The Journal of Heredity 68:297-300. 1977.
Heritability in retrospect
A. EARL BELL
WHO COINED the term "heritability"? This
question recently evoked considerable debate
in an interdepartmental genetics seminar,
organized around the general theme "Genetics in
the Bicentennial." As chairman of the seminar 1 had suggested
the concept of heritability for one of six sessions
devoted to population and quantitative genetics. The
graduate student selecting this topic, Stephen Rich, had
little difficulty in documenting the tremendous contributions
made by J.L. Lush to the understanding and application
of both broad and narrow sense heritability, especially
as these concepts relate to the genetic improvement
of farm animals. In due course, the student traced
the basic concepts to the classical papers of Fisher5
and
Wright21
. In fact, no problem arose until he sought to
establish the origin of the term "heritability."
Since my own graduate student days I had assumed
that Dr. Lush originated the term. Who else? But when
challenged for documentation, I became apprehensive
after checking unsuccessfully Lush's Animal Breeding
Plans'5
, his mimeographed The Genetics of Populations,
and my personal notes from his animal breeding classes.
Then I turned to more general information sources.
For example, a couple of genetics dictionaries"'12
define
heritability without giving its origin. A third18
cites
Lush's 1939 paper16
which simply lists the various
methods of estimating heritability. Cook's2
chronology of
genetics gives the origins of many genetic terms and concepts,
including the major milestones of biometry and
animal breeding, but has no mention of heritability. Many
textbooks (e.g., Kempthorne'sA/i Introduction to Genetic
Statistics ">, p. 464) cite the 1937 edition of Lush's
Animal Breeding Plans as their source reference. Much to
my surprise, I could not find the word heritability in
the 1937 edition (the concept is accurately presented in
terms of the relative importance of hereditary and environmental
sources of variance). Yet Lush's second edition
(1943) uses heritability extensively without any
historical reference.
My previously stated conviction as to the originator of
the term was reinforced by Sewall Wright's 1961 com-
ments22
: "There has been considerable drifting in the
meaning of the term heritability and the symbol h2
. I
think I introduced the latter in a paper in 1920 as the degree
of determination by heredity. . . . I think Lush introThe
author is professor of genetics, Department of Animal
Sciences, Purdue University, West Lafayette, Indiana 47907. He
wishes to acknowledge and thank Dr. J.L. Lush for permitting
the quotation from personal correspondence. Journal Paper
Number 6548 of Purdue University Agricultural Experiment
Station.
duced the term heritability and the narrow sense of the
additive component. . . . "
Following Wright's suggestion, I traced Lush's usage of
the term to a 1936 publication14
where he titled a major
section "Summary of Evidence on Heritability." Yet
the word heritability was not defined or used again in
this section or elsewhere in the bulletin. In this and
later papers from that period, Dr. Lush frequently used
heritability without a hint that he might be coining a new
term, but with the implication that its usage was understood.
Nevertheless, in these early papers he obviously
preferred such terms as "hereditary in the narrowest
sense" and "portion of the individual variance due to additive
gene effects."
At this dead end and with the initial question unresolved.
I gave up the search and posed the question to Dr. Lush.
His carefully considered reply went far beyond the immediate
question and detailed historical background in a
manner that merits a wider audience than my seminar
group. Any attempt to paraphrase his remarks would distract,
so I quote directly.
My answer won't be as clear-cut and definite as you
might wish, but that is the way things happen. The ideas
themselves overlap. Especially when they change a little
with time. And the operations, which really define them,
may vary at least a little from one user to another.
The following comments I take from a little 3" by 5" card,
such as I keep, in a file headed "definitions." On such
cards I put details about various definitions whenever
I have occasion to consider the history and discordances
of some of those. These comments come from the card entitled
"heritability."
Darwin, in the chapter on "Natural Selection" in the
ninth edition of his book The Origin of Species, says
". . . as well as the strength of the hereditary tendency."
This seems to come close to heritability as we use it today
although, of course, it implies nothing about the discrete
nature of the units, or their segregation or how the variation
is conserved.
Galton must have said many things about this general
subject after about 1885. He wrote voluminously about
heredity and always tried to quantify his ideas as much as
possible. I have not, however, noted that he used
"heritability" specifically.
