GIFT OF

BIOLOGY

LIBRARY

G

THE CATEGORIES OF VARIATION

PROFESSOR S. J. HOLMES

[Reprinted from THE AMERICAN NATURALIST, Vol. XLIII., No. 509, May, 1909.]

[Reprinted from THE AMERICAN NATURALIST, Vol. XL1IJ., May, 11)09 )

THE CATEGORIES OF VARIATION

PROFESSOR S. J. HOLMES

UNIVERSITY OF T&iaeassstN <L/)t-tforNt A

IT is a well-established fact that what are commonly
called variations include modifications of quite different
import in relation to the process of evolution. Whether
or not the variations that are induced in the soma, either
by its own activities or through the influences of the en-
vironment, have any effect in shaping the course of evolu-
tion as they were held to do by Lamarck and his followers,
it is evident that they do not count in this process in the
same way as variations that arise in the germ. But
among the germinal variations themselves there are
classes of unequal significance. Variations differ mark-
edly in regard to their stability or permanence. Many
variations after their first appearance persist with little
modification for an apparently indefinite time., Of these.
what are commonly called mutations afford conspicuous^
examples; these are abrupt variations which breed true.
or nearly so from the start, having their own fluctuating
variability, to be sure, but around a mean which does not
approach that of the parental type in successive genera-
tions. Other variations behave quite differently. They
may be selected generation after generation, modifying
the stock up to a certain point, after which, if the variety
is left to itself, there is revision towards the original par-
ent. It is held by many "Chat these two classes of varia-
tions are fundamentally distinct, and that only the first,

257

258 THE AMERICAN NATURALIST [VOL. XLIII

so-called discontinuous variations, play an important role
in the origination of new species.

A considerable proportion of what is described as
fluctuating variability is, in many cases, simply somatic
variation, having no relation to the germ plasm. It is
evident, however, that all fluctuating variability can not
be such, otherwise species could not be modified by ordi-
nary methods of continued selection. Our mathematical
curves represent two kinds of variability lumped together
and which it is in most cases practically impossible to
separate. The character of height, for instance, in
human beings is to a certain extent an inherited one, but
it is determined to a marked degree by influences opera-
ting after birth. The usual curves of variation represent
both and may even include also variations in the nature
of mutations which fail of discrimination from the rest
of the aggregate.

De Vries distinguished three kinds of germinal varia-
tions, elementary species, retrograde varieties and fluctu-
ations. These three kinds he conceives to be sharply
distinguished and produced in different ways. /All con-
genital variability is regarded by him as resting upon
qualitative or quantitative changes in the pangens or the
organic units of which he conceives living matter to be
built up^The pangens form the basis of the unit char-
acters, or independently variable elements of the organ-
ism, there being a special kind of pangen for each such
character. Variations in the number of pangens cause
variations of the fluctuating type which obey Quetelet's
law of chance frequency distribution. I De Vries main-
tains and attempts to prove by the citation of several
examples that through the selection of such variations
modification may be carried to a certain point, but soon
a limit is reached beyond which selection is incompetent
to effect further improvement. Moreover, continued
selection must be practised in order to maintain the con-
dition which has been reached, else the stock will in the
course of a few generations revert more or less completely
to the ancestral mean. I

No. 509] THE CATEGORIES OF VARIATION 259

Retrograde varieties, according to De Vries, are
sharply distinguished from fhictuationst They are, as a
rule, constant from the start, and differ from the type
in only one or at most a very few respects.

They originate for the greater part in a negative way by the ap-
parent loss of some quality and rarely in a positive manner by acquir-
ing a character seen in allied species." " By far the greatest part
of the ordinary garden-varieties differ from their species by a single
sharp character only. In derivative cases, three or even more such
characters may be combined in one variety, for instance, a dwarfed -
variety of the larkspur may at the same time bear white flowers or even
double white flowers, but the individuality of the single characters is
not in the least obscured by such combinations.

These varieties, says De Vries, "do not possess any-
thing really new." The loss of a character is merely
apparent. "On a closer inquiry we are led to the as-
sumption of a latent or dormant state. The presumably
lost characters have not absolutely, or at least not per-
manently disappeared. They show their presence by
some slight indication of the quality they represent, or
by occasional reversions. They are ..„ ilot wanting, but
only latent. ' ' In other words^tj^eTonly difference be- '
tween retrograde varieties an^Rhe types i§ the latency
or patency of certain charJicters. The same kinds of
pangens are present in the^erm plasm of both.

Elementary species, on the other hand,

are distinguished from their nearest allies in almost all organs. There
is no prominent distinctive feature between the single forms of Drab a
verna, Helianthemum or of Taraxacum; all characters are almost
equally concerned. The elementary species of Draba are characterized,
as we have seen, by the forms and the hairiness of the leaves, the
number and height of the flower stalks, the breadth and incision of the
petals, the forms of the fruits, and so on. Every one of the two
hundred forms included in this collective species has its own type,
which it is impossible to express by a single term. Their names are
chosen arbitrarily. Quite the contrary is the case with most of the
varieties, for which one word ordinarily suffices to express the whole
difference.

The most important distinction which De Vries draws
between retrograde varieties and elementary species is
a physiological one. They

260 THE AMERICAN NATURALIST [VOL. XLIII

behave in quite different manner, when subjected to crossing experi-
ments, and the hope is justified that some day crosses may become
the means of deciding, in any given instance, what is to be called
species, and what variety on physiological grounds.

When varieties are crossed with the parent type the
character that is active in one or the other of the forms
will usually be patent in the first generation of offspring.
In the second generation there is, according to De Vries,
a segregation of characters which takes place in con^
formity to Mendel's law. Ordinary sugar corn, for ex-
ample, differs from the usual type in having a part of
the starch replaced by sugar in the kernels, which conse-
quently become wrinkled when dry. When these two
forms are crossed the active character of starchy kernels
is present in all the members of the first generation. In
the second generation there is a segregation of these
characters, one fourth of the offspring being wrinkled
kernels, and three fourths smooth ones. Approximately
one third of the latter produce only smooth kernels in
subsequent generations, while the other two thirds split
up again in the expected Mendelian ratio.

