Am. ZOoLocisT, 4:139-145 (1964).

JAN 8 1965

BREEDING ANALYSIS OF NATURAL UNITS
IN BEHAVIOR GENETICS

Jerry Hirscu

Department of Psychology, University of Illinois, Urbana

On receiving my assignment for this re-
fresher course, my self-appointed task be-
came an examination of the term behavior,
and consideration of the kind of knowl-
edge genetic analysis has to contribute to
the understanding of behavior and behav-
ior analysis has to contribute to under-
standing in biology.

At a reception following a conference
two summers ago, Joshua Lederberg asked
me “How do you define behavior?” It’s a
disconcerting experience at any time to be
caught off guard harboring an unanalyzed
premise. It’s doubly so when it happens
on one’s first encounter with a Nobel lau-
reate. A survey of textbook and dictionary
definitions proves very unsatisfactory and
suggests a possible reason for his question.
In one form or another, most definitions
hinge upon response to stimulation. As
Skinner (1938) has so appropriately
pointed out, while response to stimulation
certainly does occur, so does behavior occur
in the absence of antecedent events that
bear an easily demonstrable stimulating
relation to a given behavior.

BEHAVIOR

In the physical sciences, ever since J. W.
Gibbs, a system has been defined as that
part of the universe chosen for consider-
ation. In the behavioral sciences we can
define behavior as whatever an organism
does or, by analogy to physical science, that
part of what an organism does that we
choose for consideration.

The study of behavior employs the de-
sctiptive methods of the naturalist, clas-
sificatory taxonomic methods such as are
used by the medical diagnostician and the
biological systematist, and the analytic ex-
perimental methods of the laboratory sci-
entist. In Ernst Mayr’s words “it is the
basic task of the systematist to break up

the almost unlimited and confusing diver-
sity of individuals in nature into easily
recognizable groups, to work out the sig-
nificant characters of these units, and to
find constant differences between similar
ones” (1942, p. 9). Similarly, the student
of behavior must select out of the continu-
ous stream of activity that is behavior,
those units which are suitable for study
and which can be classified.

The importance of a “suitable” choice
of descriptive features and taxonomic cate-
gories will be appreciated when it is real-
ized that the units of behavior chosen and
the criteria by which they are recognized
will strongly influence the future course of
research with respect to both the kinds of
analyses performed and the kinds of inter-
pretations made. It is convenient, there-
fore, to consider the study of behavior as
consisting of three distinguishable phases:
description, taxonomy, and analysis.

Description. To be part of science, non-
introspective observations must be formu-
lated in terms that will place them in the
public domain. An adequate description
of a behavior presents a statement of its
differentiating properties and of its rela-
tions to other events, so that other ob-
servers can identify it.

Descriptions of behavior may refer to
either means or ends—what Hinde (1959)
has called “physical description” and “de-
scription in terms of consequences.” Ideal-
ly, physical descriptions would be made in
terms of the means by which a behavior
is executed, and thus would refer to the
intensity, duration, and pattern of the
physical activities of the body. In practice,
though, to avoid cumbersome detail, only
the grosser aspects of behavior are usually
described, e.g. the startle reaction, the
amount of time spent in crouching by a
bird, the rate of intromission in sexual

(139)
140

behavior, etc. Descriptions in terms of
consequences, made by reference to the
goals of behavior, do not always include
an account of the particular means by
which those goals are sought. This type of
description groups into broad categories
(such as reproductive behavior, homing,
maze-learning) all those performances that
achieve the same end state.

The two types of description, however,
are not always alternative. In some cases,
for example, description in terms of mus-
cular contractions is undesirable because
of the unwieldy complexity of the data
that result. In other situations, description
by consequence is equally undesirable.
“Sometimes—as in the threat and courtship
posture of birds, which involve both a rela-
tively stereotyped motor pattern and an
orientation with respect to the environ-
ment—both must be used” (Hinde, 1959,
p. 571).

Taxonomy. An ideal behavioral taxono-
my would provide a theoretical system un-
der which classes of behavior could be ar-
ranged in some meaningful and consistent
way. The classes in the taxonomy would
be defined in terms of the behaviors they
include. In such a system the subjects of
classification are behaviors, and the sub-
jects of taxonomy are classifications (King
and Nichols, 1960; Simpson, 1961). As yet,
no taxonomic system has been devised
which attempts to do for behavior what
the Linnaean system attempted for species.

