromise solution of the problem is
possible. In some flocks six weeks of age, and in others five months, are the
points selected. Whether these are the best possible choices is difficult to say.
It is probably wisest to use a date or age not later than the beginning of lay
of the earliest maturing bird in the flock. From the commercial producer's
standpoint, on the other hand, comparison between different flocks may best
be made on an overall basis: from the day-old stage to the final disposal of
the birds.

What we are saying about culling applies, of course, to flocks on which
data are being gathered for future use in selection procedures. For other
purposes culling may be economically desirable. For instance, in the breed-
ing scheme described on page 22, there is little reason why pullets should not
be culled after January 1. Trapnest records beyond that date will not be

30 PRINCIPLES OF COMMERCIAL

taken. Culling in this case does not destroy the sample on which decisions
about the genotypic merits of a bird or a family are to be made. It is true
that the annual production index by which the breeder judges his progress
in genetic improvement cannot be considered an accurate measure under
such circumstances. But the breeder must choose here between perfect ac-
curacy and lower costs of operation.

POULTRYBREEDING 31

Mating

Mating Systems

After the breeder has selected from his flock the birds which are
to become the parents of the next generation, he has to decide what combina-
tions he is going to use in mating them. There are several systems he is free
to follow.

First, he can mate birds at random, and assign the chosen females to the
pens headed by different males simply by chance.

Second, he can mate females he considers to be best to his best males, and
the poorest of the selected birds to the poorest. In other words he can mate
like with like. This system is known as somatic assortative mating.

Third, he can use the opposite of the previous process and mate the best
birds of one sex to the poorest of the other. This is somatic disassortative
mating.

Fourth, he can mate together closely related individuals, that is, he can
use inbreeding. The fifth method, genetic disassortative mating, is the op-
posite of inbreeding. In its limit it cannot be used within ordinary isolates
and calls for crossbreeding. In a closed flock of a single breed, genetic dis-
assortative mating reduces itself to random mating with restrictions. This
will become clear as we go on.

All of these systems can be used in individually pedigreed, pen-pedigreed,
or flock matings. Our main interest is with individually pedigreed and to
some extent pen-pedigreed matings. But the genetic consequences of the first
kind in the main apply to the others as well and we shall not consider them
separately.

Each of the five systems can be varied and modified. For example, re-
stricted random mating may be based on the chance combination of males
and females with a proviso that no full sisters be placed in a breeding pen
headed by their brother. This restriction can be extended to half sisters,
cousins, or whatever other degree of relationship the breeder wishes to set
as a limit between mates, without unduly affecting the essential randomness
of his breeding pens. A practical method of restricted random mating which
has been used in some flocks is based on pedigrees to two generations (ending
with grandparents). In this scheme only birds having no common grand-
father or grandmother are to be mated together.

Somatic assortative mating can have a variation where the best are mated
to the best not on the basis of the overall criterion of selection, but from a
consideration of component factors. For instance, in the case of breeding

32 PRINCIPLES OF COMMERCIAL

for improvement in the production index, early maturing birds may be mated
together in one pen, non-pausing birds in another, and so forth without re-
gard to their production index (after they have been selected on its basis).

Likewise, the somatic disassortative method can take the form of compen-
satory mating, in which the defects and excellences of the two mates are
balanced. Thus, a bird possessing good egg size but showing winter pause
could be placed in a pen headed by a male judged to come from a small-egg-
size but non-pausing family.

Inbreeding can be of varying degree or intensity. Full brothers and sisters
can be mated together, or half brothers may be mated to half sisters. Other
variations on the same theme may be adopted in a systematic fashion, or less
formal schemes of mating relatives together may be used. What is known as
linebreeding is also a form of inbreeding, in which an attempt is made to
increase the number of times a given bird appears in a pedigree by the mat-
ing together of its descendants.

We need not consider genetic disassortative mating in any detail. When
it becomes wider than restricted random mating it usually stops being a
method of genetic improvement. Our remarks about migration (page 12)
apply here as well. Crossbreeding as a system may produce superior stock in
the first generation of a cross, a fact recognized and used by producers of
birds for meat. But the improvement from generation to generation must
depend on the selection and mating schemes followed within each of the
breeds. Only when an attempt is made to synthesize a new variety from crosses
between two or more previously existing ones is genetic improvement in-
volved. Such attempts are the concern of a very small group of adventurous
breeders. They are subject to special requirements and conditions in each
instance. We can give no general instructions for the best methods to cover
all cases, and by and large the whole problem lies outside our subject.