In the crossing of varieties it is possible, according to
De Vries, for all the corresponding characters of the two
forms to become paired. As the distinguishing feature
of the variety is the latency or patency of one or more
characters, these characters "will unite as well as though
they were both active or both dormant. For essentially
they are the same, only differing in their degree of ac-
tivity. From this we can infer that, in the crossing of
varieties, no unpaired remainder is left, all units com-
bining in pairs exactly as in ordinary fertilization." As
the varieties differ only in the dominance or latency of
certain characters, offspring obtained through crossing
them differ only in the same way. For such unions De
Vries gives the name "bisexual crosses," inasmuch as
there is "complete bisexuality, all unit characters com-
bining in pairs. ' '

In the crossing of the elementary species of (Enothera
De Vries found that in the first generation there was a

No. 509] THE CATEGORIES OF VARIATION 261

splitting up of the progeny in various ratios, but that the
second and subsequent generations bred true to type,
thus presenting a condition just the reverse of Mendelian
inheritance. For instance when the mutant rubrinervis
was crossed with the parent type Lamarckiana the first
generation of hybrids were either rubrinervis or La-
marckianas, the proportion varying greatly in different
lots. The two kinds of hybrids did not split up in the
second generation but bred true to type. Similar results
were obtained by crossing several other elementary
species of Lamarckiana but this kind of behavior does
not seem to be generally characteristic of the elementary,
species of other forms.

In the crossing of elementary species there is, accord-
ing to De Vries, one unit character which is -not mated,
since

the differentiating mark is present in one of the parents and not in
the other. While all other units are paired in the hybrid it is not.
It meets with no mate and must therefore remain unpaired. The
hybrid of two such elementary species is in some way incomplete and
unnatural. In the ordinary course of things all individuals derive
their qualities from both parents; for each single mark they possess
at least two units. Practically but not absolutely equal, these two
opponents always work together and give to the offspring a likeness
to both parents. No unpaired qualities occur in normal offspring;
these constitute the essential features of the hybrids of species and
are at the same time the cause of their wide deviations from ordinary
rules.

These differences between variations were predicted
by De Vries on the basis of his pangen theory, and in his
essay on ' i Intracellular Pangenesis" published in 1889
he expresses the opinion that fluctuating variability which
rests upon numerical variation of the pangens plays but
a minor part in the modification of species. The "art-
bildende" or species forming variability, is dependent
upon the appearance of an entirely new kind of pangen.
When categories of variation are anticipated a priori on
the basis of a theory of the constitution of living matter
there is naturally produced a temptation or bias towards

262 THE AMERICAN NATURALIST [VOL. XLIII

reading the classification into nature and to the over-
looking of transitional stages, and we shall therefore en-
quire if the distinction which is made between elementary
species and varieties is a valid one.
\ In the first place, there does not seem to be any very
good reason why on the pangen theory elementary species
should differ in numerous characters from the parent
form. A pangen is the basis of a single unit character.
Elementary species are produced through the origination
of a new kind of pangen. / If the becoming latent or
dominant of a pangen affects only one unit character of
an organism, it is not evident, when a new kind of pangen
is produced, why the whole organization of the plant
should be so profoundly influenced. Why should not the
awakening of a dormant pangen produce as great a
change as the production of a new pangen of a somewhat
different quality. Says De Vries :

There can be little doubt but that all the attributes of every new
species are derived from one principal change. But why this should
affect the foliage in one manner, the flowers in another and the fruits
in a third direction, remains obscure. To gain ever so little insight
into the nature of these changes, we may best compare the differences
of our evening primroses with those between the two hundred ele-
mentary species of Draba and other similar instances. In doing so
we find the same main feature: the minute differences in nearly all
points.

De Vries nowhere gives us a much clearer explanation
as to why elementary species and varieties should differ
in this way and we must probably be content with re-
ferring the matter to different degrees of "correlation."

It is evident that there are allied groups separated by
small differences throughout the entire organization, and
there are other groups which differ apparently in single
characters only, such as the presence or absence of hair,
spines or certain colors. Hornless cattle and six-toed
cats do not seem to present any general or constitutional
differences from the other members of their species, but
this is a subject upon which we should exercise great
caution, as very slight differences in the rest of the organ-

No. 509] THE CATEGORIES OF VARIATION 263

ization may be correlated with pronounced differences
in a single part.

Among the forms arising by mutation from (Enofhera
lamarckiana De Vries distinguishes three varieties,
lavifolia, brevistylis and nanella. These forms are called
varieties instead of elementary species because they dif-
fer from the type in a few characters only and because
of their different behavior when crossed. But even ac-
cording to De Vries ' own description the points of differ-
ence are not limited to a single character. Lcevifolia, for
instance, is

chiefly distinguished from Lamarck's evening primrose by its smooth
leaves, as its name indicates. The leaves of the original form show
numerous sinuosities in the blades, not at the edge, but anywhere be-
tween the veins. The blade shows numbers of convexities on either
surface, the whole under surface being undulated in this manner. It
lacks the brightness of the ordinary evening-primrose or (Enothera
biennis. These undulations are lacking or at least very rare on the
leaves of the new Icevi folia. Ordinarily they are wholly wanting, but
at times single leaves with slight manifestations of this character may
make their appearance. They warn us that the capacity for such
sinuosities is not wholly lost, but only lies dormant in the new variety.

The leaves of Icevifolia are also ' ' a little narrower and
more slender than those of the Lamarckiana." But
Icevifolia also shows differences in the flower. "The
yellow color is paler and the petals are smoother. Later
in the fall, on the weaker side branches these differences
increase. The Icevifolia petals become smaller and are
devoid of the emargination at the apex, becoming ovate
instead of obcordate."

Brevistylis is characterized by its short style. The
stigma is different in shape from that of the parent form ;
there are differences in the ovaries, and there are only
a few seeds produced. These differences maj possibly
depend upon a single varying character, although

the leaves of the 0. brevistylis are more rounded at the tip, but the
difference is only pronounced at times slightly in the adult rosettes; but
more clearly on the growing summit of the stems and bitches. By
this character the plants may be discerned among the others, some
weeks before the flowers begin to show themselves.

264 THE AMERICAN NATURALIST [VOL. XLIII

Nanella is a dwarf plant, but it is not distinguished by
its smaller size alone.

