THE VARIATION OF
ANIMALS IN NATURE

13

a

CL

>3

-2

a>

o

5.

«

— *

«

B>

• -H

"3

£

' P " H

o

fa

—

01

- —

°

■«H

SI

CB

r«

<;

en

Ch

cc

en

o

«D

CO

SJ3

^

W

to

C/3

CO

OS

^^

e-i

J

*#

fa

t *

>-

03

C/3

cfi

Q

a

U

CQ

cS

!-5

fa

— '

o

£H

o
o

>

o

SI

£

, 1

^

u

CS

p

i— i

£

p

1— 1

H

-

<o

^

_*

=+H

o

<J

*r-

5

's

e-i

„ — .

CD

">

PQ

- —

■

-0

tn

• —

> — i

. —

-

(fl

*j

<

c

t»

t»

>

CD

13

^

•-*

4-9

o

e«

■ —1

Cfi

&,

o

-

o

+J

o

JH

O

"£»

U

X

THE

VARIATION OF ANIMALS
IN NATURE

BY
G. C. ROBSON, M.A.

Deputy Keeper of Zoology, British Museum (Natural History)

AND

O. W. RICHARDS, M.A., D.Sc.

Lecturer in Entomology,
Imperial College of Science and Technology

With 2 Coloured Plates and 30 Illustrations in the Text

LONGMANS, GREEN AND CO.

LONDON ♦ NEW YORK ♦ TORONTO

LONGMANS, GREEN AND CO. LTD.

39 PATERNOSTER ROW, LONDON, E.C. 4

6 OLD COURT HOUSE STREET, CALCUTTA

53 NICOL ROAD, BOMBAY

36A MOUNT ROAD, MADRAS

LONGMANS, GREEN AND CO.

114 FIFTH AVENUE, NEW YORK

221 EAST 20TH STREET, CHICAGO

88 TREMONT STREET, BOSTON

LONGMANS, GREEN AND CO.

480 UNIVERSITY AVENUE, TORONTO

First published February 1936

Printed in Great Britain

PREFACE

In 1928 the authors of this work commenced to collect and
arrange data on the variation of animals in Nature. Any
naturalist, particularly the systematist and the student of
geographical distribution, will realise that there are many
methods and subjects of inquiry which might be usefully
adopted in analysing the vast amount of detail which has
accumulated on this subject. We felt, however, after some
time that we could make our analysis most useful if we tried
to show the relation between natural variation and the main
problem of the causes of evolution. We came to the conclusion
that, in spite of the many valuable contributions to this
subject, a review which was both synthetic and critical was
still necessary. The subject has become so complex of recent
years, so many special lines of research have been opened up
and the accumulated literature relevant to the subject has
become so intractable, that a synthesis of the sort we have
attempted is an urgent necessity. To exemplify the need for
such a synthesis we would point out that of the observations,
experiments and theories made by workers a generation or more
aso some have become the matter of text-books and current
biological teaching, some have been neglected and forgotten,
and others again are still the subject of ill-informed controversy.
There is a great need for an overhaul of our heritage of research
and observation and for an exact valuation of much that is
either summarily neglected or accepted without question or
scrutiny of the original publications.

We do not claim that in this work we have produced either
an exhaustive survey or a novel viewpoint which might illu-
minate an old and contentious problem. The method we
have adopted differs very little from that elaborated by
Darwin, though we have tried to formulate the problem in
accordance with the many generally accepted changes which

vi PREFACE

have taken place in biological thought and procedure since
his time. We do not suggest that the attempts at a synthetic
treatment that have been made in recent years are to be lightly
disregarded. To some of these, indeed, we are deeply indebted.
We feel, however, that many new and fundamental questions
are left still unrelated one with another. Moreover, no
adequate attempt has been made to see how far the data
of variation in structure . and behaviour confirm particular
theories of evolution. 1

We are under obligation to numerous fellow workers who
have given time and trouble in assisting us, and tender our
thanks for their generous help and the trouble they have taken
on our behalf.

G. C. R. O. W. R.

1 Owing to the illness of one of the authors the publication of this book was
delayed, and no references to literature later than 1 933 are included.

CONTENTS

Preface ........

.

PAGE
V

Precis of Contents ......

.

ix

List of Illustrations .....

.

XV

Chapter I. Introduction ....

.

i

„ II. The Origin of Variation .

.

18

„ III. The Categories of Variant Individuals

58

„ IV. The Distribution of Variants
Nature .....

IN

77

„ V. Isolation .....

129

„ VI. Correlation ....

160

,, VII. Natural Selection

181

„ VIII. Other Theories of Evolution .

3i7

„ IX. Adaptation .....

348

„ X. Conclusions .....

368

Bibliography .......

377

Index ........

403

4:

PRECIS OF CONTENTS

CHAPTER I

INTRODUCTION

Use of the term ' variation.' The study of variation in nature as opposed
to that of domesticated animals.

Object of the work. Variation and evolutionary problems. Causes of
variation (general). Various other aspects of the problem — mode
of occurrence of variation, frequency, limitations, etc. Individual
variation, groups and special categories contrasted. Variation and
taxonomy. The ' natural population.'

Formulation of the chief problems involved in the study of natural varia-
tions : (i) Causes of variation ; (2) The characteristics of groups and
their mode of occurrence in nature ; (3) The origin and causes of
isolation ; (4) The causes of correlation ; (5) The spread of new
characters ; (6) The relation between variation and the main ten-
dencies of evolution. General remarks on the methods of evolutionary
study.

CHAPTER II

THE ORIGIN OF VARIATION

The three types of variation : (1) Fluctuations ; (2) Effects of genetic
recombination ; (3) Mutations.

( 1 ) Fluctuations. Difficulty of distinguishing between these and heritable
variation by inspection. Plasticity.

(2) The basis of heritable variation. Mutation and the genetical determina-
tion of hereditary characters. Haldane's six modes of genetical
representation.

(3) Recombination. General. Evolutionary value of variation due to this
cause. Crossing between species in nature. Recombination princi-
pally of value in ' trying-out ' new combinations.

(4) Gene-mutations. Their origin (general). Difference between produc-
tion of new mutations and alteration of mutation-rate.

Experimental evidence for alteration of mutation-rate. The alleged
' spontaneity ' of mutations.

(5) ' The inheritance of induced modifications.'' Introductory and historical.
Preliminary difficulties discussed. The difficulty of experimental

x PRfiCIS OF CONTENTS

proof — question of the stock used in experiment, elimination of
selection, etc.

(a) General.

(b) Experiments.

(c) Circumstantial evidence.

(d) Habit-formation.

(e) Summary.

(6) General Conclusions.

CHAPTER III

THE CATEGORIES OF VARIANT INDIVIDUALS

Historical. Various types of categories and the terms used to designate
them. The individual, (i) Taxonomic categories. Use of the
various terms. Taxonomy and population-analysis. (2) Palaeonto-
logical categories. Lineages and bioseries. Palaeontology and neon-
tology. (3) Geographical categories. Terms used to designate
various kinds of groups. Concept of the ' Rassenkreis.' (4) Genetical
and reproductive categories. Various terms employed. The recon-
ciliation of genetical and taxonomic categories. (5) Physiological
categories. General conclusions.

CHAPTER IV

THE DISTRIBUTION OF VARIANTS IN NATURE

Preliminary considerations. Methods of distinguishing heritable variation
from fluctuations apart from experiment. Intermediacy. Variation
and size of area. The chief modes of occurrence of variants.

( 1 ) Individual variation, principal modes of occurrence and examples.

(2) Polymorphism. Discussion of the term. Examples of the phenomenon
in land snails, Lepidoptera, etc.

(3) Geographical variation. Introduction and general discussion as to
whether geographical variation is characteristic of some groups and
not of others. Rensch's views. Conclusions on this subject. Examples
of geographical variation.

(4) Physiological races. Degrees of differentiation. General summary.

CHAPTER V

ISOLATION

Two main kinds of isolation — geographical and topographical separation
and the prevention of sexual intercourse. General discussion on their
effects and interaction. Correlation of the degree of isolation with
that of divergence. Time necessary for the establishment of new
species.

Topographical isolation. General. Capriciousness of endemism on islands.
Difficulty of the problem. Peculiar characteristics of endemic species.
Relation between numerical abundance and rate of evolution.

PRECIS OF CONTENTS xi

The establishment of permanent isolation. Analysis of the various methods
of establishment. Discussion of (a) Seasonal occurrence ; (b) Time
of breeding ; (c) Loss of means of dispersal, etc. (indirect methods) ;
and (d) various bars to intercourse, and (e) various sources of
sterility (direct methods). Conclusions : Importance of biological
races.

CHAPTER VI

CORRELATION

Use of the term ' correlation ' ; Diirken's analysis. ' Physiological ' and
* gametic ' correlation (Graham Kerr). The correlation of specific
characters very variable in degree, probably on the average rather
low. The two fundamental types of correlation (' causal ' and ' coin-
cidental '), their causes and importance. Methods of deciding the
basis of the correlation of specific characters. Segregation and specific
characters in relation to correlation and variation. Highly correlated
characters are probably those for which large populations are homo-
zygous. ' Lineages ' and the correlation of specific characters.
Independence of characters as revealed by the study of ' lineages.'
Difficulty of reconciling the apparent independence of characters in
phylogeny with our conception of development and organisation.
Specific characters as mosaics of fortuitously associated units. Their
incorporation in the general unity of the organism and transformation
of their basis from a fortuitous to a permanent one.

CHAPTER VII

NATURAL SELECTION

Introduction. Darwin's presentation of the evidence. Subsequent
modification and development of the theory. Conditions of proof
required and procedure to be adopted in this work. The origin of
domesticated races and its relevance to the problem of Evolution.
Selection experiments with pure bred stock. Pearl's requirements of
proof that selection has altered the character of a race.

Direct evidence for the selective incidence of death-rates in nature. Twenty
cases of direct observation on the selective incidence of death-rates
examined. Summary of the examination (p. 212). Direct observa-
tion on the alteration of natural populations. Summary (p. 215).

The nature of variation considered in relation to natural selection. The
mutation-rate and survival-value of mutations. Mathematical treat-
ment of the subject. The problem of random mating. Summary
(p. 229).

Indirect evidence for and against the efficacy of selection.

I. Standard cases. Protective coloration. Mimicry.
II. Less intensively studied cases. The adaptations of torrent-living
animals. The colour and pattern of Cuckoo's eggs. The adaptations
of (a) abyssal animals and (b) cave-dwelling animals.

xii PRfiCIS OF CONTENTS

Difficulties raised by the theory, (i) Specific differences in colour and
structure ; (2) The problem of secondary sexual characters ; (3) The
origin of habits ; (4) Complex organs and co-adaptation. General
summary and conclusions.

CHAPTER VIII

OTHER THEORIES OF EVOLUTION

(1) Lamarckism and ' the inheritance of induced modifications.'

(2) Evolution by hybridism. Origin of new characters and character
combinations already discussed (Ch. II). These theories now reviewed
in their wider evolutionary application. Transformation of popula-
tions. Progressive modification.

(3) ' Chance survival.' Survival of non-advantageous mutants. Elton's
theory. The occupation of new habitats. Rapid spread of introduced
species.

(4) Orthogenesis. Use of the term. Historical. Parallel variation.
Main groups of evolutionary phenomena which are treated as ' ortho-
genetic' (a) Normal evolutionary series. Haldane's explanation.
(b) Recapitulatory series : (i) Haldane's explanation. (ii) The
racial life-cycle and the theory of racial senescence, (c) Abnormal
growth. Excessive size of parts distinguished from excessive com-
plexity. Various explanations, (i) The direct adaptive value of
excessive size. Haldane's theory. Certain special cases examined.
(ii) Huxley's explanation (heterogony and selection), (iii) Fisher
and Haldane's theory of the effect of selection of metrical characters
determined by numerous genes, (iv) Theory of an internal momentum.
Conclusion on the various explanations of ' orthogenetic ' phenomena.

(5) Theories involving an internal impulse of a non-physiological nature,
(i) Bergson's theory, (ii) ' Psycho-biological ' theory of Russell and
others, (iii) Smuts's holistic theory. General conclusions.

CHAPTER IX

ADAPTATION

Use of the term. (1) Useful characters. (2) Specialisation. (3) Statis-
tical adaptation. (4) The organismal concept of adaptation. The
adjustment of the organism to environmental pressure : (i) modi-
fication, (ii) compensation, and (iii) independence. Closeness
of adaptation. The conception of optimum conditions. Internal
optima. Optimum density. Self-regulation. Organisation and
development. Organisation and specialisation. Difficulty of ex-
plaining the origin of organisation by random mutations. Initiation
of variation by the organism itself.

PRECIS OF CONTENTS xiii

CHAPTER X

SUMMARY AND CONCLUSIONS

The fundamental divergences in evolutionary theory — the organism as the
product of variation guided by environmental change and as endowed
with an internal momentum. Limitations of our knowledge. The
origin of groups and the production of adaptation are the outstanding
features of evolution. The apparent unitary nature of the evolutionary
process ; are group-formation and adaptation produced by the same
process ? The evolutionary relationship between specialisation and
organisation. Natural selection and the unitary concept of evolution.
Discussion of the evidence for the efficiency of natural selection. The
role of Lamarckism and ' induced ' mutations. The importance of
certain ' orthogenetic ' phenomena. Summary of the main theories
of evolution. The ' spread ' of variants an acid test of evolutionary
theories. Difficulty of regarding organisation as a product of natural
selection.

ACKNOWLEDGMENTS

For kind permission to make use of illustrations from already
published works the authors are indebted to : —

The Director of the Carlsberg Laboratory, Copenhagen, for
Figure i from Schmidt, J., in G.R. Trav. Lab. Carlsberg, 18, 1930 ;

Firma Julius Springer, Berlin, for Figure 2 from Zimmermann
in Zeitschrift fur Morphologie und Oekol. Tiere ;

The American Museum of Natural History for Figures 3,
10, 11, 12 ;

Firma Johann Ambrosius Barth, Leipzig, for Figure 4 from
a paper by Sikora, H., in Arch, fur Schiffs-und Tropenhygiene ;

The United States Government Printing Office for Figure 5
from Mickel, C. E., in Entomological News, 35, and for Figure 22
from Journal of Mammalogy, 7 ;

The United States Department of Agriculture, Bureau of
Biological Survey for Figures 6a, b and c from Howell, A. H.,
in Bulletin, 37 ;

The Carnegie Institute, Washington, for Figures 7, 8, 18
from Crampton, H. E., Studies on the genus Partula in publica-
tions 228 and 410, and for Figure 25 from Lutz, F. E., in publication
101;

Brighton and Hove Natural History and Philosophical
Society for Figure 9 from an article by H. Toms in Report,
1920 ;

The Wistar Institute of Anatomy and Biology, Philadelphia,
for Figures 19, 19A and 21 from the Journal of Experimental
Zoology ;

The Royal Entomological Society of London for Figure 13,
a plate from the Presidential Address, Proceedings, 5, 1931 ;

The Zoological Society of London for Figure 14 from
Ingoldby, C. H., in Proceedings, 1927 ;

The British Museum (Natural History) for Figure 15 from
the ' Handbook of British Mosquitoes ' ;

Messrs. Ernest Benn, Ltd., for Figure 16 from J. R. Norman's
' A History of Fishes ' ;

Professor Nuttall and University Press, Cambridge, for
Figure 26 from Parasitology, 4, 191 1 ; and

University Press, Cambridge, for Figure 30 from Himmer
in ' Biological Reviews,' 7, 1932.

Beneath each illustration is the reference to a source which is
set out in full in the Bibliography at the end of the volume. The
Table on page 97 is translated from Rensch, R. B., in Arch. Naturges.
1, by permission.

LIST OF ILLUSTRATIONS

Coloured Plates

PAGE

I. Mimicry of bees by flies in Brazil. (From Study,
1926) ...... Frontispiece

II. Polymorphism in the moth Acalla comariana Zeller.

(From Fryer, 1928) .... to face 102

Illustrations in the Text
fig.

1. Distribution of average number of vertebrae in the

Atlantic Cod (Gadus callarias L.). (From Schmidt,

193°) 49

2. Correlation of yellow markings with climatic conditions

in the wasp Polistes foederata. (From Zimmermann,

I93 1 ) ; •_ 50

3. Map of distribution of Eumenes maxillosus De G. adapted

from Bequaert (1919) . . . . . .68

4. Body- and head-lice. (From Sikora, 1917) . . . 75

5. Frequency curve of variation in size of male and female

Dasymutilla bioculata Cresson. (From Mickel, 1924) . 80

6. (a) Map of distribution of races of the marmot, Marmota

caligata ........ 83

(b) Map of distribution of races of Marmota flaviventris . 84

(c) Map of distribution of races of Marmota monax . . 85
(From Howell, 1915)

7. Distribution of primary varieties of Partula otaheitana

on Tahiti. (From Crampton, 19 16) . . . 86

8. Comparison of means of colonies of Partula in Tahiti.

(From Crampton, 19 16) . . . . .98

9. Variation in the Pointed Snail in its colonies in Sussex.

(From Toms, 1922) ...... 100

10. Variation in the finch, Buarremon. (From Chapman,

1923) ......... 106

1 1 . Distribution of Buarremon brunneinuchus and B. inornatus.

(From Chapman, 1923) . . . . . .107

u8

122

xvi LIST OF ILLUSTRATIONS

PAGE

F it Distribution of S. American wrens of the Troglodytes
musculus group. (From Chapman and Griscom,

1924) ..■•••* 3

13. Male genitalia of races of Ctenophthalmus agyrtes drawn

on a map of Western Europe to show distribution of
races. (From Jordan, 1931) !I 5

14. African squirrels of the genus Heliosciurus. (From

Ingoldby, 1927) '

15. Respiratory siphons of larvae of Culicella morsitans and

C.fumipennis. (From Lang, 1920) .

16. Scapanorrhynchus owsteni. (From Norman, 193 1) . ' * 3 l

1 7. A group of endemic Hawaiian insects . . • • 1 3 6

18. Distribution of the species of Partula on the island of

Moorea. (From Crampton, 1932) . . . . 138

19. Peromyscus maniculatus. Histograms showing distribution

of frequencies for the various values of relative tail-
length and relative width of the tail-stripe in eight
localities. (From Sumner, 1920) . . • -165
19A. Variation in seven characters in Peromyscus maniculatus
showing general failure of correlation within the race.
(From Sumner, 1920) x "7

20. Specific differences between the queens of Vespa ger-

manica F. and V. vulgaris L l 73

a 1. Individuals of two different clones of Hydra, kept under

similar conditions. (From Lashley, 1916) . ■ 19 1

22. Map showing localities in which Peromyscus pohonotus

albifrons and P. p. leucocephalus were trapped by Sumner.
(From Sumner, 1928) 2 37

23. Leptodirus hohenwarti Schmidt (Silphidae) . . .271

24. Specific characters of the Psammocharidae . . -275

25. Gryllus. Polygons of frequency for ratio of ovipositor to

tegmina. (From Lutz, 1908) 28 4

26. Hypostomes of larval and adult ticks of the genus

Argas. (From Nuttall, 191 1) 286

27. Forelegs of some male Crabronidae . 2 95

28. Horns of Ovis poll (male) 333

29. Oligolectic and polytrophic bees . . • -349

30. Internal temperatures of bees' and wasps' nests. (From

Himmer, 1932) ..-••• 3

THE VARIATION OF ANIMALS
IN NATURE

CHAPTER I

INTRODUCTION

The term variation is generally used in biology to connote
the differences between the offspring of a single mating or
between the individuals or groups of individuals placed in a
single species, subspecies, or race. It is sometimes used in a
more general way to connote, e.g., the differences between
genera and other groups above the rank of species (cf. Pel-
seneer, 1920 ; Gardner, 1925). The former usage, which is
more common and is regularly used in evolutionary, genetical
and taxonomic studies, is the one employed in this work.

A division of the study of variation in animals according
to whether they are living under natural conditions or in
domestication is arbitrary from one point of view. We have
no reason to believe that either the origin of variation or its
mechanism of hereditary distribution is different in any essential
as between wild and domesticated animals. Nevertheless the
various procedures employed in the mating of domesticated
animals have, in the mixing or isolation of hereditary strains,
such different effects from the matings of animals in
nature that the distribution and evolutionary fate of variant
characters in domesticated and wild forms can rarely be
comparable. Whether the study of variation under domestica-
tion has the importance in evolutionary studies that Darwin
originally assumed is very doubtful : but if the study of
variation is to yield any results of value in assessing the causes
of evolution, it should primarily be conducted in natural
populations.

2 THE VARIATION OF ANIMALS IN NATURE

The facts of variation impressed themselves on the early
systematists and the collection and utilisation of such data are
a part of systematic zoology. The analysis of the vast body
of facts thus accumulated and the extraction of general prin-
ciples from them were, of course, stimulated by the work of
Darwin and Wallace and became important items in the
technique of evolutionary studies. Various aspects of the
problem have been dealt with in a number of synthetic
works : Bateson (1894), Woltereck (191 9), Philiptschenko
(1927), Rensch (1929). The origin of variation and its
hereditary distribution has become one of the common-
place matters of biological literature. The majority of the
synthetic works are concerned with the special problems
presented by what is after all a very extensive subject. The
object of this work is, like that of the majority of its prede-
cessors, a special one. It does not set out to review the problem
of variation in all its aspects, but to gather together all the
leading facts and principles that emerge from a study of varia-
tion and have any bearing on the causes of evolution.

On account of the vast numbers of books and papers that
have been produced on evolution, some word of excuse is
perhaps needful in adding to the number. In spite of all
that has been written on this subject and the fresh prestige
which, after a period of intense criticism, the doctrine of
Natural Selection has acquired from mathematical and gene-
tical studies, we believe that the causes of evolution are still
obscure and the relative importance of the presumed causative
agencies is still to be assessed. We further believe that many
principles and much recorded data still need to be worked into
the general scheme of inquiry and that in a number of direc-
tions much more research is still necessary. Even such a sub-
ject as geographical distribution and variation, which might be
thought to be worn threadbare, is still in need of systematic
study.

As we are mainly concerned in this work with the causes
of evolution it may well be asked whether a survey of this
subject based only on zoological data can be of much
assistance. We think that a comprehensive work including
both botanical and zoological data and principles of the kind
brought together here is eminently desirable. At the same
time we do not feel that such conclusions as we have formulated

INTRODUCTION 3

are in any way invalidated because they are based on zoo-
logical data alone. We are concerned with the evolution of
animals and are content to let our conclusions speak for them-
selves. It is very probable that there are certain evolutionary
principles and phenomena that are peculiar either to animals
or to plants. Polyploidy and certain other chromosomal
phenomena seem at present to be almost restricted to the
latter. We do not, however, believe that the truth or falsity
of any theory of evolution is likely to be decided by an
acid test provided by exclusively botanical or zoological
data.

The importance of variation in the study of evolution is
too well known to require much explanation. Whatever we
may hold to be the cause or causes of the evolutionary process,
it is almost invariably recognised that it has proceeded by the
progressive accumulation of changes of the same dimensions
as are found in the variation within a species. In spite of the
considerable changes that have taken place in evolutionary
inquiry, the fundamental idea enunciated by Darwin and
Wallace that evolutionary divergence is the summation of a
series of changes having the status of individual differences
is still almost universally accepted. Students of evolution are
still concerned with the questions — how do such variations
arise and by what means are they amplified so as to give
progressive change in given directions ?

Some measure of variation is of universal occurrence among
all living organisms, and the capacity to display this phenome-
non might be given as one of the attributes of living matter.
It is doubtful indeed whether it is an exclusive property of living
organisms or even of organic compounds (Reichert, 1919) ;
but it is far more marked in them than in inorganic bodies.

The origin of variation is fully discussed in Chapter II.
The most generally accepted view, of course, is that, while
the somatic tissues are readily modified by environmental
factors, heritable variation is due to spontaneous changes at
single loci in the chromosomes (gene- or point-mutations),
to various kinds of chromosomal abnormalities, or to the
combination of maternal and paternal genes. In certain
conditions, however, it seems that mutations may be induced
by environmental factors. Whether this is a correct view and
whether all heritable variation may not in the last resort be

4 THE VARIATION OF ANIMALS IN NATURE

due to modification by the environment will be discussed in
Chapter II. The term mutation is used in the narrow sense of a
change at a single locus (e.g. cf. Hammerling, 1929, p. 1) or
in a less restricted sense for both gene-mutations and the results
of chromosomal abnormality (Morgan, Bridges and Sturtevant,

Though it is quite certain that part of the variation induced
by the action of the environment is not heritable (somatic
variation, 'modification,' 'fluctuation'), such variation is
not to be distinguished by inspection of its visible effects from
heritable variation and it is quite common to find that a given
variation (e.g. in size) is heritable in some cases and non-
heritable in others. The heritability or non-heritability of a
character can be determined only by experiment, and even
the argument as to the status of a given character based on
analogy with other cases in which heritability has been
experimentally proved, is insecure.

Somatic variation is a very widely occurring phenomenon
and is due to a great diversity of environmental factors. It
ranges from minute changes in size, shape and colour to
excessive and ' monstrous ' changes. The causes may be
operative over large areas and whole populations may be
affected by them, or they may be local and operative only in
exceptional circumstances. There is an unfortunate tendency
to use the term ' purely phenotypic ' for such variation, but
' phenotypic ' has a precise and totally different meaning, so
that this usage is undesirable. The term ' Dauermodifikation '
(for which no English equivalent is in common use) has been
given by Jollos and others to temporary and reversible altera-
tions of the hereditary constitution.

In distinguishing between hereditary and non-hereditary
kinds of variation we touch on what is the most important
distinction from the evolutionary point of view. We ought,
however, to remember that hereditary variation may be either
due to the combination and recombination of pre-existing
factorial material or to the introduction of new hereditary
material. Moreover, as is well known to systematists, variation
may be due to the divers combinations into which the characters
of the zygote enter. Thus series of species are known which
represent the permutation and combination of a common stock
of characters, e.g. :

INTRODUCTION

Species a may have the constitution ABCDEF

BCEFGH
ABDEGH

b „ „ „ „ BCEFGH

)) v J> J> )> J5

The nature of variation may be further studied according to
whether we are considering (a) the part of the organism affected,
(b) the extent of the deviation from the norm, or (c) the mode
of its occurrence having regard to (i) its spatial distribution,
(ii) its frequency of occurrence, and (iii) its limitations.

(a) By far the greatest part of our knowledge of variation
relates to the structural characters of animals. Herein it
appears to be practically universal and it affects the size,
form and arrangement of parts and also appears in the form
of meristic as opposed to substantive variation (Bateson)
as well as in the phenomena of homeosis (replacement of one
part by another). It is much open to discussion whether
certain parts or areas of the animal body are more subject than
others to variation. For example, Pelseneer (1920, p. 409)
holds that ectodermal derivatives are more subject to variation
than those derived from the other germ-layers. This opinion
has been combated by Robson (1928, p. 48).

Variation is also seen in the various functions and activities
of animals. Our knowledge here is more scanty and in need
of systematisation : but there is ample evidence, e.g. from the
data in ' Tabulae Biologicae ' and such a work as Winterstein's
' Vergleichende Physiologie,' that variation occurs in the
majority of the vital activities and their products. It is hardly
necessary to state that variation in ' performance ' is a familiar
phenomenon in applied genetics. Finally, there is evidence
of very considerable variation in habits, food- and habitat-
preferences and similar activities.

(b) It was originally customary to draw a distinction between
continuous and discontinuous variation. The former were held to
consist of the slight differences found between individuals,
even when they are of identical genotypic constitution. The
latter were the clearly marked and uncommon variations
sometimes alluded to as ' sports.' Genetical study has tended
to minimise the importance of this distinction. Originally
held to be distinct in kind the first were thought to be non-
heritable, the latter to be heritable (' mutations ' of de Vries).
It is now realised (cf. Chapter IV) that there is no essential

6 THE VARIATION OF ANIMALS IN NATURE

difference between the two ; both marked and slight variations
are known to be heritable.

(c) (i) Variant individuals are not distributed in space at
random and in a chaotic fashion. In the first place there is
a very marked correlation (more marked in some groups of
animals than in others) between the ecological background
and the type of variation, which is one of the most obvious
effects of the susceptibility of the living organism to its environ-
ment. There is also a tendency for variant individuals which
demonstrably do not owe their peculiarities to their environment
to be distributed in certain specific ways. The most familiar
example of such distribution is the geographical race.

(ii) The frequency of heritable variation is one of the most
important topics of modern evolutionary study. It is now
generally agreed that gene-mutations are of the greatest im-
portance, as they are regarded as the only source of new
evolutionary material. It is usually stated that they occur
very infrequently, and this conception is of prime importance
in the modern statement of the theory of Natural Selection
(Fisher, 1930 ; Haldane, 1932). How true this conception
is it is impossible to say, as the subject has only been intensively
studied in two species kept in artificial conditions. However,
it is desirable to keep in mind the probability that the very
great profusion of variation among animals in nature is due
mainly to somatic differences and factorial recombinations.

This conception has introduced a rather different outlook
on the role of Natural Selection. Darwin in no place in ' The
Origin ' or any other of his works, as far as we know, committed
himself to any pronouncement as to the frequency of heritable
variation. He repeatedly insisted indeed on the slowness of
the selective process. This we imagine was due to his belief
in ' blended inheritance ' and his realisation of the smallness
of the individual steps and the comparative infrequency of
serviceable ones, rather than to any idea of the infrequency
of any heritable variation. Nevertheless he conveys the
distinct impression that he thought that the stock of heritable
variation was plentiful. We are now confronted with the
suggestion that any kind of mutation is very rare, so that the
additional qualification that it must also be serviceable
renders it highly necessary that Selection must act with great
efficiency ; it also introduces the question — how frequently will
such rare mutations coincide with the selective circumstances

INTRODUCTION 7

that confer on them an advantage ? This matter will be
discussed at greater length at a later stage in this work.

(iii) On surveying the general field of variation in all its
aspects the first impression one gains is of the very great
plasticity of animals. This is, it is true, more clearly seen
in some groups than in others, but marked variability
is very general. Nevertheless variation is subject to strict
limitations. The living organism is not capable of variation
in all degrees and directions. Pantin (1932, p. 710), in an
interesting essay, refers the limitations of variation to the
fact that protoplasmic materials comprise a limited number
of standard parts of limited properties. In spite of the seem-
ingly infinite plasticity of morphological parts the variation
of the living substance is limited by the character of its mole-
cular structure. Thus Pantin (I.e. p. 709) cites the fact that
only four respiratory pigments have been evolved capable
of combining reversibly with oxygen. He suggests that
the same limitation affects the capacity for morphological
variation. We might explain on these lines the very notable
occurrence of parallel evolution and the development of
similar variation in allied species.

The limitations of variability in a particular group of
animals (Dinqflagellata) has led Kofoid (1906, pp. 251-2) to
stress the analogy between the variation of a group of ' ele-
mentary species ' and a group of related organic compounds.
' The seeming reversion in these mutants (?) of Ceratium to
old and fundamental subgeneric types, the occasional rever-
sibility of mutations elsewhere and the limitations in the range
and number of mutant types appearing in nature and under
culture suggest that the chemical nature of living substances
. . . place certain rather definite restrictions upon the number
and amplitude of the departures which mutants make from
their sources . . . the relation which exists among the mem-
bers of a group of elementary species . . . presents a striking
analogy to that which is found to exist among the various
radio-active substances or members of a chemical series of
related organic substances.'

In the preceding paragraphs we have considered the origin
and nature of variation, and for the purpose of defining our
particular problems it is now desirable to discuss a little more
fully the way in which variants occur in nature.

At the offset the exact study of natural variation is rendered

8 THE VARIATION OF ANIMALS IN NATURE

obscure by the relatively slight amount of precise knowledge
as to which variants are heritable and which are mere fluctua-
tions. Every population will contain a certain element of
individual forms having the latter status and sometimes
(possibly quite often) large sections of a population will be of
this nature ; this is particularly true of plastic organisms, such
as corals and hydroids, in which ' ecological types,' the
products of the peculiar environmental conditions found in
various habitats, have been often reported.

When fluctuations have been allowed for, as far as possible,
we are left with the important heritable elements. Of these
we may distinguish three kinds — (i) individual variants ;
(2) groups ; and (3) special categories of various types.

(1) Individual Variants. — Individual variants occur in
nature with very different frequencies and there is every
gradation between the variant which occurs sporadically
throughout a population and groups of appreciable size. In
some classes and orders sporadic individual variation is common;
in others, group-formation is more characteristic. The diver-
gence of such individual variants may be in one or several
characters.

(2) Groups . — Although no two individuals are ever exactly
alike in all their characters, it is a commonplace that indi-
viduals can be classed together in assemblages or groups of
various kinds. For the study of the origin of variation the
constitution and status of such groups are irrelevant, but, inas-
much as we find that variant individuals tend to form groups
characterised by the possession of a set of common and peculiar
characters and that such group-formation seems to be the
initial stage of evolutionary divergence, it is clearly part of
our business to inquire into the process by which recognisable
groups are formed.

These groups differ among themselves not only in their
degree of distinctness, but also in the nature of the distinction.
Thus a clone is a different kind of assemblage from a physio-
logical race. The various kinds of groups recognised are
discussed in Chapter III. For the moment we are concerned
only with a single general question, viz. the relation between
taxonomic units and the concept of the natural ' population.'

The facts of variation, and indeed all the phenomena with
which the biologist deals, are most often given in association

INTRODUCTION 9

with a specific name. The very idea of variation assumes
deviation from a norm which is invariably the character of a
group defined (whether as species, subspecies, or race) by taxo-
nomic procedure. To anticipate the discussion on the species
(Chapter III) we must point out that the latter is not a group
with standardised properties by which it can be invariably
recognised as such. It is an abstraction from a number of
individuals varying in such a way that any group or groups
defined must do some violence to the natural divergences that
certainly have always occurred in time and very frequently
occur in place. There are further difficulties to note which
arise from the actual imperfections of taxonomy. The vast
literature of taxonomy and the categorical nature of its
definitions obscure the incompleteness of our knowledge
in this branch of zoology. In certain limited groups in which
abundant series have been collected and studied critically the
status of the species at least rests on a solid foundation. In
many groups, however, particular species are known only from
a few individuals, sometimes of one sex only. Sometimes our
knowledge of the range of variation of a species depends on
whether two forms found in different areas are really identical
and no adequate comparison of them has ever been made.
Often purely nomenclatorial difficulties intervene, e.g. where
one species is known under more than one name in different
countries. All these difficulties are intensified when we are
dealing with the finer taxonomic units, such as very closely
allied species or geographical races. Many generalisations
about the variation of particular species are still rendered
dubious in this way, probably many more than is usually
supposed. The imperfections of taxonomy in this respect
are doubtless temporary, but they are at the present time a
great practical difficulty in the investigation of variation in
nature and not uncommonly they produce an element of
doubt in generalisations as to distribution and similar matters.
A species, like a molecule, is a statistical summary, and a
comparison of its properties with those of related forms can
most efficiently be made with the aid of statistical methods
involving tests of significance. When simple measurements,
such as those of size, are being made or when the material
studied consists of numerically small samples, these tests are
often indispensable, but in a broad survey like the present

io THE VARIATION OF ANIMALS IN NATURE

one we are limited in two ways. First, we are bound to give
some weight to statements not verified by these methods, when
the author alone has had, and perhaps can have, access to
the material. Secondly, many problems in the study of
variation appear at present to be outside the field of statistics,
because it is not yet possible to obtain sufficiently accurate
measurements for statistical tests to be applied, e.g. to differ-
ences in habit or to some of the finer structures. Often those
characters which are most easy to measure have no biological
significance, while those for which measurement is most needed
are least susceptible to it. Finally, all taxonomists are familiar
with differences between races and species which depend on
a general ' facies ' ; the individual characters which go to
make up this facies can be measured singly and the correlation
between any pair of them determined, but no single formula
can express the whole.

We have laboured this point in order that at the offset it
may be amply clear that the study of variation within groups
is bound up with systematic procedure and is liable to errors
arising out of the inevitable defects of the latter. We do not
wish to minimise the risks to which theories of evolution are
liable through defective systematics. But although species and
other systematic categories are important reference points and
significant episodes in the course of evolution, with modern
intensive collecting-methods and the intensive study of large
numbers of individuals, the centre of interest is passing from
the systematist's species to the ' natural population ' from
which the species is abstracted.

The term natural population (cf. Chapter III) is given to
any assemblage of individuals of a species living in nature irre-
spective of its systematic relationships, i.e. whether it is homo-
geneous or whether it contains diverse genotypic elements.
A ' population ' consists of a number of more or less geno-
typically similar individuals which are better able and have
more opportunity to interbreed with one another than with
the individuals of other populations. Such populations
considered taxonomically may be only a group of individuals
isolated topographically (e.g. on an island) from other struc-
turally identical individuals, or they may form a definite
variety, geographical race or species. The taxonomic name
given to the population depends on a variety of circumstances,

INTRODUCTION n

but we are concerned with the character of the population
rather than with the name given to it. In the study of such
populations we can use for convenience any name that may
have been given by a taxonomist, even though groups put into
the same taxonomic category are not necessarily equivalent
in degree of isolation or divergence. Nevertheless the dis-
tribution of variants in nature does not, in general, appear
to be at random ; they are arranged so that different types of
populations can be recognised. Populations may be distin-
guished by a varying number of physiological and structural
characters which may be correlated in different degrees with
one another. Further, the size of the area inhabited and the
nature of the factors limiting the area may differ.

Topographical groups. — By far the most striking manifestation
of natural variation is the occurrence within a population of
larger or smaller groups of some measure of homogeneity.
Usually these are denned by at least partial isolation and they
range in size from a small patch of individuals (colony), pecu-
liarly characteristic of small sedentary animals like land snails,
to a group occupying an extensive area (geographical race).
Such groups may be rigorously isolated from neighbouring
races, or they may overlap. In the growth of these assemblages
we may note as in (i) that the divergence of one or more
characters may be involved.

(3) Special Categories. — The terms polymorphism and
dimorphism are sometimes used without any general agree-
ment as to their meaning and it is necessary to clear up this
ambiguity. In its clearest and most usually employed sense
dimorphism is applied to the occurrence within a species of
two strongly marked and discontinuous phases, such as we
see in the difference between the colour, etc., of males and
females, between seasonal forms, or between mimetic forms
{e.g. the East and West African female of Acraea alciope (Lepi-
doptera), Eltringham, 1910, pp. 44-45). The term has also
been given to other contrasted types within a species, whose
occurrence is not apparently related to bionomic needs, e.g.
by Bouvier (1904 : dimorphism of the Atyidae).

Polymorphism has been used either in a general way to
denote that a population is very variable (cf. Coutagne, 1895)
or with a special significance to denote the occurrence of
several well-marked phases which inhabit the same area.

12 THE VARIATION OF ANIMALS IN NATURE

The latter meaning is the one used in this work. The pheno-
mena to which it is applied are best exemplified by the mimetic
phases of certain Lepidoptera. Rarely, seasonal variation
may also be found to produce a polymorphic series, e.g. in
Daphnia acutirostris Woltereck (1928) found an unusual cycle
consisting of spring, summer and winter forms.

Over and above the variation just described a population
may contain other special elements such as castes (e.g. in
Hymenoptera), ' high ' and ' low ' males (Scarabaeidae) and
developmental phases.

General theories of evolution have usually concerned
themselves with questions as to the origin and importance of
new characters and the processes by which the continuous
transformation of such characters is brought about. The
reference to group-formation in the previous paragraphs
stresses an aspect and a result of the evolutionary process
which, though they are universally recognised, are perhaps too
little regarded. Darwin has been taxed for naming his most
important work ' The Origin of Species.' We may admit that
he thus gave undue prominence to the species as opposed to
other systematic categories ; but the implication that the
problem of evolution is closely bound up with that of the origin
of groups shows that he realised what to our minds constitutes
one of the essential problems of evolution. The formation
of groups having some degree of distinctness seems to be a
universal property of living organisms, and the whole scheme
of animate nature reveals itself as a hierarchy of groups begin-
ning with simple aggregates of the status of the pure line, the
clone and the colony and developing in distinctness and indi-
viduality through the local race to the species and higher
categories.

The main qualitative changes in evolution no doubt begin
with changes in single characters, and for the essential features
of the process, the linear changes in the history of organs and
of one individual type into another, the occurrence of groups
is perhaps at first sight irrelevant. As long as the necessary
changes occur, the question as to whether they occur in one
or 1,000 individuals might seem unimportant. But evolution
does not proceed by the transformation of single organisms,
but by the mass changes of populations. The outstanding
feature of the process as it is seen in palaeontological and

INTRODUCTION 13

systematic data is the continued break-up of populations,
the divergence of the groups thus formed along different paths,
and the replacement of groups having one kind of constitution
by other groups having a different constitution. What we
have to account for is not only the changes in single characters
or groups of characters in single individuals, but also the means
by which they become characteristic of populations. We
stress this obvious and generally accepted truth, because in the
generalisations based on experiments and observations in the
laboratory, or in the genetical and mathematical treatment of
the subject, emphasis is usually laid on the origin of new
characters and their chances of survival and the fact of group
formation are neglected. Moreover, various authors (e.g.
Kinsey, 1930, pp. 34-35 ; Hogben, 1931 ; Guyenot, 1930,
p. 211 et seq.) have suggested that any mutant might spread,
if it was not actually harmful to its bearer. Darwin also was
evidently of the same opinion and seemed to think that ' neu-
tral ' characters might survive. Haldane and Fisher, however,
have clearly shown that the mere fact of re-emergence from
a cross does not confer on mutations the power to spread
through a population. The spread of variants is, indeed, one
of the most crucial problems in the study of evolution.

We will now proceed to formulate what we believe to
be the chief problems which a study of natural variation
raises.

(i) A population inhabiting a definite area may gradually
change in the course of time, or two populations, originally
similar and practically homogeneous, but inhabiting different
areas, may diverge so as to become two distinct groups. The
two processes are probably much the same, though in the
latter case it may be possible to point out definite differences
in the environment of the two areas to which the divergence
might be due. In either case we have to explain the origin
of the new characters by which the diverging groups differ
from those they used to resemble, i.e. we have to consider the
causes of variation.

(ii) As indicated on pp. 8-1 1, variants are not found dis-
tributed chaotically but in groups of various kinds. It is
necessary to define what these groups are and how they occur in
nature.

(iii) It is evident that our definition of the term ' population '

14 THE VARIATION OF ANIMALS IN NATURE

depends not only on a morphological criterion but also
on differences in ability to interbreed, populations being
more or less isolated from one another. We may distinguish
between populations separated by temporary topographical
barriers, populations which, if the barriers were removed,
would interbreed freely and soon become homogeneous, and
those separated by permanent reproductive barriers either of
instinct or due to sterility. In the latter case the populations
remain distinct even when inhabiting the same area. The
study of variation, therefore, is much concerned with the
origin and causes of isolation.

(iv) In different individuals of a population are many
more or less peculiar characters, but only those will be called
specific which are found in association in the bulk of the
members. Thus the specific characters are more or less
correlated with one another and we have to investigate the
origin and causes of this correlation.

(v) The divergence of populations depends not only on
the occurrence of new variations but on their accumulation
to give rise to those groups of characters by which species are
recognised. Any new character to become specific, if it does
not first appear in a number of individuals simultaneously,
must arise in one or a few individuals and then spread through
the species. We must further consider, then, the spread of
new characters within the population.

(vi) We have finally to consider what is the relationship
between the establishment of groups and the main tendencies
of evolution. It is almost universally held that the main
adaptive divergences which constitute the most striking feature
of evolution are merely group-divergences progressively raised
to a higher power by the continued operation of the same
processes that produced group-formation. We consider that
this is a questionable doctrine. Chapters IX and X are
devoted to a consideration of the relation between variation
and organisation.

The problems enumerated above will be treated in the
following order. In Chapter II we consider the origin of
variation. In Chapter III we enumerate the types of
groups recognised as the result of various methods of study
(systematic, genetical, etc.), and in Chapter IV we detail how
variants and groups of variants are actually found in nature.

INTRODUCTION 15

The action of isolation in producing discontinuity is dealt
with in Chapter V and that of correlation in Chapter VI.
The efficacy of Natural Selection as the most generally ac-
cepted theory of the spread of new characters is examined
in Chapter VII. It is shown that the scope of this process
is questionable. In Chapter VIII we examine the other
theories of evolution, and in Chapter IX the nature of adap-
tation and the special difficulties of explaining its origin
are detailed. In a general summary (Chapter X) we
attempt to define the relationship between adaptation, varia-
tion and group-formation and to distinguish between their
presumed causes.

We may conclude this chapter with some remarks on
procedure in evolutionary inquiry in so far as our methods
are involved.

Many of the subjects mentioned above can be investigated
experimentally. The origin and mode of inheritance of
variation are almost exclusively to be treated in this way.
The validity of the selection hypothesis, as an explanation of
the spread of variants, has been likewise tested by experiments
in the field and in the laboratory, and the formation of new
habits, food preferences, reactions to the environment, etc.,
have been similarly investigated. The behaviour of animals,
their interrelationships, seasonal occurrence and the incidence
of actual environmental pressure on animal populations are
most profitably studied by direct observations in the field.
For the study of the distribution of variants in nature, the
formation of groups and the incidence of correlation we fall
back on the methods of taxonomy and statistical analysis,
though the findings of genetics are of service here : of supreme
importance is the method of population-analysis, which is a
combination of statistics, field observation and taxonomy.
This has been much in vogue during the past thirty years.
It dates further back indeed, viz. to the pioneer work of
Coutagne, Gulick, Duncker and Heincke, and to other studies,
particularly of economically important animals (fishes).
More intensive and critical work supported by modern gene-
tical and statistical methods has been conducted by such
workers as Crampton, Schmidt, and Sumner.

In this work we are approaching the subject of evolution
primarily as taxonomists. We believe that all theories of

1 6 THE VARIATION OF ANIMALS IN NATURE

evolution should be tested by the results of taxonomy (dealing
with both living and fossil forms) and population-analysis.
These two studies, more than any others, bring the theories
of evolution into contact with the gross facts of nature. We
realise their specific limitations and in particular the need to
supplement them by observations on habits and behaviour,
but we feel that they constitute an acid test of evolutionary
theories based on other studies. This test has been insufficiently
applied in the past. It is well worth while to try to describe
the facts of nature as they actually are and to see what are the
simplest deductions suggested. There has been a tendency
to ignore or distort certain observations because they fail to
fit in with the theories, e.g. some of them seem to suggest a
neo-Lamarckian explanation of evolution, but this idea has
been nearly always ruled out on a priori grounds. The occur-
rence of non-adaptive specific characters, and certain palaeon-
tological and other evidence suggest that variants can spread
without any adaptive qualifications. But recently mathe-
matical theories have been invoked to prove that this is im-
possible. We believe it is advisable to make new contacts
between theories so obviously developed by deductive methods
and the large body of recorded observations from which they
have been so long divorced.

It appears that on the whole modern writers on evolution
fall into three classes. The first are impressed by the obvious
facts of adaptation. They take variations for granted and
tend to describe the assumed effects of selection. The second
argue from a relatively few animals which have been studied
under laboratory conditions. They tend to assume that,
when once a mutation has occurred, it can look after itself
and that, as long as it is not harmful, it can spread through
a population. The third class, recognising that the spread
of variants needs explanation, have given exact mathematical
expressions for the efficiency in this respect of Natural Selection
without, however, first showing that that process is actually
operative in nature.

In our attempt to evaluate the evidence put forward on
behalf of the various theories of evolution we discuss the
logical conditions for an exact proof of certain theories and
in particular (p. 186) Woodger's account of the stages by
which a theory attains the status of an accepted truth. It is

INTRODUCTION 17

unfortunate that along with the development of theories as to
the causes of evolution no serious methodology has been
developed and very little attention has been paid to the
logical requirements of such inquiry. The ground is partly
covered by Woodger's admirable ' Biological Principles '
(1929) ; but there is still need for an inquiry into the methods
of evolutionary research and the logical procedure by which
the main and subsidiary theories may be tested.

CHAPTER II

THE ORIGIN OF VARIATION

It is generally held at the present time that there are three
main types of variation differing in their mode of origin, viz. :
(i) fluctuations or non-heritable somatic variations, (2) the
effects of recombination of existing genes, and (3) mutations in
the wider sense (Chapter I, p. 4). Most biologists believe
that there is a real distinction between spontaneous germinal
change, which is heritable, and non-heritable fluctuations,
and they experience great difficulty in accepting any evi-
dence that changes wrought either on the body cells or on
the germ cells by external agencies, by use or by changed
habits, are inherited. It is our object in this chapter to examine
the evolutionary importance of the different modes of origin
of variation. After estimating the importance of those pro-
cesses we consider whether fluctuations can ever become
hereditarily fixed. We deal with these questions in the
following order :

1. Fluctuations.

2. The basis of heritable variation.

3. Recombination.

4. Mutation in the restricted sense — Gene-mutations.

5. The inheritance of induced modifications :

(a) General considerations.

(b) Experiments.

(c) Circumstantial evidence.

(d) Habit-formation.

(e) Summary.

Finally, we attempt to summarise the data and to evaluate
their importance in the study of evolution.

Before proceeding with this programme we may consider
what importance the origin of variation has in the study of

THE ORIGIN OF VARIATION 19

evolution. An intelligent layman once observed to one of us :
' Why do you worry how variations arise : surely it is their
fate that matters ? ' Up to a point this is a valid criticism.
But, if we anticipate what is discussed in later chapters, it is
of considerable importance to decide whether new variants
arise only in a few scattered individuals or whether in some
cases whole populations are changed simultaneously. In the
former case we have to explain how the rare variants spread.
Again, any factor seriously affecting the rate of mutation might
have some influence on the chance of establishment of mutants,
especially in a rare species. In fact, apart from its logical
value in completing the theory of evolution, some knowledge
as to the origin of variations is necessary to form any theory
at all.

1. Fluctuations

That animals are more or less ' plastic ' or modifiable by
the environment in their structure, reactions and physiological
properties and activities is a fact of general knowledge. 1 We
do not propose to describe the many and varied effects which
external factors produce. They have been sufficiently detailed
in a number of works, and the varying action of temperature,
salinity and other chemical factors, humidity, etc., is familiar
to most biologists. Surveys of the subject have been made by
Hesse (1924), Cuenot (1925), and others, and studies of the
effects of all known environmental factors on a single group of
animals have been made for the Mollusca by Pelseneer (1920)
and less fully for the Insecta by Uvarov (1931) and Chapman

(i93i)-

In actual practice the proof of the non-hereditary nature
of a variation is relatively infrequent and the great bulk of
' fluctuations ' is diagnosed as such on a priori grounds. Yet
no variation, as far as we know, declares its origin by its mere
' appearance ' (p. 78). Whether it is a fluctuation or of fixed
heredity can be determined with certainty only by experiment.
Nevertheless many systematists and other writers proceed as if
it were possible to determine the nature of a variant by mere

1 The ease with which some animals are experimentally or otherwise modified
by their environment should not lead us to ignore the marked constancy with
which others retain their specific characters. Nabours (1929, p. 55) lists a long
series of environmental factors which have no effect on the colour-patterns of the
grouse-locusts (Tettigidae).

20 THE VARIATION OF ANIMALS IN NATURE

inspection and write-off many forms as ' mere fluctuations '
or ' due to the environment.' It may be claimed that this
procedure is justified by analogy with effects known to be
produced by experiment. But actually a number of experi-
ments has been claimed to show that certain effects are due
to the environment, though no examination was made of the
behaviour of the affected characters in heredity. Further, the
amount of variation that is treated as non-heritable is far in
excess of the number of cases that have been experimentally
verified.