We will therefore compare the first four mating systems only. None of
these is perfect for all purposes. There are advantages and disadvantages to
each, and most likely the best procedure will be a flexible approach, using
various combinations between them. This statement is not likely to satisfy
breeders who want practical instructions. On this account we will briefly
show the merits and dangers of each system, aiming at a practical recom-
mendation not for the best method, but for the most workable one at the
present time.

Inbreeding

Inbreeding for years has had and still has many enthusiastic sup-
porters, who may not always see the consequences of their recommendations.
To start with, intense inbreeding is the only one of the mating systems which

POULTRYBREEDING 33

has any powers of fixing traits that depend on many gene pairs for their ex-
pression. But we have already pointed out that when characters of relatively
low heritability are dealt with, fixation in the sense of achieving complete uni-
formity is not possible even if desirable (which also is questionable). Fur-
ther, we have noted that under inbreeding non-additive genetic effects enter
the picture. Under random mating the knowledge of the selection differential
and of heritability permits prediction, within the limits of sampling error, of
the average level of a trait under selection in the next generation. But when
non-additive genetic action enters the picture, this is not possible.

Particularly distressing is the fact that non-additive effects in inbred stock
lower the performance in productive characters. Numerous experiments con-
ducted with poultry have shown that intense inbreeding with or without
selection cannot raise the productive level of economically valuable traits.
Some supporters of intensive inbreeding as a method of fixing characters
may dispute this. We suggest that they look at the recent literature on genetic
theory and experimental inbreeding in species other than poultry, literature
which leaves very little room for doubt on the matter. Or let them search
their own experience and find a single instance where inbreeding has been
able to fix egg production, viability, or hatchability at a high level of per-
formance in a flock or strain.

Why, then, do we need to consider inbreeding at all? For two reasons. One
concerns forms of inbreeding less intense than those involving full brother x
sister or half-brother x sister. The second is the production of inbred lines
for the purpose of subsequent crossing.

Moderate Inbreeding. At present it seems that the degree of inbreeding
to be gotten by restricted random mating (page 31) is not necessarily detri-
mental to the performance of the birds. There may be some question whether
more homozygous birds are produced by such a method than by ordinary
random mating. The fact is that so far as pedigree relationships are con-
cerned, some inbreeding does occur under this scheme. In the breeder's hands
it may be reasonably useful for special purposes.

For instance, a breeder may find that without previous warning an ana-
tomical defect has appeared in his flock. This may be a character easy to
identify, like crooked toes in day-old chicks, or it may be a more serious
condition which kills the embryos before hatching (a lethal gene or combina-
tion of genes) . In such cases the breeder may try to eliminate the genes
responsible by applying intense selection to his flock against the undesirable
trait. But if the trait is recessive, it may be that the culprit genes have spread
widely throughout the population before being noticed. The breeder's prob-
lem may then be to fix the desirable counterparts of the undesirable genes.
This is of course what is meant by selection against an undesirable trait. But
the point is that fixation as noted may call for some inbreeding if the unde-

34 PRINCIPLES OF COMMERCIAL

sirable gene is present in the majority of the flock. A judicious combination
of selection with mild inbreeding may have to be brought into action under
these circumstances.

Inbreeding may serve a further purpose as an incidental process in selec-
tion. A breeder who has identified a superior genetic combination by what-
ever means (sib or progeny testing) should naturally try to take advantage
of its existence in the flock. He may do this by favoring in his selection the
individual or family carrying this genotype. If he practices family selection
he will find that the number of different ancestors present in the pedigree of
his flock will be materially reduced from the number found under random
mating. In fact, selection itself will result in a certain amount of inbreeding.

There is a dissipation of genetic worth in every generation by which birds
are removed from the desirable ancestor. It is caused by the tendency of
superior genotypes to produce offspring which will regress to the flock aver-
age, owing to the fact that the superior individual contributes only half of
the inheritance received by its immediate descendants. Nevertheless, the
only practical method of partial conservation of desirable genotypes in the
flock lies in the type of selection, which may automatically involve linebreed-
ing in one form or another.