Analysis. Analysis breaks behavior
down into components that can be related
to antecedent and contemporary events, as
well as to the physical properties of the
organism. The responses of any behaving
organism can be represented as points in a
four dimensional space whose axes may be
labeled: response, time, stimulus, and indi-
viduals. The experimental analysis of be-
havior may be thought of as the study of
relations among variations along these four
dimensions, because behavior shows: (1)
temporal variation or differences in re-
sponse over time; (2)  stimulus-response
covariation or changes in response as a
function of changes in stimulation; and

JERRY Hirscu

(3) individual variation or response char-
acteristics dependent upon the particular
organism observed.

Terms like conditioning, learning, matu-
ration, fatigue, adaptation, sensitization,
etc., refer to those aspects of behavior in
which, for a given individual observed
under constant stimulus conditions, a re-
sponse changes over time. ‘Tropisms, taxes,
kineses, preferences, and all of the phycho-
physical relations refer to those aspects of
behavior in which, for a given individual
observed over a given time interval, the
response changes as a function of changes
in stimulation. Mental tests and psycho-
metric methods refer to techniques for the
measurcment and analysis of individual
differences in response; for all behavior,
under a given set of stimulus conditions
over a given time interval and in a speci-
fied population, the characteristics of the
response will depend upon the particular
individual observed. The three major
methodologies employed in the experi-
mental analysis of behavior are condition-
ing, psychophysics, and psychometrics
(after Hirsch, 1962).

LEVELS OF ORGANIZATION AND DIMENSIONS
OF ANALYSIS

The most characteristic feature of all
biological material is “organization,” by
which I mean that when two or more
things are put together something new ap-
pears, the qualities of which are not always
additive or expressible in terms of the
qualities of the constituents. Electrons and
nuclei combine to form atoms, atoms to
molecules, molecules to amino acids, amino
acids to proteins, proteins and nucleic acids
to nucleoproteins, etc. Chemical and col-
loidal structures are integrated into the
specialized organelles, different organelles
are integrated into cells, cell populations
into tissues, different tissues into organs and
systems of organs, these into multicellular
organisms, bisexual organisms into Men-
delian populations, and the last finally into
ecological communities.

These levels of organization are proper-
ties of nature, some of which are obvious
BREEDING ANALYSIS IN BEHAVIOR GENETICS

and readily observable as in the case of an
individual organism, while others, as in the
case of the chemical constituents of the
hereditary material, have only been dis-
cerned as the result of decades of research.

‘The strategy of the scientist’s approach
to the study of behavior involves the use
of dimensions of analysis. Stimulus con-
ditions are manipulated as independent
variables and response measures are ob-
served as dependent variables. The meth-
ods of research, however, must not be con-
fused with the objects studied. Upon the
presentation of radiant energy of different
wavelengths, properly instructed human
subjects respond with different hue names
(Graham and Ratoosh, 1962). The data
which are collected in this way can be used
to test: (1) genetic theories about the in-
heritance of genes controlling the develop-
ment of color vision; (2) physiological theo-
ries about the nature and functioning of
the color receptor apparatus (Fuller and
‘Thompson, 1960); or (3) behavioral theo-
ries about psychophysical relations. Cer-
tainly, dimensional analysis is a powerful
tool which can yield much reliable informa-
tion about behavior. The interpretation
of that information, however, requires both
an understanding of biological organiza-
tion and an appreciation of the levels of
organization for which the information
from a particular analysis may be relevant.

BIOLOGICAL STRUCTURE AND GENETIC
ANALYSIS

In the discussion that helped set Maurice
Wilkins on the road leading to the Watson-
Crick model, the physicist Erwin Schrédin-
ger points out how “It is well-nigh unthink-
able that . . . laws and regularities should
happen to apply immediately to the be-
havior of systems which do not exhibit the
structure on which those laws and regu-
larities are based” (1946, p. 3).

We now possess considerable information
about the master plan of most organisms
and about the genotypic structure of bi-
sexual species populations. The many im-
plications that this knowledge has for the
study of behavior are just beginning to be

141

appreciated. In the first half of this cen-
tury, it was learned that organisms have
properties that are as fundamental to be-
havioral science and biology as thermody-
namic properties are to physical science.
Mutation, recombination, and meiosis are
properties of organisms that generate an
endless succession of unique genotypes,
upon which natural selection and genetic
drift exert the influences that mold the
populations these genotypes comprise.

