Sayce, Archibald-Henry. Principles of Comparative Philology – T01

The principles
of
comparative philology.

Chapter I.
The sphere of comparative philology, and its relation
to the other sciences.

Among the many new departments of study which
have been called into existence by the extension of
the scientific method, there is none that possesses
greater interest than Comparative Philology. It
is, on the one hand, so closely bound up with the
history of mankind in general, while, on the other,
it enters so largely into the life of the private
individual, that there are none whose attention it
ought not to excite. We have no need, as in ethnology
or botany, to collect from outside the preliminary
facts upon which the science is built:
the facts of Comparative Philology are literally
in the mouths of every one; they are the words
which we speak, the thoughts which we clothe in
1articulate language, the indispensable links of
union which bind together a civilised society.
While we are thus favourably placed in regard to
the materials of our study, the study itself appeals
at once to our reason, our imagination, and our
curiosity. Language peculiarly belongs to man;
it is the most characteristic mark of distinction
between him and the brute; and the careful
examination of it seems therefore especially appropriate
to him. The object of all science, it may
be presumed, is twofold: to obtain such a knowledge
of nature and its laws as shall enable us to
combine and control them practically for our
future use and benefit. Years have not deprived
the old Delphic oracle of its truth, and a knowledge
of ourselves is still the most important
that we can acquire. The improvement of the
species, the amelioration of society, the well-being
and happiness of the individual, are the
most pressing questions of our age. But to
answer these satisfactorily, we must know the
laws which govern the race and the individual,
and the way in which we have arrived at our present
condition. Every new discovery confirms the
theory of progressive development: man was not
once what he now is; and the long series of centuries
that lie behind us have seen him slowly
changing with changing circumstances, and gradually
2moulded by the experiences and habits of
former generations. It is not the outward form
of man that concerns us now; that, indeed, may
have altered, and the long ape-like jaw of the
primitive savage have contracted into the mouth
of a Cleopatra or a Mary Queen of Scots. But the
change of external shape has little interest for the
politician and philanthropist; it is rather an index
of deeper spiritual changes than the cause of them;
and with these more secret and subtle changes
the student of society has alone to do. It is the
development of the moral and intellectual life of
mankind a knowledge of which is so necessary
if we would understand the present state of society,
and rightly set about its improvement. But
over the first beginnings of this moral and intellectual
life — the very foundations of it, without
which all the superstructure is but half intelligible
— there is drawn, as it were, the veil of Isis, and
the veil can only be lifted by the interpreter of
the symbol. Such an interpreter is language, the
mediator between the spiritual and the physical,
which, records the varying phases of human
thought in enduring symbol and sensuous metaphor,
like the rocks which bear witness to the
climate and zoology of remote geological eras. If
we are to look anywhere for the solution of some
of the highest problems connected with the history
3of our species, it must be, above all, to the
science of language. Already much has been
done by it; not the least good being the clearance
of many old prepossessions and beliefs that blocked
up the path of inquiry, and distorted all the evidence
that might be presented to the mind.

But it must be remembered that the science is
still in its infancy, and more has been expected
from it than its present stage of advancement
would legitimately allow us to demand. Many
causes have combined to give an impetus in the
present day to the investigation of the historical
sciences — those, namely, which deal with man and
his works, and to the study of social phenomena.
Ready conclusions and rapid generalisations are
wanted; answers to the many questions which are
starting up on all sides cannot be waited for; and
the Comparative Philologist is accordingly called
upon to furnish the key or suggest the solution of
numerous difficulties. His situation is a tempting
one. Knowing, as he does, how much certain
ground has already been won, and acquainted
with a further range of data from which he is
only too well disposed to draw hasty inferences,
he is ready to take his seat on the tripod, and
deliver dogmatic statements which, are received
by the general public as so many ascertained
facts. If put forward as provisional hypotheses
4only, intended to represent the sum total of the
evidence upon some particular subject which the
inquirer had at his disposal, such statements
would have great value; but the mischief done is
immense when they are made and received as of
equal authority with the ground-principles of the
science, and become so many propositions which
may not be contradicted. It is, indeed, the lot of
all new sciences; but none the less necessary on
that account to be foreseen and guarded against.

A man's foes, it has been said, are those of his
own household. Comparative Philology has suffered
as much from its friends as from its opponents;
and now that it has at last won its way to general
recognition and respect, there is a danger that its
popularity may lead to the cessation of sound and
honest work, and to an acquiescence in theories
which, however plausible, are not yet placed upon
a footing of scientific certainty. The great names
to whom the scientific study of language owes its
origin are passing away from among us, and there
is reason to fear that their places may be taken by
patient plodders, content to work out small details,
and to walk in the paths already traced for
them, rather than to criticise and re-examine the
magnificent generalisations of their masters, and
to further the progress of the study by fresh hypotheses
of their own. Newton, was followed by a
5century of stagnation, and Aristotle's successors
were the grammarians of Alexandria. Geniuses
are rare, and it is much easier for the ordinary
man to fill in by patient elaboration what has
already been sketched for him in outline, than to
venture upon a new line of discovery, in which
the sole clue must be the combinative powers of
his own imagination and comprehensive learning.
And yet, now as much as ever Comparative Philology
has need at once of bold and wide-reaching
conceptions, of cautious verification, and of a
mastery of facts. It is true the science is no
longer struggling for mere life, and the time is
gone by for proving the possibility of its existence
But it is still young, scarcely, indeed, out of its
nursery; a small portion only of its province has
hitherto been investigated, and much that is at
present accepted without hesitation will have to
be subjected to a searching inquiry, and possibly
be found baseless after all.

Scientific hypotheses do not pretend to do more
than explain all those phenomena which are known
at the time of their formation; they supply the
mind with a clue for further researches; they
serve to connect the isolated facts, and to simplify
the bewildering maze in which we find ourselves;
and however erroneous they may eventually
turn out to be, they will yet be of use, like
6will-o'-the-wisps, in warning future students from
what has been proved to be wrong. But they can
do no more than this: with the accession of
further facts and the enlargement of the boundaries
of the science, they have to be continually
modified, and often to be given up altogether.
A science consists of hypotheses more or less
nearly related; and its aim is to make these
hypotheses correspond more and more closely with
the observed facts. It is evident, therefore, that
while a science cannot progress without the formation
and testing of hypotheses, a young science,
like that of Philology, will put forward many
which maturer knowledge will show to be untenable.

Now, it is necessary to bear in mind what is
meant by science and scientific knowledge. Scientific
differs from the ordinary knowledge of practical
life in being comparative. In order to know
an object or be conscious of a sensation, we have
to compare and contrast it with some other object
or sensation. The more accurately this act of
comparison is performed, the more nearly shall we
approach to scientific certainty. For this purpose
a standard of comparison is required, some third
term with which we may compare our two other
terms. In other words, to use Mr Herbert Spencer's
language, the distinction between scientific
7and unscientific knowledge is, that the one is
quantitative, the other qualitative. The primitive
savage knew the qualitative difference between
hot and cold water; his senses told him that
much: a scientific knowledge of the matter began
with the thermometer, which enables us to measure
the amount of heat in each case.

It is easy enough, then, to see wherein a scientific
treatment of language differs from that haphazard
charlatanry at which Voltaire directed one
of his epigrams. Language is the expression of
thought and feeling through mechanical means;
and just as it has been found possible to construct
a science of thought and feeling, so with greater
reason should we expect to discover law and order
when that thought and feeling has been subjected
to the restraints of physiological conditions, and
expressed in articulate speech. Every sound
emitted by the human voice is the result of the
physical formation of the vocal organs, and of the
manner in which these are brought into contact
with the breath; while, on the other hand, the
laws which govern the development of the human
mind will necessitate the expression of thought
and its relations in a particular way. Language is
limited as much on the psychological as upon the
physiological side: a knowledge of this twofold
limitation will constitute its science. And inasmuch
8as the two sides can be as little separated
from one another in actual speech as oxygen and
hydrogen in water, or colour from the objects about
us, the general laws of the science must relate to
the combination, although, for analytical purposes,
it may be advisable to investigate the two separately.
But we must never forget that such a
separate investigation is preliminary only. Neither
linguistic metaphysics nor phonology by themselves
represent philology, but a combination of
both. We may have laws of phonology like that
of Grimm, or laws of linguistic metaphysics, such
as that every predicate must have a subject. But
these are only empirical, subordinate, and partial,
forming the scaffolding of the higher and more
comprehensive generalisations of the master-science
itself. This, however, is a truth often forgotten,
and more will be said about it further on.

Now, in being scientific, Philology must be
comparative; and it is simply the application of
the comparative method to the phenomena of
language that has brought the new science into
being. The attempt to study a language without
reference to any other is futile. A certain number
of empirical rules may indeed be found peculiar to
the language in question, but the reasons of the
existence of these, and the more important and
general laws to which the language conforms, can
9only be discovered by a methodical comparison
with other dialects, while many of the fancied
facts of “scholarship” will turn out to be the
most portentous errors. Hence originated such,
beliefs as the derivation of Latin from the Æolic
dialect, the misconception of the locative case,
the idea of the priority of the passive to the middle
voice in Greek, the identity of καλέω and “call,”
ὅλος and “whole,” or the grotesquely wrong
meanings assigned to such Homeric words as διερὸς
and μέροπες, 11 not to speak of Buttmann's endeavour
10to get ἄφνος out of ἄφθονος. 12 Until a common
quantitative standard was applied, until it was
recognised that language, like everything else in
this world, obeys undeviating laws of its own,
excessively complicated though these may be,
such mistakes were inevitable. As in other things,
so in language. We cannot really understand a
single dialect unless we study it in the light of
others. For literary and artistic purposes this
may not be necessary, but then we must not confound
such a study with philological knowledge,
and believe that we know a language because we
can successfully imitate the idiosyncrasies of a few
of its literary men.

Dr Wagner, in the President's Annual Address
to the London Philological Society for 1873 (p. 33),
says (speaking of German scholars), “We have
passed the stage of a sentimental admiration of
the ancient authors such as we find it in the
editions of Heyne and his school. Our eyes are
fully open to the shortcomings and failings of
Latin literature when considered æsthetically, nor
11do we any longer attribute to this literature the
‘humanising’ influence so naïvely believed in by
former centuries. There is among ns very little
of that which may be termed elegant scholarship —
which is all very nice, but perfectly useless; in
fact, we do not work like ladies, but like men
mindful of a serious purpose, which is, in the first
line, to trace the intellectual life of the great
Roman nation in its literature; and secondly, to
show and follow the connecting links between this
literature and the other nations of Europe and
Asia. To attain this end, it is necessary to pursue
the most minute investigations, but not to generalise
without sufficient data and foundations. But
the days in which it was held the height of Latin
scholarship to write a splendid Ciceronian style,
and to turn neat Latin verses, are past, and will
never return.”

In calling Comparative Philology a science, we
must not, of course, think of it as an exact science
like astronomy. Here the phenomena are comparatively
but little complicated, and have been
studied for a considerable number of years. The
generalisations obtained from them by a comparison
of instances have been so far simplified as to
be resolved into the one primary law of gravitation,
which serves as the starting-point of deductively
determining the relations of new astronomical
12phenomena. Other sciences have not yet reached
this exactness; chemistry seems likely soon to
find its primary law, but meteorology, sociology,
and many more in which the phenomena are extremely
complex, are very far indeed from such
perfection. Here we can only collect, compare,
and classify, thankful if we can bring the isolated
phenomena under some general heads which may
bear more or less relation to one another. The
process is strictly inductive; we assume the uniformity
of nature, and generalise from the facts in
accordance with what we see at present going on
around us, testing these generalisations by fresh
instances and combinations of instances. Thus in
Philology the facts with which we have to deal
are thoughts expressed in speech. So far as these
will carry us, we can proceed with our generalisations.
We assume that the same mental processes
were involved in the first attempts at language
that are involved now, and that, given a certain
arrangement of the vocal organs, the same sound
will always have been produced. In other words,
we assume the uniformity of nature in regard to
language. With this assumption we proceed to
our comparisons, classifying the like together and
separating the unlike. It is the object of the
science to discover the limits of this classification,
and to create an ideal type, as in natural history,
13around which we may group the several phenomena
which resemble one another. Thus we put the
so-called Indo-European languages into a class
apart by referring them to an imaginary Aryan
parent-speech; and we throw together a number
of derivatives, or a series of ideas, by assuming a
common root or a common primary notion. In
this way we come to know the typical marks by
which similar instances may be recognised. The
analogy of the other sciences would lead us to
infer that these typical marks are by no means
those which first meet the superficial view; and,
in fact, one of the first results of Comparative
Philology was to lay down that mere similarity of
sound could constitute no basis for a sound comparison.
Language is not phonology only; if we
would seek the true marks of difference and resemblance,
we must penetrate below the surface,
and find some surer guide to our first attempts at
grouping than the shifting modifications of sound.
Speech is uttered thought; grammar and structure
therefore must lead the way to the examination of
the lexicon. When we have formed our groups
by comparing the grammatical characteristics of
the languages under review, we may complete the
process by comparing the vocabularies, knowing
the limits within which the resemblance of letters
is due to identity of origin, and not to accident.
14The groups thus formed will then have to be compared
with one another, and the general laws of
the science determined from them. It is evident
that such a comparison must be as wide as possible;
the greater the number of facts brought together,
the more diversified in time, and space, and circumstances
the languages compared, the safer and
more general will our conclusions be. To confine
our attention to a single family of speech, much
more to two or three members of the family, will
lead us into many errors and false generalisations.
No idiom, however obscure and barbarous, can be
despised by the comparative student. The most
precious facts of the science will often lie in dialects
whose very names are almost unknown, and
whose speakers stand upon the lowest level of
humanity. It is in these, however, and not in
the polished periods of a classical literature, that
we can trace the fundamental laws and working of
primitive speech, and detect those simple contrivances
which have elsewhere been obliterated.
Science desires truth, not beauty, although in the
end the true is always the beautiful.

The laws or generalisations which we are called
upon to observe are of two kinds — empirical and
ultimate, or primary. So long as we confine our
attention to one part only of the subject, we meet
with a number of rules which are always complied
15with, though we cannot account for their existence.
Thus we find that a Gothic g almost invariably
answers to a Greek χ, a Latin h or f, a
Sanskrit gh, and a Slavonic g or z; but why
this should be we cannot at present tell. 13 We
only know that such is the case; it is an empirical
law, the immediate result of observation, which
will have to be explained by some higher and more
comprehensive law. These subordinate laws have
to be made out before the primary can be deduced
from them by comparison; but the primary laws
alone belong to Comparative Philology properly
so-called, the subordinate ones relating to the
preliminary subdivisions of the science, such as
Phonology. But both kinds of laws are alike
affected by two great principles or causes of
change in language. These cannot be called laws
themselves, since they do not act in an invariable
manner; but they make a science of language
possible, by preventing it from being stationary,
and by bringing about that constant movement
and development in speech which allows the action
of the several laws to take place. These two
principles may be named laziness and emphasis.
16The first of these has been made familiar to every
one by Professor Max Müller under the name of
Phonetic Decay. Words become clipped and
shortened in the course of time, until it may
happen that nothing is left of the original, some
secondary termination alone remaining. Thus the
Latin pilus has passed the various stages of the
Spanish peluca, the Italian perruca, the French
perruque, and the English perwiche, periwig, into
the modern wig. Rapid speaking, an imperfect
ear or pronunciation, and the common desire to
save time and trouble, will inevitably wear away
the words of everyday life. Where little care is
taken of language, where there is no literature,
and no standard or court dialect, the vocabulary
will be like the drift boulders that line our valleys,
or the sand and gravel of ancient beaches. The
lower we descend in the scale of culture, the more
rapid and extensive will be the process of decay.
The Berlin workman has contracted ich into i, like
our own countrymen; and the waggoner's wo! and
way! are the last relics of withhold and withstay.
In spite of artificial attempts to preserve the full
forms of the words, and the adoption of Greek
metres by a literary coterie, the curt colloquialisms
of the plays of Plautus and Terence, or the
cauneas with which the contemporaries of Cicero 14
17scandalised the purist, became the models after
which the Romance languages shaped themselves. 15
It may not be a distant period at which the don't,
the I'll, and the isn't of conversation take their
authorised place in books; and if ain't is never
able to lose its taint of vulgarity, it will be due
to the printing-press and the schoolmaster. Children
are the best representatives that we have of
the infantile and barbarous state of society, and
the language of childhood is one of maimed and
half-pronounced words. Such nursery names as
Tom, Harry, Bob, Peggy, have become so many
household terms. But Phonetic Decay is especially
accelerated by the contact of two languages. The
attempt to speak a foreign idiom leads to the rejection
of all difficult sounds. Thus the final guttural
in our enough, through, though, has been softened
and lost; and languages such as the Hawaian, which
do not suffer two consonants to follow one another,
turn words like “steel” into kila (for tila). Indeed,
contraction and decay may be carried so far as
to become an idiosyncrasy of a particular language.
18This is pre-eminently the case in French, which
persistently modifies the pronunciation of every
foreign word which it has to adopt, in accordance
with, its principle of rejecting the final letters.
Thus London must be Londres, and Biarritz
Biarri' in spite of local usage. 16 The terminal consonants
have been lost in the majority of words,
and the rest of the vocabulary has had to follow
the general fashion. Analogy has immense power
in language, and whatever once becomes a distinguishing
feature of a dialect forms a type after
which every exception is gradually forced to model
itself. 27 As poetry is better remembered than
prose, so the rhythm of analogy fixes itself upon
the memory, and the ear and will, once accustomed
to a particular association of sound and
idea, instinctively demand the sound when the
idea has to be expressed. Irregularities constantly
tend to disappear, more especially if no
artificial means are employed to perpetuate them.
The Æolic dialect assimilated the accentuation of
19every word to the general rule which threw the
accent back upon the antepenultima; our own
tongue is replacing the strong preterites of our
verbs by the secondary perfect in -ed, originally
dide (did), the reduplicated past tense of do; and
an English child whom I knew, born and brought
up in France, and speaking French only, conjugated
all the verbs regularly, saying, for instance,
avrai for aurai, and allerai for irai. In fact, we
may lay it down as an universal rule that the
oldest grammatical forms are those which are
rarest in a language; so that when we find in
Sanskrit only a very small number of verbal roots,
such as as-mi, ad-mi, which affix the pronoun without
any intervening element, we may regard them
as going back to the most ancient period of the
speech. The influence of analogy had been continually
narrowing the province of the formation,
until only those verbs which constituted the most
necessary stock-in-trade of everyday life were able
to resist the encroachments of other later but
more popular forms. If, however, we really want
to see the principle of Phonetic Decay in its full
activity and importance, we must turn our eyes to
unwritten dialects rather than to that particular dialect
which has accidentally been stereotyped into the
standard language of literature. Here the various
processes which change and develop language go
20on unchecked; and unless we can compare dialect
with dialect, it is often obviously impossible to
settle the original form, and therefore the true
etymology, of some word in the special idiom we
are examining. The wear and tear of time alters
so completely the face of words, that where we are
not able to apply the scientific method of comparison
by the help of cognate dialects, our
attempt at derivation is likely to be nothing more
than an unscientific guess. It is this want of
allied dialects that makes Latin etymology a
matter of such difficulty and uncertainty; and we
have to be thankful for the fragments of Oscan,
Umbrian and Sabellian which we can recover
from a few inscriptions or the scanty notices of
grammarians. When we remember that it is only
our extensive knowledge of the languages which
are, as it were, the daughters of Latin that enables
us to trace such a word as the French même, for
instance, through the Portuguese mesmo, the Old
French meïsme, the Provençal medesme, and the
Old Provençal smetessme, to the Latin semetipsissimum,
we may well despair of making out the
true ancestry of words when such an assistance is
not available. Even the Turanian or Ugro-Altaic
languages, 18 which do not so readily admit of
21phonetic decay as the inflectional families of
speech, are kept living by the same active principle,
and without dialectic comparison we should
be altogether unable to penetrate their secrets. It
is thus that we can analyse the verbal forms in
Magyár, Mordvinian, and Vogul, which incorporate
the objective pronouns, or trace the original
forms of the Turanian numerals; and if Basque is
to be added to the group, the importance of an
acquaintance with a variety of dialects becomes
still more manifest. The Basque verb presents
the phenomenon of incorporation to an astonishing
degree; not only the objective cases of the
pronouns, but the datives and the index of the
plural as well, are inserted into the body of the
22word, and the whole has been fused together by
the influence of phonetic decay into a hardly distinguishable
unity. A comparison of the several
Basque dialects — Labourdin, Souletin, High and
Low Navarrese, Guipuscoan, and Biscayan — is
equally indispensable for the vocabulary. Basque
has existed for centuries as an unwritten language,
separated from the rest of its kindred, and struggling
for existence in a small tract of country. If
we are to discover the affinities of its lexicon, it
must be by knowing what were the primitive forms
of its words. The larger part of the dictionary is,
indeed, derived from Spanish or French; but
when we find that the natives of S. Jean de Luz 19
ordinarily drop r and d between vowels, without
even the substitution of the aspirate, thus making
harits (oak) haits, 210 aditu (heard) aïtu, baduzu
23(have you) bauzu, emadazu (give me) emāzu, we
may well be cautious even when we are dealing
with a member of the agglutinative family of
speech.

Of course, Phonetic Decay attacks principally
those portions of a word or sentence upon which
no emphasis falls. 111 The accented syllable remains
untouched, and when this is a secondary derivative
and not part of the root, sometimes causes
the entire loss of the root itself, as in age, from
ætaticum, where the first letter only claims connection
with ævum, αἴων (Etruscan aiv-il), our
ever, Sansk. âyus, “life.” 212 In a case of this
kind we have another principle besides laziness
brought into play. It is the striving after
clearness and distinctness, the second cause of
change in language, which I have called the principle
of Emphasis. It works in the contrary
direction to Phonetic Decay, and, as it were, counterbalances
the latter. The use of language is to
make ourselves intelligible to others; and the
more intelligible we wish to be, the more careful
we are in our pronunciation, and the greater stress
24we lay upon those words or syllables to which we
would particularly direct the attention. If we find
that a foreigner does not understand us, we instinctively
raise the voice and speak with slowness
and precision. There can be little doubt that
the principle of Emphasis loses in force with the
progress of culture and intelligence. Education
makes us readier in catching the meaning of those
with whom we are conversing, and our mastery
over ideas gives us the clue to many of which only
a fragment, only a suggestion, has reached the
ear. 113 The modern Englishman of the upper
classes, particularly if he belong to the south of
the island, is notorious for closing his lips and
lazy indistinctness of speech. It is quite otherwise
with savage races. They lack that quickness
in seizing the signification of what is set before
them which is characteristic of the civilised man,
even though they do not display that hopeless
bewilderment which Mr Galton's African Dammaras
showed when required to count beyond
25three. 114 The meaning of their words has to be
eked out by gesture and gesticulation, and the
muscular effort called forth by these necessarily
extends to the elocution also. If we would speak
clearly we must take the trouble to exert our
muscles in the endeavour. 215

Now the principle of Emphasis acts upon language
in many ways. First of all, it lies at the
bottom of what Professor Max Müller entitles
Dialectic Regeneration, which he seems to set up
as the counterbalancing principle to Phonetic Decay.
The words, however, and still more rarely
the grammatical forms, which from time to time
find their way from the so-called dialects into the
literary language, are too few and unimportant for
the process to be raised into a principle, much less
a principle co-extensive with that of laziness.
We want one which is the same in kind; one,
namely, which is due to the general constitution
of our nature. Moreover, Dialectic Regeneration
principally applies to literary languages only, not
26to the mass of human speech. And even in these
its action is extremely limited, and, unless we can
find the motive of it, at once accidental and capricious.
The motive, however, is the desire to give
additional strength and clearness; to make the
language employed more forcible, and therefore
more distinct and plain. A new word, taken up
from the fresh fountain of living speech, carries
with it new ideas, and impresses itself upon the
mind more vividly than the familiar expressions
which have become nothing except dead insipid
symbols. We read of the “four points of the
compass,” with a full understanding of what is
meant, but without picturing it to ourselves in any
way; but when Carlyle talks of “the four airts,”
at once our attention is aroused and our imagination
engaged. Our mechanical association of
sound and sense can alone be broken through by
novelty, and the excitement of realising the complete
force of a term which has come up from the
patois where the life of language is still vigorous,
and words have not become mere counters and
conventional signs. Another mode of arresting our
attention and giving distinctness to the thought
which has to be expressed is by setting two synonymes
side by side. This is especially frequent in
a language like English, the vocabulary of which
owes as much to Latin as to Saxon; and much
27of the charm of the authorised version of the Bible
is due to the fact that the translators have usually
tried to bring out the meaning of a Greek word
by using two English equivalents, one from a
Romanic, the other from a Teutonic source. In this
way we are obliged to dwell upon the conception
intended, and to contrast and define the two synonymes.
Somewhat similar is the origin of that
analytical tendency which distinguishes our modern
European languages. The difficulty of mutual
comprehension on the part of the Roman provincials
and their Teutonic conquerors necessitated
the distinct expression of each grammatical shade
of meaning by a separate word. The old broken-down
inflections would no longer suffice. The
idea had to be clearly marked out, not merely
suggested, for a people whose ear and mind were
unaccustomed to the language with which they
came in contact. Amabo did not sufficiently convey
the conception of futurity to the Frank: the
termination allowed him neither time nor opportunity
to consider what it intended to signify; and
in order to have the tense-distinction clearly presented
to his view, it was necessary to go back to
the definite representation of futurity — ama-fuo —
out of which amabo had grown, and analyse the
concept into aimer-ai, “I have to love.” Even
this was not enough; the personal pronoun had to
28be prefixed, and no longer implied only in the
form; and when j'aimerai itself had become familiar
and conventional, a new mode of expression, in
which the attention might be fixed upon the fact
that future time was denoted, had to be invented
inje vais aimer. The influence of Emphasis will
again show itself not only in the preservation of
sounds that would otherwise be subject to Phonetic
Decay, but also in the introduction of expletive
ones. The insertion of the dental and labial in
such Greek words as ἀν-δ-ρὸς and μεσημ-β-ρία may
indeed be ascribed to the first principle rather than
to the second, since their addition facilitates pronunciation;
but this cannot be said for the final d
in our own sound, lend (A.-S. lænan), riband (Fr.
ruban), and the like. The same letter has also
crept into thunder (A.-S. thunor), tender, and
jaundice (Fr. jaunisse). The effort to be distinct
has again produced thumb out of thum-a, behest out
of behæs, amongst out of amonges, tyrant from the
Old Fr. tiran, parchment from parchemin, ancient
from ancien. So, too, citizen has come from
citoyen, though this may have been due to an orthographic
mistake. Hardly so, however, the
inserted letter in impregnable, from the French
imprenable; and the cases of an intrusion of an n
or r into the middle of a word are numerous.
Thus nightingale represents the A.-S. nihtegale,
29messenger, passenger, and popinjay are the Old
Fr. messagier, passagier, and papigai; groom
and horse are the A.-S. guman and hôs, cartridge
is the Fr. cartouche, corporal is caporal, culprit
comes from culpa. Similarly n has been added in
bittern, A.-S. butore, and marten, A.-S. mearth,
and the Fr. perdrix (our partridge) goes back to
the Latin perdix. 116 The same principle is at work,
but in conjunction with phonetic decay, whenever
the loss of a sound is compensated by the lengthening
of the adjoining syllable, as in mollis for
molvis (stem madhu?), μᾶλλον for μᾰλyον, or feci
for fefeci. But the principle appears by itself in
lengthened forms, such as μανθάνω, λαμβάνω, where
the secondary inserted syllable — αv — arises from
the wish to attach greater clearness and emphasis
to the action of the verb. Much the same account
must be given of the expletive w and y, which, like
our vulgar kyind for kind, or the Italian luogha
from locus, have played so great a part in Greek
grammar, and in bringing about phonetic changes.
The extension of πόλις (Sansk. purû) into πτόλις
and of πόλεμος into πτόλεμος is a further illustration
of the same tendency. But perhaps the chief
exhibition of the power of Emphasis is to be found
in its regulation of accent and intonation. We
30naturally accentuate the syllable or word to which
we would give prominence and definiteness; and the
less cultivated the language, the more important is
the employment of accent. As has been well remarked, 117
accent and tone vary inversely as syntax;
and we may gauge the development of syntax in a
language by its use or disuse of accent. Chinese
depends almost wholly upon tone, and its syntax
may be compressed into a few lines; English, on
the contrary, which is so rich in syntax and idiom,
is correspondingly poor in intonation. We may
say that tone or accent is to the primitive man
what syntax is to his civilised successors. In
other words, what civilisation expresses by intellectual
processes, barbarism expresses by the physical
management of voice and muscles. Accent
goes along with gesticulation; and action is still
needed by the orator who has to appeal to the
passions, and not to the reason of his hearers.
The important part played by accent in the early
history of speech is still but inadequately recognised.
The guna and vriddhi of the Sanskrit
grammarians are as much the result of it as those
diacritical marks which were invented by Aristophanes
of Byzantium. A considerable proportion
of the phenomena which we observe in Aryan
grammar is the effect of accentuation; and many
31of the changes undergone by the flections are due
to the attempt to lay the accent on the modifying
element of the word. Why, for instance, we may
ask, do we have οἶδα, οἶσθα, οἶδα in the singular,
but ἴστον, ἴσμεν, ἴστε, ἴσασι in the plural? and
why is this distinction in the length and quality
of the vowel in the two numbers preserved in all
the cognate languages, so that Sanskrit gives us
vèda, vêttha, vêda, — vidmá, vidá, vidús; and Gothic,
vait, vaist, vait, vitum, vituth, vitum? Accent
alone can answer the question. When the vowel
of the singular was gunated, that is, raised in
clearness and emphasis, the terminations of the
singular had grown into such common and familiar
use as to convey the ideas which they denoted
without the aid of any distinguishing sign or stress.
It was otherwise, however, with the terminations
of the dual and plural. These still had a somewhat
strange sound, and required a greater effort of
intelligence to connect them with the conceptions
they denoted; consequently they were brought out
into distinct relief by placing the accent upon
them. 118 Something not unlike this has been the
32procedure of those languages which, like the Tibetan
dialects, form the present tense out of the
aorist by doubling the last consonant and adding
a firm vowel; as in ngo gyed do, “I do,” from
ngó gyed, “I did.” Here the indefinite time of
the aorist is made definite by a prolongation of
the syllable, and the distinctness of the idea of
present time marked out by an emphatic dwelling
upon the uttered word.

