The Dominance of Traits in Genetic Analysis

Author(s): Raphael Falk

Source: Journal of the History of Biology, Vol. 24, No. 3 (Autumn, 1991), pp. 457-484
Published by: Springer
Stable URL: http://www.jstor.org/stable/4331190
Accessed: 19/08/2009 12:22

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=springer.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact support@jstor.org.

Springer is collaborating with JSTOR to digitize, preserve and extend access to Journal of the History of
Biology.

http://www.jstor.org
The Dominanceof Traitsin GeneticAnalysis

RAPHAEL FALK

D)epartmentof Genetics
The Hebrew University
9l1904Jerusalem, Israel

In the modern interpretation of Mendelism, facts are being

transformed into factors at a rapid rate. If one factor will not
explain the facts, then two are invoked; if two prove insuffi-
cient, three will sometimes work out. The superior jugglery
sometimes necessary to account for the results may blind us, if
taken too naively, to the common-place that the results are
often so excellently "explained" because the explanation was
invented to explain them.1

In his introductory comments to the English translation of

Mendel's paper, Sir Ronald Fisher points out that Mendel repre-
sented his work not as an attempt to solve a great problem, but
rather as a contributor to the methodology of research into plant
inheritance. Mendel had studied the earlier writers, and he tells us
in just what three respects he thinks their work should be
improved upon: proper care should be given, he suggests, to "the
distinction between generations, to the identification of genotypes,
and, to this end, to the frequency ratios exhibited by their
progeny, when based on an adequate statistical enumeration."
Consulting Mendel's own (translated) words reveals, however,
that he referred to "forms in which progeny appear" - that is, to
phenotypes, as we have come to denote these forms, rather than
to the hereditary potentials, or genotypes: "Whoever surveys the
work in this field will come to the conviction that among the
numerous experiments not one has been carried out to an extent
or in a manner that would make it possible to determine the
number of different forms in which hybrid progeny appear, permit
classification of these forms in each generation with certainty, and

1. Thomas H. Morgan, "What Are 'Factors' in Mendelian Explanation?"

Amer. Breed. Ass., 5 (1 9 09), 365.
2. J. H. Bennett (ed.), Experiments in Plant Hybridization (Edinburgh: Oliver
and Boyd, 1965), p. 3.

Journal of the History of Biology, vol. 24, no. 3 (Fall 1991), pp. 457-484.
C) 1991 Kluwer Academic Publishers. Printed in the Netherlands.
458 RAPHAEL FALK

ascertain their numerical interrelationships."3 Although Mendel

did not conceive the difference between genotype and phenotype,
I will suggest that in the framework of his experimental design
such as distinction was not necessary, exactly because he devised
another important methodological procedure - namely, the
analysis of individual, discrete, easy-to-follow traits. In such an
experimental context, there was a one-to-one relation between the
trait and the unit of inheritance, except for the phenomenon of
dominance. This is the reason why dominance - a correlation of
the phenotypic expression of the genotype - played such a
central role in Mendel's work.
In a sense Mendel was a methodological reductionist. Method-
ological reductionists endeavor to identify in their experiments
entities that may be heuristically treated in future calculations and
tests as if they corresponded, in a roughly one-to-one relationship,
to discrete entities of a relevant theory. However, when Mendel's
work was rediscovered, his definition of traits was given a pro-
found theoretical meaning that played a significant role in the
discussions on the importance of discontinuous versus continuous
variation in the origin of species. I argue that even after the
distinction between phenotype and genotype was established,
Mendel's methodological reduction was maintained as a concep-
tual, even ideological, reduction of questions of heredity to traits
that are mappable to genes, and vice versa. This had far-reaching
consequences in genetic theory and its applications. Furthermore,
I contend that this attempt to superimpose a conceptually mean-
ingful content upon a methodologically helpful reduction is not
limited to genetic analysis: it has been tried, for example, also in
taxonomy. There, as in genetic analysis, it is the case that the
reduction is heuristically helpful only as long as the specific frame
of reference, or the system, in which the entities are treated is
appropriately narrowly delineated.

INTRASPECIFIC VERSUS INTERSPECIFIC HYBRIDS

Contrary to most of his predecessors, Mendel focused his

attention primarily on intraspecific hybrids.4 His predecessors,
studying predominantly interspecific hybrids, were inevitably en-

3. Gregor Mendel, "Experiments on Plant Hybrids" (English trans.), in C.

Stern and E. R. Sherwood, The Origin of Genetics (San-Francisco: Freeman,
1906), p. 2.
4. See Hans Nachtsheim, "Gregor Johann Mendel," Erbarzt, /0 (I1942).
147-154.
The Dominance of Traits in Genetic Analysis 459

gaged primarily with problems of fertility of the hybrids. Only a

low proportion of the cross-pollinations were successful, and most
of these did not produce fruits. Even among those that produced
some fruit, only few were normal in appearance, and only on rare
occasions were seeds recovered for further cultivation of the
progeny of the hybrids.
The hybridization experiments of Mendel's admired predeces-
sor, Carl Friedrich von Gartner are typical: in his opus magnum
of 1849 Gartner published data on thousands of hybridization
experiments performed in the years 1825-1848, all of which he
believed to be interspecific (Table 1).5 It is hardly surprising that
of the fifteen subsections into which Gartner divided the chapter
"Of the Identifying Markers and Characteristics of Hybrids" (57
pages long), only one is devoted to "Change and Stabilization of
the Hybrids in Their Progeny" (6 pages long). He mentions that
"some plant physiologists suggested the hypothesis that those
hybrids that were endowed with fertility and that maintained a
constant type may be transformed to stable forms." However, the
problem that concerned him was the role played by the fertility
(or sterility) of hybrids in the classification of "systematic distinct
species, and the possibility of the origin of new species through
hybridization."6 Gartner mentions the few examples of such an
acquisition of stability in hybrids described in the literature, as
well as the few cases observed by him - notably that of the
hybrid Dianthus armeria X D. deltoides, in which the type was
maintained for ten generations: during six to eight generations, he
claims, these hybrids maintained themselves through self-propaga-
tion, but due to "successive reduction in fertility of the seeds" they
were lost after ten generations. He stresses, however, that "except
for these few examples, the fertile hybrids produce in the second
generation, and to a larger extent in successive generations, no
uniform type, but rather varieties, variants, and deviants, part of

5. There is no doubt that many of his hybridizations would nowadays be

classified as intervariety, or cultivar, crosses. It is very difficult to sort out the
taxonomy of the hybrids in modern terms. It seems that at least some infertility
was produced by crossing varieties of different ploidy (I am grateful to Prof. Ed
Klekowski for his opinion). Mendel himself notes that even for his peas "the
systematic classification is difficult and uncertain.... In the opinion of experts,
however, the majority belong to the species Pisum sativum; while the remaining
ones were regarded and described either as sub-species of P. sativum, or as
separate species, such as P. quadratum, P. saccharatum, and P. umbellatum
(Mendel, "Experiments," pp. 4-5).
6. Carl F. von Gartner, Versuche und Beobachtungen uber die Bastarder-
zetugungim Pflanzenreich (Stuttgart: K. F. Hering, 1849), pp. 684-728.
460 RAPHAEL FALK

Table 1. Summary of hybridization experiments of C. F. Gartner

in the years 1823-1848

Number Percentage

Hybridizations 12964
Flowers hybridized 62688 11001
Fruits obtained 23335 37.22
Normal fruit and seed number 257 0.41
Nearly complete fruits with:
almost normal seed number 2584 4.12
reduced seed number 1410 2.25
only few good seeds 1852 2.95

Source: Summarized from Carl F. von Gartner, Versuche und

Beobachtungen uber die Bastarderzeugung im Pflanzenreich
(Stuttgart:K. F. Hering, 1849), pp. 684-728.