E.D. Davenport, in 1907 on pages 486-490 et seq. of his
book Principles of Breeding, uses as "the coefficient of
heredity" the correlation between relatives, although I
think he was not discriminating clearly between different
kinds of relatives nor was he using any factors to multiply
those correlations. Also, he seems somewhat confused
about the difference between regression and correlation
for interpreting "the strength of heredity." He draws
freely on Galton. Perhaps he generalizes even more
broadly than Galton did?
Wilhelm Johannsen deserves credit for distinguishing
297
Downloadedfromhttps://academic.oup.com/jhered/article-abstract/68/5/297/881010byMendelUniversityinBrnouseron13December2018
298 The Journal of Heredity
clearly between variance caused by differences among
the individuals in their genotypes and variance due to
differences in the environments under which they grew.
This is close to what I call "heritability in the broad
sense." I take it that "Erblichkeit" can be translated
fairly as heritability. He did not subdivide clearly the
genetic variance into that which is additive, that which is
due to dominance, and that which is due to epistasis. He
had little need of the latter two in his own material
because he worked mostly with populations of pure lines.
His most pertinent warning against generalizing too
freely is in the second edition (1913) of his text book
where, on page 210, he warns against, "Die falsche Vorstellung,
das Ahnlichkeit ohne weiteres ein Austruck von
Verwandtschaft sein muss—Oder dass gar "Erblichkeit"
durch "den Grad der Ahnlichkeit zwischen Verwandten'
ausgedruckt werden konne." On page 358 (the 1913 edition
of his text "Elemente derexakten Erblichkeitslehre")
he does use "Erblichkeitsziffer" for regression on midparent.
On page 122 he uses regression of offspring on
mid-parent to estimate "Grad von Erblichkeit."
Much in Johannsen's text is still worth reading,
even today. When he put his ideas and evidence on "pure
lines," "genotype" and "phenotype" in German (1903)
for circulation in other countries, he left out some
interesting material about why he undertook these experiments
in the first place (this is in the slightly earlier report
which was printed in Danish). Briefly, he ws trying to
see how far he could generalize "Galton's Law" and
whether it could be reconciled with the Mendelian ideas
of DeVries. He dedicated the 1903 paper to Galton as the
creator of genetics (Schopfer der exakten Erblichkeit-
slehre).
Lauprecht (in a letter to me dated February 3, 1962)
thinks that Jacob Peters in 1913 was the first to use the
idea of heritability in agriculture practice. Peters worked at
Konigsberg in East Prussia. He was, I think, the first man
to use sire indexes, although he didn't carry them far.
He used them mainly to correct the sire's progeny test
for the phenotypes of his mates.
Apparently my own usage of the word "heritability"
began with the second edition of my book, rather than
with the first. At least I do not find it in the first edition
on a casual scanning with that in mind. Nevertheless,
pages 63 to 70 in that first edition are entirely
about what we would now call heritability. I was, at that
time, using variance components or various expressions
such as "the genetic portion of the variance in . . ."
(page 477 in the Journal of Heredity, Vol. 23, 1932);
". . . the relative importance of the different causes of
variance in production records"; "to what extent are the
observed differences an accurate guide to the hereditary
differences . . ."; "how much of the observed variation
is hereditary . . ."; and as long ago as 1923 on page 32 of
Texas Bulletin 311, "therefore the hereditary benefits to
be derived from culling are going to depend entirely and
the present benefits will depend somewhat on the amount
of hereditary variation in wool production in the flock
which is to be culled."
The first place I notice where heritability was used
systematically was in Research Bulletin 204 on the Danish
progeny testing stations published by the Iowa station in
1936, but actually written in 1935. There on pages 182 to
184 is a systematic section entitled "Summary of Evidence
on Heritability." There is also a fairly good
statement on the limitations and qualifications of the ideas
which were derived from these operations.
From looking at these bits and from memory I think
I must have been systematically avoiding the use of a
single word, lest the readers oversimplify it and apply it
too widely to conditions for which it was not suited.
That is, I was trying to guard against a single term being
overworked and used illegitimately, as happened often
with statements like those by Davenport. At least I used
longer sentences or clauses than I would now use. This
amounted to using the definition of the word repeatedly
instead of the word itself.