From its first leaves to the rosette period, and through this to thd
lengthening stem, the dwarfs are easily distinguished from any other
of its congeners. The most remarkable feature is the shape of the
leaves. They are broader and shorter, and especially at the base they
are broadened in such a way as to become apparently sessile. The
stalk is very brittle, and any rough treatment may cause the leaves to
break off. ... The stems are often quite unbranched, or branched
only at the base of the spike. Strong secondary stems are a striking
attribute of the Lamarckiana parent, but they are lacking, or almost
so in the dwarfs.

So far as morphological evidence is concerned, the dif-
ference between the above forms and elementary species
are not so sharp as to inspire one with much confidence
in the essential distinctiveness of the two classes. All
of these so-called varieties differ in various parts of their
organization. It may be said that these differences are
dependent through correlation upon the variation of a
single character, but if any one maintains that smooth
leaves and paler flowers, or small size, brittle stem and
short leaf stalks are related in this way, the burden of
proof is on his shoulders. If a half dozen characters in
different parts of the plant vary it would indeed be diffi-
cult, amid a considerable amount of fluctuating varia-
bility, to separate on morphological grounds a retrograde
variety from a true elementary species, especially since
experts are sometimes troubled in distinguishing some
of the elementary species from one another. Indeed, De
Vries admits that it is often very difficult to decide
whether a given form belongs to one or the other of these/
two groups, but he states that in such cases we have a
means of testing the matter experimentally by the forma-
tion of crosses. Let us see, therefore, how the test of
crossing works out.

In the case of the varieties of (Enothera lamarckiana
there is in the second generation a splitting according to
the Mendelian ratio when the variety is crossed with the
parent form, but with varieties of other forms this does

No. 509] • THE CATEGORIES OF VARIATION 265

not seem to be an invariable occurrence. Davenport has
shown that albinism in poultry is in some cases a non-
Mendelian character, and the same is probably true, ac-
cording to Castle, for the lop-eared condition in rabbits.
In the inheritance of long and short hair in guinea pigs
there is also a marked departure from Mendelian ratios.
In silkworms Kellogg has shown that while most features
are Menaelian, cocoon color in some cases follows Men-
delian ratios, but in others it proved to be " inconstant
as to dominance and recessiveness and numerical propor-
tions, and may even break down and blend." Deaf-
mutism also refuses to come under Mendelian categories
according to the statistics compiled by Bell. The fore-
going are cases of apparent retrograde variations which
form an exception to Mendel's law, but it must be ad-
mitted that the majority of such variations which have
been investigated show a fair approximation to Men-
delian ratios.

In the crosses between the elementary species of
(Enothera lamarckiana there is commonly a splitting
up in the first generation with absence of splitting in the
second and subsequent ones. Hybrids of 0. Lamarckiana
and 0. biennis, however, have nearly the aspect of the
latter species and remain true in the second and subse-
quent generations without reversion or splitting. Crosses
between 0. muricata and 0. Mrtella produced hybrids
showing the characters of both parents. These were
propagated through four generations and remained "true
to /this type, showing <&nly slight fluctuations and never
reverting or segregating the mixed characters." In-
stances of constant hybrids between different species are
very common and it is unnecessary to specify them here.
Such constancy according to De Vries is ' i one of the best
proofs of unisexual unions" or unions between distinct
elementary species in which there is always one unpaired
pangen.

The attempt to make a general rule for the hybridizing
of elementary species leads to many difficulties. In

266 THE AMERICAN NATURALIST [VOL. XLIII

poultry, as Davenport has pointed out, such characters
as pea and rose comb, extra toes and the presence of
muffs and beards on the head, are acquisitions which
developed since the domestication of the original an-
cestral species. They certainly can not be regarded
as the outcropping of latent characters which are rep-
resented in allied forms, but are in the line of progres-
sive variation and therefore according to<rtKebry" de-

ire

Mendel's law, the extra toes do not seem to come under
-any definite rule of inheritance, and none of them follow
the rule for the hybridization of elementary species.

Consider the forms of the common potato beetle studied
by Tower. These arise suddenly, breed true to type and
differ from the parent form in many characters, some of
which are apparently in the line of progressive evolution.
They seem to be as truly elementary species as the
mutants of (Enothera lamarckiana. Yet when crossed
with the parental type they produce hybrids which in
most cases give a mixed progeny segregating according
to Mendel's law. If we can not call these forms ele-
mentary species there is no way of distinguishing such
except through breeding experiments, and the distinction
De Vries draws between elementary species and varieties
amounts to nothing more than the fact that crosses be-
tween certain groups follow Mendel's law, while crosses
between others do not. There is no correlation between
any structural criterion of species and the criterion
afforded by breeding experiments.

Now when we attempt to make a classification on the
basis of breeding experiments alone we fare little better.
/ With blended inheritance in the first and all subsequent
generations, partially blended inheritance, total resem-
blance of hybrid to one or another parent with or without
subsequent splitting, incomplete segregation of char-
acters, splitting of offspring of hybrids in various incon-
stant and non-Mendelian ratios, and many other irregular

No. 509] THE CATEGORIES OF VARIATION 267

manifestations of heredity, the difficulty of maintaining
a sharp distinction between varieties and elementary
species on the basis of behavior in inheritance is apparent.
Are we to classify a six-toed cat as a variety or an ele-
mentary species? The variation is apparently limited
to a single character and it has therefore one of the marks
of a variety, but the variation is doubtless a progressive
one and not due to an awakening of a latent character, and
hence possesses one of the features of an elementary
species. When crossed some of the offspring of the •
first generation may inherit the variation and some not,
and the same is true for the following generation; but
there is apparently no splitting according to the law of
Mendel. So far as our knowledge goes the situation is
the same in respect to polydactylism in man.

The second volume of the Mutationstheorie, which
seems to have been little read by most expositors of De
Vries, affords several examples of irregular behavior of
the hybrids between elementary species which are very
difficult to classify. Crosses between (Enothera nanella
and 0. rubrinervis, for instance, the one *a retrograde
variety of Lamarckiana and the other a distinct elemen-
tary species, gave very variable results, with splittings
in the first and succeeding generations in very inconstant
ratios, and the occasional production of blends which
bred fairly true. We have here a curious combination
of the characteristics of unisexual and bisexual crosses.