which pull back to the maternal type, others stepping forward

toward the paternal type.7
Why did Mendel decide to abandon the traditional methodol-
ogy of interspecific hybridization? Was it because he wished to
study hybrids quantitatively - or was it that he decided to study
intraspecific hybridization, and consequently found he could
perform his studies for many generations and obtain quantitative
results? Unlike Gartner and others of his predecessors who wrote
learned books about nature, Mendel profoundly integrated his
scientific work with the practical problems of his Moravian
countrymen.8 To respond to the demands of the expanding
economy, new methods to improve crops were imported. The
breeding method of producing hybrids between strains with
desired qualities, followed by intensive inbreeding as a device to
"stabilize" the new combinations, which was popularized in
England by Robert Blakewell, was brought to Moravia by Ferdi-
nand Geisslern. Although the experts dismissed as "city-folk's
talk" the time-honored claims that inbreeding causes degenera-
tion,9 breeders were apparently plagued by the repeated segrega-

7. Ibid., p. 553.
8. See V. Orel, "Selection Practice and Theory of Heredity in Moravia
before Mendel," Acta Mus. Moravia 42 (1 977). 179-200. See also E. B.
Gasking, "Why Was Mendel's Work Ignored?" J. Hist. Ideas, 20 (1959), 60-84.
9. It is reasonable to assume that they overcame the adverse effects of
The Dominance of Traits in Genetic Analysis 461

tion of the prized hybrids. The fertile hybrids, in spite of the pro-
mises of the promoters of the breeding-by-hybridization method,
were not as stable as expected."' I suggest that it was in an attempt
to understand this problem encountered by breeders that Mendel
came to focus on (intraspecific) fertile hybrids - and it was his
interest in intraspecific hybrids that led him to develop his new
methodology in the study of hybridization.

MENDEL'S DISCOVERY OF SIMPLE TRAITS

It is important to note that once Mendel was on the path of

studying intraspecific rather than interspecific hybrids, the whole
nature of the problem of hybridization changed. The accent on
infertility as the characteristic of hybrids disappeared. The over-
ruling attribute of the hybrids that caught the investigator's atten-
tion was not so much the differences between the hybridized
parents, as the identity in most of their traits.'' Inevitably, atten-
tion was (or might have been) directed at the few conspicuous
discrete traits that varied in the otherwise identical parental
strains: "When two plants, constantly different in one or several
traits, are crossed, the traits they have in common are transmitted
unchanged to the hybrids and their progeny, as numerous experi-
ments have proven; a pair of differing traits, on the other hand,
are united in the hybrid to form a new trait, which usually is
subject to changes in the hybrid's progeny." '2
This assertion of Mendel about the union of differing traits
forming a "new trait" sounds odd when one remembers that the
essence of his thesis was that no new traits were found in hybrids,
but rather that what happened was the reappearance of the
original traits in specific numerical ratios. Such a phrasing, makes
sense, however, when he indicates his aim to study the (stable)
production of hybrids of crop plants, in which case the "new trait"

inbreeding to some extent by inadvertent selection for those recombinant

variants that were better adapted to the prevailing conditions.
1(0. V. Orel, Mendel (Oxford: Oxford University Press, 1984); see pp. 1)0-
13.
11. Gasking ('Why Was Mendel's Work Ignored?" p. 64) suggested that
Mendel's decision to study the inheritance of particular characters led him to be
indifferent to the classificatory position of his plants, while I contend that once
Mendel became involved in intraspecific hybrids, he was led to study individual
traits. It is noteworthy that Darwin's results on hybridization and breeding in
domestic species also focused on the inheritance of individual traits. Darwin, too,
was often nearer to breeders and horticulturalists than to institutional scientists
in his experiments and observations back on his estate at Down.
12. Mendel, "Experiments," p. 5.
462 RAPHAEL FALK

would be improved (overall) yield. Once liberated from the con-

straint of the infertility of hybrids he could concentrate on the
prospect of producing plants with new traits, as far as the breeder
was concerned.
But the great accomplishment of Mendel was to view the
hybrids in a different way: rather than focusing on hybrids as
exemplars with new traits, he envisioned the complex trait of the
hybrid as the combination of a small, discernible number of easily
recognizable simple traits. It is this ability to redefine the new
complex-trait of hybrids as nothing but a combination of a few of
the parental traits that led Mendel to his first great discovery,
which he expressed in one sentence: "Thus the study breaks up
into just as many separate experiments as there are constantly
differing traits in the experimental plants."'3 Such a reducing
redefinition of traits was probably upheld when Mendel charac-
terized the twenty-two varieties that he selected for his hybridiza-
tion experiments. It is most likely that at least some of the traits
thus defined also manifested their integrity as entities in different
combinations among the various varieties.'4
Mendel went one step further in his reductionist definition of
traits. Five of the seven traits that he chose to work with in his
hybridization experiments are described as discrete morphological
characteristics of the pea plants. However, the description of each
of the remaining two traits, numbers 3 and 5, includes a combina-
tion of two apparently unrelated morphological phenomena. Trait
No. 3 relates to "the difference in coloration of the seed coat. This
is either white, in which case it is always associated with white
flower color; or it is grey-brown, leather-brown with or without
violet spotting, in which case the color of the standard lof the
legume flower] is violet, that of the wings is purple, and," he adds,
"the stem bears reddish markings at the leaf axils." Trait No. 5
relates to "the difference in color of the unripe pod. It is either
colored light to dark green or vivid yellow," but it relates also to
"the coloration of stalks, leaf-veins, and calyx."'5 Mendel realized,
probably by examining the results of his hybridization experi-
ments, that one phenomenon - the coloration of the seed coat, in
one case, or the color of the unripe pod in another case - always
goes in association with another phenomenon - flower color, or
stalk coloration, respectively. Consequently he decided to pool

13. Ibid.
14. See Federico Di Trocchio, "Mendel'sExperiments:A Reinterpretation,"
in idem, Legge e caso nella genetica Mendeliana (Milan:Angeli, 1989).
15. Mendel,"Experiments," p. 6.
The Dominance of Traits in Genetic Analysis 463

two different pairs of morphological phenomena as (pleiotropic)

effects of the same unit-trait. Thus he not only introduced the
concept of simple discrete traits as the variable entities of experi-
ments of heredity, but also established the heuristic of defining a
unit-trait as that which is transmitted as a unit in hybridization
experiments.
Indeed, when Mendel carried out his experiments with Phaseolus
hybrids, he went still further and suggested that the "simple" trait
of flower and seed color "is composed of two or more totally
independent colors that behave individually exactly like any other
constant trait in the plant,"'6 exactly because these did not fit the
single trait pattern. Thus Mendel not only reduced the unit-trait to
a "genotypic unit" but, at this point, circularly, he also interpreted
it as that entity that behaves in breeding experiments as an entity.'7

DOMINANCE OF TRAITS OR OF FACTORS?

It is significant that Mendel developed the concept of heredi-

tary factors, Faktoren or Anlagen, without conceiving any mate-
rial particulate theory of heredity.'8 Indeed, Mendel explained his
results without employing invisible particulate determiners. Once
he had discovered the device of studying discrete, practically non-
overlapping individual traits, he was in a position to study the
laws of heredity of traits.
In his discussion of the reproductive cells of hybrids Mendel
notes that "constant progeny can be formed only when germinal
cells and fertilizing pollen are alike, both endowed with the poten-
tial for creating identical individuals, as in normal fertilization of
pure strains.""9Therefore, he forms the hypothesis that

in a hybrid plant also identical factors are acting together in the

production of constant forms. Since the different constant
forms are produced in a single plant, even just a single flower,
it seems logical to conclude that in the ovaries of hybrids as

16. Ibid., p. 35.

17. Such a procedure may be tolerated as long as it is clear that the reducing
as well as the reduced entities are meaningful only within the limits of the experi-
mental setting. In later years William Castle criticized the concept of the gene as
a material unit of genetics exactly on this issue. Vide infra and see R. Falk,
"What Is a Gene?" .Stud.Hist. Phil. Sci., /7:2 (I 986), 143.
18. See F. V. Monaghan and A. F. Corcos, "The Real Objective of Mendel's
Paper," Biol. Phil., 5 (1990), 267-292; and R. Falk and S. Sarkar, "The Real
Objective of Mendel's Paper: A Response to Monaghan and Corcos,", Biol. Phil.
(in press).
19. Mendel, "Experiments," p. 24.
464 RAPHAEL FALK

many kinds of germinal cells (germinal vesicles), and in the

anthers as many kinds of pollen cells are formed as there are
possibilities for constant combination forms and that these
germinal cells correspond in their internal make-up to the
individual forms.2"