But these are not all the results that may be
traced to the principle of Emphasis. The origin
of poetry itself may be referred to the wish to set
33forth in clear and distinct language the ideas which,
possess the mind. The more primitive language
is, the more rhythmical we discover it to be; in
fact, early speech may be called a lyric. It is not
surprising, therefore, that verse should be the first
form in which literature clothes itself. The deep
strange thoughts which, with the force of a new
revelation, are struggling to find expression in the
soul of man, must be invested with all the strength,
and distinctness of which language is capable;
and as language itself is poetry, symbolising the
impalpable things of the spirit under the veil of
metaphor, so the earliest form of conscious language
must be poetical. Now poetry at the outset
possesses melody, and not harmony; the notes
must follow one another, each distinct, clear, and
independent; and the monotonous rhythm which,
meets us in the verse of uncultivated tribes is
generally characterised by alliteration. But alliteration
is not only useful as an assistance to the
memory; it serves to force a particular sound upon
the attention, and to afford so many resting-places,
as it were, in which the mind may take in clearly
all that lies between. Throughout the course of
its development, literature remains true to its primary
instinct. So long as books are recited or
read, not to convey knowledge solely, but to communicate
thought and feeling, distinctness of
34pronunciation will be of the highest moment. It
is only in an age of science, when we read not for
the sake of the style, but of the matter, that the
principle of Phonetic Decay takes the place of the
principle of Emphasis. While thought and its
expression are but the two sides of the same prism,
while the language is regarded as an end in itself,
and not a mere instrument for the imparting of
scientific truths or statistical facts or commercial
instructions, every syllable will be watched with
jealous care, and its due weight and meaning
assigned to each. It is in this way that we can
explain the precision and crystallisation of the
literary language of Rome, so different in this
respect from the ordinarily spoken Latin dialects
amongst which Phonetic Decay reigned supreme.
The pronunciation of Virgil and Horace was regulated
by the spelling; and the tendency of Latin
poetry was more and more to avoid elisions. It
was this stereotyped, unreal condition of literary
Latin, as has been acutely remarked, 119 which has
caused the same phenomenon to reappear in modern
literary Italian. Modern Italian is the dialect
of Tuscany, and Tuscany, screened as it is by
mountains, was the part of the peninsula least
affected by the inroad of the Teutonic nations.
The Tuscan population long preserved the relics of
35the old Roman literature and civilisation, and “the
studied accuracy with which the Romans of the
Augustan age pronounced their Græcised poetry”
still lingers in that standard Italian language of
which it has been so truly said that it cannot be
pronounced both well and quickly. We must go
to the other dialects of Italy to find Phonetic
Decay in unrestricted action.

Both Phonetic Decay and Emphasis, however,
have their root in the same utilitarian object:
both are intended to aid the memory. As laziness
would save trouble not only to the breath but
also to the recollection, so the effort to be distinct
has the same end in view. As we should burden
the memory by a needless string of sounds which
are not wanted as soon as the understanding has
seized the idea, so we should burden it equally
were we not to furnish it with the means of easily
determining what idea it is that is intended. To
give too much or too little to the comprehension,
in order that it may take in and remember the
meaning of what is suggested by symbolic speech,
is alike contrary to the economic provisions of
nature. Hence arise the two great principles which
underlie all those laws of language which it is the
business of our study to ascertain by careful observation
and accurate verification.

Comparative Philology, then, must be defined
36as an inductive science, pursuing the same method
of inquiry as geology or biology, and engaged in
the discovery of laws or regulative generalisations
which may possibly be some day applied deductively.
But there is one point in which, in common
with the other sciences which concern the
human mind, Comparative Philology differs from
geology. It is an historical, as distinguished
from a physical, science. In the one case, the
sum of the forces at work remains always the
same — the same processes and the same results
operate still upon the surface of the earth that
operated millions of years ago; in the other case,
the sum of the forces increases in an accelerated
ratio. Every new generation is influenced by the
preceding one; and that influence is a fresh element
of motive power introduced into our calculations.
Human volition is the result of so many
obscure and complicated causes, as to appear at first
sight mere caprice and chance; and an historical
science like Philology is eminently subject to the
will of man. Then again we have to admit the
influence of the individual, who may invent and
give currency to new words, or change the social
condition of a country, though, strictly speaking,
this is only another way of regarding the element
of volition. In short, instead of the simpler, unvarying
processes of nature, which, for the most
37part, can be tested by experiment, we have to deal
with the infinitely complicated developments of
human thought and action, in which observation
alone can be our guide. Language, as we find it,
is as much the creation of man as painting or any
other of the arts; and thus all possibility of forming
a science out of what would be dependent upon the
arbitrary caprice of the individual would seem to
be out of the question. Such, however, is not
the case. It may be true that the individual
exercises some influence upon speech; that individual
writers — for instance, such as Neckar and
Reichenbach — have brought in new words like
sepals and od force, but this influence after all
is infinitesimally small. Language belongs to
the multitude; it is the medium of communication
between man and man; and consequently
must be the combined product of causes and influences
which affect all alike. Now, these causes can
only be general; and if on the one side they are
psychological, they are on the other side still more
physical. The constitution of the human mind is
fundamentally the same at all times and in all
places; every one, be he savage or civilised, must
become conscious of objects in much the same
way, and must express his first needs in a similar
manner. Once grant the power of forming articulate
speech, and there can never be much difference
38in the attempts to realise it. All men have
at bottom the same primary instincts and passions,
otherwise they would not be men; and the primitive
experiences of all races must have been almost
identical. The life and necessities of the barbarian
of to-day differ but little from those of the
barbarian of yesterday. Even greater than the
psychological similarity is the physical similarity.
We are all cast in the same mould. We are all
given the same physical machinery for producing
sounds; and that machinery has everywhere the
same restrictions. We cannot speak without
opening our lips. How far this machinery may
be modified by food, climate, and education, is a
question which will have to be considered hereafter;
in this place it is sufficient to notice that
it can only be modified, never radically changed.
Such modifications, moreover, cannot be individual;
they must affect a whole people, for
language is social and national, not individual.

Language exists for the sake of society: the
self-sufficient man would have no need of such an
instrument of intercourse with his kind. We
speak in order that we may be understood; and
consequently we are obliged to say what is intelligible
to those around us. The child learns the
idiom of his parents, and cannot unlearn it if he
would. It becomes part of himself and his nature
39before he has arrived at an age to think about it;
and so long as he remains a member of a particular
society, he is bound to talk the language of that
society. The invention of a new language would
be an useless waste of labour; he could not expect
any one else to learn it, and so the whole raison
d'être of language would be lost. The individual,
as such, has no language: language is the product
and instrument of society, whose fortunes it represents,
whose laws it obeys, and whose progress it
shares. As particular societies tend to lose their
insulation, and to be more and more assimilated to
each other with the advance of civilisation, so also
it is with the dialects which severally belong to
them.

Thus it is that the element of individual uncertainty
is eliminated from the study. Although
in one sense the creation of man, language is yet
the outgrowth of general causes, and governed by
general laws, partly mental and partly physical.
By extending the area of our comparison, we are
enabled to make these laws more and more
general, and thus more and more to exclude the
caprices and idiosyncrasies of particular nations.
It is true that these idiosyncrasies will have to be
explained; but it can only be done by the light
of the general laws: we can only recognise and
understand the exception by knowing the rule.
40Hence our inductions ought to be as wide as
possible, and our collection of facts of the most
extensive character.

Now, these facts are words, or rather judgments
expressed in words; and since these are the outward
embodiments of thought, the reflections of
the passing phases of the mind subjected to the
restrictive conditions of our physical nature, it is
clear that, just as thought is progressive, and can
only be studied historically, so words also must be
subjected to an historic treatment. In so far as
thought is stationary, it is unconscious, and must
be treated physically like the rest of brute nature:
with consciousness, history begins. It is the same
with language: consciousness first shows itself in
the period of roots, and with this period accordingly
Comparative Philology commences. Behind
lie the unconscious, instinctive beginnings that
led to articulate speech, but our linguistic data do
not carry us so far; the investigation of this
primeval age of humanity belongs to physical
science, not to Glottology.

Here, then, is one of the boundaries of the
science to which I have alluded. Our data are
limited to the words that can be collected from the
mouths of living speakers, or have been committed
to the safe keeping of writing. It is only where a
group of cognate languages has changed but little
41that we can go back much beyond the invention of
writing. Practically, therefore, we are bounded,
so far as time is concerned, by the earliest written
records which we possess, whether in Egypt, Babylonia,
or China, or by a literature like that of the
Rig-Veda, which has been stereotyped by traditional
recitation. It is absolutely necessary that
our facts should be accurate, that is, that we should
know the exact forms and meanings at any given
period of the words with which we are dealing;
and this can only be done by the help of cotemporaneous
evidence, or by the inductions built
upon this. It has been found possible to construct
a dictionary of the primitive Aryan language; but
this is only because the cotemporary evidence
we possess of the different branch-languages of
the Aryan family of speech is sufficiently large to
enable us by the use of the comparative method
to determine what must have been the parent
sound, which alone could have given rise to the
several varieties of the same word. And, after
all, much in this dictionary must remain uncertain;
we cannot always be sure of the original
form of a vocable, and words possessed by the
parent language may often have been lost altogether,
or have left but slight traces behind them.
Of course, in this work of reconstructing parent
languages, or of probing language in general to
42its roots, we obtain additional light and assistance
from other sciences, such as psychology, prehistoric
archæology, or physiology.

From all this it will be evident to every one
what is the object and scope of Comparative Philology. 120
It is an historic science, which traces
the gradual evolution of human thought and
action as photographed in the enduring monuments
of language — the outward expression of
that thought and action — and which has its roots
far down in the dawning consciousness of primitive
man. So far as man is man, so far, that is
to say, as he has emerged from a mere brute life,
43and has awakened to consciousness, he has a
history, and that history may yet be recovered
either wholly or in part from a scientific study of
language. The facts with which this study deals
are words or stereotyped thoughts; these it has
to compare and classify, and thus determine the
general laws to which they are subject. The
general laws, made up of a variety of subordinate
ones, belong partly to psychology, partly to phonology;
the first lays down the conditions under
which the awakening and developing mind views
objects and their relations; the second the conditions
under which sounds are produced by the
human voice, and the mind is enabled to express
itself. Phonology is of the highest importance for
getting at the laws of speech, since it ascertains
the relation of sounds one to another, and thus
explains the changes and kinship of words; but
it must not be made synonymous with Comparative
Philology, as is so often implicitly done. It is
one of the chief and most valuable instruments of
the science, but it is not commensurate with the
science. The outward and physical is the most
accessible to observation, and, therefore, to comparison,
but words may often be phonetically
identical, which yet have nothing to do with each
other, like the sounds set apart by most languages
to denote “father” and “mother,” or the roots
44dhā, “to suck,” and dhā, “to place,” in our own
family of speech. This mistaken conception of
the place of phonology is the modern representative
of the notion that etymology is the
beginning and end of philology, and that when a
word had been tracked back through cognate
dialects to the most original form attainable,
nothing further was needed. This was the error
of the lexicographer, just as the phonological
misconception is the error of the grammarian.
Words are of no value in themselves except to a
dictionary-maker; they are only valuable in so
far as they reflect and embody thought; and the
object of a true philological etymology is to illustrate
or discover the laws which have governed the
evolution of thought, or rather the way in which
that evolution has been determined by material
and social circumstances. It is hardly likely that
we shall ever attain to a perfect knowledge of
these, and lay bare the whole mystery of the
origin of roots and the history of grammatical
relations. Should we do so, Comparative Philology
would become an exact deductive science,
and we should be able to predict the future destiny
of language and languages. Meanwhile, we have
to be content with an examination of the past
or the present, so far as this is open to us, testing
our conclusions by the facts of history and
45psychology, and by the laws which control the
utterance of sounds.

To explain more clearly what is meant, we may
quote, by way of example, the general law that all
languages have a period of roots, in which the
several distinctions between the parts of speech
lay undeveloped in a kind of embryonic common
sound. The empirical laws of phonology enable
us to trace the words of a civilised community
back to this common source; and the law itself is
verified by what psychology teaches us of the
gradual growth of the mind, and by the facts of
ethnology, with its illustrations of modern savage
intelligence, and of prehistoric archæology, with
its rough-hewn flints and other evidences of childish
ignorance.

Thus, on all sides, Comparative Philology is
brought into contact with its sister sciences. If
language is the reflection of common thought, it
is at once the product and the mirror of society.
It will, therefore, bear the impress of every movement
of society, and its phenomena consequently
will in large part be explicable only by means of
the social sciences. Why, for instance, is Lithuanian,
one of the least advanced members of the
Aryan family, more conservative in its retention
of many primitive grammatical forms than even
Sanskrit; while, as a general rule, tribes in a low
46state of civilisation, like the Ostiaks or the Bushmen,
are continually changing the character of their
idioms, so that in the course of a single generation
two neighbouring villages become mutually unintelligible?
Why, again, did the Northmen give
up their language in France, and retain it in
Ireland? Comparative Philology alone cannot
furnish, the answer. Similarly we must go to
physiology, if we would investigate the influence
of food and climate upon the organs of speech,
important as this question is to the philologist,
who finds that every Polynesian syllable must end
in a vowel, or that the Chinese have to turn every
foreign r into l before they can pronounce it, or
that Portuguese is more closely related to French
than the intervening Spanish, or that the Teutonic
coast population from Denmark to Flanders
drops the final d of a syllable, while English, on
the contrary, tends to introduce an expletive one, as
in sound and compound. One of the most important
problems which now awaits solution is to
explain the causes of that regular shifting of sounds
which words undergo in different cognate languages.
Why, for example, must a Latin d answer to an
English t and a High German z? 121 or what brought
47about the loss of a guttural before a labial in some
dialects, and the retention of it in others? Some
common cause must have been at work to produce
apa-s in Sanskrit, eau in French, and aua in the
Romansch of the Engadine, by the side of the
Gothic ahva, Latin aqua, Italian acqua, and
Spanish agua. 122 It is scarcely an answer to say
that this cause was laziness, the general principle
of phonetic change, because we want to know why
this cause should have acted in some cases and
not in others?48

It may be said that the reason will be furnished
by history. This is perfectly true. If we had a
complete history of the movements of society, we
should have a key to the changes of language
which are its expression and reflection. But such
a history would be nothing more than an exposition
of the laws which govern society; and
as we do not and cannot possess it, we must
endeavour to find out these laws by some other
method. When once the laws have been discovered,
that fragmentary and superficial series of
biographies which we term history can be applied
for the purpose of verification. It is thus that
the generalisations of an historic science are tested.
As in the physical sciences we verify our conclusions
by an appeal to experiment, so in philology, our
inductions can be verified by a reference to the
known facts of history. The clear traces of a
Teutonic influence in French point to a German
occupation of the country, and this we know
from history was actually the case. Arabic words
in Spanish afford evidence of a contact with the
Moors; and the relation of the Romance languages
to Latin necessitates philological conclusions
which are borne out by the statements of annalists.
Such general principles even as the ascription of
phonetic decay to laziness may be confirmed by
historical instances like the Norman conquest of
49England, where the loss of inflections was accelerated
by the attempt of a foreign population to
speak the language of the country with the least
possible trouble to themselves. From cases like
these, which can be tested by a direct appeal to
history, we may proceed by analogy to others in
which such a test cannot be applied. But it is
evident that the further we recede from cotemporaneous
history, and the more unable we are
to verify our inductions by its means, the more
hazardous and provisional will our conclusions be.
Hence some of the primary laws of the science
can best be obtained from a study of modern
European languages, though we must be upon our
guard against applying the results gained from
these to languages which are not occidental, or
which do not stand upon the same level of civilisation
and religious progress.

History is especially valuable in corroborating
the empirical laws which we discover, those, namely,
of which the reason cannot be given, but which
fall under some higher and more general law.
Psychology has more to do with the general laws,
in so far as these relate not so much to the external
accidents as to the inner meaning and structure of
language. In fact, just as a philosophy of history,
in which the attention is turned to the motives and
connection of outward events, depends upon
50psychology, so also does philology, which displays
the laws that govern our mental development, not
in action, but in speech. Physiology, on the other
hand, deals with the external, and is therefore
mostly applicable to phonology alone. Here we
have to ask it to help us in determining what
sounds may pass into one another, and under
what conditions they may do so. To look too
exclusively at this side of the science, however, is
to repeat the mistake of the last century, and
to see nothing but mechanical materialism everywhere.
We require the aid, not only of those
sciences which are concerned with the external
framework and circumstances of man, but yet more
urgently of those which trace the growth of his
spiritual life, like jurisprudence or history, however
much these may lead us back to a dim starting-point,
where the distinction between matter
and spirit, between nature and consciousness,
seems almost imperceptible.

But dim as it may be, we must remember that
it is a starting-point. Comparative Philology
cannot get beyond the range of its facts, beyond
the commencement of conscious articulate speech.
Language for it is not the language of gesture,
but the language of articulated utterance. The
investigation of language, in the wider sense, as
including looks, play of features, modulation of
51voice, and gesticulation, to say nothing of the
finger-language of the deaf and dumb, must fall
under a more comprehensive science. The examination
of this inarticulate speech belongs to physiology;
and Mr Darwin, in his work on the
“Expression of the Emotions in Men and Animals,”
has already broken ground in this direction.
But it is one of those sides of physiology which,
most directly bear upon our science, and from
which we may hereafter expect the most important
aid. In fact, if ever we are to solve the problem
of the origin, not of language in the philological
sense of the word, but of articulate speech itself,
the subject-matter of philology, it can only be by
special physiological researches upon this head.
Häckel has endeavoured to trace the earliest utterances
of man to the cries of the ape; and, as
Professor Benfey points out, the physical accessories
of speech, as we may call them, “make the
purely human origin of articulate speech more
easily intelligible; for we must certainly credit
them with the power of assigning to any sound or
combination of sounds the meaning which the first
man who joined together these articulations and their
accessories was impelled or intended to express by
them.” Looks and modulation of voice seem to agree
in all nations, gestures only in part, thus forming
the bridge by which we may pass over into spoken
52language, the dividing element in human history.
The first three are common to man and the lower
animals; articulate language alone, whatever may
be its ultimate source, draws the impassable line
between us and the beasts, and makes man man.
This is the justification of a science of Comparative
Philology taking its rank among the historical
sciences, and not being merged in a general science
in which the brute and the human are coupled
together.

In applying its laws, the practical rules to
be deduced from them lie upon the surface.
If the facts with which we start are judgments
expressed in words, it is obvious that
the grammar and structure of a language will
afford the only sound basis of comparison. It
is not mere sounds that we have to compare,
but the processes of thought involved in them.
Thought is relative, and these relations may be
viewed in different ways. Only those languages
which agree in their mode of viewing these
relations can be grouped together. When once
agreement in grammar and structure has determined
the connection of two tongues, we may
proceed to compare their lexicons. The first words
to be brought under scientific treatment are the
pronouns and numerals, which constitute a link
between grammar and vocabulary. They are the
53earliest attempts to reduce the abstract to the concrete,
to embody thought; and the need of their
frequent use will better preserve them than is the
case with, other words. At the same time, the
very frequency of their use subjects them all
the more to the influence of phonetic decay, and
so renders a knowledge of their history the more
necessary. Now, the history of a word can only
be made out by a comparison of dialects, and an
acquaintance with the older monuments of the
language; so that until we have traced back a
word to the most ancient form attainable, we have
no right to employ it for the purposes of comparison.
We may compare roots, but not derivatives.
Words derived from the same radical will
often assume different forms in different languages,
or even in the same one; while words derived
from different radicals will, on the other hand,
often assume the same form in different languages,
or even in the same one. Before we compare, we
must know the history of a vocable. It is equally
important that the words should be found in some
written language. In no other way can we obtain
documentary evidence of their older forms, and
compare the latter with the forms of the same
words in modern dialects. We shall never know
the roots of the Polynesian idioms, since we can
only bring dialects together which are still spoken,
54and the most primitive forms to which such a comparison
will conduct us are relatively modern. Similarly,
our area of comparison must be wide and
varied, and not confined to a group of dialects
which all flow from one and the same mother
speech, like the manuscripts of Sophokles from a
single tenth-century original. Unless we are aided
by the sub-Semitic dialects of Africa and the old
Egyptian, our comparative researches into the
Semitic family will remain as unsatisfactory as
would be the case with the Romance languages
were all the cognate idioms, past and present,
utterly extinct and lost. Written languages, moreover,
guarantee a systematic pronunciation. We
are not obliged to take our materials from one
observer who represents the French un by a, and
from another who represents it by one. But above
all, we must not compare roots together, unless
the meaning as well as the sound agree, 123 or apply
55to one group of languages the phonetic rules and
possible interchanges of letters which belong to
another. The last error is a fatal one, but is not
unfrequent under the disguised form of attributing
a phonetic law peculiar to a special language to
allied dialects or the common parent of them all.
Thus, because Sanskrit may drop an initial short
a, Pott assumes, in his theory of roots, that the
primitive Aryan could do the same; and the Latin
habit of changing s into r has been quoted by K.
0. Müller and others to support an extraction of
πελασγοί in Greek out of πελαργοὶ (from πέλω
and ἄργος = ἄγρος). 124

In conclusion, a few words must be said about
the name of the science with which we are concerned.
“Comparative Philology” is at once
long and misleading; it perpetuates the idea that
its subject-matter belongs to a higher and more
comprehensive philology. Apart from Comparative
Philology, however, there can be no scientific
study of articulate language; and if philology
means something other than this, it would be
absurd to rank the scientific under the unscientific.
56But this is what is popularly done — philology signifying
sometimes a dilettante acquaintance with
the canons of taste and polite literature, and, in fact,
with everything that is not the science of language;
sometimes classical scholarship, in which the correction
of a MS., or the close imitation of an
Augustan writer, is the highest result aimed at.
Now these are all very good things in their way;
but it cannot too often be repeated that they have
nothing in common with Comparative Philology.
Classical scholarship may, indeed, contribute much
valuable material to the science, so far at least as
Latin and Greek are concerned; bat even here
its supposed discoveries often tarn out to be erroneous
when investigated by the light of the comparative
method, and can seldom be received without
further examination, unless the facts are very
plain and self-evident. The particular can only
be understood in the light of the universal; and
the empirical rules derived from a careful comparative
study of some special language, indispensable
as such data are to the scientific philologist,
are still narrow, unexplained, and questionable.
We are often told that a comparative philologist
must be thoroughly acquainted with some of the
principal languages with which he deals, otherwise
the inner structure of the language will be
concealed from him, and he will be obliged to
57take his facts at second-hand, and thus be often
led into error. This is quite true; and the more
numerous the typical languages that are thoroughly
known, the better and more accurate will be the
work of the scientific student. But it must be
remembered, firstly, that if a specialist takes up
Comparative Philology as a merely subsidiary
matter, the minor details of his specialty, whether
it be Greek, or Sanskrit, or Hebrew, will assume
an unreal importance in his eyes, and the main
phenomena be correspondingly dwarfed; and,
secondly, that it is impossible for the student to
have anything like a close acquaintance with the
large number of languages with which he is
obliged to deal. As in the other sciences, so here;
division of labour is imperatively required, and
much of the material has to be received on trust.
Where this is done cautiously and scientifically,
where the authorities are critically chosen and
weighed, and where the comparison of facts is
large and wide-reaching, the chances of error are
minimised, and the single wrong fact is neutralised
by the many accurate ones. We do not require
a linguist, but a philologist in the true sense
of the word. As this sense, however, is unfortunately
misunderstood, I should prefer to use the
term Glottologist, and in the rest of these chapters
I shall speak of Glottology rather than of Comparative
58Philology. Glottology will be the science
of language, by which we are enabled to trace
the gradual growth of the mind of man, whether
displayed in the creation of language generally as
an instrument of intercommunication, and the
embodiment of the conception of the relations
between thought and the world, or in the triumph
of the will over the mechanism of the bodily
organs and the limitations imposed in turn by
them upon it, or lastly, in the evolution of the
religious idea — in other words, in Comparative
Mythology and the Science of Religions.59

Chapter II.
The idola of glottology — the laws of the science
determined from the aryan family alone.