Hybrid Vigor. The second broad purpose of inbreeding, producing
crosses between inbred lines, is part of a vast and complex subject which
would take a book to treat. Most poultrymen are now familiar with the ideas
behind this scheme, and yet our knowledge of either the proper techniques
of putting it into operation or of its efficiency as a method of poultry improve-
ment is incomplete.

The basic fact is that when inbred lines are crossed, the first generation
of offspring show what is called hybrid vigor or heterosis for some traits. The
term heterosis refers to the condition when the crosses are superior to their
parents in performance. Usually, heterotic behavior is shown in traits such
as hatchability, viability and growth rate, but not in others, like egg size. Of
course, one of the reasons why the hybrids are superior to their parents is
that the parents, being inbred, do not themselves excel in these particular
traits. The question is whether the hybrids are indeed better than the superior
strains of birds produced by other selection and breeding methods.

This question has not yet been adequately answered for chickens. We
know definitely that in some plants, like corn, the method of crossing inbred
lines is an efficient technique for raising yield. But in the light of present in-
formation it is too early to transfer this conclusion from corn to chickens,
and it will probably be some years before we have a clear solution. In the
meantime, even if inbred crossing is as usable and efficient as many believe,
it will call for a vast expense of money on the part of the breeder, for spe-
cialized direction by highly trained geneticists, and for other requirements

POULTRYBREEDING 35

which most present-day breeding establishments would find it hard to meet.
Wholesale adoption of it would cause revolutionary changes in the structure
of the whole poultry industry.

It is well known that the recent extensive commercial exploitation of the
techniques of hybrid chick production has already made heavy inroads into
the hatchery business. This is probably all to the good if it leads to the re-
placement of unimproved flocks by better stock no matter how produced.
But whether hybrids or crossbreds will completely drive out efficient closed-
flock breeding programs is still very much a question. Likewise, we still do
not know how promising are several other recently proposed methods, such
as the so-called "reciprocal recurrent selection." The answers, of course, are
to be sought in impartially conducted research, such as is now being under-
taken by many agencies, and not in high-pressure commercial promotion.

Other Mating Systems

The remaining three mating systems in their various forms are of
more immediate interest to breeders now engaged in poultry improvement.
The genetic consequences of random, somatic assortative and somatic dis-
assortative mating schemes are somewhat different. The first promotes genetic
variability, but probably leads to a slower approach to the extreme top levels
of performance for the selected characters. The second permits a more rapid
extension of the range of performance. The best birds in each generation may
be better than the best of the preceding one. The rise in average performance
does not, however, keep pace with the extension of the range. The third
method, somatic disassortative mating, works in the opposite direction : the
upper limit is not extended at a rapid rate but the lower limit of performance
may be raised.

The extent of the differences between the three methods depends on a series
of factors, such as heritability, the number of gene pairs involved in the
inheritance of a trait, and the degree to which heritable resemblances in
performance between the potential parents can be recognized.

There are special uses for each method. For instance, in the progeny test-
ing of males, which involves comparisons between the performance of several
sires each mated to a group of dams, random mating provides fairer estimates
than the other types. When the dams are randomized (when mates for each
sire are selected by chance) , it is more likely that genetic differences between
them will cancel out, so that the average of each sire's offspring may be used
as an estimate of his genotype with greater assurance of accuracy than when
other mating methods are involved.

Somatic assortative mating should probably be practiced in at least part
of the flock if extension of range of performance is wanted. Somatic dis-

36 PRINCIPLES OF COMMERCIAL

assortative mating may, on the other hand, be useful when the middle ex-
pression of a trait is preferred to both extremes. For instance, since neither
too small nor too large eggs are commercially desirable, it may be worth while
to keep a happy medium by mating large-egg females to males with poten-
tialities for small eggs, and vice versa.

The choice of a specific mating scheme must depend on the particular ob-
jectives of the breeder. As a rule, it is best not to keep to any formal scheme
too rigidly, but to approach the problem in a flexible way. Perhaps in general
a combination may be recommended, which will include a random mating
system with restrictions so that no female with a grandparent common to the
male be placed in his pen, somatic assortative mating for the upper half of
the selected flock, somatic disassortative mating for the lower half, and
occasional excursions into inbreeding for special purposes.