It is not easy in fact to obtain more than relatively few
instances of characters which have been shown experimentally
to be non-heritable. Among the Mollusca, the form albo-
lateralis of Arion empiricorum (ater) (Collinge, 1909), the carinate
and ecarinate forms of Paludestrina jenkinsi (Robson, 1929), and
various forms of Limnea peregra (Boycott, Oldham and Waters-
ton, 1932) seem to be definitely fluctuations. Pelseneer (1920,
p. 641) catalogues a list of 'variations non hereditaires ' in
the Mollusca ; but in all his cases, except that of Arion ater,
there is no evidence that the character in question was not
acting as a simple recessive, since the breeding test was not
extended to more than one generation. In the insects, which
have been so much used for genetical research, rather more
cases are available. Some of the naturally occurring colour-
variations of the bug Perillus bioculatus (Knight, 1924) and of
the parasitic wasp Microbracon brevicomis (Genieys, 1922) are
certainly not inherited. As for variations known only under
artificial conditions, we may mention a white variant of the
moth Ephestia kiihniella (Kiihn and Henke, 1929) and a number
of variants in Drosophila, especially reduplications of various
organs (Morgan, Bridges and Sturtevant, 1925, p. 71 et seq.).
Amongst birds, Beebe's (1907) experiments on the effect of
a humid atmosphere on doves of the genus Scardqfella are
well known. In the rotifers, Kikuchi (1931) shows that in
Brachionus pala lateral spines are developed when the animal
is fed on the alga Scenedesmus ; the spines are lost when it is
fed on Polytoma, and the action is completely reversible.

A point worth remembering in discussing this question is
that a given character may be heritable in one form and not
in another. This is especially evident in the matter of the total
size of an organism which is determined not only by the

THE ORIGIN OF VARIATION 21

available food and the temperature at which development
occurs, but also by numerous genetic factors. It seems also to
be the case in some of the naturally occurring strains oiDaphnia
studied by Woltereck (1908) ; e.g. the low-helmed form from
the Lund See could be easily transformed into a high-helmed
form, but the apparently similar variant (mutant E) of the
Frederiksburg See could not be modified by the same conditions.

These facts are of some importance. In the minds of most
workers there is a general idea that animals live in a variety
of places and are exposed to a diversity of environmental factors
that produce a great amount of merely somatic modifications
— that all animals are in varying degrees plastic and receive
a more or less marked amount of modification from the food
they eat, the soil on which they live, and so on, and that much
variation is without moment in evolution, because it is not
heritable. The assumption that animals are plastic is no
doubt a sound one ; but each case ought to be considered on
its own merits and tested by experiment.

In practice what is done, in taxonomy at least, is to
proceed by no particular principle except some such idea as
that, if a short form of a marine Gastropod (e.g.) is found in
brackish water, it is a ' stunted ' (somatic) form. The result
is that species and their variation are described according to
the systematist's very varying knowledge of experimental
work. This is, of course, a matter of systematic procedure ;
but it is important, as to a certain extent the work of the
systematist is taken as evidence of the plasticity of animals.
As we suggest later (p. 55) we do not know if this plasticity is
actually without evolutionary significance. Moreover, most
workers would probably agree that more of the alleged
fluctuations are hereditary than was at one time supposed.

The role of intrinsic and extrinsic factors in the production
of fluctuations deserves considerably more attention than it
has yet received. Investigations are often carried out under
insufficiently standardised conditions and there is a consequent
tendency to attribute variation to unknown differences in the
environment. Again, there is usually a considerable probability
that the species studied are genetically very diverse. The two
loopholes so provided are quite sufficient to prohibit much
generalisation. It would, however, be a matter of some interest
to discover how far. variation can be eliminated by rearing

22 THE VARIATION OF ANIMALS IN NATURE

stringently selected strains under thoroughly controlled con-
ditions. It appears by no means impossible that a certain,
not altogether negligible, range of variation might remain
under the most severe precautions. The complex organisa-
tion of the higher animals would appear to be inherently
unstable and liable to irreversible changes. The data with
regard to conditioned reflexes suggest that this may be the
case in the nervous system and it is likely that other
organ-systems may be liable to similar ' habit-formation.'
Under severely controlled conditions it might still be possible
for permanent ' deformations ' to result from intrinsic causes.
There are, of course, good grounds for believing that
physiological rhythms may be permanent in at least the
lifetime of the individual. Thus Payne (1931) found that
in the parasitic wasp Microbracon hebetor, adults taken from
cultures reared at high temperatures lived a shorter time at all
temperatures than those taken from lower temperatures. In
the future it may be hoped that the large amount of research
now being conducted into the effects of controlled conditions
of temperature and humidity on insects will provide significant
data.

2. The Basis of Heritable Variation

The nature and distribution in heredity of the visible
characters of an organism are to an important extent deter-
mined by the way in which they are represented in the
chromosomes of the germ cells. Thus some characters are
determined by a single gene, others by several genes, and others
again by complementary genes. Or again the distribution of
certain characters will depend on whether linkage occurs or
not. The way in which characters are genetically determined
will thus influence their variation.

In discussing the origin of variation we have to distinguish
carefully between the origin of new hereditary material and
the occurrence of variation due to differences in the way in
which characters are genetically represented. The latter
includes, for example, the effects of recombinations and com-
plementary genes. We have, therefore, to examine the
various ways in which characters are genetically determined
in order to distinguish the sources of new evolutionary steps
(mutation) from other forms of hereditary variation.

THE ORIGIN OF VARIATION 23

Haldane (1932, p. 37 and foil.) has distinguished six modes
of genetic representation which are tabulated below, though
it is by no means clear that all are found among animals.

(1) Characters determined by extra nuclear factors (plas-
mons). Haldane thinks that some of Goldschmidt's results
(1923) on sexuality in moths illustrate this (cf. also Boycott,
Diver and others (1930) ; Toyama (191 2) on heredity of
voltinism in silkworms) .

(2) Characters determined by a single gene.

(3) » » » several genes.

(4) ,, ,, ,, genes which undergo re-

arrangement (but not alter-
ation in number and
quality).

(5) „ „ ,, genes some (but not all) of

which are represented more
or less than twice in aber-
rant types of individual, e.g.
non-disjunction.

(6) „ „ j, genes the total diploid num-

ber of which is increased
by one or more whole sets
(polyploidy).

Before proceeding to discuss these various modes of genetic
representation we ought to remind the reader that the term
' mutation ' is applied either in a narrow sense to changes in
a single gene or to the various phenomena of chromosomal
abnormality and other variations dependent on variation in
the genetic basis of characters. It seems clear that in Haldane's
list the differences enumerated under 2, 3 and 4 are chiefly
related to differences in the distribution of characters and
to recombination. Differences in sex and fertility are also
associated with 4 (attachment of X to Y chromosome).

Morphological change seems to be associated with 5 in
plants, and Haldane states (I.e. p. 52) that the presence of an
extra chromosome generally produces a very unhealthy type
(cf. production of intersexes possessing the second and third
chromosomes in triplicate and the X in duplicate in Drosophila
(Morgan, Bridges and Sturtevant, 1925, p. 156) ). It is not
clear if any morphological changes are associated with this

24 THE VARIATION OF ANIMALS IN NATURE

abnormality. As to 6 the position is uncertain. Polyploidy is
not completely absent from animals, but according to Gates
(1924, p. 177) there is nothing comparable to the condition
found in plants. Varieties univalens and bivalens with 2X and
4X chromosomes have been recorded in Ascaris megalocephala,
Artemia salina, etc. In three out of the four cases noted
by Gates ' no particular significance seems attached to
the bivalent or tetraploid conditions' {I.e. p. 177). In the
Phyllopod Artemia salina it appears to be associated with
differences in reproduction, a tetraploid form of that species
being parthenogenetic. Tetraploids have been found in
Drosophila (Morgan, Bridges and Sturtevant, I.e. p. 21), but
' as yet their chromosomes have not been studied.' As regards
the appearance of entirely new characters from any of the
various modifications of chromosomes (either those treated
here as abnormalities or those figuring in 4 to 6) in Haldane's
list, it seems clear that new characters or at least new com-
plexes of characters have arisen, e.g. as seen in the appearance
of the ' Diminished ' mutant due to the loss of a ' fourth
chromosome' (Morgan, Bridges and Sturtevant, I.e. p. 136).
But, owing to low viability (I.e. p. 137), it certainly seems that
this type and probably other similar ones are of small
evolutionary importance.

Up to the present we have had little opportunity outside
the study of Drosophila to distinguish between the various causes
of mutation (in the broad sense, p. 4), so that it is not
possible to distinguish between gene-mutation and chromosomal
abnormality, etc., from the evolutionary point of view. On
the other hand, in the many experiments on induction, etc.,
that have been carried out, we do not know what kind of
mutation is involved. From Mavor's experiments (1922) it
seems clear that X-ray treatment causes non-disjunction of the
X-chromosome.

3. Recombination

It is sometimes not realised what an enormous scope for
variation lies in the permutation of a relatively small number
of gene-differences. Fisher (1930, p. 96) points out that in
a species with a hundred segregating factors the number of
different true-breeding genotypes would be so large as to
require thirty-one figures to express it, or forty-eight if the

THE ORIGIN OF VARIATION 25

heterozygotes are included. Thus, even if thousands of
millions of individuals are produced in any one generation
and no two individuals are genetically alike, only a small
fraction of the possible combinations would actually be
realised. The possibilities of recombination are much en-
hanced by the variation in the expression of genes when
combined with different gene-complexes.

Outside domesticated forms it is not very easy to find good
examples of the effects of recombination. Permutations of
specific characters within a genus are, of course, very familiar,
but owing to the occurrence of sterility, etc., are rarely capable
of genetical investigation. Amongst domestic animals recom-
bination, leading to novel forms, was early recognised in
poultry, rabbits and pigeons. In a wild insect we may
mention the cases of Papilio polytes investigated by Fryer (19 13)
and of Aricia medon studied by Harrison and Carter (1924).
In the latter species two forms meet on the Durham coast and
a wide range of variants, many not known elsewhere, is
produced. More usually the meeting of two geographical
forms leads merely to the production of simple intermediates
(see p. 89). It is probable that recombinations of numerous
small gene-differences in wild populations are responsible for
a considerable part of the continuous range of variation in
size, colour, etc., often alleged to be fluctuational.

It is very difficult to assess the actual evolutionary value
of the variation arising in this way. Some authors, such as
Lotsy, have supposed recombination, especially after crosses
between very different varieties or species, would supply all the
variation required for evolution. This theory is more plausible
in the case of plants — in which interspecific sterility is not so
much developed — than in that of animals. The problem is not
one of very easy direct approach, for genetical experiment on
a sufficient scale is lacking, but some indirect evidence may
be obtained. If the individuals of a species often differed from
one another in a large number of genes we should expect that
crosses of such individuals would give rise to a wide range
of variation, including some forms perhaps quite distinct from
either parent. Continued inbreeding of such a stock would give
rise to a large number of distinct lines. Apart from domesti-
cated forms, which are in quite a different category (cf. p. 188,
Chapter VII), it is not easy to find good examples. In some of

26 THE VARIATION OF ANIMALS IN NATURE

the most obvious cases, such as polymorphic butterflies or snails,
the evidence suggests that the various forms differ in a rather
small number of genes and the range of variation on crossing
is not very great. If we except geographical races and poly-
morphic species, crosses within the species rarely give rise to
a large series of variants. We are not aware, however, of any
serious attempt to discover by prolonged inbreeding how many
genes might be present. Duncan (1915) crossed specimens
of Drosophila from widely separated localities, but found that
no unusual amount of variation resulted. Unfortunately, the
flies of this genus are so largely spread by commerce that they
are not suitable material for such an investigation. Timofeef-
Ressovsky (1927) obtained seventy-eight wild females of
Drosophila melanogaster from a house in Berlin. It was supposed
that each of these had already mated with more than one male.
As a result of interbreeding it was deduced that eighteen of the
females and thirty-four of the males were heterozygous for at
least one mutant. Ten different genes were identified, some
of them already known in cultures.

Geographical races when crossed often give a consider-
able range of variation, usually intermediate between the
parents. If the types produced by recombination are few,
the chances of a beneficial variant are smaller, while the
larger the number of types, the fewer the individuals of each
that will appear. As far as the evidence goes, it would seem
that most individuals of a species are homozygous for a large
common stock of genes, so that little or no recombination
would occur on crossing. The geneticists' idea of a ' wild
type ' is partly based on this assumption. Of course we cannot
say how far this is true of genes producing only very minute
external effects, but we must judge by what evidence we have.
When forms differ considerably, so that recombination would
be expected to produce much variation, sterility in one form
or another seems usually to intervene. It is quite possible
that the majority of animal species have always been homo-
zygous for most of the genes carried at any one time.

No doubt some crossing between species, subspecies, etc.,
occurs in nature. How far such unions are fertile is a very
debatable point. When we consider the diversity of means by
which isolation is brought about (Chapter V) it does not seem
likely that successful crossing is very common or that it occurs

THE ORIGIN OF VARIATION 27

between individuals of markedly contrasted genetic constitu-
tion. In view of this, Lotsy's speculations as to evolution by
crossing appear unlikely to have a wide application in the
animal kingdom. There is a further difficulty in the way of
Lotsy's theory. If it has been something more than a minor
factor, we would have to admit that all the material of variation
was in existence in the earliest forms of life, and evolution has
consisted in the allocation to the forms which diverged from an
ancestral stock of various portions of this fund of material and
the recombination of parts of it to form new genetic groupings.
That a good deal of factorial recombination (with the appear-
ance of ' novelties ' due to this cause) has taken place we
do not doubt. But, if recombination is the only or even the
main source of variation, we have to imagine evolution as
merely the revelation of latent possibilities — a picture very
difficult to harmonise with the facts, for, looked at in the
broadest way, evolution undoubtedly leaves the impression of
the continuous emergence of new types of organisms. Thus,
while recombination has an obvious importance in trying
out all the permutations of the material lying to hand, we feel
the need of another process which will provide new material.

4. Gene-mutations

In spite of the vast amount of genetical research carried
out during the past thirty years our knowledge of the origin
of gene-mutations is still extremely slight. In the first place,
if a given variant is a mutation and not merely a recom-
bination, it should appear suddenly in an inbred stock. Thus
only in very quick-breeding forms can much information be
accumulated.

In the second place a distinction must be made between
agencies which actually produce mutations and those which
accelerate mutation-rates. We may illustrate this distinction
by recalling the effect of temperature on growth in inverte-
brates. Here, while differentiation, within wide limits, proceeds
independently of temperature, the actual rate at which it
goes on is directly dependent.

In actual practice there is no known treatment which
regularly produces a high proportion of any definite type of
mutant. Such agencies as X-rays induce variation in all

28 THE VARIATION OF ANIMALS IN NATURE

directions, while other treatments which have been supposed
to produce ' one way ' mutation have given only a very low
percentage of mutants. It is possible, therefore, that all these
agencies merely alter a mutation-rate which, even without
special treatment, would slowly lead to the production of
mutations which the treatment makes more numerous.

Before considering the experimental evidence for alteration
of the mutation-rate, there is one other point that must be
considered. Those who do not believe in the possibility of the
inheritance of acquired characters sometimes write as if the
experiments carried out in this connection were designed to
investigate the factors controlling the mutation-rate. Thus
Sonneborn (1931), commenting on Macdougall's experiments
on rats (see p. 40), writes as if Macdougall had produced
a series of adaptive mutations (i.e. assuming Macdougall's
claim to be technically sound). In our view this is a confusion
of the point at issue. The question is rather whether there
is not a special process, in addition to mutation, by which
characters gradually become inherited under prolonged environ-
mental influences. We have to distinguish between (a) induced
mutations which are hereditarily stable from the start and do
not revert back to type except by a jump as sudden as that
by which they arose, and (b) induced modifications which
gradually become more intensified and more stable as the
stimulus lasts longer and are often slowly lost when the stimulus
is removed. Variation of this second category is considered
in our next section. At the moment we shall consider only
examples of what is clearly induced mutation.

It was long thought that gene-mutations were spontaneous
because they are so rare, so erratic in occurrence, and appar-
ently so unrelated to any known factor in the environment.
It has been held that mutations observed in animals kept
under standard cultural conditions cannot be related to an
environmental cause, and the mode of origin of the Drosophila-
and Gammas-mutations has been regarded as evidence of
first-class importance. It may be noted that a great deal of
the evidence relates to mutations in eye-colour and develop-
ment (20 per cent, in Drosophila, 100 per cent, in Gammarus)
and nearly all the mutants are more or less of the nature of
defects. This cannot but arouse suspicion that some dis-
turbing external agency may be involved.

THE ORIGIN OF VARIATION 29

In so far as the vital activities are physico-chemically
determined it is impossible to imagine that mutations can be
truly spontaneous. Doubtless all that this term has meant in
the writings of those who have thought out its implications, is
that the agencies responsible for gene-constancy or gene-
mutation are so numerous that it is difficult or impossible to
speak of any one as the cause. A theoretical discussion has
been given by Schmalfuss and Werner (1926) with reference to
the hypotheses that the genes are enzymes (Goldschmidt) or
autocatalytic substances (Hagedoorn), and the conclusion is
favoured by them that mutations are produced by the action
of external factors on specific catalysts.

More recently good experimental evidence has been put
forward to show that high temperature, (3-rays (of X-rays) or
y-rays (of radium) have a marked effect on the mutation-rate.
We shall mention these experiments briefly in the order
indicated.

A. Effect of High Temperature. — Goldschmidt (1929), Jollos
(1930) and Rokizky (1930) have shown that the mutation-
rate of Drosophila is very much raised when the late larvae
are subjected to a temperature so high (35°-37° G.) as to kill
most of them. The attempts of other workers (e.g. Ferry and
others, 1930 ; cf. also Muller, 1932) have been partially or
completely unsuccessful. Apparently the mutations produced
are all types that have already been recognised. Jollos obtained
evidence that the mutations were largely in one direction and
the effect cumulative. This is very suggestive of the actual
causation of mutation, but more evidence is required on this
point. The results should be compared with the Dauer-
modifikation-experiments (p. 35).

B. Effect of X- Rays. —Muller (1928) showed that the
mutation-rate of Drosophila was raised about 150 times by
subjection to X-rays. Hanson, Heys and Stanton (1931)
have recently shown that the increase in mutation-rate, as
measured by the number of sex-linked lethals, is directly
proportional to the X-ray dosage. Similar results have been
obtained by Little and Bagg (1924) and Dobrovolskaia (1929)
with mice. Most of the mutations are not unknown in normal
cultures, though some of those in mice are apparently novel.
The effect would seem to be one of general disturbance, since
Mavor (see Morgan, Bridges and Sturtevant, 1925, p. 116)

30 THE VARIATION OF ANIMALS IN NATURE

found that the amount of non-disjunction of the X-chromo-
some in Drosophila was also materially increased. Many papers
have been published on this subject during the last few years,
but these seem to be the essential facts.

Huxley (1926) and Haldane (in Robson, 1928) at one
time suggested that naturally occurring radiations might
cause the apparently spontaneous mutations. But Muller and
Mott Smith (1930) have shown that this is highly improbable.

C. Effect of Radium. — Hanson and Heys (1928) obtained lethal
mutations in Drosophila by exposing the males to the whole
radiation of radium or to the y-rays only. Similar results have
been obtained in plants. On the whole it appears much more
difficult to obtain positive results with radium than with X-rays.

D. Experiments with Salts of Lead and Manganese. — Harrison
and Garrett (1926) and Harrison (1928a) claimed to have
produced melanic mutations in certain Lepidoptera by
feeding the larvae on food-plants which had absorbed these
metallic salts. Plunkett (1927) criticised the 1926 results
chiefly on the score of the low number of individuals
involved in the experiments. Recently, Hughes (1932) and
Thomsen and Lemche (1933) have repeated the experiments
on a very large scale without producing any melanics. It
appears probable either that melanic mutations occurred as a
very rare coincidence in the stock that Harrison was using or,
as suggested by Haldane (in Hughes, I.e.), that the original
parent was heterozygous and the recessive melanic factor is
linked with a lethal. (Cf also Harrison, Proc. Roy. Soc,
London, 117 B, 1935.)

We see therefore that in a few cases the mutation-rate has
been directly affected by external agencies. It must not be
forgotten, however, that some of the agencies used (e.g. X-rays)
are not likely to be influential in nature. In the same way we
should disregard the experimental induction of hereditary
defect by such toxic agencies as alcohol (Stockard and Papa-
nicolaou, 1916) and lead acetate (Cole and Bachuber, 1914),
which really amount to a direct poisoning of the reproductive
organs.

5. The Inheritance of Induced Modifications

(a) General Considerations. — This subject has been dis-
cussed almost ad nauseam and there are numerous critical

THE ORIGIN OF VARIATION 31

summaries. The most judicious and well informed, though
by now a little behind the time, is that of Dctlefsen (1925),
which is admirable in its judgment and analysis. It omits
some important experimental work (viz. that of Agar, Sumner
and Woltereck) and does not discuss some of the circumstantial
evidence (e.g. that based on geographical distribution) in
detail. The analysis given by Robson (1928), which is largely
based on Detlefsen's summary, contains a more detailed
reference to these subjects, though the question of ' Dauer-
modifikationen ' (p. 35) is only lightly touched on, and it
does not include mention of Woltereck's work. The following
discussion is largely based on the two studies just alluded
to, with an extended consideration of certain circumstantial
evidence in addition.

There is no need for a long account of the historical con-
troversy as to the origin of variation. It is enough to say that
in the period up to and including the first acceptance of the
theory of Natural Selection the heritable effects of environ-
mental change or of use-inheritance were freely held, and
Darwin himself, as is well known, accepted the idea.

The theoretical delimitation of the germ-cells from somatic
tissues and the idea of the organic integrity of the former were
due to Weismann, though he made a concession in favour of
* parallel induction ' as the result of his acceptance of Fischer's
experiments. Thus the matter stayed (with a few exceptions,
mostly among the palaeontologists) until the past two decades,
when the matter has again been called into question by the
work of Kammerer, Harrison, Przbram, Woltereck and Rensch,
and by the advocacy of MacBride in this country.

Opponents of the theory of the ' inheritance of acquired
characters ' and even those who were prepared to accept the
possibility that induced variation might be heritable have
always found a serious objection in the difficulty of explaining
how a modification of the parental soma might be transferred
to the germ cells. The experiments of Castle and Phillips
(191 1 ) on ovarian transplantation in guinea-pigs have been
held to show that germ cells having a given hereditary con-
stitution are not modified by being transplanted to a new
' somatic ' environment. These conclusions have been criti-
cised by Detlefsen (I.e. p. 257). The latter goes on to show that
there is much evidence to prove that our present cytological

32 THE VARIATION OF ANIMALS IN NATURE

knowledge of the origin of germ cells suggests that they are not,
at least in their early stages, likely to be immune from in-
fluences affecting the somatic tissues, inasmuch as they are,
in many cases, morphologically indistinguishable from the
latter (cf. also Gatenby, 191 6). However, the fact remains
that no mechanism by which a true Lamarckian effect could
be brought about has as yet been demonstrated. It is very
easy to imagine that a new habit or a far-reaching somatic
modification involving both structural and physiological re-
organisation and readjustment might have a profound effect
on the constitution of an animal. But the proof is still
lacking that such readjustment would have a specific effect
on the hereditary material of such a kind that the original
somatic modification was reproduced.

It is customary to attach very great importance to the
experimental evidence on this subject. Now the value to be
set on experiment in such a matter is open to some doubt. It
has the unfortunate limitation of being incapable of dealing
(as Caiman (1930) has pointed out) with the historical back-
ground of animal morphogenesis. This question becomes
crucial when we consider the negative evidence brought
forward to disprove the inheritance of induced variation. If
such-and-such a stimulus repeated for a few months or a few
years on a few generations fails to modify the germ cells, is
there any reason for assuming that it will have no effect if
the stimulus is applied, as it may well be in nature, for many
years and decades and over innumerable generations? We
cannot point to any case in which the duration of an induced
effect is proportionate to the time-intensity of the stimulus ; but
that such a contingency is possible ought not to be ignored and
negative results have to be accepted subject to this reservation.

Before considering the experimental evidence we shall
briefly set out what appear to be the essential conditions
for a really convincing experiment. It is one of the mis-
fortunes of the controversy that so much of the evidence is
equivocal. The following are, we believe, the necessary
precautions to ensure definite results.

1 . The Use of Inbred Stock. — In our section on natural varia-
tion (Chapter IV) we show how often species consist of a
mixture of strains. It is the universal experience of those who
breed animals under artificial conditions that inbreeding for

THE ORIGIN OF VARIATION 33

several generations sorts out the strains. These may differ from
one another in all sorts of characters, both morphological and
physiological. If an environmental factor modifies the appear-
ance or physiology of an animal, it is always necessary to make
sure that similar modifications, if not perhaps of the same
degree, do not occur in certain strains in nature.

There are two ways of guarding against this source of error.
The most satisfactory is to use an inbred stock. Ten genera-
tions of close inbreeding will probably isolate a reasonably
homogeneous strain. In many case:>, however, this pro-
cedure would be very lengthy or even impossible. The only
method is to employ adequate controls, which indicate that
the modification does not occur normally in untreated portions
of the same stock. It is impossible to say how many control
animals should be maintained ; in a variable species the
number necessary for stringent experimental procedure might
be so large as to make some preliminary inbreeding almost
essential. Even with large numbers of controls a mutation by
a coincidence may happen to arise in the experimental animals,
but the reduplication of experiments with different stocks
reduces the risk of misinterpretation.

2. The Elimination of Selection. — The experimental treat-
ment to which animals are subjected frequently causes con-
siderable mortality. If the survivors show some modification,
it is always possible that the mortality has been selective
and the survivors are that part of the original stock which was
genetically fitted to live in the novel environment. The
' modification ' of the survivors may be therefore only the
expression of their particular genetic constitution. Such
forms will be especially liable to lead the investigator to wrong
conclusions, because their characters will of necessity be
inherited.

The safest way of guarding against this error is to bring
to maturity every individual of every family throughout the
course of the experiment. If the experimental treatment
necessarily leads to considerable mortality, it may be almost
impossible to arrive at any convincing result, though the use
of highly inbred stock would be a great safeguard. In certain
cases (many insects) the size of the family is so great that the
stock would rapidly become unmanageably large if every
specimen was allowed to breed. In these circumstances it is

34 THE VARIATION OF ANIMALS IN NATURE

necessary to kill off part of each family, but the greatest care
must be taken to avoid any selection. With adequate statis-
tical treatment such material may still lead to a definite
conclusion.

This difficulty arises in its most acute form when only
some of the experimental animals show a modification. It
has often been the practice to carry on the stock only from
these modified individuals, thus introducing a stringent
selection in the direction of the modification. Two suggestions
may be made in this connection. First, repeated experiment
with different strains may show that the modification always
tends to arise in the experimental animals and never in the
controls. If the experiment stops when the modified indi-
viduals first appear, no selection can have been exercised in
that particular direction. If repeatable results of this sort can
be obtained, the effect of selection in later experiments is
relatively unimportant. Secondly, if the modification is an
induced mutation and is permanently heritable from the start,
selection is evidently only a secondary issue. To prove that
there has been an induced mutation is chiefly a matter of
reduplicating experiments with different stocks.

3. Persistence of the Modifications. — It is necessary to
distinguish at the offset between induced mutations and any
other sort of induced modification. Induced mutations
resulting from subjection to high temperature or to X-rays
are now well known in Drosophila and in mice. The dis-
covery of other equally effective agencies would be a matter
of great interest ; but it is evident that experiments of this
sort throw no light on the point at issue here. If one admits
that it is unlikely that mutations are really ' spontaneous,' the
discovery of agencies which raise the mutation-rate need not
excite great surprise, even when the mutations tend to be in a
particular direction. The question is whether there is any
process by which modifications gradually become hereditarily
stable. There is a sharp distinction here from mutations
which are stable from the start.

To prove that an induced modification gradually acquires
stability is certainly a difficult matter and there is a danger
that experiment will lead to a vicious circle in interpretation.
If the process alluded to can occur, then the modification
induced by experimental conditions must be expected to be

THE ORIGIN OF VARIATION 35

lost when the animals are returned to the control environment.
It is very difficult to decide what degree of permanence in the
modification must be established to prove the possibility of
the process. It is at least necessary that the modification
should be partially maintained for at any rate one generation
after the return to control conditions. Actually, in quite a
number of experiments no return to the control environment
was ever attempted.

4. The Value of Negative Evidence. — No amount of un-
successful experiments can prove that modifications do not
gradually become hereditarily stable. Under natural condi-
tions it might require many thousands of years for the
modification to become permanent.

On the other hand, the experiments should not entail
subjecting the animal to conditions very unlikely to be met
with in nature. If many thousands of years are required to
produce a stable modification, it is probable that only a few
simple agencies, such as low or high temperature or changed
salinity in the sea, can be effective. Few other environmental
factors are likely to operate steadily for long periods.

(b) Experimental Evidence, (i) Experiments on Protozoa. —
This work has been summarised critically by various authors
(see references in Robson, 1928, p. 168 ; and Hammerling,
1929). The bulk of the work (Jennings, Jollos and others)
concerns such forms as Paramoecium and Arcella and consists
in their habituation to altered temperature-conditions or to
doses of arsenic or calcium salts. Reversible modifications
(' Dauermodifikationen ') are frequently found. Some (e.g.
' calcium-dauermodifikationen ') are in all probability deter-
mined by changes in the cytoplasm and reversion follows on
the return to normal asexual reproduction after conjugation
(e.g. in Paramoecium). In Bacteria also the changes are still
manifested after transplantation to a new medium.

(ii) Experiments on Metazoa. — There is substantial evidence
that lesions are not inherited. We need mention only such
practices as circumcision, modification of shape of head or feet,
docking of tails, etc., which produce no heritable effect after
hundreds of generations (cf. also Agar, 1931).

There is a large number of experiments which may be
set aside or regarded as so questionable as to be practically
worthless as evidence. These are dealt with very briefly.

36 THE VARIATION OF ANIMALS IN NATURE

11.

in.

IV.

VI.

Vll.

Vlll.

IX.

Author

Experiment

Criticism

Ferroniere

Tubifex ; change

No controls. ? Direct

(190O

of medium

adaptation.

Kellogg and

Philosamia ; re-

? Direct weakening of

Bell (1904)

duced diet

P and F x genera-
tions.

Pictet (1910)

Lymantria ; change

? as ii. Possibly a

of diet

'Dauermodifi-
kation.'

Schroder

Gracilaria ; change

Lack of information

(1903a)

of habit

as to natural varia-
tion in habits.

Phratora ; change

Low number of

of food plant

cases (cf. Detlefsen,
p. 262).

Fischer(igoi,

Arctia ; effect of

? Genetic purity of

1907)

low temperature

stock.

Standfuss

Vanessa ; effect of

? Genetic purity of

(1898)

low temperature

stock.

Schroder

Abraxas ; effect of

? Genetic purity of

(1903a)

high tempera-
ture

stock.

v, vi and vii

are suggestive of induced mutation, but

there were no

adequate controls.

Guyer and

Cavia ; modifi-

Repeated unsuccess-

Smith (lit-

cation of lens by

fully by Silfrast

erature in

sera

(1922), Finlay

Guyer,

( 1 924) , and Huxley

1923)

and Carr-Saunders

Kammerer
(i9 J 9)

Alytes ; modifica-
tion of male
thumb

(1924).

(Experiments not
identical in the
first two cases.)

Diverse interpreta-
tions are possible
(see Detlefsen, I.e.
p. 266).

Procedure questioned
(cf. Noble, 1926).

THE ORIGIN OF VARIATION 37

x. Kammerer Ciona ; truncation Repeated by Fox
(1923) of siphons (1924), who did not

obtain the same
result.

xi. Tower (1906). Colour changes induced in Leptinotarsa
by alterations in temperature and humidity.

This very extensive series of experiments brings to light the
fact that, if the stimuli were applied to the eggs or larvae,
little or no change was effected. If they were applied to the
pupae, changes were induced which were not inherited. But
if the adults were exposed to the stimuli during the period of
maturation, the offspring alone were modified and the effects
were inherited. Tower's results have been very adversely
criticised, unfortunately on the score of the actual accuracy of
the results claimed. It is difficult to judge whether the criti-
cisms are finally destructive or inspired by prejudice. The
work has not been repeated, so that in all fairness it cannot be
used as evidence.

More recent work on the effects induced by temperature
and humidity in insects suggests that Tower's results must be
at least very exceptional, though sublethal temperatures may
induce mutation (p. 29).

xii. Diirken (1923) and Harrison (1928a). Colour changes
in PzVm-pupae.

Diirken studied P. brassicae ; Harrison, P. napi. In the
former species under normal conditions about 4 per cent, of
the pupae are green, in the latter about 2 1 per cent. If the
pupae are exposed only to orange light a much higher per-
centage becomes green — in P. brassicae, 69 per cent, in the first
generation, 95 per cent, in the second ; in P. napi, 93 per cent,
and 95 per cent, respectively. In Diirken's experiment
offspring of the first generation reared in normal light gave
41 per cent, green. Harrison's broods of the second genera-
tion gave 100 per cent, green in the third generation and
58 per cent, in the fourth. In both experiments the initial
stock may have been somewhat mixed and there was con-
siderable mortality, which may have involved some selection.
Further, in both experiments only green pupae were bred from
to obtain the pupae which were returned to normal conditions.
In both experiments, and especially in Harrison's, the in-
herited modification occurred in far more of the offspring

38 THE VARIATION OF ANIMALS IN NATURE

than would be expected if the result was entirely due to
selection, considering the small amount of elimination in-
volved. Further, in another experiment of Durken's (see his
fig. 8) selection of non-green pupae did not eliminate the
individuals with power to become green, so that there is no
reason why reverse selection should have given a pure line of
green. We believe a prima facie case has been made out for
the inheritance of this modification.

xiia. Wladimirsky (1928). Colour of pupa of Plutella
maculipennis.

In this moth the amount of black pigment in the pupa
case appears to depend jointly on temperature, light and on
hereditary constitution. In view of this complicated relation-
ship it is rather difficult to come to certain conclusions.
Wladimirsky's experiments, which were carried on over twelve
generations, gave results not unlike those of Diirken and
Harrison, though the author himself does not regard them as
evidence for the inheritance of induced modifications, selection
being at least partly responsible. The question how far selection
was exercised in this case is a difficult one to decide, owing to
the heterogeneous nature of the material.

xiii. Kammerer (1913). Induced colour-change in Sala-
mandra.

These experiments were carried out and the results are
presented in such a way as to make it impossible to draw any
conclusions as to the inheritance of induced modifications.
They were initiated with wild material, which may well have
been genotypically diverse. No exact numerical data are
given, so that it is impossible to discover whether any form of
selection may have been practised. The number of individuals
in which the induced changes were supposed to have been
inherited is not explicitly stated.

xiv. Metalnikov (1924). Immunity of Galleria larvae to
the Cholera Vibrio.

The account of these experiments is not sufficiently detailed
to enable one to draw any certain conclusions. There is no
description of the stock used, no detailed lineages are set out,
and the system of mating adopted is not stated. As far as
can be gathered, larvae were immunised against the Vibrio
and the survivors in each generation were bred from. There
was thus a stringent selection in favour of immunity and it is

THE ORIGIN OF VARIATION 39

not surprising that the percentage of immunity eventually
rose.

xv. Agar (191 3). Effect of temperature and medium on
Simocephalus.

Agar succeeded in inducing heritable changes in the size
of Simocephalus vetulus (Cladocera) by raising the temperature
of his cultures. He also experimentally induced an outward
flanging of the edges of the carapace by keeping his cultures
in Klebs' solution. These modifications were reproduced in
F x individuals, the mothers of which had been restored to
normal conditions just before the eggs were laid, and per-
sisted for some generations, though they became progressively
modified, i.e. they behaved as ' Dauermodifikationen.' Agar
interprets them as effects of ' parallel ' modification. As
reproduction was parthenogenetic, inheritance may have been
through the cytoplasm.

xvi. Woltereck (1908, 191 1, 1921, 1928). Modification of
the ' helm ' in Daphnia.

The work of Woltereck on the modification of the ' helm '
of Daphnia stands in a rather different category from the
work just described. Woltereck claimed to have induced a
temporarily heritable change in the form of the ' helm '
by transplantation to a different medium and to have found
natural races exhibiting characteristics similar to those which
he induced, living in appropriate natural conditions.
Woltereck's conclusions have been seriously challenged by
Wesenberg-Lund, who supplies a totally different explanation,
and the matter must be left very largely in abeyance, with the
qualification that as far as Woltereck's experiments are con-
cerned they bear a striking resemblance in the results to those
of Agar.

xvii. Sumner (1932, summary). Geographical races of
Peromyscus.

Sumner conducted for many years an extensive series of
observations and experiments on the species and races of
Peromyscus (deer-mice of N. America). He has summarised
the work in a survey which involves the modification of views
previously published. As he states (1932, pp. 2-3), he started
the investigation ' with a distinct bias in favour of the cumula-
tive effect of climatic influence.' This bias was due to the
results of certain experiments on white mice. The animals

40 THE VARIATION OF ANIMALS IN NATURE

were subjected to different temperatures and it was found that
in ' warm room ' temperature there was an increase in tail-,
foot- and ear-length. The offspring of these were born and
reared in normal temperature and had longer tails, ears, and
feet than the progeny of animals kept in ' cold room ' tempera-
ture. This was found in three out of four lots. In the fourth
lot the relations as regards tail and foot were reversed. F 2
animals were not studied. For various reasons the experi-
ments were not very satisfactory (see Robson, 1928, p. 170).
It should be pointed out that similar results were obtained by
Przibram (1909).

Transplantation experiments were undertaken with
Peromyscus (1932, p. 27) and it was found that mice trans-
planted from one environment to another {e.g. from the
Mohave Desert to La Jolla) showed no change over six to eight
years and that there was no convergence in transplants of
various races under the influence of a common environment.
This fact and others (e.g. p. 58, the wide range in an un-
modified condition through a diversity of environments of
P. maniculatus gambeli) induced Sumner to abandon his belief
in the effect of climate in producing subspecific characters, at
least over a few generations.

xviii. Macdougall (1927, 1930). The inheritance of train-
ing in rats.

Macdougall has presented evidence to show that rats
trained over twenty-three generations may be definitely modi-
fied. The animals had to escape from a tank full of water.
They could attempt to escape either at a lighted platform (in
which case they received an electric shock) or at an unlighted
one (without a shock). There was evidence that the number
of mistakes made by the rats before they chose the exit where
they did not receive a shock was gradually reduced with each
generation. The data are not treated statistically, but seem
convincing. They have been criticised by Sonneborn (1931)
on various technical grounds, especially that there may have
been unconscious selection 1 or that the strength of the shock
varied. We believe that Macdougall has made out a good
prima facie case, but confirmation is required. Somewhat

1 But cf. Rhine, J. B., and Macdougall, W., 1933, Brit. J.Psychol. (Gen. Section),
24, pp. 2 13-235. (The authors show that in fourteen generations selected adversely

^4., pp. •£ 13— ^35. ^ i. 11c auuiuis snuw uiai 111 il
for ability, marked improvement took place.)

THE ORIGIN OF VARIATION 41

similar results claimed to have been established by Pavlov
have now been withdrawn by the author (see Macdougall,

1927)-

xix. Harrison (1927). Oviposition of the sawfly Pontania

salicis.

Harrison found that the galls of this sawfly in any limited
area tended to occur on only one species of willow, though in
the whole range of the sawfly many species of willow were
attacked. He therefore took sawfly galls from one willow and
exposed them in a locality where only another species was
available. In the most convincing experiment, in the first
year there were few galls and many of these aborted, but later
the sawfly became entirely attached to the new host, and, when
tested five years later, the original host was no longer attractive.
Harrison regards these experiments as a proof that an induced
habit-change is inherited. It is possible to regard them as an
evidence for ' larval memory ' (see remarks on biological races,
pp. 50-52), the oviposition of the females being influenced
by the nature of the site in which their larval life was spent.
It has been suggested that crosses between certain moths show
that oviposition-response segregates as a typical unit character
and that therefore such responses must always be germinally
fixed. We do not doubt that in many cases the oviposition-
response is germinally fixed and it is possible that the temporary
fixture by means of ' larval memory ' is a sort of ' half-way
house.' But this is by no means proved, and, indeed, any
attempt at direct proof would be likely to meet with invincible
technical difficulties.

xx. Tornier and Milewski (literature in MacBride, 1924).
Experiments with ' fancy ' types of Carassius (Goldfish).

Certain domesticated ' fancy ' breeds of Goldfish have been
cultivated for a long time in China and Japan. They are
characterised principally by abnormal development of the
fins and the snout and head and by certain colour-changes
(Ryukin and Ronchu types, e.g.). In the course of a long
period of culture these aberrant types have been detected, iso-
lated and bred from for ' fancy ' purposes. It is said that they
breed true, but how far this is a fact is uncertain. In experi-
ment {e.g. Milewski's) they seem to be relatively unstable.
Tornier discovered by experiment, both on Carassius and other
forms, that the abnormal structural features were due to

42 THE VARIATION OF ANIMALS IN NATURE

scarcity of oxygen, and it is a fair inference that the original
' mutants ' were produced by the unhygienic conditions of the
culture practised in China, in which oxygen-starvation in
particular was inevitable. We need not detail the particular
action on the growing embryos of the oxygen-starvation, its
specific effect on particular structures (to which Tornier
devoted some very interesting study), nor Tornier's special
theory of ' plasma-weakness,' which he held to be ultimately
responsible for the malformations in question. What is not
apparent from Tornier's and Milewski's experiments is that
specific malformations produced under observation by a
verifiable environmental factor are regularly transmitted to
the offspring. It seems quite clear from some of Milewski's
experiments (MacBride, I.e. p. 8) that embryos of one of
the ' mutant ' types (the Ryukin) born in conditions of oxygen-
starvation, but reared in oxygenated water, give a high per-
centage of ' Ryukin ' types (80 per cent.). It is nevertheless
by no means apparent how far the experimental animals were
genetically pure. If the regular causation of the abnormal
condition by specific environmental factors is established, and
if the original abnormal breeds were indeed produced by this
cause, we might be disposed to admit that the inheritance of
the character in control conditions suggests that an induced
modification had acquired some degree of stability. But it is
very uncertain how far we can eliminate an original selection
in the production of the ' fancy ' types. The relative in-
stability of these forms under experimental conditions makes
it very difficult to judge the value of this work.

We may sum up this survey of the experimental work by
concluding that there is a small amount of evidence that
induced modifications of certain types may be inherited. We
shall defer any further discussion until we have considered
the circumstantial evidence.

(c) Circumstantial Evidence. — There is a large body of
observations and suggestions for consideration under this
head. There are two principal groups which are available
for examination.

(A) The effects of use and disuse. — So much evidence (of a
sort) is available from human heredity that the effects of use
and disuse are not inherited that it might seem superfluous
to discuss this question. Nevertheless the matter cannot be

THE ORIGIN OF VARIATION 43

dismissed without some discussion. A single case will make
the difficulty clear. Duerden (1920) has shown that the
sternal, alar, etc., callosities of the ostrich, which are un-
doubtedly related to the crouching posture of the bird, appear
in the embryo. The case is analogous to the thickening of
the soles of the feet of the human embryo attributed by Darwin
(190 1, p. 49) ' to the inherited effects of pressure.' As
Detlefsen (I.e. p. 248) points out, this would have to be ex-
plained on selectionist grounds by the assumption that it was
of advantage to have the callosities, as it were, preformed at
the place at which they were required in the adult. But it is
a large assumption that variations would arise at these spots
and nowhere else.

Detlefsen (I.e. p. 248) reasonably asks ' why it is necessary
to have these anticipatory hereditary callosities appear in the
embryo before there is any demand made upon the organism
. . . why do they not recur in each lifetime entirely through
individual adaptation, as indeed it appears they can ? . . .
What advantage . . . would an inherited callosity . . . have
over an equally effective ontogenetic one ? ' We cannot see
what selective advantage is involved in having them formed
so early, unless we appeal to some principle of ' developmental
convenience.' Moreover, Detlefsen asks (p. 250), ' is it not
extremely improbable that chance variations in the germ plasm
would arise to determine such callosities at exactly those
points and nowhere else ? Why not fortuitous variations for
callosities elsewhere or almost anywhere on the body — which
should persist, for they would have little or no lethal effect in
the process of natural selection ? '

This is a particularly interesting, but at the same time a
very baffling case. On the one hand we have the general
lack of evidence to support the ' Lamarckian ' explanation ;
on the other the apparent absence of any advantage in having
the formation of the callosities pushed back in ontogeny to a
stage preceding the period at which they come into use. One
might construct a purely hypothetical explanation in terms
of ' developmental convenience ' to make a case for selection,
but it would have very little weight if it were not supported
by an exact and intimate study of this particular ontogeny.

The blindness of cave-animals and deep-seaforms (cf.p. 269,
Chapter VII), and the atrophy of limbs in aquatic mammals are

44 THE VARIATION OF ANIMALS IN NATURE

examples of this kind of difficulty. In general the principle of
physiological economy on which the selective explanation of
atrophy from disuse is based, seems to us very unsatisfactory.
Where an organ or structure is definitely inconvenient in a
new mode of life its disappearance may be expected ; but
when it is merely useless it is very difficult to see how slight
variations in the direction of reduction could be effectively
selected, more especially if they are infrequent.

We do not consider that this line of evidence is particularly
helpful, as it seems incapable of exact examination. Experi-
ment may show us that a given organ does or does not atrophy
through disuse and, if it could be experimentally proved that
complete atrophy took place in conditions in which selection
could be excluded, it would go a long way to proving that the
results of disuse are progressively inherited. Up to date no
such experiments are available. It may be pointed out that
Payne (191 1) subjected sixty-nine generations of Drosophila
ampelophila (melanogaster) to total darkness without any modifi-
cation of the eyes or the reaction to light.

(B) Correlation of environmental differences with structural
divergences known or presumed to be hereditary. — Under this
heading we have a very large body of facts summarised in a
very able fashion by Rensch (1929). This author maintains
that there is much evidence tending to prove that lines of
structural differentiation are very frequently correlated with
environmental ' trends, 5 that there is a ' parallelism ' between
' phenotype ' and genotype of such an order that ' modi-
fications ' can be artificially induced which are the same
as, or more or less the same as, characters known to be ' geno-
typic,' and that there is an inference that such externally
induced * Phanovarietaten ' become genotypically fixed.
He admits (p. 161) that the latter stage in the thesis has never
been quite unexceptionally proved ; but he holds that this
parallelism is of the highest significance. He gives great
weight to the production of identical variants in natural and
comparable experimental environments, but actually there
are very few instances of such parallelism. His evidence,
indeed, consists only of Sumner's (1915) experiments, already
shown (p. 39) to be of dubious value. He does not, however,
refer to the later experimental work of Sumner (cf. 1923), in
which it was shown that ' environmental ' forms of Peromyscus

THE ORIGIN OF VARIATION 45

taken from the desert to a new environment remained un-
changed for two to ten generations. Sumner, in fact (1932),
abandons the theory of the direct environmental origin of
' desert ' pigmentation. On the other hand, the proof that
the racial characters are now germinally fixed does not show
that they were always so. Still, for the present at least, the
Peromyscus experiments cannot be held to favour Rensch's
views. We shall mention later other instances of characters
which are racially diagnostic in some species and known to be
due to external causes in others (cf. Robson, 1928, p. 166).

The alternative explanation to Rensch's hypothesis would be
that all the races whose differences are correlated with trends
are merely ' somatic ' forms or produced by selection. Rensch's
theory is, we hold, no more than suggestive, and in the light of
Sumner's conclusions as to the intensive study of gradients,
perhaps less impressive than is at first sight apparent. It is,
however, more fully examined later on (p. 46). In the same
category is Ekman's theory (191 3) of the origin of the lacustrine
crustacean Limnocalanus macrurus, which, it is claimed, has
arisen in many places from the brackish-water L. gnmaldii
(cf. Gurney, 1923) owing to the progressive freshening of the
lakes which it has occupied since the Glacial Period. This is
a case for which we would require experimental evidence,
especially as the various lacustrine forms are not all similar,
though, as Gurney admits in his critical review {I.e. p. 428),
the tendency has been in the same direction.

Rensch's data (with some supplementary evidence) may
now be considered in detail. Some of his most important
conclusions in the present connection are summarised in the
following three rules :

(1) Bergmanrfs Law. — In nearly related warm-blooded
animals the larger live in the north, the smaller in the south.
This is also true to some extent of invertebrates, provided they
are compared within their optimum range, outside which
dwarfing may appear.

(2) Allen's Law. — The feet, ears, and tails of mammals
tend to be shorter in colder climates, when closely allied forms
are compared.

(3) Gloger's Law. — Southern races tend to be black, brown,
grey and especially rust-red ; northern forms are paler and
greyer. Humidity here has an important modifying influence.

46 THE VARIATION OF ANIMALS IN NATURE

The whole weight of Rensch's argument depends, of course,
on accumulating a large number of examples which it is not
desirable to reproduce in the present chapter.

His first point is the extremely gradual changes shown by
geographical races arranged along a climatic trend, e.g. in
the five races of Parus atricapillus between North Siberia and
the Rhine district the mean wing-length changes regularly
from 66-5 to 60 -5 mm. The changes are quite regular and
even the two extremes vary enough to overlap. A number of
similar examples is quoted. He stresses the fact that geo-
graphical variants are normally distinguished by several
characters rather than by one major character. Next are set
out numerous instances of parallel geographical variation,
showing that in any one district related forms tend to be all
modified in one direction. Examples in the Vertebrata,
Crustacea and Mollusca are given. Allen's Law concerning the
relative lengths of projecting parts is illustrated by a number
of tables. It is seen to hold for the tail-length in a variety of
mammals (chiefly rodents), when Alpine or northern races
are compared with the representative race occurring in
warmer districts. Interesting tables (pp. 149-15 1) show the
same relation between wing- and body-length in North
American Picidae, Bubonidae, etc. In 80 per cent, of the
species the wings are longer in the southern races. On p. 152
he turns to Gloger's Law, and in a table on p. 155 he shows its
application to twenty-five races of nine species of European
tits and tree-creepers. It is naturally more difficult to grade
species accurately according to colour.

Rensch has collected together a bigger body of information
of this sort than has ever been presented before. For more
detailed information his book and bibliography must be
consulted.

Certain analogous or additional examples, not mentioned
or not fully treated by Rensch, may be added. Alpatov ( 1 925,
1929) has recorded some interesting investigations on the Honey
Bee [Apis mellifera) in Europe. The data are very extensive
and have been subjected to rigorous statistical treatment.
He has been able to show that southern forms are smaller on
the average and have longer tongues, wider wings, longer
legs and small wax glands. The number of hooks on the
hind wings is greater and the colour is yellower. The change

THE ORIGIN OF VARIATION 47

from north to south is extraordinarily gradual, and except
when extreme forms are compared, it is only the averages that
differ. The comparisons must be made between individuals
occurring on the same line of longitude, the change occurring
more quickly in the east than in the west, the lines of equal
change probably corresponding with the summer isotherms,
which have a similar slope. Exceptions to this regularity are
found only in the Caucasus, where development of a special
geographical race of the Honey Bee introduces a fresh compli-
cation. Experiments on the effects of temperature are as yet
not very extensive, but they, like the seasonal changes observed
in Apis, show that cold produces artificially the same effects as
those found in nature. Nevertheless transportation experi-
ments have shown that the various naturally occurring types
are to a large extent hereditary. Alpatov's attempt to explain
his results is so characteristic of the orthodox way of dealing
with such facts, that it is worth setting out at some length.