It is this hypothesis that he is going to test in what follows (making

the additional assumption that "the different kinds of germinal
and pollen cells are produced on the average in equal numbers").
Iris Sandler has suggested that the explicit distinction made by
Mendel in the above-quoted paragraph between factors that are
acting in the production of forms, indicates that he realized that
his characters were only the external markers of unobserved, yet
real, hereditary units.2' But this suggestion is belied by the rest of
Mendel's paper. Both before the above quotation and later in his
paper, Mendel denotes by the "genotypic" notation the "pheno-
typic" traits: "If A denotes one of the constant traits" (p. 16); "the
differing traits of the seed plant will be indicated in these experi-
ments by A, B, C' (p. 17); "every pollen form A and a will unite
equally often with every germinal-cell form A and a" (p. 30).
More important, as I have already mentioned, Mendel did not
need to make this distinction between what we call today the
genotype and the phenotype.22
Before trying to substantiate this claim, it will be helpful to ask
first why Mendel paid so much attention to the phenomenon of
dominance. Dominance does not concern the rules of inheritance,
but rather the problem of the expression of the inherited proper-
ties. It would seem that the job would have been facilitated
considerably if he had worked with traits that do not show
dominance (i.e. those where the heterozygote's phenotype is
intermediate between that of the two parents). Mendel, however,
is very explicit about deliberately choosing traits with dominance,
rather than following "the intermediate form of some of the more

20. Ibid.
21. Iris Sandler, "PierreLouis Moreau de Maupertuis- A Precursorof
Mendel?"'J.Hist.Bio., 16 (1983), 101-136.
22. Richard Lewontin, in a paper on "Genotypeand phenotype"(Manu-
script for E. F. Keller and E. Lloyd, eds., 1989), claims that "it was Mendel,
twenty years before Weismann,who made this distinction[betweenthe change-
able soma and the underlyingconstantgerm plasm]before Weismann."I cannot
agree with this. I thinkthat Lewontinrelegatedto Mendelthat whichvery-much-
latter-dayMendelists would conceive as the distinctionbetween genotype and
phenotype.
The Dominance of Traits in Genetic Analysis 465

striking traits."23The answer to this query should be examined at

three levels:

(1) The practical level. As noted before, Mendel was inter-

ested in thie methods of breeding new agriculturally significant
forms through hybridization. Intermediate forms of the desired
traits were not interesting enough.
(2) The methodological level. A quantitative analysis of the
data demands that the classification into types should be as clear-
cut and distinct as possible. Although Mendel endeavored to
choose among the 34 varieties at his disposal the 22 that showed
most constancy in the traits among their progeny, he notes that
"some traits listed do not permit a definite and sharp separation,
since the difference rests on a 'more or less' which is often
difficult to define."24Any classification of the range of variation of
the traits into three categories would be bound to result in more
ambiguous overlaps than the classification into two alternative
categories that he adopted. Fortunately, his practical interests
directed him away from traits with intermediate dominance.25
(3) The theoretical level. This may perhaps be better formu-
lated as the lack of theoretical level. Mendel had no reason to
differentiate between the rudimentary undeveloped trait in the
germinal cell (and the pollen cell) and the trait observed in the
developed plants and their fruits and seeds. It is because of this,
that the unobservability of a trait that "was there" attracted so
much of Mendel's attention. Had he understood the significant
role that development may play in the realization of the Faktor
into a Form, the phenomenon of dominance would still have
posed practical and methodological problems to his study of
transmission genetics, but it would not have been part of the
theory of transmission.

In order to appreciate this, it is necessary to recall the impact

that the concepts of materialist chemical and physical research
had on biological research in the first half of the nineteenth

23. Mendel, "Experiments," p. 9.

24. lbid, pp. 5-6.
25. Prof. R. C. Lewontin pointed out to me the irony that had Mendel
chosen to work with the related species of sweet peas (Lathyrus) rather than
garden peas (Pislm), he would have found that hybrids between plants with
white flowers and plants with red tlowers produce in the second generation three
practically nonoverlapping classes of offspring with red, pink, and white flowers,
respectively. For this and other helpful comments I am most grateful to him.
466 RAPHAEL FALK

century. Against such a background, preformationistic concepts of

development and heredity could gain their place beside the still-
prevalent epigenetic ideas - this time, however, detached from
the theological connotations of the previous century. This replace-
ment of epigenetic concepts by modern preformationism was of
course extended especially toward the end of the century, with the
corollary of August Weismann's concept of the separation of the
soma and the germ line, the latter alone being effective in the
transmission of variation to the next generation.26 For Mendel,
whose main training was in physics, a preformationist approach to
the concept of heredity was very much at hand. Physicists main-
tained essentially that the observed phenomena could be directly
explicated as the unfolding of causal interactions of well-defined
material entities and the forces exerted on them.
With such a preformationist frame of mind there was no need
for Mendel to distinguish between the character proper and the
"potential for the character" - or, as we would say today,
between the phenotype and the genotype of the trait. Ontogenic
development had no creative role to play, since the trait was
already somehow present in the germ cells. There was no con-
ceptual difference for him between the trait and the factor. The
only problem was to deternline whether the trait (= factor) dis-
appears in the hybrids; or, more accurately, to establish that it
disappears only from sight, and to formulate the laws of its dis-
appearance and reappearance. Dominance was an important
concept, just because it referred directly to the thing that needed
explanation: the existence of the trait, even though it could not be
observed. Thus Mendel's preoccupation with the problem of
dominance in the study of trait transmission in hybrids was a
direct consequence of his inability to conceive of a difference
between a trait and a gene for a trait, which, through an epigenetic
process, might or might not be expressed as a (phenotypic) trait.
However, when Mendel's work was rediscovered at the begin-
ning of the century, his practical and methodological reducing
definition of traits was given profound theoretical interpretations.
In the disputes over the role of continuous versus discontinuous
variation in the origin of species by natural selection, Mendel's
heuristic was taken as support for the advocates of the discon-
tinuous concepts. It was not accidental that one of his redis-
coverers, Hugo de Vries, was the promoter of what was soon
thereafter to be denoted the mutation theory of evolution, and

26. See P. J. Bowler, The Mendelian Revolution (Baltimore: Johns Hopkins

University Press, 1989), pp. 23-32.
The Dominance of Traits in Genetic Analysis 467

that its most avid evangelist was William Bateson, a longtime

advocate of discontinuous variation.
Dominance could become a part of genetic theory only because
no clear distinction was made between the inherited potential for
a trait and the trait proper.