In every science we must advance from the known
to the unknown. This can only be done by the
aid of hypotheses. These bridge over the gulf,
and are, as it were, so many imaginary circles,
half of which is filled up by facts already known,
while the remaining half is a purely mental conception,
which will, however, turn out to correspond
with objective phenomena should the hypothesis
prove correct. The younger the science,
the smaller will be the amount of known facts,
and therefore the greater the number of hypotheses
required. Now, in so far as these are the
product of the imagination, it is clear that wide
scope is given for subjective prejudices, false analogies,
and a distorted view of the evidence. This
tendency to error will increase with the increased
meagreness of the facts, and can only be checked
by enlarged knowledge and a critical comparison
of the theory with what is actually known. Hence
60as a science grows older, its ascertained laws become
more numerous, its provisional hypotheses
either passing into laws by a process of verification,
or being thrown aside for something that
will better stand the test of facts. Even discarded
hypotheses, however, have done a good
work. In so far as they had any facts to support
them, they helped to unify a string of isolated
phenomena, and to set the student on a definite
path of research. We cannot collect facts to any
solid purpose, or compare them afterwards, without
having some theory to guide us in our selection.
But good care must be taken to place all
such hypotheses upon their proper footing, to remember
their provisional character, and to compare
them again and again with the phenomena
that come before us. Too often they become
unverified assumptions, which we accept without
questioning, and thus exalt into scientific laws,
thereby vitiating our further investigations, and
falling into numberless false conclusions. In this
way what were intended to be mental landmarks
become what Bacon expressively called Idola,
empty assumptions and misconceptions, which
take the place of the true conceptions that correspond
with the order of existing things. Glottology,
I think, like all other new studies, will be
found to offer a plentiful crop of these idola.
61Partly the science is still overshadowed by the
false associations connected with the word Philology,
to which I alluded in the last chapter;
partly the interest of special portions of the science
— phonology, for example — are allowed to obscure
the several interests of the whole; partly it has
been forgotten how large is the mass of materials,
and consequently the results obtained in one department
have been supposed to be of universal
application; partly opinions which were necessitated
by the only evidence available when the
science was in its infancy have been adopted
without criticism, and regarded as so many first
principles which no one would dream of disputing.
It is time, however, that such questions
should be fully discussed. We have now become
accustomed to the idea of applying the scientific
method to language; a large body of classified
facts has been brought before our notice, which
is being increased every day, and the sister
sciences of ethnology, prehistoric archeology, and
comparative law, not to speak of psychological
and physiological discoveries, are ever throwing
fresh light upon the problems of Glottology, and
assisting us to verify the conclusions to which it
comes. Hence we are in a position to examine
anew the foundations of the science, and to determine
what are to be accepted as really the principles
62of Comparative Philology, and what is of
doubtful authority or altogether erroneous.

One of the first assumptions of the glottologist,
either openly avowed or unconsciously implied,
is that a scientific investigation of the Aryan
family alone will give a full and complete solution
of all the problems of the science of language,
helped out perhaps by a few illustrations from
non-Aryan dialects. 125 The causes of such an
63assumption lie upon the surface. Not only did
Comparative Philology begin with the Aryan
family; not only are its students members for the
most part of that family, and best and primarily
acquainted with some one or more of its dialects;
not only does the historical position of Europe
give to this group of languages an immediate and
practical interest; but still more it is here that
the facts of language are most numerous, and its
vicissitudes most accurately known, from the oldest
hymns of the Rig-Veda down to the newspaper
of to-day. When the great discovery of the affinities
of this group dawned upon Schlegel and
Bopp, and the commonest inflections of grammar
were traced from dialect to dialect and from century
to century, it was impossible not to believe
that what held good of the Aryan would hold
equally good of all other tongues. We can
only work by means of analogy, and there
seemed no reason for supposing that the phenomena
would differ in the two cases. Moreover,
there was the continual striving of the human
mind after unity, which would tend towards the
belief, unless disproved by fact, that all languages
64have radiated from a single centre; and tradition
and religious prepossessions had fixed that centre
in the East. In the enthusiasm of a new discovery,
bewildered by the vagueness of Indian chronology,
it was hard not to fancy that the primeval language
had been found in Sanskrit, or at least in
the parent Indo-European speech. It is to this
that we must ascribe the attempt of Bopp to
attach, the Polynesian idioms to the Aryan family.
Already the world had been accustomed to derive
all the languages of the earth from some common
ancestor, whether that were Hebrew as orthodoxy
ruled, or Basque with Erro, or Dutch with Goropius.
It was the Christian spirit that saw the
same blood, the same origin, and the same hope
in all men, in contradistinction to the pagan spirit
of classical antiquity, which localised its gods and
its institutions, and could discover in a foreign
language nothing but a “barbarous” jargon.
Everything seemed to favour the belief that the
new science had made its way back to the sources
of all living speech, or, at all events, to something
very near those sources, at a single leap. Every
day brought fresh proofs of the close affinities of
Greek and Sanskrit, of Latin and Gaelic; while,
on the other hand, it became increasingly evident
that many of the inflections, the origin of which
had hitherto been ascribed to nature or convention,
65had primitively been independent words. Was it
not clear, then, that Aryan speech itself had once
been in a condition similar to Turkish, if not to
Chinese? Here, therefore, the common starting-point
of all languages had at last been reached,
that plain of Shinar which ended in its Babel of
confusion. The idea was strengthened by the
fossilised antiquity of the Chinese Empire itself;
it was like some pterodactyl or ichthyosaurus happily
preserved in the rocks to tell us the character
of animal life in the liassic period. Accordingly it
was assumed without further debate that the Aryan
group of languages was the model of every other;
either they were all descended from a common
source, or, at any rate, were subject to identical
laws. Philology could offer no difficulty which a
fuller knowledge of Aryan would not solve.
Where, for example, was an explanation of the
Etruscan inscriptions to be found? In some Aryan
dialect, of course. 126 What was the original form
of all articulate speech? The verbal monosyllables
66to which the Sanskrit grammarians had reduced
the lexicon. How was the idea of action first
expressed? By attaching a pronoun to one of
these verbal roots. These and suchlike were the
answers readily given to the inquirer; and time
was required to learn that the inner mysteries of
a science cannot be so easily penetrated; that it
is not the first solution that comes to hand which
is necessarily the true one; but that the truth is
only to be gained by slow degrees, by the labours
of many students, and by the orderly succession
of hypothesis after hypothesis, until the right one
is at length hit upon. We are still too far from
seeing this. We inherit the opinions and idola
of our predecessors along with their method, and
it requires an effort to criticise what has been consecrated
by great names, and has become part and
parcel of our belief. Above all, the glottologist
has still to be trained for his work in the Aryan
family. Here alone are the materials sufficiently
large, clear, and certain; here alone have we the
immense advantages offered by a preparatory
knowledge of some of the languages to be studied,
and by the possession of monuments at once so
old and so perfect as the Rig-Veda; and here
alone have the facts been classified, their conclusions
drawn out in their full extent, and the
whole brought into scientific shape. The Semitic
67family is at once too small and too compact; its
branches do not differ more among themselves
than do the Romance languages in Europe; and
until its Sanskrit has been found, as it may yet
be in the old Egyptian or the sub-Semitic idioms
of Africa, we cannot get back beyond a parent
speech which is philologically late, and which fails
to offer that facility for comparison which is
needed by the young glottologist. As for the
other languages of the world, they are still, for the
most part, awaiting their Bopp. Something has
been done for the Ural-Altaic or Turanian family,
which embraces Finnic, Tataric, and Mongolian,
especially by Schott, and the cuneiform records
from Babylonia and Susiana are likely to lead to
important results by revealing the character of this
group of tongues at an early date. 127 Bleek, too,
has worked at the Ba-ntu of South Africa, 228 and
Chinese has been more and more attracting attention
68to itself. As yet, however, but little has been
done outside Aryan beyond the determination of the
most general conclusions, and much of that little
will probably have to be revised. Consequently,
just as Latin and Greek are still the basis of
popular education, it is in the Aryan family that
the glottologist will have to receive his training
for some time to come. Hence, when he begins
to deal with other classes of languages, his mind
is filled with certain prepossessions and beliefs,
which are likely to colour his researches more or
less. He naturally expects to find the same phenomena
and obtain the same results in his new
field of inquiry as those with which he is already
familiar. It is only after considerable experience
that he comes to see that the Aryan family is but
one out of many, and that in several respects its
character is altogether exceptional. The languages
of civilisation are not numerous. The case is still
worse if the student be unacquainted with any
non-Aryan dialect, or, at all events, only uses these
to illustrate the views he already holds. Unfortunately
this is what is only too common. Glottology
has for the most part been confined to Aryan
scholars, and consequently the laws they have
formulated, however true they may be of the
Aryan group itself, are not necessarily of universal
validity.69

Then, again, these laws are not always obtained
from a survey even of the whole Aryan family.
The modern languages of Europe, whether Romance
or Teutonic, afford us the most numerous
and the most certain data for our studies that we
can find. It is these, moreover, that furnish us
with the best means of verifying our theories.
They have, therefore, especially attracted the
notice of glottologists, and some of the most valuable
results of the science have been gained from
them. But it must never be forgotten that the
phenomena they represent are in large measure
unlike any that have ever occurred before throughout
the history of language. As we saw in the
last chapter, the subject-matter of an historical
science is continually incorporating fresh elements
with the process of time, like some organic growth;
and this is particularly the case with the languages
we are now considering. These modern dialects
have grown up in the midst of literature, and of
the influences inspired by the Roman Empire and
the Christian Church. The latter worked on the
side both of law and of religion — the most potent
influences to which society is open — and thus not
only filled the vocabulary even of Teutonic tribes
with Latin and Greek terms, but perpetuated a
popular knowledge of the Latin tongue itself, and
gave a Latin shape to the expression of popular
70thought. Literature kept up an artificial standard
of linguistic purity and excellence, and to some
extent prevented the natural progress of phonetic
decay and the rank growth of dialects. Shakespeare
and the Bible have stereotyped English not
less than Dante has determined classical Italian, or
than the railway, the telegraph, and the daily press
will arrest the further development of European
speech.

These considerations will explain how it has
come to pass that eminent philologists have committed
themselves to general theories which will
not bear a very close examination. Every one can
see the absurdity of supposing that the history of
the Aryan family faithfully represents in all particulars
the history of all other families of speech or
of language generally. No one, for example, would
argue that all civilised languages must be inflectional;
but when the opinion is not stated in this
broad way, it is very liable to escape notice, and
to be unconsciously assumed and acted upon. I
shall give two or three instances of this, in which
theories have been put forward, and are still commonly
held, which rest entirely upon the above
assumption. No canon is so often laid down by
glottologists as that the roots of all languages are
monosyllabic. And yet this assertion rests simply
upon the fact that such is the case in the Aryan
71family. It is true that Chinese may sometimes be
called in to corroborate, or rather to illustrate, this
belief; but then we are too little acquainted with
the primitive form of Chinese to say what was the
original nature of its radicals. And indeed, so far
from confirming the canon, the present character
of Chinese would rather tell against it, seeing that
the tendency of all languages is towards phonetic
decay and the loss of syllables; while Mr Edkins
would lead us to infer that the existence of longer
roots can still be detected in the living language. 129
The recovery of Accadian from the cuneiform records
of Babylonia — the importance of which, for
philological purposes, will make me often refer to
it — enables us to go back to a very remote period
of Turanian speech; and here, though the majority
of roots are monosyllabic, dissyllables like
dugud, “heavy,” gusur, “wood,” are by no means
unfrequent; and not only are there no data for
reducing them to monosyllables, but their obedience
to the law of vocal harmony would seem
absolutely to prevent such an analysis. Bleek's
investigations, again, into the Ba-ntu of South
Africa led him to the belief that polysyllabic roots
were rather the rule than the exception, many
72combinations of sounds which seem to us most
difficult being really the most primitive, while
mimetic roots — those, for instance, which denote
sneezing — would most naturally take a dissyllabic
form. These few facts are sufficient to show the
worth of the attempt made to pare down the Semitic
radicals to monosyllables in accordance with
the supposed law of monosyllabic roots. The task
is a hard one; and the disagreement among the
many eminent scholars who have tried it as to the
way in which the desired result is to be secured,
would of itself indicate the worthlessness of the
whole proceeding. One would slice off a letter at
the end of the word, another in the middle, another
at the beginning, while a fourth, with an
arbitrary eclecticism, would cut out letters in all
three places according to his fancy. 130 No one,
73however, can enter into the spirit of the Semitic
languages without seeing how entirely they are
built upon the principle of triliteralism. It is
implied in the whole theory of their grammar; and
74to imagine that it has grown out of something
essentially unlike, is to admit the possibility of a
change of mental view, which is inconsistent with
all the experiences of psychology. Triliteralism is
not the invention of Jewish-Arabic grammarians
of the tenth century; long before this, it was recognised
to the fullest by the literati of Assurbani-pal,
the son of Essar-haddon, whose lexical
and grammatical tablets are now in the British
Museum; and so clearly was the principle felt by
the people, that foreign words of one syllable,
which were borrowed by the Assyrians, had to be
Semitised by the addition of a consonant or semi-consonant.
The so-called biliterals are either the
result of phonetic decay, or else, as I think we
now have materials for proving, were loan-words. 131
The concave roots were really of triliteral origin,
and are primarily used as triliterals in Assyrian,
which possesses the inestimable advantage of a
syllabary; while such few compounds as really
exist go back to triliteral elements. The same
holds good of quadriliterals, which for the most
part have extended a vowel into a liquid; 232 and
75the occurrence of words of similar meaning which
differ in having letters of cognate sound merely
shows that certain letters interchange, not that
the word was originally triliteral. 133 No argument
can be drawn from old Egyptian, because, whatever
may be the relationship of the grammar, the
bulk of the lexicon is certainly non-Semitic, while
those few archaic words, like p'takh, “to open,”
and kh'tam, “to close,” which have Semitic analogues,
are triliteral. This is one example of the
false conclusions, the hasty neglect of evidence,
and the wasted ingenuity that have resulted from
the attempt to apply a law peculiar to Aryan to
other families of speech.

We may take another example from what has
been called the doctrine of roots. From an analysis
of Aryan it has been inferred that all roots
were originally verbal. This is certainly the case
in the Indo-European family, so far as our facts
allow us to see; and it seems to have psychology
in its favour. Language is the expression of
thought, but it is equally the expression of will;
and this was true more especially at first, when it
76was used in the service of the primitive wants of
mankind. Now will, as realised in action, is
essentially of a verbal character; hence it might
be supposed that the verbal nature of radicals was
a fact which held good not only of Aryan, but of
all other human languages. Not so, however. In
this case we cannot appeal to Turanian; for though
Accadian seems to have nominal as well as verbal
roots, our data do not carry us back to their original
content and meaning, and they may have
been a confused combination of nominal and verbal
elements, in which neither of the two had the predominance.
But the Semitic languages refer us
to nominal roots as decidedly as the Aryan do to
verbal ones. 134 The Semitic verb presupposes a
noun just as much as the converse is the case in
Aryan. Here, then, the conception of the object
lay at the bottom of the language — an intuition in
which the subject was ignored, or rather absorbed
into the object; subjective action and the development
of will being left out of sight. A similar
explanation seems necessary in regard to idioms
that have few, if any, abstract general terms, like
77Tasmanian, which could express an abstract idea,
such as “round,” only by saying, “like the moon,”
or some other round object. 135 The same deficiency
of abstract terms, that is, of words in which the
subjective predominates over the objective element,
marks many barbarous languages. The Malayans,
for instance, have words to signify different sorts
and parts of trees, but none to signify “tree”
itself; while the Algonquin can localise special individual
acts of loving, but cannot express the act
regarded in the abstract, when it is removed from
the category of space to that of time — in other
words, becomes an action which can be repeated
any moment, instead of being a definite objective
fact. 236 Similarly the Cherokee possesses thirteen
78different verbs to denote particular kinds of
“washing,” but none to denote “washing” in a
general sense. 137 Perhaps the verbal conception
upon which the Aryan languages are built pointed
out from the beginning the active, self-conscious,
nature-subduing character of the Aryan race, just
as we seem to trace the features of Judaism in the
determinate objective Semitic root and the resignation
of the subject which it implies.

The last example of the Idolum, or rather of its
effects, which I shall select, is the expectation of
finding elsewhere the same similarity of grammar,
if not of vocabulary, that exists among the several
members of the Aryan family. But the striking
unity of form that meets us in this family is really
exceptional, and will have to be explained hereafter.
The rule is rather change and diversity. 238
79The dialects of barbarian tribes are perpetually
altering. There is nothing to preserve them —
neither traditions, nor ritual, nor literature. The
savage has the delight of a child in uttering new
sounds, and exhibiting his power and inventiveness
in this manner, with none of the restraints
by which civilisation confines the invention of
slang to the schoolboy and the mob. In some
cases, among the Caribes of the Antille Isles, for
instance, where the wife was generally stolen from
an alien tribe, the language of the women and the
men is essentially different; and this, of course,
exercises considerable influence upon the language
spoken by the next generation. 139 Then, again, the
80barbarian is especially open to all the influences
of external nature, climate, food, and so forth,
with nothing to check the disintegrating effect
these may have upon the combination of sounds;
hence we are not surprised at finding the same
word, orang, “man,” appearing in the Polynesian
idioms under the various forms of rang, olan, lan,
ala, la, na, da, and ra. 140 Sometimes, moreover,
the custom known under the name of tapu among
the Pacific Islanders will have acted upon language,
according to which every word which contains
a syllable identical with that forming part
of the name of the reigning chief has to be
dropped or changed, and a new word adopted in
its place. Thus mi has been substituted for po,
“night,” in Tahitian, since the reign of Queen
Pomare; and a king with the name Tu caused
fetu, “star,” to be transformed into fetia. Professor
Max Müller 241 points out that a similar custom,
81called ukuhlonipa, prevails among the Kaffir
women, who are forbidden to pronounce a word
which happens to contain a sound similar to one
in the names of their nearest relations. This
usage, however, is but one phase of the way in
which the barbarian will play with language,
regarding it at once with superstitious awe, as
though the word in itself had an ominous power,
and as an opportunity of displaying his wit and
imagination. 142 Nothing is really harder than to
keep a language from changing where it is not
protected by the habits of settled life, especially
when men meet but seldom together, and when
the transparent uninflectional character of the
language allows every word, however formal, to
retain its full force and independent meaning.
The comparatively stationary nature of Eskimaux,
82which seems to have changed but slightly since
the time of Egede, and the astonishing identity
of dialect, more especially among the eastern
tribes, may be ascribed to the long winters, which
oblige the different communities to live closely
packed together. At all events, we are told that
since the institution of an annual fair among the
Rocky Mountains, the idioms of the eastern and
western portions of the nation, who at first were
hardly understood by one another, have become
more and more assimilated (Gallatin's “Synopsis
of the Indian Tribes of North America,” in the
Archæologia Americana, vol. ii.); while the phenomenon
noticed by Messerschmidt among the
Ostiaks, where villages a mile or two apart are unintelligible
to each other, 143 will be explained by the
agglutinative framework of the language. Where
the plural is expressed by an independent word
signifying number, one word will do as well as
another; for such a purpose we might use indifferently
“many,” or “multitude,” or “company.”
Nor must we forget how rapid are the social
changes that take place among savages, and language
is the expression of an existing state of
society. A tribe may be decimated by famine or
disease, it may amalgamate with another, or still
oftener it may be conquered and enslaved, and so
83forced in the course of a generation or two to
adopt the dialect of its conquerors. The vocabulary
of a savage is never very large; the strain
upon the memory of the learner, therefore, is not
great. All this is abundantly sufficient to show
that the persistency of form which we observe in
the Aryan family is altogether exceptional, due
partly to the semi-civilised life attained before the
first emigration set out, partly to a common
stock of traditions, partly to the inflectional character
of the language; and we cannot argue from
this to other families of speech where the rule will
be change and not fixity, variety and not similarity. 144

Besides these negative instances of the misconceptions
and erroneous generalisations which arise
from too narrow a view of Glottology, and from
the false belief that all its problems can be solved
by a study of the Aryan languages alone, an
affirmative instance will be needed to show how
the converse holds good, how the particular can
only be explained from the universal, the part
from the whole. We cannot understand even the
Aryan group aright, unless we put it in its proper
84place, and examine it in connection with the
general facts of philology. The original form of
verbal expression — that is, the representation of
the carrying out of will into action in time — is
ordinarily said to have been the immediate addition
of a pronoun to a root. This would hardly
be an adequate explanation, even were it true that
all radicals were verbal; and this, as we have
seen, is by no means the case. In Accadian,
lulim-mu is both “my king” and “I am king,”
though for the latter mu-lulim is the more usual
form, and in old Egyptian ran-i means indifferently
“my name” and “I name.” What is it,
then, that constitutes a verb? or rather, since
Glottology is an historical science, what is the
origin of the verbal idea? Now, the different
words and tenses of the Aryan verb have been
created by suffixing various pronominal and verbal
radicals, some of which belong to an older period
than others. Those moods and tenses which have
been formed by the help of another verb, such as
the Romance futures, the Teutonic perfects, the
Latin imperfect or future, or the Greek and Sanskrit
future and optative, are clearly of secondary
antiquity, and presuppose already existing verbal
forms. The aorists, again, and the presents with
extended bases, can hardly go back to the first
beginning of the verb. The reduplicated perfect
85affords room for doubt, and it may have been
coeval with such simple presents as ad-mi or
as-mi, in which the pronoun is attached to the
root without any intervening syllable. Granting,
however, that these simple presents are the oldest
forms of the verb — and their rarity and simplicity
of meaning point to this — we have not advanced
towards a solution of our question, What was the
original purport of the verbal idea? From the
Aryan alone we should be inclined to conclude
that it expressed present time, the most definite
possible conception of action, however, and one
which philosophy teaches us is among the latest
arrived at. Present time, moreover, implies a
knowledge at least of the past, if not of the future,
with which it may be compared; and some of the
lower races, like the New Caledonians, who cannot
be made to understand the abstract notions of
“yesterday” and “to-morrow,” are equally unable
to express the notion of “to-day.” The primitive
Aryan, therefore, if he began with the expression
of present time, must have stood on a high level
of culture. Here, then the study of the Aryan
family cannot give us the answer we require.
With Semitic, however, it is altogether different.
The Semite, who never had that sense of individual
freedom of will and action which distinguished
the Aryan, preserved with but little
86alteration the primitive vague conception which
underlay the verb. The so-called future or imperfect
of the Semitic languages is not a tense in
the Aryan acceptation of the word. It does not
express time at all, merely relation. Now, this
was originally the sole Semitic verbal form. The
other so-called Semitic tense is nothing else than
the participle, the nomen agentis, from which the
third person singular masculine can still be only
artificially distinguished, 145 and it did not take its
rise until what Ewald calls the Aramaic or second
period in the growth of the Semitic family. In
time this perfect, as it is commonly termed, came
to acquire a kind of present force, but though
more verbal in character, according to our Aryan
ideas, than the imperfect, it never was a tense in
the true sense of the word. 246 Where intercourse
87with a foreign people, as in the case of the Assyrians,
and to a lesser degree of the Gheez, brought
about something like a conception of verbal time
and mood, the varying vocalic forms of the imperfect
were appropriated for this purpose, but even
here with no very great strictness. A similar
device has been adopted in Arabic, helped out by
the use of other words like kad, “now.” In the
88Semitic family, accordingly, the original purport
of the verb was purely indefinite: it had no reference
to any particular time or mode; it did not
even denote action in general, but regarded the
act of the will as an affection of the object, not as
an exertion on the part of the subject. We may
compare the use of the Greek aorist in similes,
where it is a little remarkable that the verbal
89form which best exhibits the bare root should be
set apart for this aoristic or indefinite purpose.
More remarkable still is the usage of the polysynthetising
languages of North America, where the
idea of time or mode is altogether absent from the
verb, and personal relations are alone indicated.
For this purpose a most intricate and elaborate
machinery has been devised, and according to the
Baptist missionary Edwin James, the Chippeway
90Indian possesses no less than from six to eight
thousand verbal forms. So, too, in Eskimaux we
have such monstrosities as aglekkigiartorasuarnipok,
“he goes away hastily and exerts himself
to write.” (See Gallatin, “Trans. Amer. Antiq.
Soc.,” vol. ii., p. 176; Crantz, “History of Greenland,”
vol. i., p. 224.)

Much the same phenomenon reappears in
Basque, a different form being employed for
addressing a superior, an equal, a child, or a
woman, and in reference to an object in the first,
second, or third persons singular and plural.
Thus det is “I have it;” ditet, “I have them;”
dizut, “I have it for thee;” at, “I have thee;”
zaitustet, “I have you;” dizquizutet, “I have
them for you;” daunat replacing the last when a
woman is meant, and dayat when an equal. Here
the forms originated in the incorporation of the
objective and oblique cases of the personal pronouns,
for the most part before the root, which is
followed by the postfixed subject, a noun of number (it) being actually intercalated into the root
itself when the plural has to be signified.

The same fact meets us again in Accadian.
Here we have but two tenses, an aorist and a present.
The first is formed by the immediate addition
of the pronouns to the root; the second by a
vocalic prolongation of the root: thus, in-gin, “he
91made;” in-gine, “he makes;” in-gar, “he did;”
in-garra, “he does.” The present is formed in
the same way in the Tibetan dialects, and clearly
points out the priority of the aorist, from which
the idea of present time was obtained, with the
growth of experience and civilisation, by dwelling
upon the sound of the aorist. With the creation
of a present the aorist ceased to be aoristic, and
became a past tense. Thus Turanian bears the
same testimony as Semitic, and explains the original
nature of the Aryan verb; while the observation
of actually existing cases, like that of the
New Caledonians mentioned above, supplies the
historic verification of the theory, and throws a
new light upon the development of mankind.92

Chapter III.
The idolum of primeval centres of language.