The possibility of the inheritance of a long-continued
environmental effect is dismissed ' because of the lack of any
credible experimental evidence of the inheritance of acquired
characters.' He goes on to attempt to show that the observed
characteristics of the southern forms may really be adaptive.
Thus ' the longer tongue of southern bees is probably con-
nected with the peculiarities of nectar secretion in the south
as compared with the more northern localities. Michailov
suggested that the longer tongue of southern bees is an adapta-
tion to dry conditions which lead to a lower level of nectar in
the south, and thus compel the bee to have a longer working
organ. We expressed the hypothesis that the southern bees
are obliged to have a longer tongue, not only because of a
lower nectar level, but also because of a probable difference
in the composition of the whole nectar-secreting flora. It
has been reported by many beekeepers that the southern, and
particularly the Caucasian, bees can fly longer distances
gathering nectar, and it is probable that in consequence their
wings are more developed and have a larger number of hooks.
The smaller size of the wax glands is probably connected with
the condition that the bees in the south have perhaps less
need to work upon the reinforcement of their nest. Hence
the difference in the tongues, the wings and the wax glands
(also probably in the first joint of the tarsus of the last pair of

48 THE VARIATION OF ANIMALS IN NATURE

legs) may be considered as adaptations of different biological
ends. It is probable that these characters have been developed
by means of natural selection. Other characteristics like the
general size of the body and coloration cannot at the present
moment be even hypothetically evaluated as having any
biological importance for the organism.'

Allen's Law is corroborated in the diminished length of
the tails of the island races of certain British mice. It is
possible, however, that there is a further special effect due to
island life. In all the British mice and shrews which have
races in the Shetlands, Orkneys and Hebrides, the proportion
of the tail-length to that of the head and body is almost
invariably less in the island races (figures taken from Barrett-
Hamilton and Hinton, '1910-21). Unfortunately, the majority
of the insular forms are found in the north, while the measure-
ments of the mainland forms were based on southern speci-
mens, so that it is not possible to separate the effects of latitude
from those of insular life.

Le Souef (1930) has published an interesting note on the
changes of three species of wallaby and an opossum imported
sixty years ago from Australia to New Zealand. All have
varied in the same way, the fur being now longer, more silky,
and less dense.

Rensch's ornithological examples illustrating Gloger's Law
may be supplemented by the data brought forward by Banks
(1925). Here a number of subspecies or of specimens from
different parts of the range of a species were compared and
their colours correlated with the average meteorological con-
ditions obtained during the breeding season. A very general
positive correlation was found to exist between temperature and
dark colours. The relation between pigmentation and humidity
is not nearly so simple, being sometimes inverse, sometimes
direct, but it appears in any case that the darkening effect of
higher temperature is evident only in the presence of a moderate
humidity. In areas which are very dry the colours tend to
be pale in spite of a high temperature. This general result
agrees with the well-known results which Beebe (1907) found
by experiment. Dealing with doves of the genus Scardafella
he found first that, in nature, there was a regular increase in
dark pigment as one passes from Mexico to Brazil, the centre
of least pigmentation being the driest area and the pigment

50 THE VARIATION OF ANIMALS IN NATURE

increasing in either direction as the humidity increased. By
exposing the lightly pigmented form to very humid conditions
he was able to show that pigment was slowly acquired through
a course of moults, till finally a stage was reached darker than
any known in nature. Of other examples of the correlation

Fig. 2. — Correlation of Yellow Markings with Climatic Conditions in
the Wasp Polistes foederata (? above, o* below). The Yellow Markings
increase in Warm, Dry Areas.

(From Zimmermann, 1931.)

of structural characters and environmental trends one of the
best known is that of the number of vertebrae which is associated
with a temperature trend in the Atlantic Cod (Schmidt, 1930).
The correlation of colour and temperature is also known in
insects, e.g. in Polistes (Zimmermann, 1930, 1931).

(d) Habit -formation. — The habits and instincts of
animals are largely responsible for bringing their heredi-
tary make-up into play with the environment. No account

THE ORIGIN OF VARIATION 51

of the evolution of the more highly organised animals can be
complete that does not explain the evolution of instinct as
fully as that of structure. Unfortunately, this is a matter of
which we are largely ignorant. Instincts are less fixed than
structures and their heredity and modifications are much more
difficult to study accurately. Undoubtedly in the vertebrates
it becomes difficult to distinguish inherited aptitudes from
traditions handed directly from one generation to the next.
Even in Arthropods there may well be a bigger element of
tradition (of rather special sort) than has usually been
allowed.

As an introduction to the subject we will consider the
predacious habits of the wasps of the family Crabronidae
which have been discussed by Hamm and Richards (1926).
Most species, in the store of dead or paralysed prey laid up
for their offspring, include flies, but particular species capture
insects of most of the more important orders. In a few species
members of two or more orders are mixed, while in others
there is great specialisation, the prey being sometimes practi-
cally restricted to one sex of one genus. From the present
point of view the most interesting species are those which,
while tending to specialise on one kind of prey, always capture
some or many individuals of a widely different systematic
category. Such a species is Crabro leucostomus, which always
captures a high proportion of Stratiomyid flies, but includes
also Diptera belonging to a large number of other families.
This habit is independent of the habitat in which the wasp
is nesting. The behaviour of C. leucostomus suggests rather
a special form of ' larval memory,' the insect having a
tendency to capture the food that it received during its own
larval life, as has been suggested by Wheeler (1923, p. 57), who
points out that the central nervous system is almost the only
larval structure not radically modified at metamorphosis.
Such a larval memory would not at first be hereditary in
the ordinary sense, but may well have become so in those
species which are now strictly specialised.

The general question of biological races in Arthropods is
reviewed elsewhere (p. 119) and the facts need only be briefly
dealt with here. The most important conclusions are the
following :

(1) There are numerous instances of species which are

52 THE VARIATION OF ANIMALS IN NATURE

divided into one or more strains differing little, if at all,
morphologically, but with different habits, e.g. different larval
food, host of parasite, etc.

(2) Experiment has shown that, if one race is, for instance,
forcibly maintained on the food of another, there is at first
little oviposition or breeding and a heavy mortality. Often
a few individuals manage to perpetuate the race on the new
food, to which it eventually becomes adapted. This would
suggest at first sight that there has been a selection of a suitable
stock, but this interpretation breaks down (cf. (3) ).

(3) It has been possible in several cases to take a race A,
normally feeding on a food a, and adapt it to the food b of
race B. In these circumstances it may be as difficult to make
the survivors of A (on b) return to a as it was to make the
original change. This does not fit in with the theory of strain-
selection, and Thorpe (1930) definitely postulates a process of
more or less permanent habit-modification. It is by no means
necessary that this change should at first be incorporated into
the normal hereditary mechanism as claimed by Harrison

(1927)-

(4) There is some evidence that there is a tendency for these

biological strains to mate within the race and therefore to
stabilise their constitution.

From the evolutionary point of view, instinct is, as we have
said, a particularly important subject, but unfortunately we
do not know how new instincts arise nor how they are in-
herited. When an instinct is very firmly established it is
naturally handed from parent to offspring like any somatic
character, but this is by no means necessarily the case in the
early stages of instinct-acquirement. In the songs of birds,
for instance, while there is a large hereditary element, there is
also much that is local and individual. Yet it is very plausible
to suggest that the former element was originally built up from
the latter without, in all cases, the actual selection of individuals
with a particular song-type.

A further phenomenon which may be considered under
the heading of habit-formation is that of voltinism in insects.
The subject has recently been discussed in an interesting
paper by Dawson (1931), who summarises the main theories
and presents some very valuable experimental data (see also
Baumberger, 19 17). The problem is seen at its clearest in

THE ORIGIN OF VARIATION 53

temperate countries in the many insects which have two or
more broods a year. In these the pupae from the early broods
produce adults in the same year, whereas the pupae of the
last brood hibernate. It is difficult to imagine how any such
system could keep in step with climatic seasonal changes, if it
were not ultimately controlled by temperature or some other
climatic variable. When, however, the determinative factors
are investigated experimentally, a very perplexing state of
affairs is laid bare, recalling in detail the complex problem of
seasonal variation in colour. There is little doubt that the
gradual sinking of the mean temperature in the autumn is the
main controlling factor. Pupae which have been exposed to
such a gradual cooling tend to become dormant. But even in
one family (of brothers and sisters) the effect is not uniform ;
in a number of experiments some individuals become dormant,
while others do not. Probably genetic factors partly determine
the response to temperature, but Dawson was unable to find
any simple scheme of segregation. Previously Toyama (1912),
in the Silkworm {Bombyx mori), had suggested matroclinous
inheritance.

In the Cornborer (Pyrausta nubilalis), Babcock (1927) and
Babcock and Vance maintain that ' the seasonal rhythm
is to a certain extent persistent and is due to the formation
of a physiological condition which forces the insect to
develop a certain type of seasonal cycle. This physiological
condition is formed by continued impress of a particular type
of normal environment and persists after the impress of the
environment is removed ' (1929, p. 53). The whole question
of seasonal rhythms in animals is still in urgent need of
experimental investigation.

Since the genesis of instinct is still so obscure there is some
value in putting on record a number of instances of aberra-
tions in instinct. Some of these appear to be merely individual,
but others have been more widely manifested. In birds and
mammals, where social tradition has some weight, even
individual aberrations have importance.

Insects. — One of the best known instances of a sudden
change in habits is that of an English bug, Plesiocoris rugicollis,
which, prior to 191 8, was known to feed only on willow, but
since that date has increasingly turned its attention to apple,
so that it is now a serious pest. The flies which ' blow ' sheep

54 THE VARIATION OF ANIMALS IN NATURE

in Australia did not become a serious pest till about 1895,
apparently owing to a definite change in habits (references in
Carpenter, 1928, pp. 111-113). Manhardt (1930) records
that a beetle, Luperus xanthopus, after stripping all the willows
on the banks of the Elbe, made its way inland in large numbers
and attacked fruit trees. In some parts very serious losses
resulted. In view of what has been recorded as to the forma-
tion of biological races, such invasions have some significance.
Still more individual aberrations are seen in the genus Vespa,
where species normally subterranean sometimes nest above
ground and vice versa (see Stelfox, 1930).

Mollusca. — An octopus (Bristowe, 1931 ; Robson, 1932a)
was found eating spiders, though the diet is normally restricted
to Crustacea.

Limax maximus (Taylor, 1907) is usually found in gardens
or near houses, but in Ireland is never found in cultivated
ground or gardens.

Reptilia. — Lacerta muralis according to Eisentraut (1929)
is found on the shore in the Balearic Islands, feeding on
Halophytes because the normal supply of insects and snails
is reduced.

Birds. — The Black-headed Gull (Larus ridibundus) (Lack,
1933) sometimes feeds on land in spite of its adaptations to
aquatic feeding. The same species (Gray, 1930, p. 170) has
been observed flying in a V-formation like geese. This is
very unusual for the species.

The Reed Bunting (Emberiza schoeniclus) (Lack, 1933) is
typically a marsh form, but is very occasionally found nesting
in typical Yellow Bunting habitats.

The Great Tit {Parus major) (Darwin, 1884, p. 141) some-
times behaves like a shrike and kills small birds. Darwin
gives further examples of habit-anomaly on the same page.

The New Zealand Parrot (Nestor notabilis) (Buller, 1888,
pp. 244-5) was originally insectivorous, but relatively recently
began to attack sheep.

The Barbet (Trachyponus emini) (Loveridge, 1928, p. 41)
nearly always nests in burrows, but was once found nesting in
a tree.

Mammals. — The African Buffalo (Bubalis coffer) (Elton,
1927, p. 145) used to be a diurnal feeder, but after the rinder-
pest epidemic of 1890 became a much more nocturnal feeder.

THE ORIGIN OF VARIATION 55

In a highly adaptable mammal like the Grey Squirel (Sciurus
carolinensis) (see Middleton, 1931) almost endless variations in
habit are recorded, e.g. in food, use of burrows instead of trees,
etc.

These data suggest that the fundamental genetic basis of
behaviour is very easily modified by the environment. It also
appears to be subject to spontaneous change, though the
origin of this change is obscure. It is similarly difficult to
distinguish the various roles of heredity and tradition. Some
authors have suggested that ' traditions ' ultimately become
hereditarily fixed.

(e) Summary of Data on the Inheritance of Induced
Modifications. — Much of the experimental evidence is un-
satisfactory, but it is difficult to avoid the impression that some
types of impressed modifications are in certain circumstances
inherited.

The indirect evidence appears to require one of three
possible hypotheses :

(a) That the modifications are all mere fluctuations. This
is scarcely tenable.

(b) That where the modifications are inheritable, it is due
to the selection of adapted variants.

(c) That acquired modifications, long impressed, have
become inherited.

A serious objection is brought forward by those who hold
that in any particular case the correlation between the varia-
tion and the environment may be due merely to the selection
of variants best suited to that environment. This objection is,
quite literally, unanswerable, but it assumes what can never
be proved, at any rate with our present knowledge. It is a very
large assumption to maintain that a graded series of variations
in a species corresponds to a parallel gradient of adaptation
to the altering environment, if only because of the extra-
ordinarily discriminative selection required. It appears to us
that neither of these rival theories can be dismissed by a priori
argument. Both are possible, both are at present incapable of
final proof and must in each case be judged by the balance of
the evidence. The extent to which the discriminative power
of Natural Selection is developed is discussed in more detail
elsewhere (Chapter VI I). We shall merely record our opinion
that an adaptive explanation of much of the data on pp. 44-50 is

56 THE VARIATION OF ANIMALS IN NATURE

unconvincing. At the same time we do not pretend that the
evidence available suggests that any ' Lamarckian ' process is
very important as a source of new heritable variation, except
possibly in the matter of habits. There is certainly a very large
body of evidence (Chapters IV and VII) suggesting that the
bulk of the morphological differences between species and races
is not in any way correlated with a particular environment ;
and conversely that many species and subspecies range widely
without any modification. Although this seems in conflict
with the evidence for geographical trends (p. 46), yet such
trends are relatively uncommon (i.e. compared with the
number of races and species not arranged in trends) and
further usually only some of the characters of a species exhibit
the trend.

Conclusions.

In this chapter we have considered the origin of the various
types of variation that may be encountered in a natural popu-
lation. Fluctuations certainly form a large element, but
quantitative data as to the importance of these are hard to
obtain. Genetically determined variations include (in addition
to gene-mutations) changes due to fragmentation, etc., of
chromosomes, polyploidy and recombination. The first two
phenomena seem to be of minor importance in animals. Re-
combination is certainly responsible for much of the normally
wide range in phenotypes. We have not much evidence
yet whether species in nature are often heterozygous for
more than a few characters. If they are not, the results of
recombination are strictly limited, especially in any particular
direction. In any case Lotsy's theory of evolution by crossing
cannot have much application in the animal kingdom, where
successful interspecific crosses are relatively uncommon.
Gene-mutations are certainly a very important source (or, as
some would have it, the only source) of new hereditary material.
The real cause of gene-mutations is quite unknown, but it is
theoretically improbable that they are in any real sense
spontaneous. The rate at which they occur has now been
influenced by X-rays, radium-rays and high temperature.
Even under these influences the rate is still relatively low.

The problem of the inheritance of induced modifications
appears to be ultimately reducible to the question whether

THE ORIGIN OF VARIATION 57

there is a process by which the hereditary basis handed on to
the next generation may be gradually altered, as opposed to the
apparently sudden induction of mutants. The actual experi-
mental evidence is not very conclusive, except in so far as it
shows that lesions and mutilations are not inherited. The
problem of the degeneration of disused organs requires further
consideration. There is no positive evidence that disuse has
a direct effect, but the alternative selectionist explanations are
equally unsatisfactory.

In a few cases there is experimental evidence which suggests
that induced modifications are inherited, but confirmatory
experiments are much to be desired. There is also a con-
siderable body of indirect evidence which may be held to
support the experiments. In a number of instances alternative
adaptational explanations of the data have been (or could be)
put forward. Such explanations depend on very large as-
sumptions as to the closeness of the adaptation of the organism
to its environment. The prime difficulty of the assumption
that induced modifications are inherited lies in explaining how
the modified character comes ultimately to be represented in
the germ cells.

CHAPTER III

THE CATEGORIES OF VARIANT INDIVIDUALS

Biological inquiries in general involve recognising that
individual animals may be grouped in various ways, and
in investigations of variation, heredity and evolution the
characteristics of such groups are the subject of inquiry and
the measure of divergence. Investigation of the nature and
status of these groups and their relationship one with another
is an indispensable preliminary to the study upon which we
are engaged.

The levels of evolutionary divergence most usually indi-
cated by the species and variety have been subjected since
Darwin's time to a careful scrutiny from divers points of view
and numerous categories have been proposed to designate
groupings of individuals other than the traditional species
and variety of taxonomy. Historically we may date the
commencement of serious analysis to Alexis Jordan's publi-
cation of his work on elementary species, and to such pioneer
work as Waagen and Neumayr's studies of ' Formenreihe.'
The conception of geographical races may be dated to earlier
workers (Kant, Pallas, Gloger [cf. Rensch, 1929) ).

An admirable study of the lowest systematic categories
has been published by du Rietz (1930), who discusses criti-
cally the status of the various groups proposed and the syno-
nymy of the terms used, du Rietz's list is defective in one
or two important respects. He discusses neither palaeonto-
logical categories nor physiological differentiation, nor does
his survey, which is mainly based on botanical data, include
such divisions as colonies, etc.

The most commonly recognised categories are, of course,
those used in taxonomy. In addition there are a number of
others in regular use in various branches of zoology, which
either have not been absorbed into the hierarchy of systematic

THE CATEGORIES OF VARIANT INDIVIDUALS 59

terms or are only rarely used by systematists. But, although
the majority of systematists still maintain the traditional
Linnean categories, many feel impelled to supplement them
with other terms devised to fit special groups revealed by
systematic analysis or to attempt to substitute for the older
categories of species and varieties fresh ones designed to bring
systematic procedure into line with new methods of analysis,
(e.g. Linneon and Jordanon (Lotsy), ' Formenkreise ' and
' Rassenkreise ' (Kleinschmidt, Rensch) ) .

The following appear to be the chief types of category
that have been proposed :

Taxonomic.

Palaontological (lineage, gens).

Geographical (local race, colony, ' Rassenkreis ').

Genetical and Reproductive (e.g. pure line, biotype, clone,
syngameon, sibship).

Physiological (strain, physiological race).

Although for the purpose of convenient discussion we have
adopted the above distinctions, it will be noticed that a hard
and fast division between, e.g., genetical and geographical
categories is fundamentally arbitrary. All we wish to imply
by these distinctions is that various methods of research have
led to the adoption of various categories which we have to
define and relate one to another.

Over and above these we have the various terms which
perhaps could be classed as genetical by which heritability,
partial heritability or non-heritability is implied, such as
forma, alteration, Dauermodifikation, genotype and phenotype.
There is also a category of groups, partly of geographical,
partly of habitudinal significance such as the school, rookery,
shoal, etc. Some categories are based on more than one
concept, e.g. the ecotype, and ecospecies are groups recognised
on account of genetical behaviour and ecological relationship.
Lastly we may point out that some categories are strictly
classificatory, i.e. they form part of a system and designate
a more or less closed group, though they are not all in current
taxonomic use, while others, such as lineage, are taxonomically
neutral, i.e. they involve no recognition of a classificatory
system. Of the same order is the term population or
natural population, which is used to designate any number
of closely related and interbreeding individuals occupying

60 THE VARIATION OF ANIMALS IN NATURE

a given area, without any taxonomic specification of the status
of the variants it contains. 1

We are thus presented with very many different kinds of
groups, which seem to reflect various modes of divergence in
nature and it is desirable to ascertain what is their relationship
one with another, and what light they throw on the actual
process of divergence itself.

du Rietz in the paper mentioned above suggests (p. 337)
that the most elementary unit of taxonomy is the individual.
He points out that the limits of the individual are not always
easy to define, but he thinks that the soundest definition
involves the recognition of physiological autonomy. We
believe, however, that the analysis might be pressed further.
To suggest that the character is the most fundamental unit
is to open the door to all kinds of complications, chief
among which is that the limits of characters are usually
very hard to define ; but the suggestion has a particular value
from our point of view. Evolution is essentially a matter
of character-changes. Individuals are bundles of characters
which have each a history of their own, and the divergent
groups manifest a progressive accumulation of character-
divergences. It is a matter of more than academic or formal
interest to keep the individual character before our minds
throughout this discussion (cf. lineages, p. 65) and to re-
member that the individual maybe resolved into its constituent
elements ('structural units' — Swinnerton, 1921, p. 358).
The organism has its peculiar autonomy and ' wholeness,'
but each of its structural units has an individual history of
change which, though related to the needs of the whole or-
ganism, can be treated as a separate evolutionary episode.
It is also of very great importance to remember the individual
character in considering the processes by which we recognise
groups of individual organisms such as species, etc. It is
not perhaps sufficiently realised how much variation is attain-
able, if all the possible characters are taken into account.
A. Agassiz (1881, pp. 18-19) pointed out that in the Echinoids
the number of variable structural items is at least twenty and
that the permutations and combinations of the most restricted

1 ' Population ' is sometimes used in the sense of ' sample ' in describing local
collections made from a larger assemblage. Thus Schmidt (1930, pi. 1) alludes
to the population of the Atlantic Cod, though he uses the word ' sample ' in the
text.

THE CATEGORIES OF VARIANT INDIVIDUALS 61

types of variation are 2 19 . Henry (1928, p. 65) has shown
that the chance that two human individuals will have the
same finger-print pattern for a given digit of one hand is of
the order of over 1,000,000 : 1 [cf. p. 24, supra).

I. Taxonomig Categories

The Linnean hierarchy of morphological groups of which
the species and variety are members is still the system by
which we express an animal's relationships. We do not wish
to discuss the general principles according to which this system
is constructed and its capacity to express animal relationships.
We may suspect with Bather (1927, p. ci) ' that the whole of our
system is riddled through and through with polyphyly and
convergence,' and we may agree that the chief and most philo-
sophic duty of the systematist is to ' free it from this reproach '
(Bather, I.e.), even if this task presents difficulties which
may be occasionally insuperable (Robson, 1932).

Nevertheless the species and the variety or subspecies are
the most frequently used categories, and they are the reference
points round which all the data as to habits, distribution and
variation have been assembled. It will be as well, therefore,
to commence our survey with them. The status of the species
has, of course, been subjected to long and painful inquiry.
It has been challenged on two principal counts — (a) that it is
an arbitrary abstraction from a number of individuals which
vary so much inter se that any grouping must do violence to
the natural divergences that are found both in time and place ;
and (b) that it is not a group having regularly definable proper-
ties and a standardised status vis-a-vis other groups. The
first of these objections questions the capacity of the systematist
to designate any part of a more or less continuous natural
assemblage, the second criticises the status of the species in
a hierarchy of classification.

Most biologists are now agreed that the latter objection
is valid and that the species has no standardised attributes by
which it can be distinguished from the variety and the genus.
Such a standardisation, it is true, might be defined by the
acquisition of some qualities constituting critical upward and
downward limits in the process of evolutionary divergence

62 THE VARIATION OF ANIMALS IN NATURE

(e.g. at the lower limit, the intervention of mutual infertility).
But, as organisms diverge in many characters, and as these
are not correlated in any universal scheme of divergence, any
attempt to fix a downward limit fails.

The first objection is far more cognate to our problem.
The universal occurrence of individual variation has led
certain writers to assert that the individual is the only real
unit and that species and similar groups are devoid of any
significance. This view is worth dwelling on for a moment,
as its importance is not fully recognised. Finding agreement
between the members of his species in a limited number of
characters the systematist has perhaps given undue prominence
to them. When the term similarity is introduced into the
definitions of systematic units, we may well ask if any two indi-
viduals, even of a moderately complex phylum, are ever alike in
all their characters (cf. p. 60, supra) . If this is never the case, we
may also ask how it is that any discrete groups, such as species,
have come to be recognised and what may be the value of
a classification that recognises such crude groupings. The
answer to this may be given briefly. In spite of very extensive
individual variation (a great part of which is of unknown
hereditary status and may be non-heritable), the systematist
tends to find certain regular correlations, associations of a
limited number of characters that occur regularly in individuals,
and it is this correlation that, amid a very great amount of
individual variation, constitutes the basis of species-diagnosis.
Such correlations are, of course, of very varying intensity and
can involve a greater or less number of characters of various
kinds ; but, though they cannot be standardised as a univer-
sally recognisable grade, the taxonomic procedure is justified.
It is necessary to make the proviso that a number of species
in each group are founded on inadequate statistical data.
Indeed so great is the disparity between the number of species
described by the systematist and the knowledge of natural
variation of the populations from which species are abstracted,
that some systematists (e.g. Ramsbottom, 1926, p. 28) have
been impelled to draw a distinction between ' the natural
species ' and ' the taxonomic species,' and one of the authors
of the present volume has suggested that forms which, by
reason of the poverty of material, imperfect preservation,
or the lack of adult specimens, are of uncertain status, though

THE CATEGORIES OF VARIANT INDIVIDUALS 63

seemingly distinct species, should be referred to by a symbol
rather than by a specific name.

It must be remembered that not a great deal is known
concerning the hereditary stability of species. It has always
been assumed, since the contrast between hereditary and non-
hereditary characters was realised, that the characters of the
species were hereditarily stable. Naturally few taxonomists
have had the time or opportunity to breed out the members of
groups which they have confidently described as species. A
substantial number of described species are forms of dubious
hereditary stability. ' Environmental forms ' are often given
distinct specific names, as in the case of Artemia salina and
A. milhauseni and in various groups of Cladocera and Mollusca
(e.g. cf. Miller, 1922). Finally, in claiming a general validity
for taxonomic procedure in the treatment of species as distinct
groups, we recognise that this claim must be limited by the
admission not only that such groups are of various degrees
of distinctness in the number of divergent characters, but also
that sometimes intergradation between the various elements
in a population may be so complete as to render the limits
between species purely arbitrary.

Within the species itself systematists are accustomed to
recognise certain subdivisions — the subspecies, the variety,
and less frequently the form and the race. At the present
time the terms variety and subspecies are both used for the
major subdivisions of the species, but speaking generally they
have a different connotation. The subspecies is a term in
regular use among mammalogists and ornithologists, and it
is used essentially to denote a geographical entity, the major
subdivisions of the species of birds and mammals having
usually distinct geographical ranges. The term variety, 1 on
the other hand, though it is used for a major division of the
species of invertebrate animals, has no such geographical
implication. In many invertebrate groups the subdivisions
are types which occur sporadically throughout the range of
the species, and though in morphological status they correspond
to the subspecies of birds and mammals, the accidental

1 Rothschild and Jordan (1903) have used the term variety not for any
particular category of the components of a species, but for ' all the members of
a species indiscriminately.' The different categories of varieties are given special
names or symbols.

64 THE VARIATION OF ANIMALS IN NATURE

difference in terminology conceals a real difference in the
type of variation (i.e. in distribution).

Below the level of varieties and subspecies the ordinary
task of the systematist is not pursued. All that we have
said concerning the validity of the species-concept applies
with equal truth to the subdivisions of the species itself, viz.
the uncertainty as to their genetic status and the difficulty of
standardising the concepts.

It remains for us to notice the various attempts that have
been made to incorporate the results of population-analysis
into taxonomy. A good account of this is given by du Rietz
(I.e.), who reviewed and attempted to harmonise all the various
terms proposed. It is enough to state that intensive popula-
tion-analysis (dating from Alexis Jordan's pioneer work) has
revealed the presence within systematic species of various
subordinate elements which are imperfectly represented by
the old terms variety and subspecies. It is clear that there is a
basic distinction, now generally recognised and described in
detail by du Rietz (I.e., pp. 349-354), between a population
forming a local (variety) as opposed to a regional (subspecies)
element in a species. The extent to which the Jordanon
(Lotsy), microspecies and elementary species (Jordan), natio
(Semenov-Tian-Shansky), etc., are merely synonymous with
one or the other of these is an academic point, and it is similarly
obvious that the line between ' local ' and ' geographical '
race is quite arbitrary. The differentiation of populations
into a large number of intercrossing ' biotypes ' and the way
in which such subordinate elements are distinguished by
isolation lead to a very finely graded hierarchy of local
groupings (cf. Crampton, 1 916-1932 ; Gulick, 1905 ; Heincke,
1898), and it would be undesirable to attempt to define
these by a rigid terminology. Some taxonomists have recog-
nised a finer distinction under the name ' forma ' to designate
a purely fluctuational type (— 'modification') or, with a
more non-committal connotation, to designate a type ' occur-
ring sporadically in a species-population and not forming a
distinct local or regional facies in it ' (du Rietz) .

Finally, we would draw attention to the attempt which has
been made by Fenton (1931, p. 30) to remodel the traditional
Linnean system so as to suit the findings of palaeontology.
His definitions of ' subspecies ' and ' form ' are not to be

THE CATEGORIES OF VARIANT INDIVIDUALS 65

commended, as they introduce fresh connotations for terms
which are beginning to acquire a fairly regular meaning.

II. Pal^eontological Categories

Perhaps the most important principle to which we should
refer under this heading is the palaeontological ' time-charac-
ter ' concept. The status of the species in time is as significant
as it is in its modern relationships and is often neglected by
neontologists. Of recent years some noteworthy studies have
been made on series of fossils in which evolutionary change
can be studied intensively through successive horizons. The
technique of this study was formulated by Neumayr and
Waagen ; but its application to series of closely allied forms
has been developed by Carruthers, Rowe, Swinnerton and
Trueman in this country. The essence of the procedure is
the study through a series of successive horizons of series of
closely related forms in terms of their individual characters.
The result of such studies is the concept of the lineage and the
bioseries. The first is a racial complex of lines of descent,
which on account of crossing and biparental reproduction
must, as Swinnerton (1930, p. 387) points out, prove to be not
a series of parallel evolutionary lines, but a finely meshed
network. The bioseries is the historical sequence formed by
the changes in any one character and relates to the modifica-
tion of any single heritable feature. Each line of descent
and each lineage will be composed of numerous bioseries
evolving at different rates, just as each individual is composed
of different characters. In such developmental series ' tran-
sients ' (i.e. individual modes) at stages remote from one
another are as distinct as taxonomic species, e.g. in one
such lineage the Cretaceous sea urchin Micraster has a stage
M. praecursor which could be rated as a distinct species from its
successor M. coranguinum.

There exists some ambiguity as to the relationship between
the ordinary systematic concept of species and the lineage.
But this much is clear — that although within a given lineage
the concept of species is difficult to apply (Trueman, 1930)
because of the difficulty of disentangling the series of ' anasto-
mosing ' lines of descent, yet a given horizon will contain
discrete entities corresponding to systematic species, each of

66 THE VARIATION OF ANIMALS IN NATURE

which represents a stage in a particular lineage. Thus at the
stratigraphical level of the Millstone Grit, Carruthers found
two distinct species of coral, ^aphrentis constricta and £. disjuncta,
though each of these at this horizon represented a stage in an
individual lineage in which the individuals cannot be speci-
fically delimited from individuals that occur in earlier and
later horizons. It seems that the character-complexes, in
which the individual characters in any one lineage are modified
at different rates and so afford no regular correlation by
which species may be recognised, do in fact diverge so that
one lineage may differ from another at a given moment in the
same way as the species of the neontologist differ. In other
words, the investigations of lineages have revealed distinct
divergences equivalent to species, but these divergent groups
show no discontinuity in time from their predecessors or suc-
cessors. The criticism that the forms on which such studies
have been carried out are peculiarly plastic (Robson, 1928)
and therefore apt to be misleading has, we think, been suffi-
ciently answered by Trueman (I.e. p. 307), although there must
always exist some element of doubt as to the relationship
between groups diagnosed on certain plastic characters of the
shell and those founded on more stable characters. Finally,
it must be observed that the existence of lineages could be
suspected from the distribution of variants in modern popula-
tions (cf. p. 176, Chapter VI).

III. Geographical Categories

The subordinate units within the species recognised in
taxonomy and associated with the intensive study of geo-
graphical distribution are somewhat diverse and no standard
usage obtains. There are some outstanding works on the
geographical variation of single species or on allied forms,
such as those of Heincke (1898), Duncker (1896) and Schmidt
(191 8-1 930) (fishes) ; Sumner (1932) (Peromyscus) ; Crampton
(191 6-1 932) (Partula). Alpatov (1924, 1929), Semenov-Tian-
Shansky (1910), Rensch (1929) and others have attempted
to define the terms used.

Mammalogists, ornithologists and, to some extent,
herpetologists regularly subdivide the species into subspecies
or smaller units such as races, all of which are characterised

THE CATEGORIES OF VARIANT INDIVIDUALS 67

by their members occupying a more or less clearly delimited
geographical area. Among the students of invertebrate
groups no such regularity of usage obtains and there is evi-
dently no general tendency, easily detected, for the subordinate
groups to be spatially segregated. We discuss at some length in
Chapter IV the question whether there are any real grounds
for this difference in procedure and its implication. For the
moment we are concerned only with the categories themselves.
How different the procedure among students of invertebrate
groups may be will be seen from the following extracts.

Pilsbry (1919, p. 277), in treating of the subordinate divi-
sions of species of African land snails, distinguishes between
' those of racial value or subspecies in the sense of forms charac-
teristic of geographic areas or habitats,' and ' the different
forms (mutations of de Vries (?) ) occurring together in the
same colonies and doubtless interbreeding.' These he calls
mutations. This usage of ' subspecies ' is found largely among
lepidopterists (but cf. Wheeler, 191 3 (ants) ).

Bequaert (19 19, p. n), who evidently feels that it is not
possible to recognise geographic units of the same status as
those in other groups, uses the term variety for his subordinate
divisions in a ' neutral ' sense, i.e. without any presumption
as to their true status as geographical races or individual aberrations
or elementary species. His varieties oiEumenes maxillosus (African
wasp ; p. 59) seem to occupy separate parts of the range of
the species (p. 60), but they are not to be considered geo-
graphical races, as they ' do not inhabit a given country to
the exclusion of all others.' Here we see geographical units
less distinctly segregated than in other cases, but still perhaps
deserving that status.

The term variety is generally used in dealing with inverte-
brates in the ' neutral ' sense of Bequaert for anything from
a single rather distinctive individual in a limited number of
specimens representing a species to the kind of group seen in
Eumenes maxillosus. It is given regularly to clearly marked
and distinctive groups numerically well represented, the
individuals of which occur as a certain percentage in any part
of the range of a species, but are not restricted to a particular
locality (colour-classes of land snails). There seems to be a
fairly well-established practice of distinguishing between sub-
species and varieties in the sense outlined above according

45

50

15

SO

45

15 30 45 60 75 90 105 120 150

15

30

>S E 30

60

120

1 35

150

Fig. 3. — Map of Distribution of Eumenes maxillosus De G. adapted from
Bequaert (191 9). Fourteen Areas can be distinguished according to
the Colour-variants present, as shown in the following List : —

Area i . Maxillosus, reginus.

2 . Maxillosus , pulchen imus, fenestralis.

3. Maxillosus, fenestralis.

4. Alaxillosus.

5. Maxillosus, pulcherrimus.

6. Maxillosus, fenestralis, tropicalis.

7. Maxillosus, fenestralis, dimidiatipennis.

8. Maxillosus, dimidiatipennis.

9. Dimidiatipennis.

10. Dimidiatipennis, conicus, xanthurus, circinalis,petiolatus.

1 1 . Conicus, petiolatus.

1 2 . Conicus, xanthurus, circinalis, petiolatus.

13. Xanthurus, petiolatus.

14. Petiolatus.

THE CATEGORIES OF VARIANT INDIVIDUALS 69

to whether intergrades occur between the groups. Sub-
species are groups between which intermediates occur only
rarely or not at all (see Dice, 1931 ; Merriam, 191 9, for
conflicting views on this subject).

We have thus quite clearly established the recognition of
more or less distinct geographical groups on the one hand and
groups or types not spatially segregated, but appearing either
as individual variants sporadically throughout a population
or as larger local elements not segregated into geographical
units. We have now to inquire concerning other subdivisions
of this kind.

Races. — The term geographical race is used as a complete
synonym for subspecies by several authors (cf. Alpatov, 1929).
But it is also used for a smaller unit not of the same dimensions
as the subspecies. Local race and local forms {cf. Duncker, 1896)
are used in the same loose way. In fact it will be readily
recognised that such a hierarchy might exist within the species,
that the boundaries of the various groups would be difficult
to draw and there would be some confusion of terminology.

That such a hierarchy of local or geographical groups does
exist is, we think, quite clear. This is perhaps best seen in the
work of Schmidt (1920), who finds that the ^oarces population
is divided into numerous ' races ' and each of these can be
again split into still smaller elements. In this case (p. 114)
the averages of the smaller groups combined give the average
of the race. A similar example is seen in Duncker's studies
of the Flounder and Plaice (1896).

In Sumner's investigation of the local variation oiPeromyscus
maniculatus it is quite clear that the local populations within
the three chief subspecies are not identical (1920, p. 388, fig. 2),
but exhibit significant statistical differences. He says (191 7,
p. 173), ' subspecies themselves are far from being elementary.'
They are composite groups comprising in numerous cases a
number — perhaps a great number — of distinguishable local
types. Similar groups which are the result of intense localisa-
tion in segregated populations are recorded by Gulick (I.e.),
Crampton (I.e.), Mayer (1902), Boycott (1919), Aubertin,
Ellis and Robson (1931) for ' colonies ' of land snails (general
discussion of the problem in the last-named paper). Many of
these colonies are found in valleys or on ridges. A still more
acute form of local differentiation is seen in the ' forms ' of

70 THE VARIATION OF ANIMALS IN NATURE

rats found in different houses in India by Lloyd (191 2) and
the statistical differences between communities of ants found
in different nests (Alpatov, 1924) and in the ' races ' otPartula
found on single trees by Pilsbry, Hyatt and Cook (191 2).
For such ' besondere kleine lokal geographische Einheiten '
Semenov-Tian-Shansky (1910) has proposed the name ' natio?
We might even include here such groups as are produced by
a gregarious instinct and appear as centres of attraction
in populations not broken up by topographical obstacles
(' schools,' shoals and rookeries). In the majority of cases the
groups under discussion represent mere statistical divergences
from the mean of the population, such as are seen in the per-
centage-difference of colour- and band-classes of land snails
and in the different combinations of ear-, tail- and foot-length
of Peromyscus.

How far the groups which we have been discussing are
hereditarily stable it is impossible to say. Experimental proof
is available to show that the races of ^oarces and Lebistes
(Schmidt), Peromyscus (Sumner, 191 5), Cerion (Bartsch, 1920),
moths (Goldschmidt, 1922, 1923) and bees (Alpatov, 1929)
are stable. We would, however, surmise that a good many
alleged racial distinctions are of the nature of ' fluctuations '
(cf. Woltereck on non-inheritable racial characters of the
Cladocera, 1928). Much valuable work remains to be done
in this field. Crossing experiments have been undertaken
by Sumner (191 7), who finds that some subspecies of Pero-
myscus maniculatus can be successfully crossed, while others
are sterile inter se.

The fact that populations are divisible into distinct geo-
graphical groups such as we have been describing and that
some taxonomic species are constellations of geographical forms
has led certain students to seek some means of distinguishing
such composite groups. They were first called ' Formen-
kreise ' by Kleinschmidt ; but Rensch (1929) has recently
proposed the term ' Rassenkreise ' for them and has thoroughly
examined the subject. He suggests that the term ' species '
should be restricted to groups of mutually fertile and struc-
turally similar individuals which exhibit only individual,
ecological or seasonal variation, having heritable differences
but not divisible into geographical races. Rensch's definition
{I.e. p. 15) has to be taken in conjunction with that of his

THE CATEGORIES OF VARIANT INDIVIDUALS 71

geographical race which ' geht gleitend in die Nachbarrassen
uber.' He suggests that groups of geographical races which
may or may not correspond with taxonomic species should
be called ' Rassenkreise.' x Now Rensch's Rassenkreis, as
far as we can see, can scarcely be treated as a classificatory
unit, but rather as the name of a principle of divergence. It
denotes the tendency to form constellations of geographical
races. At times the Rassenkreis appears to us to be clearly
conterminous with the taxonomists' species. Rensch does
not hesitate to give some of his Rassenkreise binominal names
(e.g. p. 29, the Rassenkreis of Troglodytes troglodytes). The
suggestion is of value in pointing the differences between
groups of races connected by transitional forms and more
homogeneous and geographically undiversified groups ; but
it has a disadvantage in that two terms are applied to what
are in practice equivalent degrees of morphological divergence.
We are left, in short, with the general result that there is
a principle of geographical divergence manifest within the
systematic species, and at all early stages in evolutionary
divergence, of such a nature that groups very slightly different
in structure (often only in a single character, e.g. coat- or
plumage-colour) are also distinct in their topographical
range. That such divergence is, according to our present
knowledge, more clearly seen in some groups than others is
quite apparent. But we would point out (a) that it is by no
means a universal feature in mammals and birds and (b) that
we are a little uncertain as to how far it may not be exaggerated
in those groups by the relatively low numbers used in the
discrimination of mammalian and bird races. Finally, it is
uncertain to what extent many of the subspecies and geo-
graphical races described by taxonomists are hereditarily
stable.

IV. Genetigal and Reproductive Categories

It is convenient to consider here not only the strictly
genetical categories, such as the biotype, pure line and the
' petite espece,' but also the clone and the syngameon which
depend on the type of reproduction (whether sexual or asexual,
interbreeding or not), and the aberration, form, modification and

1 In all probability the Rassenkreis corresponds to Waagen's ' Collectivart '
and the gens of certain modern palaeontologists (cf. Bather, 1927, p. Ixxxviii).

72 THE VARIATION OF ANIMALS IN NATURE

exotype which depend on the recognition that a given form is
non-heritable. Perhaps we might also include the ecotype
and ecospecies (Turesson, Alpatov), which are combinations
of genetical and ecological concepts. Even in motile animals
such as ants Alpatov (1924) has been able to recognise
analogous ' subspecies ecologicae truncicolae ' in the European
and Japanese subspecies of Formica rufa. We are dealing
here, however, with a category having primarily an ecological
basis, some members of which are physiologically differentiated
(cf. Chapter IV, p. 119).

In categories such as the clone and the pure line one may
say that the logical classificatory ideal of a category having
standardised characterisation is attained. These units are
defined not by their degree of morphological divergence, but
by their mode of reproduction and degree of genetical homo-
geneity.

Some of the genetical units are obviously subdivisions of
the species. It has long been realised that taxonomic units
may contain numerous intercrossing strains (? = petites
especes),just as, considered in the time-relationship, the lineage
consists of interwoven and anastomosing lines of descent which
at any one horizon seem to have a similar status. Other such
categories have less to do with the content of the species. The
pure line is indeed an expression of differentiation within the
species, but as it is (sensu stricto) the result of a particular mode
of reproduction (autogamous), it is only of importance in
certain groups. It must also be noticed that a pure line may
consist of individuals homozygous for only one pair of allelo-
morphs. The term pure line is sometimes inaccurately given
to a genotypically homogeneous group, without reference to
the mode of reproduction, e.g. a homozygous biotype. Clone-
formation, on the other hand, seen in the Protozoa will be
characteristic only of such parts of a species-population as are
reproducing asexually. 1

The term biotype (' a population consisting of individuals
with identical genotypical constitution ' (du Rietz) ) is a
recognisable entity among both autogamous and allogamous
forms, but, as du Rietz (I.e. p. 340) points out, there is little
chance that in regularly allogamous forms any biotype will

1 The term clone is sometimes applied to the broods of parthenogenetic
animals.

THE CATEGORIES OF VARIANT INDIVIDUALS 73

be represented by more than one individual on account of the
great number of possible gene-combinations.

Just as Rensch attempted {I.e.) to reconcile the systematic
and geographical concepts by a new terminology, so Lotsy
has attempted to synthesise systematic and genetical results.
He pointed out that the homozygous biotype is the only real
fundamental taxonomic unit (191 6) and therefore the only
unit worthy of being called species. He proposed the term
Jordanon to denominate the smaller character-groupings that
Jordan had detected within many Linnean species, and Linneon
for the larger composite groups. A considerable literature
has accumulated around Lotsy's suggestion. We do not
venture to discuss what is primarily a feature of plant popu-
lations. But there seems to be this much of common ground
between botanical and zoological results. As we have seen
in discussing Rensch's proposal, there are homogeneous and
heterogeneous species (' simple ' and ' compound,' Cockayne
and Allan, 1927) and the lines between a group consisting of
a single biotype and a Jordanon and between the latter and
a Linneon are quite arbitrary. What we seem to be dealing
with is the progressive formation of groups differing in more
and more characters.

Genetical analysis has revealed a process of differentiation
partly produced by the mechanism of heredity, partly the
result of some other factor or factors. At the lowest level,
populations have their characteristics determined by the
processes of heredity and methods of reproduction — they are
homozygous or heterozygous, pure lines or heterogeneous
assemblages. Some characters may keep together in pairs
according to the amount of linkage. Imposed on this funda-
mental character-distribution is the process usually recognised
by the taxonomists by which larger and more substantial
character-groups are formed, either with or without geo-
graphical or ecological differentiation.

V. Physiological Categories

Of recent years it has been increasingly apparent that in
certain classes taxonomic species are subdivided into races,
characterised by slight or no morphological differences, but
by marked differences of habitat, food-preference and even of

74 THE VARIATION OF ANIMALS IN NATURE

function and occupation. Such units are generally known
as biological or physiological races. They have, of course,
been for a long time familiar to bacteriologists and have been
detected in Protozoa among which structurally indistinguishable
strains are found in different hosts. Similar ' host-specificity '
accompanied by morphological differentiation is a well-known
phenomenon in various groups of parasitic Metazoa. The
whole problem of physiological differentiation involving such
phenomena as immunity, certain aspects of interspecific
sterility and graft-specificity has been recently reviewed by
Robson (1928, Chapter III), and Thorpe (1930, p. 177) has
given a survey of the special phenomenon of biological races
in insects, nematodes, etc. It should be noted (a) that it is
not always easy to distinguish ' physiological races ' from
those separated by habitat-preferences which may be
determined by other factors than physiological idiosyncrasy,
and (b) that ' physiological ' is sometimes used in a very broad
sense. Thus Fulton (1925) and Allard (1929) allude to the
stridulation of Orthoptera as physiologically differentiated.

How frequent this phenomenon is it is not easy to say.
It may be that in every phylum the species are composed of
subordinate groups diversified in regard to their ' physio-
logical ' characters. The ground has not been sufficiently
explored from this point of view. A list of the features of this
order that seem in one group or another to be the basis of
racial diversification is sufficiently impressive to lead us to
believe that it must be of very frequent occurrence.

While in practice it would be undesirable to give separate
names to the various physiological races within a species, it
should be noted that some botanists have definitely adopted
the practice of naming ecological subspecies and that Alpatov
(1924) has recognised similar subspecies (' truncicolae,' etc.)
in ants.

Just as the taxonomist's species may contain divers struc-
tural, geographical and genetical subdivisions, it also seems
to contain elements that are diversified by habit, habitat-
preference, physiological reactions, food-preference and so
on. Such differentiation may or may not be accompanied
by structural differentiation and its occurrence must always
constitute an interesting starting-point for evolutionary inquiry,
as it invites the obvious query — do initial differences in food,

THE CATEGORIES OF VARIANT INDIVIDUALS 75

habits, etc., lead to structural change ? The demonstration
by Nuttall (19 14), Bacot (191 7) and Sikora (191 7) that the
human head-louse could be transformed into the body-louse
by transference from the head to the arm is interesting in
this connection.

The physiological race presents no special difficulty in
our scheme of categories. How far they are (a) regularly
distinguished as discontinuous populations and (b) hereditarily
fixed are more difficult questions, and there are not sufficient
data to answer them. Races habituated, e.g., to different
food-plants will obviously be dis-
continuous, but some contrasted
types of habitat-preference are
certainly not. As regards the
hereditary fixation of such racial
characteristics little can be said
at present. The experiments with
Pediculus [anted) and Thorpe's ex-
periments (1929) with Hyponomeuta " pitn ?
seem to suggest that physiological
preferences are not germinally fixed.
Harrison's claim to have induced
a new germinally fixed habit of
oviposition in Pontania (1927), in-
volving the acquisition of a pre-
ference for a new host-plant, does
not seem to be justified (see p. 41).

Pi/estimen/i 6

P.vcstiminh' ^

Fig. 4. — Body-Lice (larger
specimens) and Head-
Lice {Pediculus).

(From Sikora, 191 7.)

Different methods of analysing the variation of natural
populations have shown that it is not without order and the
most obvious tendency is for individual variants to form groups
of various kinds. These groups are aggregates of individuals
resembling each other usually in a number of correlated
characters. The simplest and most fundamental manifestation
of this tendency is seen in the homogeneous stocks produced
by vegetative or autogamous reproduction. The mechanism
of heredity produces another kind of group in the biotype and
combined with autogamous reproduction, the pure line. A
third kind is produced by topographical and other barriers
to intercourse, and here it is customary to indicate the degree
of divergence by a hierarchy of grades beginning at the colony

76 THE VARIATION OF ANIMALS IN NATURE

and passing through the local race to the subspecies. 1 In this
system we see groups progressively diverging either in more
characters or in the amplification of individual differences.
So far the bulk of our knowledge of these processes is concerned
with structural divergence, but there is strong evidence for
the occurrence of ' races ' which differ from one another in
single features of habit, food-preference and physiological
activity. Still further divergence is seen in the groups usually
recognised as species which contain a number of distinct
but intergrading subordinate elements of the various kinds
described above. Species may be more or less homogeneous
or they may be markedly diversified by sharply cut constituent
elements (Rassenkreise). Palaeontological evidence suggests
that historically considered the various individual character-
sequences within a group do not develop at the same rate.
This principle can probably be harmonised with the results
of neontology by reference to the observed fact that different
elements (e.g. colonies) exhibit different proportions of the
same stock of variants and the theoretical assumption that
new mutations occur at different parts in and spread slowly
through a population.

1 Sometimes a form is given subspecific rank because it covers a wide area,
although it differs from its nearest ally in very minute details. On the other
hand, a well-marked variety with a very restricted range might not be given the
same rank, chiefly because, on the whole, fewer workers will be interested in
a form found only in a small area.

CHAPTER IV

THE DISTRIBUTION OF VARIANTS IN NATURE

In this chapter we propose to consider the manner in which
variations are distributed in nature. As indicated in Chapter I
the distribution is not purely random. Groups of various
kinds are manifest on the most superficial inspection, and it
is our object to describe the various kinds of aggregates found
and the mode of their occurrence, and to indicate any general
inferences which may be drawn from the latter.

As a preliminary to this inquiry we have to discuss certain
general principles and facts which have an immediate bearing
on this subject.

i. In Chapter II we have given certain data relating to
the susceptibility of the living organism to its environment
and have discussed how far we can form an opinion as to the
likelihood that the effect of such susceptibility is heritable.
Apart from the latter all-important question, it is clear that
some part of the variation (both in individuals and in popu-
lations) in nature is causally related to the factors of the en-
vironment. How far we are entitled to consider the characters
of any variants and groups as heritable and how far our
knowledge is embarrassed by ignorance in this respect will
be discussed in 3.