It would not be difficult to bring forward many other j
cases which refuse to fall within the scheme of classifica-
tion propounded by De Vries. There are many kinds
of variations which are inherited in many kinds of ways.
The pangen theory of the celebrated botanist has proved
a deceptive guide and has led its author to do scant justice
to many classes of facts which do not fall in line with it.
Hypotheses about paired and unpaired pangens have de-
termined De Vries 's classification of the different kinds
of variations and profoundly influenced his interpretation
of his extensive and valuable researches. The doctrine

268 THE AMERICAN NATURALIST [VOL. XLIII

of intracellular pangenesis has never received the logical
development that characterizes Weismann's theory of
the germ plasm and is considerably inferior to the latter
as a scholastic production. The explanation it affords
of the alleged distinction between varieties and elemen-
tary species is, as we have seen, practically no explanation
at all. The theory may be consistent with the facts of
Mendelian inheritance and the supposed independent
variability of parts, but why it should lead one to antici-
pate that elementary species differing from the parent
throughout their organization originate by a single sud-
den leap is not so clear. Rather it would lead one to
expect that organisms would be modified, a part here and
a part there, corresponding to the independently variable
elements determined by particular pangens, of which there
are numerous kinds, until the whole was slowly trans-
formed. De Vries, however, is careful to explain that a
single pangen may be responsible for certain characters
found in various parts of the organism, such as the color
of leaves, flower and fruit, and that pangens are supposed
to influence each other's manifestation so that a variation
in a single pangen may have a far-reaching effect. In
a chapter on the association of characters in his recent
book on Plant Breeding a great deal of emphasis is laid
upon the value of a study of the correlation of the dif-
ferent parts of the plant. He says :

We come to the conception of a general interdependency of all
parts, organs and qualities of an organism. They are governed more
or less by the same laws which cause them to undergo corresponding
changes when subjected to the same influences.

It seems to me that the author is here upon treacherous
ground. Through the assumption of manifold correla-
tions De Vries attempts to account for a change in a
single pangen which has to do primarily with one inde-
pendently variable part of the organism producing a
modification of the organism as a whole, but in so doing
he is taking the foundation away from the argument upon
which the justification of the pangen assumption rests.

No. 509] THE CATEGORIES OF VARIATION 269

To the extent that the organism is a whole of interde-
pendent parts, to just that extent it gives evidence of not
being a piece of mosaic work and hence removes the neces-
sity for a hypothesis of discrete germinal units. De
Vries even goes so far as to say that "in order to be
correlated the characters must begin by being independ-
ent entities which through some later means may come
into relation with others. ' ' At one time it is argued that
the existence of pangens is proved by the fact that the
parts of the organism are independently variable; and
now it is said that we must assume that pangens must
exist to account for the parts being correlated; that is,
for the fact that they are not independently variable !

The most salient feature of the mutation theory is that
the process of evolution is conceived to take place by
sudden steps of considerable magnitude. This has been
heralded with eclat as enabling us to get rid of certain
difficulties inherent in the Darwinian theory, such as the
assumed absence of intermediate forms between existing
and fossil species and even the contention that the geo-
logical history does not afford time enough for the process
of evolution as it was formerly conceived to take place.
But it is not correct to say that a mutation is neces-
sarily a large variation. Mutations may be very small
steps, falling far within the limits of ordinary fluctuating
variability, as is especially emphasized by De Vries in
his later writings.

In groups (such as brambles, roses, buttercups, willows and many
others) where large numbers of species are closely allied, the differ-
ences between any two of them become smaller, and the number of
distinct forms increasing, the distinction in the end may become re-
duced to a single differential mark for each two neighboring types.
Such differences must be assumed to be produced each by a single
mutation.

In the light of experiments made at Svalof, De Vries
now concludes that "ordinary varieties of cereals are
built up of hundreds of elementary forms which with few
exceptions have hitherto escaped observation. The high
variability which is commonly attributed to our ordinary

270 THE AMERICAN NATURALIST [VOL. XLIU

varieties of cereals consists only in the differences among
the constituents of the mixtures. ' ' Much of the improve-
ment of grains that was formerly obtained by continued
selection De Vries ascribes to the unconscious selection
of elementary species and the gradual improvement of an
originally mixed stock. Kimpau's rye, a stable race
obtained by gradual selection, is thus accounted for, but
the burden of proof is here on the part of the mutationists.
It is apparently not so easy to test the role of mutations
versus fluctuations in the improvement of species, because
if one should secure a stable race by the usual process
of selection, the mutationist might urge that, after all,
amid the confusion of seemingly fluctuating variability,
there were some mutations which escaped notice, and,
through the unconscious selection of these and their off-
spring, the stock was gradually purified and converted
into an improved stable form.

Where ordinary varieties include "hundreds of ele-
mentary forms," separated by characters which in many
cases are so small that they * * may be scarcely perceptible
to the inexperienced eye," how is one to tell whether he
is dealing with mutations or ordinary fluctuations ? The
latter may be much greater in extent, and, as we have seen,

\J there is no structural criterion by which a mutation may
be recognized. Crossing experiments give us no certain
test and we have therefore to fall back upon the criterion
of stability and class as mutations those variations which

y /breed true from their first appearance. Here the oppor-
tunities for begging the question are excellent. If by
the ordinary process of selection a stable race is pro-
duced we can of course ascribe it to the unconscious choice
of one or more undetected mutations. To be sure, stable
races can be produced only on the basis of stable varia-
tions, and if this is all that the mutation theory neces-
sarily implies its divergence from Darwin's teaching is
not very wide.

If now it should turn out that stability is a matter of
degree the last distinguishing feature of the mutation

No. 509] THE CATEGORIES OF VARIATION 271

theory would be destroyed. This is a question upon
which we are sadly in need of light. Some of De Vries 's
own mutations are, however, quite inconstant and show
a strong tendency to revert to the parent species.
(Enothera scintillans, when self -pollinated, produced from
8 per cent, to 52.9 per cent, of Lamarckianas and 34 per
cent, to 69 per cent, of its own kind and a variable number
of other mutants. 0. elliptica and 0. linearis repeat their
kind in still smaller ratios. The reversion of these
mutants like their origin is sudden, but it shows an un-
stable condition of their germ plasm and it may be ques-
tioned if this reversion is essentially different from the
slower reversion which often follows the cessation of the
selection of ordinary fluctuations.