Plato has laid down that the end of science, as
of philosophy, is unity; and he attempted to
anticipate the slow processes of modern induction
by discovering a master-science from which all
the others radiate. It would seem nowadays as
though the dream of the Greek thinker were in a
fair way to be realised. The physical sciences are
becoming more and more metaphysical with the
increasing transcendentalism of their highest laws,
while the historical sciences are growing more and
more physical as the interdependence of the two
is more clearly recognised. Science is beginning
to deal almost exclusively with force, in itself a
metaphysical conception; and the doctrine of the
conservation of forces, that is, of one invariable
whole which manifests itself under various interchanging
forms, is the keynote of modern research.
Whether, however, an ideal unity will
ever be attained, is a question which admits of
grave doubt in the face of the opposition and contradiction
93which lie at the foundation of the world,
of the separation of our several senses, of the
deficiency of our data and the limitation of our
positive knowledge, and of the mysterious but
impenetrable background which appears to lie
beyond the highest and primary laws. Nevertheless,
unity is the goal of every inquirer; it is
necessitated by the very constitution of the mind;
and in so far as thought is one, or rather, as the
way in which we are compelled to regard the phenomenal
world is the same, a certain kind of unity
is not only attainable, but necessary. We cannot
help believing that under all the variety that we
see there lies a hidden unity, and that that variety
is itself but a way of producing unity. If we are
to think at all, we must sum up the isolated
phenomena under general heads, we must discover
some similarity and order in them; and the more
nearly the mental order corresponds with objective
sequence, the more fully shall we satisfy the requirements
of science. But we must not forget
that the so-called laws of science are, after all,
only so many mental conceptions, the imaginative
framework which we fill up with the results of
our experience, or rather of the manner in which
we are obliged to look at things. Now these conceptions
are all alike in so far as they are thought,
and we can ideally sum up one conception under
94another, until at last we reach the highest and
most comprehensive unity. It is this that we call
the world; and the day on which this general
unifying conception was first struck out was a day
of importance in the progress of the human race.
The Greeks ascribed the discovery to Pythagoras;
and whether or not it was really invented by the
semi-mythical Samian, the Greek word was well
worthy of a nation of philosophers. Κοσμός, or
“order,” is the best and truest conception of the
universe that can be arrived at; it is the summing-up
of civilisation and civilised reflection in contradistinction
to the unreflective fetichism of the
savage, who can see nothing except caprice and
disorder around him. Unity must be found in
order, if it is to be found anywhere; it is just that
orderly arrangement of our conceptions, that successive
sequence and co-ordination of thought
which impresses itself upon the outside world,
that enables us to detect and name an unity
amid the everlasting flux of things. The Romans,
in this, as in most other intellectual matters, the
pupils of the Greeks, were content to translate
κοσμός, by mundus, in which, however, the reference
to well-disciplined arrangement was lost, and
replaced by an allusion to the neatness of personal
adornment. It was only for the needs of Cicero's
amateur philosophising, under the subsequent pressure
95of a dry scholastic philosophy, that the Latin
language yielded universum, universe; where all
the vivid concrete metaphor of fresh Greek thought
has had to make way for the barren abstraction
which simply affirms that unity is “one.” Our
own world is of far humbler parentage. It is
merely wer-alt, “men of antiquity,” from alt, the
Gothic alds, “antiquity,” and the Old Saxon wer,
“man,” which appears as a Gentile suffix under
the form ware, “men,” in words like Rôm-ware,
“Romans,” and has its kindred in the Gothic
vair, the Latin vir, the Greek ἥρ-ως, the Gaelic
fear, the Welsh, gŵr, and the Sanskrit vîr-as.
It is the same root, vri, that has produced virago
and virgin, as well as vires, “strength,” in Latin,
and vrîhi, “rice,” in Sanskrit, and whose primary
meaning is simply, “to grow.” The same idea
is contained in the word which is used instead of
the representative of wer-alt in the Gothic of
Ulfilas, mana-sedhs, “man's seed.” It is characteristic
of the practical, domestic, conservative
Teuton to have found his world in the past
generations of mankind, just as the richly-gifted
Greek, with his keen sense of the “golden mean”
of proportion and beauty, found his in the unchanging
order that underlies the whole course of
nature.

This instinctive desire to discover unity has had
96its effect upon the science of language. Here, as
elsewhere, the aim of science is to generalise, and
to show that there is order, and not caprice, among
the phenomena — classification, and not isolation.
But in this search it is bound not to go beyond
the facts and the strict inferences which may be
drawn from them. However tempting an assumption
may be, it must at once be set aside if our
data fail to give it plausibility, much more if they
actually tell against it. Now this, I think, is the
condition of a very common philological hypothesis:
that all languages are descended from one
original centre, or at most from two or three
centres. The assumption runs through a great
deal of our modern glottological reasoning. It is
implied in the ordinary classifications of languages,
which assume that families of speech analogous
to the Aryan are to be found all the world over.
Every idiom, ancient or modern, has to be brought,
willing, nilling, under some “family;” the admission
that a language may be sui generis is never
even dreamt of. We have even had a “Turanian
family” invented, into which everything that is
not Aryan or Semitic has been thrust, from Turkish
and Tamulian to Chinese and Red Indian. Now,
however, that the term “Turanian” is more properly
confined to the chain of dialects which
extend from the North Cape to Tungusia,
97embracing Finnic, Tatar, and Mongolian, to which
Basque also is probably to be added, and which in
some measure exhibit the same marks of resemblance
as the members of the Aryan group, a new
family has been brought into existence, to be called
Agglutinative, or Allophylian, or heaven knows
what. Scholars of the highest reputation have
endeavoured to derive Aryan and Semitic from a
common source; and, when all else failed, have
had recourse to the desperate expedient of making
them separate down the opposite slopes of the
same chain of mountains shortly after the invention
of a common tongue. Nay, attempts have been
made to show at least the possibility of one
primeval language, or embryonic language, on the
basis of the theory that would make a language
develop out of an isolating into an inflectional
stage, through an intervening period of agglutination;
and this, too, in a scientific spirit, and on
professedly scientific grounds, not after the manner
of Mr Forster, who has discovered the language of
Eden in the combination of a modern Arabic lexicon
with a rudimentary Chinese grammar; or of
Mr Edkins, who has lately left his special province
of Chinese to enlighten the world on the origin of
language. We are still too much under the influence
of early prejudices; we remember that
there was one speech before the confusion of Babel,
98and that in the old days of etymology nothing was
easier than to derive any one language from any
other according to fancy. A few such instances
as the resemblance of sanguis to the Mongolian
sengui, “blood,” or sex to the Hebrew shêsh, “six,”
were sufficient to settle the question. Then,
again, there is the analogy of the Aryan languages,
which all emanate from one source; and, as we
observed in the last chapter, the ordinary procedure
of Glottology has hitherto been to predicate
of language in general what has been found
true of Aryan in particular. The other sciences
have aided in the matter, tending as they do
towards a common point of agreement, and returning
to the primeval world-egg of Egyptian philosophy,
out of which all things have been generated
by a continuous process of differentiation. No
utterance of science is clearer than this, that all
which is now in being is the result of evolution or
development; that look where we will, to the
most distant horizon of space, or the dimmest
antiquity of time, there is no break, no void,
nothing but an unvarying, unchangeable continuity
of progress. Darwinism is the most fashionable
hypothesis of our day; and Darwinism is
supposed to imply a common type and a single
pair of ancestors. But the most advanced supporters
of the Darwinian theory have themselves
99been obliged to resign the homogeneity of the
human race so far as origin is concerned. The
very fact of the variation of species demands it,
as different varieties would have the best chance of
succeeding in the struggle for existence in different
parts of the earth; and sexual selection alone cannot
explain the black skin of the negro, whose
brain also contains the colouring pigment, or the
small stature of the Andamanner, or the curious
fact that the population of a continent corresponds
with the typical characteristics of its brute animals.
We have all been cast in the same mould, or, as
St Paul puts it, we have all the same blood; but
it does not follow that we all come from the same
ancestry, still less that all languages have radiated
from the same centre. In fact, if we are to believe
that articulate language began with the period of
roots, remote as this period is in the history of the
Aryan race, it is still not remote enough to allow
for the vast changes that have taken place in the
distribution of earth and water, in the fauna and
flora that inhabit the land, and in man himself in
all his variety of form and colour. The human
remains found in the upper levels of the Seine near
Abbeville, or the geological alterations that must
have happened since the entrance of the Papuan
race into their present habitat, supposing they had
migrated from a common cradle of mankind, seem
100irreconcileable with the limited antiquity of the
root-epoch of the Aryan languages. When the
latter first make their appearance, it is in the highlands
of Middle Asia, between the sources of the
Oxus and Jaxartes. Is it likely that the Dravidian
races, the “Dasyus,” whom they encountered
in India, or the tribes which they found existing
in India, in Asia Minor, and in Europe, could
have once belonged to the same race with themselves?
All things, of course, are possible in
science, and we are often called upon to believe
what is far stranger than the strangest fiction;
but where this is not the case, where there are no
facts to support the assumption, we must abide by
the ordinary analogies and conclusions of experience.
The class of languages nearest akin in
appearance to the Aryan is the Semitic; and here,
if anywhere, upon the received theory, we should
expect to find the most convincing proofs of relationship.
On the contrary, everything is against
it: the structure of the language, 147 the phonology
of the speech, 248 the conception of the grammar, the
101character of the lexicon, alike forbid the supposition,
unless we can imagine a psychological miracle,
by which the same mind was capable of originating
two things so contrary as the Aryan and Semitic
conception of the verb. 149 Add to this, that while
we first meet with the Aryan in the Hindu Kush, the
earliest revelations of Semitic speech point unmistakably
to the deserts of Northern Arabia. The
theory of common primitive centres breaks down
at the very threshold.

I have more than once said that, in studying
Glottology, we must not go beyond our facts; and
the statement, simple as it seems, cannot be too
often repeated. Now our facts, scientifically considered,
are, firstly, similarity of general structure
in language; secondly, similarity of grammar both
in form and meaning; and thirdly, a regular and
uniform interchange of phonetic sounds between
the languages we are comparing. When once a
sufficient number of instances have shown that a
102certain letter in one dialect is replaced by a certain
other letter in another dialect, we must never
admit any violation of the rule unless it can be
explained by the action of subordinate laws; and
the explanation of these interchanges of sound and
their mutual relationship is part of the duties of
philology. In addition to these facts, which
belong for the most part to the province of
phonology, a fourth fact will be similarity of
signification. Two words may conform to all the
requirements of Grimm's law, and yet have
nothing to do with each other. Ὁδὸς and ἕδος,
solea and sella, for instance, both point back to a
root, sad, but there is no common idea that will
allow us to bring them together, or from which we
can derive them; and the attempt to do so is as
futile as to reduce the various incompatible meanings
of a Semitic radical under one head, or to find
some single fundamental conception for the numberless
significations attached to the same sound in such,
languages as Chinese or old Egyptian, where yû
means at once “me,” “agree,” “rejoice,” “measure,”
“stupid,” and “black ox;” or ta, “thou,”
“gift,” “direction,” “corn,” “drop,” “type,”
“tear,” “heap,” “stick,” “health,” “head,”
“throne,” “man,” “assembly,” “wicked,”
“navigate,” “steal,” “burn,” “carry,” and
“give an account.” Such, then, are the facts
103with which Glottology begins, and the lower
empirical generalisations so derived furnish the
means for arriving at those higher and wider
laws which are the ultimate object of the science.
Beyond the facts we can never get, at least if we
wish to obtain valid conclusions. But similarity,
the comparison of the like with the like, is what
lies at the bottom of them all, and hence, where
he cannot find a similarity which can be scientifically
proved, the glottologist must resign an
opinion, however plausible. This is precisely the
case with the subject of this chapter. The amount
of likeness in sound, meaning, and relation which
is sufficient to establish a common origin between
various dialects is the exception and not the rule
in language. A general likeness, of course, there
must be, otherwise the science of Glottology would
be impossible, since the subject-matter of each
science must be of the same character; but this
general likeness results from the fundamental
identity of the human mind and human experience,
and of the physical organs which determine
the limitations of articulate speech. 150104

When we come to look into the facts, we find
that, so far from supporting the hypothesis of a
small number of primitive centres of speech, they
are all, so far as they go, on the opposite side.
We have already disposed of the alleged common
origin of Aryan and Semitic in the last chapter;
we need only add the significant fact, that a closer
analysis, instead of confirming the belief in the
original identity of the Aryan and Semitic numerals — one
of the chief arguments in favour of the
idolum we are now discussing — has shown that
they are of wholly different origin. The coincidence
of sound between the Hebrew shêsh, “six,” and
shêba, “seven,” and the Sanskrit shash and saptan,
had led to their being identified; and to the further
attempt to compare the Heb. ékhad, “one,”
with the Sansk. êkas, and kam-esh, “five,” with
the Sansk. pan-chan (quin-que). But the Arabic
sittuñ and Eth. sedestu prove that the primitive
form of shesh contained a dental, derived probably
from sad-sad, a by-form of sal-sal, which appears
105in shalos, “three,” while the Zend kshwas points
as clearly to an original initial guttural, justifying
Professor Goldshïcher's view that it stands for kakatwar,
“(two) and four.” Saptan seems a participial
form from the same root that gives us ἕπω in
Greek and sequor in Latin, and thus to have
signified “following,” while no amount of reasoning
can ever get rid of the final guttural of the
Semitic numeral, which is best traced back to
arba, “four.” Ekhad, I believe, is from a foreign
(Accadian) source; at all events, the vowel at the
beginning is prosthetic, and cannot be compared
with the initial syllable of ê-ka, which, when compared
with u-nus, ἑ-ν-ο-ς, Gothic âi-n-s, and the
Sanskrit pron. ê-na, “that,” would appear to be a
principal part of the Aryan word. To connect
kamesh and panchan is comparison run mad. The
whole argument rests upon the same unscientific
comparison of words superficially alike that was
the staple of the etymologising of the last century,
and the conclusions arrived at are equally valid.
As well might we join the Basque sei, “six,” with
sex, or bi, “two,” with bini. 151

When we pass from the Semitic to other groups
of languages, the difficulty of connecting these
with Aryan becomes even greater. First of all
they are lumped together in one mass, or at best
106divided into agglutinative and isolating, and then
it is asserted that the parent Aryan language had
passed through both these two stages before it
reached the inflectional stage, and that it was
during the first of these periods, in other words,
during the epoch of roots, that it formed one with
all the known languages of the world. But, passing
by the assumption of this graduated development,
which we shall examine in a future chapter,
we may well ask how such a fact, if fact it is, can
possibly be known? Nothing is more deceptive
and dangerous, it is agreed on all hands, than the
comparison of words only, unless we are guided
by rules like Grimm's law, more especially when
the original meaning of the words is vague and
obscure. In order that our conclusions shall be
sound, we must begin by the comparison of the
grammar; and in the present instance, such a
comparison is excluded by the nature of the case.
In fact, the whole attempt rests upon air; its sole
basis is the inherited prejudice in favour of a
common primeval tongue. It cannot be urged
that the readiness to change which distinguishes
savage dialects, as we saw in the last chapter,
gives any countenance to the maintenance of the
theory. In the first place, the extent and nature
of the changes are unknown, and science does not
allow us to spin theories out of what may be; in
107the second place, however great may be the change
in the vocabulary, the manner in which the mind
views objects and their relations, that is to say,
the structure and grammar of the language, remains
unaltered; in the third place, war and pestilence,
the chief instruments of change, do not introduce
any new language, they only bring about the
extinction of one idiom and the destruction or
diminution of another; and lastly, the peculiar
language of the women and of the nursery is at
once conservative and confined to the lexicon.
Where two Manipuran villages are unintelligible
to one another, it is on account of changes in pronunciation,
in idiom, and in vocabulary, not in
the grammatical forms. It may be doubted, moreover,
whether we should not always be able to
recognise some, at least, of the ordinary terms of
daily life in two dialects which were once closely
united, however great their divergence may have
been. In spite of the wide interval in time, space,
and social relations, we may still detect several
words of this sort which are common to Accadian
and Basque. Thus aria, “water,” and Basque
ura, eri, “city,” and Basque hiria, seem to claim
relationship. This is still more true of Accadian
and the semi-barbarous idioms of Northern Russia
or Tartary: pi, “the ear,” for example, reappears
in the Wotiak pel; kats, “two,” is the
108Esthonian kats; dingir, “god,” is the Turkish
tengri, “heaven.”

How far grammar is changeable, how far it may
be affected from without, is a matter which we
shall have to investigate hereafter. For the
present, we may acquiesce in the received doctrine
that a grammar is never borrowed, even though
the dictionary may almost entirely consist of foreign
words.

However, it is not enough to overthrow the
arguments brought forward by the homogenists;
we require positive instances on the contrary side;
and these, I think, we have. How else can we
explain, consistently with the given facts, such
phenomena as the ancient languages of Etruria
and Lykia? It is said that our inability to decipher
the Etruscan inscriptions is a disgrace to
philological science. So it would be if they fell
within the province of Comparative Philology, if,
namely, there were any other known language with
which they could be compared. If such does not
exist, the taunt is undeserved. And it seems to
me that this is the conclusion to which every unprejudiced
thinker must be driven after the vain
attempts that have been made to find the key in
every possible or impossible language. The latest
decision is that they belong to the Indo-European
family, because the language of them is inflectional;
109but surely the decision refutes itself. Were they
Aryan, they would have been explained long ago.
If any one thing distinguishes an Aryan language
more than another, it is its persistency of type,
its general fixity of grammatical form, its common
residuum of roots, which allows us to determine its
character at a glance, whether among the valleys
of the Caucasus or on the shores of the Atlantic.
The characteristics of the European portion of the
family are even more distinct; and we may well
ask, whence did the Etruscan acquire its peculiar
features? We know that it descended into Italy
from the north, and hence, if Aryan, could only
be connected either with Keltic, Teutonic, Slavonic,
or Thracian. With the three first every one allows
that it has nothing in common, in spite of Sir W.
Betham and Dr Donaldson; and little as we know
of the last, we know enough to deny its kinship to
Etruscan. The Rhætian Alps are now inhabited
by a population which speaks Romansch and
Ladin; but these are Romance dialects, and in
spite of many strange-sounding Etruscan-like local
names — Velthins, and the like — all the researches
of Dr Freund and Mr Ellis have failed to discover
a single Etruscan word in the modern idioms.
The Etruscans may have been the bronze-men of
the Swiss lakes, or their predecessors of the
Neolithic age whose pile-dwellings in the north of
110Austria have yielded wheat and coral, evidences of
Eastern intercourse; at any rate, except in Italy,
where they had the good fortune to come into
contact with Greek civilisation, they have passed
away and left no trace behind them. Unlike the
Aryans, they were unoriginative and receptive;
and not only did they receive into their vocabulary
Greek words like βροντὴ (in the phrunt-ac of the
Pisaurum inscription) or αἰών (in aiv-il, “age”),
but even the Latin inflections of a proper name,
Velthina, Velthinas, in the late inscription of
Perugia. The native inflections, however, were of
a very different character; the patronymic al, the
termination isa to express “the wife of,” the
verbal e and ke, and the nominal l, ls, n, k, are
all non-Aryan either in form or use. 152111

Unlike the Etruscan, a few of the shorter
Lykian inscriptions can be read, thanks to the
Greek legends attached to them. Here again we
have an inflected language; which has accordingly
been added to the Aryan stock, with the support
of such forms as prinafatu, “he made,” by the
side of prinafūtu, “they made.” The nearest Aryan
language fixed upon is Zend; but a certain admixture
of Semitic is also assumed! The general
character of Lykian, however, so far as we know it,
as displayed in the nature of the vocabulary and
grammar, is so clearly and widely removed
from that of the Aryan family, that an endeavour
has been made by Mr Ellis to attach it to certain
of the Caucasian idioms, but with scant success.
The language is inflectional, it is true; but the
inflections are not those of the Indo-European
group. If the supposition be hazarded that it
branched off from this group, or rather from
some remote ancestor of this group, long before
the days to which Fick's dictionary and
Schleicher's grammar of the parent-speech refer,
we can only reply that there is not a single
112fact to support the belief. Behind that period we
know only of the period of roots (if, indeed, such
ever existed), and an intermediate epoch during
which, the inflections of the parent-speech were
being settled; but neither these roots nor these
inflections are to be found in Lykian. The root-period
must in any case have preceded the branching
off of Lykian or its presumed ancestor: how
is it then that the Lykian radicals are not Aryan?
Besides, we may ask, as in the case of Etruscan,
whence did Lykian come, and what are the genealogical
links by which its application to Aryan
are to be established?

Another inflectional language not comprised
in the Indo-European family is the Georgian.
This is still spoken, and consequently we are not
reduced to the allowance of forms and words that
can be extracted from inscriptions. The language
has a fair antiquity, if it can be shown, as M.
Lenormant believes, that the cuneiform inscriptions
of Van are written in a cognate dialect. However
this may be, Georgian on its inflectional side differs
remarkably from the Aryan in several particulars.
Thus the sign of the plural, bi or ni, is
inserted between the root and the case-endings,
as thavi, “head,” genitive thavisa, plural thavebi,
thavebisa, assimilating the language to the Turanian
family; the pronouns have a demonstrative
113and a copulative case; the ordinal numbers are
formed from the cardinals by the prefix me; and the
verbs incorporate the objective pronouns, and are
able to increase themselves by means of unmeaning
letters. Like the grammar, the roots of the
language show no affinity to the Aryan. Georgian,
with its allied idioms, is sui generis; and if we
abide by the simple facts, instead of following
delusive analogies and prepossessions, we shall
recognise here also a new independent class of
languages. The same must be said of the Caucasian
dialects. Anomalous groups of speech as
distinct as Abasian and Mingrelian exist side by
side with an Aryan dialect so nearly allied to
Persian as the Ossetian of the Iron; and in spite
of Dr Hyde Clarke's attempt to compare them
with the dialects of Tibet, the Caucasian group
remains a mixture of languages that bear no,
resemblance either to one another, or to the other
known idioms of the world. In the words of the
homogenists they are still “unclassified.” The
only inference that can be legitimately drawn
from the facts, without stepping beyond them, is
that the mountain fastnesses of the Caucasus —
the “snow-white” peaks, as Isidore interprets the
word — afforded refuge to the last relics of many
old tongues which have otherwise disappeared,
just as Basque has preserved itself in Biscay and
114Gaelic in the Highlands. The social revolutions
to which barbarous and semi-barbarous tribes are
exposed, particularly through their limited numbers
and the reverses of war, more than account
for the entire loss of languages; and when we
consider the great antiquity of man, as revealed
by geology, by ethnology, by glottology itself,
together with the vast extent of area over which
he had spread himself at a remote period in scattered
isolated bands, with no protection against the
beasts of the forest except miserable chipped flints,
no protection against the excessive cold of winter
except the skins of wild animals and the shelter
of a cave, our sole wonder must be, not at the
diversity of languages, but at the paucity of the
wrecks of ancient speech that still remain spread
over the face of the earth. The modern races of
mankind are but the selected residuum of the
infinitely varied species that have passed away:
the same surely will hold good of language; and
we ought no longer to be surprised at the multitudinous
variety of dialects found in North and
South America, in Australia, in the islands of the
ocean, or in the continents of the Old World, but
be content to believe that they represent but a
small part of the extinct essays and types of
language which have gone to form the language-world
of the present day, like the numberless types
115that nature has lavished since the first appearance
of life upon the globe. Manifold must have been
the earliest attempts to form articulate speech, to
utilise the mouth for the purpose of supplying daily
wants. Man is a social animal; comparative law
and comparative ethnology first introduce him
leading the communistic life of bees, out of which
the idea of individualism grew up with the progress
of civilisation. Intercourse by means of
gestures and signs could not long have been adequate
to the needs of the community; the hands
were wanted for other purposes, even if, as Helvetius
held, it was through them that man became
man; and accordingly the natural powers of producing
sound that lay in the voice would have
been employed to procure what one man required
from another. Whether or not, however, language
was at first communistic, like everything else, and
not individual, is a question which we have no
means of determining. This much is clear, that at
a certain period of social life, the impulse towards
the expression of articulate speech must have
become irresistible, and primitive man would have
delighted in displaying his newly found power, as
much as the modern savage or the modern child,
the best representatives we now have of primitive
man. The child is never tired of repeating the
words it has learned; the savage and the schoolboy
116of inventing new ones. Indeed the slang of the
school is the reaction of the still unextinguished
feelings of primeval barbarism against the restraints
of civilisation, and the strange interjectional
“tongues” of religious enthusiasm are
the return, under the pressure of strong emotion,
to the original state of productive energy. After
all, the barrier between interjectional utterance
and articulate speech is very slight, and it must
have been slighter when both were but the outburst
of natural feelings and the expression
of wants differing in degree only, and not
in kind. Can the emotion that prompts the
savage to shout be said really to differ from the
sense of power and life that makes him turn his
shout into a significant word? In both the object
is the same; in both the means of attaining that
object by the use of the lungs is the same. Surely
language originated in the desire to speak, in the
pleasure felt in the very act of inventing sounds;
and to limit such invention, such desires, to a
single body of men, is as reasonable as to hold
that the manifold songs of different species of
birds have all developed out of some original one,
or at most out of two or three.