In addition to the significant and universal occurrence of
groups already noted (Chapter I), it is known (Chapter II)
that there is another broad principle of distribution of which
the essential characteristic is the correlation of some progressive
modification or series of modifications with a climatic or
environmental ' trend ' or ' gradient.' Such a series is often
represented by a number of subspecies or races, as in the
subspecies of the Fox Sparrow {Passerella iliaca) of N.W.
America (Swarth, 1920). Many cases of single-character
modifications are seen in the data brought forward in support

78 THE VARIATION OF ANIMALS IN NATURE

of the so-called ' Laws ' of Allen, Bergmann and Gloger. In
some instances these ' trends ' are not obviously correlated
with environmental gradients (Swarth, I.e. pp. 98-100 ;
Hewitt, 1925, p. 263; Snodgrass, 1903, p. 411). The two
last-named writers attribute the series (in scorpions and birds)
to successive waves of migration. Hewitt (I.e. p. 274) speci-
fically states that the series he studied are phylogenetic.
Hutchinson (1929, p. 444) records an interesting trend from
west to east in South Africa among the Notonectidae, in which
three subspecies of Micronecta piccanin form a series, though
the typical form M. piccanin piccanin is found unmodified along
the whole trend. Swarth (I.e. p. 92) notes that a trend may
be composed of successive areas of subspecific or racial stability
separated by narrow areas of intergradation.

2. The very general occurrence of local and geographical
races is discussed later on (p. 104). It should, however,
be pointed out here that into the formation of some groups
more than one factor probably enters, viz. differentiated
environments (the effects of which may be inherited or not),
isolation, mode of reproduction and inheritance. How far
adaptation to local conditions enters into their formation is
considered in Chapter VII.

3. It has been shown (Chapters I and II) that there
is a great lack of knowledge as to how far the variation of
animals in nature is heritable or not and whether the very
obvious plasticity of form and habit is of any moment in
evolution. It has also been noted that there is among taxono-
mists and other students a rough-and-ready acceptance of
the distinction between fluctuations and heritable variation,
though there is no criterion for deciding between them other
than the very small number of experiments and rather dubious
analogies (Chapter I). All generalisations based on the facts
of local and geographical variation labour under this initial
disadvantage. There have, it is true, been cited a number of in-
stances in which the heritable or non-heritable nature of variants
has been satisfactorily determined. But it is reasonable to
ask— what inferences are to be drawn from perhaps 20 or 30
experiments, when our generalisations should cover the whole
range of recorded variation ? If modern Biology elects to
stand by the criterion of experiment in what, after all, consti-
tutes one of its most important fields of evolutionary research,

THE DISTRIBUTION OF VARIANTS IN NATURE 79

it is obviously thrown back on a relatively small number
of experimentally tested cases and the great bulk of the data
on local divergence (often associated with valuable ecological
and bionomic data) is worthless !

We have given in Chapter II a general survey of the facts
concerning fluctuations ; but it is desirable here to define
how far the deficiency in experimental evidence may be
remedied by other means. The following means of inferring
whether we are dealing with fluctuations seem to be available.

A. Certain characters such as size and colour are some-
times determined by the amount and type of food available
and, though the non-heritability of such variation is only
very rarely demonstrated, it is a fair inference that they are
not inherited.

(a) Size. — The adult size of insects obviously depends on
the food available for the larvae. In forms with a fluctuating
food-supply, such as carrion-feeding flies, adaptability in this
respect is very marked (cf. Salt, 1932). Mickel (1924, pp. 15-16)
has given a summary of a number of cases, in addition
to his own definite evidence that in the wasp Dasymutilla
bioculata adult size is dependent on the quantity of food available
for the larva. Especially significant is the experiment of
Wodsedalek (191 7), who was able to vary the size of the larvae
of a Dermestid (Trogodenna tarsale) from large to small by
starving them and from small to large by feeding them again.
Amongst molluscs, Hecht (1896) records that Elysia viridis
grows to a much larger size when its diet is changed from
Codium to Cladophora.

(b) Colour.- — Pelseneer (1920, p. 485) gives a long list of
colour-changes in molluscs wrought by differences in diet.
In insects which feed on different plants the colour likewise
varies with the food. Thus Waters (1928) notes that the
moorland form of the moth Coleophora caespititiella, which
feeds on Juncus squarrosus, can be distinguished fairly easily
by its darker colour from the specimens bred from J. communis.
Eisentraut (1929a) attributes the darker colour of certain
littoral forms of the gecko Hemidactylus to their feeding on
Halophyta. In general it may be noted that there is a tra-
ditional suspicion among taxonomists that colour is an unsafe
systematic index. This is partly because it is extremely plastic.
In some instances, however, experiment is against this view.

1

>

(

f

1
/

^

•

,'''

►

„--

— -

,,--

" ™"

V

*

V

%

>

\

i

•,

\

1

/

V

/

\
\
1

\

V

/

/

s

N

s

X

s

f"

«

X

H**"*

"*"

>

--..

1

I
1

1
1

■^

V

%

>

V

\

1

1
1

■b

o+

i

1

x> «

* «

\J 4

=> <

r> <

a ■

CT <

\* *

9 (

D <

a "

» <

Nl .<

=> c

o <

£> «

* «

M

«M *M —

r: co <o

3 -a

Z
O

IA

CQ

•±

cu o»

a

-*

m «v»

a

CM

2 ^

3

rr r—

-S M

tilla
ckel,

£2

S2

in

a •*»

d <2

<

t\J

£i m

in

«•> _

2 S

—

< g

O

V3

V

U

w

N

a,

172 «

3

M CM

Z °

z P

m
erf

»*» —

2 t
< £

ao

s ~

in

CO

w S

CO

oo _

MCY C

eaks ar

in

2^

SH

^

on

Pn

IV N N N M

1

£

to

Of \

vo

CO

5

THE DISTRIBUTION OF VARIANTS IN NATURE 81

Sumner (191 8) says that in Peromyscus ' it [colour] is less subject
to erratic local influences than the length of body parts.'

B. Certain mechanical stresses, such as wave and current
actions, produce on forms with hard external parts (e.g. corals
and molluscs) modifications of a particular type which we
may fairly infer are not hereditary. Thus we find that Limnaea
andersoniana of N. India (Annandale and Rao, 1925) exhibits
a still-water form, a stream form and a current form recog-
nisable by the shape of the shell. Similar habitat-forms of
corals are described by Wood Jones (1910). We may include
here such modifications as are imposed on sedentary organisms
by the character of their substrate (sponges (Burton, 1928) ;
Anomia (Jensen, 191 2) ). There is no direct evidence that
these forms are not inherited ; all we can say is that they
seem to show that accommodation to external stresses which
we have come to associate with non-heritable plasticity. It
is known that certain variations in mollusc shells less obviously
related to environmental conditions (e.g. dwarfing in Crepidula
(Conklin, 1898) and the ' abyssicola ' form of Limnaea palustris
(Roszkowski, 191 2) ) are non-heritable.

G. A good number of variations associated with other
external factors are probably of a fluctuational nature. These
include (a) the effects of the chemical differences in the medium
(soil or water) (e.g. modifications of the shell of molluscs in
brackish water (Bateson, 1889), the stunting of marine molluscs
in water of low salinity (Pelseneer, 1920, p. 565), and the
modification of the shell of terrestrial forms on soils deficient
in lime-salts (id. I.e. p. 577) ), (b) the action of humidity and
dryness, (c) of temperature and (d) of sunlight.

In all the cases enumerated in A-B it is necessary to make
a distinction between the action of intermittently changing
factors and long-sustained environmental pressure, as we have
already suggested the possibility that the time-factor cannot
be altogether disregarded in the induction of heritable
variants.

We ought to consider the converse question — to what
extent are natural variations known to be heritable ? A very
considerable literature is, of course, available on this subject.
The experimental results are, however, very unequally
distributed among the various phyla, largely because all
animals do not lend themselves to experiment with equal

82 THE VARIATION OF ANIMALS IN NATURE

facility. A great deal of work has been done on Protozoa
and insects, a less amount on Mollusca, and still less on
birds and mammals (wild), fishes and Crustacea. Among the
other groups our knowledge is defective. No general inferences
can be made from these results as to what characters are
especially prone to be heritable nor as to the likely incidence
of such variation in the vast number of described species. As
regards the heritability of the characters which distinguish
local races it is still more difficult to generalise. From the
work of Sumner (mammals), Schmidt (fishes), Harrison,
Tower, Goldschmidt (insects) and Woltereck (Crustacea) it is
evident that some races tend to breed true, though the racial
complex is dissociated and broken up on crossing.

4. There are certain special types of local variation which
are more properly considered in relation to the causes which
are presumed to have encouraged or given rise to them.
Prominent amongst these is the occurrence of special insular
forms. These include not only normal divergences from the
adjacent continental forms, but also certain abnormalities,
such as melanic, dwarf and giant types, which have repeatedly
been noted as characteristic of insular faunas (see Chapter V).

5. In the intensive study of local variation involving the
comparison of distinct races or subspecies there is sometimes
available data for estimating the relative size of local groups.
Such data have often given us the impression that in a group
of closely related groups (races or subspecies) one particular
group will tend to occupy a larger area or otherwise tend to
predominate over the others. This is usually recognised in
taxonomy as the typical form. The means for judging how
frequent this predominance of one or more forms within a
species may be, are not very extensive, as the appropriate
data are not often given. If it is, as we suspect, of general
occurrence, it is a phenomenon of some consequence and
might conceivably be adduced as evidence for the operation
of selection. Instances are seen in the distribution of the
subspecies of American marmots (Howell, 191 5) and Glaucomys
{id. 1 91 8) and also in the races ofPartula (Crampton, 1 916-1932).

It may be pointed out here that the suggestion put
forward by Willis (1922) that the size of the area occupied by
a species is an index of its age (more recent species occupy-
ing smaller areas) has been in some measure confirmed for

Fig. 6a.— Map of Distribution of Races of the Marmot, Marmota caligata.

(Fig. 3 in Howell, 191 5.)

Fig. 6b. — Map of Distribution of Races of Marmota flaviventris.

(Fig. 2 in Howell, 1915.)

I.M. monax ochTacea
2. " *> petrensis

3."

» »

canadensis

4. '*

»»

ignava

5. "

»»

Tufescens

6."

» ?

preblor~u."m

7."

>♦

monaac

Fig. 6c. — Map of Distribution of Races of Marmota monax.
(Fig. I in Howell, 191 5.)

86 THE VARIATION OF ANIMALS IN NATURE

animals by Riley (1924, p. 77). We hardly believe it feasible
to test that hypothesis with reference to the area occupied
by related subspecies. Willis's theory seems to have a partial
validity ; but, as Robson (1928, p. 114) has suggested, we are
not justified in dealing with it as of prime importance in
explaining differences in distribution.

Fig. 7. — Distribution of Primary Varieties of Partula otaheitana on Tahiti.

(Text-fig. 7 in Crampton, 191 6.)

6. All taxonomists and probably very many other students
know that closely allied species are frequently united by
' intermediates ' or, to put it in another way, that they have
different means but overlapping ranges of variation in some
characters. Other closely allied forms appear to be sharply
distinguished in all the characters investigated, though, of
course, the analysis is rarely pushed far enough to enable us
to say if such distinctions are found in every character.

That all species have a certain, if sometimes very limited,
range of ' continuous ' variation is too well known to require
documentation. The notion of ' continuous ' variation is

THE DISTRIBUTION OF VARIANTS IN NATURE 87

largely an arbitrary one and in practice merely implies that
the differences between individuals are sometimes so slight
that they can be arranged (in a graph or diagram) in a more
or less imperceptibly graded series. Similarly ' discontinuity '
merely implies that there is a more or less perceptible break in
such a series of variates. The sizes of the steps in a continuous
series and of the breaks in discontinuous series are of course
incapable of standardisation. It is largely held that differences
of environment (e.g. the amount of nutrition received by
individuals) contribute very largely to ' continuous ' varia-
bility, though it is now known (e.g. from the work on Drosophila
or that on Ephestia kiihniella (Kiihn and Henke) ) that the
smallest and least sharply distinguished variants may have a
discontinuous hereditary basis.

One of the most important applications of elementary
genetics to the field of taxonomy is to break down the distinction
between ' continuous ' and ' discontinuous ' variation. This
is still insufficiently realised by taxonomists. When we are
dealing with a single character, the occurrence of continuity
or discontinuity is determined by how two contrasting charac-
ters happen to interact in a particular species. It is well
known that the expression (as opposed to the inheritance) of
hereditary characteristics may depend on the environment
to which the individual is exposed. Thus with a given heri-
table basis deciding the main lines of, e.g., colour-pattern, its
actual degree of development may depend on the environment,
heredity determining only the mean. This principle is doubt-
less very important in considering the numerous examples
of pairs of species having a different mean but overlapping
range in some character. Where a complete gradation can
be found over a certain range of variation, it is not sufficiently
realised by taxonomists that very simple statistical treatment
will often demonstrate that the continuous range of variation
really masks a fundamental discontinuity. Taxonomists
usually content themselves with saying either that ' inter-
mediates are rare ' or that ' the forms are connected by all
intergradations,' in each case deciding summarily to separate
or ' lump ' together the two forms. If one makes a table
showing the frequency with which the character appears in
different degrees of development (e.g. as prepared by Sumner,
1923), the true nature of the variation-range may become

88 THE VARIATION OF ANIMALS IN NATURE

apparent. The same method may be applied to discontinuity
in a complex of characters, by means of a table showing the
extent to which they are correlated with one another. In
simple cases it may not even be essential to apply actual
statistical calculation. It is hardly necessary to point out
that discontinuity may be found between single characters
and between groups of characters and that, as Robson (1928,
p. 11) has shown, the attempt to formulate an exact standard
of specific distinctness based on the degree of discontinuity
in structural characters breaks down on account of the very
varying number of characters which may show discontinuous
differences.

The question which really affects our present discussion
is the cause of this continuity and discontinuity of variation
and its usual mode of occurrence in nature. These two
subjects have been discussed by Bateson (191 3) and Robson
{I.e. p. 28 and foil.), and the following brief statement of their
views may be given with some expansion.

Intermediate forms may be of two kinds — (i) ' mid-inter-
mediates,' which are a blend of the characters of two divergent
groups and represent a condition half-way between the two,
and (ii) various combinations of the characters of the two
groups. The former may be due to environmental causes or
to such genetic phenomena as imperfect dominance. The
latter are almost certainly due to genetic causes. Where a
genetic basis for intermediacy between species is involved,
it must arise from crossing or the intermediates may represent
the residuum of a stock from which distinct groups are being
evolved. It should be noted that between two species which
occupy the same area there may be intermediacy in one
region and none in another. This is noted for Cepea hortensis
and nemoralis by Coutagne (1895) and for Notonecta by Delcourt
(1909). It is even seen in such a restricted area as a single
lake, as has been recorded in the pond snail Vivipara of Lake
Garda by Franz (1928). The extent to which intermediacy
in nature is brought about by crossing is very uncertain. That
a great deal has the appearance of being due to this cause is
undoubted, and many systematists (e.g. Pictet, 1926, p. 399 ;
Ruxton and Schwarz, 1929, p. 571 ; Lowe, 1929, p. 29) are
of the opinion that particular intermediate populations are
produced by this cause. Crampton (1932, p. 160 and passim)

THE DISTRIBUTION OF VARIANTS IN NATURE 89

evidently holds that there is good evidence that much of the
interracial intermediacy in Par tula is due to crossing. In a
later chapter we give an account of the factors that make for
isolation between species in nature, and it will be seen that
they are many and varied. Though it may amount to a
truism, we must content ourselves with the conclusion that,
wherever opportunities for crossing are available, a good part
of the intermediacy (notably in respect of recombinations of
characters) found in nature, is due to this cause.

Although all degrees of intermediacy are found in nature
there are certain broad lines which can be recognised in their
mode of occurrence.

Observations in nature suggest that there are three main
tendencies recognisable at the meeting-point of allied species
or races occupying distinct areas.

(1) The groups occupy distinct areas with few or no
intermediates — Lepidoptera (Clark, 1932, p. 8), Peromyscus
maniculatus and P. blandus (Dice, 1931), Eumenes maxillosus
typicus and tropicalis and fenestralis (Bequaert, 1919). This
may occur either with topographical discontinuity (Thomas
and Wroughton, 191 6 (squirrels) ) or without (Dice, I.e.).

(2) There is a narrow area between the two groups occupied
by an intermediate type — Tisiphone species (Waterhouse, 1922),
Passerella iliaca (Swarth, 1920), Peromyscus albifrons and P. polio-
notus (Sumner, 1929). It is interesting to note that the area
of intergradation is very narrow in the last-mentioned case,
although the species in question are known experimentally
to be quite fertile inter se (Sumner, I.e. p. 114). In the case
of Passerella the subspecies mentioned may be broken up into
separate populations {i.e. there may be no continuity of popu-
lation).

(3) A number of subspecies may occur over a larger
or smaller area with complete intergradation between the
various groups — Troglodytes musculus (Chapman and Griscom,
1924), Heodes phlaeas (Ford, 1924). It is of importance to
note that these tendencies may be observed in one and the
same group. Thus Clark (1932, p. 8) states that ' while
some species pass by a series of minute intergradations from
one geographical form to another, others do not, the N. and
S. form occurring together with one or perhaps two well-
marked intergrading types.' So, too, one may note the sharp

go THE VARIATION OF ANIMALS IN NATURE

contrast between Peromyscus polionotus and albifrons and the
gradual transition between P. leucopus and noveboracensis de-
scribed by Osgood (1909).

It is also worth while, from the genetical point of view,
to summarise briefly at this point some of the data with
regard to intergradation in specific characters.

(a) If two species meet but do not interbreed, then there
is no tendency for their character-complexes to break down
more frequently in the area where they meet than elsewhere.

(b) When the intervening area is inhabited by a more or
less definite intermediate form, there is considerable break-
down in correlation. But the breakdown is of a predictable
sort and not altogether at random, some of the more strongly
correlated characters remaining in association.

(c) When there is complete intergradation, correlation
between specific or racial characters is completely broken
down over an area of varying size. Specimens can be given
only a conventional taxonomic name on the basis of the
majority of the characters exhibited. Numerous instances of
such intergradation are noted in our examples (pp. 102-1 19).

Grinnell and Swarth (191 3) also recognise these three
types of intergradation and see in them, probably correctly,
three stages in the fixation of specific type.

7. Darwin (1884, p. 42) was the first to point out that
there is a relationship between the extent of the range of a
species and its variation. Most zoologists probably believe
that 'widely ranging species vary the most' (Darwin, I.e.).
By ' widely ranging ' Darwin clearly meant ' having a wide
distribution ' (as species) and not ' having a wide individual
range,' a distinction of some importance. Obviously, if we
take ' variable ' to involve merely the number of mutations
Darwin was at least theoretically correct, because there will
be a larger chance of mutation in a large population than in
a small one. If he meant that such forms tend to throw more
numerous varieties or regional forms, the statement is only
true in a very general way. We shall see later on (p. 105)
that the amount of regional variation is determined by a
variety of factors, among which habits play a very large part,
and that there are many cases of widely ranging species (e.g.
the Common Heron) which show very little or no regional
differentiation.

THE DISTRIBUTION OF VARIANTS IN NATURE 91

We now proceed to consider the actual mode of occurrence
of variants in nature, in so far as they form recognisable parts
or assemblages within natural populations.

As we have stated (p. 8), all stages can be traced from a
variant which occurs sporadically in a population or occurs
in a small local enclave to a well-marked local or geographical
assemblage. Any attempt to isolate and classify particular
types of occurrence . must necessarily be arbitrary ; but it
seems to us that the following scheme illustrates the chief
stages in the process :

I. Sporadic individual variation usually involving a

single character.
II. The local combinations formed from a stock of variable

characters.
III. The emergence of qualitatively distinct groups involving
large sections of a population. This embraces all
the divergences usually alluded to under the terms
polymorphism and geographical variation.

Of these three stages the phenomenon usually known as
polymorphism includes both II and III, while geographical
variation illustrates III.

These differences are seen in physiological as well as
structural characters, and the former will be discussed at the
end of this chapter.

It will be understood that precise knowledge as to the local
distribution of variants (either in single characters or in
several) ought to be based on a very large array of specimens
collected at all points over the range of the species. Such
intensive studies are unfortunately uncommon. Population
analyses have been conducted on a large scale upon commercial
fishes, though it is at present uncertain to what precise extent
the characters studied (size, number of vertebrae and fin-rays)
are influenced by the environment. The population analyses
of Sumner (Peromyscus) are not sufficiently intensive and are
more concerned with the causes of local divergence. By
far the most valuable data are those based on the population
of land snails (Crampton, etc.), to which allusion is made
under II.

It should finally be noticed that practical experience as
well as a more refined study of natural populations has revealed

92 THE VARIATION OF ANIMALS IN NATURE

that they are often broken up into small self-contained com-
munities such as ' schools,' colonies, rookeries and shoals.
The statistical constitution of such communities is very little
known and only the colonies into which the populations of
land snails are divisible are at all well studied. Some
progress has been made with the study of the shoals of
commercial fishes (Schnakenbeck, 1931). The distinction
between such intimate subdivisions of a population and,
e.g., the races of ^joarces described by Schmidt, is not easy to
draw.

A. Sporadic Individual Variation. — There seem to be
two main tendencies to be recognised under this head according
as a sporadic variation occurs throughout the range of a species
or is more restricted in its occurrence. The most obvious and
commonest type of individual variation of this kind is seen in
colour-phases of various sorts. We ought also to include
certain pattern-forms which occur rarely and sporadically,
e.g. in populations of land snails, in which the main pattern-
types show local statistical differences.

The following are the principal ways in which individual
variants are distributed :

(a) A typical form and a variation occur sporadically
throughout the range.

1. Albinism. — The majority of species of mammals
which have been adequately investigated are found
occasionally to produce albinos in nature. Twenty-
one out of the forty-three British mammals dealt
with by Barrett-Hamilton and Hinton are known to
have produced albinos sporadically within the British
Isles. Similar sporadic variation is widespread in
birds and in some Lepidoptera. Though it is rare
in fishes, Norman (1931, p. 227) says that it is
common in flatfishes.

2. The variety caeruleopunctata of the Small Copper Butterfly,
Heodes phlaeas. — Ford (1924) shows that this variety,
in which the upper side of the hind wings has marginal
blue spots, occurs sporadically through the greater
part of the range. It tends to occur in different
proportions in different places ; the ratio may remain
constant over a number of years.

THE DISTRIBUTION OF VARIANTS IN NATURE 93

3. ' Xanthochroism ' in fishes. — The black and brown pig-
ment is lost more or less entirely in certain groups
and the golden and yellow is left. The Goldfish is,
of course, a cultivated variant of this type. This
condition is found in nature in the Trout and Eel
(Norman, I.e. p. 227).

4. Variation in sculpture in the water beetles, Dytiscidae
{Kolbe, ig2o). — In many species two forms of the
female occur, a smooth form and a sculptured form.
The latter may have deep striae or merely denser
microscopic sculpture, according to the genus and
species. In most cases the proportion of the types
varies locally and one or other form may be found
almost exclusively in certain parts of the range.

5. Colour and pattern forms of land snails. — Many species
of Helicidae are extremely variable in colour and
pattern. It seems at present that the variation is
subject to some measure of local statistical divergence ;
but certain pattern combinations are rare and occur
in single individuals in most local assemblages.

6. Sinistral varieties of normally dextral snails {Crampton,
igi6, ig2j and 1332). — These are somewhat similarly
distributed, but are normally rare, while areas of
high frequency are very localised.

7. Colour variation in the wasp, Synagris cornuta L.
(Beguaert, igig, p. 204). — The species is practically
confined to Engler's Western Forest Province of
Africa. There are eight distinct colour forms.
Many of these occur together in any one district and
several of them have been found in a single nest.
Many intergrades occur. The ground colour is black
with black wings, and variation consists in the pre-
sence or absence of varying amounts of orange on the
thorax and base of the abdomen. These occur in
all combinations.

(b) There is a typical form and a variety or varieties are
localised in definite parts of the range, where they
occur with the typical form.

8. The black variety (var. nigra) of the Rose Chafer, Cetonia
aurata {Blair, igy, p. 121 2). — In Great Britain this

94 THE VARIATION OF ANIMALS IN NATURE

is confined to the Scilly Isles, where it is rare. It
is also known from Corsica and certain parts of
the Mediterranean. The type-form ranges all over
Europe.
9. The greenish female variety (var. valesina) of the Silverwashed
Fritillary, Argynnis paphia. — Goldschmidt (1922) has
shown that the variety is the expression of a single
dominant sex-limited gene. In England the variety
is confined to the New Forest, though the species
has a much wider range. The variety also occurs
sporadically on the Continent.

10. The ' blue ' and ' white ' phases of the Arctic Fox [Elton,
I 93°> P- 8° and foil.). — These two forms often exist
together and interbreed with perfect fertility. The
proportions in which they occur are subject to much
local variation. In certain areas one or the other
form is found exclusively.

11. Colour phases in birds. — Stresemann (1925) records in
birds a type of variation much like that seen in
the Arctic Fox. Thus the Indo-Australian Accipiter
novaehollandiae occurs in a white and a dark form.
In Tasmania, however, only the white form is found.

B. Polymorphism. — This term has been applied, as we
have shown (p. 11), to variation in general and also in a
more restricted sense to the occurrence of strongly marked
phases within a species, whether they are geographically
distinct or occur in the same habitat. We propose to use the
term in the latter sense and to use ' Geographical Variation '
for the occurrence of geographically isolated groups.

1 . Colonial divergence in land snails.

A great deal of intensive study has been devoted to the
statistical investigation of ' colonial ' divergence in land
mollusca. As the results are of considerable value we give
a more detailed analysis than usual and provide a summary
of the results.

(a) Alkins (1928) studied two characters (altitude and
major diameter of shell) in Clausilia rugosa and
C. cravenensis in 19 loci distributed over an area of
8x4 miles.

THE DISTRIBUTION OF VARIANTS IN NATURE 95

(i) Each colony has a rather wide range of variation
in the two characters, rather more in respect
of altitude. In altitude the ' spread ' is very
wide, i.e. no one class is very frequent. In
diameter there is a distinct tendency for a
high grouping about one phase. This is
very well seen in the figure of ' polygons of
variation ' (op. cit. pp. 59 and 61).
(ii) In general the series from neighbouring loci are
more or less alike, but the converse is not true.
No two loci have exactly the same mean.
The shell-characters are not correlated with
the ecological characters of the various loci.

(b) Boycott (191 9, 1927) also studied the shape of the

shell in C. rugosa. He found the same amount of
variation in each locus and that there was no relation
between the former and the character of the locus,
though he suspected some relation between shell-
altitude and environment. ' Significant ' differences
were found in 5 out of 6 pairs of contrasted characters.

(c) Aubertin (1927) studied a number of colonies of

Cepea nemoralis and C. hortensis both for shell-colour,
etc., and anatomical characters. The former only
are considered here. The number of specimens used
is rather low.

(i) C. hortensis. — Each colony has a wide range of
variation and in three out of four ' adjacent
colonies ' there was good ' spread ' for ground
colour. In one (' Hedge Lane ') yellow was
90 % of the total. For three types of banding
the spread as between type 12345 an d 00000
was equal. Some colonies lack a particular
ground colour-class altogether.

Adjacent colonies tend to be different signifi-
cantly in ground colour, less so in banding.
A Buckinghamshire colony closely resembles
one Wiltshire colony, though it differs in the
absence of a colour-class found in the latter.
(ii) C. nemoralis. — In colour some colonies lack
certain classes altogether, as in (i) ; but this

96 THE VARIATION OF ANIMALS IN NATURE

is far more marked in nemoralis. The ' spread '
of variation is more limited ; actually two
out of the four colour-classes only are repre-
sented, though a few ' brown ' occur at three
colonies. In one colony (Maiden Castle)
one class is 77 % of the population. In banding
the spread was fairly wide, though usually
one or two classes tend to be more highly
represented.

(d) Rensch (1932) calculated the percentage frequency

in 16 colonies of Cepea nemoralis (mostly remote from
each other) for 7 colour- and band-classes. The
statistical significances of the differences were not
worked out. From his table we may give the following
results on ' spread.' In four colonies one class was
found in over 90 % of the specimens ; in four, one
class was over 70 % , and in one a class was over
80%. In the rest the tendency was for two classes
to be well represented and the others to be numerically
inferior. Very often three or four classes are entirely
absent. Two classes, yellow 00000 and yellow 12345,
have a very high frequency and are about equal in
frequency, and the others are all very low.

(e) Crampton's work (1916, 1925 and 1932) is on a

much larger scale than the rest. It is, in fact, so
extensive and the details are so manifold that one
awaits a summary and analysis by the author and
only the following points can be noted here :

(i) The spread of variation tends to follow the same
lines as in (d), i.e. there is a tendency for one
or more classes to be preponderatingly frequent
and some colonies may lack a whole series of
classes,
(ii) Adjacent colonies tend to be alike, but the same
percentage of a given class may be found in
remote colonies. Abrupt change in the number
of classes and their percentage frequency is
found between adjacent loci, and the latter
may differ in the absence and presence of
whole classes.

THE DISTRIBUTION OF VARIANTS IN NATURE 97

SS

■a

oo

W
oo
oo

<

h-l

u

I

Q

<

< 2

1

P

co
o

o
1-1
o
U

o u

oo

u w

&

w

00

u
w

<

U

o

u

c

3
W

fa

m

CO

3 O

13

ID

CO

o

o
13

m
o

CO

o

J

13

m

o

CO

o

s

—
v

J

13

u

u

o

H-l

CO CO

CO

r^

r^

TtH

TjH

loco

CO

CN

I-"

I-"

CO

•-"

CO

JN

01

CD

r^

t-H

LO

OJ i-c

(M

I —

r»

1— <

HH

m

^

CO

Of

CO

lO

CO

TjH

co

r^

CD

Of

o

T}H

LO LQ

^ O

CD

o

cn r-

CO

CD
CN

CO TjH

o 6

LO LO

CD CO
<N (N

lO

tN

CO

co

O

(N

LO
LO

(N

(N
<N

CD

O) CI N CO G

CO

CD O CN f^ i-h

CD

co cd r^

OJ D r^ O O

CO CO N tN h

CO

rf Of CO lO c

CD lO lOD

CD

fNCO CO
t^ O O)

CN CO CD Tf CO ""CD
CO CONNN Tf

CN >-.

G) CO m N C) -f< N >tco o

r^ r- lod d o co lo CD co

O Of X^ LQ

bo

P4

cu

be

1

N
C
CU
Sh

Ph

C+h

o

P4

o

n" C

00

00

CCS

co 00

D
-D

3

<U

c
£

OJ

c

Sh

<U

N

CJ

PQ

5 &

P4 pq

00
O

Sh

OJ

3

ctj

oi

(U

Sh

Sh

ccj
OJ

C

00
u
O

PJ

Sh

OJ

Sh
CJ
CU

a

o

a

D

cd
Ph

Sh
<U

c

<u

CvJ

cj

c

OJ

5 §

cu >

u

C3

<u

Sh

<u

O

cu

o

>

<u

OJ

^H ^

.CJ h^

^*^ Sh

>^ " £ ^

OO .rH OJ Sh

£ ^ >(J fcJDw^

~> O >- cj o

M-T

Sh

Sh T3
CJ 00

7: CJ C

> hc: c

4H

c73

W

11

98 THE VARIATION OF ANIMALS IN NATURE

tfi

ro

c*

TV 943JO0O

A enoy

A anqedBj

a m^v

A nrutfunj

A BundBBj^

ir:

in

I - — 1

1 1

i
i

>o

lf-1

in

in

in

IO

lO

ir

ir

5 iO

1/5

IO LO

us

!ip

i.O

LO

<J:

eg

ni

CO

< — i

-r

l>

o

so

CO

a

■5 CO

IO

CO »-l

T

c-

o

C3

C^}

CO

TO

ro

rr

-r

-+

1-1

lO

m

ii-

:* tc

to

to c—

t -

t>

CO

V)

i I

r— <
<

. — i
j i

« — i

■-'

J

_. 1 .

1 ...

•—1

1— 1

1 1

JZL

(•uulu) qjSuai 'jjaiiS

<:

-
-

ta

O

2

;;

/

X

<

2

o

a

U

o

a

S

.-

K

u

<

G

m

a

m

i;

•<

W

—

n

rt

«

-H

Z

o

j

*J

<

PL,

a

■-— *

x

<■

x

X

o

h

o

O

w

V

J

w
ij

z

►J

■j

CI

w

<-o

X

■^

c/3

Cl

~

2

>— i

W

a

v.
o

w

h

a.

«j

n

<

>

r/l

W

2

z

o

<

J

o

U

cd

o

-.

THE DISTRIBUTION OF VARIANTS IN NATURE 99

(/) Aubertin, Ellis and Robson (193 1 ) studied colonies
of Cochlicella acuta in W. Sussex in respect of three
main types of shell-colour.

(i) 21 comparisons were made (I.e. p. 1042), and of
these 2 only showed equal distribution of the
types. In the rest no regularity of incidence
was found, but either one class or two
tended to preponderate at the expense of the
third. In each colony, however, all three
types are usually well represented, and in 63
cases there were only 7 instances of a colony
having less than 20 °/ of any one type.

(ii) The various colonies differ significantly in 36 %
of the possible comparisons. The authors
say that on the whole (p. 1047) very little
relation exists between the distance separating
the colonies and the differences in shell-pattern.
But this is not quite true, as nearly all imme-
diately adjacent colonies tend to show very
little difference one from another. Neverthe-
less it is true that some adjacent colonies may
differ significantly and distant ones may be alike.

From these summaries we may form the following conclu-
sions.

(1) Populations of land snails tend to occur in colonies
having a different facies, the differences having little correlation
with differences of environment (Alkins, Crampton, Aubertin,
Ellis and Robson) except perhaps in size (Boycott).

(2) Continuous populations (/) may be divisible into sub-
ordinate areas with a statistically different composition.

(3) In two cases ( (b) and (/) ) these differences are main-
tained with a tolerable degree of uniformity over a limited
number of years (up to ten) .

(4) While each colony tends to show a fairly wide range
of variation, certain classes of variants tend to preponderate
and often whole classes may be absent. One gets the impression
that colonies exhibit the results of obligatory selective mating.

(5) That certain classes tend to occur in a high percentage
might suggest that selection may be at work ; but we think
that this is unlikely, as (e.g. in Rensch's observations) we find

23 24 25

27 28 29

O JO 20

SCALE OF lu ub u J u u U u U MILLIMETRES

Fig. 9. — Variation in the Pointed Snail in its Colonies in Sussex.

(From Toms, 1922.)

THE DISTRIBUTION OF VARIANTS IN NATURE 101

reversal of frequency, e.g. yellow ooooo is very numerous at
the Viennese locus, very rare in the Bohemian, 12345 1S
numerous at ' Berlin-Buch,' very rare at Ratzeburg.

(6) (a) In populations intensively studied over a limited area
(up to 15 X 15 miles) there is an initial tendency for adjacent
colonies to be alike, but

(b) the converse is not true.

(7) 3 (above) suggests that colonies once they have diverged
might give rise to races.

(8) The extent to which boundaries are broken down (e.g.
by specimens being carried about by birds, wind, etc.) is
unknown.

(9) It is indeed a little surprising how much community
there is over wide areas, and this suggests that homogeneous
races and colonies differing significantly in several characters
are not likely to be very often produced in such populations.

2. Polymorphism throughout the range of the species.

(a) Slugs. — The variation of the commoner European slugs
is not completely known ; but it has been recorded in
sufficient detail to enable us to state that in polymorphic
species such as Limax maximus and Arion ater some of the colour
varieties are widely spread over the range and certainly occur
together very frequently.

(b) Spiders. — Bristowe (1931) has described the colour-
variation of the spider Theridion ovatum, on which there are
three types of abdomen-colour, viz. : white, striped and red.
Details are given of the various proportions of these characters
in different parts of England.

(c) Fishes. — Norman (1931, p. 220) states that the fish
Epinephelus striatus has eight colour-phases, none of which can
be called more normal than any other. Some of the forms are
strikingly different.

(d) Beetles. — Hauser (1921) has described the extraordinary
variation in the Asiatic beetles of the genera Damaster and
Coptolabrus (Carabidae). In most of the species-groups, the
characters which elsewhere define species and races are
variable. Thus in one local race of a species — e.g. in the
coelestis group — very plump, moderately short-legged and
very long, long-legged forms are found ; the elytra may be
parallel-sided with strongly marked shoulders, or elliptical

102 THE VARIATION OF ANIMALS IN NATURE

or egg-shaped with no shoulders. The pronotum and other
parts vary in the same way. There are about forty-six types
of variation (such as long- and short-legged ; long-, short- or
a-mucronate-forms, etc.), which are liable to turn up in the
races of any species. The colour also varies, but may be
directly correlated with climatic conditions. In most European
Carabus variation within the species consists of many local
races, each of which is pretty constant. In Coptolabrus each
race is very variable and not nearly so sharply defined.

(e) Lepidoptera. — Doubtless some of the most remarkable
cases are complicated by the phenomena of mimicry, but many
non-mimetic species are quite sufficiently remarkable. In
the mimetic forms the discontinuity between the various types
tends to be more marked. Of mimetic butterflies Heliconius
melpomene (Eltringham, 1916) is one of the most remarkable.
Eltringham united ten reputed species and 60-70 named
colour-forms, all of which are structurally indistinguishable.
Some of the forms are geographically limited, but often several
are found in one restricted locality.

Fryer (1928) has studied the variation in England of the
moth Acalla comariana Zeller (Plate II). At Wisbech there are
six main forms which differ sharply from one another in colour,
the fundamental pattern being the same. Genetical inves-
tigations suggest that there are probably three allelomorphs
for ground colour and a factor for the colour of the costal
blotch which is strongly linked with the ground colour. The
proportions of the various forms were of the same order in
1926 and 1927 at Wisbech, but in Lancashire the proportions
were quite different and an additional type was discovered.
Other species of the genus are even more polymorphic, but
have not been investigated genetically. Sheldon (1 930-1 931)
has shown that there are almost innumerable varieties, many
of them sharply distinct from one another, in Acalla (Peronea)
cristana.

3. Polymorphism combined with constancy in particular areas.

Probably most polymorphic species are really of this
character. We rarely have enough data to show that all the
various forms occur throughout the range. There is always
a tendency to form non-variable colonies or even larger
populations.

I

wm&

3

I

y

£

i>

!

Polymorphism in the moth, Acalla comariana Zeller
(From Fryer 1 928)

THE DISTRIBUTION OF VARIANTS IN NATURE 103

(g) Humble-bees. — Many species of humble-bees, besides
geographical variation, show marked polymorphism in parts
of their range. Certain species, such as Bombus solstitialis
and B. soroensis, which arc extremely variable in Central
Europe, are almost constant in England.

(h) The Coccinellid beetle, Harmonia axyridis. — Dobrzansky
(1924) shows that this species varies in colour from yellow to
black, the colour-pattern of the elytra falling into eight main
classes. Most of the variations can be found all over the
range in different proportions, with the exception that in the
western part of its range (Russia to Japan) there is a tendency
for a single form, H. axyridis (typical), to dominate the
others.

(i) The previous type of variation may be compared with
instances of local specific intergradation, which give rise
to a similar distribution of variants. Thus von Schwep-
penburg (1924) notes that the sparrows Passer domesticus and
P. hispaniolensis, in various subspecific forms, inhabit most of
Europe, N. Africa and Asia without interbreeding, but in
large areas of Algeria, Tunisia and in Malta they interbreed
so much that it is hardly possible to find specimens true to
either type.

Barrett-Hamilton and Hinton (191 5, pp. 545~6) record
that the mice Apodemus jlavicollis and A. sylvaticus, which
occur more or less commonly together in England, occupy
different habitats in Norway. In the latter country the
lowland mice of the south are nearly all sylvaticus, while those
of the high upland pastures are jlavicollis ; in intervening
areas intermediate forms occur, almost certainly as a result
of cross-breeding.

(j) Fernald (1906) shows that the American sand-wasp
Chlorion cyaneum and its race C. c. aerarium differ in average
size and in colour. The typical form is mainly southern and
aerarium mainly northern, but they overlap over a wide area
and occasional specimens of aerarium are found very far south.
He records the same type of variation in C. thomae (with var.
bifoveolatum) , and Porter (1926) found a similar relation in
Sceliphron cementarium between the forms servillei (southern)
and Jlavipes (northern). In other species of Chlorion Fernald
found more complicated variation. Thus in C. ichneumoneum
there are three forms, one found in U.S.A. between Maine and

io 4 THE VARIATION OF ANIMALS IN NATURE

Mexico, one found in Florida, Mexico, Cuba and Venezuela,
and a third found in Florida and the Greater Antilles. In
C. flavitarsus there are four forms, which overlap in a rather
similar way, but there is a main type in U.S.A. and another
in S. America.

C. Geographical Variation. — Under this heading we
propose to deal with instances illustrating the tendency seen
in the species of certain groups to be divisible into subordinate
groups occupying separate or overlapping areas. Such groups
are usually alluded to as subspecies (p. 63).

We have already had occasion to contrast the frequent
occurrence of this kind of geographical variation in Vertebrates
with the irregular and more complex distributional phenomena
in the Invertebrates. This point required some further dis-
cussion. So far as we are aware Rensch (1929) was the first
to point out and to stress the fact that species of certain groups
are more obviously divisible into geographical races than
those of others. Admitting the inadequacy of taxonomic
study and the slight amount of attention paid so far to the
study of geographical variation in some groups, he considers
that mammals, birds, reptiles and Amphibia, Coleoptera,
Lepidoptera, Hymenoptera and Orthoptera display this
tendency markedly. The other insect groups, Arachnids
and Myriopods, probably show the same tendency, but the
available knowledge is defective. The tendency is seen in
land molluscs, but is largely masked by individual and ' ecolo-
gical ' variation. Freshwater and marine groups show it
in some measure ; but it is less marked here. Rensch's actual
survey of the chief groups of the animal kingdom is not
exhaustive, but it includes the more important groups.

He explains (p. 79) the difference in the incidence of
geographical variation by pointing out that in certain groups
the habits, size and mode of reproduction are of such a nature
as to prevent the establishment of barriers and so of isolation
between the parts of a population. Migratory habits, as in
many seabirds and fishes, small size which facilitates accidental
transport, as in land snails, Tardigrada, Nematoda, etc., and
the occurrence of ' resting eggs ' as in Cladocera (but cf.
Lowndes, 1930 ; see on p. 135) are all factors which make
for the homogeneity of a population.

Rensch contrasts the uniformity of the widely ranging

THE DISTRIBUTION OF VARIANTS IN NATURE 105

heron x with the acute local differentiation of the sedentary
wren. Schmidt (191 8, p. 112) in the same way contrasts the
homogeneity of the Common Eel population with the acutely
diversified races of the localised Blenny (^parces), and Burck-
hardt (1900) shows that the cyclic and acyclic species of
Crustacea in Swiss lakes exhibit analogous differences in
the degree to which they form local races.

It will at once strike the critic that, if Rensch's theory is
correct, the proneness or inability to form local or geographical
races must be the resultant of a number of conflicting tenden-
cies. Thus animals like land molluscs by their sedentary
habits should be especially prone to form local races, yet this
factor may be more than counterbalanced by a marked
liability to accidental transport arising from their small size
and mode of life. Small mammals, on the other hand, which
are more or less localised and have a limited range, are less
prone to be transported, so that they should form conspicuous
local races. Finally, large mammals and certain kinds of
birds, though they have a wider range, are obviously not
prone to wide accidental dispersal, so that they should
form larger, but still distinct (geographical) groups. Probably
Rensch would hold that the greater range of the last two
groups is set off by their localised breeding habits. It will
be noted that the contrast is not between, e.g., birds and
molluscs, but between widely ranging and sedentary forms
even of the same group. It will also be seen that wide-ranging
habits in the Mammalia should have the same effect as small
size in the Mollusca, viz. the restriction of local variation.
As the distinction between forms which vary geographically
and those which do not must be based on a resultant of the
kind just suggested, we would expect to find very considerable
differences in the degree in which local or geographical races
are formed, according as one or another of the conditioning
factors is paramount. We must also take into account a
tendency to which little attention has been given, viz. the
inherent tendency of a species to vary. We will now review
some of the salient facts from these points of view.

1 The example chosen is perhaps not very fortunate. The Common Heron
has a remarkably wide range, is migratory and shows little or no regional varia-
tion. Nevertheless, it is a bird of otherwise sedentary habits and evidently
conservative in its breeding habits, as many of the English heronries date back to
an ' immemorial antiquity ' (Nicholson, 1929, p. 270).

106 THE VARIATION OF ANIMALS IN NATURE

In birds there is a very noticeable tendency to form geo-
graphical races (cf. Troglodytes musculus, fig. 12), and this is
probably connected with the tendency of migratory species to
return to the same spot to breed. Exceptions occur to Rensch's
rule that habits condition race-formation. Thus Chapman
(1923, p. 252) states that Buarremon brunneinuchus, though it
ranges from Mexico to Peru and is essentially sedentary in
habits, ' shows no appreciable variation which can be corre-
lated with any given area.' This is all the more striking when
it is realised that a species, B. inornatus, has been evolved in and

Fig. 10. — Variation in the Finch, Buarremon. a and c, B. brwmeinuchus ; bandd,
B. inornatus from the Chimbo Valley and Los Llanos, Ecuador.

(After Chapman, 1923.)

is restricted to a single valley in Ecuador. The case of the
Common Heron (p. 105) has already been discussed. We
believe that these instances must be referred to some inherent
inability to vary.

As far as the recorded facts go, Rensch's rule holds for
mammals, though some exceptions should be noted. Roosevelt
and Heller (191 5, p. 570) show that the Steinbok (Raphicervus
campestris) is remarkably uniform throughout its range and
is not separable into geographical races. Christy (1929)
finds that the African Buffalo (Bubalis coffer) is undifferentiated
over all its range, while the Congo Buffalo (B. nanus) has many
local races. The remarkable differences in local variation

THE DISTRIBUTION OF VARIANTS IN NATURE 107

Fig. 1 1. — Distribution of Buarremon brunneinuchus (i) and B. inortiatus (2).

(From Chapman, 1923.)

of the species of Dicrostonyx (Hinton, 1926, p. 148 and foil.)
should be noted. Grinnell (191 8, p. 241) points out that,

108 THE VARIATION OF ANIMALS IN NATURE

in spite of their powers of flight, bats are as much prone to
form subspecies as other mammals. Possibly this is explicable
on the grounds of localised range, though no facts can be
produced to support this suggestion. Among reptiles the
common African tortoise, Testudo pardalis, ' extends practically
over the whole continent . . . and is everywhere uniform
as regards its colour-pattern ' (Duerden, 1907, p. 74).

Land molluscs tend to fall into well-marked local races
in spite of Rensch's statement. These are especially well
marked if the terrain is favourable to isolation (Gulick,
Crampton, Bartsch, Mayer, Sarasin, Simpson). Even when
these conditions are absent, races may be formed as in Murella,
Helicogena and Iberus (Kobelt), Otala (Boettger), and in sundry
African species (Pilsbry). On the other hand, certain forms
such as Carychium (Thorson and Tuxen, 1930) show no such
tendency. Again, in such forms as Cepea and Cochlicella, though
statistical differences occur in the percentage incidence of
colour-patterns in various colonies, there is no regional differ-
entiation worth mentioning. In contrast with the acute
local polymorphism of Achatinella and Partula in the valleys of
the Sandwich and Society Islands, the land snails of the valleys
of Valais (Piaget, 1921) show no such variation, and although
numerous insular races are found in Liguus on the Florida
Keys (Simpson, 1929), Ampkidromus in the Philippine Islands
(Bartsch), etc., the land snails of the Hebrides and Scilly Isles
are, as far as they are known, quite like the mainland forms.

Amongst insects, taxonomy is still, as a rule, insufficiently
advanced to allow certain conclusions to be drawn. It is
probably significant, however, that in the minute, wingless
Collembola the species often have a very wide range without
any apparent signs of local differentiation. Uvarov (1924)
records an interesting example in the grasshoppers of the
genus Cyrtacanthacris. C. tatarica is found over the whole of
South and Equatorial Africa, including Madagascar, Seychelles,
Comoro Is., Khartum, Massourah, Sokotra, India, Siam and
Ceylon. There is no geographical variation and the species
is extremely constant, though very common. On the other
hand, C. aeruginosa, which is purely African, has three races,
a southern, a western and an eastern race.

It is a remarkable fact (and one which might seem to be
easily explicable on the grounds that there are no barriers

THE DISTRIBUTION OF VARIANTS IN NATURE 109

5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

Fig. 12. — Distribution of S. American Wrens of the Troglodytes musculus
Group. 5, Troglodytes m. atopus ; 6, T. m. slriatulus ; 7, T. m. columbae ;
8, T. m. albicans ; 9, T. m. tobagenis ; 10, T. m. musculus ; 1 1, T. m. rex ;
12, T. in. carabayae ; 13, T. m. puna ; 14, T. m. audax ; 15, T. tecellatus ;
16, T. m. chilensis and, from the valley of Copiapo northward, T. in.
atacamensis ; 1 7, T. m. magellanicus ; 18, 7". m. bonariae ; 19, T. cobbi.

(From Chapman and Griscom, 1924.)

no THE VARIATION OF ANIMALS IN NATURE

to intercourse) that many species of marine Crustacea (Cope-
poda — Scott, 1909 ; Euphausiacea — Hansen, 191 1) are homo-
geneous throughout very extensive areas and pass practically
round the world within certain isothermal limits. It is perhaps
curious that there is no gradual regional differentiation of such
species and that such mutations as occur are so rapidly and
effectively extinguished or spread throughout the population.

Doubtless many of these exceptions may be ultimately
explained by reference to differences of habit, etc., which so
far are unknown. In some cases this seems to be very unlikely.
The contrast between the Oligochaeta and the land Mollusca
is a case in point. We are indebted to the late Lt.-Col. J.
Stephenson, F.R.S., for pointing out many facts in connection
with the slight variability of earthworms. He informed us
that undoubtedly many species are ' peregrine ' and are
carried round the world either as cocoons or adults, probably
in agricultural and horticultural produce. Michaelsen {fide
Stephenson) also postulates the action of winds in dispersing
the cocoons, but Benham criticises this view. Peregrinal
species like Allolobophora caliginosa are remarkably constant and
exhibit very slight or no variation over an enormous range,
and it would seem that the means of intercourse must be fairly
regular if local differentiation is so easily effaced (cf. marine
Crustacea). But there are also many species of earthworms
which are not thus peregrine and have a more localised range,
and these are invariably homogeneous. Lt.-Col. Stephenson
did not think that these species are accidentally transported
from place to place. Moreover the means of transport either
of cocoons or of adults (human agency, birds, winds) should
be also similarly operative in the case of land snails.

It remains to notice some theoretical considerations which
have a bearing on the interpretation of these facts. In the
first place we must emphasise the difference between small
local assemblages having a distinct statistical expression and
larger ' geographical ' groups. The greater average size of
Vertebrates must be of importance here, as it tends to involve
a wider range and less isolation. Small size, on the one hand,
facilitates accidental transport (Nematoda, Tardigrada), and
on the other makes for a homogeneous local population. In
the majority of cases the former influence seems to have been
paramount.

THE DISTRIBUTION OF VARIANTS IN NATURE 1 1 1

Certain other facts are also relevant. In the first place
the relatively small number of species in birds and mammals
has enabled much greater advance to be made in the study
of subspecific differences ; the definition of a large number of
geographical races does not of itself prove that this type of
variation is more common in these groups than in others in
which the number of species is very much greater. Again,
where the number of species is small, the systematist will tend
not to hesitate to introduce a new name for any apparently
stable local form. In such groups as the insects the species
are already so numerous that considerable evidence is needed
before definite named races will be published. In the Lepido-
ptera, in dealing with which authors have been less cautious,
considerable confusion has resulted. The local variation
is so great that it is a difficult and lengthy task to deal ade-
quately with even a single species, and, where species are
numerous, it is unlikely that more than a few have been
sufficiently studied for so-called ' races ' to be very clearly
defined.