POULTRYBREEDING 37

Practical Applications

The Heritability of Economic Traits

We have so far considered the fundamental basis of inheritance,
the principles of selection, and the forms of mating systems. These matters
in a general way cover the fundamentals of the breeder's trade. A more con-
crete application of these principles to actual breeding practice calls for
further consideration of the most fundamental of genetic constants— the de-
gree of heritability.

Our information about its size for different economic characters is quite
incomplete. It should be realized that the genetic approach to breeding prob-
lems which we are talking about is a relatively new one. Less than a handful
of experiment stations and institutions have so far contributed to it. Precise
information is lacking and in our discussion of actual heritability values we
must limit ourselves to estimates, some of which are fairly sure, while others
are merely first approximations.

In general, heritability values for different traits of economic use can be
classified simply as high, middle, and low. Characters classified as high
readily respond to mass selection. Middle characters require a combination
of mass and family selection. Characters with low heritability also depend
on a combination method of selection, but sometimes it may be found
simpler to use family selection exclusively.

Traits with High and Middle Heritability. The breeder interested in
improving meat quality is by and large dealing with traits with a high h2.
Both body weight (at various ages) and rate of growth have heritabilities
of from .40 to .50. It would seem then that family selection for these char-
acters is no more efficient than individual selection. Note, however, that body
weight in the early stages of life is to some extent determined by the size of
the egg from which a given chick emerges. This will be recognized as a C
effect, since full sibs are likely to arise from eggs resembling each other in
size. The magnitude of the C factor is gradually reduced as the birds grow
older, until only a small C residue is present in mature body weight.

Conformation comprises traits with either high or middle heritability. Thus
shank length has an h2 of about .5, keel length roughly .3, and breast width
.2. Shank length, it must be understood, bears a strong genetic correlation to
body weight. This means that efforts to increase body weight and reduce
shank length within a flock may not only be very difficult, but even in some
cases impossible. Selection of equal intensity for both characters is likely to
result in a standstill performance for both, and even the best selection index

38 PRINCIPLES OF COMMERCIAL

may not resolve this difficulty. Whether or not there are strains in which the
genetic correlation between body weight and shank length is low or non-
existent is not known. Only in such groups could the combination of large
body size and short shanks be attained.

The negative genetic correlation between keel length and breast width
is not high. We have already noted that a selection index to combine the two
traits has been constructed for one flock of New Hampshires. The particular
point of interest about these two characters is that their heritabilities are
low enough to make combined selection preferable to individual selection.
This is especially true of breast width.

Another high-heritability trait is egg weight. Its value is about .6, which
makes possible improvement by individual selection without regard to family
averages. About the heritability of other egg characters we know less. In
some cases these seem subject to partial control by non-genetic forces relating
to the dam. The best established example is shell thickness, but there are also
indications for albumen quality. The generally high heritability of egg char-
acters accounts for their rapid response to selection.

We have no exact information about other high or middle h2 traits, with
one exception: egg production of survivors. For this character it is almost
certain that h2 is near .3. This figure applies equally well to the annual record
and to part-year records (when only birds surviving the first laying year are
considered). It is likely that rate of feathering in heavy breeds where there
is a mixture of slow- and rapid-feathering types in a flock has relatively high
heritability. Likewise, the h2 of sexual maturity is in the middle range, some-
where about .2 to .3.

Traits with Low Heritability. Of characters with low h2, particularly
important are the two for which our information is reasonably precise : the
production index and viability. Both are below .10. The most acceptable
present figure for the production index is about .05 and for viability about
.08. Both these characters (viability is, of course, a component of the produc-
tion index or hen -housed average) are extremely important to commercial
poultry breeders, whether primarily interested in eggs or in meat. We have
seen that mass selection applied to them is not very efficient. The superior
method is probably combination selection, with emphasis on family averages
but with some attention paid to individual performance. But family selection
alone may be profitable sometimes, particularly if precautions are taken
about the C factor.

We will say no more about the production index except to repeat that the
breeder, whose first concern is not with meat qualities but with eggs, should
rely on it as a selection criterion. The formulas given on page 19 can be
used to determine what weight is to be given to the family record as compared
to the individual one.