That the differences between mutations and fluctua-
tions are not so fundamental as the pangen theory implies
is indicated by several facts, some of the most suggestive
of which have been furnished by De Vries 's own experi-
ments. For a number of years De Vries carried on a
series of experiments on the corn-marigold Chrysan-
themum segetum, with the purpose of creating a double-
flowered variety. De Vries chose a garden variety of
this form, grandiflorum, and raised several generations,
selecting the seed each time from heads which contained
1.3 rays florets. After four years of propagation, when
he was satisfied of the purity of the isolated strain, De
Vries began to discard all plants with less than 21 rays
in the terminal head. The selection was continued for
a number of generations when a plant appeared which
seemed to form a promising one for the production of a
double variety.

It was not remarkable for its terminal head, which exhibited the
average number of rays of the 21-rayed race. Nor was it distinguished
by the average figure for all the heads. It was only selected because
it was the one plant which had some secondary heads with one ray more
than all the others. This indication was very slight, and could not
have been detected save by the counting of the rays of thousands of
heads. But the rarity of the anomaly was exactly the indication
wanted, and the same deviation would have had no significance what-

272 THE AMERICAN NATURALIST [VOL. XLIII

ever had it occurred in a group fluctuating symmetrically around the
average figure. On the other hand, the observed anomaly was only
an indication, and no guarantee of future developments.

From this slight indication De Vries selected for three
years more and found that the

average number of rays increased rapidly and with it the maximum
of the whole strain. The average came up from 21 to 34. ... The
largest numbers determined in the succeeding generations increased by
leaps from 21 to 34 in the first year, and thence to 48 and 66 in the two
succeeding summers.

Up to this time, while there was a great increase in the
number of ray florets, there was no trace of doubling, but
in a few of the best heads "the new character suddenly
made its appearance. " If sudden, the step was certainly
a very modest one. A single plant was found in which
careful inspection revealed ' ' three young heads with some
few rays in the midst of the disk." "Had the germ of
the mutation, " asks De Vries, "lain hidden through all
this time? Had it been present, though dormant, in the
original sample seed ? Or had an entirely new creation
taken place during my continuous endeavors? Perhaps
as their more or less immediate result?" It is stated
that "The new variety came into existence at once"—
but when? While certain that a mutation must have
appeared, De Vries is uncertain when and where it ap-
peared. ' ' The leap, ' ' he says, ' i may have been made by
the ancestor of the year 1895, or by the plant of 1899
which showed the first central rays, or the sport may have
been gradually built up during these four years" (italics
mine).

During the next two years improvement by selection
was kept up.

The average number of rays which had already risen from 13 to
34 now at once came up to 47 and 55. . . . The maximum numbers
came as high as 100 in 1900 and even 200 in 1901. . . . Real atavists
or real reversionists are seen no more after the first purification of
the race.

A variety which is pronounced "permanent and con-
stant" was produced whose lower limit of the number

No. 509] THE CATEGORIES OF VARIATION 273

of rays was raised to about 34, "a figure never reached
by the grandiflora parent."

The unbiased reader who has carefully followed the
account of the production of this double flower can
scarcely escape the feeling that the interpretation of the
facts according to the mutation theory is at times some-
what strained. The starting point of the whole process is
the selection of fluctuations. Now and then the selection
of a somewhat more pronounced variation was made ; but
the so-called mutations had very small and weak begin-
nings, and De Vries is uncertain just where they occurred
and even suggests that they may have been built up grad-
ually! The selection of fluctuations seems to have had
the effect of inducing variations of greater stability, if
not greater extent, in the same direction. It is not unrea-
sonable to suppose that the appearance of florets with
ligulate corolla on the disk is due to the same factors
which cause the increase in the number of ray florets, and
the variation may be in reality not so discontinuous as
it seems. In fact we are ignorant of the stability or the
real discontinuity of many of the steps in advance towards
the production of the double flower. If a mutation can
make its appearance as an extra ray floret in one case,
and by the occurrence of two or three ligulate corollas
on the disk of a few flowers of the plant in another, and
if both these characters can be increased by selection until
they reach a stable condition that is far more highly
developed than their original one, the facts do not lend
much support to a theory of the saltatory origin of
species. Rather they would indicate that species have
been formed along lines determined by selective processes
much in the same way as Darwin conceived them to be.

By a similar method of selection Burbank has produced
a scarlet variety of the California poppy, Eschsclioltzia
calif ornica. He noticed one flower with a fine scarlet line
on one petal. From the seed of this plant other poppies
with scarlet lines were produced, but only to a slight
extent. After selection was practised for some years

274 THE AMERICAN NATURALIST [VOL. XLIII

a race of pure scarlet poppies was finally obtained with
no indication of their yellow ancestor. This case is cited
by De Vries as one of mutation, but certainly it required
more than one mutation to bring about the result.1

Discontinuity may often be more apparent than real,
the discontinuous variations in the soma being the out-
come of continuous variability instead of abrupt changes
in the germ. Let us consider from this point of view
the occurrence of digital anomalies such as polydactylism,
cleft hand, etc., which are frequently cited as illustrations
of discontinuous variability. These anomalies are often
strongly inherited, but in most cases which have been
fully studied, the inheritance, while partly alternative,
is not Mendelian. In the race of polydactylous guinea
pigs which Castle has produced and bred for a number
of generations the anomaly appeared in different indi-
viduals in various stages of completeness. The parent
of the group, a male, bore an imperfectly developed toe
on his left hind foot. The extra toe contained a claw and
probably the phalanges, but it was loosely attached and
hung limply down on one side. This male produced 27
young, of which 15 were polydactylous. Of the latter some
had an extra digit on both hind feet, others had it on but
one, and in a few individuals the digit was more fully
developed than in the father. In subsequent generations
the anomaly appeared in very different degrees of de-
velopment, some animals having a fully developed digit,
others having a loosely hanging toe with or without a nail,
while in extreme cases there was only a fleshy bag of skin
without bones or claw which often shriveled up and dis-
appeared a few days after birth. The variation, when
appearing on one side alone more frequently was limited
to the left side (1. 630, r. 589), and when unequal on the
two sides the left one was usually the better developed.
Normal and polydactylous individuals did not segregate
in Mendelian ratios. In some instances the normal condi-

aSee also the experiments of MacCurdy and Castle in relation of con-
tinuous and discontinuous variation in rats. Publications of the Carnegie
Institution, No. 70, 3907.