Morphologically, the fundamental units
of life may be chromosomes. Each chromo-
some, we know, is linearly subdivided into
local units of function called genes. The
image of beads on a string, such as we see
in an abacus or a lady’s necklace, provides
a sufficiently precise physical analogy for
this discussion.

Since the chromosomes come in pairs,
the genes also come in pairs. The geno-
type is the name for the set of all gene pairs
on the total number of chromosome pairs
in the karyotype of a species. By means of
breeding analysis, each pair of genes, in
principle at least, is distinguishable from
every other pair; and the allelic alterna-
tives are frequently, but not always, dis-
tinguishable from one another. The alleles
of a gene pair are chosen from a set of
alternative forms which seems to have no
fixed number. For some genes, only two
alternative forms have been distinguished,
for others many have been recognized. A
familiar example of multiple alternatives
is the ABO blood system in man.

In a geometry that facilitates generaliza-
tion, I shall represent each gene by a
dimension in a multidimensional coordi-
nate space, where the permissible values
along each dimension are the allelic alter-
natives of the gene it represents. Each
chromosome contains hundreds, possibly
thousands, of genes. Hence the haploid
chromosome complement of the gametes,
called the genome, can be represented by
a point in an n-dimensional space, where
the letter n represents the total number of
genes and is fairly large, say anywhere from
10,000 to 50,000 in man. Since the geno-
type results from the union of two inde-
pendently formed gametic genomes, it can
142

be represented by a point which is the
intersection of two n-dimensional genome
spaces. Therefore, genotypically an indi-
vidual is a single point in a hyperspace
that defines the sample space of possible
genotypes for a species.

Since the dimensions in the hyperspace
are independently variable and their num-
ber is very large, the probability of any two
genotypes having all the same coordinates
and therefore occupying the same point in
a sample space is infinitesimally small.

The population is the effective breeding
unit of the species. It consists of all those
members of a species that comprise a geo-
graphically and temporally continuous
group, and among whom reproduction is
possible. The members of a population
can be represented by a distribution of
points in the hyperspace of genotypes. The
set of genotypes from one generation is
replaced by a set of different genotypes the
following generation. Consideration of the
nature of the genotype has shown that,
although a population consists of sets of
substitutable and replenishable points, each
point is unique and non-replicable, There-
fore, the population can be represented
over the generations by a cloud of ever-
shifting points. We now realize that an
understanding of the structure of hetero-
genic, biparental populations with the man-
ifold multivalued dimensions of allelic var-
iation that comprise their gene pools is of
fundamental importance in the study of
behavior. Since the lawful relations which
make up the content of science are an ex-
pression of the structure of the events
studied, the more intimate our knowledge
of that structure becomes the clearer will
be our ideas about what relations are pos-
sible.

Behavior has been defined as whatever
an organism does that we choose to con-
sider, Furthermore, the activities of an
organism are lawful events—the lawful
functioning in a specific environmental con-
text of a biological system with a specific
structure.

Szent-Gyérgyi and his associates isolated
actin and myosin from muscle tissue. When
they recombined these two chemicals in

Jerry Hirscu

the presence of adenosinetriphosphate and
certain ions, the threads of the highly vis-
cous actomyosin contracted before their
eyes. “To see them contract for the first
time and to have reproduced in vitro one
of the oldest signs of life, motion, was per-
haps the most thrilling moment of my
life” said Szent-Gy6rgyi (1963). He was ob-
serving the emergence of one element of
behavior when the proper constellation of
constituents had been arranged.

If we put organisms into the appropriate
conditions, much about their behavior can
be predicted from knowledge of their phy-
logeny, their genotype, and their ontogeny
—most Drosophila seem to be negatively
geotactic; by selective breeding, however,
we have produced genotypes which show
a predominantly positive geotaxis. The im-
portant point of the discussion by Schrédin-
ger, cited above, is that lawful relations
derived from one system cannot be ex-
pected to apply immediately to other sys-
tems of different structure. Genetics ex-
plains the simultaneous existence, within
a species, of both similarities and differ-
ences in structure among individuals.

While it is true that we have some knowl-
edge about the master plan of most organ-
isms, it is also true that with respect to
understanding how behavior occurs we
have barely started to map the functional
anatomy of any organism. Obtaining such
a map is a very difficult task because the
usual methods of physiology and biochem-
istry involve insult to the very systems
under observation. An operated or drugged
animal is a different organism from the
one whose behavior we wish to understand.