POULTRYBREEDING 39

About first-year viability more must be said. In general, a breeder who
uses hens two years old or older for reproduction permits nature herself to
exercise individual selection. There are only two possible phenotypes a bird
can have for viability : either she dies during the first laying year or she sur-
vives. The birds which die are not available for breeding purposes in their
second year. Family selection is therefore the only tool the breeder can use
under the circumstances. If his flock mortality is very high, he may find that
there is no room left for any selection on his part— he may have to breed from
all of the survivors if the flock is to maintain its size. But in general the
breeder will have some opportunity of rejecting certain families. The most
useful criterion for this purpose is the percentage of survival.

We may note here that when full-sister families differ in size, an equal
percentage of mortality does not indicate equal genetic merit. Thus if one
out of a group of three birds dies, the survival rate of 67 per cent is not
genetically equivalent to the same ratio of living to dead birds in a family of
nine pullets. A "conversion formula" to equalize the information for families
of different sizes may be applied. It is given as

l+(n-l)rhs

which contains symbols we have already used. For the case of full-sister
families (r = .5) and viability (h2 - .08) , the expression is

25n
24 + n'

where n is the number of sisters in a family. Thus if we tried to decide which

of the two families given above is the better, when both come from a flock

with an average viability of 40 per cent, we would multiply their respective

superiority over the flock as a whole (.67 - .40, or .27 in both cases) by

25 x 3 25 x 9

— — - in one case, and by — — — in the other. Obviously the larger the family

the more accurate is the information on its genetic merit available from its
average performance. Preference should be given in selection to larger fami-
lies over smaller families with the same mortality incidence when it is lower
than the average.

Often there are situations where particular attention must be paid to a
specific given disease rather than to mortality from all causes. An example is
the case of lymphomatosis, which at one time caused such severe losses as to
have forced its singling out as a specific breeding objective. The heritability
of resistance to this disease is about .05 in the flocks in which it was studied.
Resistance to other diseases may have lower h2 values. For instance, the
proneness to develop unspecific disturbances of the reproductive system has

40 PRINCIPLES OF COMMERCIAL

such a low heritability (.02) that little progress in the breeding control of
this defect may be expected from any but very intensive selection.

Two problems connected with breeding for high viability are of utmost
importance but are still not solved. The first is the problem of the correlation
between viability and egg number. It seems well established that the pheno-
typic correlation between the two is positive, that is to say that high egg rec-
ords are more likely to be obtained where mortality is low. There is a pos-
sibility, however, that the genetic correlation between viability (particularly
that of embryos, that is hatchability) and production may be negative in
flocks previously subjected to intensive selection for a high production index,
so that intensive selection for improving one of these traits may decrease the
average performance of the flock in the other. This, if true, is another reason
why total-score selection as exemplified by some form of the production index
is to be preferred to selection on the basis of each trait alone.

The second problem for viability is the correlation between resistances
to different causes of death. There is some conflict of opinion on this point.
In one of the flocks studied, there was, for instance, a reasonably high corre-
lation between resistance to lymphomatosis and resistance to other diseases.
In another flock the correlation was rather low. If the first situation is more
typical it may be argued that general resistance or vigor genes are involved
in the inheritance of the production index. This conclusion may not be true
if the low correlation is found to hold. Uncertainty about the question shows
how our knowledge of the whole subject is still in its infancy. Much intensive
effort and experimentation will have to come before we are able to provide
satisfactory answers to all the problems faced by breeders.

Breeding from Pullets

The breeder who wants to use pullets for reproduction is faced
with a special problem. In selecting the mothers of each generation when the
candidates for motherhood are themselves less than a year old, he runs the
risk of choosing individuals which, after producing offspring but before the
end of the first laying year, will themselves die. This has often deterred
breeders from putting pullets into breeding pens. Such fears are unfounded
as long as family performance is the primary criterion of selection.

Similar fears do not seem to have disturbed most breeders who use
cockerels in their improvement plans. Yet the risk of a male dying before
reaching his second birthday is not necessarily smaller than the risk for
females. A more decisive point is that the heritability of viability is only
.08. This means that 92 per cent of variation in the fate of the birds in their
first year of life is non-genetic in nature. So it matters little if the individual
bird herself lives or dies. What does matter is whether she is a member of a

POULTRYBREEDING 41

family with or without high genetic resistance to death-inducing causes. By
the time pullet selection for breeding takes place, a reasonably good estimate
of family viability is available. There are exceptions in which certain fami-
lies may exhibit a characteristically late date of death in the first laying
year. But as a rule the genetic correlation between viability to one year and
viability to 18 months is high. Because of the gains provided by the shorter
period between generations, pullet breeding despite these objections still
seems to be a sound method.