If there is one lesson that modern savage life
teaches more emphatically than another, it is that
in a so-called natural state separation and hostility
117are the rule. Mankind live apart in numberless
small groups or families, which have no connection,
except perhaps a hostile one, with one
another, and which continually tend, unless
checked by other circumstances, to become narrower
and smaller. We see them, too, in a constant
state of flux and migration, exposed to all
the dangers of famine, disease, and want of wives.
Language, the product and mirror of society,
faithfully reflects this state of things. In Colchis,
Pliny says (vi. 5), there were more than 300
dialects. Sagard in 1631 states, that among the
Hurons of North America, not only is the same
language hardly to be found in two villages, but
even in two families in the same village, while
each of these multitudinous dialects is changing
every day. Waldeck asserts that a dictionary compiled
by Jesuit missionaries in Central America
became useless within ten years; while Captain
Gordon tells us that “some” of the Manipuran
dialects “are spoken by no more than thirty or
forty families, yet [are] so different from the rest
as to be unintelligible to the nearest neighbourhood.”
Spix and Martius bear the same testimony
in regard to the languages of South
America, in reference to which Humboldt (i. 298)
writes, that together with a great analogy of
physical constitution, “a surprising variety of
118languages is observed among nations of the same
origin, and which European travellers scarcely
distinguish by their features.” Now we may
fairly take modern savage life as representing the
condition of man when he first comes under the
notice of philology, remembering, however, that
all the traits we have just been alluding to must
have been exaggerated at that early period, when
the human race was on a lower level of culture
than the most degraded barbarian of to-day, and
necessarily existed in scantier numbers. But we
need not even go to savage life to exemplify what
is the normal condition of spoken language.
Every dialect that we meet with gives its evidence
against the actual reality of those ideal centres to
which we would relegate the various languages of
the world. Dialects and diversity are the natural
order of things; and as soon as the coercive hand
of a literary civilisation is taken off a language, it
at once breaks out into a plentiful crop of dialects.
Thus in Greece alone there are (or were a few
years back) no less than seventy different dialects.
Many of these, no doubt, were new creations;
that is, they have originated among the isolated
communities of Greece since the tenth century,
testifying to the perpetual creativeness of language
when left alone. But others will go back
beyond the rise of the literary language, which is
119but one out of many dialects, though, selected by
circumstances as the standard of the rest. Dialects
are the material out of which the idioms of the
court and of the book-writer have been formed;
they reach back as far as Comparative Philology
allows us to carry our investigations. We may,
indeed, conceive of a time when, in the Aryan
family, for instance, they did not yet exist, and
picture to ourselves some parent-speech which,
held them, as it were, in embryo; but we must
not forget that such a parent-speech is altogether
ideal; that, so far as our data go, they presuppose
the existence of dialects, and that the attempt to
explain the laws of lautverschiebung by original
indeterminate sounds, out of which the various
letters which correspond to each other in the
several branches of our race were derived, although
possible, is neither demonstrable nor satisfactory.
In fact, as soon as a language ceases to be confined
to a single household, it breaks out into
varieties. Every family has its peculiar pronunciation,
its favourite words. And the period in
which language first becomes an object of study
to Glottology is one of scattered and isolated
communities. How these first acquired articulate
speech is not for the glottologist to ascertain;
for him the origin of language means the analysis
of the words that we at present possess, the
120determination of the monuments that have come
down to us. This much, however, is clear; that
the beginning of articulate speech, the beginning
of that language with which he has to deal, is not
coeval with the physiological beginning of man;
that it is a product of society; and that as society
in those primitive times was infinitely numerous,
so also were languages. To derive one language
from another is to derive one community from another;
and where we find all living at once separately
and simultaneously, without any mark of
priority or derivation, such a procedure can obtain
no scientific sanction. We may make ideal centres,
like the ideal types of natural history, to which to
refer the different members of one or two so-called
“families;” but these centres will ever remain
ideal: for the philologist dialects exist from the
beginning. Nor can we exclude the possibility
that some at least of these dialects never had
philologically any connection with each other; but
that a common climate, common food, and common
conditions of life produced similar linguistic
phenomena among the isolated communities of a
given area, and that a similarity of germ necessarily
brought about a similarity of development.
The physical descent of certain tribes from a
single household must be kept carefully distinct
from any philological descent, since, as far as
121Glottology is concerned, language is posterior to
the physical beginnings of man. When we consider
the immense multitude of savage idioms,
and the changes which, these are always undergoing,
we may form some idea of the infinite
number of tongues that have been spoken since
the first epoch of language, and have left
no trace behind them. Here and there a few
have been stereotyped and preserved by a happy
selection; here and there relics of others may be
detected; but the large majority have perished more
completely than the animals of geological antiquity.
When mankind first awakened to linguistic consciousness,
each isolated community had its own
means of intercommunication, its own dialect if
you like; and from a combination of some of
these which lived near one another, or were
brought together by war or migration, the dialects
which make up a “family” were originated.
Instead of deriving the latter from a common
ancestor, a common centre, the truer account
would be that they were slowly evolved out of an
amalgamation of pre-existing dialects. Even the
imaginary root-period of Aryan speech cannot
disguise this; we need not go further than Greek
to discover roots which exist in no other cognate
dialect, or which, like ῥαχις, σιγὴ, θεάομαι, τρέμω,
νεφρός, τέμνω, are found but in one or two. How
122can this be explained upon the hypothesis that
all the linguistic material of the Indo-European
tongues has been developed out of one original
common stock-in-trade of radicals?

Indeed, if the history of language shows anything
clearly, it is the exact converse of the theory
usually maintained on this subject. The tendency
of time is to unify what was originally separate,
not to multiply what was originally one. Every
war among savages which ends by the subjugation
of a tribe and the extinction of its language justifies
this assertion. But its full truth is not seen
until we come to examine the records of civilisation
— that is, until we pass beyond the period of
barbarism in which language arose. Were the
ordinary hypothesis correct, barbarism would show
union, civilisation disunion. But the contrary is
the case. All the social conditions of civilised
life tend to break down dialects, to assimilate
languages, and to create a common medium of
intercourse. The Macedonian Empire spread a
common language through the East; the Roman
Empire still more effectually stamped out the
various idioms of the West, until a second period
of linguistic disunion was brought about by the
return of barbarism with the invasion of the German
nations. The Church alone, the sole representative
of civilisation, continued to have a common
123language. In fact, the more intense and
extended the civilisation, the more impossible is it
to keep up a diversity of tongues. The one unites,
the other disjoins. A common government, a
common literature, a common history, a common
law, all require a common language. The material
triumphs of the present century — the railway,
the steamer, and the telegraph, with increased
facilities of travelling and intercommunication —
all emphatically tend in the same direction. Above
all, the same holds good of commerce, the mainstay
of our modern civilisation, which is gradually
absorbing the whole world, and carrying with it,
wherever it goes, the languages of the chief
trading nations. “One coinage and one language”
is a cry now heard often enough. Small
nations, like the Dutch, find it absolutely necessary
to be bilingual, if not trilingual; and the
children in the schools are regularly taught to
speak some other language besides their own,
commercial reasons making English especially
favoured. Politics, too, look the same way. The
desire of unification, which has been satisfied in
Italy and Germany, aided by compulsory education,
is fast destroying the local dialects of Europe;
and already the vanguard of democratic socialism
and sameness have disowned the distinctions of
language, just as they have disowned the distinctions
124of race. Such is the end of that cry of
nationalities which shook Europe so short a time
ago: in the midst of their successes, the migration
of his countrymen to America is practically
disclaiming the words of Arndt, “So weit die
Deutsche Zunge klingt;” and the Teutonic population
of Alsace has preferred exile to reunion with
Germany. In spite of the efforts of philologists,
Welsh is rapidly disappearing from Wales, and
Gaelic from Scotland; while German alone is
heard in the schools of the Engadine, and French
in the schools of Brittany. The fact that the
revival of Flemish has been the work of the literary
classes, shows its artificial and hollow character,
and is of itself a proof how thoroughly the
attempt is contrary to the spirit of the age, when
the language is preserved, not on account of its
utility — the sole foundation of the continuance of
a language — but because it is regarded as a literary
curiosity, a philological plaything. The cry
of nationalities was really a backward step; it
was the reaction against the bourgeoisie of the
French revolution, and a revolt against the old-world
diplomacy that parcelled out anew the empire
of Napoleon.

To sum up: Instead of maintaining the existence
of a few original centres of speech, the truer
view would be that languages at first were infinitely
125numerous and diversified, being the natural
and spontaneous outcome of the powers, the feelings,
and the needs of primitive man, just as
much as the formation of flint tools or the ornamentation
of sun-baked pottery, and that they
have gradually diminished and disappeared through
the course of ages by a long process of natural
selection, civilisation finally threatening them
with utter extinction, and tending to reduce their
number to the smallest possible cipher, if not
finally to one universal medium of intercourse
between man and man.126

Chapter IV.
The theory of three stages of development in the
history of language.

We are often told that one of the chief results of
the science of language has been to show a continuous
and regular development in the history of
speech: first an isolating stage, or period of roots,
when the position of the word alone denoted the
meaning of a sentence without the assistance of
any (auxiliary) signs of relationship; then an
agglutinative stage, when these auxiliary marks
were added, each, however, remaining a fully
significant, independent word; and lastly, an inflectional
stage, when the auxiliary marks have
lost all independent meaning, and have become so
many inseparable signs. The final stage has a
further tendency to analysis: the inflections are
broken down, and the decayed compounds are used,
as in English, to express singly and independently,
by the aid of position, the various relations into
which a sentence may be resolved. The analytical
period differs from the isolating, in that in
the latter each root is a sort of germ which contains
within itself every kind of mode and relation,
127while in the former the germs have been,
broken up into their elements, and these are
represented by words, each of which is a relic of
a preceding era of inflections. 153 The three stages
are supposed to answer to the solitary individual
life the first men are imagined to have led — contrary,
however, to the communistic beginnings
which comparative inquiries now assign to them
— to the family and tribal life of nomads, and to
the social life of the civilised citizen. Chinese is
taken as an illustration of the first, Turkish of the
second, and Sanskrit of the third.

I have said, in a former chapter, that the starting-point
of glottology, the ultimate fact with
which it has to deal, is thought expressed in
speech. This is more accurate than the ordinary
view, which makes philological science begin with
the word. We may ask, “What is a word?” and
the only answer we shall get which will cover all
words alike is, “Meaning combined with form.”
128This is nearly the same as saying that it is thought
expressed in speech; but then what becomes of
such words as auxiliaries and conjunctions? It is
certainly difficult to detect much meaning in such
particles as “and” and “or;” and logic tells us
that the copula “is” represents simply the act of
mental comparison. Again, are interjections to
be considered words? In this case it would be
very hard to define the significations of “oh!”
or “alas!” By a word, therefore, a definite conception
must be intended to be understood; and a
conception must be subject to the relations of time
and space. Now every conception is the result of
a judgment, the decision that such and such particulars
are compatible with one another: when
expressed in language, it is the shorthand form of
a sentence or proposition. A difficulty, however,
arises in the case of the verb. The verb, as its
name implies, is pre-eminently the subject of philology:
it is emphatically the word; and yet we cannot
say that we have any very definite conception
of its meaning in the sense that we have of the noun.
The fact is, that the idea we form to ourselves of
a verb is an idea of action, whether that is restricted
to a definite single act or extended to an
indefinite succession of acts. But in either case,
the conception of action implies the conception
also of a subject and of an object; and neuter
129verbs, which throw the object into the background,
or even seem to obscure it altogether, are as rare
in an early period of language as verbs of a purely
abstract signification. Hence we find the middle
voice, where the subject is also an object, preceding
the passive; while languages of a more
primitive type than the Aryan, such as the
Accadian, the Basque, or the Mordvinian, insert the
object-pronoun between the subject-pronoun and
the verb, even in cases where it seems to us superfluous;
just as the Algonquin has no verbs to
express “to be” and “to have,” or the Semitic
languages preferred to denote existence by the
paraphrase, “something is an object to him.”
Analysis, again, is leading us to the conclusion that
the prepositions are, for the most part, old substantives,
while even the conjunctions, such as and, the
Greek ἔτι, and Sanskrit ati, 154 or que (Greek τε,
καὶ, Sanskrit cha), were originally demonstratives.
Thought must have a beginning and an end as
well as a middle, and to seize upon one of these
alone apart from the rest as the starting-point of
Glottology is manifestly absurd. It would be as
if geology were to concern itself with the individual
pebbles of a sea-beach, and endeavour to draw
conclusions from them, instead of comparing them
130one with the other, and with the texture of the
neighbouring rocks. Language is based upon the
sentence, not upon the isolated word, for the latter
can mean nothing except interjectional vagueness.
It is merely a bundle of syllables and letters, or
rather of animal sounds; merely the creation of
the grammarian and the lexicographer. To become
language, it must embody thought and emotion;
it must express a judgment.

It is the conception of the sentence, therefore,
wherein languages will resemble or differ
from one another. In Chinese the sentence is
summed up in a single word; the mind has not
yet clearly marked off its several parts, and
analysed what we may call the early communism
of speech. This is done in Turanian; but here
the sentence is of the most simple character, each
portion being of the same hue and force. It is
not till we come to the inflectional stage that the
parts are duly subordinated; co-ordination of
function gives place to a fitting correlation, and
makes possible the long compounds of Sanskrit,
or the exquisite periods of a Greek writer. In
the terms of the Hegelian philosophy, thought first
lies implicit, indetermined, and confused, in a
kind of rough block; then it becomes determined,
but by means of an opposition which equalises the
contradictory elements; and the opposition is
131finally removed by making each term supplement
the other, according to the laws of a relative subordination.
But the conception which underlies
each form of the sentence, each stage in the
development of language, is as essentially different
as the idea or principle which lay at the
bottom of the national life of those races who,
according to Hegel, have successively worked out
the problem of history. Regarding them from
the point of view of science or philosophy, we can
see how these stages stand related to each other in
the order of thought; but we do not see how
the gulf between them could practically have been
bridged over, or how it is psychologically possible
that the same race which conceived its sentence
as consisting of co-ordinate elements could also
have been potentially able to conceive it as consisting
of subordinate elements. There is no
question here of growth or evolution. The Aryan
languages may or may not have originally been in
a state not very unlike that of agglutination, the
Finnic group may or may not have come to offer
many of the phenomena of inflection: the agglutinative
idioms are still agglutinative, and the
Aryan family, so far as Glottology has cognisance
of it, has always been inflective. We may resolve
the Aryan verb into root, or base, and pronoun, but
we can never point out a time when the two were
132of full, equal, independent power; we may show
that the suffix tar, whence we get father, mother,
and numberless other words of agency, is the root
which means “to cross,” or “get through” (with
a thing), as in trans and through, but we prove no
more than when we demonstrate that the last
syllable of kingdom is the same word that we get
in our doom and the Greek δέμα, or that knowledge
and wed-lock are compounded with the old
English lâc, “sport,” “gift,” the Gothic láiks.
The Teutonic languages were inflective before these
suffixes were added, and they remained so even while
these suffixes still retained their original independent
power. In fact, had they not already been
inflectional, the suffixes would not have become
so, but have continued agglutinative and independent,
since mere outward phonetic change
cannot produce an internal mental change, and,
without the inflectional instinct to precede it,
cannot alter the manner in which the sentence and
its several parts are regarded by the mind. In
the same way, however nearly some of the Turanian
dialects may approach to the perfection of
inflectional speech, the character of the language
still remains fundamentally agglutinative. The
wearing of time may, indeed, have wasted away
the personal terminations of the Votiak verb adzo,
“I see,” adzi, “I have seen,” or even have acted
133upon the earliest-known Accadian, so that the
origin of the participial affix á is obscured, and
the termination of the third person plural of the
past tense in -es and -us exhibits but slight trace
of the primitive mes, “many,” out of which it
arose; but the example thus accidentally set has
not been followed, and the most “Europeanised”
of the Turanian tongues keeps true to its original
conception of objects and actions. Time will do
much, but it will not bring about an entire change
in the mode of thought, in the whole constitution
of the mind, through the external accident of
phonetic decay. We, whose idea of language and
manner of viewing things in thought belong altogether
to the inflectional stage, naturally fancy
that it will be with other races as it is with us,
and that when certain antecedents are given, the
phenomena of inflection will necessarily follow.
But how is it possible for one to whom objects
and actions and relations are all equally concrete
and important to be brought to regard them otherwise,
at all events without the help of education?
Can we expect a “Principia” from the Negro, or
an “Organon” from the Arab? The Ethiopian
cannot change his skin, nor the leopard his spots.

The advocates of this theory of development,
who would cast all men in exactly the same mould,
instead of admitting that different races have
134started in history with different tendencies, different
potentialities, are obliged to lay down that
each successive stage in the evolution of language
marks a successive progress in civilisation, and
that as men became more civilised, so did they
approach more nearly to the inflectional level. But
this is to ignore the facts. Chinese civilisation is
the oldest in the world; its origin is lost in myth,
and its continuity is unbroken. And yet its
founders spoke an isolating language, while their
barbarian neighbours on the West were in the
more advanced and civilised stage of agglutination;
and not only so, but all their long unbroken civilisation,
all the meditations of Confucius or Mencius,
all the desperate contrivances of writing, all
the intercourse with an Aryan population that
Buddhism introduced, have not made the Chinese
language advance one step beyond its first isolating
stage. Phonetic decay has been at work in the
vocabulary, dialects have sprung up in the empire,
new words have been applied to denote the relations
of grammar (more especially in writing), and
yet the sentence is still confined to the individual
vocable, and position and tone must determine the
meaning of the speaker. 155 It is the same with the
135other Taic languages in the south, where the King
of Siam, at all events, is said to be the most
learned monarch in the world. Nor is it otherwise
when we look at the western side of Asia.
Civilisation there began in the valley of the Tigris
and Euphrates; and the cuneiform monuments
have informed us that the first known inhabitants
of the country, the inventors of writing and arithmetic,
the builders of cities and temples, the
observers of the phenomena of the sky; nay, it
would also seem, the instructors of the “inflectional”
Semites in the rudiments of civilisation,
were a people whose language was agglutinative in
the highest degree. Is it not strange that throughout
their long career, in spite of the example given
them by their Semitic neighbours, the Accadians
should not have improved upon the inflectional
136character of their language in the slightest, although
their Elamite kindred, less advanced in culture, it
would appear, than themselves, had reached just
that amount of semi-inflection in the verbs which
strikes us in the Finnic dialects? It is equally
remarkable that the latter, which bear a very close
resemblance to this Elamite idiom, should have
made no further progress in the direction of inflection,
notwithstanding their longer period of existence
and their contact with the Aryans. All goes
to show that an isolating or agglutinative stage does
not imply civilisation or the reverse, and that no
amount of culture, no amount of years, and no
amount of foreign intercourse, has been able to change
the radical character of a language. Surely, if the
three stages of language mean mental progress, that
progress would have been attained more or less
by those who were capable of originating civilisation;
and if the circumstances of civilised life
are able to alter the conception of the world and
its expression in language, Chinese and Accadian
would have afforded marked illustration of the
power. To say that Chinese is artificially fossilised
is not only to beg the question, but also
to assume that civilisation stereotypes an early
expression of thought; an assumption contrary
to facts, as well as to the theory of continuous
development itself; and then what becomes of a
137barbarous isolating language like that of the Ainos?
The Aryans were not very highly civilised when they
herded together on the plateau of the Hindu Kush;
yet according to the common hypothesis they had
already passed through the stages of isolation and
agglutination, which their more civilised cotemporaries
in China and Babylonia were never able to
transcend. Nothing can show more clearly the
baselessness of a theory which asserts that every
language, with sufficient time and civilisation,
must pass through the three epochs of development.
What was sufficient for the Aryan or
Semite was surely sufficient for the Chinaman or
Accadian. The civilisation of the latter may have
been defective and inferior, but it has the merit of
origination; and the superiority of our own shows
only the superior intellectual capacities of the
race, that is, that the mind of the primitive Aryan
was potentially superior to that of the Chinaman,
and accordingly potentially conceived of things
and their relations and embodied its conceptions
in speech in a superior way. We are apt to underrate
the extent of the psychological change that is
implied in the passage from one of these modes of
expressing thought to the other. It is little short
of a radical metamorphosis of the mind. And
when we think of the impossibility which the Jew
of Alexandria, and afterwards the Arab of Spain,
138had in understanding the primary truths of Greek
philosophy, in spite of education and culture as
well as the fact of their all using inflected languages,
we may gain some idea of the impossibility
the unassisted primitive savage would have
found in changing his mental point of view in
the concerns of everyday life. What Philo and
Averrhoes could not do on a small scale, could the
early Aryan or Semite have done upon a large
scale?

The theory, moreover, does not take account of
the forms of speech which do not strictly fall
under one of the three heads. The so-called polysynthetic
languages of North America, for instance,
are extremely important, characterising as they do
a whole continent. Here the sentence is fused
together into a sort of long compound, the several
words of which it is composed being cut down to
monosyllables generally, by the same kind of
accentual instinct that makes the French drop
their final letters in pronunciation, though each
fragment still remains an independent word of
equal force with the rest. Thus in Mexican, a
priest is addressed as notlazomahuizteopixcatâtzin,
compounded of no, “my;” tlazontli, “esteemed;”
mahuiztic, “revered;” teo-pixqui, “god-keeper;”
and tatli, “father:” and in Delaware, kuligatchis
signifies “give me your pretty little paw,” from
139k, the inseparable particle-pronoun of the second
person; wulit, “pretty;” wichgat, “paw,” and
shiss, “littleness.” Compound words are, of
course, formed in the same way, like the Delaware
pilápe, “a youth,” from pilsit, “chaste,” and
lenápe, “a man.” Are we to class these languages
under the isolating, since the sentence is reduced
to one long word pronounced at a breath, or under
the agglutinative, since the elements continue co-equal
and independent, or under the inflectional,
since they have been subjected to a species of
phonetic decay? Again, if we consider the incorporating
languages, those, namely, which insert
the objective pronoun into the verbal form, we
shall have to admit two possible ramifications of
the agglutinative group. Incorporation appears
in its simplest form in Accadian: thus in-bat, “he
opened;” in-nin-bat, or bat-nin-in, “he opened
it;” and we may even have the root used as a
substantive thrust between the pronoun and the
same root used as a verb, as in-śub-śube, “he
builds a building.” The same phenomena show
themselves in Basque, where the endless forms of
the two auxiliary verbs are due to the wearing of
time, which has amalgamated the incorporated
pronouns, and sometimes even (as in Accadian)
an incorporated noun. Didac, “you have it for
me,” for instance, is decomposed into the acc. d,
140the dat. id, the root a or an, and the nom. c; dizut,
“I have it for you,” into the acc. d, the dat. iz,
the root au, and the nom. t; while the characteristic
of the pl., it, is intercalated into the root, thus
cutting it in two. Certain verbal forms in Magyar,
again, enclose the objective pronoun; and the
Finnic Mordvinian of North Russia, where m + ak =
“me + thou,” or m + am = “me + he,” sets the contrivance
before us as plainly as Accadian and Basque.
When, therefore, we are told that language must
pass through an agglutinative stage, we may inquire
whether that means incorporation or not, or whether
it is necessary that every agglutinative language
should once have been incorporating? Of course,
polysynthetic and incorporating are to be kept
carefully apart: in the one, the words of a whole
sentence are cut down to their shortest form and
fused into a kind of long word; in the other, a
few words are loosely attached to the verbal root,
unimpaired and independent. There is much more
difference between incorporation and polysynthetism
than between incorporation and inflection.
In some respects, indeed, Basque might almost be
considered to have entered upon the road of
inflection.

This leads us to what first suggested, and has
since been the chief support of, the theory under
discussion. An analysis of the Aryan inflections
141seems to take us back to a period when the primitive
language was purely agglutinative, and to a
still earlier period when it consisted of rough
isolated roots alone. The inflections of the verb
in Aryan as well as in Semitic can be traced to the
attachment of the objective cases of the personal
pronouns to the root or base, while many of the
verbal forms seem to be the result of a combination
of the root with other verbs, ya, “to go,” dha,
“to place,” or the substantive verbs, as and bhu.
Other forms, however, such as the reduplicated
perfect or the optative (Sanskrit bhavey-am = φύοι-μι),
have had a different origin, not unlike that of
the Teutonic ablaut, which represents unmeaning
vowel changes (caused by the accent) in Sanskrit,
or the use of the vowels in Semitic to distinguish
different parts of the verb. The modern languages
of Europe have returned to the simplicity of the
primitive Aryan verb, though the pronoun has become
subject instead of being semi-objective. All
this would apparently tend to show that flection did
not originally belong to the verb, and that there
was a time when its several relations of time and
mode and person were each expressed by independent
words. The analysis thus successfully carried
out in the verb has been applied to the noun, but
the results here have not been so decisive. One or
two of the case-endings have been identified with
142prepositions, or in this case postpositions, the
locative (primarily -in, as in Sanskrit tasmin, “in
that”) with in, and the instrumental with bhi,
“by,” and an attempt has been made to compare
the sibilant of the genitive and of the dual
and plural with the adverbial sa (sam, sahá), and
that again with the demonstrative pronoun. The
other cases are referred to pronominal roots; but
however well a demonstrative may suit the nominative,
it is difficult to see how it could express
the other cases, or how the other cases could have
arisen out of it. How, for instance, could the third
personal reflexive pronoun swa, se, produce the
plural locative, or the idea of the accusative be
obtained from mâm, mâ, “me,” or ama, “that,”
used to denote “a suffering object”? Moreover,
the pronominal ta, which plays so important a part
in the ordinary analysis of flection, is, as Ludwig 156
points out, a nonentity, since t is always followed
by the vowel i. In fact, the whole pronominal
theory rests upon a very narrow basis, as we shall
see further on; and the primitive Aryan must have
been at once supernaturally clever and supernaturally
stupid to extract the various cases of
the noun by ringing the changes on a row of
demonstrative suffixes. 257 Against the whole assumption
143lies the fact that no such vague generality,
such clumsy confusion, appears in the attachment
of the pronouns to the verb. Can we suppose that
144the same people who so distinctly marked out the
meaning of mi in the verb can have employed
it to express the sense of the accusative? If it be
145replied that the pronouns were all of indefinite
signification, and might be attached to the roots
at haphazard to express the various relations of
the sentence, out of which the different cases
gradually grew in some unexplained way, and
appropriated the several pronominal roots to themselves;
we must answer, firstly, that the whole
hypothesis is unsupported by facts, and therefore
beyond the range of Glottology; secondly, that
even the communistic inhabitants of a bee-hive
would find it hard to be mutually intelligible
with such conversational machinery; thirdly, that
the growth of the idea of the several cases out
of such a chaos, much less their selection,
is inexplicable, since the accidental terminations
would have confused the mind, not led it
towards analysis; fourthly, that there was no
difference between the nominative, the genitive,
and the dual and plural, so far as suffixes go,
and yet these are among the most important
distinctions; and fifthly and chiefly, that even
supposing we grant all that is required of us,
we shall still be no nearer to an agglutinative
condition of the primitive Aryan, since the
agglutinative languages do not form their cases
by the help of pronouns, but of postpositions, or
rather verbal and nominal roots. The relation of
cases, like all other relations, is with them an independant
146word; and from the Accadian down to
the latest and most barbarous dialect, we find words
like lal, “filling,” ge, “deep,” ra, “inundating,”
employed to express the several cases. In fact,
to represent these by indefinite pronouns is the
characteristic of a language inflectional from the
beginning, in which the suffix is weakened and
subordinated to the radical. It points to a primary
inflectional instinct, which shaped the sentence
accordingly as soon as the period of conscious speech
arrived. When the conception of the locative case,
for example, first arose in the mind of the Aryan,
he selected some formally existing but hitherto
meaningless suffixes to express the near relation,
and so turned a mere phonetic complement, a
mere formal sound, into a grammatical inflection.
It is the same with Semitic. Here the original
machinery of cases was elaborated by the adaptation
of the three primary vowels, u, i, and a,
though a might have been the earliest sound
shading off into u, the sign of the nominative, by
slowly closing the lips, and into i, the sign of the
genitive, by raising the tongue towards the palate.
It was not until later times that the case-terminations
were confused together, and replaced,
as in English or Persian, by prepositions. It is
plain that, before the setting apart of the three
primary vowels, the Semites had no cases; as soon
147as they became conscious of the want of them, the
cases came into existence, and this by purely inflectional
means, in which there can be no question
of agglutination with pronouns or aught else. In
Aryan, likewise, we must believe that case and
flection — whatever may be the origin of the latter
— are co-existent. As far back as the Aryan had
any conception of the relations of a sentence, he
expressed them by subordinate suffixes, not by the
help of independent agglutinations. More complex
nominal relations might be represented, as in the
Latin gratiâ, or the Greek χάριν, or the German
wegen, by a kind of postposition; but whenever
the latter ceased to be a separate word which could
receive inflections of its own, and became simply
the sign of a case, it was forthwith assimilated to
the other merely flectional cases, and its individuality
lost. The clear flectional growth of the verb
shows only that it took place during the historic
period, when the structure and tendency of the
language were already inflectional, and that it was
of later origin than the noun.