TRAITS DENOTE TYPES

As I have indicated, at the rediscovery of Mendel's work in

1900, the discreteness of traits had both a theoretical biological
and a philosophical significance. The debate over the role of
continuous small changes in the evolution and the origin of
species by natural selection was at its climax. Many leading
biologists, notably de Vries and Bateson, were advocating the
need for large-step variants as the raw material for Darwinian
evolution. Mendel's studies with discrete traits were perceived not
as just methodological tools to uncover the rules of heredity, but
as inherently indicative of a type of variation of nature.27 This
view of the living organism as a mosaic of discrete traits complied
with the prevailing philosophical realism that the entities that we
perceive correspond more or less to reality, and that we can
reduce the entities and processes of living systems to finite,
discrete, basic units of biological structure and function.28
When Wilhelm Johannsen performed his experiments with the
quantitative traits of beans, such as the length of the seeds, their
breadth, and their weight, he soon discovered that by intensive

27. See, for example, A. Brannigan, "The Reification of Mendel," Soc. Stud.
Sci, 9 (1979), 423-454.
28. Such an approach was in sharp contrast to the organicist, holistic
concepts in biology that were especially promoted by students of embryology.
The confrontation between holistic science, particularly holistic biology, and the
strong mechanistic materialist conception must, of course, be seen in the wider
context of the sociology of science. It involved not only power struggles within
the community of biological scientists, but also struggles for the authority of
science in the wider political and ideological context. It had repercussions not
only in the construction of the sharp distinction between genotype and pheno-
type, but also in such circumstances as the role of scientists - primarily "race-
hygienists" - in the politics of the Third Reich. Further discussion of the
sociological, political, and ideological aspects of mechanistic materialism versus
organicist holism in beyond the scope of the present paper - but see Scott F.
Gilbert, "The Embryological Origins of the Gene Theory," J. Hist. B3iol., 11
(1978), 307-351; Jan Sapp, Beyond the Gene (Oxford: Oxford University
Press, 1987); and Garland Allen, "T. H. Morgan and the Split between Embryol-
ogy and Genetics, 1910-1935," in History of Embryology, ed. T. J. Horder and
J. A. Witkowsky (Cambridge: Cambridge University Press, 1985), pp. 1 3-146.
468 RAPHAEL FALK

inbreeding for a number of generations, they too could be treated

as discrete (i.e., nonoverlapping in their distribution) traits. Much
of the observed variability that remained was "irrelevant"as far as
the inheritance of the traits was concerned. Once he inbred the
progeny of specific plants, he increased the genetic "purity"of the
inbred lines by reducing the frequency of heterozygotes among
the progeny in a geometric series, as Mendel had already shown."'
The only variability left was the environmental "noise" of classifi-
cation; otherwise, phenotypic variability converged on genotypic
variability. As pointed out by Nils Roll-Hansen, Johannsen, in a
genuine Aristotelian typological spirit, claimed that the genotype
of an individual inbred plant and its progeny embody the truthful
species concept, the "genospecies."'3"On such an account Johann-
sen distinguished "individual" or "fluctuating" variability from
"mutuation (stepwise changes, spontaneous variation)," which,
according to de Vries's mutation theory, were the raw material of
natural selection. Only stepwise mutations could contribute to
evolution, and these could be identified as traits - that is, traits
were real, discrete entities of living beings. Thus, although Johann-
sen did not perform Mendelian segregation analyses with the
variables that he measured, so as to determine that the progeny of
hybrids of two inbred lines segregated in the discrete Mendelian
ratios, he implied that variables such as seed breadth or seed
weight conform to a Mendelian discrete trait concept. For
Johannsen, the concepts of the phenotype and the genotype
explicitly served to emphasize the need for a statistical distinction
between the hereditary type and the environmental noise, without
any reference to the developmental process that intervenes be-
tween the genetic potential and the phenotypic execution.

TRAITS BECOME DISTINCT FROM THE GENES

With a concept of units of heredity that was completely

oblivious to the processes of embryological development and the
role of extranuclear cytoplasmic components in the determination
of the fate of the cells of organisms, the denial by embryologists
and physiologists of the Mendelian conception of the inheritance
of traits is hardly surprising. Scott Gilbert emphasizes that T. H.
Morgan's early resistance to Mendelian inheritance and the

29. Mendel, "Experiments," p. 16.

30. Nils Roll-Hansen, "The Genotype Theory of Wilhelm Johannsen and Its
22 (1978).
Relation to Plant Breeding and the Study of Evolution," Cemitaluruis,
201-235; see pp. 221-227.
The Dominance of Traits in Genetic Analysis 469

chromosomal theory of inheritance was profoundly grounded in

his antipreformationistic, epigenetic approach to problems of
embiyology.31 From 1906 to 1909 Morgan, the experimental
embryologist, conducted a purely cytological study of some
parthenogenetic species of aphids in order to see whether or not
the chromosomal basis of sex determination, as proclaimed by
Nettie Stevens and by E. B. Wilson, would hold for such organ-
isms. In females, eggs destined to become females developed
differently from those destined to be male-producing. In the latter,
after the polar body was extruded, an additional pair of chromo-
somes entered it. Also during male meiosis, two types of sperm
were formed: one type, with six chromosomes, survived and
eventually fertilized the eggs, resulting in females; the other type,
with only four chromosomes, degenerated. Morgan saw in these
observations strong support for his conception that "the sex
determinant was whatever cytoplasmic factor moved the chromo-
somes."32 Even when he published his first results on the inherit-
ance of the "sex-limited" white eye color in Drosophila in 1910,
positing that all the sperm from white-eyed flies carried the factor
W for white eyes, and admitting that "the fact is that this R [for
red eyes] and X [for sex] are combined, and have never existed
apart," Morgan avoided any conceptualization of these factors as
nuclear ones.33 The interdependence did not necessarily mean
that these were linked physically together on the chromosome:
Our general conclusion is, therefore, that the essential process
in the formation of the two kinds of gametes of hybrids in
respect to each pair of contrasted characters, is a reaction or
response in the cells, and it is not due to a material segregation
of the two kinds of materials contributed by the germ cells of
the two parents. . . . The general point of view that underlies
this conclusion is epigenetic, while the contrasting view, that of
separation of materials, is essentially one of preformation.
As noted by Garland Allen, "Iblehind all of Morgan's objection
to Mendelism in 1910 lies one fundamental problem. This is his
confusion of the phenotype and genotype of an individual."'35But

31. Gilbert, "Embryological Origins of the Gene Theory."

32. Ibid., p. 342.
33. Thomas H. Morgan, "Sex Limited Inheritance in Drosophila," Science,
32(1910), 122.
34. Thomas H. Morgan, 'Chromosomes and Heredity," Amer. Nat., 44
(1910), 479.
35. Garland Allen, "Thomas Hunt Morgan and the Problem of Sex Deter-
mination, 190 3-19 10," Proc. Amer. Phil. Soc., 110 (1 966), 5 3.
470 RAPHAEL FALK

by the summer of 1911 Morgan had shown that at least three

other characters and sex all segregated together with the
X-chromosome. "To explain these drastic deviations from random
assortment, Morgan was forced to espouse the Mendelian prefor-
mationism against which he had contended for over a decade."36
There was, however, no need to return to the preformationist
concept of the genetic trait. By that time Morgan and his school
were reinterpreting Johannsen's statistical, horizontal distinction
between phenotype and genotype of a population in terms of a
vertical, embryologists' developmental concept of an individual.37
There was no identity between the germinal-cell factor and the
trait of the organism. The phenotype was just the way that the
genotype was revealed (insofar as it was revealed) under given
environmental (internal-developmental, as well as external) cir-
cumstances. There was no conflict between a particulate theory of
information-transmission, and a physicochemical theory of devel-
opment. It is remarkable that Morgan the embryologist and his
students, acknowledging the distinction between the trait and its
hereditary potential, turned their attention to the aspect of the
transmission of information, essentially leaving the epigenetic
aspect to be treated by a later generation. As noted by Leo Buss,
"fundamentally .. . the success of Morgan's group was the result
of the explicit program of the laboratory to divorce problems of
transmission genetics from problems of cell physiology and onto-
geny."'38The fact that Morgan, a central figure in embryological
research at the time, turned his back so absolutely on the prob-
lems of the phenotypic expression of the genotypic potentials,
must have played a decisive role in the hegemony of the gene that
has prevailed ever since in genetic research.
The role of the environment in the implementation of the
genotypic potential as a phenotypic trait was admitted. Terms like
"expression" or "penetrance," which were introduced by Bateson
to explain (away) deviations from the expected Mendelian ratios
at the phenotypic level, could now be understood in the epigenetic
context as related to environmental factors that are necessarily
involved in the execution of the (genotypic) potential for develop-
ing a trait and the (phenotypic) trait proper.3: However, in the