Secondly, a geographical race is commonly defined by the
average size, proportion or colour of certain parts. No one
(p. 69) supposes that geographical races are normally uni-
formly homozygous for merely a single differential factor,
so that the variation within the race cannot be regarded as
purely somatic. This implies that the race could be broken
up into a number of varieties differing slightly from one another
in the diagnostic race-characters. The average of these
varieties gives the race, because, being quantitative, these
characters can be given a mean value. But, in other cases,
as often in insects, a species consists of several rather sharply
discontinuous varieties. If these differ qualitatively, they
cannot be averaged : it is possible only to give the proportion
in which the different varieties occur in different parts of the
range. Difference in these proportions evidently defines a
race of exactly the same nature as described in the last para-
graph. But in normal taxonomic procedure the race described
there would receive a name, whereas in the second case each
of the distinct forms would receive a name, but there would
be no name for the various populations defined by consisting
of different proportions of the named forms. The example
of Harmonia axyridis ( (h), p. 103) exhibits this difficulty.

ii2 THE VARIATION OF ANIMALS IN NATURE

Thirdly, we are a little doubtful if the data for various groups
are really comparable and whether samples of populations
consisting of a few individuals, such as are used in mammals
and birds, afford a sound basis for distinguishing local races.
Sumner (191 8, p. 292) seems to express this doubt concerning
the races of small mammals. We do not as a matter of fact
think this vitiates the general principle, for there are groups
(e.g. the Cephalopoda) in which the numbers used are equally
low and yet few races are recorded. What we feel is that
comparable data are required and that some modifications
of the alleged incidence of race-formation might result, if
large numbers were regularly used.

In conclusion, it seems likely that geographical variation
will ultimately be found to be as frequent in groups like
terrestrial arthropods and molluscs as in vertebrates. Where
all the present evidence is against the likelihood of such races
being discovered, it will be usually found that special habits
and other factors that prevent isolation and colony-formation
are mainly responsible. Again, in some species we must
look to the inherent capacity for variation as a cause. It must
always be recalled that our knowledge of variation is at present
very unequal in its incidence in the various groups and is less
easily obtainable in some than in others.

Examples may now be given of the occurrence and dis-
tribution of geographical variation in various groups.

1. Geographical variation in Lygaeus kalmii (Hemiptera, Hetero-

ptera) .

Parshley (1923) has shown that there is a clearly marked
eastern and western race in the United States. These meet
at a line joining Winnipeg to Brownsville, Texas. Along this
line intermediates occur, which cannot be referred to either
race. Since the species is, in addition, highly variable in
colour, it is only possible to recognise the geographically
significant characters by careful study.

2. Geographical variation in water beetles.

Omer Cooper (1931) summarises the evidence for two
examples. The extremes in each case are treated as species,
but they correspond to what are called geographical races in
other groups. Thus in Deronectes depressus and D. elegans there

THE DISTRIBUTION OF VARIANTS IN NATURE 113

are differences in size, shape, colour, tarsal claws and width
of penis. In the south of England only elegans is found, while
in the north of Scotland only depressus and in Ireland only
depressus or approximating intermediates. But in N. England
and S. Scotland a completely intergrading series is found.
There is a similar relation between Gyrinus natator and G. sub-
striatus, except that the overlap appears to be wider.

3. Geographical variation in butterflies.

In this group geographical races have been more studied
than in any other order of insects. Frequently, however,
races have been described from too little material and their
geographical limits are often very uncertain. A well-studied
example is described by Waterhouse (1914, 1922), who deals
with the races of an Australian butterfly, Tisiphone abeona.
This species is found on the S.E. coast to the seaward of the
main dividing range. Five races follow one another in
succession down the coast. Two of the races have been proved
to be interfertile with a third, which is not in direct contact
with either of them. Another two races appear to interbreed
and produce peculiar forms not known elsewhere. A similar
outburst of peculiar forms where two races meet is recorded
by Harrison and Carter (1924) in Aricia medon in England.
Doubtless a variety of genetic conditions will determine
whether the recombinations resulting from an interracial
cross shall produce an intergrading series or an unexpected
new type.

4. Geographical variation in fleas.

Jordan (1931) gives an interesting example in the variation
of the common mouse flea Ctenophthalmus agyrtes. This species
is represented in Western Europe by five races — one in England
and N.W. France, one in E. France, Germany and Switzer-
land, three in Switzerland and N. Italy (separated by various
mountain ranges). There are several peculiarities in this
distribution. First, the presence of the English Channel has
not led to the formation of a peculiar English race. Secondly,
the environment of fleas is unusually constant and wide
differences in external conditions do not appear to affect them
(e.g. in the Alps they occur without modification right up to
the tree limit). It is difficult, therefore, to see why races

ii4 THE VARIATION OF ANIMALS IN NATURE

should evolve- where there are no very definite barriers, e.g.
races of E. and W. France. On the other hand, the existence
of several races in Switzerland, where mountain barriers are
numerous, suggests that isolation alone may account for the
changes observed. The identity of the English and W. French
races, however, is in disagreement with this view. Possibly
a survey of the hosts most commonly affected in different
areas might be important, though the variety of hosts appears
to be unusually great.

5. Geographical variation in fishes.

Examples are available of intense ' local race formation '
in the sedentary ^oarces viviparus (Schmidt, 1918) and in
species, such as the Atlantic Cod (id. 1930), which have a
wider range. The latter is split up into ' a mosaic of popula-
tions,' each of which has a peculiar statistical facies in respect
of the two characters (number of vertebras and fin rays)
studied by Schmidt.

6. Geographical races in squirrels and mouse-deer.

It is well known that the squirrels of the Old World tropics
provide examples of some of the most extraordinary racial
complexes. The data are worth some consideration, since
they raise the question how far the variation of other
animals would prove equally refractory to schematic treat-
ment if more material were available. The races in squirrels
are largely separated by colour-pattern, differences in which
are sharply marked and easily studied. In such forms
as the smaller Muridae, where the study of each individual
requires a far more tedious technique and the characters
cannot be seized at a glance, a similar complexity might more
easily be masked.

Evidence as to the African squirrels (Heliosciurus) may be
found in Ingoldby (1927) ; certain Burmese forms are dealt
with by Oldfield Thomas and Wroughton (1916), and Banks
(1931) discusses the Bornean races of Sciurus prevostii. The
last-named species has numerous races in Malaya, Sumatra
and Borneo. The latter island has about eight races, one of
which is also found on Sumatra or at least represented by a
closely similar form. Where the races overlap, intermediates
are found, almost certainly as a result of intercrossing. Some

| V— ' M^f*

\ \? Bordeaux Wt*-»"

Figs. 13A and 13B. — Male Genitalia of Races of Ctenophthalmus agyrles drawn on a

Map of Western Europe to show Distribution of Races.

6 and 6a, Race celticus ; 7 and 8, agyrtes ; 9, provincialis ; 10, oreadis ; 11, verbanns.

(From Jordan, 193 1 .)

n6 THE VARIATION OF ANIMALS IN NATURE

of the races, however, are sharply isolated from one another
by rivers. Oldfield Thomas and Wroughton also note the
importance of rivers as barriers to the Burmese forms. Banks,
further, finds that individual variation within the races is
extreme and appears partly to produce forms which might be
called races were it not that they do not form definite popu-
lations. Thus in S. prevostii borneensis, according to Banks
(I.e. p. 1336) — ' No two specimens are alike, and the
variation is endless.' Both colour and pattern are affected,
and Banks shows it is very difficult to correlate the characters
of the races with any known feature of the environment.
Apart from one mountain race, most of them appear to live
under very similar conditions, the island being tropical through-
out. It is also interesting that certain races appear to have
a discontinuous distribution, such as has already been noted
in the flea Ctenophthalmus agyrtes. A similar example of dis-
continuous geographical groups is found in the Carrion Crow
(Kirkman and Jourdain, 1930, p. 2). An E. Siberian form
of this species is separated from the main area of the species
by the whole distributional area of the Hooded Crow. It
cannot, of course, be proved without elaborate genetic experi-
ments that apparently similar forms are really identical, but
the formation of similar races in different areas within a larger
patch of uniform conditions is strongly suggestive of the
convergent establishment of the same chance combinations
of genetic factors. It may be mentioned that Bequaert (1931)
has shown that the geographical race of the Hornet (Vespa
crabro) inhabiting the British Isles, resembles a Chinese race
far more closely than it does the adjacent continental form.

In the African Heliosciurus, Ingoldby has shown that similar
races tend to be found on each side of the equator, with races
of a different type lying between them. Here there is a greater
possibility of a direct environmental effect and, according to
this author, the races in two localities with identical ecological
conditions are the same. It is not difficult, however, to find
instances where there is no obvious correlation with the
environment ; in fact such correlation appears to be the excep-
tion rather than the rule. Thus Miller's study of the Malayan
mouse-deer (Tragulus) (1909) shows that numerous races have
been developed under conditions as nearly uniform as possible.
In this genus races are more often developed on the smaller

THE DISTRIBUTION OF VARIANTS IN NATURE 1 1 7

islands than on the larger ones and on the mainland, suggesting
that isolation has been the most important factor. It is curious
that some of the races occur on more than one small island.
Admittedly these islands are usually close to one another,
but not always closer than other islands which bear distinct
races. Further, the most similar races do not usually inhabit
the closest islands. Taking the islands as a whole we see
a progressive change in colour from the mainland form, but,
as the various changes are scattered at random amongst the
islands, it is unlikely that the series represents the actual line
of evolution, which was probably polyphyletic.

In considering geographical races it is a matter of some
importance to examine the normal size of the racial population.
Many races of course exist over enormous areas and include
millions of individuals, but in the case of smaller units taxonomic
practice becomes somewhat arbitrary. It is evidently con-
venient to have a name for any race which covers a large area,
even if structurally it is little differentiated from its closest
allies. But in more localised races a higher degree of divergence
tends to be demanded. Thus a statistical examination of the
populations of a species inhabiting a number of small islands
might show that each had a different mean character, but it
might be taxonomically very inconvenient to give a name to
each. On the other hand, unnamed variations tend to be
ignored, and in making any such survey as the present only
the most general information about such forms can be obtained.

We may give examples. Perhaps a record for smallness
of racial area is held by Lacerta simonii (Cott, 1932), which
inhabits a small rock with a surface of perhaps 1,000 square
yards in the Canaries. Cott estimates the total population
at not more than a few scores of individuals. The Skomer
Vole is confined to an island only a few square miles in extent,
and the same is true of many other island races. Isolated
colonies of the Rabbit (Oryctolagus cuniculus) are known which
are quite distinct in colour, e.g. a mouse-coloured race on
Sunk Island in the Humber (Barrett-Hamilton and Hinton,
I.e. pp. 196-9). The Skomer Vole is given a name because
it is a relict form whose nearest allies live in the Hebrides,
while the Rabbit is unusually variable and there are too many
trifling local variants for a name to be given to any one. In
the moths of the genus J^ygaena, particular colonies have often

1 2

Fig.

I. " 2-

[4. — African Squirrels of the Genus Heliosciurus .

1 and 2, Forest forms ; 3 and 4, Grassland forms.
(From Ingoldby, 1927.)

THE DISTRIBUTION OF VARIANTS IN NATURE 1 19

a distinctive pattern ; possibly in some of them the mean of
the colony would not actually be repeated anywhere else in
the range of the species. But such colonies are so numerous,
and so often show a considerable range of variation, that it
is useless to name them all. Thus, while taxonomic procedure
has very good practical arguments in its favour, it tends to
exhibit geographical variation more distinct from other types
of variation than it really is.

Physiological Races (see also Chapter III, p. 73). — There
is no theoretical reason to suppose that the physiological
(instinctive, psychical, etiological, etc.) characters of species
should be less variable than the morphological except in so
far as variation in the latter is less likely to impair viability.
In the Protozoa, strains differing in various physiological
properties (immunity and virulence) have long been known.
The literature of entomology, ornithology, etc., is full of
descriptions of individuals with aberrant habits or instincts.
In most cases, however, the previous history of the individual
was unknown, so that little can be concluded except that
instinct is capable of modification. It is easier to study the
phenomenon when a whole population exhibits such a change.
Such populations are termed ' biological races ' or ' physio-
logical strains ' of the species concerned. If physiological
characters are inherited in the same way as morphological,
the same tendency to group-formation and subdivision of the
species might be expected in them, some groups being charac-
terised mainly physiologically, others mainly morphologically.
A very much more complete knowledge of animals than we
possess might perhaps break down the distinction.

Some of the data as to biological races are considered
elsewhere (Chapters II, III and VII), so that we shall en-
deavour here mainly to establish that physiological differen-
tiation occurs in all degrees. As an instance of the asso-
ciation of minute physiological differences associated with
almost equally small structural ones, we may mention the
work of Bodenheimer and Klein (1930), who deal with three
subspecies of the ant Messor semirufus in relation to temperature.
It was found that each race had a different optimum tempera-
ture for normal activities (viz. 18-4°, 19 , 20-3° C). This
and similar evidence that is now accumulating show that
at all grades of morphological differentiation physiological

120 THE VARIATION OF ANIMALS IN NATURE

differences are likely to be present as well, even if requiring
refined methods for their detection. Food- or host-selection
is the feature in which physiological differentiation has been
most studied, but Thorpe (1930) also notes differences in
the susceptibility of scale-insects to fumigation, and differences
in song may also be mentioned. Owing to the difficulty of
the investigation not very many examples have been really
exhaustively examined, but it is clear that various stages can
be traced from forms which differ only in physiology to those
which also differ morphologically, eventually to such an
extent that they are regarded as closely allied species.

Hachfeld (1926) records that in the bee, Trachusa byssina,
different individuals use different plant-leaves with which to
build their nests. In different localities different plants are
the main source of material.

Hackett and Missiroli (1931) have investigated factors
leading to the reduction of malaria in various areas in Europe.
It is practically certain that the disappearance of this disease
in some localities [e.g. parts of Italy) is due not to preventive
measures but to the establishment of definite zootrophic races
of Anopheles which attack domestic animals but not human
beings. Another instance of purely physiological races may
be found in the wasp Tiphia popilliavora. This is being im-
ported into the United States from the East to control the
introduced Japanese Beetle {Popillia japonica) , which has proved
a serious pest. Hollo way (193 1) finds that the forms of this
wasp found in Korea, China and Japan respectively cannot
be separated into geographical races on the basis of their
structure, but that they are so different physiologically that
three strains must be recognised if economic measures are
to be successful. The strains differ principally in their tem-
perature-relations and their consequent fitness to survive in
the climate of the United States. The strains differ, for
instance, in their length of life, developmental period and in
the minimum temperature for mating. As a result of such
differences the Chinese race is able to maintain itself only at
the extreme southern border of the area now infested. For
control in the greater part of eastern U.S.A. the Japanese race
is alone suitable.

Fulton (1925) finds races of tree-crickets, Oecanthus, which
differ in song, method of oviposition and habitat, but not in

THE DISTRIBUTION OF VARIANTS IN NATURE 121

structure, while Myers (1926) states that the song is the most
stable single character in the cicadas, though here morpho-
logical differences are also fairly conspicuous. In grasshoppers,
taken as a whole, structure would appear to be more distinctive
than song, though the latter is difficult to define owing to
environmental effects (temperature, presence of other indi-
viduals, etc.). Promptoff (1930) records statistical local
differences in the song of the Chaffinch in two different areas
of Russia. Again Kinsey (1930), in his valuable revision of
the genus Cynips {Spathegaster and Dryophanta, auctt.), finds
several pairs of species or races which are only to be distin-
guished by their galls. His actual summary for the genus
(p. 38) is as follows : 52 species have structure more distinctive
than the galls ; 24 species have galls more distinctive than
structure ; 17 species have the two equally distinctive. The
formation of the galls is known to be due to the action of the
larval gall-wasp.

Thorpe's account (1929, 1931 ) of the races of the Small
Ermine Moth (Hyponomeuta padella) shows that structural
and physiological differences are about evenly balanced,
neither being very great. There is a distinct food preference,
indicated by oviposition-response of the female and even more
by larval choice ; members of one race cross with one another
more easily than they do with members of the other, and there
are slight overlapping colour-differences between the adult
moths ; the larvae construct different types of cocoons. The
two forms of the Human Louse (Pediculus) are somewhat more
distinct and crossing is liable to lead to abnormalities in the
hybrids.

Unfortunately data as to selective mating between races are
very scanty. If we knew more it might be possible to regard
species differing only in the male genitalia as a special type
of biological race. In a number of forms (Lucilia, the blow-
fly ; Chironomus-midgcs, etc.) the females are morphologically
indistinguishable and the maintenance of the species must
depend on the reactions of the male, perhaps to a scent emitted
by his mate.

In connection with biological races it is interesting to
consider the differences which may be found in the develop-
mental stages of animals, especially in larval forms. If we
eliminate species which are still imperfectly known, it is probable

122 THE VARIATION OF ANIMALS IN NATURE

that of the remainder the majority are more easily recognised as
adults than as larvae. But this is not always true. Thus Edwards
(1929) points out that a number of Chironomid midges are

Fig. 15. — Respiratory Siphons of Larvve of Culicella morsitans (above) and
C. fumipennis, of which the Adults are almost indistinguishable.

(From Lang, 1920.)

almost indistinguishable as adults but have totally different
larval habits or structure. In some mosquitoes two different
types of larvae have been found to produce identical adults
(Lang, 1920; Culicella morsitans and C. fumipennis). In this
case the larvae are said to be dimorphic, because it is usual
to lay most stress on adult structure. The egg-rafts of some

THE DISTRIBUTION OF VARIANTS IN NATURE 123

mosquitoes are similarly dimorphic. The common British
moths Acronyctapsi and A. tridens may also be mentioned. The
larvae differ sharply in colour, though the adults arc separable
only by the genitalia. In all such cases it is logical to claim
that evolution has progressed further in the larvae than in the
adults, just as in biological races evolution has been in the
direction of physiological rather than structural divergence.
It is of some interest to show that the tendency to form local
populations does not affect only structural characters. The
existence of biological races evidently provided partial proof
of this, but we may add a number of other instances of local
segregation of what may be called ' non-taxonomic ' characters.
Local variation in the extent of sexual dimorphism is not
at all rare, but is best considered a special case of normal
group-formation in structural characters. There is much
variation in seasonal occurrence in most insects with a wide
range. It is usually unknown to what extent this character
is due to the direct action of the environment. Probably the
genetic element is larger than is commonly supposed. While
often the number of broods gradually increases as one goes
south, in other cases closely allied forms have a different life-
cycle in the same district. Sometimes the effect of temperature
is reversed. Thus, in gall-wasps, Kinsey (1930) finds that
the species emerge earlier in the north, and Willey (1930,
pp. 79-80) records a comparable condition in Copepods, in
which growth is faster in the north. Many butterflies which
have more than one brood a year show marked differences
between the spring and summer broods. Such seasonal
change is much subject to local variation and may be almost
absent in some parts of the range (cf. Ford, 1924).

Gurney (1929) shows that some Copepods are locally
dimorphic in size, while elsewhere this character is distributed
in a normal curve. In some species one sex alone shows the
dimorphism. This may be compared to the dimorphism in
the males of the Common Earwig (Forficula auricularia) . Bateson
and Brindley (1892) showed that in some localities high males
were much more prevalent than in others. Stephenson (1929)
records various methods of reproduction separating species of
Sagartia. Amongst eight species there are five methods.

Local variations in the sex-ratio are also well known.
The subject has been dealt with at some length by Yandel

124 THE VARIATION OF ANIMALS IN NATURE

(1928), who finds that in many Hymenoptera there is a
tendency for the species to be parthenogenetic in the northern
part of their range, but to reproduce normally in the south
(cf. also Brues, 1928). Poulton (1931) described similar local
anomalies in the sex-ratio in the Fijian butterfly, Hypolimnas
bolima. There appear to be a good number of instances of
insects which possess two types of females, male-producers
and female-producers, but the two types are not often geo-
graphically segregated.

Summary

Any account of variation is unfortunately limited by the
inability to present more than a small selection from the vast
mass of available data. It has been usual in the past (and
the practice is difficult to avoid) to construct all-embracing
theories on the basis of selected species or genera which supply
favourable data ; the theories based on the genetics of Droso-
phila or of Oenothera are cases in point. Obviously the best
method would be to treat all doubtful points statistically and
to state definitely that a particular type of variation occurred
in such-and-such a percentage. In the present state of taxo-
nomy no numerical statement of this sort is possible except
perhaps for a few well-worked groups. For, in the absence
of experimental investigations, it is often quite uncertain
whether particular variations are inherited, and moreover
the diverse types of variation encountered are very numerous
and difficult to classify, so that statistical treatment might in
any case be liable to serious errors. In the preceding account
we have tried to choose our examples fairly and not to pick
out merely those which support views we already hold on
other grounds.

Up to the present we have not considered the effects of
isolation and the different ways in which it can be brought
about. Evidently isolation of one sort or another is a prime
factor in the process of group formation. Geographical
isolation is the type most easily recognised, and it is on this
account that taxonomists have evolved the conception of the
' geographical race,' a term applied to minor categories,
whose ability to interbreed with their closest allies is held in
check only by more or less marked spatial separation. Other

THE DISTRIBUTION OF VARIANTS IN NATURE 125

categories, of a similar structural grade, have been termed
' subspecies ' by some entomologists. These subspecies, unlike
geographical races, live side by side ; but they can be called
species only if we give up all attempts to indicate (in any
one group) the same degree of divergence by the latter term.
It is probable that these subspecies occur in some groups more
than in others owing to differences in the mode of reproduction,
particularly in the length of the breeding season, in the way
in which the sexes find one another and in the degree of
development of gregarious habits. Subspecies tend to occur
in any group in which non-geographical methods of isolation
are easily effective. The great possibilities of such isolation
have often not been sufficiently realised and undue weight
has been given to geographical effects.

The following are the more important general results
which emerge from our survey :

1 . We have discussed at some length the antithesis between
individual and regional and geographical variation. In
some cases the antithesis stressed by Rensch and others between
populations broken up into clearly defined regional or geo-
graphical groups and those in which the variants are more
universally distributed is clear and can be shown in some
instances to be due to differences in habits, size, etc. We
believe, however, that the distinction is more apparent than
real and that no particular significance is to be attached to it.
To begin with, there seems to be a likelihood that geographical
variation will be found to be less clearly cut when the relevant
forms are more exhaustively studied and knowledge of their
distribution is based on more material. Series of geographical
races are easy to demonstrate when the samples are not too
large. Secondly, while we admit that clearly-cut qualitative
divergences on a geographical basis are not so typical of
groups such as terrestrial molluscs and arthropods, it is quite
evident that the proportions of the variant types in these
groups define populations quite as definitely as average
dimensions, colour, etc., define those of vertebrates. It is of
secondary importance that the regional divergences among,
e.g., populations of land molluscs tend to be smoothed over as
a result of the size and habits of these animals and in certain
(but by no means all) of their characters by reason of their
plasticity. When many characters of vertebrate populations

126 THE VARIATION OF ANIMALS IN NATURE

are examined statistically (Sumner, Schmidt), the same
quantitative local divergences are discovered as those observed
in populations of land snails. It seems to be true on the
whole that there is a lack of innumerable individual variations
in vertebrates that requires explanation, though the obser-
vations of Fowler and Bean (1929) on variation in fishes of
the order Gapriformes must prepare us to realise that indi-
vidual variation is far more frequent than Rensch has allowed ;
but perhaps the wider range and consequent less susceptibility
to minor isolating influences render their populations more
homogeneous. It is also possible that a more highly evolved
physiological control makes them less susceptible to external
factors. A study of the variability of sedentary mammals
(such as small rodents) contrasted with that of more widely
ranging forms (carnivores and ruminants) is much to be
desired.

2. The very frequent occurrence of variants established as
a small percentage of a population and at the same time living
along with the typical forms seems to us of some importance.
Many more examples are available of this phenomenon than
those which we have cited.

3. The frequent occurrence of statistical divergences calls
for attention. It is not without significance that, when
populations are broken up by divergences of this kind (p. 99),
the latter can be maintained over periods of about ten years,
at least as far as the admittedly imperfect records allow us to
judge. As to the origin of these divergences it seems most
unlikely that they are due to selection. They sometimes occur
under identical ecological and bionomic conditions and,
unless we appeal to the argumentum ad ignorantiam, are most
unlikely to be produced by selective adaptation to local con-
ditions. For a similar reason they do not appear to be produced
by the direct effect of the environment. We are thus forced
to conclude that they are produced by the effects of local
isolation or obligatory preferential mating working on available
stocks of hereditary material.

4. We have introduced somewhat cautiously the idea that
certain species have a more marked proneness to local and
regional variation than others, apart from any habits, etc.,
which might promote this feature. The contrast between
the South American Wren and Buarremon (p. 106) is an

THE DISTRIBUTION OF VARIANTS IN NATURE 127

instance of this. It seems evident that all animals are not
equally prone to receive the impress of their environment nor
in the same state of mutational activity.

5. The general impression that one gets from a survey
such as the foregoing is that groups are formed by the spread
of individual variants rather than by mass transformation.
What we find is a gradation from single variants, or variants
represented only by a low percentage in the population, to
larger and more distinctive assemblages and eventually to
distinct regional geographical groups. We do not know, of
course, how many of the smaller groups may not be on the
way to extinction ; but we may assume that at least half of
them are not and that this possibility does not vitiate the
general conclusion that there is a process at work in nature
which facilitates the multiplication of single variants. If the
latter were spreading from single loci the mosaic of poly-
morphism is exactly what one would expect to find. Rensch's
attempt to show that variants are distributed in ' Rassenkreise '
under the influence of differentiated environments seems to
us to break down on three counts :

(a) The very general occurrence of polymorphism is
a proof that the environment is not the direct trans-
forming agency. The only way in which those who favour
that view could explain the occurrence of differentiated forms
living side by side in the same habitat is to suggest that they
acquired their differences elsewhere and have subsequently
met. But, as Robson (1928, p. 174) has pointed out, this
involves explaining (1) the frequent lack of epharmonic con-
vergence and (2) the means of spreading.

(b) In numerous cases variants are not arranged with
reference to environmental gradients and many races range
unmodified through a variety of environments {cf. Sumner,
1932, etc.).

(c) To argue that many of the observed changes that are
correlated with environmental differences may only be somatic
is but a negative objection ; but it is a great weakness of
Rensch's case that there is so little experimental evidence
that local races, etc., are of a fixed heredity. We do not wish
to ignore the many and striking cases of structural and en-
vironmental trends. We would even admit that in such cases
mass transformation of populations may be possible. But

i 2 8 THE VARIATION OF ANIMALS IN NATURE

we hold that the occurrence of the various grades of poly-
morphism is far more widespread and far more significant,
and whether we are considering groups such as colonies of
land snails which are distinguished by the varying proportions
of a number of characters or the statistical differences in the
occurrence of single characters, we cannot fail to be impressed
by the evidence for a process of multiplication of certain types
rather than their production en bloc. Nevertheless, if the
evidence from the facts of distribution suggests such a process,
it does not justify any conclusions as to how it took place.

CHAPTER V

ISOLATION

The importance of isolation in evolution was first strongly
insisted on by M. Wagner (cf. summary of his work, 1889).
Darwin also allowed its influence to be considerable, as, for
instance, in the production of island races. Both these authors
regarded adaptation to the local conditions as of fully equal
importance (cf. Wagner, I.e. p. 401). In Chapter I it was
indicated that isolation may be regarded as playing two
opposing roles in the process of group-formation, viz. the
maintenance of the identity of groups and the splitting up of
large groups into smaller ones. In the present chapter this
matter is considered more fully.

The more general problems of geographical distribution
need not be given special attention. They have been dis-
cussed at length in many works wholly devoted to the subject.
For the same reason actual dispersal mechanisms are only of
secondary interest. These also have been much discussed,
but well-authenticated data are somewhat meagre and scarcely
sufficient to enable us to formulate any general relation
between powers of dispersal and race-formation. Allusion
has already been made to this difficulty in Chapter IV
(p. 104), and it may be added that any such relation
might be obscured by innate tendencies to race-formation
which appear to be independent of dispersal. Two main
types of isolation itself may be recognised. Geographical or
topographical isolation is operative when two populations are
separated by uninhabitable country. Sections of a species
isolated by such a barrier would, for some time after their
separation, be able to interbreed if they could be carried
across the barrier. Isolation of this kind is temporary, since
without changes in the animal itself it is always liable to break
down as a result of modification of the barriers themselves
(e.g. movements in the earth's surface). Jordan (1896, p. 442)

130 THE VARIATION OF ANIMALS IN NATURE

indeed states that, if in the course of divergence a point is
reached after which it is impossible for the diverging form to
coalesce with the parent stock, we are given by this point a
definite means of distinguishing varieties from species.

The changes in animals themselves which make inter-
breeding actually impossible form the second or permanent
type of isolation. Permanent isolation may be the result of a
variety of factors, and an important consideration is to determine
whether it can ever be developed in the absence of some degree
of geographical separation. The establishment of geographical
isolation might often be due to geological changes within the
area of a widely ranging species, but we must also recognise
the importance of the wanderings of the animals themselves.
The continual invasion of all countries and habitats, however
apparently uncongenial, is a commonplace of natural history.
Where the invaders have to overcome great difficulties, we
usually find the formation of isolated colonies, as in oceanic
islands.

Permanent isolation may arise frequently from ' accidental '
changes in the structure and habits of populations no longer
in a position to eliminate or assimilate the variant individuals
by free intermixture. The actual mechanism which prevents
allied species from interbreeding is rarely understood in detail,
but very often there seems to be a great difficulty in explaining
how the mechanism can have been perfected, since the charac-
ters on which it depends appear to be of little use to in-
dividuals or even to the species as a whole.

Although we now suspect that some measure of permanent
isolation may be developed amongst individuals inhabiting
a continuous area, yet it is probable that geographical isolation
is more often than not a necessary preliminary. The temporary
nature of the latter type of isolation makes it important for
us to examine the rate at which topographically isolated
populations diverge from one another. It may be admitted
that the degree of permanent isolation is only very roughly
correlated with that of the resulting morphological divergence,
but in so far as the latter is likely sooner or later to entail
permanent isolation, the rate of divergence under geographical
separation becomes relevant. We shall therefore digress to
consider the available evidence as to the time necessary for
the establishment of a new species or subspecies.

ISOLATION

*3 l

In this inquiry we are obliged to depend on the relatively
few groups which both provide suitable material and have
been subjected to sufficient taxonomic study. We are not
so much concerned with the maximum as with the minimum
time which such a change may take. We can never know
whether a fossil form which is identical in structure with a
modern one would, in fact, be able to interbreed with it.
But even in the majority of living species we do not know
whether interbreeding is possible, and we are endeavouring
rather to estimate something which has a meaning in present-
day taxonomy, viz. how long it has taken to evolve differences
which would be considered sufficient to separate races or
species, if they characterised recent forms.

Modern species known to have persisted since pre-Tertiary
times are rare. An interesting example is the shark Scapanor-
rhynchus owsteni, which was first described from fossil teeth in

Fig. i 6. — Scapanorrhynchus owsteni.
(From Norman, 1931.)

the Upper Cretaceous but has since been found living off
the coast of Japan (Norman, 1931, p. 124). In other
instances, as perhaps in the Brachiopoda, the characters
available for study in the fossil state are so few that the com-
parison with recent species could not be expected to be very
enlightening. But it appears that, just as some species with
discontinuous range soon form numerous races while others
remain relatively homogeneous, so the rate of evolution,
judged by palaeontological evidence, must be variable from
group to group, and probably depends on innate potentialities.
Wheeler (191 3, chapter x) has discussed the fossil history
of the ants. Many of the amber fossils are perfectly preserved
and are as capable of exact study as recent specimens. In the
Sicilian Amber (Lower Oligocene) nearly 69 per cent, of the
genera are still living. Three species, belonging to different
genera, are not separable from well-known living forms. There

1 32 THE VARIATION OF ANIMALS IN NATURE

is some evidence that even the main features of the habits of
ants were established at this early date, though it appears that
the polymorphism of the workers was not developed till the
Pleistocene. Apparently the species and genera of ants were
established at a much earlier date than those of several other
groups. If such a species as Ponera coarctata (Wheeler, I.e. p. 174)
has really existed with little change from the Lower Oligocene,
then only the most permanent geographical barriers would
have any effect on its divergence. Unless permanent biological
isolation was set up, there would be ample time for two isolated
races to be joined together again in the course of so prolonged
a specific history.

Lapouge (1902) has given some account of the beetles of
the genus Carabus found in the Mid-Pleistocene of Belgium.
In this genus the surface sculpture of the elytra is highly
distinct and provides some of the most important characters
for separating species and races. The fossil elytra could all
be referred to existing species, except in one case ; but the
sculpture was nearly always somewhat different, to an extent
which in a modern form would be considered deserving of a
varietal or racial name.

Borodin (1927) has published some data on the Clupeid
fishes of the Caspian Sea and a neighbouring lake. Certain
subspecies have probably been isolated from one another since
the second interglacial period (ca. 350,000 years) . The changes
they have undergone are not yet very great. Analogous data
are recorded of another fish (Cottus) in the Swedish lakes
(Lonnberg, 1932) and of the prawn, Limnocalanus (Ekman,

1913)-

The British mammals provide perhaps the best material

for an inquiry of this nature. The evidence for each species

is given by Barrett-Hamilton and Hinton (1911-1921). Two

main types of evidence are available. First, in numerous

instances, an existing species is found fossil in the Pleistocene

as an identical or a scarcely different form, and we have some

idea as to the length of time the species has remained unaltered.

Secondly, in a few specially valuable instances, a species which

is now represented by a purely British race does not occur in

the British Pleistocene, and must have evolved to the extent

to which it differs from its continental representative since that

period.

ISOLATION 133

The following data for British insectivores and rodents are
derived from Barrett-Hamilton and Hinton (1911-1921).

(1) Adequate fossil data not available : 9 species.

(2) Species not known in the Pleistocene, but now repre-

sented by a distinct British race : 3 species (Common
Hare, Field-mouse (Microtus hirtus), and Water-rat
(with two races) ).

(3) Form apparently identical with the modern repre-

sentative known from at least Late Pleistocene :
(a) No British race : 4 species {Epimys rattus, Shrew,
Pigmy Shrew, Rabbit), (b) With a British race :
3 species (Irish Hare, Northern Field-mouse (Microtus
agrestis) , Apodemus flavicollis) .

(4) Late Pleistocene form racially distinct : (a) No British

race: 2 species (Mole, ? Water-shrew), (b) One
or more British races : 4 species (Apodemus sylvaticus
(2 races), Skomer Vole (3), Bank Vole, Orkney
Vole (5) ).

The examples under (2) are particularly instructive, since
it is almost certain that fossils would have been found had the
animals been present in the Late Pleistocene. On the other
hand, since there is now a distinct British race, or, in the
Water-rat, two races, we can say that this degree of evolution
has taken place since the Pleistocene. 1

In the six species included in (4) evolution has been rapid
enough to produce new races since the Late Pleistocene, while
in the seven species under (3) there has probably not been
much change since the Pleistocene.

Evidently the data are not sufficient to support much
speculation, but they do at least suggest that in the rodents
and insectivores, of which at least the former group appears
to evolve very rapidly, the evolution of a new race normally
takes an interval of time not much shorter than that intervening
between the end of the Pleistocene and the present day. This
period of time is well known to have been sufficient for con-
siderable changes in geographical barriers and we may surmise
that, with evolution working at this rate, intrinsic methods

1 An alternative hypothesis would be that the British form had remained
unaltered, and that it was the continental representatives that had changed.

K 2

i 3 4 THE VARIATION OF ANIMALS IN NATURE

of isolation are a very necessary supplement to any purely
topographical isolation.

With this preliminary conclusion, we shall now return to
the main theme of the chapter and consider first topographical
isolation in somewhat greater detail, before passing on to the
intrinsic factors. The mere fact that most species have a more
or less extensive range automatically introduces a measure of
isolation between the more widely separated individuals. We
have already reviewed this question in Chapter IV, where we
came to the conclusion that, while habits and mode of repro-
duction may predispose a species to race-formation, the latter
process is not a very good index of the extent to which the
species-range is broken up by topographical barriers. Intrinsic
factors exert an important effect, which is at present largely
unpredictable. Possibly some of the anomalies might be ex-
plained away if we knew more of the minor migrations of
individuals that occur within the range of many species.

An important point is that relatively slight barriers often
appear to be sufficient to determine the limits of races or
species. Thus in the Central Arabian desert, two races of
the rodent Meriones syrius (Cheesman and Hinton, 1924)
inhabit different stream valleys separated by only a mile of
bare limestone plateau. The intervening area is inhabited by
two quite distinct species. The habitat barrier is here much
sharper than would be normal in ordinary temperate regions.
Again, Wagner (1889, pp. 53-7) gives some instances, in various
groups of animals in N. Africa and Syria, of rivers acting as the
boundaries of races or species. In Chapter IV we have also
noted this in the case of squirrels (p. 116).

Probably far more ecological knowledge of particular
species is required for a profitable discussion of topographical
isolation on continuous areas. It is possible, however, briefly
to review the problem of ' island-races,' since here the same
difficulties arise but in a more clear-cut form.

When once a population has been cut off or immigrant
individuals have succeeded in reaching an isolated area, there
is much evidence in favour of the view that sooner or later the
fauna will undergo larger evolutionary changes. Probably
the oceanic islands, such as the Hawaiian or the Galapagos
groups, are the best examples of a high degree of geographical
isolation. Under these conditions it is well known that the

ISOLATION 135

proportion of endemic species is very high, and often what was
probably a single immigrant species is at the present day
represented by a large genus (cf. Perkins, 191 2).

The effects of isolation in these extreme cases appear
sufficiently striking, but there is a danger of overestimating
the part that geographical isolation has played in the evolution
observed. The enormous area of continuous tropical forest
covering the larger part of northern South America is probably
proportionately quite as rich in endemics. The distribution
of the fauna of South America is still very imperfectly known,
but it appears likely that an enormous number of species have
developed under the relatively constant rain-forest conditions
without the intervention of any very definite barriers. Some
species would appear to occur over the whole area, while others
are apparently definitelylocalised ; but much more information
is needed on this point. Again, in the Hawaiian Islands with
their singularly stable and relatively uniform environment
(especially before the arrival of Man), numerous allied species
have often been evolved on one island. Further, while islands
as a whole are characterised by the endemism of their fauna,
there are a good many exceptions. We may instance the
following :

Crustacea. Lowndes (1930) records that, in a collection
of Copepods from the New Hebrides, practically none
of the species are endemic. Many are identical with
British species, though in this group dispersal powers
would not be expected to be very effective. The
Ostracods, on the other hand, are nearly all endemic,
though special dispersal mechanisms (resting eggs, etc.)
are developed.

Spiders. No peculiar forms occur on the Scilly Isles, Lundy
Island or Channel Islands (Bristowe, 1929a, 1929^,
1929c). On the whole, dispersal power (by gossamer) is
good, but the incidence of this power throughout the
order requires investigation (cf. Bristowe, 1929c).

Hydracarina. Lundblad (1930, p. 24) records only one
endemic variety on the Faroes.

Myriapoda. No endemics * on the Faroes (Hammer and
Henriksen, 1930).

1 i.e. definite subspecies.

B.

D.

Fig. 17. — A Group of Endemic Hawaiian Insects. All belong to Large
Endemic Genera (except the Odynerus) .

A. Plagithmysus blackburni Sharp (Cerambycidae). B. Omiodes anastrepta Meyr.
(Pyralididae). C. Odynerus nigripennis Holmgr. (Vespidae). D. Anomalo-
chrysa blackburni Perk. (Chrysopidae). E. Megalagrion blackburni Macl.
(Agrionidae). Photo W. H. T. Tarns.

ISOLATION 137

Mollusca. No endemics x on the Scilly Isles (Richards
and Robson, 1926). Probably no endemics on the
Hebridean Islands (Robson, MS.). This may be
contrasted with the high degree of endemism in the
mammals.

A somewhat similar phenomenon is the capricious occur-
rence of endemism in archipelagoes. We have already given
a few examples (e.g. mouse-deer, p. 116). Simpson (1929),
in his study of the species of Liguus (land snails) on the Florida
Keys, finds that they are broken up into numerous varieties,
but that there is no regular localisation on particular keys
(contrast with ' ridge ' forms of Partula (Crampton) ). A
given variety may occur on several keys, and a given key
may have only one or else several varieties. There appears
to be no obvious correlation between topographical isolation
and varietal differentiation.

Similarly Riley (1929) finds that the birds of the Sumatran
Islands are on the whole more differentiated on the remote
islands than on the less remote. But this is not invariable, and
in the W. Sumatran Islands the relation between differentiation
and spatial separation is not nearly so obvious (cf. Robson,
1928, p. 139 (Hebridean mammals) ; also Aubertin, Ellis and
Robson, 1 93 1 (colonies of Cochlicella acuta) ).

We are led, therefore, to inquire as to the circumstances
in which some species change or remain stable ; and, secondly,
as to whether numerous smaller factors tending to produce isola-
tion on a small scale are not just as important as the high degree
of isolation produced by marked geographical separation.

The relative stability of some species and the high degree
of variability in others provide one of the most curious and
baffling problems in biology (cf p. 106, Chapter IV). It
is remarkable to what an extent certain species of a genus
may vary, when others are quite constant. The same differ-
ences are found in the frequency with which geographical
races are formed. It might be supposed that such differences
in variability depended on whether a species was exposed to
constant and homogeneous or varying and heterogeneous
conditions. But in fact all who have analysed such cases
agree that no such detailed relation can be found. With one

1 i.e. definite subspecies.

- fV 4

I Partula mooreana

Partula exigua

Partula mirabilis

*S&

->

Partula aurantia

■ '■'""

Partula dendroica

Partula olympia

Partula tohiveana © Partula solitaria

Fig. i 8. — Distribution of the Species of Partula on the Island of Moorea.

(From Crampton, 1932.)

ISOLATION . 139

exception (see below) there appears to be little really con-
vincing evidence that differences in rate of evolution are
determined by the environment. In this matter, however,
one positive example is probably worth several negative ones.
The exception referred to above is provided by island
races. We have already noted that endemism, though not
uniformly developed, is considerable. Not only are endemics
numerous, but they are sometimes of a peculiar type. Rensch
(1928, p. 174) has already noted that on small islands there
is a ' Neigung zu Excessiv-Bildungen in Grosse, Form und
Farbe.' We may note particularly :

Dwarfing. Birds. (Rensch, I.e. pp. 174-5 ; Dwight, 1918,
p. 269.)
Tiger. (Pocock, 1929, p. 505.)
Mollusca. (Sturany, 1916, p. 137.)
Lizards. (Kammerer, 1926, p. 88.)
Giant forms. Lizards. (Kammerer, I.e.)
Mollusca. (Rensch, I.e.)

(Not observed by other describers of
insular variation, e.g. Bristowe, Lundblad,
etc.)
Melanic forms. Reptiles. (Kammerer, I.e. ; Mertens, 1931,

p. 205.)
Spiders. (Bristowe, 1929a, p. 164.)
Hydracarina. (Lundblad, 1930, p. 24.)
Mollusca. (Pelseneer, 1920, p. 561 ;
Aubertin, Ellis and Robson, 1931, p.
1049 5 Kammerer, I.e.)
Mammals. (Kammerer, I.e.)

These rather striking consequences of life on islands require
further investigation.

Another factor, viz. the numerical abundance of the species,
has been supposed (Darwin, 1884, pp. 42-3 ; Fisher and Ford,
1928 ; Ford, 1931, p. 100) to be important. Abundant
species are or tend to be more variable. A good example of
this is given by Fisher and Ford (I.e.) in the species of British
Noctuid moths. Greater variability will on the whole mean
quicker evolution. According to this idea evolution will
proceed by the fission of a few common, widespread variable
species, while the rarer, less variable species will become

1 4 o THE VARIATION OF ANIMALS IN NATURE

extinct, and will not contribute to the evolution of the group.
It seems doubtful whether this principle is very helpful, except
in comparing fairly similar forms, and it can scarcely explain
the anomalies of differentiation in archipelagoes, etc.

Apparently much more importance must be ascribed to
innate differences in species, which we have to allow for but
cannot at present explain. When once we admit that some
species may have an innate tendency to unusual variability, we
make it very difficult to study the effects of isolation. A high
degree of innate variability will increase the chance that any
isolated parts of a population will have a composition differing
from the norm of the species. If permanent isolation depends
on the cumulative effect of various small accidental dishar-
monies, then geographically isolated populations of a variable
species may be expected to reach a state of permanent isolation
more quickly.

Later in this chapter there is a discussion of whether
permanent isolation is most often gained by the accumulation
of numerous small differences rather than by one substantial
change. It can be shown that relatively slight differences
sometimes maintain a significant degree of isolation, and it is
much easier to imagine the evolution of the isolatory mechanism
by several small steps than by one big step. In larger animals,
on the other hand, geographical isolation may be very im-
portant, but, as body-size is reduced, it becomes progressively
less significant. This is probably a natural result of large
animals x wandering over extensive areas, which often include
numerous types of habitat, while smaller animals can maintain
themselves in a population of efficient size, within the much
smaller limits of perhaps a single restricted habitat.

Our argument, then, runs as follows : in large animals
geographical isolation is probably an important factor, though
the degree to which the inherent variability of the species is
developed is no less important. Unless a population changes
enough to become permanently isolated, it will be liable to be
recombined with the parent stock by subsequent topographical
changes. We do not know how long it takes to evolve per-
manent isolation, but, at least in some species, evolution is

1 von Schweppenburg (1924, p. 143) says he knows of no clear case in birds
in which subspecies are in the least likely to have arisen in one place. Where
there is considerable overlap it is likely to have arisen by spread since the races
originated.

ISOLATION 141

slow enough to allow considerable land-changes to occur
during the establishment of a race. In small animals geo-
graphical isolation becomes, on the whole, less important, for
even on continuous areas there are numerous ways in which
populations can be isolated from one another. While it might
be thought that none of these ways was sufficiently absolute
to allow permanent isolation to set in, the recent studies of
biological races point to another view.

Methods of permanent isolation.

These have been analysed by Robson (1928, pp. 122-33).
We recognise a primary division into two chief methods, each
of which may be subdivided.

I. Indirect methods :

(a) Seasonal occurrence.

(b) Time of breeding.

(c) General habitat.

(d) Differences in breeding habitats.

(e) Loss of means of dispersal.

II. Direct methods :

(1) Prevention of copulation.

(a) Psychological or physiological.

(i) Differences in specific recognition marks.
(ii) Differences in epigamic characters (scents,
courtship behaviour, secondary sexual
ornaments) .

(b) Mechanical.

(hi) By differences in the mechanical relations
of the copulatory apparatus.

(2) Prevention of effective fertilisation.

(c) By failure of the sperm to reach the egg.

(d) By disharmonies in development, including and

leading up to sterility in hybrids.

In sedentary animals and aquatic animals with externa
fertilisation only I and II (2) can be effective. In motile
animals with internal fertilisation II (1) may also operate.
In this respect plants are in the position of sedentary animals.

142 THE VARIATION OF ANIMALS IN NATURE

I (a) and (b) . Seasonal occurrence and breeding season.

In short-lived animals the breeding season of a species is
usually almost coextensive with the seasonal occurrence.
With longer-lived animals a definite season tends to be set
aside for breeding. In either case, one of the simplest ways
in which varying degrees of isolation may be brought about
is by specific differentiation of this season. Besides being
simple, separation in this way appears to be important because
the seasonal occurrence or breeding period is likely to express
the summed effect of the reaction of the organism to its environ-
ment. The various small characters by which we separate
species can be regarded as the visible expression of differences
in growth-rates and in various physiological processes. The
species must develop in a different way, and the length of
the period necessary to complete the life-cycle is one of the
most obvious ways in which developmental differences may be
expressed. Where the species live in more or less separate
habitats even greater disparities might be expected. It is,
indeed, surprising that species are not more often separated by
differences in breeding season, but it may be supposed that
the fluctuations of the environment make it difficult for any
species to have a sharply defined breeding season, and further
that the rhythm is much modified to fit in with other periodic
features in the environment, particularly the food-supply.
The latter factor becomes more important as species diverge
more and more widely from one another. Where an insect,
e.g., depends on one or a few species of plants there is often a
very close correlation between their life-cycles.

Specific differences in breeding season or seasonal occurrence
are extremely common in insects and are not rare in other
groups, though complete isolation by this means is probably
rather rare. We can only mention here a few typical examples.
One of the most striking instances is seen in the Seventeen-year
Cicada (Tibicen septemdecim) of the United States (Marlatt,
1907). To begin with there are two races, the 17-year race
(mainly northern) and the 13-year race (mainly southern).
The number of years refers to the time spent as a subterranean
nymph. These two main races scarcely differ in structure,
but do not appear to interbreed where they meet. Almost
every year a brood of each race emerges in some part of the

ISOLATION 143

range of the species, but some broods are very localised. Other
broods are discontinuous, but the 17- or 13-year period of the
broods in any one locality has been well established over the
last 200 years. Occasionally some individuals come out a
few years late or early, and it is probably by this process of
retardation or acceleration that the different broods originally
became established. This accounts for the discontinuous
distribution of some broods and also for the general rule that
broods adjacent to one another in space are also adjacent in
time.

von Schweppenburg (1924, p. 151) notes that Lasiocampa
quercus and L. quercus callunae scarcely interbreed because
their times of emergence are different (May and June in
callunae and July and August in quercus). There is also a
more or less marked difference in larval food and in habitat,
but the moths are structurally almost identical. Tutt (1910)
states that the only known barrier between the butterflies
Agriades thetis and A. coridon is that the single brood of A. coridon
falls between the two broods of A. thetis. Dietze (191 3, pp. 134-
136) gives an interesting account of the relation between the
moths Eupithecia innotata and E. unedonata. The larvae feed
on Artemisia and Arbutus respectively and the moths have a
non-overlapping seasonal occurrence, unedonata appearing
much earlier. By cooling the pupae of unedonata he was able
to obtain a late emergence, and the resulting moths paired
freely with the production of fertile hybrids. Lackschewitz
(1930) has recently revised the crane-flies of the oleracea group
of Tipula. The seven species now recognised were all ' lumped '
together until recently, and even now are distinguishable
mainly by minute differences in the male genitalia. The
females are mostly still inseparable. Of the three species occur-
ring in Western Europe, T. oleracea has two broods — one in the
summer and one in the autumn. T. paludosa has one brood
between July and September, while T. czizeki occurs only in
mid-September and October. The Morrisons (T. A. and L.)
(1925) have shown that there is in addition a preferential
mating reaction between T. oleracea and T. paludosa. Peacock
(1923) records a difference in seasonal occurrence between
the very closely allied sawflies Thrinax mixta and T. macula.
The former emerged between April 29 and May 8, the latter
between May 8 and May 17. The species are exceedingly

i 4 4 THE VARIATION OF ANIMALS IN NATURE

alike both as larvae and adults, and the food-plants are
identical.

Differences of this type seem to be fairly common in
phytophagous insects, but there is usually some overlap between
the seasons. Where the female of a species is always impreg-
nated immediately after emergence and the male emerges
before the female, very small differences in the total period of
occurrence may have considerable effect.

In other cases seasonal occurrence appears to play no
part in isolation. Thus Schubert (1929), in his account of
the dragon-flies of the neighbourhood of Neustadt, records
that all the 18 species (6 genera) have overlapping periods,
with the possible exception of the two species of Orthetrum.
Richards (1930, p. 321), in his account of the British flies of
the family Sphaeroceridae, shows that most of the species occur
throughout the year, and many of them seem to have no
restricted breeding season.