Only the breeding techniques of genetics
—selection, inbreeding, hybridization, and
test-crossing—permit analysis of both the
structure and the functioning of biological
systems without insulting the integrity of
the behaving organism to be observed. Gen-
etic methods, however, involve the study
of populations over several generations.
Selection has been defined as the non-
random differential reproduction of geno-
types. This method, and the others, permit
us to control the composition of popula-
tions. By selectively breeding among in-
BREEDING ANALYSIS IN BEHAVIOR GENETICS

dividuals rank-ordered on one or more
dimensions of analysis, it is possible to ef-
fect drastic reorganizations at many levels
of biological organization. Breeding meth-
ods, in fact, accomplish an elegant dissec-
tion of the total organism under study and
a subsequent re-assembly of its component
parts into an array of alternative organi-
zations. Never before has the student of
behavior had available a technique power-
ful enough to influence without insult every
cell in the body of the behaving organism.

In Benson Ginsburg’s well chosen words
“Just as the parsing of a sentence gives us
only a certain kind of understanding about
it, so the analysis of some aspects of an
organism into gene-controlled reactions
tells a limited story, but it is a story that
adds immeasurably to our understanding
of living systems in their behavioral, as well
as morphological and physiological, capaci-
ties” (1958). It is in this way that we can
discover the natural units of biological or-
ganization. They are natural in the same
sense that Ginsburg describes phenylketo-
nuria in man as a natural entity. It is an
abnormal allelic alternative to a gene-con-
trolled link in a metabolic chain that has
been produced by evolutionary selection
pressure. Tyrosinosis and alkaptonuria are
also genetic alternatives to the normal.
They involve adjacent links in the same
chain. Through these genetic deviations
from what natural selection has produced
as the norm, the chain and its links are
being identified. A phenotype on which
evolutionary selection could have acted as
well as its substituent genotypes—especially
those genotypes showing deviations from
the norm, as in the examples just reviewed
—constitutes what some of us mean by a
natural analyzable entity. If such a pheno-
type and its genetic-metabolic correlates
can be studied in a broad comparative
series of taxonomically related organisms,
its evolutionary history may also be eluci-
dated. Several laboratories have been at-
tempting to do this with the audiogenic
seizure syndrome.

‘There are presently available to us two
well known series of studies which illus-
trate how genetic analysis yields informa-

148

tion that can be obtained in no other way.
The studies of the effects of domestication
in the rat, which Richter reviewed, show
that in the domesticated animals activity
as measured in a revolving drum was
mainly dependent upon gonadal secretion,
whereas in the wild animals it was mainly
dependent upon adrenal secretion. Appar-
ently, the selective influences of domestica-
tion reorganized the endocrine balance, in-
creasing gonadal output relative to adrenal
output.

About a decade ago a team of psycholo-
gists and a biochemist began studying the
chemical basis of learning (Rosenzweig
et al., 1960). When they compared two
reproductively isolated strains of rats, they
found that the two strains differed be-
haviorally in their maze performance, and
chemically in their cholinesterase concen-
tration. They reported their finding of a
correlation between the expression of these
two traits, and then embarked upon an
ambitious research program exploring
many ramifications of this empirical rela-
tionship. Years later when the appropriate
breeding controls were instituted, the chem-
ical-behavior correlation broke down in
the progeny of hybrids between the two
strains.

It has taken a very long time for it to
be appreciated that there is no surgical or
biochemical substitute for proper breeding
analysis. In genetic terms, the correlation
between two traits can represent either the
pleiotropic effects of an integrated genetic
system or the accidental combination of
two or more independent genetic systems.
It is only the breeding of F, hybrids and
their subsequent Fy, and backcross test
crosses which affords us the chance, not the
certainty, of distinguishing between these
two alternative possibilities.

Quite recently I was consulted about a
physiological study that was looking for a
chemical-behavior correlation. When I
pointed out that, even though a different
behavior and a different biochemical were
involved, the problem was analogous to the
cholinesterase-maze—performance fiasco and
must therefore remain inconclusive with-
out appropriate breeding controls, my ad-
144

vice was met with the remonstration “But
we haven’t enough time or facilities for
genetic experiments; we hadn’t planned to
study the genetics of this problem!”