36. Gilbert, "Embryological Origins of the Gene Theory," p. 345.

37. Frederick B. Churchill, "William Johannsen and the Genotype Concept,"
J. Hist. Biol., 7(1974), 5-30.
38. Leo W. Buss, The Evolution of Individuality (Princeton: Princeton
University Press, 1987), p. 10.
39. "Penetrance" is the proportion of individuals with a given genotype that
The Dominance of Traits in Genetic Analysis 471

context of the mechanistic conceptual dichotomy of genotype and

phenotype, "environment" remained for a long time a nuisance
that had to be dealt with, rather than being accepted as an integral
part, equal to that of the gene, of the concept of the trait.
Bateson's suggestion that the dominance of a trait represented the
presence of the genetic factor, and recessivity its absence, was
essentially an attempt to save a preformationist concept of the
trait. But it could not be maintained. Examples of traits with
partial dominance, codominance, and other dominance-recessivity
relationships have been recognized.4" Such relationships were
eventually interpreted, exactly as in the case of penetrance and
expressivity, in terms of the role of the environment in the realiza-
tion of the genotypic potential. "Environment," however, had to
be interpreted in the case of dominance-recessivity primarily as
intraorganismal, or even intracellular, environment; and, as it soon
turned out, in order to maintain the reductionist concept of the
Mendelian trait (or "unit-character"),it had to include the concept
of other genes' being the environment of specified genes.
There was, however, a great difference between the interpreta-
tion of dominance and recessivity and that of the other terms
related to the phenotypic expression of the genotypic potential; a
difference that to a large extent has been maintained until today.
In contrast to other concepts and terms that denoted explicitly the
impossibility of mapping the phenotype on the genotype without a
detailed knowledge of the specific environmental conditions under
which an organism developed and lived,4' the notations of domi-
nance and recessivity were never disconnected completely from
the concept of the gene and its alleles. This is the case even
though both R. A. Fisher and Sewall Wright, in their models of
the evolution of dominance as due to the selection of appropriate
"#modifier"genes, viewed dominance essentially as a trait indepen-
dent of the trait it was dominant for. Suffice it to look, for
example, at the standard notation of the mututants of Drosophila
- capital letters for dominant alleles, and small ones for the
recessive alleles - and their "official" description.42 Dominance

reveal the specified phenotype. "Expression" is the degree to which the pheno-
type appears among those individuals where it has penetrated.
40. The discussion about Bateson's "presence-absence" model of dominance-
recessivity continued for some time, at least until N. V. Timofeev-Resovsky
claimed in the 1930s to have proved experimentally that recessive alleles could
mutate back to wild-type alleles.
41. See Lewontin, "Genotype and Phenotype" (above, n. 22).
42. See D. L. Lindsley and E. H. Grell, Genetic Variations of Drosophila
melanogaster(Washington,D.C.:CarnegieInstituteof Washington,1968).
472 RAPHAEL FALK

and recessivity are still considered a dichotomous fundamental

qualitative property of alleles of every discrete gene. I do not
know of a single textbook containing the suggestion, for example,
that one of the two terms should be dropped because they repre-
sent nothing but reciprocal relationships on a single, continuous
scale (100% dominance equals 0O/o recessivity, etc.).43Even in the
descriptions of genes of haploid prokaryotes, such as bacteria and
viruses, the dominance-recessivity relations of a gene are con-
sidered an inherent property of the alleles. Significantly, also at
the level of molecular genetics, when the terms cis- and trans-
dominance were introduced in a totally different context from the
original one, they too denoted the property of a gene to be able to
affect an adjacent gene or physically distant genes, respectively,
rather than an epigenetic concept related to the environment in
which the allele may or may not exert its effect.
This "anomaly" in the status of dominance in the genetic inter-
pretation of the concept of traits seems to me to be the most overt
expression of the ideological, rather than merely the heuristic,
reductionism that has dominated genetic theory since its initiation
in 1900. It is no accident that nearly every philosophical discus-
sion of the applicability of the reductionist thesis of the life
sciences adduces the case of genetics. As we have seen, by decid-
ing to study intraspecific hybrids, Mendel - having no theoretical
qualms - was able to carefully choose traits that were easy to
classify. Notwithstanding dominance, there was no reason for him
to distinguish between traits and the potential for those traits.
However, when modern genetics set out on its way, traits, for
geneticists, were discrete entities by definition. There seemed to
be no hindrance to mapping the genotypic factor from the pheno-
typic trait, once the environmental noise had been taken care of
through statistical means or by careful experimental design. The
unit-character of early genetics was coreferential with the unit that
was inherited. Even when the tortuous path from the gene to a
trait was recognized by Morgan and his school, the pragmatic
heuristic needs of using traits as "markers"for genes in the effort
to elucidate the rules of transmission-genetics prevailed. Thus the

43. The practical aspect of using two different terms of the extremes of a
continuum ("good - evil," "full - empty") in daily language cannot be denied.
Admittedly, this is also the case for "dominance - recessivity." Think, for
example, about the phenotypic relations of the ABO blood types: A and B are
codominant, but A and B are dominant over 0. which is recessive to A and B -
rather than A and B are 1)000%dominant over each other and over 0, whereas 0
is 0% dominant over A and B.
The Dominance of Traits in Genetic Analysis 473

conceptual one-to-one relation between the two was further

entrenched.

ONE TRAIT - MANY GENES

Epistemologically, if genotypes are defined by their phenotypic

effect, phenotypic traits should be defined prior to and indepen-
dently of the breeding results that lead to such identifications.44
Unit-characters were entities by morphological, physiological, or
behavioral considerations; yet it was maintained that, as a rule,
they prove amazingly efficient also as unit-entities in Mendelian
transmission and segregation experiments.45 This was achieved at
the price of invoking - often ad hoc - factors that were sup-
posed to vary with environmental conditions. The unit-character
was upheld also in the face of evidence that the phenotype of even
many simple morphological traits, such as the form of a hen's
comb, must be interpreted as due to two independently segre-
gating genes. It was simply stated that what was considered a
simple trait was, on further empirical evidence, two or three traits.
Thus in spite of Edmund B. Wilson's comment that "the whole of
this apparatus, the entire germinal complex, is directly or in-
directly involved in the production of every character," the reduc-
tionistic pattern was maintained.46
This one-to-one relationship of unit-character and the geno-
typic unit led to a head-on collision when William Castle observed
that in extensive hybridization experiments, such as those between
rats of different hair-pattern, that were followed by selection, the
clean segregation of the unit-character in hybrids was violated.
The progeny of matings between white and hooded (having a
black "hood" on their hair) rats could be selected for various
degrees of whiteness contaminated with black, and vice versa (the
phenotype of the black hooded rats being diluted with white).
Castle suggested that, contrary to Mendel's postulate, when the
alternatives of the unit-character are present in the hybrids, they
may affect each other.47This claim was strongly opposed by H. J.
Muller, who insisted that all the unexplained variation was due to
additional (modifier) genes that contributed to the inheritance of

44. Sec n. 17, above.

45. But see Morgan's comment, in the epigraph to this paper.
46. Edmund B. Wilson, "Some Aspects of Cytology in Relation to the Study
of Genetics," Amer. Nat., 46 (1912), 60.
47. See W. E. Castle, H-eredityini Relation to Evolutioni and Animal Breed-
ing (New York and London: Appleton, 191-1), pp. 123 ff; idem., "Piebald Rats
and the Theory of Genes," Proc. Nat. Acad. Sci. Wash., 5 (1919), 126-130.
474 RAPHAEL FALK

the rats' fur-color. According to the developmental conception of

the genotype-phenotype dichotomy, a clear distinction must be
made between the genes, which were there, and the traits, which
were only an expression of the genes. The phenotypic expression
of one gene could be perturbed not only by external environ-
mental factors, but also by internal factors, such as the function of
modifier genes.4Y This sounded to Castle very much like an ad hoc
explanation.49 And it was: Castle's colleague, Edward M. East,
contented that as many genes should be "invented" (rather tahn
"discovered") as were necessary to explain the data.i0 The genes
are instrumental entities predicted by the geneticist for explana-
tory reasons; they are just methodological devices of the experi-
menter.5'
Once the unit-character was conceived as having no more
reality than the genes if identifies, it was only a question of which
alternative had more convincing explanatory power. A similar
argument had been put forward already in 1902 by George Yule,
who pointed out that the conflict between biometrical and Men-
delian concepts of variation of quantitative characters was re-
solved if one assumed that many factors may cooperate in
determining one character.52 Instrumental reductionism was not
incompatible with conceptual holism. But to no avail. Neither
the advocates of continuous variation in evolution nor those of
discontinuous variation were ready for the kind of explanation
that resolved the conflict more than a decade later.53 Indeed,
Castle's (ad hoc) contention that, contrary to Mendel's conclusion,
unit characters may affect each other's individual identity in the
heterozygotes, was directly related to his acceptance of de Vries's
and Johannsen's conception that evolution occurs only by dis-
continuous variation. Once Castle adopted such a conceptual
framework, his observations of gradual changes demanded addi-
tional means of achieving change by selection, beyond that of de
Vries's mutations - that is, beyond that of segregation without the
interaction of unit-characters or the speculative genes. This was