No. 509] THE CATEGORIES OF VARIATION 275

tion gave evidence of being recessive, but this was not
borne out by many other cases in which crosses be-
tween normal individuals produced polydactylous young.
Crosses of normal individuals both of which were of
polydactylous ancestry yielded a much higher per cent,
of polydactylous young than did crosses in which one
individual came from a normal breed, thus showing a
certain tendency in the blood towards polydactylism even
when it did not manifest itself by any outward mark.
Different males of the same amount of polydactylous
ancestry often showed great variation in the potency
with which they were able to transmit the anomalous
character.

The evidence goes to show that we are dealing here
with a tendency which, whatever may be its basis, varies
continuously and not abruptly, although producing varia-
tions which, taken alone, would naturally be classed as
mutations. The extra toe is a new character, but the
polydactylous breed behaves neither like an elementary
species nor like a retrograde variety. The character fluc-
tuates to the vanishing point and even beyond (as shown
by crossing experiments with individuals of different an-
cestry) and shows varying degrees of fidelity of trans-
mission in different strains. Do we not have a condition
intermediate between the abrupt discontinuous variations
which breed true with great fidelity and ordinary fluctua-
tions ? It might be said that we have to do with a muta-
tion which fluctuates to -an unusual degree, although it
originally depended upon a sudden change in the germ
plasm; but the assertion would have no evidence to rest
upon. It might be said, on the other hand, that the varia-
tion is dependent on the undue activity of some of the
factors of normal development, an expression, for in-
stance of increased growth tendency in the part at a cer-
tain period, and that this tendency is kept from definite
expression until it reaches a certain strength, when it
manifests itself as a sudden variation. This conclusion
is warranted, I believe, not only by the great variability

276 THE AMERICAN NATURALIST [VOL. XLIII

of the anomaly, but by the fact already cited that normal
individuals of abnormal ancestry are more apt to produce
abnormal offspring than are normal individuals of an-
other strain.

The studies of Lewis and Embleton and of Pearson on
the inheritance of split hand and split foot in man yield
results in many respects similar to the preceding. Al-
though the normal condition seems to be recessive, segre-
gation does not occur in Mendelian ratios. Often both
hands and both feet were abnormal, but frequently not
in the same way, and in many cases there were marked
differences in the variations on the two sides of the body.
As Pearson remarks, it is difficult to specify in such cases
what the unit character may be. With this, that or the
other bone present in some individuals and absent in
others and represented in very varying degrees of de-
velopment, the inheritance gives little evidence of definite
units of any kind. What is inherited appears to be a
condition which manifests itself in varying ways and de-
grees and which can not be accounted for by any theory
of the sharp segregation of characters.

Why certain germinal variations are strongly inherited
and others not is a problem of much interest, but the solu-
tion of it may lie, not in the supposed behavior of distinct
morphological entities representing certain parts, but in
the physiological relations of the basis of the variation
to the organized structure of the germ plasm. The sex
cells are organisms as well as the bodies that arise from
them; they have the same capacity for self regulation;
and it is not at all probable that all kinds of variations
that may arise in response to the various influences to
which they are subjected should be retained to the same
degree. Weismann has made the suggestive comparison
between the variations of an organism and the oscillations
of a polyhedron on one of the faces upon which it rests.
If the oscillations are small the body tends to come to
rest in the same situation as before ; if they are larger it
may topple over upon a new face about which it mav oscil-

No. 509] THE CATEGORIES OF VARIATION 277

late as around a new center of equilibrium. Weismann
postulates a self-regulating power in the germ plasm
which keeps numerous minor fluctuating variations from
producing any essential modification of the stock. If,
however, a certain variation forms a new center of stabil-
ity it may be permanent. The various mutants of (Eno-
thera lamarckiana,2 most if not all of which contain the
potency of giving rise to any of the others and which pre-
sent very different degrees of stability, may be due to
more or less stable forms which the germ plasm may as-
sume rather than the creation of new kinds of germinal
units. The stability of a variation may be due, however,
not so much to its extent as the analogy with the polyhe-
dron might lead us to expect as to its kind. Variations
which are physiologically congruent with the organized
structure of the germ plasm form stable races; those
which are not tend to become reduced sooner or later to
the norm through the regulatory activity of this sub-
stance.

The germ plasm may be conceived to exercise, in regard
to its variations, a kind of selective activity which may
manifest itself as a proneness of the organism to vary
along certain lines. It is well known that there are partic-
ular types of variation which crop out independently and
mo¥e or less frequently and may be faithfully perpetu-
ated. Polydactylism, split-hand and split-foot, albinism
and melanism, the appearance of races of hairless animals
and glabrous plants, the development of nectarines from
peaches and peaches from nectarines, the origin of peloric
flowers, etc., have occurred many times in independent

2 It is of course possible that the mutations of (Enothera Lamarckiana
result from the impurity of the stock. The species has been for a long
time cultivated as an ornamental flower, and we have nothing but conjecture
regarding its origin. Should it turn out to be derived from a mixture of
two or more forms the mutation theory would be deprived of some of its
best evidence, but there would still remain a considerable number of muta-
tions from pure ancestry. In (Enothera gigas the interesting fact has
been discovered that there is double the normal number of chromosomes.
Whether this is the cause or the effect or merely one instance of the dif-
ferences between this mutant and the type is unknown.

278 THE AMERICAN NATURALIST [VOL. XLIII

strains. These phenomena may be compared to various
anomalies which take their origin in the soma. The em-
bryo in its development is liable to certain accidents
resulting in the production of teratologies! phenomena
such as hare-lip, double formations, anencephaly and
many others. These anomalies fall into certain classes
and in many cases can be attributed to particular defects
of development. The germ plasm also may be regarded
as liable to certain classes of accidental modifications
which produce heritable variations of more or less clearly
defined types. No one would think of attributing anom-
alies of somatic origin to the development of a new kind
of organic unit. If the same mutation appears time after
time, would it not be more reasonable to suppose that it
arose after the fashion of somatic anomalies than that it
depended upon the creation each time of the same kind
of a new pangen? The fact that mutations can be in-
duced through the influence of the environment certainly
favors such a view. Tower found that in Leptinotarsa
eertain variations or mutations arose repeatedly in inde-
pendent strains and that by subjecting the beetles to un-
usual conditions during the period of active development
of their germ cells the proportion of these sudden varia-
tions could be very greatly increased. The variations
thus produced belonged to a few well-marked types, and
while it would be hazardous to set bounds to the possible
number of mutants the species may produce, it is probable
that the number is subjected to a certain limitation im-
posed by the peculiar organization of the germinal sub-
stance.