To the cautious breeder we might suggest taking out some insurance by
raising more chicks than he needs, and discarding or selling the surplus be-
fore placing his pullets in the laying house on the basis of the mother's fate
at that time. His discards would be the chicks from dams which died between
the hatching season and the moving of their offspring to the laying house. In
this way the breeder will have two advantages : he can capitalize on the gains
produced in the rate of improvement by using pullets, and yet he need not
include the offspring of birds which die before completing their first laying
year in his flock. It is only fair to point out, however, that the selection
differential in such a scheme will be below the maximum possible.

Other Details of Breeding Plans

The heritability of other productive traits is not too well known.
Fertility, hatchability, persistency, broodiness and winter pause do not seem
at best to have high enough heritabilities to make mass selection efficient. We
may hope that in the not too distant future more precise information will be
available. At present we suggest that breeders proceed on the assumption
that a combination of family with individual selection provides the best op-
portunities of improving these traits, even if the exact weighting of the two
selection bases cannot be provided.

Above all, the breeder should not rely on any rigid idea of selection and
mating procedures. His greatest efficiency of operation will come when he
can adapt himself to the conditions of the moment. Breeding objectives may
change, the price structure of the poultry industry may be modified over-
night by changes in market requirements or by government regulation. Fami-
lies or individuals of exceptional merit may be discovered in a flock. Only
when the breeder's system is flexible can he on the one hand protect himself
against sudden shifts, and on the other, take full advantage of opportunities
which may come up.

Specific details about factors other than heritability could possibly be
brought into our discussion. The actual variety of details that each breeder
faces is, however, too large for treatment. Often such details call for indi-
vidual decisions which cannot be made in a blanket fashion. For instance,

42 PRINCIPLES OF COMMERCIAL

there is no best system of record keeping. Each system must answer each
man's needs and facilities. It is likewise impossible to say outright whether,
as an example, improvement is egg size should be pursued in preference to
improvement in persistency, or vice versa. Questions of this sort depend on
the particular conditions in each flock and each chick-marketing area. It is
difficult to foresee a time when such details can be given rule-of-thumb treat-
ment. The important thing is to understand the basic principles. Once the
breeder has understanding, he will be a more competent judge of the best
procedure to use than any recognized authority who lack information on the
flock in question.

We now turn to some general remarks on the immediate application of our
principles to the commercial poultryman's problem of finding good stock.
For reasons already given we shall not concern ourselves with chicks pro-
duced by crossing inbred lines.

POULTRYBREEDING 43

Zhe Commercial Poultry man

Breeding Methods and the Commercial
Poultryman

The commercial poultryman who operates either a specialized
egg-producing farm or a broiler plant must choose, on the basis of many
attractively presented claims, the source of supply of stock. The general
farmer of course has the same problem, but his investment in poultry may
be modest, his requirements may be met with cheaper chicks, and his annual
income depends only to a small extent on the Tightness of his choice of stock.
To the commercial operator the question is much more serious, for he cannot
withstand a succession of serious mistakes and still make his living from
poultry.

It is surprising, then, that among the welter of information available to
the poultryman on nearly every management problem he may face, there is
so little on the vital subject of the choice of supplier. There have been many
recommendations made in the past on the choice of breed, but the choice
within the breed is at least as important and little has been done about it. The
reasoning behind this deficiency seems to be that there is nothing the com-
mercial operator can do about the genetic potentialities of his flock. He may
build new houses, change diets, vaccinate his birds, but the inheritance of
his birds is fixed before the chicks ever reach him. This is of course the very
reason why he should have information on their breeding. Only when he is
in a position to discriminate between the claims made for the different sources
of supply, can he form any kind of beforehand judgment as to where to buy
his stock.

In general there are three types of suppliers on the market : the hatchery,
the hatchery-breeder, and the breeder. There are complex degrees and grada-
tions within this simple classification. It is not uncommon for chick producers
to call themselves breeders. Should a poultryman simply accept such self-
classification? Obviously not, since it is not what the chick producer calls
himself, but the breeding policy he follows which classifies him.