48. Herman J. Muller, "The Bearing of the Selection Experiments of Castle

and Philips on the Variability of Genes, Amer. Nat., 48 (1914), 567-576.
49. W. E. Castle, "Mr. Muller on the Constancy of Mendelian Factors."
Amer. Nat., 49 (1915), 37-42.
50. E. M. East, "The Mendelian Notation as a Description of Physiological
Facts," Amer. Nat., 46 (1912), 633-655; see p. 635.
5 1. See Falk, "What Is a Gene?" (above, n. 17).
52. G. U. Yule, "Mendel's Laws and Their Probable Relation to Intra-Racial
Heredity," New Phytologist, 1 (1902), 193-207, 222-237.
53. R. A. Fisher, "The Correlation between Relatives on the Supposition of
Mendelian Inheritance," Trans. Roy. Soc. Edinburgh, 52 (1918), 399-433.
The Dominance of Traits in Genetic Analysis 475

provided to him by such experiments as those of asexual propaga-

tion in Daphina (and also in other organisms), in which "among
the offspring developed from the unfertilized eggs of the same
mother Daphina, variations do occur which are heritable, so that
if one selects extreme variants he obtains a modified race."54He
later found that when "the residual heredity J= genetic back-
ground] was equalized, the hooded character appeared substan-
tially the same in the two races," consequently, he was convinced
that "the single gene is not subject to fluctuating viability, but is
stable like a chemical compound of definite composition."55
It is noteworthy that although the unit-gene replaced the unit-
character, even Muller could not rid himself completely of the
unit-character concept of the unit-gene. When he put forward his
explanation of the observation that most wild-type alleles were
dominant over their mutant alleles (in the normal range of envi-
ronments they encounter), it was one of selection of those alleles
that inherently produced an excess of their product thus saving
dominance-recessivity as a mandatory concept of the gene, rather
than of the conditions in which it is found.56
Once the gene became the entity to which geneticists referred,
the classical, commonsense classification into traits, it would
seem, should have become inconsequential for them. However,
insofar as practicing transmission-geneticists tried to follow such a
path, the basic problematic of genetics - that of explaining the
similarity and differences of phenotypes between relatives - was
only augmented. As Richard Lewontin points out,

the explanation or prediction of the hereditary passage of

phenotype requires a three-stage inferential and deductive
process involving both phenotypic and genotypic spaces of
description. Beginning with the phenotypic description of the
parents, the first step is to infer their correct genotypic descrip-
tion. The second step is the deduction of the genotypes of
offspring, given the genotype of the parents and the Mendelian
laws of genetic inheritance. Finally, the genotype description of
the offspring must be mapped back into the space of pheno-
typic description by forward application of the epigenetic rules.
The only unproblematic part of this inferential process is the
second step.57

54. Castle, Heredity in Relation to Evolutiont (above. n. 47), p. 1 7.

55. Castle, "Piebald Rats" (above n. 47). p. 130.
56. H. J. Muller, "Evidence of the Precision of Genetic Adaptation," Harvey
Lect., 43 (1950), 165-229.
57. Lewontin, "Genotype and Phenotype" (above, n. 22), pp. 3-4.
476 RAPHAEL FALK

Efforts to break the circularity imposed by the need to identify

the phenotypic unit-trait by inference from the results of the
genotypic analysis were repeatedly pursued "by inferring geno-
types from phenotypic manifestations that do not have intervening
developmental processes,"" - that is, by approaching as much as
possible the units of heredity proper. The advent of biochemical
genetics, and later, of molecular genetics, appeared for a time, to
provide the necessary immediate genotype-phenotype relation-
ships. George W. Beadle and Edward L. Tatum's slogan of "one
gene - one enzyme" best represents the effort to restore an
independent one-to-one relationship between a phenotype and a
corresponding genotype.59 When the structure of DNA was
elucidated, this slogan was "improved" by Seymour Benzer's
concept of "one cistron - one polypeptide."6"The ultimate solu-
tion of this reductionist effort to achieve complete correspon-
dence between a phenotype and a genotype would seem to have
been the elucidation of the genetic information at the level of the
sequence of DNA base-pairs.6' However, as was soon realized,
at no level is there a context-independent one-to-one relation
between the traits and the genetic determinants.
Thus, the old problem of whether "primary pleiotropism" is at
all possible, which was answered to the satisfaction of reductionist
genetics by a loud "no" in the early days of molecular genetics,
surfaced again. Even after the mapping of one trait to many genes
was accepted, it was not clear whether there is ever a one-to-one
mapping of the gene on a primary phenotypic product. Many
genes were known to be "pleiotropic" - that is, they affected many
traits that superficially were unrelated. Only detailed physiological
and developmental analyses could reveal whether such pleiotro-
pisms were primary or secondary. However, the issue was con-
sidered definitively closed when James V. Neel and his associates
proved that all the symptoms of the highly pleiotropic sickle-cell
anemia syndrome could be ascribed to a single molecular com-
ponent of the hemoglobin unit, and eventually to a single base
change at a specific site in the DNA sequence encoding this
molecule.
Still, findings of the prevalence of alternative splicing patterns

58. Ibid.
59. See G. W. Beadle and F. L. Tatum, "Genetic Control of Biochemical
Reactions in Neurospora," Proc. Nat. Acad. Sci., Wash., 27 (1941), 499-506,
for the exposition of the concept.
60. See S. Benzer, 'On the Topology of the Genetic Fine Structure," Proc.
Nat. Acad. Sci., Wash., 45 (1959), 1607-20.
61. See Falk, "What Is a Gene?" (above, n. 17).
The Dominance of Traits in Genetic Analysis 477

of given DNA sequences, and the different overlapping (some-

times contradirectional) reading frames of such sequences, which
result in different messages being produced by presumably the
same piece of hereditary information, should have provided
reminders to even ardent genetic reductionists that although the
abandonment of the unit-characters for genes (units of heredity)
was an important heuristic step, it was hardly justified to embrace
the inverse route, which ascribes genes to traits. It was, however,
precisely this route that was followed by many geneticists, espe-
cially in reference to aspects of the structure and evolution of the
gene-pool of populations.