I think I was moved out of that position by the practical
necessities in summarizing the Danish work. That is, the
point of view, that one should use only the variance
components or their ratios, works well enough when comparing
two rather similar populations, as when one asks
whether the heritability of milk production in DH1A
records differs from that in semiofficial records, or the
heritability of rate of gain in pigs is larger when the
pigs are about five months old or is larger when they
are about two months old. It gets almost hopelessly confused,
however, if one tries to compare the heritability
of traits which are different in kind, or are expressed on a
different scale, such as I was doing with the Danish
records, or one might do concerning percentage of
butterfat in milk and quantity of milk and so forth.
There may, of course, always be a question of whether
it "makes real sense" to compare the heritabilities of
different traits or in widely different populations. Doing
that does often give us some information we would miss
otherwise, however.
I used diagrams showing the partitioning of variance
in fractional form at least as long ago as 1933 (Journal
of Dairy Science 16:516). Also I used a tabular form to
express this at least as long ago as 1934 (Genetics 19:341).
Leaving it in variance components, rather than as fractions
of those, seems to have certain technical statistical
advantages when one is concerned only with statistical
significance. Fisher and Snedecor were stressing unduly
the testing for significance in the early 1930's. Many of us
were going along with that. (I'm not sure we have yet recovered
completely but that is another story!) I merely
mention that as part of the explanation for the widespread
preference for expressing our findings in variance components,
rather than to express them as fractions of variance.
Perhaps this was only a minor skirmish in the
sometimes-heated expression of our opinions about the
relative merits of correlation coefficients and regression
coefficients.
I tried to summarize my thoughts on heritability for
the Genetics Congress at Stockholm in 1948. I still regard
that as one of my best summaries on the topic although,
of course, there are still other ways for saying many of
these things.
In none of my work did I go much into variance due to
interaction between heredity and environment. I merely
mentioned that. Perhaps Falconer has so far made the best
stab at that? [Perhaps Wrights's circular causation should
point the way?]
I think none of us has yet taken a detailed, hard look
at variance due to cooperation or to competition. It is
imaginable that these could be highly important in some
populations and for some traits. Obviously, the whole
field of the genetics of such traits as altruism, patriotism,
aggressive self-interest and so forth could be very
important in species living under semi-social conditions
and in experimental plants and animals which are grouped
in physically separate plots or in pens and must use resources
which are always somewhat limited. I have never
pursued this much further than to get the feeling that these
things would act like negative intraclass correlations between
members of the same social or competing subgroup.
Since it is mathematically impossible for an intraclass
correlation to be both negative and large when there
Downloadedfromhttps://academic.oup.com/jhered/article-abstract/68/5/297/881010byMendelUniversityinBrnouseron13December2018
Bell: Heritability in retrospect 299
are many members in each group, the very existence of
such a situation would be hard to measure experimentally
with statistically significant results. Yet, if they do compete
for several different limited elements in the environment,
the total effect could be large. I don't see how to
break through this logjam experimentally.
In summary of what is in this rather rambling and long
letter, I would say:
1) The idea of heritability is so old and so general that
establishing the history of its use is almost impossible.
2) So many people have had the general idea, since
even long before Darwin's time, that it almost certainly
emerged many times in the literature and submerged again.
3) I got my impetus for whatever I did in this area primarily
from the demands of the practical problems on
which I was working. Starting out with repeatability
of fleece weights in sheep and Angora goats, I extended
these to problems of breeding dairy cattle and pigs.
The most useful steps naturally varied a little according
to the material.
4) I got my biggest boost in the area and general nature
of the problem from following Wright's analysis of the inheritance
of amount of white spotting in piebald guinea
pigs in 1920. Then I extended it by trying to understand
thoroughly his classical work on mating systems in 1921
and by auditing his class in statistical genetics at Chicago in
the spring of 1931. Had I known in detail Johannsen's
articles from 1902 until 1914 I might have been five or ten
years further along the road. I think we give him less
credit in this area, than any other man, compared to what
his work really deserved. Probably Fisher should be given
credit for first separating the variance due to dominance
from the rest of the genetic variance in 1918, although
others, such as Jennings, had made small steps in this
direction. However, Fisher's paper, like the 1902-1903
Danish papers of Johannsen, was not very widely read;
its main influence came much later.
5) Wright, more than anyone else, clarified the effects
of systems of mating on estimates of, and indeed on
the very idea of, heritability.
6) If I and my colleagues deserve any special credit it
would be more for having assembled and tried to bring
under a common interpretation the estimates and ideas of
heritability of many traits in various farm animals.
The relation of heritability to what is ideal in breeding
plans in a given instance was never far from our thoughts.