Isolation by means of differences in seasonal occurrence
has a special interest because of its relation to the environment.
It is a general rule for insects to have more broods in the south
than in the north and, although partial broods, in which only
a few individuals of a given generation emerge, are often
found, there is a natural tendency for a species to fix on a
definite reproductive rhythm. The intermediate state, where
partial broods are formed, would appear to be one of unstable
equilibrium. A species which is single-brooded in the north
will be double-brooded in the south and, if the range is suffi-
ciently great, even more broods may develop still further south
— e.g. Agrotis segetum (Filipjev, 1929), Pyrausta nubilalis Hb.
(Babcock, 1927).

Owing to climatic conditions there will be a tendency for
the single-brooded form to occur between the broods of the
bivoltine form in time. If we knew more as to how such
rhythms become fixed, we might see a way in which the two
forms could remain permanently isolated, even if their ranges
came to overlap. This subject has been ably reviewed by
Uvarov (1931, pp. 104 ff), who concludes that rhythms
originally induced by climatic conditions are eventually
hereditarily fixed. Pictet (1913), experimenting on Lasiocampa
quercus, obtained results suggestive of such a process. (See
also Chapter II.)

ISOLATION 145

In longer-lived animals with a definite breeding season a
comparable state of affairs exists, but isolation appears to be
much more partial, except sometimes between races of one
species, e.g. Rana esculenta (Cuenot, 1921), Sepia (Cuenot, 1917)5
Crangon and Orchestia (Plate, 191 3). In addition long-lived
animals appear to be largely those which also evolve mainly
through geographical races, in which the breeding period is not
likely to be an important factor in isolation. When the races
have evolved so far that their ranges overlap, and we find two
species living side by side, other factors often override any
original differences in the breeding period. It may be sus-
pected that any environmental pressure tending to reassimilate
two rhythms would sooner or later be effective and, if the two
forms were not by that time intersterile or isolated in other
ways, they would be reunited. We might, therefore, expect
that differences in seasonal occurrence in the breeding season
would usually be found as specific only between forms still
quite closely allied.

I [c] . General habitat.

It must be very rarely that two closely allied species have
so sharply different habitats that no crossing could occur. In
a country like England, where no one habitat covers an exten-
sive area without interruption, this is obvious, but in some
continental areas habitat-differences may be much more
important, though no clear distinction can be drawn in this
case between restriction to one habitat and to one geographical
area. Even on a much smaller scale, however, habitat-
differences will lead to some degree of selective mating,
especially with forms with low powers of dispersal. This small
contribution towards the establishment of isolation is important
because some degree of differentiation in habitat preference
must be regarded as one of the commonest of specific characters.
As the general facts are well known to most zoologists, we will
give a few instances, confining our attention mainly to pairs
of closely allied forms.

von Lengerken (191 7) and Macgillavry (1927) record that
the tiger-beetle Cicindela hybrida L. is restricted to the part of
sand-dunes which is fixed by vegetation. The subspecies (or
species) C. maritima Latr. occurs only on stony places on the
actual strand. In Holland, however, a darker race of maritima

146 THE VARIATION OF ANIMALS IN NATURE

occurs on alluvium inland, where it is associated with C.
campestris. The two moths Lasiocampa quercus and L. quercus
callunae, already noted as differing in emergence period, also
differ in habitat, the former being a lowland species, the latter
inhabiting moors and mountains. The two habitats in this
case are subject to considerable overlap. Fulton (1925) has
described two races of the common N. American cricket,
Oecanthus niveus. The races differ in song and habits of ovi-
position and also in habitat, one living on trees, the other on
bushes. Myers (1929, p. 50) records that the various New
Zealand species of Cicada are strictly confined to different
plant-associations. It is probably not usual, however, in
England for a species to be strictly confined to a plant-associa-
tion. Many species have a single food-plant, but few plants
are rigidly confined to a single association. Again, allied insect
species not rarely feed on the same plant, e.g. many Chryso-
melid beetles and weevils. It is not easy to find numerous
genera in which both taxonomic and ecological studies are so
advanced that we can say with certainty which species are
closely allied and what range of habitat is occupied. It is
certainly quite impossible to give a numerical estimate of the
frequency with which allied forms are found together or in
different habitats. We only know that both conditions may
be encountered.

Amongst the vertebrates, closely allied forms tend to be
geographically isolated, so that this method of separation can
hardly arise. Amongst more distinct species, of course,
habitat-differences are common, but are probably not very
important in preventing interbreeding.

I (d) . Differences in breeding habitats : minor geographical units.

Differences of this type are best known in forms with a
definite breeding season. In migratory birds, for instance,
there is a well-known tendency for individuals to return to
breed in the locality where they were reared, and this tendency
makes possible the formation of geographical races, since races
which may mix in the winter, sort out and return to their own
areas in the spring. Though this phenomenon is largely part
of geographical differentiation, it must also lead to the forma-
tion of smaller units. Thus Schmidt (1931) has shown that
species of eel which breed in a single restricted area are

ISOLATION 147

relatively uniform and are not separable into subspecific units,
while other species which breed over a large area are much less
homogeneous, being formed of a number of separate strains.
From the point of view of isolation, it is difficult to distinguish
between the action of geographical barriers and of differences
in migratory instincts.

I (e) . Loss of means of dispersal.

The examples cited in the previous paragraph lead us
naturally to consider animals in which the power to migrate
has been lost. Our ignorance of this matter is much greater
than would appear at first sight. The high percentage of
endemism on islands is well known, as is the tendency for
island forms of winged species to be apterous. Evidently, if
the species had not been winged originally their chances of
reaching an oceanic island would have been small. Once an
island has been reached, loss of powers of dispersal will aid the
formation of local colonies, though it will not aid in isolation
from fresh immigrants. We need not consider at this point
the theories that have been put forward to account for the
winglessness of island species, but, however produced, apterism
will tend to multiply the numbers of endemic species on an
island. At the same time very numerous examples of complete
or nearly complete loss of wings are known from continental
areas. In the beetles these facts have been summarised by
Jackson (1928), and for Diptera by Bezzi (1916, 1922). The
former author, working on the weevils of the genus Sitona,
found short- winged, long-winged, and dimorphic species.
Wherever the power of flight has been lost we might expect
some degree of isolation to arise between colonies that pre-
viously were able to interbreed, if only because the ordinary
habitat of the animal is not likely to be continuous. But we
have to be very careful not to assume that the species with
apparently the best means of dispersal are necessarily the most
active species in getting about. Thus Richards (1926) points
out that the wingless beetle Helops striatus is one of the first
insects to re-invade heaths after fires. The wide range of
many other wingless forms suggests that detailed knowledge of
actual methods and powers of dispersal is necessary before we
can assume very much about their significance in isolation.
The loss of eyes in cave insects is a parallel phenomenon.

148 THE VARIATION OF ANIMALS IN NATURE

Jeannel (191 1 and 1926) has shown how cave species tend to
be confined to one or a few adjacent caves. Doubtless blind-
ness is not the only agency confining a species to its own cave
(for, just as some apterous species are widespread, some
blind species also occur in the open), but it probably plays
an important role. This subject is discussed elsewhere
(Chapter VII, p. 269).

The truth is that we know extremely little about the
powers of dispersal of animals, apart from the more sensational
migrations, and it is possible that some of the anomalous
differences in the variability of different species might become
clearer if we knew more. It is especially difficult to trace the
minor wanderings which occur within the normal area inhabited
by the species. The frequency of such wanderings must largely
determine the homogeneity (or the reverse) of the species, and
this, in turn, has an important effect on the significance of
geographical isolation, since any isolated population has a
greater or less chance of differing from the norm of the
species. The converse, however, is equally important, viz.
that no true random mating can occur in any species, because
the chance of an encounter between individuals separated by
a few miles of country is relatively low. Even in long-lived
and wide-ranging forms this must have some effect, and in
small, short-lived species, unless the specific range is extremely
small, the results must be very significant.

II (i). Recognition marks.

We have very little information as to the function of
recognition marks (or odours) amongst animals, apart from
structures (or odours) specifically connected with mating.
Something of the sort is evidently present in most gregarious
animals. Thus Ward (1904) has shown that the bats in
certain caves in Mexico roost according to their species.
Feuerborn (1922) has given some evidence which suggests
that flies of the family Psychodidae recognise the species as
well as the sex of other individuals. He suggests that certain
glands, present in both sexes, produce odours on which this
faculty depends. Seitz (1894) long ago suggested that some
Lepidoptera may produce both specific and sexual odours.
Colour must also play a part in species recognition, as Eltring-
ham (191 9) has shown in certain butterflies in which the sexes

ISOLATION 149

are alike. Again, many insects, just prior to» mating, form
swarms of one sex only : the attraction here cannot be strictly
sexual, although it is a preliminary to mating. While we know
little in detail as to the influence of recognition marks, we
can see that any tendency to form aggregations will lead to
some degree of isolation. We cannot yet say whether such
recognition marks often come to differ in the early stages of
species evolution. In many animals with more than one
colour-form the various types all interbreed (cf. Elton, 1927,
p. 182 et seq. ; Richards, 1927). Probably recognition marks
grade insensibly over into what must be classed as epigamic
characters, but the former would include stimuli not acting
only during a brief period before mating.

II (ii). Differences in epigamic characters.

The enormous mass of data concerning the epigamic
characters of animals is not very helpful from the present
point of view. An examination of the literature shows that
the greatest number of papers describe the morphology of
epigamic structures ; a less number describe the mating
behaviour, including the use of such structures ; still fewer
provide any evidence as to the significance in isolation of
specific differences in epigamic structure and behaviour.

It is well established that species do very often differ in
secondary sexual characters. In only a small fraction of the
total number of species have these differences been shown to
have a significance in mating. In many cases (e.g. Saturniid
moths (Mayer, 1900) ) the characters are probably only indi-
cators of important differences in metabolism. In other cases
the female may have been modified in connection with her
maternal duties (development of brood-pouches, etc.).

Where the sexual characters are known to play a part in
courtship their exact significance is nearly always doubtful.
There is not enough experimental work to prove that particular
structures or types of behaviour are actually essential if the
male is to be successful. Usually the most that is known is
that some conspicuous structure is exhibited in a provocative
way during courtship. We may give a few examples in which
the significance of epigamic structures or behaviour is fairly
certain.

Sturtevant (191 5) has shown that the wing- waving of male

150 THE VARIATION OF ANIMALS IN NATURE

Drosophila has a significant effect in reducing the time taken
by the male to succeed in copulation. Males with their wings
removed are able to mate sooner or later, but in them the
pre-mating period is longer. In the Lepidoptera the experi-
ments of Fabre, Mayer and Freiling have shown that the scent-
apparatus of the female is frequently (probably nearly always)
an essential element in pairing. The males are normally
attracted to the scent of their own female, who distributes it
until pairing has been effected. In some fireflies, different
species of a genus emit light of different colours or in flashes of
different frequency. Where both sexes are luminescent, each
sex may respond only to the signal of its mate (Coblentz,
191 1 ; Macdermott, 1910, 191 1, 19120, 1912^). In the
Mollusca, Diver (quoted by Robson, 1928, p. 126) has shown
that the two common English banded snails (Cepea hortensis
and C. nemoralis) differ in the energy with which mating
individuals stimulate one another with their darts. This
difference, which appears to have no connection with the
actual structure of the dart (which is also specific), is normally
sufficient to keep the species apart if they attempted to
pair.

Standfuss (1896) was able to show that the females of the
Italian subspecies persona of Callimorpha dominula (Lepidoptera)
are scarcely attractive to the males of the normal form.
Grosvenor also (1921) has found local variations in the
attractiveness of the female in ^ygaena (Lepidoptera).

In the Orthoptera, where sound-production plays an
important part in courtship, Fulton (1925) has shown that
two biological races of the tree-cricket, Oecanthus niveus, differ
in their song. Faber (1928), however, in his study of the
German Orthoptera, found that by no means all species could
be separated by their song, which, further, was very variable
owing to the influence of temperature and the rivalry of other
males. Whether a species responds only to the song of its
own kind appears still to require much more confirmation.

In spiders, Bristowe and Locket (1926) show that courtship
antics and male decorations may have a real value as recog-
nition marks. It appears that unless the female recognises the
male as belonging to her species she will often eat him, and the
peculiar dances of the males assist the females to avoid mistakes.
Tactile stimuli may play a similar part in families where sight

ISOLATION 151

is little developed, and it is probable that the dances are also
stimulatory in their effect. The female behaviour has two
phases, an amatory and an aggressive one, and when the
former holds sway she is much less likely to attack the male.
Thus courtship dances, besides giving the female a chance to
recognise her mate, also put the female into a state in which
attack is unlikely. After copulation, when the aggressive
phase reasserts itself, she may devour the male, though she can
scarcely be said not to recognise him.

Against these examples we may set others in which the
epigamic characters are not yet known to play a part in
isolating species. Among birds, as Huxley (1923) and others
have pointed out, the exhibition of coloured parts and the
performance of special antics, flights and songs take place
usually after the birds are already mated up for the season.
The displays are supposed to have a purely stimulatory effect.
It is possible that male epigamic characters may play a minor
part as recognition marks, though on this point we have no
evidence. The stimulatory function seems likely to be impor-
tant in many groups. Species which hybridise naturally also
provide important evidence, since they show that no single
element in the isolationary complex is necessarily and always
competent to produce its normal effect.

II (iii). Differences in the mechanical relations of the copulatory
apparatus.

We may, in the first place, mention a rather exceptional
example amongst the fish. In the genus Anableps (Nor-
man, 1 93 1, p. 296) the male genital orifice is prolonged into
a tube. The genital aperture of the female is covered by a
special scale, free on one side only. The opening may be on
either the right or the left and the males may have the intro-
mittent organ turned in either direction. Copulation takes
place sideways and a right-sided male always pairs with a
left-sided female and vice versa.

The whole problem becomes much more complex when we
consider the more usual type of specific differences in the
genitalia, which are so often found, especially in the males,
in a number of groups (see p. 296). It is important to dis-
cover how far these elaborate structures act as a mechanical
means of isolating allied species. When we find that the male

152 THE VARIATION OF ANIMALS IN NATURE

genitalia (as often happens in insects) differ sharply in charac-
ters whose degree of variation is not enough to make them
overlap, in a species in which most or even all other structures
intergrade from species to species, it is tempting to assume that
we see here the actual agency for permanent isolation in these
forms. The essence of this theory, the well-known ' lock-and-
key ' theory of L. Dufour (cf. Perez, 1894), is that the females
should also differ in some way from each other ; differentia-
tion in the male alone would not be effective. Whether the
females do differ and whether the male armature really is
effective in isolation have for many years been matters of
controversy. The argument has chiefly lain amongst the
entomologists, and a decision for the insects would probably
also be valid in the case of many parasitic worms, Crustacea
and Arachnida.

The chief supporter of the ' lock-and-key ' theory has been
Jordan (1896, 1905). Boulange (1924) has reviewed the
subject and takes the opposite view. Jordan, in his first
paper, dealing with the swallow-tail butterflies {Papilio), showed
that the differences in the male genitalia are quite manifest,
sometimes in geographical races. The females sometimes
differ markedly in their genitalia, though they were much less
thoroughly investigated. The actual proof that the male
structures coincided so accurately with those of the female that
copulation between different species would be difficult or
impossible was not very convincing, and the evidence put
forward was derived from a few species only. In his second
paper the correlation between differences in genitalia and in
other characters is examined. His main thesis is that local
and seasonal * polymorphism in colour and wing-shape is
quite independent of variation in male genitalia. In one geo-
graphical area the genitalia vary only slightly and at random,
but as soon as a distinct geographical race becomes recognisable
the variation in the genitalia tends to be correlated with the
size and colour characters defining the race. It may be
admitted that the male genitalia are easily modified in the
evolution of species, but it is much more uncertain what part
they actually play in that process.

In the Lepidoptera as a whole interspecific crosses are not

1 Mercier (1929) claims to have demonstrated seasonal variation of the male
genital organs in the fly, Cynomyia. Jordan also records one case in Papilio.

ISOLATION 153

very rare and there is little evidence that differences in the
male genitalia are often a very serious barrier between species,
except when the structures are extremely different, as between
species belonging to different genera or families. In a number of
species of Diptera the male genitalia are extremely diverse, and
there appear to be no corresponding differences in the female ;
sometimes (Lucilia) it is only with great difficulty that the
females can be distinguished, if at all. In many Hymenoptera
the male genitalia differ greatly, with little or no differentia-
tion in the females (Richards, 1927a, p. 262 ; Boulange, I.e.).
In Bombus, where the female genitalia do to some extent vary
specifically, it is largely groups of species which differ and the
structures showing differences come into contact only with
part (and that not the most complicated) of the male armature.
Further, in some species the male genitalia, though nearly
identical, show certain minute but constant differences, too
small to have, with any probability, any functional significance
(Richards, I.e.). Although there seems to be usually no
detailed co-adaptation between the male and female, there are
some exceptions. Edwards (1929, p. 40) records a correlation
between the length of the male penis and of the female sper-
mathecal duct in the flies of the family Blepharoceridae. A
similar correlation is sometimes observed in beetles of the
family Chrysomelidae (Harnisch, 19 15), but how far this is
specific rather than generic requires investigation.

A more rational explanation would appear to be that
differences in instinct — possibly {e.g. in insects) in the nature
of the scent produced — are the first stage in the permanent
isolation of species ; later, differences in genitalia may arise
and may sometimes, incidentally, make the isolation more
perfect. In this way it is possible to explain the occurrence
of groups (families, genera, etc.) in which the genitalia are
scarcely specifically differentiated. All who have studied
insect genitalia agree that the value of these structures to the
taxonomist varies greatly in different families, in some pro-
viding characters of little more than generic value, in others
differing very greatly in species otherwise very similar.

In the preceding paragraph we have advisedly used the
phrase ' permanent isolation ' to describe the result of changes
in instinct, for temporary isolation may result from geographi-
cal barriers. It is a matter of controversy whether some

i 5 4 THE VARIATION OF ANIMALS IN NATURE

measure of geographical isolation is necessary for divergence
to begin. This view has been strongly maintained by Jordan
(1896, 1905), and is implicit in the ' Formenkreis ' theory of
Kleinschmidt and Rensch, according to whom geographical
races alone are the starting-point for new species.

In the case of birds and mammals there would appear to
be good evidence for this idea. The lowest systematic cate-
gories (geographical races) never occur together except in a
minimal part of their range (cf. von Schweppenburg, 1924,
p. 143) and, generally speaking, only rather widely divergent
forms live together in the same habitat. It is true that the
geographical barriers between the races are not always abso-
lute, but imperfect barriers combined with the usually dis-
continuous occurrence of suitable habitats may be sufficient
to allow divergence. The chief lack at the moment is the
accurate study of the distribution and nature of the forms
occurring where two races meet.

With insects the necessity of geographical isolation is
much more difficult to maintain, as might be expected from
the relative complexity of the way in which the sexes are
normally brought together. If selected cases are examined
(cf Jordan, 1896), it is easy to show the importance of geo-
graphical isolation, which in any case must always be operative,
even if it is not the only agency responsible for divergence.
Thus Jordan found in certain Oriental swallow-tail butterflies
that forms differing in colour, shape of wings or seasonal
occurrence never differ in genitalia unless they are restricted
to geographically separated areas. Since Jordan maintains
that mechanical isolation as a result of differences in the
genitalia is the chief means of making divergence permanent,
he argues that in these swallow-tails it is only the geographical
races and not variants which occur together in one locality
which will (or may) give rise to new species. It is possible,
however, in other groups to find examples which suggest the
opposite point of view. Thus species or races with genitalia
so similar as to differ from one another no more than do the
geographical races of swallow-tails, may occur together over
wide areas, as in the butterfly Satyrus huebneri (Avinoff, 1929),
in many Tortricids (compare male genitalia of species of the
genera Cnephasia or Epiblema, Plates v and xxiii (and p. 68)
in Pierce and Metcalfe, 1922) or in some Hesperiids (Warren,

ISOLATION 155

1926, p. 40). While it is possible to assume that these forms
evolved in geographical isolation but have since crossed their
barriers, it is doubtful whether the evidence for the necessity
of geographical isolation is so cogent that it is necessary to
make so big an assumption.

We may summarise the outstanding points in this con-
troversy as follows :

1. The male armature differs specifically much more

often and usually more markedly than the female.

2. There is often, perhaps usually, no close specific corre-

lation between the male and female structures. At
least such correlation has not been established.

3. The numerous interspecific crosses, mostly artificial but

some natural, between species with very different geni-
talia, show that the male and female armatures by no
means necessarily impose an insuperable barrier.

4. The vast mass of species with different genitalia prob-

ably do not try to interbreed. They are in fact
separated by other types or combinations of types
of isolatory factors (especially those included under
I and II (a) ).

5. As a corollary to (4), large groups of species exist in

which the female genitalia differ but little from
species to species. There is no evidence that such
forms hybridise more readily than those in which
the differences are marked.

6. There appears to be no very high correlation between

degree of differences in genitalia and the fertility of
hybrids if a pairing does take place — e.g. Sturtevant
(1920) shows that Drosophila simulans and D. melano-
gaster have identical mating habits and hybridise
freely, but the hybrids are quite sterile. The male
genitalia differ, but not those of the females.

We can only conclude that the genital armature may
sometimes provide a bar to interspecific crosses, but the bar is
by no means universal or incapable of being surmounted.
This is particularly true of the smaller differences which
characterise very closely allied species. The value of specific
differences in the genitalia lies rather in their relative
constancy. Thus, while variation does occur {e.g. marked

156 THE VARIATION OF ANIMALS IN NATURE

variation in the Magpie Moth, Abraxas grossulariata, recorded
by Kosminsky, 191 2), it is not usually of a type to make
species overlap.

If small differences in the genitalia are not in themselves
enough to isolate species, it becomes a matter of importance
to decide whether the degree of difference commonly found
between species is likely to have been built up in several stages.
One series of observations made by Foot and Strobell (1914)
suggests that the specific differences must be due to the action
of several independent hereditary factors. In crossing two
bugs of the genus Euschistus they found that the length of the
penis (a specific character) was intermediate in F x and only
rarely reached either parental type in F 2 . This suggests that
more than one factor for penis-length is involved, and we
may suspect that the first stages in this divergence cannot
have been very important as means of isolation.

Apart from the genital armature, difference in size in itself
might be expected to play some part in isolation. This would
be more important if really closely allied species did more
commonly differ markedly in size. We have little very definite
information on this subject. Mickel (1924) has shown that
the Mutillid wasp Dasymutilla bioculata Cress, has a bimodal
variation in size owing to its having two main hosts. A male,
however, could mate with a female which was only half his
size, so that there was not much chance of the size difference
leading to isolation. In insects generally size-variation does
not appear to be very important. In a sea slug, however,
Crozier (191 8) has shown that mating individuals tend to be
of about the same size. But even in the molluscs this is not
universal (Robson, 1928).

II (2). Prevention of effective fertilisation.

Some degree of sterility on crossing is well known to be
a common type of difference between species. The term
' sterility ' is in fact employed to describe a number of dis-
tinct phenomena. Only exceptionally do we know exactly
what occurs in a particular case. After an apparently
effective pairing, we may distinguish between the following
possibilities :

1 . The sperm fails to reach the egg.

2. The egg is fertilised, but development ceases at an

early stage.

ISOLATION 157

3. Development proceeds further, and a feeble or mal-

formed F x may be produced.

4. Well-developed, more or less vigorous hybrids are

produced which are sexually abnormal — e.g. one sex
missing from brood, spermato- and ovo-genesis
abnormal, production of intersexes, etc.

5. F x more or less fertile — e.g. fertile with one sex of one

of the parent species.

6. F x fertile, but F 2 infertile or weakly.

7. Complete fertility.

In noting this wide range of possibilities, it is important
to remember that some degree of intraspecific sterility is
always met with. Sometimes sterility between certain types of
individuals is very marked — e.g. in some Ascidians (cf. Plough,
1930, 1932). The nearly fertile interspecific hybrids, there-
fore, grade over completely into species in which intraspecific
sterility is normally present in some degree.

The essential question is whether any of these forms of
sterility provides commonly the first important stage in isola-
tion. At least one case is known of extreme sterility between
a species and a mutant differing only in a single character,
viz. Drosophila obscura (Lancefield, 1929), in which a naturally
occurring race, with a very large Y-chromosome, will not cross
with the normal form. It is difficult to see how a mutant
determining sterility could establish itself in the population ;
the process is not likely to be very common. On the whole,
however, the facts do not suggest that sterility is commonly
the initial method by which isolation is established ; at any rate
it is unlikely to have been the only important factor. This
subject has been discussed at greater length by Robson (1928).

Conclusions

This survey of the factors which promote isolation suggests
the following generalisations :

(1) There is no one predominantly important way in

which isolation becomes established in the early
stages of species-formation.

(2) Geographical or temporary isolation is undoubtedly

very important, but it cannot be claimed that this

158 THE VARIATION OF ANIMALS IN NATURE

is the only way in which new species arise. The
permanent isolation of geographical races must be
established in much the same way as permanent
isolation between species inhabiting the same area.
(3) The establishment of isolation is probably due to the
interaction of a number of different factors, none
of which would be effective by itself.

The third generalisation is the one which appears to be
most useful. The study of geographical races is not likely to
be helpful, except in the narrow zone where two races meet,
and not here if, as often happens, the races interbreed at this
point. It is rather in the study of biological races of animals
that our hope lies. These closely allied taxonomic groups,
differing more in habits than in structure, show us where the
fission of species is just beginning. Since the races often occur
together without much intercrossing, isolation must have been
developed and may be analysed with some likelihood of
reaching definite conclusions (cf. Thorpe, 1929, 1930). In
1896 Jordan was able to make out a case for the theory that
permanent isolation would be developed only between species
already geographically isolated. It seemed at that date
that a difference sufficient to isolate two forms could not arise
at one step without a new species also arising suddenly, and
this appeared to contradict the widely accepted generalisation
that specific change was gradual. The more recent study of
biological races demonstrates that these a priori arguments
are unsound. Whether it seems probable or not, biological
races more or less isolated from one another do appear to
arise from an originally homogeneous species.

The occurrence of local breakdowns in a normally effective
isolatory mechanism also suggests the complex nature of the
process. Delcourt's (1909) study of Notonecta shows that
species isolated in part of their range may interbreed in a small
area, von Schweppenburg (1924) records the same thing
in Passer domesticas and P. hispaniolensis, and Tutt (1909, 19 10)
in Agriades thetis and A. coridon. If we take an imaginary
example in which two species are separated almost completely
by the time of their breeding season, and if we suppose that
the onset of the breeding season partly depends on climate,
but that the two species do not react to climate in precisely

ISOLATION 159

the same way, then it is easy to see that at some point in their
range the breeding seasons might coincide. Or again, two
species with different habitat preferences might be brought
into close proximity in certain areas where only an intermediate
type of habitat was available. We are in great need of accurate
analyses of actual concrete examples.

The most important conclusion in relation to our more
general argument as to the course of evolution is that, in so
far as the isolation of species from one another depends on the
combined effect of several agencies, it is likely that the same
agencies produce some degree of isolation between populations
within the species. The likelihood that species are much
broken up into populations which are to a considerable extent
isolated from one another must be fully allowed for in any
theory as to the spread of variants.

CHAPTER VI

CORRELATION

In a previous chapter we have endeavoured to show that
throughout the animal kingdom there is a tendency for indi-
viduals to be capable of arrangement in a hierarchy of groups,
each group being defined by an association of characters
which are more or less correlated together. It is evident that
whatever the cause or causes of evolution may be, one of its
most characteristic effects is the divergence of groups distin-
guished by blocks of characters which tend to hang together.
Much of this correlation is far from unexpected and calls for
little comment. It is not surprising, e.g., that a given mammal
of carnivorous habits should have teeth adapted for tearing
or crunching, a skull with suitable muscular attachments and
limbs appropriate to a raptorial habit. The regular association
of characters whose functional significance is far from apparent,
such as we see in species and subspecies, is quite another
matter and is the main theme of this chapter.

From a restricted point of view the origin of such correlation
appears a relatively simple problem, but a full treatment in-
volves the examination of some of the most difficult problems
in biology. It is easy to suggest how a group of characters
(each regarded as the expression of a single genetic factor)
could come to be correlated together, even if we cannot actually
verify our hypothesis in any concrete example. There is,
however, a tendency to treat the separate characters as some-
thing apart from the fundamental organisation of the living
animal (cf. Chapter IX). While this may be a justifiable
simplification for the practical purposes of genetics and
taxonomy, as we shall show at the end of this chapter, it
comes into conflict with another conception of the living
organism.

The term correlation has, since Darwin first made the

CORRELATION 161

phenomena an object of study, been applied to a variety of
relations which are not of the same nature. The credit of
distinguishing them seems to be due to Durken (1922). He
recognised three distinct types of association :

(1) Relation. — The 'unilateral' dependence of a structure
for its full expression on an internal factor on which the
structure in question itself has no effect (e.g. the dependence
in development on the optic capsule of the embryonic lens
in the Vertebrata).

(2) Correlation. — The reciprocal dependence of two asso-
ciated parts of such a nature that alteration of the one leads
to the alteration of the other (e.g. the reciprocal depend-
ence of the extremities and nervous system in vertebrate
development) .

(1) and (2) include all causal associations.

(3) Combination. — The ' static ' coincidence of variables
without any reciprocal or unilateral dependence (e.g. special-
isation of several parts for the same function ; dependence
of several structures or organs on sex hormones or on an
external stimulus (cf. Sumner, 1915) ).

Graham Kerr (1926) distinguished primary or gametic
correlation from secondary or physiological correlation. This
is a fundamental distinction of considerable practical value,
and forms the basis of our discussion. Robson (1928) discussed
the various kinds of correlation in so far as they are contributory
to the process of group divergence, pointing out some of the
difficulties that are encountered in explaining the origin of
groups by the current theories of evolution. In particular
he dealt with Pearson's contention (1903, p. 2) that Natural
Selection is probably the chief factor in causing correlation.
The fact that correlation may be fluctuating or stable according
to the degree in which the variables are affected by environ-
mental factors, was pointed out by Love and Leighty

(1914).

Darwin's views on the importance of correlation in relation
to selection and the data which he assembled are discussed in
Chapter VII. It is true that in the course of his examination
of a large series of cases of correlation he touched on the causes
of the phenomena — e.g. he discussed the correlation of variation
in homologous parts (1905, vol. ii, p. 389) and the effects of
selection (I.e.). He did not, however, take his discussion

M

1 62 THE VARIATION OF ANIMALS IN NATURE

on the causes very far, nor did he attempt to distinguish
the various phenomena to which the name ' correlation ' is
given.

The distinctions made by Diirken and Kerr can be harmo-
nised, if we realise that Diirken's ' relation ' and ' correlation '
are causal types of association and correspond to Kerr's ' physio-
logical correlation ' ; while Diirken's ' combination ' includes
Kerr's ' gametic ' correlation as well as other phenomena.
Thus we can include in it (a) character associations produced
by the mechanism of heredity in its distribution of segregating
characters (e.g. effects of linkage, strains homozygous for
several characters, etc.), and (b) equally fortuitous association
produced by the coincident effects of external causes operating
simultaneously on the individual.

It is desirable, before proceeding further, to obtain some
general idea as to the extent to which the characters distin-
guishing species and races, etc., are correlated. Were such a
measure obtainable, it would give us an idea as to the extent
to which these groups are homogeneous for their diagnostic
characters. Taxonomic experience, of course, prepares us
for the result that the degree of correlation is very varied,
probably on the whole rather low. The value of the available
data is rather dubious, as what we obviously want to know
about is the correlation of hereditary characters, and in sys-
tematic data little attention is paid to the discrimination of
fluctuational from hereditary characters.

A great deal of statistical information is available as to the
correlation of miscellaneous characters, but very little con-
cerning those which distinguish groups. The exact analysis
of the variation — e.g. of pairs of related species or races — from
this point of view has been very little studied, and more work
of this kind is desirable. The facts we give are slight in amount,
but we believe they may be typical of a larger array. It must
be borne in mind that such studies as are available are made
on limited sections of populations, and we have no means of
saying how far the correlations indicated are characteristic
of the groups over their entire range. Lastly there is available,
as far as we know, no analysis of all the diagnostic features of
a pair of allied species.

We will first give (a) some data concerning the correlation
of characters within species, and then (b) examples of the

CORRELATION

163

correlation between characters diagnostic of pairs of related
species.

(a) :

Species

Authority

Characters Correlation

Clmisilia itala .

Alkins (1923a)

Length

X width of shell

0-39

Ena obscura

5, (1923)

53

*■ 35 33 55

036

Rhynconella cf. boueti

,, (1923*)

33

* 55 53 33

o-86

5 3 35 55 "

„ (l-c)

35

X depth ,, ,,

030

Terebratula punctata .

„ (l-c)

35

X width ,, ,,

o-94

t> 33 '

„ (l-c)

55

X depth ,, ,,

0-94

35 55 ■

„ (/.*.)

Width

X 55 ,; ,,

o-88

Portunus depurator

Warren (1896)

Total breadth X frontal breadth

o- 14

(carapace)

j j )»

53 55

55

,, X R. dentary
margin

0-56

55 >y

55 55

R. antero-lateral length X L.

dentary margin

0-74

>> >>

55 55

R. antero-lateral length X L.

antero-lateral length

086

Gryllus sp.

Lutz (1908)

Length of body X tegmina

061

jj )>

' 55 55

55

,, posterior femora X
tegmina

080

j> >j •

■ 55 55

55

,, ovipositor X tegmina

073

33 3) •

» 55 55

33

,, body X posterior
femora

o-53

>} 3> •

• 55 55

35

,, ovipositor X posterior
femora

0-77

35 55 •

• 53 55

33

,, ovipositor X body

0-70

Carbonicola qffinis

Trueman (1930)

Height X length

028

It will be seen that in these examples, which have been
collected at random, the average correlation is o • 56, which is
a fairly high figure.

In all probability this figure is rather in excess of the general
average. Thus, in his analysis of the variation of various
groups of invertebrate fossils, Trueman (I.e.) emphasises the
tendency for the characters of species to vary independently
of each other, and the consequent low correlation.

(b) Alkins (1928) has analysed the variation of the land
snails Clausilia rugosa and C. cravenensis in a study which is
particularly valuable on account of its being based on samples
taken from different colonies (though from a restricted region).
He studied two of the diagnostic characters, viz. length and
major width of the shell. He gives no statement of the corre-
lation of those characters in the two species over the whole

1 64 THE VARIATION OF ANIMALS IN NATURE

area investigated, as his work is centred on the analysis of the
correlations in each species in each colony. But he states
(p. 68) that ' the mean altitude and mean diameter of C.
cravenensis always exceed those of C. rugosa . . . individually
their altitude ranges may overlap to some extent, but their
diameter ranges hardly ever . . . doubtful cases (shells of un-
certain specific identity) are rare.' From this one may infer,
though without a definite measure, that the correlations between
length and width and between shortness and narrowness are
marked enough to render it easy to decide at once to which
species a shell must be assigned. Within the range of each
species, however, the correlations are low, in a selected series
of colonies (p. 68) never exceeding o • 50 and sinking as low as
OT, the mean being for rugosa 0-31, and for cravenensis 0-39.
This is interesting as showing that, though the two species
tend to reveal two regularly contrasted characters, the latter
do not maintain an absolute identity of association within the
species.

Alkins (1921) and Alkins and others (1921) also studied
the correlation of various proportion-indices in Sphaerium
lacustre, corneum and pallidum. They find that in all three
species the correlation of length and width, length and thick-
ness, and width and thickness has a high value, never falling
below 0-9. In S. lacustre and S. corneum length and width are
certainly diagnostic.

We owe to Sumner (see his summary, 1932, and bibliography
of a long series of papers) a valuable study of interracial
diagnostic characters in the deer-mouse (Peromyscus) . He states
(1928, p. 183) that ' there is no general tendency for the
elements which distinguish one race from another to vary
together within the single race.' He does not state what the
figure for the total range of variation is, but from this paper
and a later one (1929) we may infer that the distinctive
interracial correlations may be fairly well marked : indeed in
the forms dealt with in the latter paper the amount of inter-
mediacy is very slight (I.e. p. 112). He sums up the situation
in his final review as follows : ' Interracial correlations, so
far as these concern the length of body parts, are altogether
erratic. While within single populations certain parts (e.g.
tail and foot) tend to vary together in their relative size, such
concomitant variation may or may not be encountered when

CORRELATION

165

Hj L CMLO

r ' HwnKH2S U

CARLOTTA

FORT BRAGG

90 CASES
Mean 93.31

Ha

DUNCAN MILLS
LL_D

££i

CALISTOGA

(U US CASES
jjnj HtH 81 29 S

VICTORVILLE

I

CARLOTTA

i07DOE5
hOD4l

+

^^

FORT BRAGG

DUNCAN MILLS

n-i

CALISTOGA

_ fiyiie cases Uh , n

on

.Ji CASES
MKn32.ii

n

U

LAJOLLA

"i n

[u

■1

OR

133 CASES

Mean 28.0

VlCTORYlllF.

Hi

65 to 7s eo es 90 9s 100 >«4 «o 115 i?o 70 25 jo is « « 50 SS
1 ■ ... 1 ... . 1 1 .... 1 .... I .... 1 .... I .... i .... I ■■ 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Fi 1 1 1 1 1 1 1 1 1 1 1

Fig. 19. — Peromyscus maniculatus. Histograms showing Distribution of Fre-
quencies for the Various Values of Relative Tail-length (left) and
Relative Width of the Tail-stripe (right) in Eight Localities. The
Broken Lines connect the Means of the Various Series.

(From Sumner, 1920.)

1 66 THE VARIATION OF ANIMALS IN NATURE

we examine a series of geographic races. Throughout con-
siderable tracts a positive correlation may hold : in other
territories the correlation may be entirely dissolved. Intra-
racial correlations in pigmental characters, on the other hand,
are even more pronounced than are interracial ones. Darker
races, like darker individuals, tend to have more extended
coloured areas in their pelages, deeper pigmentation in the
skin of their feet, broader (and longer) tail stripes, etc'

Other papers of Sumner's (e.g. 1918, 1920, 1923) make it
quite evident that the character complexes which distinguish
subspecies are by no means highly correlated, and certainly
his evidence concerning the behaviour of these complexes on
crossing shows (1923) that they fail to behave as units. The
systematic analysis of species and geographical races has
yielded similar results, and there is a good deal of evidence
that the characters distinguishing such groups vary inde-
pendently (cf. Swarth, in Linsdale, 1928, p. 257 ; Mertens,

193^ P- 205).

The discussion as to the kinds of correlation (p. 161) shows
that they may be reduced to two fundamental types : (1) one
in which the characters stand in relation to each other as
cause and effect, and (2) one in which their association is
coincidental (' combination ').

(1) This includes (a) the dependence of one part on another,
and (b) the reciprocal dependence of two parts on each other.

(a) A structure may depend, as we have seen, on another
structure on which it has no effect itself. Certain of the
phenomena of development have been interpreted as due to
various kinds of stimuli (chemotaxis, thigmotaxis) exerted
by one part on another. The classical example is the failure
of the lens of the vertebrate eye to develop if the optic
capsule fails to make contact with it. Other examples are
discussed by Jenkinson (1909, p. 273 and foil.).

The dependence noted here affects the main architecture
of the parts rather than the characters which distinguish
species. But certain characters of proportion are obviously
influenced by growth principles, and Huxley (1932, passim)
in particular has applied the principle of heterogonic growth
to explain certain differences between species. (See especially
the case of the Lucanid beetle, Cyclommatus tarandus.) It may
therefore come about that correlated specific differences

CORRELATION

167

4+-H-

■H— ¥

a

CO C_>

II I I I I I

•> m

2;

o

Bi
O

U

t.
o

w

5

►J

<

w
z

w

O

o

z

o

X

<3 W

S w

3

a ^

£ E

o

3

o

d a

I— +

CO O

: <_> u.

■HH H— I"

Bi

W

F

o
<
b;
<
x
O

z

w
>

w

C/2

w-

+H-

z

o

— '

2

>

*/>- o

H-H-

I I I I

c

■JCJ •—

1 68 THE VARIATION OF ANIMALS IN NATURE

consisting of proportional differences of parts might arise, if
the characters in question were related to the absolute size of
the animal and if the latter were of selective value.

Another kind of correlation of this type is seen in the
dependence of a structure on the specific activity of a gland —
particularly those of internal secretion.

(b) As regards the reciprocal dependence of the parts of
the living organism little can be said. There is some evidence
that in the course of development various parts are dependent
for their expression on each other. This fact was indeed made
a prominent feature of Driesch's theory of development.
According to Jenkinson {I.e. pp. 75—7), the dependence
diminishes with age ; correlation is only high during periods
of rapid growth, and there is an increasing power of self-
differentiation. That certain relations of this kind persist
into later life is seen in the dependence of the extremities on
the nervous system in the Vertebrata.

Correlation has often been invoked to supplement the
theory of Natural Selection. The modification of apparently
non-serviceable structures has thus been attributed to their
being correlated with characters influenced by selection.
The nature of the correlation has not seriously been studied.
It was probably some kind of causal association such as we
have been discussing that was in (e.g.) Darwin's mind when
he stressed its evolutionary importance. Not only, however,
does this kind of correlation require much more study and
exploration, but also the efficacy of selection itself (Chapter VII)
is open to question. Possibly some differences of size and
proportion between species have been produced by selection
acting on characters correlatively associated in this way.
Whether differences of colour, ornamentation and the arrange-
ment of parts are influenced by it is far more problematical.

(2) We have already seen (p. 162) that we have to deal
here with two types of correlation, viz. (a) one due to the
coincident effects of various external causes, and (b) another
due to the mechanism of heredity.

(a) There is a variety of ways in which such correlations
can arise. Thus Hubbs (1926) shows that low temperatures
tend to make fish large, small-headed and small-eyed. High
temperatures make them small, large-headed and large-eyed.
Schmidt (1930, p. 28) finds that in the Atlantic Cod (as in

CORRELATION 169

the Salmon and Lebistes) external factors (? salinity and
temperature) can alter the average numbers of vertebrae and
fin rays.

It is probable that the groups of characters employed in
diagnosing species are not usually held together by a corre-
lation of this sort. How far the coincident effects of several
separate factors or the multiple effects of single factors of this
order may have been influential in evolution must be left for
a later discussion (p. 172). Theoretically at least groups of
correlated specific characters might arise as the direct effect
of environmental causes or from simultaneous selective
processes. The value of this suggestion depends on the
evolutionary importance we attach to these processes. There
is, however, some evidence of a convincing nature that charac-
ters of the same kind as distinguish taxonomic species are
altered in association as the result either of single environ-
mental factors or of several such factors acting concurrently.
Thus it is known that in the Baltic Macoma baltica and Mya
arenaria are both smaller (Brandt, 1897) and have thinner
shells than usual (Mobius, 1873). Bateson (1889) found that
the proportions and shape of Cardium edule are modified in the
brackish water of the Sea of Aral. Sumner (191 5) experi-
mentally induced lengthening of tail and foot in white mice by
high temperature, and such differences are known to differen-
tiate the wild races of rodents. Perhaps we should draw
attention to Sumner's point (1932, p. 53) that, though in some
of his experimental cases we might expect ' parallel modifica-
tion by the environment, the latter cannot account for
correlations which increase in segregating generations of
hybrids.'

There is a theoretical possibility that all the characters of
a species may be produced by several coincident selective
processes or by a single selective process affecting several
characters. The wing-pattern of a mimetic butterfly would be
an example of the latter. The pattern is composed of several
elements, all of which are associated in the mimetic effect and,
on the selection hypothesis, must have been produced co-
incidentally by a single selective process. As an example of
the modification of several quite distinct structures in rela-
tion to a special mode of life we may cite Hora's (1930,
passim) demonstration that in torrent-dwelling species several

170 THE VARIATION OF ANIMALS IN NATURE

characters may be modified in the same species as a result of
adaptation to the particular habitat.

Any attempt to explain such correlation as the expression
of single or multiple effects of Natural Selection must, of course,
depend on whether Selection is a vera causa. The occurrence
of correlation should not be held to be a proof of the action of
Selection.

It is possible in some cases that isolation may make for
correlation, as, for instance, when a few individuals of an
aberrant form are isolated on an island, so that an association
of characters originally accidental is prevented from returning
to the normal distribution (by the lack of facilities for crossing
with the parent form). Hagedoorn and Hagedoorn (192 1),
in particular, have stressed the point that isolation will lead to
inbreeding of the isolated stock, with a considerable likelihood
of the establishment of a new mean.

(b) The occurrence of correlation due to the mechanism of
heredity has been discussed by Robson (1928, p. 229), who
cites a certain number of instances revealed by genetic experi-
ment in which, on crossing, character complexes tend to hang
together, instead of being dissociated as is the usual fate
of independently segregating characters. The majority of the
instances are found amongst plants ; but a more limited
number occur in animals — e.g. Castle and Wright (19 16),
Phillips ( 1 921), Harrison (1916, p. 145 ('segregation en bloc'
of specific characters]). The actual basis of such correlation
is obscure. The relation between linkage and correlation has
been stressed on several occasions, but Robson (I.e. p. 231)
makes it clear that it is difficult to attribute the correlation of
specific characters to linkage. Sumner (1932, pp. 53-5) had
discussed this question in greater detail in connection with
his interracial studies of Peromyscus, and finds good grounds for
preferring the hypothesis of the multiple effects of single genes.
According to Haldane (1932, p. 114), 'a number of cases
of multiple action of this kind in Drosophila ' are available.
At present very little is known concerning such ' multiple '
effects in animal genetics, and certainly we are not in a
position to discuss how far they are influential in producing
intraspecific character correlation.

In any homozygous strain or pure line all phenotypic
characters are more or less strongly correlated together until

CORRELATION 171

mutation occurs. The degree of correlation will depend on
the susceptibility of the characters to environmental influences.
Again, the phenotypic expressions of dominant genes lying in
the same chromosome will be more or less strongly correlated,
depending on the amount of crossing over. We can even
invent a hypothetical case in which two characters would
show complete correlation, by assuming that each of the genes
responsible was lethal when not associated with the other.

On the whole we believe that the bulk of intraspecific
correlations is due to most members of a species being homo-
zygous for their distinctive characters. As Fisher (1930, p. 124)
has said, ' the intimate manner in which the whole body
of individuals of a single species are bound together by
sexual reproduction has been lost sight of by some writers.
Apart from the intervention of geographical barriers so
recently that the races separated are not yet regarded as
specifically distinct, the ancestry of each single individual, if
carried back only a few hundred generations, must embrace
practically all of the earlier period who have contributed
appreciably to the ancestry of the present population. If we
carry the survey back for 200, 1,000 or 10,000 generations,
which are relatively short periods in the history of most species,
it is evident that the community of ancestry must be even
more complete. The genetical identity in the majority of
loci, which underlies the genetic variability presented by most
species, seems to supply the systematist with the true basis of
his concept of specific identity or diversity.' Hagedoorn and
Hagedoorn (1921) have expressed the same idea in a rather
different way. In nearly all species the population is not of
constant size throughout the year or from one year to the next.
This is particularly obvious in all species which, in temperate
climates, have a definite breeding season. The large popula-
tion existing at the end of the breeding period is gradually
depleted till only a relatively small number is available to
breed again the next year. The survival of only a small
number to carry on the species must mean an enormous
reduction in variation each year, probably enough to account
for the observed constancy of most species. The chance that
any variant represented by only a few individuals will form a
part of the next year's initial population is very low, the
magnitude of the chance depending (apart from survival

172 THE VARIATION OF ANIMALS IN NATURE

value) on the ratio between the numbers of the variant and
the total number of individuals in the species.

As we have said already, the actual basis of correlation is in
nearly all species unknown, but there are certain methods by
which important information may be obtained, indicating
that the correlation is often of the second type.

(i) There may be considerable presumptive evidence that
the characters are physiologically independent of one another.
Thus in insects we should have no reason to suspect a direct
physiological relation between the arrangement of the wing-
nervures and the structure of the external genitalia, or, in
birds, between the shape of the beak and the colour of the
tail. How far the mere unlikelihood of a relation is significant
has to be decided in each individual case. The somewhat
anecdotal instances of correlation between apparently inde-
pendent parts which are cited by Darwin should be borne in
mind.

(2) Some specific characters are unusually variable and
cannot, therefore, show a very high correlation with more
stable ones. Wherever low correlations are observed, there
is a likelihood that the basis is not physiological. More im-
portant evidence can be obtained in species in which in some
individuals a single specific character is replaced by one
normally distinctive of another species. The identity of such
aberrant individuals may be reasonably certain, since the other
members of its character complex are still associated together.
Further, these variant forms may be quite rare, so that the
correlation of the character in the species as a whole remains
high. Such cases strongly suggest that the character (and, by
inference, similar characters in allied species) is capable of
independent segregation. 1

As an instance of this type of evidence we may mention the
Tortricid moth, Euxanthis straminea [cf. Waters, 1926, p. 159).
A form has occurred in S. Devon (and elsewhere) which, in
its large size and distinct dark wing markings, resembles the
allied species E. alternana. The aberrant specimens, however,
have typical genitalia, and the direction (though not the inten-
sity or dimensions) of the wing fascias is normal. Similar

1 Nabours (1929, p. 33) has made the interesting observation that there are
differences in the linkage relations of similar patterns in different species of
grouse-locusts.

CORRELATION

173

Cases are mentioned by Warren (1926) in his account of the
European Hesperidae (' Skippers'). One of the authors has

Fig. 20. — Specific Differences between the Queens of Vespa germanica F. and
V. vulgaris L. Yellow Markings, Compound Eyes, and Ocelli shown in
White. Black Parts shown in Black. Reddish-brown Parts dotted.

1. Differences in the markings of the head (head seen anterodorsally,

antennae removed, antennal sockets cross-hatched). In V. germanica
(A-C) the black marks on the clypeus are variable and the black
stripe between the yellow supra-antennal spot and the yellow in the
eye-emargination narrows posteriorly. In V. vulgaris (D) there is
constantly a black ' anchor ' mark on the clypeus and the black stripe
broadens posteriorly.

2. Head seen from the left side (antenna; removed). In V. germanica

(E and F) the postocular yellow stripe is normally continuous. In
V. vulgaris (G and H) the stripe is normally interrupted. Various
intermediates (F and G) occur.

3. Left half of pronotum and mesonotum (with tegula), seen from above.

The yellow pronotal stripe in V. germanica (I-K) is more or less angled
outwardly ; the tegula, typically, is yellow with a small reddish
outer spot and a small black inner one. In V. vulgaris (L and M) the
pronotal stripe is narrow and parallel-sided and the tegula is typically
reddish-brown with two yellow and one black spot. Figures K and L
show intermediates.

noted a similar phenomenon in the wasps Vespa vulgaris and
V. germanica (cf. fig. 20).

174 THE VARIATION OF ANIMALS IN NATURE

(3) A good many of the specific characters observed in a
genus may occur in different combinations amongst the various
species. In so far as we are justified in assuming that similar
characters in different species can be rated as fundamentally
the same, we may use these permutations as evidence of
independent segregation. Lutz (1924) has described a case of
this sort in the S. American stingless bees (Melipona). Kinsey
(1930) presents convincing evidence that short-winged forms
of Cynips have been repeatedly produced from long-winged
species.

(4) Crosses between distinct species may provide convincing
evidence as to the essential independence of characters.
Sometimes the hybrids show an extraordinary intermixture of
the characters of the two parents. Less commonly some of
the characters tend to remain together and segregate in blocks.
It is not actually necessary to assume that the correlation
between characters segregating in blocks is of a different nature
from that between characters segregating independently.
It might be suggested that disharmonies during cell-division in
the hybrids make normal segregation impossible.