Perhaps the phrase “the genetics of this
problem” brings out a fundamental and
widespread misconception. Jts prevalence
may be one of the reasons it has been so
difficult for both behaviorists and biologists
to talk sense about the relations between
heredity and behavior. Many discussions
of heredity and behavior have run afoul of
the nature-nurture issue. They have been
formulated in a causal frame of reference
and, therefore, may have been misleading.
It is a fallacy to ask the nature-nurture
question for any behavior. The best we
can do is to consider heritability, a concept
which refers to the correlation between
genotypic diversity and individual differ-
ences in behavior. Moreover, this correla-
tion has no fixed value; for every behavior
it must be measured in specific populations
under specific conditions, because it varies
with both.

At a conference the summer before last,
when I objected to the phrase “the genetics
of behavior,” I was overwhelmed by Benson
Ginsburg who was supported by Ernst Cas-
pari and apparently by almost everyone
else there. Those of you who followed the
advance notices of this refresher course may
have noticed that its title changed from
“Genetics of behavior’ in the summer or
autumn issues of Science magazine to “Be-
havior genetics” in its recent issues. So it
is clear that we teachers of this refresher
course are, at best, only one step ahead of
its students.

It might once have been useful to think
in the causal frame of reference and it
probably helped to establish genetics. Gen-
ctics is now well established and has a
beautifully articulated body of knowledge.
Furthermore, the broad features of the re-
lations between heredity and behavior are
rapidly becoming clear. What must next
be more widely appreciated is that it is
not a specialty to be relegated to a con-
ceptual corner. As Sewall Wright observed
some time ago, it is a fallacy to speak of
physiological effects and genetic effects as

Jerry Hirscu

though they were two different classes of
phenomena, because genes are the ultimate
physiological agents,

Breeding experiments are the only meth-
ods we have available for varying the bal-
ance and composition of physiological sys-
tems. It is in this way that we can articulate
and discover those properties and interac-
tions which are neither immediately ob-
vious nor easily revealed by surgical and
biochemical methods alone.

Behavior Analysis and Biology

In closing, I should like to call to the
attention of geneticists in particular and
biologists in general that behavior might
provide a most sensitive measure of organ-
ismic diversity. To quote an observation
by Fuller and Thompson (1960) that I have
cited many times “. . . two individuals of
superficially similar phenotypes may be
quite different genotypically and respond
in completely different fashion when
treated alike.’ As both Dobzhansky and
Mayr might say: Fuller and Thompson’s
observation is incomprehensible to typo-
logical thinking and self-evident to popula-
tion thinking.

REFERENCES

Fuller, J. L., and W. R. Thompson. 1960. Behavior
genetics. Wiley, New York, N. Y.

Ginsburg, B. E. 1958. Genetics as a tool in the
study of behavior. Perspect. Biol. Med. 1:397-424.

Graham, C. H., and P. Ratoosh. 1962. Notes on
some interrelations of sensory psychology, percep-
tion and behavior. In S. Koch, [ed.], Psychology:
a study of a science. Vol. 4. McGraw-Hill, New
York, N. ¥.

Hinde, R. A. 1959. Some recent trends in ethology.
In 8. Koch, [ed.], Psychology: a study of a science.
Vol. 2. McGraw-Hill, New York, N. Y.

Hirsch, J. 1962. Individual differences in behavior
and their genetic basis. In E. Bliss, [ed.], Roots
of behavior. Paul B. Hoeber, New York, N. Y.

King, J. A., and J. W. Nichols. 1960. Problems of
classification. Im R. H. Waters, D. A. Rethling-
shafer and W. E. Caldwell, [eds.], Principles of
comparative psychology. McGraw-Hill, New York,
N.Y.

Mayr, E. 1942. Systematics and the origin of species
from the viewpoint of a zoologist. Columbia
Univer, Press, New York, N. Y.

Rosenzweig, M. R., D. Krech, and E. L. Bennett.
1960. A search for relations between brain chem-
BREEDING ANALYSIS IN BEHAVIOR GENETICS

istry and behavior. Psychol. Bull. 57:476-492.
Schrédinger, F. 1946. What is life? Macmillan,
New York, N. Y.
Simpson, G. G. 1961. Principles of animal taxon-
omy. Columbia Univer. Press, New York, N. Y.
Skinner, B. F. 1938. The behavior of organisms: an

145

experimental analysis.
New York, N. Y.

Szent-Gyérgyi, A. 1963. Lost in the twentieth cen-
tury. In E. E. Snell et al., [eds.], Annual review of
biochemistry. Vol. 32. Annual Reviews, Inc.,
Palo Alto, Calif.

Appleton-Century-Crofts,