But even so this flectional growth of the verb
refers only to the verb as we have it in our grammars,
with all its moods, and tenses, and persons
fully worked out. There was a time when the verb
simply signified action in general, and the suffixes
which it then possessed were sufficient to denote
148this general idea. It was not until the conception
of personal relation had been struck out, that any
necessity for the employment of the personal pronouns
arose; and it is very possible that Ludwig
is right in referring the -sti of the second person
of the Latin perfect to an old infinitive termination,
σθαι, now utilised for a new purpose, like
-mini in the second person plural of the passive.
Flection, it must be remembered, is constituted
by a combination of meaning and form; it is
meaning that gives it existence and content, and
until this is furnished, the form remains a mere
phonetic sound. Now, meaning cannot be separated
from the sentence out of which each nuance
of grammar has been elaborated, and not out of
lifeless sounds which the prevalent glottological
theory would assume to be the immediate parent
of the inward and spiritual. The idea of the
instrumental case, for instance, must have been
obtained from a deeper analysis of the sentence,
which all along implicitly contained it; and then
some already existing ending or suffix was set
apart to express it. In this way we are able to
explain how it is that the same sound is not
appropriated to the same case, to the same grammatical
relation, in each and all of the Aryan
languages, but that i and bhis stand for the
locative singular and instrumental plural in Sanskrit,
149and for the dative singular and plural in
Greek and Latin. Still more significant is the
change of meaning of a form in the same language,
as in the case of tar, which characterises the
present tense in the Veda and the future in the
later epic. That such unmeaning terminations
existed in the period which lies immediately behind
that in which Comparative Philology properly
begins, has been made sufficiently clear by the
Prague professor to whom I have already alluded.
Although to the analytical lexicographer of the
nineteenth century the ultimate germ of a group
of words is a monosyllabic root, yet when we come
to regard these germs as beginning to be endowed
with life and meaning, as capable of being employed
in living, actual speech by the addition of
suffixes, we find that they are for the most part
no longer monosyllabic, but are, in the jargon of
the grammar, become bases. Thus the Sanskrit
voḍhavai, the Latin vectu (= vectuī), the Slavonic
vésti, must be regarded as independent unless a root,
vaghi-tavai, be presupposed. The same fact appears
still more unmistakably in varying forms with
identical meanings, such as rat and râjan, bhûs
and bhûmi, uś- and uśaś, sthât and sthâtar, tṛiś and
tṛiśâ, where phonetic difference is not accompanied
by difference of function, that is, where the
material outward elements of flection exist, but
150there is no flection as yet, because the inward
signification which makes flection still lies implicit
and unrealised in the sentence; or, again, in roots
of similar meaning and similar sound, but which
differ either in the initial or the final consonant.
Sthâ, stabh, stav (σταῦρος), star (στερεός) in Sanskrit,
or στεγ and τεγ in Greek, for example, like
the Semitic עול, על, טעל, טחל, םחל, all go back
to the same ultimate analytical origin; but no one
would think of discovering any diversity of signification
in the several varying forms. Each was
a form of the same unconsciously-felt type, which
lay at the bottom of the consciously-spoken word.
But the word, as conveying sense and meaning,
as filled with content and life, could not exist
apart from the sentence; of this it formed a
portion; and the relation which it expressed in
this was determined by the other words with which
it was joined. Now, it was just this determination,
and nothing else, which created flection. The unmeaning
terminations of the several words were
used as the external signs and channels of this
determination, and thereby flection, both on its
internal and its external side, became perfect.
What the primitive flections were, and whether
any of them have survived to later times, we cannot
say. It may be that all the inflections of
Schleicher's parent-speech will yet be traced back
151to independent vocables; but this, improbable as
it is in the highest degree, will only show that the
new suffixes, as soon as they became grammatical
signs, were modelled after a fore-existing pattern;
they imply that the language was already inflectional,
and inclined to assimilate everything which
modified the meaning of a sentence to the prevailing
inflectional type. In the agglutinative dialects,
they would have remained independent or
semi-independent words. There is little to be
gained on the opposite side by bringing forward
instances in which, during historical times, an
independent vocable has gradually grown into a
flectional suffix. Thus ama-fui has become amavi,
and fero has produced candela-bru-m, though even
here the flectional suffix properly so called is distinct
from the agglutinated word, and has to be
added in order to express the relation of the whole
compound to the rest of the sentence. But, in the
first place, the very possibility of thus turning an
agglutinated affix into an inflection shows that
inflection was already the characteristic and rule
of the language. Then, in the second place, we
must bear in mind that Glottology is an historical
science, and the historical sciences imply change
and progress with the change and progress of
time. Consequently what holds good of a late
period in the history of a language does not necessarily
152hold good of an early period. The Coptic,
which once formed its words by means of affixes,
now employs prefixes instead; and the rich creativeness,
the varied mobility, which distinguishes
the older Aryan dialects, disappears more and
more the nearer we approach our own day and
our own stereotyped mother-tongue. Just as
civilisation blunts the keenness of our senses and
the quick perception of the influences of nature, it
tends to dry up the springs of speech, and to confine
us to a conventional round of already existing
words. We can no more argue from the analogy
of modern Aryan languages to their early condition
than we can from the linguistic phenomena of
the Aryan family to those of other families. To
do so is to repeat in another form the error that
would make the laws deduced from an examination
of this family alone of universal validity. The last
objection that may be produced against this appeal
to historical instances of agglutination passing
into flection is, that the later meaning of the suffixes
properly so called could not be theirs if they
were independent words; and how then can we
tell if they ever were independent words? As I
have so often said before, we must not go beyond
our data in Glottology, and these present to us the
case-suffixes, for example, already in existence as
inflections or modifying affixes. When Comparative
153Philology first becomes cognisant of an Aryan
language, these suffixes are simply grammatical
forms; there is no trace, so far as meaning goes,
of their ever having been separate or agglutinated
particles. Now their meaning expresses the relation
of the several parts of the sentence to one
another; and we may well wonder how it could
come about that when the primitive Aryan first
awoke to a consciousness of these relations, and
began to distinguish between them, he denoted
them by independent words, and yet, when his
consciousness became clearer and more distinct, all
vestiges of their original nature were lost, and a
backward step was made in the analysis of the
sentences. This, of course, grants the assumption
that the independent origin of the case-suffixes has
been made out, which is very far indeed from
being the case; and that, as independent vocables,
they were actual words with real meaning and
expressiveness, not the vague and indeterminate
“pronominal” elements to which the modern
school of philologists would refer them. Indeed,
when once this useful but impalpable “pronominal
root” is introduced, the whole question is virtually
decided. Every tittle of evidence for the
theory derived from analogy is abandoned. The
agglutinative languages do not express the relations
of grammar by pronominal suffixes — indeed,
154it is hard to see how they could do so — but by the
help of postfixed substantives and verbs or participles,
each with a definite signification of its own.
Thus the postposition kyda in Ostiak is kyt, “the
middle;” the locative pir in Samoiedic Jurakish is
“height;” the possessive lal in Accadian is “to
fill.” So, again, the Bornu of Africa says “side”
for “with,” “head” for “on,” “place” for “to;”
and the Vei would express “it is within the
house” by á be keneburo, “in the house's belly.”
It is the same in the isolating languages. The
“empty words” or determinative particles of
Chinese mean “interior” (chung, nei, li, the signs
of the locative), “to use” (y, which marks the
instrumental), and so on. Nor is it different in
the Aryan family itself, wherever we can historically
trace the passage of an independent word
into a semi-flection. Either it is a personal pronoun,
as in the person-endings of the verb, or a
substantive like -dom and -head; never an imaginary
“pronominal root.” But these semi-flections
all belong to the later epoch of Indo-European
speech, when the fresh period of youth and creativeness
had passed away; and to assert that
because the High German taubheit is compounded
with heit, the A.-S. hâd, “character” or “rank,”
its earlier representative, daubitha must be similarly
compounded with a “pronominal” element,
155is to defy all the principles of scientific inquiry.
The pronominal root is a philological myth, which
owes its origin to the supposed necessity of developing
an inflectional language out of an agglutinative
one. Such forms as daubitha will have
been flectional from the first. The formal element
existed before the significant element was added to
make it a flection; and this genesis of inflection,
this rise of new flections, can be tested and confirmed
by historical instances. Thus the Teutonic
idioms have adapted the ablaut or change in the
vowel of the root to the expression of the distinction
between the tenses of the verb, thus making
it inflectional; while it remains in Sanskrit a mere
phonetic unmeaning modification of the vowel, the
mechanical result of the accent. So again the
Sanskrit verbal termination -ayâmi has been split
up into the three Greek endings in -αω, -οω, and
-εω, and these have been utilised in many instances
to set forth different shades of meaning, -οω being
appropriated to a transitive signification, -εω to an
intransitive one, and -αω floating between the two.
Πολεμέω, for instance, is to “wage war,” πολεμόω,
“to make enemies.” Such cases are more instructive
than pages of indefinite discourse on the
pronominal ta or ya, and they display the inextinguishable
instinct of inflection working in
Aryan speech, late down into the historical period.
156If we are to listen to the testimony of facts, the
agglutinative stage is a baseless dream, however
convenient it may be for the purposes of provisional
classification. 158157

The Aryan languages have always been inflectional,
so far as Glottology has any cognisance of
them. Beyond that, the Aryan must be dealt
with by physical science; and whatever the latter
may demonstrate, even that he was the eldest born
of a gorilla, we feel sure of this much, that his
brain could produce only an inflectional language,
that is, could view things and their relations only
in a particular way, as soon as he came to speak
consciously, and to be a subject for Comparative
Philology. What animal-like sounds he may have
158uttered before that time we do not know; it is
sufficient that his first endeavours to form a
language took the direction of inflection.

What has been said of the hypothetical agglutinative
stage of Aryan speech applies with still
greater force to the so-called isolating stage. It is
true that we can trace the lexicon back to a certain
number of roots, and it is assumed that these
roots, in which substantive, adjective, and verb lay
implicit in an equally vague and chaotic state, once
formed a language. Unfortunately we are not
acquainted with the exact nature of these roots.
We know them only in so far as they are the ultimate
elements of later words. But to assert that
there was a time when men conversed by means of
these roots alone is altogether unwarranted. 159 For
159anything we know, the roots might have received
flections, long since worn away; indeed Pott and
his school have endeavoured to make out that a
large number of our radicals are really compounds,
though with imperfect success. Nor do we know
whether the roots ever existed except as so many
unconscious types, after which inflectional words
were fashioned, and which were first extracted
from these by the grammarians, just as nowadays
we might take some foreign vocable and fit it to
numberless suffixes without ever using the vocable
itself. It seems clear that we must account in this
way for the numerous roots, or rather verbs, in
Semitic with similar meanings and cognate letters,
which cannot be derived either from one
another or from some common root. In any case,
we must not suppose that the imaginary isolating
stage of Aryan really resembled the phenomena of
actually existing isolating languages. In these
160the word is a sentence, and the reading of the
sentence is determined by the relation which it
bears to other sentence-words. Thus the Chinese
fú tzé, “son of the father,” or ngó tá ni, “I beat
thee,” are as truly analytical and determinate as
their English or Latin equivalents.

The root-language of the Aryans, however, did
not contain any sentences at all. A sentence implies
a mental judgment, a limitation of one idea
by another; and the vague, indefinite nature of
the root excludes a judgment altogether. As
soon as a judgment was arrived at, it was expressed
by means of inflections, or, as the advocates of the
development theory would say, of pronominal
agglutinations. Thus, in Chinese, just as in English,
the same word may be either a verb or a
substantive, or an adverb, but not at the same
time and in the same place; but this is exactly
what the Aryan root was, a kind of phonetic germ,
which contained within itself the potentiality of
becoming any one of the several parts of speech.
But until this was realised there was no language,
since Glottology begins with the sentence; there
was only an embryonic chaos of unconscious
thought. When first we find this thought becoming
conscious and embodying itself in language,
we find also the phenomena of inflection.

It is because the fact that language is the outward
161expression of conscious thought has been forgotten
that it has been supposed that the ultimate
analysis of phonetic sounds is identical with the first
beginnings of speech. It really gives us only the
beginnings of the mechanical part of speech — the
instruments of language. It is like the analysis
of colours in painting. The whole misconception
depends upon the false view that makes the bare
word the starting-point of philology, and the belief
that the history of the Aryan family is the
history of language generally. Language is an
art as well as a science; it is historical, not physical;
and in studying it, therefore, we must not
put out of sight the conscious effort exercised in
its growth by the mind of man. It is not an
organic product merely, any more than society;
and since language is the reflection of society,
whatever has influenced and determined the development
of the one will similarly have affected
the development of the other. This is the side
upon which the hypothesis of a threefold evolution
has chiefly been assailed by Pott. We
may call language an organism metaphorically,
but the metaphor must not be pressed too
far. There is no inner necessity in language
to expand like the seed into the tree, or the
caterpillar into the chrysalis and the butterfly,
any more than there is in thought and in society.
162An isolating dialect does not necessarily become
agglutinative, or an agglutinative one inflectional;
nor conversely must an inflectional dialect necessarily
have passed through the stages of isolation
and agglutination. The society of modern Europe
is not the descendant of the society of ancient
Babylonia or China: we can trace its ancestry
back through the middle ages to Christian Rome
and Periklean Greece, and far beyond that to the
herdsmen of the Hindu Kush; but its general
complexion, its fundamental principles, its innate
tendencies, have always been the same, and must
always continue to be so. External circumstances
will modify and alter; but large as their influence
may be, there yet remains an insoluble, unchangeable
residuum, which we call the character or
instincts of race. The intellectual growth of the
Negro stops at fourteen; and although he has been
brought into close contact with the civilisations of
the ancient and the modern world — with old Egypt
and Carthage, with Greek, Alexandria, and Rome,
with the Arab, the Latin, and the Teuton — he is
still in form, and colour, and nature what he was
when he first appears in the sepulchral chambers
of the Pharaohs. For racial change we need a
period of time far exceeding the miserable six
thousand years of history and civilisation; we
must go back to those incalculably distant centuries
163when our earliest progenitors trembled before the
mammoth and the cave-bear, and their animal-like
condition allowed the full play of natural
selection. But with this semi-human epoch of
mankind, Glottology has nothing to do. With
language consciousness begins, and the several
families of humanity have their characters already
formed, their modes of thought already determined
in an earlier period. Without doubt the
three stages of language mark successive levels of
civilisation: this much is proved by the subversion
of the one civilisation by the other; but each was
the highest effort and expression of the race which
carried it out, and the form which, by the constitution
of the mind of the race, each was necessitated
to assume. Mankind progresses as a whole, but
the several steps of advance are made by the appearance
of different races on the scene, each with his
mission, each with his predetermined method of
accomplishing it. The infusoria which to-day
cover the bottom of the Atlantic have not changed
since the era of the chalk; but for all that, the
world of life on the globe has been steadily improving
and growing, although the lion has always
been a lion, and the dog a dog.164

Chapter V.
The possibility of mixture in the grammar and
vocabulary of a language.

The fallacy of imagining that language is a sure
index of race still crops up occasionally, especially
in second and third-hand writers, who undertake
to acquaint the general public with the results of
Comparative Philology. We still not unfrequently
hear that we have to claim kindred with the black
Hindu of Southern India, not on the ground of a
common tongue, but of a common descent. A
very little consideration is sufficient to dispel the
illusion. The Aryan tribes of the Rig-Veda who
invaded India could not have been very numerous,
and it was long before they spread beyond the
north-western corner of the peninsula. Consequently
the chances are that a modern Hindu
will be altogether or in great part of aboriginal
blood, unless he be a Brahmin; and even the
Brahmin is to be found, according to Dr Hunter,
among the lowest castes, showing that his purity
165of parentage was not always regarded during
the disintegrating period of Buddhist democracy.
Who, again, can say how far the blood of our own
ancestors was contaminated during their distant
migrations before they entered this country? We
have only to look at such cases as the Kelts of
Cornwall, who speak English, or the Jews of
Southern Austria, who believe Spanish to be their
sacred language, to see how little we can argue
from language to race. Like the Lapps and Finns
in Europe, the Melanesians and Papuans have the
same tongue, but physiologically are essentially
different; and the only question that we can ask
in regard to them is, to whom did the language
first belong, and which of the two races borrowed
it from the other? Language is the mirror of
society, and accordingly will reflect every social
change. Wherever the social pressure is strong
enough, either through conquest, or personal interests,
or other causes, the inferior people will
adopt the idiom of the superior. Thus Keltic
disappeared before Latin in Gaul and Spain, and
social disadvantages have driven Welsh into the
mountains and the cottages. Thus, too, Slavonic
became extinct in Prussia in 1683, although five
hundred years before this date German was unknown
in the country. 160 Where the conquerors themselves
166are not numerous, or where they are less civilised
than the conquered nation, the necessities of everyday
life and the influence of literature will cause
them to adopt the language of the latter. Thus
it was with the Normans in France and England,
and the Franks in Gaul. In fact, we may lay it
down as a general rule, that whenever two nations,
equally advanced in civilisation, are brought into
close contact, the language of the most numerous
will prevail. Where, however, a small body of
invaders brings a higher civilisation with them,
the converse is the more likely to happen. 161 Visigothic
was soon extirpated in Spain, but English
flourishes in India, and Dutch at the Cape. Conquest,
however, is not the sole agent in producing
social revolutions extensive enough to cause a total
change of language. Before the Christian era,
167Hebrew, Assyrian, and Babylonian had been supplanted
by Aramaic, which was fast tending to
become the common dialect of the Semitic world,
like Arabic in later times. It was the language
of commerce and diplomacy, and this was sufficiently
strong to outweigh the conservative influence
of a sacred literature.

In all the instances just given, with one or two
exceptions, it will be noticed that a thorough-going
exchange of language has taken place only among
members of the inflectional family. There is no
example of an inflectional dialect being exchanged
for an isolating or agglutinative one, or vice versa.
The question accordingly arises, whether such an
occurrence is possible? Can an individual or a
nation, whose mind has been accustomed to regard
the nature of things from a particular point of
view, be taught to express himself under altogether
different forms of thought? Here we have nothing
to do with the possibility of an isolating speech
developing of its own accord into an agglutinative
or inflectional one. The settlement of this question
is not affected either way by an artificial
education, in which the mental faculties of one
people are domesticated, as it were, into the ways
of thought of another, to revert, like the domesticated
animal when again left to itself, into its
old nature, its original expression of psychological
168habits. A child can learn as readily the vernacular
of Canton as the language of London. The
Japanese show a singular aptitude in imitating
the externals of European civilisation. They may
yet produce a satisfactory copy of the philosophy
of Aristotle and Hegel, but I much doubt whether
they will ever be anything more than imitations
and copies; at any rate, experience is all against
it. Not to speak of the Jewish and Arabic reproductions
of Plato and the Stagirite, to which I
alluded in the last lecture, we have facts like that
of pigeon-English at Canton, where the Chinaman
has endeavoured to assimilate English, or the
Chinook jargon of Oregon, 162 or the grammarless
English of the Negro — all cases in which one race
has read its modes of thought into the grammar
of another, where it has not been able to resist
the encroachment and victory of the latter. And
yet English is, of all inflectional languages, not
even excluding Persian, the easiest to acquire;
and the extent to which it has pushed the clear
probing of analysis, and shaken off the trammels
of impractical flection, make it deserve to be, what
Grimm prophesied it would become, the language
of the civilised world. Not less striking, on the
other hand, is the preservation of the Basque;
169although driven by a Keltic invasion into the
extreme corner of Spain, it has yet lasted out all
the vicissitudes of Roman, Gothic, and Moorish
domination, instead of yielding, like its Keltic
neighbour, to the influence of the Latin tongue.
The attempt to make one race of men think according
to the forms of another is forced and
unnatural; and however much we may seem for a
time to have succeeded, yet, when the pressure of
superiority is once removed, our pupils return to
the conceptions of their ancestors, as the dog on
the prairies to his howling. Where the race has
not reached a high enough level of culture to
appropriate the language of its superior, it is a
sign that the race has done its part, and must
pass away before the coming of civilised man.
The Tasmanian and his language, in spite of every
effort to save them on the part of the Government,
have become extinct. Climate may save a tribe
and its dialect by making it impossible for the
European to settle in the country, but in that case
the dialect is preserved only because the social
conditions of which it is the expression are also
preserved through the maintenance of the original
state of nature. Civilisation inevitably kills the
natural, unless the latter is favoured by external
circumstances. Compatibility of existence on the
part of two races depends upon their being more
170or less nearly matched in culture. The greater
the distance between them, the greater will be the
influence, socially and linguistically, exercised by
the superior, until a point is reached at which it
will be impossible for the lower to live in the
presence of its higher neighbour.

Linguistically, the influence will show itself in
the shape of borrowing. We have already glanced
at the cases in which this borrowing extends to
the whole language, and have suggested the extreme
improbability of its taking place where the
ground principles of the languages are essentially
different; that is, where two civilisations, with
wholly different pasts, confront one another on
equal terms, or where the interval between two
races is morally and mentally too great to be
spanned. Borrowing, however, by no means necessarily
extends to the whole language. More often
it applies only to the vocabulary, and loan-words
are common to all dialects. No people can have
near neighbours without receiving something from
them in the shape of inventions, products, or
social institutions, and these, almost inevitably,
are adopted under their foreign names. The
French have taken meeting and turf from us,
together with the ideas which they denote; we
have had in return naive and verve. Where the
general condition of two nations is very unequal,
171the loan-words will be extremely numerous: in
Basque, for instance, more than one half the
dictionary is from a foreign source. So, again,
according to Campbell (“Teloogoo Gram.,” p. xix.),
one half the words in Telugu, as spoken in the
higher regions, come from abroad. The same is
asserted of Maráthá by Ballantine (Jour. of Amer.
Orth. Soc., iii.), and some writers tell us that nine-tenths
of the Hindu language is Sanskrit. It is
clear, however, that the borrowing will not be
entirely upon the side of the inferior; whatever the
latter is able to contribute to the superior, whether
it be a human invention or a natural product, will
generally carry its old name along with it. Thus
the Latin petorritum, “the four-wheeled,” is of
Gallic origin, and has been supposed to be of
some importance in settling the Cymric affinities
of the Gauls; 163 glæsum, “amber,” came from
Northern Germany; and our own tomahawk and
boomerang have been furnished by the Red Indians
and the savages of Australia. Maize, mangle,
hammock, canoe, tobacco, are all derived, through
the medium of the Spanish, from the Haytian
mahiz, mangle, hamaca, canoa, and tabaco (“Humboldt's
172Travels,” Engl. tr., i. 329). Indeed,
these loan-words are of the greatest use in tracing
the history of languages by revealing the geographical
and social relationships of the past.

Now, it has been much questioned whether it
is possible for a people to mix its grammar in the
same way that it can mix its lexicon, and adopt
some of the inflections or grammatical contrivances
of another speech. Before the rise of Comparative
Philology, grammatical differences went for very
little; and we still hear “philologists” of the old
school talking about borrowed grammatical forms.
Glottology, in which grammar forms the chief
fundamentum divisionis of languages, meets this
belief with a decided negative; and one of the
primary articles of faith held by the scientific
student of language at the present time is, that
if grammatical inflection be borrowed at all, it
must be borrowed throughout — we cannot have a
mixed grammar. The whole of the vocabulary
may be derived from abroad, and yet, if the foreign
grammar be not learned at the same time in
extenso, no parts of it will be adopted, and the
new words will be cast in the old moulds of
thought and expression. This is pretty much
what has happened in the case of the Negroes;
though here, of course, an attempt has also been
made to learn the English grammar artificially,
173with what success, however, is shown by the
Negro jargon of the United States. It is hard at
first to see what inducements there could be for
one dialect to incorporate fragments of the grammar
of another, as the causes which have acted
upon a borrowed dictionary — inventions, products,
social advantages — are here not applicable; and
the psychological impossibility which we were
considering in the last chapter, of forcing a race
to regard the world with the mental eyes of its
neighbour, would prevent the attempt if carried
on spontaneously, and not as the result of artificial
education. Nevertheless, the proximity of two
languages implies that a certain number of the
population are bilingual, and where this is the case
to any large extent, the idioms of the two dialects
will often be exchanged, and along with the
idioms an opening is made for the introduction of
new grammatical forms. Words like avenir and
contrée in French are the result of an endeavour
to express German idioms (zukunft, gegend) in
the Romance of the conquered provincials; and it
does not seem very difficult to stretch this process
a little further, and adapt foreign grammatical
conceptions to the contents of a native grammar.
Thus it has been asserted that the great extension
of the plural formation in -s in English was due
to Norman-French influence, though undoubtedly
174the tendency had already been felt before; and
certainly the use of the genitive and dative of the
personal pronouns in English, “of me,” “to me,”
in the place of the Anglo-Saxon min and me,
appears to be modelled after the pattern of the
French. 164 So, again, Bulgarian has imitated the
Wallachian usage which attaches the article to
the end of the word (e. g., domnul = dominus-ille),
as in Danish and Swedish, where dag-en, = “the
day,” guld-et, “the gold,” or in the emphatic
aleph of Aramaic, which is probably the postfixed
article. And still more strikingly, Persian has
adopted the Semitic order of words so repugnant
to the general structure of the Aryan group,
saying, for instance, dil-i-măn, “heart of me,” for
“my heart,” dăst-i-'Umăr, “hand of Omer.”
Conversely the Hararite is able to reverse the
Semitic order, and adopt the idiom of its non-Semitic
neighbours by writing ámir askar, instead
of askar ámir, “the Emir's army.” The so-called
175sub-Semitic dialects of Africa present us with the
further phenomenon of a grammar which is
decidedly Semitic in its main features, and which
yet makes use of postpositions. The natives of
Harar, for example, regularly employ these except
with the personal pronouns, 165 and use a postfixed
-n, which seems a relic of a primitive nunnation
to denote the accusative. 266

Here one of the fundamental principles of
Semitic thought seems to be violated, and the
attention drawn to the derivation, the ultimate
elements of an object, instead of to the immediate
presence of the object itself. In the same way a
close connection with a foreign race seems to have
suggested to the Assyrians at one extremity of
the Semitic world, and to the Ethiopians at the
other, the utilisation of existing materials to denote
more exactly the temporal relations of the verb;
and Persian, which has filled its dictionary with
Arabic since the days of Firdusi and his purely
Aryan Shahnameh or “Book of Kings,” has even
gone so far as to form one of its plurals by means
of the Arabic feminine plural in āt, jāt, as in
176niwāzĭshât, “favours,” from niwâzish; kăla'jāt,
“castles,” from kăl'ăh. Practically, however,
this plural is confined to Arabic words; consequently
it will no more be an importation of a
foreign grammatical form than our own use of the
Latin plural-ending in such words as termini. A
better instance would be the Latin and English
factitive suffix in isso (izo) and -ise, from the
Greek ιζω. But this, after all, is only a suffix, not
an inflection, and belongs, therefore, rather to
the dictionary than to the grammar; the nuances
of grammar require the true inflections of Latin
and English to be affixed or prefixed to this exotic
-ise; patrissi-t, civilise-s, το civilise, and so forth.
On the whole, therefore, the evidence before us will
confirm the absolute denial which Glottology gives
to the old notion of a mixture of grammatical
forms. Idiom may be imitated, even also the
conception of the relation of subject and attribute,
for this, as logic teaches, may be looked at in two
ways at the same time; but beyond this, language
does not seem able to go. No amount of intercourse
and familiarity seems able to transmute the
inflections of a dialect into the inflections of a
foreign one, any more than the alchemist was
able to change iron or lead into gold. He could
gild them over, but they remained iron and lead
still. The forms of grammar are the expression
177of the mental life and history of a people; they
imply, therefore, the summing-up of all that
history; and accordingly, although two nations
may have started from the same source with a
common stock of ideas and a common psychological
tendency, yet in so far as their experiences have
been different, the formative elements of their
languages will be different, and not interchangeable.
How much more will this be the case when
the two nations did not start from the same
source! The grammar of pigeon-English is not
English but Chinese; the grammar of Persian
remains Aryan. The formative part of language
must ever be the surest differentia of linguistic
kinsmanship.