We have hitherto spoken of specific characters as units
without considering their relation to a genetic basis. This
relation is of importance when we try to define the meaning
of the term ' independent segregation.' An initial complica-
tion in the discussion arises from our ignorance as to whether
apparently similar phenotypic characters in different indi-
viduals or species really are the same. We know from genetical
researches that superficially similar mutants are not necessarily
due to a mutation at the same locus. In dealing with the
mutant forms of a single species the question can be always
answered by making the appropriate crosses. But in the mass
of species such crosses have not been or cannot be made. An
individual aberrant in one specific character is not usually
recognised to possess theoretical interest until death has made
experiment impossible. The direct identification of similar
specific characters in different species is usually impossible,
owing to refusal to cross and to the rareness of hybrids. The
point we wish to make here is that practically no analysis of
specific characters in terms of genes is available. Sturtevant
(192 1, p. 1 19 and foil.) has shown that some of the mutations in

CORRELATION 175

Drosophila resemble generic or family characters which dis-
tinguish other groups. But even in Drosophila there is practi-
cally no evidence as to the genetic basis of the characters used
to separate species in that genus. The geneticist, naturally
enough, has concentrated on the mutations most easily observed
and studied. A special search for mutation in characters
known to be of specific value seems scarcely to have been
attempted.

We are in great need of information as to whether the unit
phenotypic characters are really genotypic units. We may
discard for the moment the numerous specific characters
which are not unambiguously definable as units and consider
only such differences as : number of metameric parts, pre-
sence or absence of definite spines or bristles, development of
definite coloured patches, etc. These are the sorts of characters
which appear in different combinations in allied species and
are therefore spoken of as segregating independently. Analogy
with the results of genetical studies would lead one to expect
that a number of these character differences might be due to
more than one gene difference. We are seeing here, in the
segregation of unit phenotypic characters, the transfer of
blocks of genes, and it may be asked how these blocks come to
remain as units.

When, on crossing two species, all degrees of intermediacy
are found in any character, we have a clear case of the breaking-
up of one of the gene blocks referred to. When, however,
the character acts as a unit, we do not know enough as yet to
affirm that only a single gene is necessarily involved. The
possibility of some unsuspected correlation mechanism cannot
altogether be dismissed.

The recent emphasis on the idea of the multiple effects of
single genes also raises a difficulty. The result of postulating
multiple effects is to increase the number of genes which are
regarded as contributing to the phenotypic expression of any
one character. But, evidently, the more independent genes are
concerned in the expression of characters, the more difficult it is
to explain the independent segregation of characters as units.
At the present moment this point has scarcely more than
theoretical interest, but we shall have to return to it (p. 177) in
our consideration of the validity of a unit-character analysis of
living animals.

176 THE VARIATION OF ANIMALS IN NATURE

It is instructive to compare the correlations between specific
with those obtaining between generic and family characters.
Some diagnostic characters are ' good ' and hold for every
member of the genus. Others are variable or only present in
some members. The permutations of characters amongst
related genera or families are also common. In fact, it would
seem at first sight that at all stages in divergence the correla-
tion between characters was of the same nature and depended
on the extent to which the different unit characters had suc-
ceeded in permeating populations of different sizes. Highly
correlated characters seem to be those for which large numbers
of individuals are homozygous. The position of a character
in the hierarchy would seem to depend on the extent to
which it had spread, and this, in turn, approximately on the
time that has elapsed since it first appeared.

The study of lineages by palaeontologists appears to bear
out such a view of evolution. The material studied (Bryozoa,
Mollusca, Brachiopods, Echinoids, Mammals) is restricted by
certain preliminary requirements. The organism must possess
sufficient characters (in the fossil state) to admit of establishing
correlations between groups of characters. Indeed, it may be
suspected in some phyla (e.g. certain Mollusca) that the number
of characters involved is actually too small for the results to be
very significant. Secondly, abundant material must be avail-
able of approximately the same age. Lastly, the forms studied
must occur in an uninterrupted succession of strata, so that
the fate of the character combinations may be revealed.
We do not wish to deal fully with the palaeontologists'
data in the present chapter, but only to note certain
general conclusions, of which the most important are the
following.

Each character evolves as a separate unit. In different
lineages the same character may evolve at very different rates
(cf. Trueman, 1930), so that in one case it is associated with
one set of characters and in another with quite a different level
of divergence. Correlations between groups of characters
are often only maintained at one horizon. As we traverse the
strata the associated characters alter. These conclusions, derived
from the study of actual fossils, are exactly what one would
have expected from a study of living species. In the latter, the
variation in correlation and the permutations of characters

CORRELATION 177

might have allowed us to infer that the history of species in
time would be exceedingly complex. Groups of living animals
are broken up into hierarchies of divergence isolated from one
another to a varying extent. We can, therefore, if we wish,
separate any two groups by a single differentiating character,
but only at the expense of ignoring all the other features in
which they may happen to agree or differ. The palaconto-
logical evidence that single characters evolve more or less inde-
pendently of one another is only a corollary of their failure
as group-indicators in living forms. From this point of view
evolution is a relatively simple process with two main aspects —
(1) the origin, in a relatively small number of individuals, of
new characters, some of which spread throughout large popu-
lations, and (2) the ' trying-out ' of such material in all sorts
of combinations.

It does not require a highly developed critical faculty to
see that this is a very simplified and abstract account of the
living organism. The picture of species as being built up like
houses from bricks is very hard to reconcile with any theory of
development. The phenomenon of regulation in the individual
is so like that of correlation in the species, that it is difficult to
believe that the modern genetic concept of species as mosaics
of gene interaction illuminates more than one aspect of our
problem. If the regulatory activity of organisms can deter-
mine the development of a single blastomere into a whole
rather than a fractional organism, it would be strange if the
relations between specific characters were not also in some way
controlled.

We may consider first how far it is possible to sum up an
organism as a mosaic of unit characters. Such a question
might be asked not only with regard to the relatively crude,
unanalysed specific characters, but also with regard to the
supposedly more fundamental genes of the geneticist. In
discussing specific characters as they appear in taxonomy we
have not indicated how far the taxonomic definitions are in-
complete. Actually everyone knows that specificity is not
something superficial and external, like the last coat of paint
on a new car, but something which permeates the organism
through and through. It may show itself in any part of the
organism, whether structural, physiological or psychical. It
is seen perhaps most characteristically in the apparently

N

178 THE VARIATION OF ANIMALS IN NATURE

unique character of the proteins of each species. Experimental
embryology has shown that this unique character may be
maintained even in small fragments grafted into an individual
of another species. Perhaps some taxonomists would bring
forward certain pairs of very closely allied species that seem to
differ only in one or two unit characters. But we think it can
be safely said that, even in these cases, the few unit characters
are only indicators which the taxonomist finds convenient to
use. As soon as a comparison can be made on the basis of a
sufficiently large number of individuals studied alive as well as
dead, all sorts of other differences begin to appear, sometimes
not easy to define, yet statistically significant. Sometimes it is
a slight difference in habit that first suggests to the taxo-
nomist that there may also be undetected morphological
differences.

Such considerations make it very doubtful how far the
abstract concept of species as mere collections of characters
really covers all the facts. But we may further recall that
many characters, which in taxonomy are conveniently con-
sidered as units, actually affect many different parts of the
body. Such are size, colour (especially 'ground colour'),
hairiness and sculpture. It is possible that these could be
reduced to unitary physiological effects, but this is unlikely.
As soon as we consider structure in terms of the physiological
processes that give rise to it, the whole idea of units becomes
more difficult. This is implicit in the idea of the multiple
effects of genes. A complete extension of this theory would
make every gene responsible in some degree for every part of
the whole, and the unit-character conception of heredity
would go by the board. Actually geneticists are now more
cautious than they were in the past in their theories as to how
genes affect development. As Morgan (1932a) has recently
stated, ' the earlier, premature idea, that for each character
there is a specific gene — the so-called unit character — was
never a cardinal doctrine of genetics, although some of the
earlier popularisers of the new theory were certainly guilty of
giving this impression. The opposite extreme statement,
namely, that every character is the product of all the genes,
may also have its limitations, but is undoubtedly more nearly
in accord with our conception of the relation of genes and
characters. A more accurate statement would be that the

CORRELATION 179

gene acts as a differential, turning the balance in a given
direction, affecting certain characters more conspicuously
than others.' This view certainly harmonises better with the
data of genetics, but it does not enable us to envisage the
process by which complex structures develop harmoniously.

This is the question which has been raised by Russell
(1930). He points out that there is no evidence for a qualita-
tive division of the chromosomes at any stage of development.
Each cell (in typical cases) has the same equipment of hereditary
material. The fact that different cells give rise to such varied
structures can only be explained by considering the spatial
relations of the cell to the whole. Russell is so impressed by
this antinomy that he is prepared to discard the whole unit-
character hypothesis of heredity. But this extreme attitude
appears perverse. Somehow or other the quantitative pre-
dictions which can be based on the chromosome theory must
be accounted for. The difficulty here raised has also been
considered by Woodger (1929, chapter ix, especially sect. 9).
He attempts to visualise development as a process of gradual
realisation of spatio-temporal parts, while genes are concerned
only with the characterisation of the parts. In order to include
those cases in which whole parts may be inherited on Mendelian
lines (e.g. vertebrae) he suggests that, for the purposes of
genetics, the part should be defined rather by its dimensions,
so that ' absence ' is merely the end term in a gradual process,
rather than something sharply different from ' presence.'
This idea of the relation between heredity and development
seems helpful in trying to orientate our fragmentary knowledge,
but scarcely helps us as yet in the matter of character correla-
tions. The characters do not act as separate units in develop-
ment, and we cannot help suspecting that whatever controls
the orderly unfolding of the inherited organisation must be
deeply concerned with the correlation of the characters on
which the end result largely depends.

We feel that there is a very real difficulty here. On the
one hand we have the obvious and incontestable fact that
(p. 163) the characters defining species are rather loosely
correlated, we have produced certain reasons (p. 172) for not
considering their association as of a ' physiological ' (i.e.
intimate and causal) nature, and we have definitely suggested
that it is in the bulk of cases due to the members of species being

180 THE VARIATION OF ANIMALS IN NATURE

homozygous for their distinctive characters. Nevertheless
we have shown that ' specificity ' may be a deeply seated
property of the organism, and that the facts of development
argue a close connection between the parts of the organism
and an interdependence from which even the more superficial
character expressions could hardly be expected to escape.
There is some risk, it is true, in exaggerating the degree of this
dependence, and we should remember that progressive eman-
cipation and self-sufficiency of the parts which Jenkinson
{I.e. p. 1 68) has described.

The question which we have to face is — are the complexes
of specific characters in their ultimate genetic representation
simply fortuitous mosaics associated either by the mechanism of
heredity or by the coincident effects of selection or environment,
or are they bound together more intimately by the organic
association seen in development ? It is highly doubtful whether
we know enough about the basis of specific characters to come
to any decision. Such evidence as we have certainly suggests
that the association is, on the whole, fortuitous. If this view
is ultimately found to be correct, a general question of some
importance is raised, and that is — how does it come about that
some parts are more independent of the general organisation ?
We might suggest that specific and racial characters, being
newly acquired, have not yet been incorporated in the general
unity of the organism and have not yet attained that closeness
of association and mutual dependence that is found in other
parts. How such dependence has arisen, and how exactly the
accretions produced by new evolutionary steps have their
association transformed from a fortuitous to a permanent
basis, is a matter which it does not yet seem possible to
discuss (cf. Chapter X, p. 370).

CHAPTER VII

NATURAL SELECTION

In this chapter we propose to examine as fully as possible the
validity of the theory of Natural Selection in so far as it
depends upon zoological evidence. We believe that a final
verdict on the efficacy of selection may be arrived at on zoo-
logical evidence and that there is no special category of botanical
data that is of crucial importance in determining the value of
this doctrine.

In the seventy-six years that have elapsed since its first
announcement the main framework of this theory has remained
unchanged. It has been rejected by many and held by others
to have a less universal application than was originally believed.
We have obtained a clearer insight into the various natural
processes involved and a wider knowledge of the historical
facts of evolutionary change. But no material alteration of
the basic principles has been introduced and, for those who
subscribe to its tenets, it stands very much as it did when it
was first announced. Nevertheless, the volume of evidence
that may be produced both to support and to undermine it
has expanded and it is not inaccurate to say that the accumu-
lation of data on the various issues involved has outrun the
synthetic and comprehensive treatment of the subject. It
is therefore desirable at the offset to indicate what kind of
evidence is now available and to what degree of completeness
the field of inquiry has been covered.

i. Darwin's Statement of the Evidence.— We may take
the evidence as presented in ' The Origin of Species ' (Darwin,
1884) as the chief demonstration by Darwin of the efficacy of
Natural Selection. In his letters and other works there is a
considerable mass of corroborative evidence and reasoning,
but the actual marshalling of the evidence for the operation

1 82 THE VARIATION OF ANIMALS IN NATURE

of the principle is given in ' The Origin.' As stated in that
work the proof consists of four essential parts :

(a) A demonstration of the efficacy of selection by Man.

(b) A survey of the circumstances in which Natural Selection

is assumed to work (numerical increase, struggle for
existence, variation, etc.).

(c) A consideration of the phenomena of adaptation.

(d) A survey of the facts of ' divergence ' in relation to

distribution in time and place.

The occurrence of sundry secondary phenomena of im-
portance in the theory (such as correlation and isolation) is
also dealt with.

Throughout the work Darwin does not clearly distinguish
between Evolution as an historical process and Natural Selection
as the effective agent. A large amount of his data merely
serves to prove the occurrence of the former. The following
quotation from ' Animals and Plants under Domestication '
(1905, vol. ii, p. 10) serves to illustrate this. ' The principle
of Natural Selection may be looked at as a mere hypo-
thesis, but rendered in some degree more probable by what
we positively know of the variability of organic beings in a
state of nature, by what we know of the struggle for exist-
ence, and the consequent almost inevitable preservation of
favourable variations ; and from the analogical formation
of domestic races. Now this hypothesis may be tested— and
this seems to me the only fair and legitimate manner of con-
sidering the whole question — by trying whether it explains
several large and independent classes of facts, such as the
geological succession of organic beings, their distribution in
past and present times, and their mutual affinities and homo-
logies. If the principle of Natural Selection does explain
these and other large bodies of facts, it ought to be received.
On the ordinary view of each species having been indepen-
dently created, we gain no scientific explanation of any one
of these facts.' To a modern reader, it cannot but occur
that any theory of evolution would explain, say, the facts of
homology and geological succession : Natural Selection has
no particular advantage in this respect.

In Darwin's treatment of the subject no proof is adduced
that a selective process has ever been detected in nature.

NATURAL SELECTION 183

Throughout the work such a process is suggested and assumed :
its actual occurrence is nowhere demonstrated. Stated briefly,
the argument is as follows : selection has plainly c worked '
in domesticated races, analogous results and appropriate
processes and conditions are found in nature, therefore we
may assume that selection works in nature. In short, the
proof is based on circumstantial rather than direct evidence,
and the mainstay of the case is the analogy between Artificial
and Natural Selection.

On the question of variation Darwin's mind evidently
hovered in some uncertainty. He clearly thought of it ' as
indefinite and almost illimitable ' (' Animals and Plants under
Domestication,' ii, 292). In the sixth edition of ' The Origin '
(1884, p. 648) he was still under the impression that to some
extent ' physical, i.e. environmental conditions seem to have
produced some direct and definite effect . . . with both
varieties and species use and disuse seem to have produced
a considerable effect.' Nevertheless in ' Animals and Plants '
(I.e.) he had doubted whether ' well-marked varieties have
often been produced by the direct action of changed condi-
tions without the aid of selection either by man or nature.'
Bateson (1909, p. 209) points out that Darwin originally held
that ' individual variation ' (i.e. mutation) was of high im-
portance, but subsequently abandoned the belief. With
these minor inconsistencies and changes of opinion we need
not occupy ourselves.

It is far more relevant that, though the importance of
Natural Selection is always stressed, Darwin nowhere suggests
that it is the only modifying agency. He always laid stress
on isolation and correlation and, as we have seen, on the
effect of the environment. He even goes so far as to suggest
that the modification of a species may proceed without selec-
tion — that species may arise and be perpetuated ' for no ap-
parent reason.' He carefully disposes of a (for him) too rigid
and literal application of the theory — e.g. when he shows that
Bronn's objection to it, based on the occurrence of parent
species and their varieties living side by side, may be met by
assuming that, if both had become fitted for slightly different
habitats, they might subsequently extend their ranges and
overlap (1884, P- 2D 4). It is quite clear that he thought that
varieties might arise and species might exist without having

1 84 THE VARIATION OF ANIMALS IN NATURE

any special adaptive qualifications. Recent studies have much
diminished the value of Darwin's subsidiary hypotheses.
Consequently the lack of any clear demonstration that naturally
occurring varieties do indeed experience a differential mortality
is all the more serious. Tschulock (1922, p. 290) calls ' The
Origin of Species ' ' ein logisches Monstrum,' because it
deals with the secondary issue before the primary. It seems
to us to deserve this censure far more because it fails to
demonstrate the actual occurrence of the process which it
seeks to establish as the cause of evolution.

2. Subsequent Confirmation and Development of the
Theory. — It is pertinent to inquire whether the theory has
undergone any radical modification as a result of the enlarge-
ment of the field of inquiry, and whether it needs to be restated
in a form different from that presented by Darwin.

It seems to us that the theory has persisted in very much
the same form as that in which it was originally presented.
There is no need to enlarge on the fact that Darwin's belief
in the heritable effect of ' changed conditions ' was abandoned
by most students under the influence of Weismann's teaching.
Although we do not suggest that the evidence in favour of
the environmental origin of mutations impels us to return to
Darwin's somewhat vague and naive belief in the importance
of ' changed conditions,' we think that it cannot be sum-
marily dismissed, and that more allowance has to be made
for the likelihood that mutations may be due to external
causes. There are, however, two points on which modern
investigation compels us to revise the conception of selection
itself.

(1) Fisher (1930, chapter i) has very clearly shown the
effect on the concept of selection of the discovery that in-
heritance is governed by a particulate instead of the blending
principle which Darwin — perhaps against his better judgment
(cf. Fisher, I.e. pp. 1-4) — had in mind. The point at issue
is that, with a blending principle at work, ' if not safeguarded
by intense marital correlation, the heritable variance is
approximately halved in every generation,' and ' to maintain
a stationary variance fresh mutations must be available in
each generation to supply the half of the variance so lost.'
On the particulate theory the mutation-rate may be far
smaller than that required by the blending principle.

NATURAL SELECTION 185

(2) It is implicit in Darwin's presentation of the theory
that single variants will be ' swamped ' by intercrossing, and
that the swamping of new variants is only avoided if they
happen to be serviceable and if there are enough of them to
reach maturity and breed together. Though even on the
particulate theory of inheritance a character depending on
several genes would undoubtedly run the risk of being
' swamped ' by intercrossing, much of the risk envisaged by
Darwin is seen, in the light of more exact knowledge, to be
non-existent. There is, however, at the present time an
increasing emphasis laid on the effects of wholesale elimination,
and in particular on the slight chance that a single mutant
will have of surviving unless it has some selective advantage.
A tendency has thus arisen to stress the importance of selection
in serving to multiply or ' spread ' variants, as opposed to its
value as a means of preventing the ' swamping ' process.
This valuation of selection has gained ground correlatively
with the estimation of mutation-rates based on those of Droso-
phila. Whether this estimation has any general application is
discussed on p. 220, but in all probability the revised valuation
of the selective process is a just one and failure to recognise
its cogency vitiates such criticism of Natural Selection as that
of Hogben (1931, p. 180), who, in contrasting the Darwinian
conception of selection with that of the modern experi-
mentalist, suggests that a given mutant may spread and attain
a representation in a population, without discussing how it
survives the incidence of the normal death-rate.

In addition to the important developments just mentioned,
a number of inquiries all relevant to the theory have been
developed since Darwin's time, the results of which have
enlarged the field of inquiry. It is needless to mention them
in detail, but it will be apparent that the advances in the
experimental study of heredity, in animal ecology and in the
intensive study of variation in natural populations — to mention
the more outstanding developments — have profoundly altered
our views on the efficacy of selection. It is perhaps per-
tinent to add that study of the living organism as a whole,
its development, reactions and organisation, has also modified
our estimate of selection as an important agency in evolution.

It would take us very far from our course of inquiry to
describe the changes in the attitude of students of biology and

186 THE VARIATION OF ANIMALS IN NATURE

evolution towards the theory of selection. At the present time
some students have a firm conviction as to its validity and are
prepared to offer in its support, not the naive and anecdotal
evidence offered by a past generation, but the results of critical
and intensive investigation, while to others the theory is a
' dead letter ' and an historical curiosity. It is, for example,
instructive to compare (e.g.) the attitude of Fisher in this
country, who regards the efficacy of selection as an established
fact scarcely worth verification, with that of Radl (1930), who
dismisses it contemptuously as fundamentally unsound and
unworthy of serious consideration. To cite two isolated cases
like these does not give an entirely disproportionate picture
of the divergence in the minds of biological students as a whole,
and the more this divergence is studied the more apparent
does it become to our minds that it arises just as much from
the lack of any systematic arrangement of the unwieldy mass
of data as from prejudice and bias. Candid and scholarly
examinations of the evidence have been by no means lacking.
The analyses of Kellogg (1907) and Plate (19 13) are of this
type. But of recent years their critical and unprejudiced
treatment has not been followed up and the mass of observa-
tions, inference and assumptions has grown unchecked and
little attention has been paid to the logical procedure and the
types of evidence required for the purpose of either confirming
or destroying the theory.

Woodger (1929) has indicated the stages by which a
scientific doctrine advances from the status of a hypothesis to
that of a law. If we ask if Natural Selection has attained the
status of a law, the obvious answer is that many students believe
it has and others do not. This may mean one of two things — ■
either that judgment of the doctrine is still clouded by prejudice
or that the data so far obtained are in fact insufficient to
command universal conviction. It would take us too far out
of our way to consider the steps by which a scientific theory
obtains universal acceptance, the reactions of our minds to
evidence and the part played by prejudice in scientific inquiry.
It is enough to express the belief that on the evidence available
at present Natural Selection has been accepted and its prestige
created very largely on the desire for some such hypothesis.
No other explanation of the wide acceptance of the theory is
forthcoming in face of the guarded and qualified opinions of

NATURAL SELECTION 187

Darwin himself and the imperfect nature of the evidence.
Nevertheless, the doctrine has not attained the status of a
universally accepted law, and this, we believe, is because as
strong a prejudice is brought to bear against it as for it, and
(for the relatively small body of highly critical students) because
of the intrinsic difficulty of obtaining the right kind of evidence
for either its rejection or its confirmation.

It is a very unsatisfactory state of affairs for biological
science that a first-class theory should still dominate the field
of inquiry though largely held on faith or rejected on account
of prejudice. To be just, the biologist is not wholly to blame
for this position. Any attempt to bring the method of evolu-
tionary inquiry into line with that in use in more exact branches
of science and to formulate for it a logical system of proof must
recognise that the circumstances of animal and plant life and
its transformation are peculiarly complex. The number of
variables is so large that it is doubtful whether they admit of
treatment and presentation on the same terms as the data of
other sciences. If biology is not an exact science (an accusation
often made against it), this is largely due to the nature of its
data. At the very offset the units with which zoology and
botany deal are not exactly definable as regards their morpho-
logical, physiological and bionomic properties, as the limits
of species and varieties in terms of structure, habits, reactions,
etc., are very variable. Furthermore, the background of
natural forces, which, either directly or indirectly, is held to
modify animals and plants, is homogeneous neither in time
nor in space. Finally, the phenomena of growth and numerical
multiplication introduce other variables. It is thus hardly
to be expected that a ' cut and dried ' formularisation of so
many variables would be feasible.

The fact that biological science and the study of evolution
in particular are embarrassed by the complexity of their subject-
matter affords one explanation of their defects. For the rest
it seems that the lack of the exact discipline imposed, e.g. by
mathematical procedure, has given rise to the looseness of
statement that is unfortunately characteristic of much bio-
logical thought. There is something also to be seen in the
pathetic trust in observation per se. Nothing else can explain
the fact that wholly inadequate data have sometimes been
brought forward in support of the adaptive origin of certain

1 88 THE VARIATION OF ANIMALS IN NATURE

examples of mimicry, protective coloration, etc. The extent
to which evolutionary inquiry has become a prey to histori-
cal influences is seen remarkably clearly in the frequency
with which long-discredited evidence is quoted in support of
Natural Selection (e.g.) without any reference to information
or reasoning subsequently brought to bear upon it.

Procedure. — It seems to us that the unwieldy mass of facts
and arguments that has been brought forward both for and
against this theory may, for the purposes of this analysis, be
dealt with in the following order :

I. Artificial selection. (a) Under domestication. (b)
Under experimental conditions.
II. Direct evidence for Natural Selection — studies of the
incidence of death-rates in nature.

III. The nature of variation. Do living organisms vary

in such a way that a selective death-rate would be
expected to be operative ?

IV. Indirect evidence for and against the Natural Selection

theory. Do the structure and constitution of living
organisms suggest that Natural Selection has been
an important agent in their evolution ?

It should be noted that the following discussion is concerned
with two main controversial points :

(i) Evidence for and against the existence of a selective

process in nature.
(2) Evidence for and against the theory that such a process

has been responsible for the evolution of the lower

taxonomic categories.

(1) is mainly dealt with in the second section ; until the
point at issue here is settled, any discussion of IV is irrelevant.
But as the chance of any such settlement appears to be very
remote, we have in the meanwhile to consider (2) independently.

I. Artificial Selection. — (a) The origin of domesticated
races. — It is a curious fact that the value of the major proof
brought forward by Darwin in favour of Natural Selection —
viz. that selection (either conscious or unconscious) by man
has produced forms as divergent as natural races and species —
has not been finally settled. By some it is considered worthless
as evidence and is simply neglected. Others (e.g. Goodrich,

NATURAL SELECTION 189

1924, p. 117) hold ' that Darwin's views [on this subject] have
been brilliantly confirmed by the modern work on Mendelian
lines.'

There are really two questions involved here — (i) have
domesticated races and forms been produced by the means
which Darwin considered to be influential? and (ii) is there any
analogy between Artificial and Natural Selection ?

Darwin's opinions on this subject in the sixth edition of
' The Origin of Species ' and in ' Variation of Animals and
Plants under Domestication ' are in agreement — (a) domesti-
cated forms vary more than the wild parent forms ; (b) such
variation is largely due to ' changed condition of life ' and
' perhaps a great effect may be attributed to the increased
use or disuse of parts ' (id. 1905, vol. ii, pp. 349-50) ; (c) in
some cases the origin of domesticated breeds seems to have
been due to ' the intercrossing of aboriginally distinct species '
(I.e.), though he is definitely in doubt as to how far it is really
efficacious in producing new forms, and elsewhere (I.e. p. 94)
holds that the effect of crossing has been ' greatly exaggerated.'
It is quite apparent that he held that there was a rich source of
variation for selection to draw on. There is no evidence of
his having attempted to discover how much of the variation
referred to ' changed conditions ' is inherited and therefore the
basis of new fixed races and strains, though he admits (I.e.
p. 49) that ' the greater or less force of inheritance and rever-
sion determines whether variations shall endure.' He did not,
of course, distinguish between mutations and variation due to
factorial recombination. It is clear, however, that in spite of
this somewhat ill-defined knowledge of the material available,
he held that human selection, applied to the ever-present store
of variation, had been effective. Goodrich (I.e.), in stating
the case in modern terms, holds that ' one mutation after
another is isolated and bred from, and so almost any desired
form is obtained.'

This belief in the frequency of mutation is in radical con-
trast to the view that the efficacy of selection depends on the
progressive isolation of pre-existent hereditary material and
the continuous and carefully planned crossing of stocks of
known hereditary constitution, by which appropriate combina-
tions can be formed. The husbandman has been successful,
according to this view, because in stock-rearing like can be

i go THE VARIATION OF ANIMALS IN NATURE

mated with like, which accelerates race-formation, while the
selection of parents on ' performance ' (i.e. by the quality of
their offspring) also increases the effectiveness of selection.

We thus have two distinct and opposed views as to the
origin of domesticated races. According to the first they have
been produced mainly by the action of selection applied to a
plentiful stock of variations. According to the second they are
the result of appropriate crosses combined with pedigree breed-
ing and other devices. If the second view is correct, the success
of the breeder has been due to a procedure not fully repre-
sented in nature and the analogy between Artificial and Natural
Selection breaks down. If we disregard the question of muta-
tion-rate, as mutations are perhaps liable to turn up with equal
frequency in nature and under domestication, the issue can be
narrowed down to the question — is there as much opportunity
for crossing in nature as there is in the practice of stock-raising ?
If the numerous crosses made by man are the source of the
fresh steps in the development of domesticated breeds, and if
there is nothing comparable in nature, we think the analogy
must break down. The very great diversity of the means by
which isolation is established in nature between subspecies
and species inevitably suggests that the chances of factorial
recombination must be limited. It would seem a priori that
there could be no comparison between the amount of crossing
practised by man and that which occurs between natural
groups. Nevertheless some of the data in Chapter IV show
clearly that a large number of wild forms are highly polymor-
phic, and that the polymorphism is due to genetical causes.
We very frequently find subspecies and species that exhibit
various combinations of a common stock of characters, and
even among animals with a limited range, sedentary habits and
poor means of dispersal (such as land snails), there are numerous
instances of acute polymorphism. Nevertheless we do not
suggest that this polymorphism in any way approaches the
mixture of genotypes produced in domesticated forms. We
feel that some concrete measure of the difference is desirable
before this question is finally disposed of. However, the
critical point in this train of reasoning is that those who seek
to destroy the force of Darwin's analogy do not say that
selection is powerless. What they assert is that there is
more variation for it to work on among domesticated forms,

NATURAL SELECTION

l 9*

and that there are more opportunities for the rapid achieve-
ment of results (e.g. by pedigree breeding). If this is true, the
processes of Artificial and Natural Selection differ rather in
the relative abundance of their material and the means for
rapidly producing and stabilising new combinations than in
any more fundamental difference. Though we may admit
that much polymorphism occurs in nature, there is nothing
equivalent to the judicious utilisation of suitable crosses
coupled with the isolation
of desirable combina-
tions, when once estab-
lished. It seems then
that the analogy does on
examination become di-
vested of much of its
original force. If it is
argued that selection is
nevertheless the trans-
forming agency, it is only
reasonable to admit this,
but it is a selection ap-
plied in circumstances
that can scarcely be ever
realised in nature.

(b) Experimental selec-
tion. — Since Johannsen's
classical ' pure-line ' ex-
periments several at-
tempts have been made
to modify inbred stock
by selection. Results

similar to those obtained by Johannsen have been obtained
by Ewing (191 6), Jennings (1910), Ackert (1916), Lashley
(1916), and Zeleny and Mattoon (191 5). In these experi-
ments selection shifted the mean of a given character
to some extent and was subsequently ineffective. More
definite progressive modification was obtained by Banta
(1921), Jennings (191 6), and Castle (1919). It is as well,
however, to remember that the ' residual heredity ' (*.*. the
amount of variation that a strain heterozygous for several
characters is capable of manifesting) of one stock may be more

Fig. 2i. — Individuals of two different
Clones of Hydra, kept under similar
Conditions.

(From Lashley, 19 16.)

192 THE VARIATION OF ANIMALS IN NATURE

extensive than that of another, and that more time may be
required to exhaust it. Selection may be carried on success-
fully over a certain number of generations and then stopped
before improvement has ended. All that we are entitled to
infer from this is that selection has been successful up to a point.
We are not entitled to assume that it will continue to be so.
Castle (I.e.) considered that the extensive changes in pattern
which he produced in rats were due to the effects of selection
on the ' residual heredity ' and ' not to any change in the gene
for the hooded character.' That this interpretation is correct
is shown by the result of back-crossing both the selected types
to unselected ' selfs.' But even so, the modification produced
was very extensive, whatever the underlying cause of variation
may have been. Even if selection had ceased eventually to
be effective (' the variability of the stock had not been dimin-
ished during twenty (selected) generations'), the amount of
change wrought by it was very large, and it seems quite irrele-
vant whether it was due to a change in the hooded gene or to
residual heredity. It should also be noticed that in this case
selection brought about substantial results without any fresh
stock being introduced.

The negative results cited certainly show that the initial
variability of a stock may be easily exhausted and its capacity
for improvement by selection may be very limited, unless
reinforced by new gene mutations. But it is equally clear
that in other heterozygous stocks there is a large opportunity
for selective modification. This conclusion shows that the
effect of selection is entirely a question of the initial variability
of a stock and its subsequent mutations, and that Darwin's
general assumption of unlimited variability is scarcely justified.
It also points our way to the really crucial question — viz. how
frequent in nature are species which are heterozygous for many
characters ? As we saw in Chapter II (p. 26), we are still far
from being able to give an answer.

II. Direct Evidence for Natural Selection. 1 — The inci-
dence of death-rates in nature. — The facts and arguments dealt with
in the preceding section do not, of course, cast any light on what
is, after all, the most important question — viz. Is there a selective
process in nature ? As we have already pointed out, for Darwin

1 In the present chapter we use the term ' adaptation ' in a comprehensive
sense. In Chapter IX it is subjected to more detailed analysis.

NATURAL SELECTION 193

himself, Natural Selection appeared as an inevitable conse-
quence of certain satisfactorily established phenomena, viz.
numerical multiplication, competition, etc. He did not pro-
duce evidence for the actual occurrence of a differential
death-rate.

Pearl (1930) has set out concisely the requirements of a
proof that Natural Selection has altered a race. These are :

(a) Proof of somatic difference between survivors and

eliminated.

(b) Proof of genetic differences between survivors and

eliminated.

(c) Proof of effective time of elimination.

(d) Proof of the somatic alteration of the race.

(e) Proof of the genetic alteration of the race.

(c) implies that selection must occur before reproduction
is complete.

As will be seen from the examination of the direct evidence
(pp. 196-212), most of the investigators have concerned them-
selves with (a) only.

Before considering the evidence that a selective process is
or is not actually at work, certain general considerations as to
the death-rates of animals in nature may be brought forward.
Thompson and Parker (1928) in their study oiPyrausta ?iubilalis,
the European Cornborer, find that at least 90 per cent, of the
young larvae are killed off before any predators or parasites
have begun their attack. According to these authors, ' more
individuals disappear because of their highly restricted adaptive
powers than through all the other controlling factors taken
together.' The young larvae are extremely delicate. If they
fall to the ground or into a drop of water, or if they emerge
when the food-plant is too hard, they are likely to die. A
slight injury or deprivation of food for a short period causes
a high mortality. In a rapidly fluctuating environment many
larvae, even though on the whole better adapted than their
neighbours, must succumb without a chance of justifying
themselves.

Salt recently (1 931), in a very careful study of the Wheat-
stem Sawfly (Cephas pallipes) , found that a part only of the larval
mortality accounted for 89 per cent, of the pre-adult individuals.
Thorpe (1930a) found in the Pine-shoot Moth (Rhyacionia

i 9 4 THE VARIATION OF ANIMALS IN NATURE

buoliand) that the insect parasites account for about 60 per cent,
of the larvae. In all insects death from unfavourable climatic
conditions is also very frequent in the early stages, so far as the
facts have been recorded (Uvarov, 1931). Kirkpatrick (1923)
has provided an elaborate account of the Egyptian cotton-
seed bug (Oxycarenus hyalinipennis) . At the end of the breeding
season this insect may be present at the rate of 7-12 millions
per acre, while at the end of the winter not more than 100,000
per acre are left. During the whole of his work no parasitic
or predacious enemies were discovered, and all effective control
appears to result from the operation of normal weather con-
ditions. Sunlight kills some of the eggs, and some of the young
nymphs die, possibly through lack of moisture or failure to
penetrate the boll quickly enough. Heavy rainfalls and the
harvesting of the bolls account for many more. During the
winter the death-rate from drought must be enormous, especi-
ally as many of the bugs leave their hibernacula on warm days
and probably fail to regain suitable quarters when the weather
alters. Yet, in spite of its rather imperfect adaptation, this
species can maintain itself in great abundance.

Russell (1932) has summarised some of the data as to the
fluctuations of certain marine organisms. The populations of
bottom-living Mollusca seem to undergo extreme variation,
and in certain cases it is thought that this is due to variations
in the course of currents by which the larvae are carried
passively. When the larvae settle down, only those survive
which happen to have drifted over areas of suitable bottom.
The very large mortality amongst those which have been
carried to unsuitable areas must be largely random. It would,
in fact, appear to be a general rule that the more directly
dependent an organism is on its environment, the larger will
be the element of chance in the death-rate.

In many mammals, as Elton's well-known studies have
shown, the decimation of the population is a periodic pheno-
menon. A period during which the death-rate is relatively
low culminates in an enormous increase in numbers, leading
in turn to a catastrophic reduction, often as a result of an
epidemic. Many examples are given by Elton in his book,
1 Animal Ecology and Evolution ' (1930, pp. 19-23).

It has been argued (e.g. Muir, 1931) that because 90 per cent,
of the individuals perish before reaching maturity, a selective

NATURAL SELECTION 195

process acting purely on the adult can have little effect. It
is true that selection amongst larvae (so far as this heavy
death-rate is not purely random) will tend to produce unex-
pected results in the adult stage, the most numerous types of
the latter being chosen for the characters they bore as larvae
and not for their actual facies. But this will not avert the
effect of selection amongst the adults (see Fisher, 1930, p. 134).
If there is a differential death-rate amongst the adults, a certain
genotype will be favoured, and this form will occur in an
increased proportion amongst the larvae. As long as the
incidence of larval mortality does not actually tell against the
adult character, then, on the theory of chances, the survivors
of the larval holocaust will still show on the average the same
increased proportion of the adult genotype.

The real conclusions that should be drawn from such
studies as those we have mentioned appear to be the following :

a. Most animals — all those with a high rate of repro-
duction — have a very high mortality, especially in
the early stages.

/?. This mortality often appears to be random : but the
appearance may be deceptive, and certainly a random
death-rate cannot as yet be directly verified.

y. However large the random death-rate may be, it cannot
nullify the effect of any selective death-rate, even if
very much smaller. This is at least true when two
populations in competition are both of considerable
size, and is necessarily a result of the random nature of
the main death-rate — i.e. the proportions of each
form can be influenced only by death-rates which
are not random. Actually, if one population were
very small, as when a rare mutant competes with the
dominant type of a species, a large number of trials
might be necessary before the inherent impartiality
of the random process was actually observed — i.e. the
mutation might have to occur often enough for the
mutant individuals in the aggregate to form a fairly
large population.

8. The only satisfactory way to investigate whether death-
rates are selective or not is to study in nature the
actual death-rates of competing forms, whether
species, varieties or mutants.

196 THE VARIATION OF ANIMALS IN NATURE

The view has been expressed that ' it is impossible to con-
ceive that the detailed action of Natural Selection could ever
be completely within human knowledge ' (Fisher, 1930, p. 47).
The process might nevertheless be brought sufficiently within
human observation to provide direct visual proof. Obviously
the conditions for observing an act of adaptive transformation
are very rarely available for a human observer. The coinci-
dence of several propitious circumstances, that is rarely realised,
is required : but it will be seen that the opportunity is not so
rare as Fisher suggests, and that more efforts should be made
by field workers to locate likely situations and bring them to
the notice of those able to carry out the necessary observations.

Many observations and experiments have been made
on animals living freely or in captivity which are claimed to
prove either the elimination of certain types of variant and the
survival of others, or the absence of selective elimination.
These studies are not of the same kind. 1 The problems they
set out to solve and the procedure adopted are not of the same
order, and it is necessary to show at the offset exactly what
they aim at demonstrating, before proceeding to detail the
results obtained and the criticisms that may be made as to
their interpretation.

(1) In a certain number of cases the observations (with or

without control experiments) relate to animals living
freely and exposed to a known or reasonably assumed
cause of death (Weldon, 1899 ; Harrison, 1920 ;
Trueman, 191 6 ; Haviland and Pitt, 1919 ; Jameson,
1898 ; Kane, 1896).

(2) In six cases the observations relate to animals either

subjected to laboratory or other experimental condi-
tions or experimentally exposed to natural enemies, the
cause of death being known or assumed (di Cesnola,
1904, and Beljajeff, 1927; Poulton and Saunders,
1899, and Moss, 1933; Boettger, 1931 ; Lutz,
1915; Davenport, 1908; Pearl, 191 1).

(3) In two cases the animals observed were living under

natural conditions, but the cause of death was un-
known (Crampton, 1904 ; Thompson, Bell and
Pearson, 191 1).

1 Studies comparable with some included in i— 19 below are also to be found in
our section dealing with Protective Resemblance and Mimicry (pp. 232-265).

NATURAL SELECTION 197

(4) in one case the observations involve merely a com-

parison between the variation of the natural popula-
tion (a) over a single season, and (b) over a period
of years (Kellogg and Bell, 1904).

(5) One case related to the survival or death of animals

brought into laboratory conditions after a pre-
liminary exposure to a generalised eliminating
factor, though the actual causes of death were not
controlled (Bumpus, 1899).

(6) In three cases a special procedure was adopted, viz. that

of comparing the variation of juvenile stages with
adult (Weldon, 1901, 1904 ; di Cesnola, 1907).

(1) Weldon (i8gg).

These experiments and observations are so well known that
they do not need to be explained in detail. Series of measure-
ments made by Weldon and his collaborator Thompson over
the years 1 893-1 898 on the crab Carcinus maenas in Plymouth
Sound showed that the mean frontal width of the carapace
(M.F.W. ) (expressed as a proportion of the length of the carapace
taken as = 1000) decreased in crabs of a similar carapace
length. Weldon attributed this to the elimination of crabs of
high M.F.W. through the action of silt in the gill-chamber
clogging the gills. He stated that the amount of silt in the
Sound had increased owing to the building of a breakwater
which prevented the escape of the detritus from china-clay
workings which was being washed into the Sound. Experi-
mental controls showed the following confirmatory results :
(i) Crabs were placed in vessels containing clay silt in suspen-
sion. Those that died had M.F.W. larger than that of the
survivors, (ii) Small crabs were collected on the shore and
kept in clean water. Some died — (?) from the effect of putrid
food. After the first moult the survivors were killed and
measured, and it was found that they were broader than wild
crabs of a similar size — which, on Weldon's hypothesis, is what
one would expect in the silt-free conditions.

This work has been criticised by Cunningham (1928,
summary), Vernon (1903), Pearl (191 7), and Robson (1928).
The criticism falls into three categories : (a) as to the external
conditions ; (b) as to Weldon's assumption concerning the
relation between M.F.W. and filtration of the gill-chamber ;

1 98 THE VARIATION OF ANIMALS IN NATURE

and (c) as to the interpretation of the measurements. It is
necessary to make it clear that there is a definite differential
(heterogonic) growth-effect involved in the relation of M.F.W.
to carapace length. M.F.W. decreases in proportion to the total
length of the carapace.

(a) (i) Weldon did not show that the amount of silt had

increased in the period under consideration ; he
merely assumed that it had.
(ii) He did not take into consideration the exceptional
climatic conditions of 1893, which may have had
a marked effect on growth and in consequence on
measurements correlated with absolute size.

(b) Weldon assumed that M.F.W. would affect the filtration
of water in the gill-chamber, the narrower frontal breadth
forming a better filter. It seems very strange that the actual
entrance of the gill-chamber itself was not measured. Weldon
makes no attempt to show that there is any relation between
the two dimensions. As Cunningham (I.e. p. 193) points out,
' the exclusion of particles of silt must depend on the absolute
size of the entrance to the gill-chamber, not on the proportion
which that size bears to the body-length.'

(c) (i) Vernon (I.e. p. 340) objects that to take length for

age is a dangerous procedure. Silt may retard
growth. 12-mm. crabs of 1898 may have a
narrower M.F.W. because they are older than those
of 1893. This objection assumes, of course, that
M.F.W. may be determined by age and not by
size.
(ii) A more serious objection is that of Cunningham
(I.e. p. 192). He points out that 'it would
follow from Weldon's argument that the pro-
portional frontal breadths which were fatal to
small crabs of a given carapace length, permitted
the survival of others which were only y mm.
shorter.' Thus, if in 1895 the M.F.W. of size-class
14-5 mm. has been reduced by selection from
762-00 to 754-45 in 1895, how is it that we find
all those less than 13-7 mm. size surviving in
which the M.F.W. is over 762-00 ?

NATURAL SELECTION 199

(iii) There are no control measurements given of the
wild population in silt-free conditions from which
one could see if the changes do or do not occur
there.

(iv) The control experiments are criticised by Cunning-
ham (I.e. p. 196). As regards the first, he points
out that it is not stated that the survivors were,
on the average, of the same carapace length as
the dead. As regards the second series, in which
the M.F.W. under silt-free conditions was larger
than in the wild population, it is rather difficult
to give the facts in a condensed form, because
there was a preliminary mortality due (?) to the
presence of putrescence in the water, and the
shells of the survivors at the first moult were less
than those of wild forms, which Weldon put
down to the fact that those of greater M.F.W. were
selectively eliminated. Cunningham makes it
amply clear : (a) that necessary comparisons were
not made, and (b) that Weldon omitted to con-
sider the effect of food-supply and temperature
on the size of the experimental animals.

On the whole the objections raised as to Weldon's results
are so serious that the latter cannot be accepted as good
evidence for the efficacy of selection.

(2) Harrison (ig2o).

In the Cleveland district of Yorkshire a colony of the moth
Oporabia autumnata was originally broken into two parts, one
ultimately inhabiting a coniferous wood, the other a birch
wood. The colour of the two colonies was found to differ, those
moths living in the birch wood being paler (no statistics given).
Harrison attributes the difference to the elimination by noc-
turnal birds and bats of the pale forms in the coniferous wood,
on the assumption that these moths are more conspicuous. His
proof is that of 15 pairs of wings (remains of moths attacked
by enemies [?] ) found on the ground in that wood the majority
(numbers not given) are pale, though in the total population
the dark forms outnumber the pale in the ratio of 25 : 1. He
states that owls, nightjars and bats are numerous in the pine

200 THE VARIATION OF ANIMALS IN NATURE

wood, while in the birch wood few, if any, birds occur, as the
wood is not well grown enough to afford cover.

This case is very summarily expressed. The number of
likely enemies in the two woods is not discussed in detail.
It is quite uncertain how the individuals whose remains were
found actually met their fate — i.e. whether they were killed by
birds or bats. There is no statement as to how many of the
15 pairs of wings were pale (? 14 : 1 or 8 : 7). Nevertheless if, as
he says, the population of the pine wood is preponderatingly
dark, a ' majority ' of light eliminated forms is significant.

On its surface value this case might pass as definite evidence
for selective elimination. It seems to us to be open to two
main criticisms : ( 1 ) the lack of definitely expressed evidence
as to the frequency of enemies in the two woods, and (2) the
small number of observations and the failure to state what is
meant by a ' majority,' particularly in regard to the frequency
of each variety.

(3) Trueman {1916) : alleged selection of ' banded shells of
Cepea.'

It has long been known that birds feed on the common
snails C. hortensis and C. nemoralis, taking them to stones on
which the shells are broken in order that the bodies may be
extracted. Masses of shells are often found around these
c anvils,' and Woodruffe-Peacock (1909) suggested that it
might be possible to detect from the broken shells any selection
of a particular type, e.g. as between the banded and unbanded
types. Peacock's observations did not include a survey of the
percentage occurrence of the various types in the local popula-
tion from which the victims were taken, and are therefore
useless.

Trueman compared his shells from ' anvils ' with a standard
collection, not a local one, and his conclusions are also value-
less, inasmuch as the local percentage of banded and unbanded
forms varies very much from district to district. He also fails
to give the actual numbers of shells obtained, expressing his
results as percentages, of which the following is the essential
result :

' Standard ' collection Found on * anvils '

Unbanded 25 per cent. 38 per cent.

5-banded 42 ,, ,, 23

53

NATURAL SELECTION 201

He claims that this shows preferential selection of the
unhanded.

Expressed in this form the figures are worthless from the
statistical point of view. His results have been criticised
(on the lines already suggested) by the under-mentioned
authors (4).

(4) Haviland and Pitt {1 gig).

These writers, in addition to a criticism of Trueman's work,
supply the results of their own experiments, etc.

(i) Banded and unhanded snails were tethered to pegs,
and the selection by birds was observed. It was
found that both types were taken. More of the
banded were killed, but the numbers were small.

(ii) Collections from ' anvils ' were compared with the
local population, and it was found that there was no
preference as between the banded and unhanded.

(hi) A captive Thrush was kept under observation and,
when offered the two types, exercised no discrimina-
tion.

(i) and (hi) are of little value as evidence. The com-
parison of a large series of shells from ' anvils ' with the local
population is clearly indicative of no selection.

(5) Jameson {i8g8) : colour of Mus musculus on sandhills.

Jameson observed the coat-colour of mice on sandhills on
an island in Dublin Bay. There was evidence that the island
was about a hundred years old. The mean colour of the mice
was lighter than that of the typical form. Thirty-six specimens
were examined : of these 5 were as dark as the typical form ;
5 were intermediate ; the remaining 26 were distinctly more
pallid than the typical form. Jameson states that the island
is infested by short-eared owls and hawks, and that these
' most readily capture those mice which contrast most strongly
with the sand and arid vegetation.' He does not say that
this was actually observed, and there is no statement as to what
types were actually seen captured. As there is no direct
evidence that one type is captured in preference to another,
this case cannot rank as one of direct evidence.

202 THE VARIATION OF ANIMALS IN NATURE

(6) Kane {i8g6) : melanic forms of Camptogramma bilineata.

Kane found that by 1892 the melanic form, var. isolata,
of this species had entirely superseded the lighter-coloured
typical form on Dursey Island, Ballinskelligs Bay, co. Kerry.
He points out that in this area the cliffs and islands are of a
dark slate formation. They are ' haunted by Rock Pipits,
Wheatears, Bats and small Gulls (all insectivores).' In 1893-
1894 there was a great destruction of the Silene on which the
moth lived, and a potential increase in the intensity of destruc-
tion — so much indeed that in this period the species virtually
became extinct. He thinks that the dark form under intense
competition was favoured by its colour (as against the dark
background). Some additional evidence is supplied from a
study of the distribution of dark forms on heather and peat,
and more especially of the prevalence of light forms on the
pale grey limestones of co. Clare. Other species tend to show
a parallel variation in relation to habitat. Some of the Dursey
Island melanics were shown to be of a fixed heredity (Kane,
I.e. 1897, p. 44).

There is no actual evidence as to the discrimination by
the alleged enemies. The author attempts to get round the
traditional explanation that the occurrence of melanism is
correlated with rainfall. We require far more evidence as to
the selective value of the dark colour and of the discrimination
by the alleged enemies.

(7) di Cesnola {1904) and Beljajeff {1927) : experiments with
Mantis religiosa.

di Cesnola conducted his experiment as follows :

20 green Mantis were tethered to green plants.
25 „ » » » » brown „

20 brown „ „ „ „ „ „

45 » » " » » green ,,

The insects were left exposed for 1 7 days. At the end of that
period the 40 ' harmonising ' insects had all survived. Of
the 25 ' green on brown ' all had been killed (20 certainly by
birds) ; of the 45 ' brown on green ' 10 only were left. All
the rest were killed by birds.

He concludes that the ' concealing ' background does
discriminate one type from the other.

NATURAL SELECTION 203

The results, if we allow for the rather low numbers, demon-
strate the value of the harmonising colours. As Robson
{I.e. p. 213) suggested, the selective value of the colour would
only be established for animals living freely if it could be
shown that it was accompanied by the habit of choosing an
appropriate background. Further, the contrast provided in
the experiment would be sharper than that usually found in
nature. Beljajeff (1927) repeated these experiments, using
brown, yellow and green forms of Mantis. On a brown back-
ground, out of 20 of each form, 11 green, 12 yellow and
4 brown were eaten in a fortnight. In a second experiment
some crows in 24 hours ate n green, 12 yellow and 12 brown
from the same background.