This area was much explored by the corn breeders,
notably Richey and Sprague.
7) I was helped very much by ideas I got from Fisher
when he lectured here for five weeks in the summer of
1931. He then generally dismissed mating systems as of
little importance on the ground that, if a species in nature
was small enough in numbers for the effects of inbreeding
to become important, that species was likely to be
on the verge of extinction already and would soon slip over
the edge. Therefore, inbreeding and its consequences
had little bearing on evolution, however spectacular
their results might be for a few generations with domestic
plants and animals.
8) Apparently it does little good to warn against oversimplifying
the idea of heritability. Some feel it necessary
to repeat those warnings each time they write.
Others ignore the warnings. Much of the bulk of which has
been written in recent years about heritability in man
comes from this. Of course, much of that bulk is because
people simply do not like the conclusion! No amount
of expounding about heritability would alter that by more
than a little, I think. Only time and continued efforts
of the researchers to remain unbiased seem likely to
clarify this at all.
9) Perhaps this is a basic argument for coining a new
word (preferably from Greek roots) when an idea is to
be presented precisely and in a way in which it cannot
be misunderstood. Concerning the latter aim, however,
perhaps the following "law of frustration" is pertinent:
"If you work hard to state something so precisely and
clearly that nobody can misunderstand it, somebody will!''
On receipt of the above, I suggested to Dr. Lush that
he should submit it as a historical note to an appropriate
journal. His negative response is characteristic of the
thoroughness with which he treats a question. To him,
the finding and verifying of all borderline cases would
take more time than he felt it was worth. To illustrate,
he noted in reply that by accident he had recently come on
a 1926 usage of heritability9
.
Dr. Lush's response stimulated me to spend another
afternoon in our library where I found that Kearney8
had used heritability as early as 1921 in reference to small
continuous fluctuations in cotton and the expected response
to selection. And in 1920, Whipple19
used "heritability
of fluctuations" and "nonheritability of fluctuations
due to environment" in describing selection
studies with potatoes.
These early papers by plant breeders used heritability
in a descriptive sense, rather than statistical, and reflected
Johannsen's concepts of phenotypic variation arising
from environmental as well as genotypic fluctuations, yet
no reference was made to Johannsen's well known find-
ings6
'7
. Animal breeders at the turn of this century
were actively studying quantitative traits with most following
the Galton-Pearson biometrical approach.
Among the books from this period, Dr. Lush has
commented on Darwin's use of "strength of hereditary
tendency" and Davenport's use of "coefficient of
heredity." Another, somewhat typical of that era, was
Lock's13
Recent Progress in the Study of Variation,
Heredity and Evolution which emphasizes the GaltonPearson
approach. Included is a separate chapter describing
Johannsen's pure-line concepts from which he quotes,
"Individuals which differ (in size) from the mean of the
population give rise to offspring which differ from that
mean value in the same direction but to a smaller extent,
. . . but selection within a pure line produces no effect."
However, Lock makes no effort to interrelate the opposing
points of view, nor does he translate Erblichkeit
or use heritability. Yet by 1920 some plant breeders were
using heritability as noted above. It is possible that
even earlier plant breeders had actually translated
Johannsen's Erblichkeit to heritability (as suggested by
Dr. Lush), but I have been unable to document this
hypothesis. Maybe some reader can provide the missing
evidence.
While modern German-English science dictionaries4
-20
translate Erblichkeit as heritability, such was not the case
for early geneticists. Many English publications have referred
to Johannsen's pure-line concepts without translating
Erblichkeit. A 1938 Swedish publication1
makes
the English translation as variation and Crew3
chose
heredity. This apparent reluctance of early geneticists
to translate Johannsen's Erblichkeit to heritability, combined
with the early usage of the latter term by plant
breeders without reference to Johannsen, suggested to
me that the term heritability must have had a broader
Downloadedfromhttps://academic.oup.com/jhered/article-abstract/68/5/297/881010byMendelUniversityinBrnouseron13December2018
300 The Journal of Heredity
general usage prior to Johannsen's classical findings.