Of late years, however, the attention of European
scholars has been attracted to a case of great
difficulty which apparently contradicts our conclusions.
Inscriptions of the Sassanian era have
been found in Persia, written in what seem to
be two dialects, now generally termed Chaldæo-Pehlevi
and Sassanian-Pehlevi. Greek transcripts
are added in a few instances; and we are thus
enabled to discover that the unknown dialects
closely resemble the language of books still preserved
among the Parsis of Bombay, to which the
name of Huzwaresh or Pehlevi is ordinarily given.
The writing of these is extremely hard to decipher,
178owing to the corruption of the characters; and a
comparison with the inscriptions and coin-legends
on the one hand, and the old Pazend dictionaries
on the other, shows that the traditional reading
is often very far from the truth. Now, this Sassanian-Pehlevi
is a most heterogeneous mixture
of Aryan and Semitic, and the mixture is not
confined to the lexicon alone; it dominates equally
in the grammar. Thus the great inscription of
Sháhpur I. (a. d. 240-273) at Naksh-i-Rajab,
which has a Greek translation attached to it, has
in the first line the following representatives of the
Greek βασιλεως βασιλεων, אכלמ ןיכלמ and ןאכלמ
אכלמ, while the termination of the third person
plural imperfect of the Semitic verb in -un is used
as a verbal ending for all persons and numbers.
Here we have not only a fusion of the Aryan and
Semitic order of words, but also a fusion of their
inflections. If the language were one ever spoken
by the people, the decision of Glottology would
have to be modified, and we should be compelled
to admit the possibility of a mixture of different
grammars under favourable circumstances. But
everything goes to indicate that the dialect was
never a spoken one, at all events not outside the
literary coterie of the court. How else could it
have so entirely passed away, without leaving a
trace behind it, that the language of Firdusi in
179the tenth century is the purest Aryan? — Semitic
influence, notwithstanding the Mohammedan conquest,
being as little discernible in the outward
form as in the subject-matter of the “Shahnameh.”
It was not until after this date that Semitic began
to penetrate into Persia, and even then, for the
most part, into the vocabulary alone. Many of
the grammatical forms, moreover, which are borrowed
by the Pehlevi from the Semitic are used
without any sense of their proper force and
meaning: thus the verbal form quoted above could
never have been taken from a living Semitic dialect,
or such curious hybrids as the prepositions in man
like levatman, “with,” which Dr Haug connects
with תול(הול).

On the whole, then, we must consider this anomalous
Sassanian as an artificial court-language,
invented for literary purposes from reasons now
unknown to us, but which never did, and never
could, make its way into conversation. We cannot
adduce modern Persian by way of support,
since the Semitic order of words, which it seems
to have imitated by placing the governing noun
before the governed, as in rah-i-bāghbān, “path
of the gardener,” rah-i-dānā, “path of the sage,”
may be explained by regarding it as an analysis
of the genitive conception, as in English. This is
borne out by the fact that the qualifying word
180may be left out, and that the connecting vowel i
is seldom used in familiar conversation. If, however,
Schott is right in considering rā, the affix of
the dative and accusative, as borrowed from the
Altaic postposition which we get in the Mongol
doto-ra, “inwards,” abu-ra, “to take,” Turkish
szong-ra, “to the end,” a more serious difficulty
arises. But Schott's suggestion is by no means
proved, and we have to set against it the otherwise
uniform experience of Glottology. The formation
of a case by a suffix has its parallel in the
poetic vocative, which affixes ā instead of using the
preposition yā or ai, just as rā in the dative takes
the place of the preposition ba. Until, therefore,
some more convincing example can be brought forward,
we must abide by the belief that the grammar
of a nation will always remain pure and native,
unless supplanted wholly by another through a
kind of natural selection, although under certain
circumstances foreign influences may occasion the
adaptation of existing formative machinery to new
uses. It is probably to this principle of adaptation
that we must ascribe the phenomena which
have been already mentioned as met with in the
languages of Northern India — the Bengáli, the
Assamese, the Hindi, the Khasiya, and others.
In these, the verb and pronouns are unmistakably
Aryan, while the nouns seem, on the other hand,
181to connect themselves with the agglutinative
idioms. Just as the Tamil plural affix gaḷ or kar,
the Telugu lu, has been traced by Dr Caldwell to
the common Dravidian taḷa or daḷa, “a crowd,”
so the plural suffixes of these languages, jâti,
gaṇa, dig, varga, bilak, daḷa, are separate and independent
words, which take the place of the usual
Indo-European plural flection. Indeed, Professor
Max Müller suggests that daḷa is nothing more
than the Dravidian dala, which would thus have
provided exactly the same grammatical machinery
for Bengali as for Tamil and Telugu. But the non-Aryan
character of the nominal flection in these
North-Indian languages does not stop here. The
plural affix is intercalated between the noun and
the case-ending, which thus becomes a veritable
postposition, separable from the base, and still
preserving vestiges of its original co-ordinate relation
to the noun. In this respect it resembles
the Georgian, where the plural suffix bi is inserted
between the root and the case-termination. In
Assamese, for instance, manuh is “man,” manuh-bilak,
“men,” and from this we get the genitive
manuh-bilak-or, the dative manuh-bilak-oloi, the
accusative manuh-bilak-ok, the locative manuh-bilak-ot,
and the ablative manuh-bilak-e. Not the least
striking part of the matter is that the suffixes are
none of them Aryan. It is this which creates the
182chief difficulty of the case. Otherwise we might
compare such plurals as our own man-kind, which,
joined with words like -wards, as in man-kind-wards,
are precisely analogous to the Indian forms
of which we are speaking, and which only bear
witness to the late analytic character of the
language and its loss of inflectional creativeness.
It is this view of the matter that makes Professor
Max Müller write: “We can easily imagine how
people speaking the modern Sanskrit dialects, in
which the old terminations by which the plural was
distinguished from the singular had been worn off
almost entirely, should, when again feeling a want
to express the idea of plurality more distinctly,
have fixed upon a grammatical expedient which,
for their daily intercourse with their aboriginal
neighbours, had long been familiar to their ear
and to their minds. The words which they used
as the exponents of plurality were, of course, taken
from the resources of their own language; but the
idea of using such words for such a purpose seems
to have been suggested by a foreign example.”
Now, this very passage admits a non-Aryan
influence upon the grammar; and when we consider
the remarkable fact that the case-endings are
not Indo-European, it is hard not to allow that
something more than mere influence has been at
work. Indeed, if it should turn out that the
183idioms we are discussing are at bottom not Aryan
but Dravidian, this conclusion, in view of the
verbs and pronouns, is absolutely necessary. Unfortunately
this question is by no means settled
as yet, and its determination will depend upon
whether we find that the fundamental part of the
dictionary containing the words of everyday life
belongs to Sanskrit or to an aboriginal speech.
But such a determination cannot be made until
the vocabularies of these dialects are better known.
Meanwhile we may compare the somewhat parallel
instance of the so-called sub-Semitic tongues. If
we take the Berber, the Semitic affinities of which
are unmistakable, we yet find the verbal conjugation
admitting tense distinctions, not formed, as in
Assyrian and Ethiopic, by a modification of the
vowel, but by affixes and prefixes. Thus edh prefixed
to the aorist makes the present and the
future, ere the future and potential, while the affix
-ed forms a perfect, and -an the participle. The
suffixed pronoun is inserted between the verb and
these prefixes and affixes, and consequently precedes
the verb in many cases. This is always its
position in the case of the participle, as in ey-izran,
“seeing me,” eth-izran, “seeing him.” The
definite tense-determination of these prefixes assimilates
them rather to the old Egyptian, with
its innumerable compound verbal forms, than to
184the Arabic use of cāna and kad; but their employment
is not contrary to the spirit and usage of
the Semitic languages: on the contrary, the affixes
ed and an are altogether foreign to the genius of
these tongues. Not less so is the prefixing of the
suffixed pronouns, and we can scarcely help seeing
in it the influence not only of the allied Coptic
with its developed system of prefixes, such as nen
for the plural, mad for abstracts, or ref for agents,
but also of the once neighbouring Kafir tribes,
which always prefix, never affix. Here, therefore,
will be another example of the way in which the
grammar of a people may be affected and modified
from without. We fail to see, however, anything
like the phenomenon which meets us in the North-Indian
dialects, where the case-endings appear to
have been imported, as well as the manner in
which they are applied; and it is not until we
come to the postpositions of the Hararite that we
discover any analogy to this. But the language
of Harar, like that of Assam, is as yet too little
known to permit us to come to any certain decisions
in so difficult a question. It is noticeable,
however, that in both cases it is the nominal declension
which presents the grammatical anomaly;
and when we consider that we have in English
such words as fungi, prospectus, and termini,
while German can form from Christus both Christi
185and Christo, we may perhaps conclude that the
noun does not always offer that sure criterion of
the character and position of a language which the
verbs and pronouns do, and that in certain stages
of linguistic growth, when a speech has become
more or less analytic, it is able to borrow from its
neighbours not only the form of the declension,
but even the words which compose this form.
The analytic period means the resolution of the
sentence and its grammatical relations into separate
vocables, and these can be borrowed freely by
one idiom from another.

Intimately connected with grammar is the phonology
of a language. It is a question of some
interest how far the pronunciation of a dialect
may be affected in the lapse of years by the contiguity
of another. That such an influence can
be exercised is certain. A familiar example,
which will occur to the mind of every one, is the
adoption of the Hottentot clicks by the Kafirs.
This is a very remarkable case, as the sounds are
difficult, and the superiority of the borrowing race
is very marked. So, too, the so-called “cerebral”
letters in Sanskrit, which are not found in any of
the other Aryan dialects, are commonly thought to
be borrowed from the Dravidian; and the Norman
Conquest appears to have had much to do with the
softening of the gutturals in the southern part of
186England, the Gallicised invaders finding their pronunciation
difficult, and accordingly setting the
example of breaking them down. The retention
of the gutturals in Spain, again, may be ascribed
to the long settlement of the Moors; and I remember
a Basque girl, to whom French had
become the language of everyday life, when giving
me my first lesson in Euskuara, calling egoitz
(“a house”) egoi'. Of the same nature is the
change of i to g in Anglo-Saxon, contrary to the
usual softening of consonants with vowels, of
which Professor March 167 remarks, that “the movement
(of consonants to vowels) is sometimes
reversed, as when a nation moves northward, or
northern peoples mix with a vowel-speaking
race.” It must be remembered, however, that
climate, food, and custom have much influence
upon phonology, and that where these are similar,
we may expect to find a general similarity in the
pronunciation of two languages. We are all well
acquainted with the hoarseness and roughness
that exposure to the atmosphere lends to the voice;
and the exercise and strength that a mountainous
country gives to the lungs produce a corresponding
effect in the vigour with which sounds are
emitted. Food, of course, will have an equal
influence. The vocal organs are under the command
187of the muscles and the nerves, and these
depend upon the health and robustness of the
body. A mixed race will inherit the phonetic
capabilities of its parents, and the preponderance
will lie upon the side of the stronger parent.
Particular fashions are not without their influence;
thus the loss and confusion of the labials, and the
excessive nasalisation in the languages of the
savages on the Pacific coast of America, must be
traced to the rings that are worn through the
nostrils and the lips of the people. 168 So again we
find from Bleek that the pronunciation of the O
Tyi-herero in South Africa is lisping, and is due to
the custom of extracting the four lower teeth, and
partly filing away the upper teeth. 269 Imitation
will also come into play: we acquire our pronunciation
in the mimetic days of childhood, while
the vocal organs are still plastic; and here, again,
the preference will be given to the pronunciation
which, for any reason, is the best fitted for success.
Social superiority has much to do with this: we
attempt, in school and out of school, to reflect
the pronunciation of the higher circles of society;
and just as the court dialect of Chaucer became
the universal model in England, or Parisian
188French is extirpating the Languedocian patois, so
a dissimilar pronunciation becomes the mark of
vulgarity or provincialism. And when once a
particular pronunciation has become prevalent, it
reacts upon all words that still remain exceptions:
thus in English, balcony, retinue, and contemplate
have, after a long struggle, followed the rule
which throws the accent back as far as possible.
If we cross to America, we find a similar phenomenon
taking place there. It is seldom that we
cannot detect a born and bred American by his
pronunciation. English seems in the mouths of
them all to be diverging into a sharp quick nasalisation,
which can hardly have originated in the
twang of the New England Puritan, or the commixture
of European races, but which seems due
to the influences of a dry, extreme climate, like
the hatchet-face of the aboriginal, which is being
reproduced in his white successor. Perhaps,
however, one of the best countries in which to
study this question of phonological borrowing is
Germany, with its numerous dialects and various
phases of guttural-pronunciation. Here the population
has come into contact with Slaves, Finns,
Magyars, and Latins; and Mr Howorth has endeavoured
to trace the sibilants of South Germany
to a Slavonic influence. 170 However this may be,
189imitation lies at the bottom of all pronunciation;
and it will be one of the future tasks of the glottologist
to determine how far the phonology of a
language has been modified by intercourse with
another, and how far the similarity of each is only
the result of a similarity of external conditions.
No psychological difficulty interferes here: we
have to deal only with the outward mechanism
of speech, and borrowed sounds are as natural
and as possible as borrowed words.

The latter are of immense importance in tracing
the growth and progress of the human mind.
If Glottology is the science which ascertains the
laws and successive history of that development
as embodied in the fossils of language, not the
least part of its work will be to detect the debts
owed by one race and civilisation to another. 271
190Nor is the work so easy as it seems to be at first
sight. We must find out general laws which will
allow us to determine whether words are really
borrowed or merely exhibit that accidental resemblance
which the circumscribed number of articulate
sounds sometimes brings about, as in the
North American potomac, “river,” and the Greek
πόταμος, or whether, again, they are both taken
from a common source, or one of them from the
other. Then we must have rules for knowing
191whether a word is of foreign origin or really of
native growth: and above all, when we have
actually ascertained that two words stand in the
relation of lent and borrowed, we must find out
on which side the debt lies. In the case of the
Semitic keren and the Greek κέρας, Latin cornu,
for instance, we may ask are these words of independent
origin, or are they loan-words, and if the
latter, by whom were they lent? Or again, is the
Greek χρύσος, derived from the Semitic khârûts,
“gold?” If we could learn that these were really
loan-words, much light would be thrown on the
history of early civilisation, and the relation of
Semites and Aryans under this aspect. Now, the
comparative laws of language inform us that
while, on the one hand, the final nasal of the
Semitic keren is a part of the root, the final -nu of
the Latin is a mere formative, and that the same
word appears in the Sanskrit śringam, “horn,” from
śiras, “head,” whence we have the Greek κάρα, the
Latin cervus, and our own hart. The East Aryans
of India had no such close intercourse with the
Semites as would have given the latter so common
and non-technical a word as keren, while among the
West Aryans the nasal is found only in Latin,
from which it could not have been borrowed by
Assyrians and Hebrews. Similarly, the reference
of χρύσος to the Sanskrit hiranyam (Zend zaranya,
192Slav, zlato, Phryg. γλοῦρος), and its phonological
connection with the root which signifies
“of a pale greenish-yellow colour,” — whence we
get the Sanskrit haris, the Greek χλόη and χολὴ,
the Latin viridis, bilis, luteus, and the English
green, gall, and gold, — sufficiently disposes of any
borrowing from khârûts, which, on its side, comes
from a Semitic root meaning “to grave” or
“dig.” Let us select another example from
Basque. A large proportion of the dictionary of
the latter language has been taken from Spanish
or Latin, and to this M. Bladé would add the
Basque numerals bi, “two,” and sei, “six.” But
the laws of phonology forbid this. The labial
which we see in bini is nothing more than the ν
of duo, which has lost its dental, as in viginti; and
out of the distributive the Basques could never
have got a cardinal. The only Latin form of the
numeral with which the Biscayans could have come
into contact was duo, through the Spanish dos, as
is again shown by glottological laws. And in
fact, there is no need of connecting bi with any
Latin word at all. The comparative study of the
Basque numerals has relegated them to the Finnic
family, and here both bi and sei are possible forms
for “two” and “six.” 172 Thus the laws which
193have been obtained from the comparison of phonetic
sounds in different groups of speech, by
enabling us to reach back to the earliest forms of
a word in each group, or to dialects which are
removed from the line of contact, allow us to determine
whether or not we are dealing with loan-words.
In the same way, other laws may come
into play when we are doubtful about the priority
of borrowing in any case. Thus in Accadian 'uri
meant “a city,” which at once reminds us of the
Semitic ריע (Assyrian 'uru); and we ask, Supposing
they are loan-words, on which side did the
debt lie? Now, I believe I have shown 173 that a
large number of Semitic words which denote the
first elements of a higher civilisation are derived
from Accadian, and this at once raises the presumption
that ריע is borrowed, and borrowed from
the Turanian neighbours of the Semitic nomads.
When we find, however, that not only other words
which signify settled habitations, like hêcâl
(Accadian ê-gal, “great house”) or the Assyrian
muccu, “building,” are derived from Babylonia,
but also that 'uri in Accadian enters into the
composition of other native words like murub,
194“city,” initial mu being interchanged with single
u, and b being a formative affix, and that the
Basque iri also means “a city,” Accadian and
Basque displaying startling resemblances both in
grammar and lexicon, we are induced to conclude
that it was from the old Turanian civilisation in
Babylonia, which the last few years have revealed
to us, that the early Semite obtained his first
lessons in culture. It is a contribution of the
highest importance to the mental history of mankind.
From the beginning the Semite seems to
have stood between the old and the new, between
Asia and Europe, the trader not in material wares
only, but in the far more precious merchandise of
thought and invention. I cannot do better than
conclude this chapter with two striking instances
of this.

The Greeks derived their weights and measures,
as well as their alphabet, from the Semitic East.
The standard of these was the μνᾶ, which was
handed on to the Romans, and so to the Western
world under the name of the mina. The μνᾶ is
the Hebrew maneh, and the final a proves that it
was immediately borrowed, like the letters of the
alphabet, not from the Phœnicians of Tyre and
Sidon, but from the Aramaic population further to
the north. Böckh has shown that Pheidôn, the
great king of Argos, arranged his scale of weights
195upon a Babylonian model; and clay contract-tablets
in the British Museum, written in Assyrian
cuneiform with Aramaic dockets, indicate that
from the reign of Tiglath-Pileser (b. c. 745)
downwards Aramaic was the language of commerce
throughout the Assyrian world. And not
only so, but the mana was the standard weight by
which gold and silver were weighed, and all trade
transactions carried out. There was the mana, or
“maund,” of Carchemish, whose position near the
fords of the Euphrates, on the high road to the
Mediterranean, had made it take the commercial
place of Tyre after the destruction of the latter
city by the Assyrians, as well as the mana of
“the country” (of Assyria), or “of the king.”
Thus we find Nergal-sarra-nacir (b. c. 667) lending
“four manahs of silver according to the manah
of Carchemish,” at five shekels of silver interest
per month; and in the Eponymy of Zazai (b. c.
692) a house in Nineveh, “with its shrubbery
and gates,” was sold for one maneh of silver
according to the “royal standard.”

Now the mana might seem at first sight of Semitic
origin. We have the Semitic root הנמ, “to
number,” from which comes the Hebrew mânâh,
“a portion,” and with which the Aramaic mene
that Daniel read on the walls of Belshazzar's
palace is connected, and it would seem to yield a
196good enough meaning for the mana. But this is
put out of the question by the fact that mana in
Assyrian is indeclinable when strictly used, not
even admitting of a plural, whereas, were it a
Semitic word, the nominative and ordinary form
would be manu. It must, therefore, be a loan-word,
and the similarity of the Aryan root ma,
“to measure,” which has given us “moon” and
“month,” might incline us to seek its origin here.
The Greek μνᾶ, however, comes from the Semitic;
and the Semites could not have taken a foreign
root, as distinguished from a derivative, and
formed a technical word out of it; consequently
we must look elsewhere for the home of the mana.
This has generally been supposed to be Egypt, as
the mn is found there also at an early date; not,
however, before the times when the Egyptians
borrowed freely from Palestine not only words
like sus, “horse,” and sar, “prince,” but even
marcabutha, “chariot,” and sepet, “lip.” But
a new light has of late been unexpectedly thrown
upon the matter. An old table of Accadian laws,
which has an Assyrian translation attached, orders
the man who divorces his wife to pay “half a
maneh of silver;” a mild penalty, by the way,
compared with that of the wife, who was condemned
to be thrown into the river for repudiating
her husband. Now the word mana is found in
197the Accadian column, and the vowel harmony
thoroughly suits the structure of the language.
Here, then, we seem to have lighted upon the
parentage of the word, which, after all, would
have come from Babylonia in a truer sense than
the Greek antiquary had any idea of, along with
many other Semitic names of weights and measures,
not excluding even some of the numerals. It is
interesting thus to trace the beginning and growth
of that idea of measure which lies at the bottom
of science as well as of trade; to learn that Babylonia
was its cradle, and a Turanian race its first
discoverers; that the Semites have been imitators
and mediators in the great work of civilisation,
and that the Western nations have through them
inherited the seeds of the culture which they
alone have known how to bring to its fullest
perfection.

Chapter VI.
The doctrine of roots.

All the sciences that bear upon the origin and
early history of man are beginning to point out
more and more clearly that he is a ζῶον πολιτικὸν
in a much wider sense than Aristotle ever imagined.
Instead of starting with atomistic individuals,
we must start with the converse, the
community. The individual is the last growth
and result of time; and society, as composed of
individuals, has arisen out of a sort of beehive
existence, by a process of differentiation which
holds good, as Mr Herbert Spencer has shown,
throughout the organic world. The primitive
savage was but a part of a tribe, with no ideas
beyond those which the tribe possessed in common.
Even the wives and children were common property,
thus realising Plato's Republic in a practical
manner; and special property in a wife appears
to have originated in the acquisition of the women
of another tribe in war. The captive was at the
201mercy of her captor; he might kill her or make
her his slave, in other words, his peculiar wife,
just as he chose. It was the same with other
property; the commune preceded individual possession
everywhere, thus bringing the gregarious
period of human history down to a late epoch of
development. All this throws much light upon
the earlier stage of language. Judging from
analogy, we should conclude that language also,
the artificial link between the several units of a
tribe or community, would have a communistic
origin. We must go back to the beehive era in
order to discover its beginnings. In other words,
language ought first to have been common property,
full of vague, instinctively felt signification, but
not yet differentiated into individual words with
special sounds and meanings. In fact, we ought
to start not with the word, but with a wider
indefinite whole, out of which the word, or rather
the sentence, has been elaborated; and that whole
would have conveyed the same general indeterminate
sense to the several units of the community,
whose wants and means of expressing them were
the same.

Now we have already found that this is actually
the case. If we wish to get at the primary facts
of Glottology, we have to begin with the sentence,
and not with the isolated word. It can never be
202too often repeated, that words have grown out of
the sentence, though each race has carried out
this process differently, in accordance with its
primitive tendency. Everywhere, however, the
general character of the process has been identical.
Everywhere sounds, forms, and meanings have
been differentiated; the indistinct sound, for
instance, that stood for r and l in the parent
Aryan branched off into those two consonants, just
as the obscure sound which serves for c and t in
the Sandwich Islands may yet be resolved into
these letters, and the vowel changes of the verb,
which have no meaning in Sanskrit, have become
the Teutonic ablaut, serving to distinguish the relations
of time. The compound word is pre-eminently
an example of this differentiation: two words
must be so clearly marked off and defined already
as to be able to be connected together to form a
third with determinate form and signification. It
is merely a matter of further progress in the
differentiating direction where the idea contained
in the compound has become so far fixed and
definite as to lose all reference to its original
factors, so that one or both of these are deprived
of all independent force, and convey no meaning
except when united together. Hence the existence
of compounds in a language may be considered a
mark of lateness; before it has acquired them the
203language will have advanced far beyond its period
of childhood; the vagueness of infancy, when
subject and object are blended in inextricable confusion,
will have passed away, and the judgments
that lay implicit in those first semi-conscious
expressions which I have called sentences will
have been made explicit and precise by being
summed up in an ever-increasing number of what
I have called words. The number of words, in
fact, with distinct and separate meanings, measures
the progress of a language and the culture of those
who speak it. Now it is evident that if language
continually tends to enrich itself more and more
with different words and sounds, in order to get
at its beginnings we must reverse the process of
differentiation, and discover those rude chaotic
combinations of sound and sense out of which the
manifold wealth of articulate speech has sprung.
We must go to work in the same way as the
chemist, who obtains his elemental substances by
analysing the different products of nature. Infinitely
various as these are, they have all been
obtained from about sixty simple elements, which,
by combining with one another in different proportions,
have thus differentiated the manifold
properties which each separate combination possesses.
So in Glottology we must throw our
words into the retort of the comparative method,
204break up the compounds, analyse the grammar,
simplify the signification, and trace the growth of
phonetic distinctions. It is in this manner that
we shall arrive at our simple elements, beyond
which it is impossible for Glottology — at all events
without the aid of other sciences — to proceed, just
as it is impossible for chemistry pure and simple
to advance beyond its primary substances.