(8) Poulton and Saunders {i8gg) : differential elimination of the
pupa of Vanessa urticae in different situations.

The authors exposed the pupae on backgrounds of various
kinds (tree-trunks, fences, etc.) at four stations : two in Switzer-
land, one at Oxford and one in the Isle of Wight. The
mortality was very low in the Swiss loci, which the authors
attribute to the lack of insectivorous birds. At the other loci,
where the pups were suspended to a background which con-
cealed them (from the human observer's eye), there was a lower
mortality and more of the pupae emerged. Thus at St. Helens
90 were taken by birds (?) and only 8 emerged among those
suspended on fences, whereas on backgrounds which served to
conceal better, destruction and emergence were more balanced.
The numbers in the Oxford experiment were low and of little
value.

The experiments tend to show elimination of pupae if
they are placed in conspicuous situations. Experiments in-
volving the concealing value of colour led to very ambiguous
results and the authors ' cannot make any statement ' as to
their value. Moss (1933) came to a similar conclusion after
experiments with pupae of Pieris brassicae.

(9) Boettger (1931) : observations on the selection of Gepea by
captive birds.

The author made experiments on the selection and rejection
of various colour- and band-types of C. nemoralis and Arianta
arbustorum by captive pheasants in the Berlin Zoological

2o 4 THE VARIATION OF ANIMALS IN NATURE

Gardens. The snails were put into the enclosures in which
the pheasants were kept in such a way that all four types were
accessible to the birds.

There were six experiments with six different species of
birds (including one hybrid) .

In experiments I— III and VI (Phasianus colchicus colchicus,
P. c. torquatus, Crossoptilon mantchuricum and Lophophorus impejanus)
no selection of any type was observed. In experiment IV
(hybrid of Chrysolophus pictus and amherstiae) the dark shells were
taken and the light and banded left. The author does not
state what background these forms were on, except to say that
the dark forms were difficult for the human observer to see, and
that he thought the birds revolted from the light-coloured
snails. In experiment V (Gennaeus nyctimerus) the dark forms
and red and yellow unbanded forms were taken and the banded
left alone. He says that on the pale greenish-yellow grass in
the enclosure the banded snails were inconspicuous to the
human eye.

The value of these experiments is very problematical.
The author admits that the captive birds are accustomed to
being fed by the public. He does not mention how many
experimental snails were used. In the two cases in which he
claims that selection of certain types was observed, he says
(experiment IV) one kind was taken ' zuerst ausnahmlos ' ;
in his second, that the selected types were ' grossenteils
gefressen.'

(10) Lutz {191 5) : experimental observations on Drosophila.

This author studied the effect of starvation on D. ampelophila
in relation to the duration of the embryonic period and on
two structural characters (length of first posterior cell in wing
and breadth of wing).

Two methods were adopted : (i) the comparison of the
mean of the characters of the survivors and eliminated ; and
(ii) the correlation of a given character and the ability to
survive.

(a) There was a negative correlation between the length
of adult life and the duration of the embryonic
period. Those with shortest embryonic period lived
the longest.

NATURAL SELECTION 205

(b) There was no significant correlation between the

ability to withstand starvation and the length of
the embryonic period.

(c) There was a selective death-rate in respect of the length

of the embryonic period in the fed animals, but none
in the starved ones.

(d) As regards the structural characters, there was a

positive correlation between the length of the first
posterior cell and the breadth of the wing and
ability to survive. In two cases (breadth of wing
in 6* ; length of cell in $) the correlation is statis-
tically significant ; in the other two cases it is barely
significant.

(e) As far as the difference of the means (in size) was

concerned (comparison of survivors and eliminated),
it is clear that larger flies were better able to survive
starvation.

Lutz notes ' discordant results as regards [reduction of]
variability.' The results are, on the whole, unsatisfactory —
e.g. in the difference between male and female. Also the
males which withstood starvation were distinctly more variable
as regards egg-larval period, but less so in the structural
characters. In the female the differences were insignificant.

(11) Pearl (iQii) : observations on conspicuousness in fowls.

Observations were made on a number of ' self-coloured '
and ' barred ' fowls on a poultry farm in which they were
exposed to the attacks of various carnivorous enemies. Out
of 3,007 ' barred ' fowls 290 were killed, and out of 336 ' self-
coloured ' birds 35 were killed (9-6 per cent, and 10-7 per
cent.). Only one year's results were obtained. Pearl seems
to have made careful observations as to how the eliminated
were killed. Photographs show that, as far as the human
observer is concerned, the ' barred ' birds are more incon-
spicuous than the ' self-coloured.' He concludes that ' the
relative conspicuousness of the barred colour-pattern afforded
its possessors no great or striking protection against elimination
by natural enemies during a period of seven months, during
which they were exposed to the attacks of predators.'

206 THE VARIATION OF ANIMALS IN NATURE

(12) Davenport (1908) : attacks on poultry by crows.

The author observed the attacks by crows on 300 chicks
in a poultry run. Of 300 chicks 24 were killed. The
constitution of the original 300 was as follows :

40 per cent, white.
40 „ „ black.
20 „ „ more or less like the Jungle Fowl ( c pencilled ').

If there had been no selection, the expectation would be
that of the 24 killed, 9-6 would be white, 9-6 black and 4-8
' pencilled.'

Actually of the killed, 10 were white, 13 black, and 1 was
' grey and buff.' No pencilled birds were killed.

Davenport assumes that the inconspicuous ' pencilled '
type is preserved by its colour.

We think that the extremely low total of 24 birds is quite
inadequate as a basis of estimating the effects of selection.
It seems to us extremely problematical whether the ' pencilled '
birds are in fact less conspicuous than the white and black.

(13) Crampton (1904) : death-rate of the pupa of a Saturniid Moth.

Crampton, observing that numerous cocoons of Philosamia
cynthia contained dead individuals, attempted to discover the
causes of pupal and imaginal elimination. He obtained from
trees 1,090 cocoons, of which 55 had not pupated and 93 had
left the pupal case. Of the remaining 942 pupae, 623 had
pupated, but were dead. Only 319 'selected' individuals
were alive.

Equal numbers of dead and survivors were measured for
length of antenna and various proportions of the ' bust.' The
survivors were kept till the metamorphosis was over.

(i) Pupal stage. — Estimates were made for ' type ' (i.e.
the average sizes and proportions) and ' variability '
in 8 characters. When the measurements of dead
and survivors (6*) were statistically compared, it
was found that the differences suggested that selection
must have occurred in 5/8 characters, that it was
probable in 2/8 and possible in 1/8. Where the
variability was compared, the survivors were less
variable in 1 /8 cases, possibly so in another, and not

NATURAL SELECTION 207

less variable in 6/8. The reduction of variability is,
of course, assumed to show that ' selection ' has been
operative. Thus there is definite selection for ' type,'
but very little for ' variability.' In the females there
is selection shown both for ' type ' and c variability.'
(ii) Imaginal stage. — Ten characters were examined, (a)
o* : Selection for ' type ' occurs probably in only 3
characters, possibly in 2, and is not shown in the
remaining 5. Selection for ' variability ' is certain
in 1 character, probable in 2, possible in 4, and
absent in 3. (b) ? : In the females selection for
' type ' was certain in 4, probable in 2, possible
in 1, and absent in 3. Selection for ' variability ' is
reversed (survivors are more variable) in 7, possible
in 1, and absent in 2.

Crampton points out that the actual characters cannot
possibly be of service. He thinks that the basis of selection
is ' the proper co-ordination of functional and structural
elements.' If we understand him correctly, he means that
the deviations eliminated are indices of a structural noncon-
formity and lack of developmental harmony. This is some-
what vague : but the fact remains that survivors and elimi-
nated are statistically different (significantly). There are
certain ambiguities which require explanation — e.g. why there
is selection for variability in the females and not in the males
at the pupal stage, and why there is less selection in males at
the imaginal stage than at the pupal stage.

As this work was conducted on rigorous statistical principles
and the numbers were fairly high, it is to be accepted as proving
that the survivors at each stage differed structurally from the
eliminated. The failure to find a basis for selection in the
characters studied is not necessarily a limitation.

(14) Thompson, Bell and Pearson {igi 1) : variation and correlation
in Vespa vulgaris.

These authors undertook a study of the means, variation
and correlation of certain wing-characters (dimensions of wings
and of individual cells) in the general populations of autumn
and spring queens of the Common Wasp. Their object was
to study the influence of hibernation on these characters.

208 THE VARIATION OF ANIMALS IN NATURE

They found (p. 6) that certain linear measurements of the
autumn queens are on the average 10-12 per cent, and certain
indices 18-22 per cent, more variable than in the spring
queens. They also found that there is a slightly higher
correlation between the parts of the wing in the spring, as
opposed to the autumn queens. According to the principle
that selection increases correlation, they argue that ' the only
reasonable assumption to make is that there has been a direct
selection of correlation as well as selection round a type ' (p. 4).

We assume that the authors infer that these differences in
variability and correlation were due to some selective agency
at work during the winter. What that agency was they do
not discuss. They say (p. 6) that the ' fitness for survival of
the queen during the period in which she is seeking winter
quarters, hibernating and starting to form a new colony,
seems to depend more considerably on the ratio of the parts
of the wing than on their absolute size.' The only further
light cast on this matter is the authors' analogy (I.e.) between
the wing of an insect and the parts of an aeroplane, reliability
in the latter being due to minute details comparable to those
of the insect wing !

This case is similar to that of Philosamia (p. 206), and
we should rather expect that the cell-characters of the wing
were correlated with some physiological character determining
survival rather than that it was of actual utility. We are
somewhat doubtful as to the value of inferences based merely
on the reduction of variability. To assign the latter to selection
on purely theoretical grounds seems to us dangerous, and we
think other causes reducing variability might be operative.
There is no proof that the characters ' selected ' are heritable.

(15) Kellogg and Bell [1904) : observations on the variation of
various species of insects.

The authors point out that the variation in various insects,
in spite of exposure for a season to all kinds of rigorous external
factors, is just as great as at the beginning of the season, and
none of the types of variation is eliminated. This is very
well seen in the ladybird (Hippodamia convergens) and in the
Honey Bee (Apis mellifera).

' Determinate variation ' (i.e. statistical change in the
constitution of the population) is seen in the pattern of the

NATURAL SELECTION 209

elytra of the beetle Diabrotica soror over the period 1895- 1902.
The difference consisted in the dominance in 1 901 -1902 of a
modal condition which was not dominant in 1895.

Beyond stating that it is not likely that the change in
position of the spots or the elytra would serve as a basis for
selection, the authors produce no evidence that the change
is not due to selection.

(16) Bumpus (i8gg) : alleged selective elimination in Passer
domesticus.

' After a severe storm of snow, rain and sleet a number of
English Sparrows were brought to the Anatomical Laboratory
of Brown University. Seventy-two of these birds revived :
sixty-four perished.' It was the purpose of Bumpus's study
to show that the birds which perished did not die from accident
but because they were physically disqualified, and that the
survivors lived because they possessed ' certain physical
characters ' which enabled them to withstand a particular
phase of selective elimination. He measured 9 characters
(e.g. length, weight, alar extent, etc.) of the dead and survivors.
He divided his specimens according as they were adult or young
and male or female. He found that there were differences
in some characters as between survivors and eliminated and
not in others, and he assumed (p. 213) that there were funda-
mental differences between the dead and the survivors. As
the numbers in each group thus discriminated are low (the
total which died was only 64, of which 24 were adult $ and
12 were young o*)> and as he compared the averages of the
various groups, it will strike the modern statistical biologist
that his conclusion is premature. These observations, sug-
gesting a selective elimination, have been widely cited as
proving the general occurrence of such elimination.

Harris (191 1), however, on the very full data published
by Bumpus, produced the necessary statistical constants
(standard deviation, etc.) and applied the usual tests for
significance. His treatment of the subject is rather peculiar.
He admitted that, by applying the usual statistical tests,
differences of a statistical value varying from ' significant '
to ' possibly significant ' were actually to be obtained from
Bumpus's figures for some (but by no means all) of the charac-
ters. Yet he concludes that, ' though the cautious biometrician

210 THE VARIATION OF ANIMALS IN NATURE

would hesitate to allow that Bumpus's case was proved, the
action of selection is likely.' He stresses the fact that the
number of individual variates is low, and is clearly divided
between an adherence to a rigid statistical principle (which,
when applied to the data, gives ' significant ' differences in
some characters) and an apprehension that, on account of
the paucity of data, the statistical principle may be fallacious.
Incidentally we may note that we have applied the current
tests to Bumpus's figures as a check on Harris's procedure
and find that his conclusions as to ' significance ' are
valid.

The matter might be left to remain in this rather unsatis-
factory condition, with the admission that, statistically at
least, Bumpus's conclusions are sound. But there is, however,
a further question to be decided, which we think invalidates
these observations at their source. As one of us has pointed
out (Robson, I.e. p. 214), the cause of the death of the elimi-
nated is uncertain. What Bumpus did was to compare the
birds which recovered with those which died after being
blown down. All the birds were, it is admitted, blown down
by the gale ; but those which did not recover might have
died from various causes {e.g. from dashing in their fall against
a stone or a tree, from exposure and starvation, from the
immediate effects of strain and exhaustion). In short, the
birds might, we agree, be all blown down on account of some
structural deficiency, but their survival or death after failure
to sustain themselves in the gale might very easily be determined
by quite a distinct set of causes. In short, we are plainly
dealing with two distinct phenomena- — the fact of being blown
down on the one hand, and the multiple causes of death
connected with the subsequent experience of those who were
blown down. It might be urged that the acid test is really
between death and survival — that at all events we know there
were some significant differences between those which died and
those which survived. But in reply we must, obviously, ask
how any structural character (such as weight, wing spread,
etc.) which might determine whether a bird was blown down
or not, could determine whether a bird survived or died after
it was blown down — a result which might be determined by such
purely accidental causes as whether it hit a branch or stone in
its fall, or whether it was able to withstand exposure and shock.

NATURAL SELECTION 211

Finally, if all the birds had been left out of doors, probably
all would have died, and the real selective agency was human
interference (i.e. the bringing of the birds into the laboratory).
It is most unfortunate that Bumpus did not investigate the
actual cause of death in each case, and for this reason (coupled,
of course, with the actual paucity of individual variates) we
hold that the quite clearly established ' significant ' differences
are suspect.

( 1 7) Weldon (igoi) : comparison of earlier and later whorls of
the shell of Clausilia laminata.

A series of measurements of the earlier and later whorls
of the shell (made on sections) shows that ' the mean
spiral of the young generation is sensibly identical with
that of the parental generation [earlier as opposed to later
whorls] and is not altered by any process of selective
destruction.'

As, however, the variability of younger shells is greater
than that of adults, it is inferred that there is ' periodic selec-
tion ' (reduction of variation at each generation). The fact
that the mean remains the same is held to be an indication
of the effect of selection.

We are not convinced that if a difference between
the early whorls and the later had been shown, it would
necessarily imply that the difference was due to selection as
Weldon suggests. It seems that any changes that might
have been found could have been due to environmental causes.
As for the reduction of variability in the adult stage, we
think that this might possibly have been due to greater
plasticity of the young, as well as to selection.

(18) Weldon (1904) : shells of Clausilia itala.

The same type of measurement was undertaken on the
shells of 100 young and 100 adult C. itala. No difference
between the young and adult shells was found. Weldon
suggests that this might be explained in two main ways : either
that (1) no selection was operating, or (2) the lack of selection
was due to the specimens having been collected in the spring.
If measured in the autumn differences might have been
shown (?).

212 THE VARIATION OF ANIMALS IN NATURE

(19) di Cesnola (1907) : comparison of earlier and later whorls
of the shell of Helix ( = Arianta) arbustorum.

The procedure was identical with that of the preceding
studies (17), (18). The characters of the young shells were
similar to those of the adult. ' The mean character does not
sensibly alter during growth, but is the same in young and
adult.' The same difference in the variability of young and
older shells was found as in Clausilia, and was held to prove the
occurrence of periodic selection.

The same criticism may be applied to this study as to (17).

We give in tabular form what we hope is a fair assessment
of the value of these studies.

(1)

Selection-
probable

Lutz (1915) ?
Crampton (1904)
Thompson, Bell and
Pearson (191 1)

(2)

Analogy with natural
process doubtful

di Cesnola (1904)
Poulton and

Saunders (1899)
Boettger (1932)
Bumpus (1899)
Beljajeff (1927)

(3)

Other explanations
possible

Weldon (1899)
Kane (1896)
Lutz (1915)
Weldon (1901)
di Cesnola (1907)

(4)
Procedure defective : or
numbers too low

Harrison (1920)
Trueman ( 1916)
Jameson (1898)
Kane (1896)
Boettger (1931)
Davenport (1908)
Kellogg and Bell

(1904)
Bumpus (1899)

(5)

No selection

found

Haviland and Pitt

(1919)
Pearl (191 1) ?

Weldon (1904)

(6)

Selective agency unknown
or doubtful

Harrison (1920)
Jameson (1898)
Kane (1896)
Crampton (1904)
Thompson, Bell and

Pearson (191 1)
Bumpus (1899)

It will be seen that on this analysis (which should be
checked by reference to the actual accounts) there is a little
evidence suggesting a significant difference between survivors
and eliminated. It must be admitted that any amount of
positive evidence, however slight, is of value. On the other

NATURAL SELECTION 213

hand, it is of the greatest importance that, in all the cases in
which selective elimination appears to be established, the
distinguishing features of the survivors arc not known to be
heritable.

Lastly, we think it desirable to give in a condensed form
some direct observations on the alteration of the composition
of natural populations. Sometimes, as in (4), a ' new ' character
appears to have spread ; but we do not really know that the
character is a novelty in the history of the species.

(1) Adlerz (1902a). The butterfly Polyommatus vigaureae
was very abundant in Sweden in 1896. A peculiar form of the
female (with blue spots on the light band of upper side of hind
wings) was common. In 1897 the species was not common.
The variety was relatively and absolutely rarer. In 1901 the
species was again very abundant and the variety made up
about half the individuals. Ford and Ford (1930) have found
that in Melitaea aurinia there is an increase of variation during
local numerical increase.

(2) Scudder (1889, p. 12 13). Pieris rapae, first introduced
at Quebec in i860, appeared in New York in 1868. A variety
with yellow wings (var. novangliae) first appeared in Canada in
1864. Later it was found also in the United States, where it
occurred about once in 500 specimens. It died out again by
1878. In Europe the variety is excessively rare, only one or
two doubtful specimens being on record.

(3) Probably the best instance of the appearance and
multiplication of a new variant is that of the melanic form
(doubledayaria) of Amphidasys betularia, the Peppered Moth.
The actual facts are too well known to require repetition here.
It is enough to remind the reader that (a) the melanic variety
first appeared near Manchester in 1850 and has in many
places in England now completely superseded the type form ;
(b) a similar course of events occurred on the Continent, though
beginning at a later date ; and (c) in the twenty-seven
years that have elapsed since the original study (summarised
by Doncaster, 1906) was made, the melanic forms (originally
largely restricted to the North and Midlands of England) are
now far more frequent in the South. An analogous north to
south invasion is found in France (Demaison, 1927, p. 295).
(d) Similar melanic forms occur in other genera in the same
areas, (e) We can find no evidence in contradiction of

214 THE VARIATION OF ANIMALS IN NATURE

Bateson's contention (19 13, p. 138) that between doubledayaria
and the typical form there are few if any intermediates.
Three explanations of this history are available.

(a) Protective value of the dark colour in industrial districts.

It has been suggested that the dark colour affords a pro-
tective resemblance (against birds) to smoke-darkened foliage,
etc., in the industrial districts in which it undoubtedly arose.
This has been answered by Bateson, who, reasonably enough,
points out (i) that doubledayaria is conspicuous anywhere
except on actually black materials, and (ii) that it occurs in
country districts between the towns. Bateson's criticisms
overlook the possibility that, if even 1 per cent, of the double-
dayaria were protected when on very sooty or dirty back-
grounds, it would give them an advantage. Furthermore,
Mr. A. W. McKenny Hughes informs us that Bateson very
much minimises the concealing effect of the dusky colour,
which Mr. Hughes asserts is marked. It should be noted that
Kane (supra, p. 202) claims that dark forms of moths are
protectively coloured on certain rocks on the coast of
W. Ireland, and have multiplied accordingly.

(b) Greater viability, etc., of the melanics.

Bowater (19 14, pp. 300, 303, 308) states that in the course
of breeding experiments on Spilosoma lubricipeda and other
forms the melanics are larger and stronger than the type and
are double-brooded. This was not actually observed by him
in betularia, and, in view of the capricious incidence of physio-
logical variation, we would hesitate to assert that it is likely to
be also found in that species. 1 It nevertheless remains a
possible explanation.

(c) Harrison's theory.

As far as the actual origin of the melanic character is
concerned, Harrison and Garrett (1926) and Harrison (1928)
endeavoured to show that it was due to the salts contained in
the soot-covered food in industrial areas. They did not
commit themselves to theorising how the mutants spread
beyond the industrialised area.

1 Harrison (1928) stated that the artificially produced melanics of other moths
are more delicate than the typical form.

NATURAL SELECTION 215

It is a great pity that this problem has not been attacked
more resolutely. There has been a tendency to accept some-
one's provisional hypothesis and let the matter drop. The
fact remains that we have a very clear-cut case of evolutionary
change transforming a population rapidly and under our
eyes, and the cause has not yet been ascertained. We believe
that Bowater's discovery should be followed up.

(4) Ford (1924, p. 733) states that the varietal constitution
of Heodes phlaeas is noticeably different in Madeira from that
observed by Wollaston seventy years previously.

(5) Crampton (1925, p. 17) found that the distribution of
the variants of Partula suturalis is very different from what
obtained in Garrett's collecting period (1875). The number
and range of the sinistral form have increased. So too in
P. mooreana {I.e. p. 24) : in 1904 the banded type was 44 per cent,
of the population ; in 1919 it was 8 per cent., and in 1923 it
was 2 per cent. In mooreana also a ' new ' colour-variety has
arisen since 1875.

(6) Woltereck (1928) summarises the data concerning the
appearance and spread of certain new (?) forms of Daphnia
longirostris. He (p. 39, supra) attributes their origin to environ-
mental causes — a view which is attacked by Wesenberg-
Lund (1926), who advances an adaptive explanation. This
subject is in its present stage too controversial to discuss
in detail.

(7) Stresemann (1925^. 1 63) states that the melanic variant
of Rhipidura flabellifera sixty years ago was known only in
S. Island, New Zealand. In 1864 it was taken in N. Island,
and has now spread all over N. Island.

(8) Bateson (1913^. 143) describes the spread of the melanic
form of Coereba saccharina, which was originally found on
St. Vincent (W.I.) and now is the dominant form, the typical
saccharina being ' perhaps actually extinct.'

Summary. — The value of these observations, in so far as
presumptive new characters are concerned, is not very great,
because in no instance do we really know that the new characters
are, in fact, genetic novelties and had not been previously
present in a few individuals which had escaped attention.
There is in no instance any evidence as to why the observed
increase took place, but there is very definite proof of periodic
change in the percentage representation of the classes of

216 THE VARIATION OF ANIMALS IN NATURE

variants in natural populations. Aubertin, Ellis and Robson
(1931) have studied separate colonies of a land snail over three
years in a fairly circumscribed area, and have found a rather
limited degree of change in the individual colonies in the
period of observation.

III. The Nature of Variation. — The causes, kinds and
incidence of variation are discussed elsewhere in this work
(Chapter II). What we have to ask here is whether our
present knowledge of these is consistent with a belief in the
efficacy of Natural Selection as the chief agency in evolution.

As already pointed out (p. 183), Darwin took all the facts
of variation at their face value. In the most active period of
his work at least, he believed that a substantial part of variation
was due to environmental effects, and he was at no pains to
distinguish between the somatic and germinal origin of varia-
tion. Still less did he explicitly distinguish between what are
now known as gene-mutations and the variation which is due
to factorial combination (p. 189), though he was, in fact,
familiar with the variation due to crossing. There was, in
short, available for the action of Natural Selection a large
store of variation, the hereditary fate of which he did not
seriously consider and the potentialities of which for per-
manent improvement he did not explore. This vagueness was
in some measure clarified by de Vries. on the one hand and
Weismann on the other, and the evolutionary speculations of
the period about 1880- 1920 were based on the recognition of
germinal as opposed to ' fluctuating ' variation, of which the
former alone was held to be of evolutionary significance.
Furthermore, genetical investigations revealed the distinction
between mutation (change in the constitution of a gene or of a
chromosome) and the variation due to heterozygosis in the
parents.

In Chapter II we have examined the evidence as to the
inheritance of induced modifications. We concluded that some
of the data suggest that this, at least, is possible in certain
circumstances. Although the conditions under which such a
process can operate appear, at present, to be rather restricted,
its mere possibility cannot but make the premises of all evolu-
tionary speculation somewhat uncertain. As we have pointed
out, the problem of the evolution of habit and instinct still
requires a solution, without which any theory deduced merely

NATURAL SELECTION 217

from the study of structure will be unconvincing (cf. also p. 300).
For this reason evolutionary speculation may be said to be
halting on the very threshold of its field of inquiry. Never-
theless, the following statements seem justified : (1) that
much variation in animals is seen definitely to be of the
fluctuational order, and to be of no evolutionary importance ;
(2) that some mutations arise with no apparent cause in the
environment ; (3) that a limited number are known to be
related to extrinsic factors ; and (4) that factorial combina-
tion is responsible for a good deal of variation.

It is necessary to return for a moment to the question we
have posed on pp. 28-9. We drew attention there to the highly
suggestive nature of some of the recent work on induced heri-
table variation, and we restated the doubt originally expressed
by Robson (1928, p. 254), whether ' germinal ' change is likely
to be a purely spontaneous phenomenon and entirely inde-
pendent of external stimulus. We freely admit that certain
gene-mutations appear to arise without any specific external
stimulus and in the present state of our knowledge must be
treated as ' spontaneous.' The recent work on the induction
of mutation by raising the temperature of cultures or exposing
them to radiation cannot be said as yet to explain the bulk of
ordinary mutation, and we regard the ultimate causes of gene-
mutations as highly problematical. With this uncertainty in
the background, it cannot be said that evolutionary inquiry is
ready to answer in a very authoritative fashion the questions
which it raises.

Of course it may be argued that, even if gene-mutations are
ultimately due to external stimuli, we have still to account for
their spread and multiplication. It is, indeed, theoretically
possible that a local population may be transformed en masse
by the action of the environment. There is some slight evidence
in favour of this, but it is not enough to convince us that this
is a very important factor in evolution. Moreover, the appeal
to a general environmental modification of a population
involves us in a number of difficult questions (Robson, I.e.
p. 174).

Even if we begin by admitting the possibility of some
induced variation being hereditary, and thereby acknowledge
that the general situation is obscured by doubt on a very
crucial issue, it is still possible to discuss a part of this question

2i8 THE VARIATION OF ANIMALS IN NATURE

to some purpose. In the first place, we have to-day enough
evidence from experiment to convince us that much variation
is purely somatic and non-heritable. Darwin's unlimited
variation no longer appears as an inexhaustible fund for
selection to draw upon, and the question begins to shape itself
in our minds — with this reduction made, are heritable varia-
tions frequent enough to provide a reasonable chance that they
will coincide with the crises that supposedly lead to selection ?

The initiation of a mathematical treatment of Natural
Selection was due to Pearson and his collaborators. Pearson
himself (1903) contributed an attempt to give a mathe-
matical expression to the action of selection, and studied
the special effects of selection in reducing variability and
causing correlation. As regards his main theory, ' the
calculations,' as Haldane (1932, p. 171) has pointed out, ' rest
on the particular theory of genetics held by Pearson, and the
results are not in harmony with experimental results obtained
in other organisms.' Of recent years several attempts have
been made to develop a mathematical theory of selection
which is based on our experimental knowledge of the laws
of heredity. These studies (Hardy, 1908 ; Fisher, 1930 ;
Haldane, 1932 ; Wright, 1931) do not in fact provide any
proof of the efficacy of selection, though Fisher and Haldane
imply that selection is the only means of accounting for the
spread of variants that occur as single or few individuals.
Selection is always taken as a vera causa, and the various mathe-
matical expressions of its activity are based on this assumption.
Moreover, although most authors are aware of the fact
that ' all-round adaptiveness ' cannot be neglected, the action
of selection is sometimes considered rather in vacuo as a
unitary process affecting single genes, whereas in nature
survival and extinction are probably issues in which the
organism as a whole is involved.

As we have already indicated, we do not think any
deductive argument can really replace the crucial direct evi-
dence that a selective process actually occurs in nature. But
if for the moment we neglect this point, we believe that it is
easy to be misled by concentrating too much on the genetical
evidence, which is necessarily drawn from a few intensively
studied species (for many purposes a single species of Drosophila) .
After all, what we have to explain is the normal cause of

NATURAL SELECTION 219

evolution rather than the origin of the peculiarities of a few
species.

We believe that the study of the Drosophila mutations has
led to a wrong conception of adaptation, which reacts in turn on
the present form of the Natural Selection theory. The Fisher-
Haldane modification of the Natural Selection theory requires
that animals should be extraordinarily closely adapted to their
environment. Direct evidence of this is hard to obtain. Much
use has been made of the well-known fact that most of the
mutations in Drosophila are less viable than the wild type. From
this it is argued that even the relatively slight changes involved
in most of these mutations are more than the delicate adjust-
ments of the animal can tolerate. Thus it is assumed that the
material with which Natural Selection works consists of much
smaller mutations, not large enough to upset the general
adaptation of the animal, but still big enough to affect the
chance of survival of the mutants. Small beneficial mutants of
this type have not (or scarcely ever) been observed, but Fisher
(1930, p. 19) says : ' In addition to the defective mutations,
which by their conspicuousness attract attention, we may
reasonably suppose that other less obvious mutations are
occurring which, at least in certain surroundings or in certain
genetic combinations, might prove to be beneficial.'

It seems to us a somewhat questionable procedure to
postulate the occurrence of beneficial mutations when in fact
we are so much more familiar with harmful ones. But the
argument appears to be open to a much more serious criticism.
Both the wild type and the mutants of Drosophila are kept in
exceedingly artificial conditions. The greater viability of the
wild type in these conditions provides no evidence as to close-
ness of its adaptation to natural conditions — in fact the insect
can evidently survive in a wide range. All we can safely say
is that the internal adjustments of the mutants are in some
way less perfect, and we may deduce, only, that the internal
adaptations of Drosophila are very complex and delicate (which
we might have suspected previously), not that Drosophila is
highly adapted to its external environment. We do not, of
course, maintain that animals are never selected for life in a
particular environment, but we think that in many cases it is
more important for an animal to be able to survive in all or
many environments. To accomplish this, an evolution of

220 THE VARIATION OF ANIMALS IN NATURE

internal rather than external relations is required. There
may be competition between different degrees of organisation
rather than passive selection by the external environment.
But we shall return to this question in our last two chapters.

Again, it may be questioned whether the pathological
character of many of the mutants is not a more important
feature than the small structural details by which they have
actually been identified. If this is so, the statement that
even the minute structural changes seen in Drosophila mutants
involve loss of viability, is a truism obscured by the way it is
expressed. It is possible that we ought rather to say that even
the pathological mutations of Drosophila produce visible struc-
tural variations. In its natural environment it is possible
that an animal can throw considerably larger mutations
which have no ill effect at all.

The mathematical analysis of Natural Selection and of
the multiplication of variants is necessary and desirable, and
has, we believe, already led to important results. The most
important, as must be expected from the novelty of the methods,
are a reorientation of old evidence and the indication of
new problems, rather than any far-reaching ' explanation ' of
evolution.

We are not competent to criticise from the mathematical
side the methods of the various writers, but, on general grounds,
it appears that three main assumptions have to be made before
mathematical analysis can begin. These are :

(a) A definite mutation-rate.

(b) A definite, even if only average, survival value for a

given mutant.

(c) A system of random mating.

We shall consider these assumptions in the above order.

(a) The mutation-rate. — It is much to be regretted that our
present knowledge of the frequency of gene-mutations is very
limited. Almost all our information (gleaned in somewhat
exceptional circumstances) is derived from observations on
mutation in Drosophila and Gammarus, 1 and we have no means

1 It is not quite certain how long Nabours's protracted observations on the
genetical behaviour of the colour-pattern in the grouse-locusts have been carried
on, but it seems that they have been at least twenty years in hand (Nabours, 1929,
p. 55). During that time only one mutation has been detected (Nabours, 1930,

P- 350-

NATURAL SELECTION 221

of ascertaining how far these are to be considered representative.
Now that temperature is known to affect the mutation-rate,
the actual numerical value of the observed rate must be
received with added caution. But there are more serious
difficulties. It is admitted that mutations may be easily
passed over, so that the observed rates can be only minimum
values. On the other hand, at any given moment there can
be only a limited number of directions in which profitable
mutations can occur, and it is the frequency of these rare
mutations that most interest us. Now statistical methods
are not well fitted for dealing with very rare occurrences. On
this point an interesting article by Bridgman (1932) on the
application of statistics to thermodynamics may be consulted.
He comes to a conclusion that appears relevant to the present
discussion. ' In order to establish with sufficient probability
that the actual physical system has those properties which are
assumed in estimating the frequency of rare occurrences, it is
necessary to make a number of observations so great that the
probability is good that the rare occurrence has already been
observed.' It would seem likely that the occurrence of muta-
tions in desired directions would be rare enough to make it
impossible to estimate their frequency apart from direct
observation.

Probably the most important contribution from the mathe-
matical evolutionists is the basic contention that the known
mutation-rates are insufficient to account for evolutionary
change, if they are unaccompanied by a selective process.
It had been for a long time felt by some authors, who were
inclined to discount the value of Natural Selection, that a
mutation which conferred no advantage on its possessor (or
was not correlated with an advantageous mutation) would
have little chance of surviving the normal incidence of elimina-
tion. Fisher {I.e. p. 20) has stated this difficulty clearly. He
points out that, as the mutation-rate in Drosophila is of the
order of 1 : 100,000, ' a lapse of time of the order of 100,000
generations would be required to produce an important change
in Drosophila ' at the known rate. Thus, ' for mutations
(alone) to dominate the trend of evolution, it is necessary to
postulate mutation-rates immensely greater than those which
are known to occur and of an order of magnitude incompatible
with- particulate inheritance.' There is thus held to be a

222 THE VARIATION OF ANIMALS IN NATURE

strong theoretical case against the survival of non-advantageous
gene-mutations. But at the same time, by stressing the rarity
of mutation of any sort, Fisher introduces a serious doubt as
to the fate of mutations, even if Natural Selection is operative.
For if gene-mutations are infrequent and often injurious, as
Wright (193 1, p. 143) points out, what are the chances that
a viable and useful mutation of this order of rarity will always
occur in those individuals which are allowed to survive by a
death-rate which is probably always at least 50 per cent,
random in its incidence ?

It is most unfortunate that all our exact knowledge of the
rate, nature and hereditary behaviour of gene-mutations is
founded on studies in which the mutations are mainly disad-
vantageous and even lethal (eye- and wing-mutations of
Drosophila, eye-mutations of Gammarus). Exactly how many
of the mutations in Drosophila are of this nature it is not easy
to say. We have taken the list of 389 mutations given by
Morgan, Bridges and Sturtevant (1925, p. 218 and foil.) and
analysed them as far as possible, with the following result :

Lethal
Defective

•

9° )

f2IO c
120) l r220

16 )

9

Viability poor
? Defective .

•

Uncertain or normal

114

Eye colour only

•

40

389

These figures are only approximate, as it is not possible
to be certain which should be regarded as defective ; also we
are uncertain whether the reduction of pigment in the eyes
(e.g. ' pink ') is to be treated as defects : we have accord-
ingly grouped them in a separate category. In ' Uncertain
or normal ' are included a fairly large number of types (e.g.
' ebony 3,' ' dusky ') which are plainly normal from the point
of view of their viability. Speaking generally, it may be said
that nearly 60 per cent, of the mutants are certainly defective,
and a certain small percentage is normal. Sexton, Clark and
Spooner (1930, p. 189) say of the Gammarus mutants that they
' would have but little chance, in normal conditions of nature,
of survival through the early critical period. Each new

NATURAL SELECTION 223

mutation has shown greatly lowered vitality during its earlier
generations, accompanied by marked abnormalities in breed-
ing.' Once established, however, the mutant strains ' tend to
become healthier with each generation.'

The value of calculations and theories based on the muta-
tion rates and types in Drosophila and Gammarus seems to us
to be very questionable. In these forms we are dealing with
a type of variation which is in all probability of an exceptional
order. Wright {I.e. p. 143) speaks of gene-mutations as
' generally injurious,' and suggests that they must necessarily
be of this nature. Fisher {I.e. p. 19) assumes ' that we may
reasonably infer that other less obvious mutations occur which
are not necessarily harmful or lethal.' The position, then, is
that many gene-mutations which have been exactly observed
are disadvantageous, but there may be others which are not.
Surely it is a reasonable inference that, whatever may have
been their frequency of original occurrence, very many viable
mutations of the same magnitude as those in Drosophila and
Gammarus must have occurred. Sturtevant (192112, p. 120)
even records the natural occurrence of eye colours resembling
those of the mutants observed in cultures. From the only
exact sources of information on the subject it seems that we
can draw very few useful conclusions as to either the frequency
or the nature of gene-mutations. If our theories as to the
process by which evolutionary change has been effected are
to be rigorously held to exact evidence, then we have no option
but to admit candidly that, as far as the frequency ] and nature
of observed changes in the gene are concerned, we know
nothing that entitles us to erect a general hypothesis.

{b) The survival value of mutants. — We have already discussed
the small (or negative) survival value of most of the best-
known mutants. We wish here, however, to deal more
generally with the whole conception of an average survival
value as applied to the minor variants which may arise in
any species.

Apart from the uncertainty as to mutation-rates, the
mathematical treatment of the early stages of the spread of
mutants does not seem to be very satisfactory. The particulate

1 The observations of Goldschmidt (1929), Jollos (1930) and others on the
induction of mutation by high temperatures suggest that in exceptional environ-
mental circumstances high mutation-rates might actually be observed.

224 THE VARIATION OF ANIMALS IN NATURE

theory of inheritance has been supposed to have an enormous
advantage over the blending theory held by Darwin. For
with blending, a new variant, unless isolated, is always liable
to be swamped by the excess of normal individuals in the popu-
lation. Hagedoorn and Hagedoorn (1921) have emphasised
that, even with particulate inheritance, the establishment of a
variant from a few individuals almost equally demands the aid
of isolation. In almost all animals the number of individuals
which breed in any one year is only a small fraction of those
which existed at the end of the previous breeding season.
This seasonal fluctuation in numbers means that on the
average only very common types can survive and the chance
of any particular rare variant surviving is very small. The total
variance of the population is being repeatedly reduced, and
the additional chance of survival conferred on a variant slightly
better adapted to some one feature in the environment is very
small — much smaller than would be the case in more stable
conditions. With isolation, though the same factors would
be at work, a new variant might form a far more significant
proportion of the population.

It may be argued that though the chance of survival is
small, yet, if the mutation occurs often enough, it may still
become established ; and that though the mutation-rate be
low, yet, in a species including thousands of millions of indi-
viduals, each type will occur relatively frequently in each
generation. It may be held that, even when the population
is reduced to a minimum, the numbers may still be very large
compared with those in which a mutant might be expected
to occur. In other words, as long as a mutant has a positive
survival value and the species is not a rare one, the actual
value of the mutation-rate is relatively unimportant, at least
within wide limits.

In a species with a wide range, extending over a con-
siderable variety of environments, in each of which conditions
are subject to fluctuations of daily, yearly or of longer periods,
it is somewhat difficult to assign a definite survival value to
a particular mutant. The genetic make-up of the species is
itself unlikely to be homogeneous over large areas. The
idea of an average survival value is necessarily an unreal and
artificial simplification. What is useful in one place or in
one year will be harmful or neutral in another. Survival

NATURAL SELECTION 225

value may have a more definite meaning when applied to
the population inhabiting a small part of the range, but when
the problem is numerically reduced to this extent the actual
values of mutation-rate (as distinct from survival value) and
population density become highly relevant.

The small positive or negative survival values which have
to be arbitrarily assigned to mutants for the purpose of mathe-
matical calculation can have little relation to the facts of
nature, and we may doubt whether the predictions based on
them are very likely to be fulfilled. The actual course of
evolution appears too much determined by special circum-
stances to be very amenable to generalised mathematical
treatment.

(c) Random mating. — Practically all speculation as to the
spread of mutants has been based on the assumption of random
mating. It is evident that nothing approaching real random
mating actually occurs — i.e. it is not true that within a species
any male is equally likely to mate with any female. On the
other hand, if we attempted to allow for selective mating,
our ignorance of the facts would force us to make very large
assumptions which would detract from the otherwise convincing
argument. It might be possible, for instance, to introduce a
factor relating the likelihood of mating to the distance apart
at which the individuals live, but of course it cannot really
be held that the degree of isolation would be a linear function
of the distance.

In Chapter V we considered this subject and were forced
to conclude that permanent isolation of species depended on
a variety of factors working in conjunction, and in any one
section of the population one of the factors may have a
potency which it lacks elsewhere. The species itself must
be expected to be broken up into minor populations,
and much of the evidence presented in Chapter IV supports
this.

If mating is not strictly at random, this will reduce the
effective size of the population in which any one evolutionary
step is proceeding. It may not diminish the power of selection
to spread beneficial variants, but it will make the process of
spread irregular and very difficult to predict, and once more
it is suggested that the numerical values of the mutation-rate
may not be so unimportant as has been supposed.

Q

226 THE VARIATION OF ANIMALS IN NATURE

We have hitherto considered variation in terms of single
mutants. We will now turn to the question of recombinations
of the existing hereditary material. We believe that this must
be quite a secondary problem, since the very possibility of
recombination depends, in our opinion, mainly on the prior
spread of single mutants through large sections of the popula-
tion. But, though in this sense the problem is secondary, it
demands a brief consideration.

The complex genetic basis of a combination puts it at a
disadvantage with changes in a single gene as regards rate
of establishment. This disadvantage might be compensated
for by a substantial measure of isolation. In some crosses
between plants where the parents are rather unlike, the hybrid
may be itself a new type which breeds true and cannot effec-
tively cross with the parents (polyploids) : but in animals
such a process is almost unknown.

Fisher (1930, p. 96) points out that, while it is clear that
without mutation evolutionary change must come to a stand-
still, ' it has not often been realised how very far existing
species are from such a state of stagnation or how easily, with
no more than 100 factors, a species may be modified to a
condition considerably outside the range of its previous
variation.' We have already alluded to this subject (p. 192)
in discussing the experimental production of new races by
selection, and we saw that in practice, though entirely novel
forms may be produced, selection may come to an end very
soon. We hardly think Fisher is right in speaking of residual
heredity with such confidence as a source of evolutionary change.
Moreover, it seems hardly correct to picture a typical character
as determined by as many as a hundred factors, each subject
to selection. Such a rich source of variation as Fisher indicates
no doubt exists if all the segregating characters of a species
are reckoned together : but, if the character subject to selection
is mainly dependent (as is more likely) on a few factors, the
amount of residual variability will be low and Natural Selection
would not be capable of carrying out protracted improvement.
Fisher is right in saying that there are millions of different
ways in which a species may be modified : but this does not
mean that all these are available for a single selective step or
for continued development in any one direction.

We do not deny that in the last resort gene-mutations

NATURAL SELECTION 227

constitute the basis of all new evolutionary steps. We are
inclined to counter the argument that, because they are found
in certain forms to be very rare, they must depend on Natural
Selection for their survival and spread, by suggesting that we
do not as yet know enough about the mutation-rate at large,
especially under natural conditions. But, however that may
be, we have still to discover what is the part played by factorial
recombination. We have mentioned above (p. 25) that this
is capable of producing novel forms {e.g. the numerous cases
of ' novelties produced [immediately] by recombination ' ;
Castle's production of the hooded pattern in rats). Further-
more, the species within a genus tend to comprise very many
that represent permutation and combination of a common
stock of characters, and may very well (though we do not
know of any specific instances) exhibit distinctive and peculiar
characters which arise from factorial recombination. There
are, we admit, limitations to the possibilities involved in
' evolution by hybridisation,' but, given a reasonable amount
of isolation, it seems to us likely that a considerable part of
the early stages of evolutionary divergence may be of this
nature.

The Evolution of Dominance. — Before closing this section we
propose to discuss very briefly Fisher's theory of the evolu-
tion of dominance. His case is put forward in his book (1930,
chapter iii) and in a review (1931). Ford (1930, 1931) has
also summarised the evidence. Wright (1929) and Haldane
(1932) have not accepted Fisher's hypothesis.

Fisher realises that the genetic conception of ' wild type '
is in need of some explanation. The wild type exists because
the majority of genes in animals in nature are dominant to
their allelomorphs which have been detected in the laboratory.
Fisher endeavours to explain the dominance characteristic
of the wild form as the result of selection of the gene-complex
in such a direction that any given mutant will produce the
minimum possible visible effect in the heterozygote. It is
assumed from the data on Drosophila that most mutants,
especially the easily visible ones, will be harmful, and therefore
it will be to the advantage of the species to suppress their
effects as far as possible, i.e. in the heterozygote. The argu-
ments in favour of the theory may be considered under three
headings.

228 THE VARIATION OF ANIMALS IN NATURE

(a) Observations indicating that dominance is not a fixed

property of the gene, but depends on the genetic
environment in which it is placed. We shall not
deal with this, since we consider that, as far as it
goes, the evidence is satisfactory.

(b) Observations indicating that Drosophila mutants are

recessive in their external effects but neutral in
certain slight internal ones.

(c) Observations on certain cases of polymorphism, in

which the phenomenon of dominance presents
unusual features.

(b) Ford (1931, p. 37) and Fisher (1931, p. 353) have
pointed out that certain Drosophila mutants produce a visible
effect (e.g. white eye) and an internal effect (e.g. change in
proportions of the spermatheca). In all the examples investi-
gated the external effect is recessive and the internal one is
neutral, i.e. the heterozygotes are intermediate. It is argued
from this that selection has acted only on those effects of the
gene which are harmful, visible changes such as those in eye
colour being more likely to affect the life of an animal than
minute changes in internal structures. This argument appears
to us to fail in two directions. First, the small internal effects
are just the sorts of variants which, in the case of specific differ-
ences, are assumed to be selected. Secondly, many specific
characters are admitted to be probably of no survival value
to their possessors, but are supposed to be correlated with
more important, possibly physiological, adaptations. If the
dominance of the wild type has been evolved by selection, we
can see why the adaptive characters would have been made
dominant, but the useless specific characters should have
remained neutral. So far as the conception of the wild type
has any meaning at all, this is not the case. As a rule we do
not know why the mutant forms of Drosophila are less viable
than the wild type. Sometimes, as in serious malformations,
the character by which the mutant is recognised might be
expected to have a direct effect, but in most mutants this is
not the case. We might therefore have expected the unknown
harmful effects to have become recessive, while the small
visible effect would have remained neutral. Possibly it is
wrong to assume that selection can alter one part of the effects

NATURAL SELECTION 229

of a gene and not the remainder, but in that case also this
part of Fisher's argument is invalidated.

(c) We cannot consider Fisher's evidence as to polymorphic
species (grouse-locusts, land snails, butterflies) in detail. All
the examples are highly complicated and admittedly in need
of further investigation. In order to support the theory of
the evolution of dominance it is necessary to assume that a
selective process has been favouring the heterozygotes at the
expense of the dominants. There is no direct evidence that
such selection occurs, and in the case of land snails (Cepea)
there is some evidence that the attacks of birds on the different
colour-forms are indiscriminate. The number of such poly-
morphic species is much larger than is perhaps realised (cf.
Chapter IV, p. 94), and the development of an ad hoc explana-
tion for each of them would be a thankless task.

Ford has also pointed out (1931, p. 55) that selection in
the direction of suitable gene environment will be going on
in many different directions at once, some of which may be
antagonistic. He argues that the number of relevant environ-
ments for any one gene may be relatively small, so that a
number of selective processes could proceed simultaneously
without interference. We find this argument unsatisfactory,
and must regard the theory of evolution of dominance as
still in need of verification. There is no direct evidence that
most mutants are not recessive ab initio.

Summary of Section

The preceding paragraphs may be summarised as follows.
If we examine the little we know as to the causes and frequency
of new variations, we find the data are far too scanty to warrant
any generalisation. We are not able to say whether muta-
tion-rates in nature are as low as suggested. This, of course,
has no direct bearing on the value of the Natural Selection
theory, but it does mean that extensions of the original theory
should not be made to depend on the mutation-rate of Drosophila
as observed in laboratory conditions. The data for a con-
vincing mathematical treatment of Natural Selection are not
yet available. The formulae at present proposed rely to a
large extent on assumptions which have to take the place of
the missing evidence. None of the formulae seems likely to
approximate to the actualities of fluctuating environments and

230 THE VARIATION OF ANIMALS IN NATURE

populations. This appears to hold whether they define the
conditions governing the spread of new mutations or of new
combinations. The theory of the evolution of dominance
has also been considered. It seems at present to lack
sufficient direct verification, while some of the indirect evidence
is of doubtful value.

? IV. Indirect Evidence for and against the Natural

Selection Theory. — We have seen that the direct evidence
for a selective process is inadequate both in quality and
quantity. This inadequacy is largely due to the difficulties
involved in the necessary investigations. Recent work on
insect parasites and some of the fishery investigations suggest
that the direct method of attack is not so hopeless as has been
thought. Under the stimulus of economic gain — e.g. in the
Cornborer investigations — it has been possible to breed
millions of insect larvae and to determine accurately the
incidence of some of the important causes of mortality, and
it is not unlikely that further developments of similar methods
may eventually give us a reasonably complete picture of the
death-rate in a few species.

We prefer to take this optimistic view because there are
grave difficulties in the employment of indirect evidence.
The bulk of the latter aims at showing that certain structures
or habits are ' useful.' This does not prove that they are
actually, on the balance, of survival value to their possessors.
To do this we should have to compare the death-rates of forms
with and without the structure or habit in question. But this
comparison involves the study of the direct evidence for the
selection theory.

Again, it is usually stated that the relations of any animal
to its environment are so complicated that we can never hope
fully to demonstrate the action of Natural Selection, and in
particular can never show it is not operative in a given case.
This argument is commonly brought forward to explain the
apparently non-adaptive specific characters. But the appeal
to ignorance is two-edged and cuts both ways, and cannot be
used to turn apparently unfavourable instances to advantage.
That is too much like a marksman who, seeing his birds flying
away, says that for all he knows they may belong to a variety
resistant to shot.

When Darwin wrote, it was very important to convince

NATURAL SELECTION 231

everyone that evolution had actually taken place. To that
end he endeavoured to collect a large body of evidence that
apparently could be explained only on the Natural Selection
hypothesis. To-day the much greater body of morphological,
taxonomical and embryological evidence is alone almost
enough to prove that evolution must have occurred ; and if
we admit that living organisms are always derived from pre-
vious living organisms, the picture of extinction and gradual
change presented by the palaeontological record completes the
argument without forcing us to say exactly how evolution
happened. In Darwin's day it was legitimate to ask, ' If these
structures are not the result of Natural Selection, how do you
explain them ? ' To-day we are able to answer, ' We cannot
explain them,' and yet not feel that we are betraying science.
This digression disposes of the argument that Natural
Selection must be all-important because nothing else would
explain the facts. There are many things about living organ-
isms that are much more difficult to explain than some of their
supposed ' adaptations.'

It is possible to cite a large mass of indirect evidence that
has been held to prove that the structural differences that
distinguish species and lower categories are rela