Thus the older and more general term would be avoided in
describing Johannsen's more precise concepts of genotype
and phenotype. For the same probable reason, the
term was avoided in the Galton-Pearson biometrical
studies of heredity. Support for this hypothesis comes
from a 1901 " definition of heritability as "The quality
of being heritable or capable of being inherited",
and gives 1832 as the earliest recorded use of the word in
Fraser's Magazine as "This tax, thus securing the
heritability of offices, was not perpetual." Also listed was
a 1882 quotation by the naturalist A. Gray: "The importance
of heritability, which is an essential part of
Darwinism, would seem to have had a significant illustration
in the person of its great expounder."
Summary
The origin of the word heritability remains unknown.
Its usage has evolved through three stages, becoming
more restrictive in its meaning along the way. In the initial
stage, 1832 and possibly earlier, heritability was used to
denote the hereditary transmission of characteristics or
material things, simply having the capability (legally or
biologically) of being inherited. The second stage, beginning
around the turn of this century, followed
Johannsen's classical definition of nongenetic or environmental
fluctuations distinct from genotypic differences,
and usage closely approximated "broad sense heritability"
and Johannsen's Erblichkeit. Finally, in 1936, we
come to the modern day usage of narrow sense heritability,
the ratio of additive genetic variance to the total
phenotypic variance within a population, and credit
Dr. J.L. Lush with its origin.
Literature Cited
1. AKERMAN, A., J. GRANHALL, G. NILSSON-LEISSNER,
A. MUNTZING, and O. TEDIN. Swedish Contributions to the
Development of Plant Breeding. Alb. Bonniers Boktryckeri,
Stockholm. 1938.
2. COOK, R. A chronology of genetics. U.S.D.A. Yearbook
of Agriculture, p. 1457-77. 1937.
3. CREW, F.A.E. The Foundations of Genetics. Pergamon
Press, N.Y. 1966.
4. DEVRIES, L. German-English Science Dictionary.
McGraw-Hill, N.Y. 1946.
5. FISHER, R.A. The correlation between relatives on the
supposition of Mendelian inheritance. Trans. R. Soc. Edinburgh
52:399-433. 1918.
6. JOHANNSEN, W. Uber Erblichkeit in Populationen und in
reinen Linien. Fischer, Jena. 1903.
7. . Elemente der Exakten Erblichkeitslehre. Fishcher,
Jena. 1909.
8. KEARNEY, I.H. Heritable variations in an apparently uniform
variety of cotton. J. Agric. Res. 21:227-241. 1921.
9. and R.H. PEEBLES. Heritability of different rates of
shedding in cotton. J. Agric. Res. 33:651-661. 1926.
10. KEMPTHORNE, O. An Introduction to Genetic Statistics.
John Wiley & Sons, New York. 1957.
11. KING, R.E. A Dictionary of Genetics. Oxford Univ.
Press, New York, 3 ed, Springer-Verlag, New York. 1968.
12. KNIGHT, R.L. Dictionary of Genetics. Chronica Botanica
Co. Waltham, Mass. 1948.
13. LOCK, R.H. Recent Progress in the Study of Variation,
Heredity and Evolution. 3 ed. E.P. Dutton & Co. New York.
1911.
14. LUSH, J.L. Genetic Aspects of the Danish System of
Progeny Testing Swine. Iowa Research Bulletin 204. 1936.
15. . Animal Breeding Plans. The Collegiate Press,
Ames, Iowa. 1937, 1943, and 1945.
16. . Methods of measuring the heritability of individual
differences among farm animals. Proc. 7th Inter. Genetics
Congress. 1939.
17. MURRAY, J.A.H. A New English Dictionary on Historical
Principles. Claredon Press, Oxford, England. 1901.
18. RIEGER, R., A. MICHAELIS, and M. M. GREEN. A Glossary
of Genetics and Cytogenetics. Springer-Verlag, N.Y. 1968.
19. WHIPPLE, O.B. Line selection work with potatoes.
J. Agric. Res. 19:543-573. 1920.
20. WILDHAGEN, K. and W. HERONCOURT. The New Wildhagen
German Dictionary. Follett Publishing Co. Chicago.
1965.
21. WRIGHT, S. The relative importance of heredity and
environment in determining the piebald pattern of guinea pigs.
Proc. Nat. Acad. Sci. 6:320-332. 1920.
22. . Discussion to P. Robinson's paper titled "Heritability:
A second look." In: Statistical Genetics and Plant Breeding.
NAS/NCR Publication 982. p. 613. 1963.
Downloadedfromhttps://academic.oup.com/jhered/article-abstract/68/5/297/881010byMendelUniversityinBrnouseron13December2018