ONE GENE - ONE ADJECTIVE62

While Castle's challenge to the one-to-one correspondence

between unit-characters and Mendelian factors was resolved by
his acceptance of East's instrumental interpretation of such a
reduction of traits to genes, he was as reluctant as others to give
up the conceptual reductionist interpretation of the relation
between traits and genes. There were two way by which the
conceptual reduction could be maintained, which may best be
presented by a juxtaposition of Herman J. Muller's reductionistic
philosophy with that of Charles B. Davenport. For Muller, the
genes were the true reality. Genes existed independently of our
detecting them and were the basic units - the atoms - of living
matter. Phenotypic variability was our means of identifying them.
Muller set out on a program to learn the nature of the genes and
used phenotypic traits as mere markers for them.63Not so Daven-
port. For him, the traits were the essential entities, the atoms of
living creatures:

Today biology has to recognize that its individuals are

diverse combinations of units - relatively very numerous
which ... we call unit characters, or we may use the simple
name 'characteristics." Characteristics are thus to individuals
what atoms are to molecules. As the quantities and behavior of
molecules are determined by their constituent atoms, so the
essence of the individuals of any species is determined by its
constituent characteristics. And as we may construct new sub-
stances at will by making new combinations of atoms, so we

62. This phrase was suggested by Prof. Sahotra Sarkar. I am very grateful to
him for this and many other discussions and suggestions.
63. See Falk, "What Is a Gene?" (above, n. 17).
478 RAPHAEL FALK

may produce new species at will by making new combinations

of characteristics.64

If there exist discrete real traits on the one hand, and Men-
delian units of inheritance on the other, then the job of the inves-
tigator is to correlate them: each "real" trait would be correlated
to a finite, small number of genes - genes for that trait - once
the vagaries of "irrelevant" factors that could be summed up
under the heading of "penetrance" and "expression" were taken
care of. Thus the theoretical basis for a manageable eugenics
program was established: "Studies in heredity indicate that every
man is an aggregation of large numbers of certain physical and
mental characters, and that these characters are not reducible to
simpler forms. They are therefore called unit characters; and they
are transmitted through the germ plasm as separable units.
Furthermore, the inheritance of these characters seems to follow
Mendel's law and the presence or absence of desirable and
undesirable characteristics marks the differences in the character
of the men and women about US.65
Such a modified view of unit-traits as the essential entities, the
inheritance of which should be explicated by the Mendelian
model, demanded, however, a definition of the unit-traits in non-
Mendelian terms. Davenport and his colleagues were not choosy
in selecting characteristics, whether morphological, physiological,
or behavioral, that were declared to be traits for which one, two,
or perhaps three genes were responsible. Indeed, any adjective
describing a human phenomenon that Davenport could come up
with was given the status of a unit-trait to which an effort was
made to ascribe one gene, or a small number of genes: "The
modern science of heredity ... seeks as the element of study the
'unit character.' What are unit characters can, however, be told
only by breeding experiments in which the true units reveal
themselves as relatively, if not absolutely, constant, unalterable,
indivisible things. .. . The first step in the resolution of human
traits is, then, a primary rough analysis into fairly simple traits
and, secondly, the study of the behavior of these traits in hered-
ity."66For this purpose The TraitBook was compiled; it includes

64. Charles B. Davenport, "Animal Morphology in Its Relation to Other

Sciences," Science, n.s., 20 (1904), 698.
65. A. Gartley, "A Study of Eugenic Genealogy," Amer. Breed. Mag., 3
(1912), 243.
66. C. B. Davenport, The Trait Book (Cold Spring Harbor, N.Y.: Eugenics
Record Office Bulletin No. 6, 1912), p. 1. (The quotation from Morgan's paper,
The Dominance of Traits in Genetic Analysis 479

a 33-page list of traits, starting with "01 Development - Re-

tarded" and "044 - Physical Beauty," and ending with "94542
Circumcision" and "9656 - Death from child birth."67Of course,
many of the traits thus defined were meant to serve strictly prac-
tical needs. Nevertheless, as Davenport pointed out, "The criti-
cism may be made that not all the traits, especially the diseases,
given here have any hereditary basis. It is not affirmed that this is
the case and yet it can not be denied that all have an hereditary
basis. Even tuberculosis, syphilis and the plague are the product
of a specific germ acting on a susceptible protoplasm and it is this
susceptibility that is the inheritable factor."68Given such a con-
cept of traits and inheritance, it is not surprising to find claims
such as that "'feeblemindedness"was caused by the absence of a
unit-character for normal intelligence. "Genes for traits" were
found also for such conditions as violent temper and the wander-
ing impulse, which were claimed to be transmitted strictly accord-
ing to Mendelian expectations.69
Davenport's (correct) contention that all traits, however de-
fined, have a hereditary basis draws us back to the central
problem of how we interpret the relation between nature and the
way we conceive it. The conception of traits as meaningful entities
by virtue of the existence of genes that affect them, essentially
genes for the traits, which define the traits, has received renewed
attention by claims like that of Richard Dawkins that the pheno-
type is nothing but the vehicle of the genotype to propagate itself.
Dawkins has asserted that "we need not be afraid of postulating
genes with indefinitely complex phenotypic effects," including "the
skill in tying shoelaces," as long as we can detect variability in the

given as the epigraph to this paper, was not aimed specifically at Davenport, or
at eugenicists in general. It is, however, most appropriate here.)
67. Davenport, Trait Book.
68. Ibid., pp. 2-3.
69. Davenport's eugenic convictions clearly gained the upper hand over him.
Only thus can we explain how a person trained as a scientist, and one who
published "solid" experimental results, could publish papers like "Naval Officers:
Their Heredity and Development" (Washington: Carnegie Institution of Wash-
ington, 1919), in which he asserts that "sea-lust is an inhertied, racial trait ...
that is almost wholly a male character," while "nomadism, which leads to a
fondness for travel equally on land and sea, is not rare among women." He goes
even further, to conclude that "thalassophilia acts like a recessive, so that, when
the determiner for it (or the absence of a determiner for dislike) is in each germ-
cell the resulting male child will have a love of the sea," whereas "nomadism
appears to be a simple 'unit character' whose germinal determiner is sex-linked,
i.e., is found only in such sperm cells as produce female offspring."
480 RAPHAEL FALK

traits thus defined.7"'Indeed, it is this sense in which Fred Gifford

hopes that the concept of "genetic trait" can be meaningfully per-
ceived. He suggests that a trait is genetic if it is genetic factors that
"make the difference" between those individuals with the trait and
the rest of the population. Yet, as Gifford notes, what counts as
".genetic"depends not only on the causal processes in the individ-
ual, but also on factors external to it, such as genetic and environ-
mental causal factors shared in the population.7' As Kim Sterelny
and Philip Kitcher remind us, this is not merely a confrontation
with the traditions of nomenclature: Dawkins implies that we can
sensibly speak of alleles having (environment-sensitive) effect.72
It may be the case that in a holistic conceptual context, the trait
"tying shoelaces" could provide a useful way to reduce behavior
to an empirically manageable entity that can be profitably analyzed
by the reductionist methodology, bounded by the limits of cogni-
tive underdeterminancy. Yet even in such a framework it would
probably be meaningful to refer only to the heritability of the
"trait"- that is, to the quantitative measure of the proportional
contribution of genetic factors to "making the difference" - rather
than to "genes for" the trait.73However, the move from the "genes
for P' locution to the claim that selection can fashion P indepen-
dently of other traits of the organism is perennially tempting.74

THE UNIT TRAITS OF TAXONOMY

The notion of the identification of biologically meaningful traits

that may serve as "markers"for some deeper conceptual entities,
is not limited to genetic analysis. A basic ambition of naturalists
has always been the classification and sorting of living beings, by
examining such traits as can provide a mapping of the traits on
natural kinds. It is primarily Linnaeus' conception of systematics
that has prevailed for the last two centuries. Linnaeus believed
that his system of classification fairly represented a god's-eye view
of taxonomy. While the meaning of "natural" in the Linnaen
context was reformulated with the dawn of the Darwinian theory
of evolution, the problem with taxonomy continued to be pri-