The selection of variations by the germ plasm may be
illustrated by some observations of Jennings upon inherit-
ance in Protozoa, In a few specimens of Paramoecium
it was noticed that the body was furnished with a spine-
like excrescence. During fission the spine was trans-
mitted to but one of the individuals, the acquired peculi-
arity not arising on the other. In one case the spine was
transmitted through twenty-one generations, when the

No. 509] THE CATEGORIES OF VARIATION 279

strain disappeared. In other cases the spine was
gradually diminished during successive divisions and
ultimately disappeared. Other anomalies such as crook-
edness, blunt ends, bent tip of body and various abnor-
malities, provoked by artificial mutilation, while persisting
for a variable number of generations, were eventually
regulated out, leaving a normal strain.

In like manner we may imagine that through environ-
mental changes the germ plasm becomes affected by
modifications which in the course of a few generations
become regulated out, thereby causing a reversion to the
primitive stock. Reversion may thus be conceived as but
a manifestation of form regulation. Variations to be
permanent must be accepted by the organized structure of
the germ cells, so that they may be included instead of
excluded by the processes of functional equilibration
to which these cells like other parts of the organism are
continually subjected. The congruity of the variation is
the important thing; whether the variation be large or
small, sudden or slow, is of much less consequence.

After all, it may be asked, granting that variations may
be interpreted in the manner here set forth, do not the
phenomena of Mendelian inheritance, showing as they do
that characters may be separated and combined in many
different ways prove that these characters must be borne
by some sort of units in the germ plasm! This is a con-
clusion which is adopted by a large number of Men-
delians, but, plausible as it seems, it is, I believe, a totally
.erroneous view. In the first place it is open to question
if the assumed purity of the gametes is a fact even in
typical cases of Mendelian inheritance, but, granting that
there is an absolute separation of ancestral tendencies,
it by no means follows that there is any sorting of indi-
vidual unit characters apart from the complex of tend-
encies which make for the production of the organism
as a whole. Hereditary anlagen may perhaps be shuffled
and sorted as wholes, but if the germ plasm were com-
posed of discrete parts representing the unit characters

280 THE AMERICAN NATURALIST [ VOL. XLIII

of the individuals, the result would probably be utter con-
fusion instead of orderly development. We might as-
sume that albinism is dependent upon the peculiar prop-
erties of a single chromosome, that length of hair is de-
pendent upon the constitution of a second chromosome,
that a short tail is associated with a third, and so on.
These characters may not be represented by any kind
of structural element; they may have their basis in the
general chemical constitution of the chromosome and be
produced during development in a purely epigenetic
fashion. Chromosomes probably have their individual
peculiarities of chemical constitution as might be expected
from the fact that the chromatin content of an individual
represents contributions from many different ancestors.
Each chromosome or even a small constituent of the chro-
mosome may have a relation to the inheritance of the
whole body, but the peculiarities of one chromatic element
may dominate in one part, those of another chromatic ele-
ment in another. When albinism is eliminated it does not
mean that this character alone is separated, but the anlage
of an albino organism. Even if it is shown that the num-
ber of separate characters which Mendelize is greater than
the number of chromosomes of the variety, Mendelian
phenomena can be explained on the basis of sorting out
ancestral tendencies as wholes instead of unit characters.
The facts of Mendelian inheritance at present known do
not necessarily give any support to the theory of discrete
bearers of unit characters, or the theory of the inde-
pendent variability of parts as conceived by De Vries and
Weismann. This is a point which, I believe, needs to be
emphasized on account of the uncritical acceptance of
these views by so many writers on heredity and variation.
The presence or absence of certain characters may be
independent of the presence or absence of certain others,
but this fact may very readily be accounted for without
having recourse to a particulate theory of inheritance.
The mixing up and separation of characters in inherit-
ance, far from proving the independent, variabilitv of

No. 509] THE CATEGORIES OF VARIATION 281

parts, is just what renders the proof of this theory ex-
ceedingly difficult. The alleged independent variability
of parts is Weismann's strongest proof of his doctrine
of determinants. If a pit in the e.ar or a white tuft of
hair on the head can be transmitted for several genera-
tions without involving any other change in the organism,
we are forced to assume, according to Weismann, that
there is a small part of the germ plasm varying independ-
ently of the rest which forms the basis or determinant
of this character. But the contention particularly diffi-
cult of proof is that the characters really do appear in
independence of the other parts of the organism. A
variation may conceivably depend upon a general change
in the constitution of the substance of heredity, although
manifesting itself most conspicuously in a single part.
A pit in the ear may be the most obvious sign of the very
slight constitutional differences between two individuals.
It is common to find two closely allied species or varieties
differing markedly in one or two features and much less
conspicuously in numerous other parts of their structure.
Peculiarities of horns are sometimes associated with less
noticeable characteristics of the hair, thus pointing to a
common origin of these features in some general modifica-
tion of the ectoderm which in turn may result from some
change affecting the germ plasm as a whole. Albinism
which is so often cited as a unit character is a peculiarity
of far reaching correlations, being often associated with
impaired sight or hearing, diminished fertility, and even
lessened power of resistance to disease.

To establish the independent variability of parts re-
quires a much closer study of possible correlatiojns than
has yet been made. The task is rendered particularly
difficult by the varied combination and segregation of
ancestral tendencies, which we have just considered. If
we can account for the independence of certain characters
on the ground of combining and sorting ancestral tend-
encies as wholes, one has to disprove the possibility of
applying this explanation of the appearance of a partic-

282 THE AMERICAN NATURALIST [VOL. XLIII

ular variation before the latter can be regarded as giving
evidence of a corresponding determinant. The burden
of proof is on the shoulders of the upholders of the doc-
trine of determinants and it is a far heavier one than the
champions of this doctrine commonly appreciate. Let us
suppose that among the various sets of hereditary tend-
encies that find expression in the organization of an indi-
vidual one should include the production of a particular
variation in a single small part. The constitutional differ-
ences which may go along with this peculiarity and of
which it may be regarded as one expression may be modi-
fied or kept from becoming manifest by other and rival
sets of hereditary tendencies thereby rendering it almost
impossible to detect the correlations that really exist, and
giving the character the delusive appearance of independ-
ence. The question of the independent variability of
parts is a crucial one for the particulate theories of in-
heritance, but it is one. so beset with practical difficulties
that a final answer may not soon be forthcoming.