In the final analysis the only test which poultry have to pass is that of
making money for their owners. No amount of previous information will
settle this question. Hence, when a poultryman wants to change his supplier,
the best thing he can do is to conduct an experiment. He should buy stock
simultaneously (not at different times of the year) from both the old and the
new sources, provide the two groups with as uniform environment and care
as possible, and draw his conclusions from the comparative cost accounts.

44 PRINCIPLES OF COMMERCIAL

Such experiments even on a small scale may be desirable throughout the
poultryman's career, because a single source cannot supply chicks of uniform
quality year after year, and trends of improvement or deterioration will
change the relative values of different strains in the course of several years.

But the commercial poultryman cannot afford to jump blindly at every
offer of stock. He must discriminate between the sources worth trying and
those obviously unsuitable, or his experimenting will put him out of busi-
ness. How can he do this?

The sources of information open to the poultryman are reports of other
customers, advertising descriptions of selection and mating systems, contest
results, and data from government improvement schemes. To advise on the
value of the first of these would take the combined services of a psychologist
and an economist, and not those of a geneticist. To some extent this is also true
of the second source, but here the geneticist may be of some help. The evalu-
ation of the remaining two sources depends on an understanding of the
specific details and conditions under which the data were compiled.

Evaluation of Advertising

The advertising of the breeders and the hatcherymen may be
based on a combination of the other listed sources of information. The com-
mercial poultryman should pay particular attention to the claims made for
the results obtained on the breeder's own premises, and to the description
of the selection methods used. Are the production, body-weight, and egg-size
averages based on unculled populations or on selected samples? Do the
chicks to be supplied originate from the group of birds described, or from
some of their remote descendants with selection suspended in between? What
does the breeder precisely mean by family selection and progeny testing?
Do they correspond to the systems we have described, or do they simply
indicate that the breeder collects the information necessary for such methods
without actually using them? Questions of this type are direct, and no
breeder can in good faith refuse to answer them. There are no secret meth-
ods for improving stock. The available techniques are open to all. The
breeder who claims to be in possession of methods he does not wish to reveal
very likely has something to conceal.

Many advertisements are based on the virtues of a foundation dam or sire.
Outside the fact that pedigrees based on individual performance records are
of little worth in the case of egg production (immediate ancestral per-
formance is much more useful, of course, for high h2 traits), it takes some
stretching to believe that chicks of a commercial grade can be supplied in
any quantity without a considerable dilution of the supposedly valuable

POULTRYBREEDING 45

inheritance of foundation animals. Pedigrees citing family averages are of
somewhat greater value, but even they have to be taken cautiously.

In general, an unculled production average for the supplier's breeding
flock is as good an index of merit as the prospective buyers of chicks for egg
production can have. Of course, information on factors other than the
production index, such as egg size and quality, should be obtained. Not all
breeders may be able to supply data on, let us say, the blood-spotting tendency
in their flocks, but it is quite correct for the commercial poultryman to make
inquiries on such points. The broiler producer, of course, will not be inter-
ested in such matters and if, as is usually the case, he buys crossbred chicks,
he may inquire about performance tests, not of the parents but of the crosses
he is offered.

The breeder's approach to the improvement he is responsible for will often
provide worthwhile clues. If he emphasizes high individual records or mini-
mum individual standards for commercial-chick-producing flocks, it is un-
likely that he is paying more than lip service to family selection. These words
may be unfair to some breeders who feel that high individual records have
an appeal to commercial poultrymen. But if poultrymen become convinced
that including a 150-egg hen from a 250-egg family in their flock may on
many occasions be preferable to using a 300-egg hen from a family with a
production index of 100 eggs, the breeder would no longer be under pressure
to frame advertising in terms of individual records.

What the commercial poultryman is entitled to get and should demand
from the breeder is information on performance and selection standards of
the whole flock and not of exceptional individuals in it. The problem is more
difficult when dealing with hatcheries or multipliers rather than direct
producers of improved stock. There is always some dissipation of genetic
merit when a small group of tested and selected birds is expanded into a
large chick-supplying flock. The usual methods of culling on the basis of
physical appearance do little to stop this trend. The only guide the poultry-
man then has is an indirect one : it is the breeding policy of the breeder sup-
plying the hatchery itself.