70. Richard Dawkins, The Extended Phenotype (San Francisco:Freeman,

1982), pp. 22-23.
71. F. Gifford,"GeneticTraits,"Biol. Phil., 5 (1990), 327-347.
72. K. Sterelny and P. Kitcher, "The Return of the Gene," J. Phil., 85
(1988), 339-36 1.
73. See R. Falk, "BetweenBeanbag Genetics and NaturalSelection,"Bio.
Phil., 5 (1990). 313-325.
74. Sterelnyand Kitcher,"Returnof the Gene,"p. 361.
The Dominance of Traits in Genetic Analysis 481

marily the inability to provide "objective" criteria to identify traits

and to evaluate their significance. As emphasized by P. H. A.
Sneath and R. R. Sokal, "From the time Linnaeus to our own, a
weak point in biological science has been the absence of any
quantitative meaning in our classificatory terms."75To overcome
this deficiency and to provide a satisfactory "natural"classifica-
tory scheme, Sokal and Sneath introduced numerical taxonomy
that is, "the evaluation by numerical methods of the affinity or
similarity between taxonomic units and the employment of these
affinities in erecting a hierarchic order of taxa."76In analogy to
geneticists' attempts to map the genotype on the phenotype,
taxonomists wish to map the common line of descent based on
evidence from resemblance and from homologous characters.
However, their "Iclonclusions on homologies are often deduced
from phylogentic speculations. Thus taxonomists often reason
facilely back and forth among these criteria.... Once it is looked
at critically, it becomes evident how much of taxonomic proce-
dure is circular reasoning and extrapolation."77This should sound
only too familiar to the student of attempts to map genotypes on
phenotypes, or vice versa. To overcome such circularity it was
axiomatic for Sokal and Sneath that a priori "all kinds of char-
acters are equally desirable: morphological, physiological, etho-
logical, and sometimes even distributional ones"; with such an
approach they hoped that the "strict separation of phylogenetic
speculation from taxonomic procedure" might be maintained.78
The difficulty, of course, was still the definition of traits and
characteristics and the method by which they were selected.
Therefore, Sokal and Sneath introduced the concept of "classifica-
tion from above" for making a preliminary selection of specimens.
Using this method they would avoid grouping a fly with a lizard,
or for that matter a stone with a chicken; more importantly, they
would avoid "gross mistakes":"penguins will not be excluded from
a study on birds because they cannot fly, nor will bats be included
because they do fly."79 Thus these authors introduced their
hypotheses as to kinds and principles of classification from the
beginning. To embark on the task of recognizing the basic units of
information for the study, they even introduced the concept of the

75. P. H. A. Sneath and R. R. Sokal, Numerical Taxonomy (San Francisco:

Freeman, 1963), p. vii; and R. R. Sokal and P. H. A. Sneath, Principles of
Numerical Taxonomy (San Francisco: Freeman, 1973), p. xi.
76. Sokal and Sneath, Principles of Numerical Taxonomy, p. viii.
77. Ibid., p. 7.
78. Ibid., pp. 50-55.
79. Ibid., p. 60.
482 RAPHAEL FALK

unit-character,which was defined as "an attribute possessed by an

organism about which a statement can be made, thus yielding a
single piece of information."'81' But all this was to no avail, for the
basic problem remained: the taxonomic bits of information, the
traits, must eventually be identified with, or mapped on, the
genetic information.
Nevertheless, despite the conceptual problem of the impossi-
bility of objectively characterizing a unit-character, Sneath and
Sokal's methodological notion represented a breakthrough for
taxonomy, just as Mendel's methodological notion of treating
distinct discrete traits was a breakthrough for genetics. In the
rewritten version of their book, ten years later, Sneath and Sokal
accept the inability to "objectively" define unit-characters. They
contend that "[t]he success of the intuitive approach of the past Ito
taxonomyl lay in the ability of the mind to recognize swiftly,
though inexactly, overall similarity in morphological detail."'"
Numerical taxonomy merely made possible a quantitative analysis
of this "intuitive approach." The effort to objectively map one
level of theory on another depends on these levels themselves
having some objective reality, and the unit-character of taxonomy
has as much reality as that of genetics. Indeed, in the 1973 version
of their book Sneath and Sokal introduce a more theoretical
consideration in the choice of traits, even for the allegedly purely
phenetic taxonomic construction: they turn to "classes of bio-
chemical compounds ... [whosel structures are highly complex,
and, if known, yield a great deal of information. The nucleic acid
coding of the genome is one example, and protein sequences is
another."82Just as the mapping of the genotype on the phenotype
can be approximated only when the developmental gap between
them has been eliminated or narrowed, so it is with numerical
taxonomy, which becomes a meaningful classification only when it
relates directly to the trait of interest, the genome (and, to some
extent, to its immediate product, the polypeptide sequence). When
the heuristic of numerical taxonomy arrived at a sufficiently direct
mapping of traits on phyletic landmarks it became a prosperous
scientific discipline of its own, molecular evolutionary taxonomy.

W. V. Quine, in his analysis of natural kinds, notes that "there

is nothing more basic to thought and language than our sense of
similarity; our sorting of things into kinds." Yet, there is nothing

80. Sneath and Sokal, Numerical Taxonomy, p. 72.

8 1. Sokal and Sneath, Principles of Numerical Taxonomy, p. xii.
82. Ibid., p. 91.
The Dominance of Traits in Genetic Analysis 483

more dubious than "the scientific standing of a general notion of

similarity, or kind." When one tries to define kinds in terms of
similarity, one soon discovers that to define kinds without simi-
larity is impossible. Thus, concludes Quine, "Definition of simi-
larity in terms of kind is halting, and definition of kind in terms of
similarity is unknown."83Traits are a kind of kinds that should
convey some notion of similarity, even though there are no traits
beyond those that we invent as products of our creative mind.
Notwithstanding, the heuristic of dividing nature into distinguish-
able entities is an essential step in any meaningful scientific
attempt to explain and manipulate nature. It is also indispensable
however, to realize that such "carving of nature at the joints" is a
simplifying device for pragmatic considerations. As noted by
Daniel C. Dennett," Itihe ubiquitous practice of using idealized
models is exactly a matter of trading off reliability and accuracy of
prediction against computational tractability."84
When Gifford suggests that the adherence of scientists to the
concept of "genetic traits" may reveal "insights concerning causal
and explanatory concepts more generally, more broadly than just
in the realm of genetics," he offers three aspects in which the
concept of "genetic traits" may be envisaged as having a useful
role: (a) in the controversies over the relative contribution of
genes and environments; (b) in the causal relation between the
genotype and the phenotype, thanks to the genes' "certain con-
creteness and stability in comparison to other biological causal
factors"; and finally, (c) in perspectives such as those of the
"relationship between causation and responsibility," which may
emerge "as technological advances increase our ability to control
genes" - that is, in deliberations of the sociological and cultural
aspects that theories of the material and functional nature of living
systems may have (presumably assuming the latter's independence
of the former)."SHowever, essentially in accord with the present
analysis, Gifford reveals that the discrete "genetic trait" concept as
an explanation for the first two aspects in adequate, and he shows
that it is simply socially irresponsible for the third aspect. Any
attempt to explicate a genetic or an environmental cause for a
trait is doomed to fail because, with few exceptions (such as the
traits chosen by Mendel), we cannot sort out the cause from all
the factors that shape the trait, whether genetic or environmental.

83. W. V. Quine, 'Natural Kinds," in idem, OntologicalRelativity atid Other

Essays (New York: Columbia University Press, 1969), pp. I 14-1 38.
84. D. C. Dennett, 'Real Patterns," J. Phil., 88 (1991), 36.
85. Gifford "Genetic Traits" (above, n. 71), p. 328.
484 RAPHAEL FALK

And attempts to redefine the traits (or the environment) so that

such causative relationships would be possible are begging the
argument.
The snag is that there is not, nor can there be, a syllogism to do
science, not even for such a fundamental concept as a trait. This is
a difficulty for anyone who wishes to construct science as a true,
objective description of the world as it is. All our theories and
conceptual frameworks are firmly anchored in our cultural history
and background. This is the problem, but also the challenge.
Mendel's introduction of the notion of the discrete trait into
genetics was, and has remained, a crucial tool in the development
of genetic research. It has also, however, been the basis for a more
comprehensive conceptual reductionist framework in the life
sciences. I full-heartedly share Gifford's belief (and conclusion)
that a central lesson to be learned from reflecting on the attempts
to explicate what a genetic trait is is that, at least so far as our goal
is the understanding of the causal origin of traits in individuals,
"we should abandon or de-emphasize the question of whether or
not a trait is genetic.''86

86. Ibid.