In the preceding discussion the attempt has been made
to show that the various categories of variations recog-
nized by De Vries and others are not sharply separable
either on morphological grounds or by their behavior
when subjected to crossing experiments. The attempt
was made also to show that neither the facts of variability
nor those of Mendelian inheritance give any support to
the doctrine of pangens, determinants, or other assumed
bearers of unit characters, and that unit characters, as
elements than can enter or depart from the complex of
tendencies that make up an organism probably have no
existence. It is evident that variations differ in their
stability, but the explanation of this fact may lie in the
physiological relations of the variation rather than in
some hypothetical representative unit. Whether the
variations of the discontinuous type have been influential,
in any marked degree, in shaping the course of evolution
is a question upon which we need much more evidence.
Mutations, as we have seen, may be very small affairs.

No. 509] THE CATEGOEIES OF VARIATION 283

About the only criterion by which they may be recognized
is their stability, and even that gives some evidence of
being a matter of degree. No limit has been discovered
to the minuteness of the stable modifications that may
occur, and it may happen that further study will reveal
the comparatively frequent appearance of very slight
variations of this kind. In fact, considerable progress
has even now been made in this direction by the study of
grains ; and the number of more or less stable modifica-
tions that are likely to be discovered threatens to over-
whelm systematists with the labor of naming and describ-
ing them. In many organisms not propagated by self-
fertilization the detection of these small steps is no easy
task and the attempt to describe them all would undoubt-
edly prove a futile effort. Among human beings, for
instance, what are we to designate as elementary species 1
We meet with all grades of differences from well-marked
family traits to those which separate the Caucasian from
the negro. Are we to regard the Hapsburg lip which
was transmitted with fidelity for many generations as
the mark of an elementary species'? It was apparently
a new character and therefore presumably dependent
upon a new pangen or determinant. The Celts, Teutons,
Slavs, etc., differ by more or less constant characters
which are constitutional and not confined to single parts,
and the same may be said of the various subdivisions of
these groups. The Aryan stock to which these groups
belong is separated by still greater differences from the
other subdivisions of the Caucasian race, and the latter
in turn differs still more widely from the negroes and
Mongolians. One has considerable difficulty in disposing
of these groups either as varieties or elementary species.
They can not from De Vries standpoint be considered
the results of fluctuating variability on account of their
constancy even under very varied external conditions.
If the small divisions have arisen by slow changes, as
everything indicates, there is no logical halting place
short of admitting that the greater ones may have done

284 THE AMERICAN NATURALIST [VOL. XLIII

the same. In fact a survey of the racial differences of
man in their varying degrees and kinds and their correla-
tion with geographical distribution shows us pretty
clearly that these differences have been slowly acquired
by the summation of very small variations. These groups
are not related as the so-called retrograde or digressive
varieties are, but they are based on differences in general
constitution affecting the shape of the skull, the character-
istic complexion, the general temperament, and many
other traits too numerous to specify. They may have
arisen by minute, discrete, stable variations, but to call
each step in advance an elementary species seems absurd,
and to talk of the immutability of species still more so.
We gain little by characterizing as elementary species
the small steps of which there may be a dozen or more
separating a German from a Frenchman.

Students of geographical distribution as a rule set
little store by the theory of mutation. The relation of
variation and species-forming to distribution as illus-
trated by the work of Gulick and Hyatt on the Achatin-
nelidae, the Sarasins on the snail fauna of Celebes, of
Plate on the mollusca of the Bahamas, and of many stu-
dents of the mammals, birds and fishes of North America
indicate that the steps concerned in species-forming have
been very modest ones. If sudden mutations of consid-
erable magnitude have been a not uncommon source of
varieties of domesticated animals and cultivated plants
it does not follow that the selection of comparatively
small variations has not been the predominant method
of species-forming in a state of nature.

After fifty years from the publication of Darwin's
"Origin of Species" we are still debating, and more
lively than ever, the central problem of that epoch-making
book; but it is not improbable the views of its sagacious
author will prove more nearly correct than those of most
of his modern critics. Much remains to be done before
the problem is finally solved, and there are few fields

No. 509] THE CATEGORIES OF VARIATION 285

before the investigator that are more fruitful and
alluring.3

3 Mendelism and Unit Characters. — Since this article on the Cate-
gories of Variation was sent in for publication several articles have
appeared whose contents would have been referred to had they been pub-
lished somewhat earlier. One of these is a short paper by W. J. Spillman in
"The Nature of Unit Characters/' in the April number of this journal, in
which an interpretation of so-called unit characters is given which is, in
many respects, similar to my own. Reference to Mr. Spillman 's views would
naturally be expected in an article appearing later than his in the same
journal, so that it may be well to state that the proof of my paper was
returned to the publishers in the first part of February, and that no
modification of the paper has since been made. I am glad to find myself in
agreement with Mr. Spillman at least to the extent that the facts of
Mendelian inheritance do not compel us to adopt a particulate theory of
heredity. That Mendelian inheritance can be explained by the sorting
process which is supposed to take place in the reducing divisions of the germ
cells I feel by no means assured,- there are grave difficulties in the way of
such an interpretation. But if this explanation prove to be the correct one
it would be far from justifying the commonly accepted doctrine of unit
characters with all its evolutionary implications.

Reference might also be made to some recent articles by De.Vries, espe-
cially one on the crosses of Oenothera nanella (Ueber die^Kwillingsbas-
tarde von Oenotliera nanella. Ber. Bot. Ges. 26, p. 667, '08), inasmuch as
the experiments reported tend to strengthen further the contention of the
present writer that no sharp line can be drawn from the results of crossing
experiments between elementary species and so-called retrograde varieties.

UNIV. or

THIS BOOK IS DUE ON THE LAST DATE
STAMPED BELOW

AN INITIAL FINE OF 25 CENTS

WILL BE ASSESSED FOR FAILURE TO RETURN
THIS BOOK ON THE DATE DUE. THE PENALTY
WILL INCREASE TO SO CENTS ON THE FOURTH
DAY AND TO $1.OO ON THE SEVENTH DAY
OVERDUE.

MAY 4~ 1848

IAN 5

LD 21-100m-8,'34

UNIVERSITY OF CALIFORNIA LIBRARY