Laying Tests and Official
Improvement Schemes

More or less the same things we have said about the breeder's
claims apply to contest results and to official improvement schemes. Excep-
tionally high records at laying tests are of little value to the commercial
producer unless he is certain that the stock he buys has been produced by the
same methods producing the contest winners. The birds in laying-test pens
may, for instance, represent crosses between two strains and exhibit heterosis,

46 PRINCIPLES OF COMMERCIAL

whereas the commercial poultryman may be sold chicks from one of the
strains which were used in the cross but which are unexceptional in them-
selves.

This is why random sample tests of the kind which have been inaugurated
in California are of greater value than the usual contest. These tests are
designed to measure the production qualities (egg and meat) of commercial
stock offered for sale, and not of the breeder's cream of the crop. The reports
of the California tests are further useful because they take into account not
only production but also costs. The standing of the contestants is determined
by net financial returns, not by gross, and net returns are of course the more
important returns to the commercial poultryman.

Government improvement schemes in various countries have different
bases. Data from them must be as closely scrutinized as data from private
sources. A label certifying that a hatchery or a breeder has complied with
the regulations of a particular government-sponsored scheme is meaningless
unless the provisions of the scheme itself are understood. Very often disease-
control schemes are classified under "improvement." There can be no objec-
tion to the word so long as it is understood that such schemes do not refer
to genetic improvement. Similarly, some schemes or parts of schemes are
based purely on physical selection, a method which may prevent deteriora-
tion of improved stock to some extent, but certainly cannot produce further
improvement. Other schemes may permit culling or preselection of the birds
to be entered under government supervision. The commercial poultryman
should know the definitions used before he can make use of the reports as a
guide for choosing stock.

Conclusion

Our discussion has implied many responsibilities for the com-
mercial poultryman. But it is in his own interest to have them. While many
breeders and hatcherymen have a genuine desire to improve the qualities of
their stock, they can only afford to undertake elaborate breeding methods
if their customers are willing to pay for at least part of the added costs. The
extra costs per chick are actually very small when compared to the gains which
sound breeding policies can produce. Under present conditions, a gain of
ten eggs in the production index of an average flock could be readily achieved
by investing a very tiny fraction of day-old chick prices. But the incentive
for making the investment must come from the commercial poultryman. He
is the one to put pressure on the hatchery and the breeder to adopt sound
breeding techniques. His understanding approach to the whole question will
mean his own benefit and the benefit of the poultry industry.

POULTRYBREEDING 47

With help from the poultryman, the breeder and hatcheryman may look
ahead to consequences by gradually modifying their respective breeding
methods if they are, indeed, inadequate. They can do this only by familiariz-
ing themselves with the genetic principles underlying the processes of poultry
breeding. We have noted that there is no one universal recipe that a breeder
can follow as a housewife follows a cookbook. Understanding is required
first of all. The best breeding scheme will break down if unintelligently used,
and even the least efficient one may have some redeeming features if applied
with understanding.

Reading

No books designed to acquaint the practical breeder and the
poultryman with the intricacies of modern genetics as applied to their field
are yet available. A general text, Animal Breeding Plans, written by Jay L.
Lush and published by the Collegiate Press, Ames, Iowa, discusses most of
the important aspects of the subject. But it is intended primarily for students
of livestock rather than poultry breeding and is technical. Also technical is
Population Genetics and Animal Improvement by the present writer (Cam-
bridge University Press) on which much of our discussion is based. The sci-
entific articles published in such journals as Poultry Science present the same
difficulty.

A rather more general approach to poultry breeding is given in Poultry
Breeding, by Morley A. Jull (John Wiley, New York) . It does not deal with
the genetic basis of traits of economic importance as we do, but it is particu-
larly valuable for a descriptive treatment of productive characters and for
full reference lists following each chapter.

A later book, The Genetics of the Fowl, by F. B. Hutt (McGraw-Hill, New
York) is a good source of information on characters largely determined by
single-gene-pair differences. But it does not make use of available modern
techniques in treating quantitative characters.

3m-3,'51(2528)JB

t "XVTE

Selling price $ .48545

Mate sales tax $ .01455

Total price $ .50

THIS MANUAL is one of a series published by the University of California College
of Agriculture and sold for a charge which is based upon returning only a portion
of the production cost. By this means it is possible to make available publications
which due to relatively high cost of production or limited audience would other-
wise be beyond the scope of the College publishing program.