The roots of language, then, must be reached by
comparison. The truth is of old standing, though
the scientific use of it is of such recent date. The
grammarians of India long before the Christian
era had reduced the Sanskrit lexicon to a certain
number of primitive roots by referring to one
monosyllable all those words the non-formative
part of which agreed in sound; and the Jewish
doctors of the tenth century had resolved the
language of the Old Testament into triliteral
radicals through a comparison of Hebrew with
Arabic. Every one could see that this or that
series of words presupposed the same combination
of letters; it was the root out of which the whole
series seemed to have grown, like the tree out of
the ground. But the discovery remained barren.
The Greeks contented themselves with discussing
whether language had originated by convention
or by nature, and Christian writers took it for
granted that the Semitic radicals formed the language
205of Paradise. It is only since the rise of
Glottology that it has been asked what these roots
are, and what is their relation to the words derived
from them? Now here it is necessary to bear
in mind two things, which have been too often overlooked
in the discussion of this subject. One
is that Glottology cannot go beyond its facts;
and as these are sentence-words and the ultimate
analysis of such sentence-words, it cannot go
beyond the Root-period and speculate as to what
roots themselves grew out of. The bow-wow
theory, or the pooh-pooh theory, or the ding-dong
theory, all lie equally outside the proper province
of Glottology. If we want to decide upon this
matter we must call in the aid of other sciences.
The other thing to be remembered is the loose use
of the phrase “roots of language.” There was no
one primeval language, as I have endeavoured to
show, at least so far as our data allow us to
believe; on the contrary, languages were infinitely
numerous, as numerous as the communities
which spoke them; and it by no means follows
that the roots of all these languages were of a
similar kind, or that words have been derived
from them in a similar way. Indeed, I have tried
to show that so far is this from being true, that
the chief modern races of the world have each
followed a separate and independent direction in
206reflecting their thoughts in speech. Consequently
to talk of finding the roots of language or of
investigating the origin of language is highly misleading.
What we have to deal with are the roots
of languages. The results obtained from the study
of the Aryan group are not to be applied universally,
and be made the rule for Semitic and
Turanian also. What we can do, however, is to
investigate the roots of the various families of
speech so far as is possible, and then to compare
the conclusions drawn from each. Among the
many one-sided theories produced by an exclusive
regard to the Aryan family, none is so common
as that which ascribes to roots a general abstract
meaning, as if our ancestors of the Root-period
employed nothing except abstract terms in conversing
with one another. We have only to state
the proposition, however, to see how absurd it is.
How could savages, whose vocabulary consisted
entirely of such words as “bringing,” “shining,”
“defending,” be mutually intelligible? There is
no common bond of intelligibility between such
universal ideas; language must begin with the
objects of sense, if we are to communicate our
meaning to others, and rise from these by the help
of metaphor to abstract supersensuous conceptions.
Moreover, these abstract ideas must either be the
last result of reflection, the universals arrived at
207after a long course of education, or else must be
of the vaguest and most unmeaning character.
In the first case, we are ascribing to the primitive
barbarian the mind of the civilised man; in the
second case, any language at all would be out of
the question. Two persons could not talk together
in vague generalities, more especially when their
conversation would be mostly confined to the bare
necessities of life. Even with us the same
general term bears very different meanings to two
different persons. It is what Locke called a
“mixed mode;” and with all our culture and
scientific definition it is impossible to make such
epithets as “good” or “noble” convey exactly
the same signification and the same associations
to two minds. In fact, the notion is absolutely
contradicted by what we observe among modern
savages. Here the individual objects of sense have
names enough, while general terms are very rare.
Thus the Mohicans have words for cutting various
objects, but none to signify cutting simple; and
the Society Islanders can talk of a dog's tail, a
sheep's tail, or a man's tail, but not of tail itself.
“The dialect of the Zulus is rich in nouns denoting
different objects of the same genus, according to
some variety of colour, redundancy or deficiency
of members, or some other peculiarity,” such as
208“red cow,” “white cow,” “brown cow;” 174 and the
Sechuana has no less than ten words to express
horned cattle. 275 The Tasmanians were so utterly
deficient in the power of forming abstract ideas,
that they were obliged to say “like the moon,”
or some other round object, when they wanted to
express the conception of roundness. The theory
in question has originated in a too exclusive
attention to the phenomena of the Aryan
lexicon. Here all the roots seem to bear a
general meaning only, out of which the names
of individual things have been obtained by means
of suffixes. Thus daughter (duhitâ) is merely
“the milker,” from the root which has the
general signification of “milking;” father (pater,
pitâ) is “the defender,” from pâ; brother
(bhrâtâ) is “the bearer,” from bhar. In the
same way, a large proportion of the words we use
turn out, when analysed, to be simply general
epithets which have come to be set apart to denote
some special object. Hence the conclusion
to which the Sanskritists jumped, that the general
precedes the particular, and their triumphant refutation
of the onomatopœic hypothesis of the
origin of language. But they have forgotten that
their induction has been made from a single instance
209only, and that instance altogether exceptional
in the history of speech. The parent-Aryan,
if it ever existed, was the language of
comparatively civilised men. Such examples as
duhitâ would of themselves show this, and point
to a pastoral life; and the persistency with which
the several members of the original stock have
remained true to the primitive language can only
be explained by supposing our ancestors to have
advanced considerably beyond the degree of civilisation
at present possessed by the Ostiak or Burmese
tribes. The Aryan scholar, therefore, is
dealing with a language in which we may well
expect to find general epithetic terms; but he
cannot conclude from this that there were no
individual words originally which denoted some
particular object. 176 Beyond the parent-Aryan lies a
vast unknown period, upon which Glottology casts
but little light; and the fact that in so many
unallied languages the names of father and mother
are formed by means of the labials, would seem to
imply that pitar and mâtar were chosen not without
a purpose; and although the lexicographer
210must derive these words themselves from pâ and
mâ, “fashioning,” 177 they yet point to a time when
the names given to the parents were merely the
first cries of infancy. “Father” and “mother”
must have had names before the root tar was
compounded with the roots pa and ma to denote
them. 278 But the error of the Sanskritists goes deeper
than this. They raise into a sort of pigeon-English
language the residuum of sounds which lies
at the bottom of the dictionary. Because a certain
number of vocables presuppose a common
monosyllable with a common vague meaning, it
does not at all follow that this monosyllable ever
formed part of an actual language. For anything
we know, it may be merely an archetype of
phonetic sound, presupposed by the derivatives,
but never consciously expressed in speech. Still
less can we assert that the vague general signification
given to the root was originally expressed
211by it. The root may have originally denoted an
individual object or action, which was afterwards
lost when the progress of composition and phonetic
decay had supplied the vocabulary with other
terms. There is, however, a truth in the prevailing
theory, though thus faultily expressed. The
sentence comes before the word, the indefinite
before the definite; and the root-period, as we
have seen, is characterised by the want of differentiation.
The Aryan root, consequently, while
primarily denoting an individual object, would
have done so in a very different way from that in
which we should denote the same. The individual
can only be properly understood in relation to
the general; when, therefore, the idea of the
general has not yet been arrived at, the idea of
the particular is at once vague and sensuous.
The word which denotes it is merely a mark,
nothing more; just as much as a proper name,
and with no more subjective reference than the
proper name has. So long as the object can be
pointed out sensibly, the meaning and reference
of the word is unmistakable. We know exactly,
for instance, who a particular John or Henry are
when they are indicated by the finger; but when
the object is not present, the signification and
content of the word is wholly vague and uncertain.
The judgment which is summed up in it is
212not determined by immediate reference to such
and such a thing: we cannot think “this is a
tree;” and accordingly each person forms his own
judgment, and attaches a different interpretation
to the vocable. The term is not defined by its
external object, and language has not yet arrived
at the explication of its words by other means.
In this way the Aryan roots might easily have
come to have those vague general significations
which are ascribed to them, although they properly
represented individual objects and actions.

But we must not forget that the so-called root-stage
of Aryan speech is very questionable in the
shape in which it is usually set before us. So far
as our data go, there is no reason for believing that
the Aryan was ever otherwise than inflectional,
however unlike the primitive inflections may have
been to those with which we are familiar. We
can only be certain of this much, that there was a
time when the primitive Aryan spoke a language
far simpler than that with which we are acquainted,
in which the words were for the most part
monosyllabic, few, and of indefinite meaning, and
that this earlier and barbarous condition was succeeded
by what I would call the epithet-stage.
To identify this epithet-stage, however, with inflection
is altogether unwarranted, and (as I hope
I have shown) contrary to the facts. The root-period
213is not inconsistent with a rudimentary
inflection, and the epithet-period points to a vast
series of bygone ages, to an advanced civilisation,
and to the development of the higher poetical
faculties. When the moon could be called “the
measurer,” the tribe must have left barbarism far
behind. It was still a tribe, however, and we
may perhaps assign to this communism the
general adoption of particular epithets for special
objects, and the tenacity with which they were
preserved and handed down when once adopted.
At any rate, the individual had not yet emerged
from the community; but this was inevitable
when the imaginative faculty had once made its
appearance, and the era of the Rishis could not
be long delayed.

What I have called the epithet-stage is of great
importance in the history of our group of languages,
since it supplies in great measure the answer
to the question which came before us in an earlier
chapter, why it is that the Aryan family presents
such a singular exception to the usual rule
of rapid change in language in the fixity of its
grammar and lexicon. Before the parent-tribe
had broken up, it had already entered upon the
later period of linguistic growth, in which conventional
custom sets its stamp upon spoken speech,
and consecrates its form and expression. Language
214loses its early creativeness; the very fact that
new words have to be coined out of old material
by a metaphorical use of the latter shows that
settled habits and the enlarged sphere of imagination
have to a large extent put an end to the
invention of fresh “roots,” while the common
adoption of one of these metaphors to express an
object of sense demonstrates the extinction of the
creative faculty and the stereotyped conservatism
of the speaker. Men have become at once too
highly imaginative and too narrowly conventional
to waste their energies in the pastime of the savage,
the coining of new words. In fact, language
has entered upon its ceremonial stage, when the
sounds which we utter have been made the subject
of a conscious exercise of thought, and the mind
has been called upon to compare some new object
with one whose name has already been furnished
by the ancient heirlooms of speech. Sound and
sense are no longer commingled in chaotic confusion;
sense becomes distinct and clear, and
sound is made subordinate to it. A so-called
“ceremonial language,” such as the Bhasa Krama
of Java, is but a farther development of the
epithet-stage, by definitely confining the epithets
to persons and not things. Ceremonial languages
and idioms are found all over the world, as in the
larger islands of Polynesia, or in the ceremonial
215conjugation of the Basque, or in the women's language
of South America; and they testify everywhere
to an incipient fixity of language, and the beginning
of a settled state of society. Closely akin to
these ceremonial languages is the phenomenon
which meets us in several of the North American
dialects, where the pronoun na, ni, or nu, “my,”
has become an inseparable and meaningless affix
of numberless words, just as in the Continental
“milord.” Thus “head” is in Mbaya na-guilo,
in Abiponian na-maiat, in Moxa nu-ciuti; “eye”
in Mbaya is ni-gecoge, in Abiponian na-toele, in
Moxa nu-chi, and in Mokobi ni-cote. The ceremonial
or epithet-period of language is that in
which I would place the origin of the personal
pronouns. Bleek has shown that these were originally
substantives, meaning “servant,” “lord,”
“reverence,” and the like, at least so far as the
Ba-ntu idioms of South Africa are concerned; and
the same fact appears in those languages of Asia,
such as Chinese, Malay, and Japanese, in which
the transparent character of the language allows
us to penetrate to their primary signification.
Thus the Malay ulun, “I,” is still in Lampong
“a man;” and the Kawi ngwang, “I,” cannot be
separated from wwang, “a man.” To assert that
this transmutation of expressions like “your reverence,”
or ὅδε ὁ ἀνὴρ, into personal pronouns belongs
216to a late epoch of linguistic development, is to
re-state my own position in other words; while
the attempt to resolve the nominative of the
Aryan first personal pronoun (aham, ego), for instance,
into two “pronominal elements,” ma + ga,
breaks down at the very threshold. Initial m
is never lost in the Aryan languages generally,
although it may disappear in Greek through the
medium of the digamma, as in μάλευρον by the
side of ἄλευρον, i. e., Ϝαλευρον from Ϝαλέω, or
μί-τος by the side of ἰτέα, the Latin viere, vimen;
while ga, the Greek -γε, is still found in the Rig-Veda
as the aspirated gha. To metamorphose the
singular ma into the plural nas, as has been
attempted by some over-hasty adherents of the
pronominal theory, does violence to all the phonological
laws of Indo-European speech. In my
“Assyrian Grammar,” I have suggested that a
comparison of the cognate dialects would lead us to
infer that the original form of the first two personal
pronouns in Semitic was the same, 'ecet, which
reminds us of the Ethiopic acata, “to honour” or
“thank,” while the third personal pronoun can be
proved to have originally been su'u, which may
be akin to הוש, “like,” and hence “companion.”

The epithet-stage, therefore, would have been
the closing portion of the root-period, or the
commencement of the secondary period of analysis
217(not of flection), according to the point
of view which we prefer to take. Its determination
can no more settle the nature of roots
or the existence of flection during the root-period
than those falsely-called flections like -dom and
-head, which I discussed in a former chapter.
The root remains where it was before, — the
residuum of a group of words in which the lexicographer
has discovered a common combination
of sounds and a common meaning, but which
could never have formed part of a spoken language,
and which, from the first, while denoting
the individual and the concrete, was yet vague in
meaning and indefinite in pronunciation, and
capable of being used for all the parts of speech.
This, however, was owing to the fact that language
begins with the sentence and not with the individual
word; the latter is the last growth of time,
the last result of simplification and reflection. Of
itself, the radical was as purely the mark of a
single object of sense as any of the words which
denote various sorts of tails in the idioms of the
Sandwich Islanders; but this signification was
extended by its use as a sentence-word or judgment;
for, properly speaking, the primitive Aryan had
no conception of a single object apart from the
universal: such a distinction requires comparison,
and as yet individual and, general were blended
218into one, the general being an extended individual,
and the individual a specialised universal. Without
a world there can be no individual. Now
the naming of each thing according to its momentary
impression upon the senses would necessarily
give rise to an infinite multitude of names,
not only for objects which seemed to differ in
some small particular, but also for the same object
according to the time or circumstances under
which it struck the senses. This, coupled with
the early creativeness of language, which we still
see exemplified among the lower races of mankind,
would produce an endless number of words. The
vocables, which at different times or at the same
time served to point out the same thing, would
have been as multitudinous as the dialects which
I have endeavoured in a preceding chapter to
show were the real primitive centres of spoken
speech. This is the only way in which we can
account for the existence of synonymous roots,
which become more plentiful as the language
which we are handling is less developed. Thus the
Caribs express the same notion by very different
roots, according to Adelung; 179and Professor Key
is not the only person who has been astonished at
the immense number of radicals in Sanskrit
which all mean “to go.” We have already had
219occasion to notice in a preceding page the fertility
of savage tribes in inventing new words, and the
rapid change in the vocabulary that takes place
among them. Perhaps one of the most striking
instances of this in recent times is to be found in
the island of Tasmania, where, with a population
of no more than fifty persons, there were no less
than four dialects, each with a different word for
“ear,” “eye,” “head,” and other similarly common
words. Even our own semi-fossilised language
has not altogether lost the power of striking
out new roots, as may be proved by a reference to
a slang dictionary or a scientific encyclopædia;
and this may give us some idea of the boundless
inventiveness of language before it had been
crystallised by convention and a fixed society. In
fact, just as the languages of the world with which
we are acquainted have arisen out of the wrecks
of numberless forgotten attempts at speech, so
the roots presupposed by the lexicon are the selected
relics of an infinite wealth of primitive sentence-words;
for here, too, as elsewhere, natural
selection has come into play, and the progress of
civilisation has been to unify and minimise the
inexhaustible prodigality of nature. In the same
way, the indefinite variety of meanings which it
was possible to evolve out of each sentence-word
was gradually reduced, until every idea had its
220own appropriate sound, and the sentence was
resolved into its individual words, like the word
into its individual letters. But this individualising
of the isolated word is the last result of time
and thought; and so far as our data warrant us
to infer, there was never a period when the root
existed in its naked simplicity, any more than there
was when the letter or the syllable existed apart
from the root. Both are figments of the grammarian
and the lexicographer, the convenient analyses of
the modern student. Flection in the Aryan
tongues implies a preceding flection upon which
it was modelled; and a large proportion of the
radicals, as we have seen, can only be used for the
purposes of comparison by being treated as bases.
This at once makes them dissyllables, that is, no
longer monosyllabic roots — the same conclusion
to which we are led by a consideration of such
words as bhûs, bhûm, sthât, sthâtar, with identical
significations; and when we recollect that k was
primarily always followed by u, we see that there
is a whole class of roots like loque-or, which could
never have existed in a monosyllabic form in spoken
speech. Forms such as ad-mi, which present us
with the bare radical immediately attached to the
later inflection of the epithet-epoch had better be
explained as the consequences of phonetic decay,
like our own English monosyllables, than as the
221evidences of an imaginary “root-period,” since
the tendency of language is towards attrition and
contraction, rather than extension and increase.

Another one-sided theory, which has for some
time formed part of the doctrine of roots, is, that
we must seek in them for the origin of language.
Accordingly we have had attempts to derive them
from the imitation of natural sounds, or from
emotional interjections, or again from a kind of
intuitive inspiration. Geiger believes that they
have originated in the attempt to imitate the gestures
and muscular expression of emotion; Bleek
would evolve them from the cries of animals, or
rather the inarticulate sounds made by the anthropoid
apes. The failure of these attempts, the impossibility
of supporting any one of them by the
facts of language alone, has brought about a
reaction against inquiries into the origin of
language at all, and the Societé de Linguistique
of Paris has refused to receive any papers bearing
upon the subject. But because the determination
of the matter lies beyond the boundaries of Glottology
taken by itself, it by no means follows that
it is either useless or insoluble. On the contrary,
Glottology is an historic science, and we can
never, therefore, understand the problems of
language properly until we have solved the riddle
of its origin. But this can only be done by the
222aid of other sciences; Glottology cannot go beyond
the limits of language, and physiology and psychology
must explain the rest. As glottologists, we
have to begin with roots; they are the first facts
to which we can ascend. The decomposition of
the roots themselves, the germs out of which they
have grown, belong to other branches of study.
All that we can do is to ascertain clearly the
nature of these roots and to fix their limits; to
determine, in short, where language first takes its
start, and ceases to be the inarticulate, unconscious
utterance of instinctive desire. The difficulty that
meets us here is one that presents itself everywhere
to the student of nature. There is no break, no
sudden gap in nature; all follows in a regular
unbroken order. All sharp lines of demarcation,
therefore, must be artificial; our genera and
species, our strata and our periods, in fine, our
classifications generally, exist only for the purposes
of science. Ideal types there certainly are,
around which the phenomena group themselves;
but the groups pass insensibly one into the other,
and we can only draw our lines of division to a
great extent arbitrarily. This is the case with language;
we can determine on which side of the
line language must be placed, and on which side
mere inarticulate cries, but the line itself is a
shifting one, and can only be laid down approximately.
223To believe, therefore, that roots are
simply interjections or the imitations of sounds, is
to confuse the two sides of the line of division,
and to ignore the difference between language and
inarticulate utterance. Roots are not emotional
or imitational cries, although they may have
grown out of them; but the investigation of the
process has nothing to do with the science of language.
The Aryan dictionary may be reduced to
a certain number of radicals; but after all, we
have only found the origin of the dictionary, not
of language. Consequently it is beside the mark
either to quote instances of derivatives from interjections
or natural sounds, like the Chinese ngŏ,
“stop,” and miau, “cat,” in defence of the “pooh-pooh”
and “bow-wow” theories of the origin of
speech, or to attempt to refute them by showing
that supposed examples of imitation, like thunder
or raven (corvus), turn out to have their origin in
roots of very different sound. The utmost that
Glottology can do is to show that words have
actually been derived from both these sources
within the historic period; and in that case
analogy may justify us in concluding that the
primitive man may have arrived at his roots in a
similar manner. But there is no proof of this,
so far as philology is concerned; and although
the mind may pass from vague natural cries into
224the higher forms of speech when it has come to a
state of consciousness, it is hard to see how the
process could have been performed when the mind
was yet unconscious — how, in other words, the
mind could have passed from unconsciousness to
consciousness and its expression. To say that this
happened through intuitive inspiration is merely
to state the question in a different way. We
want to know where this inspiration came from,
and how the mind first became conscious. But
this is plainly a matter for psychology and not
Glottology; and we can only see this much, that
as language is the outward expression and embodiment
of conscious thought, it must have had
much to do with the development of consciousness,
which becomes possible when thought can make
itself objective and so regard itself. Language is
the counter-side and utterance of society; with
society it begins and with society it ends. Before
society there is no language properly so called,
because there is no conscious thought, no intercourse
between man and man; and consequently
our linguistic researches will be bounded by the
limits of social science and social archaeology. As
we cannot get beyond the family in the one, so
we cannot get beyond the existing monuments
of speech in the other. It is clear, then,
that Glottology must confine itself within the
225boundaries of the period of roots, and transfer its
attention from the question of their origin to the
investigation into their nature. In a former chapter
I have endeavoured to point out that roots
are by no means necessarily monosyllabic, and
that the theory that they are so is one of the idola
generated by the over-weight given to the Aryan
family. It is bound up with the belief that the
Semitic radicals were originally biliteral. The
latter notion has been much encouraged by the
analytic character of Aryan, and the essential
difference between the two families of speech has
been overlooked. But although the attempt to
resolve the Semitic roots into more ultimate
elements breaks down, it does not at all follow
that the result is the same in the Aryan group.
Throughout this reigns the spirit of analysis, and
it is very possible, therefore, that the Aryan roots
are capable of still further decomposition. Composition
and inflection are the distinguishing features
of this family of speech, and the so-called root-period
may be only the closing era of a still older
root-period. This probability is strongly confirmed
by a fact which it is hard to explain from any
other cause, the occurrence, namely, of roots with
similar meanings which differ only in the final
consonants. Thus we find beside bhâ (φημί),
bhan (φαίνω), bhas, and bhav (φαῦος, favilla),
226beside stâ (stare), stap (stipare), stambh
(stamp), star (στερεος), stal (stellen), and stav
(σταῦρος). In accordance with this, Professor Pott
has sought to analyse the so-called roots, and to
make out that all those which enclose two consonants
are compounds, so that the earliest form
of Aryan would have resembled the Polynesian
dialects, in which each syllable must end in a
vowel. A large proportion of these compounds,
Pott believes, contain a preposition; thus pinj,
“painting,” comes from api (ἐπὶ) and anj,
“touching.” Professor Curtius 180 urges several
objections of great force against this view.
In the first place, these compounded roots are
treated in word-building just like other primitive
roots, and whereas initial api in Sanskrit may
become pi, this is never the case in Greek. The
loss of the vowel, therefore, is a peculiarity of
Sanskrit, and could not have occurred in the
parent-Aryan. Then, secondly, there was no such
close and intimate amalgamation of the preposition
and the root in early times as is necessitated
by Pott's analysis. Even in Greek and Sanskrit
the nominal and independent origin of the prepositions
is so clearly felt that the augment and the
reduplication are inserted between the preposition
and the verbal form. Latin and Greek themselves
227possess but few compounded roots in
common.

But although Pott's theory must be resigned, it
is yet certain that many of the roots are really
compounds. The radical yu cannot be separated
from yug and yudh, or the radical tar from tras
and tram, trak (torqu-eo) and trap (trepidus), trib
(τρίβω) and trup (τρύπ-ανον). Curtius 181 suggests
that the longer forms are really compounded out
of two other roots, yudh, for instance, being amalgamated
with dha (“do”), and the k in trak being
the same as the guttural which distinguishes λιθακ
from λιθο. In this case the compounded roots
would have originally been dissyllabic, yu-dha and
tar-ka. The suggestion is undoubtedly a true one.
We can hardly explain in any other way such
roots as vridh and ridh, “growing,” dâ and dam,
“binding;” and the theory would be very consistent
with the view that the root-period, as far as
it existed at all, was a period of rudimentary
inflection, which preceded the more advanced
epoch of epithet-making. The theory is also
borne out by the analogy of the Turanian languages.
This group is still getting but scanty
attention from glottologists, and until lately we
could only study it in modern idioms. Accadian,
228however, has given us a background for comparison
older than the language of the Rig-Veda;
and the clear transparent character of the Turanian
group enables us to obtain more certain results
than where we have to contend with all the
obscurities of phonetic decay. Now, the Accadian
roots, simple as they appear, nevertheless contain
compounds in which the elements are as closely
amalgamated as they would be in the Aryan roots
were Curtius's opinion correct. Thus is, “a heap,”
is compounded with ê, “house,” to form es, “a
building,” and with me, “multitude,” to form
mes, “many.” The latter word would be of great
antiquity, if, as I believe, the final s which marks
the third person plural of the past tense is a remnant
of it. We thus have composition in Turanian,
in which the two factors are so welded together
as to become practically one word, carried
back to a very remote period; and yet the genius
of the Turanian languages is thoroughly averse to
composition at all. But we must never forget
that we may easily carry analysis too far. We
cannot judge the primitive savage by our rules of
simplicity. On the contrary, simplicity is the
result of progress and culture; the further we go
back, the nearer we approach the natural state,
the more do we meet with the intricate multiplicity
of nature. Nothing can be more intricate,
229more complex, than the grammar of the Red
Indian or the Eskimaux; the simplicity of our
own grammar is the result of a long series of
comprehensive generalisations and analyses of
thought. Out of the manifold comes the simple,
out of the multitudinous the single. All progress
in philosophy and science is the reduction of the
many to the one. It is the same with the lexicon
as with the grammar. The meaning of words
begins with a confused vagueness, out of which
definite forms with definite significations are
gradually evolved. Language is the expression
of thought; and the first ideas were as much undifferentiated
embryos as the jelly-fish on the
shore or the beehive life of primeval man.
There was no unity in them; idea had not yet
been subordinated to idea; but each was the mere
individual impression of the moment, with all the
vagueness and complexity of a sensation. Accordingly,
we must not expect to find simplicity of
form, any more than simplicity of content or signification,
in the root-period; and the reversal of
this is the most serious argument against Pott's
hypothesis. As Bleek points out, many of our
involuntary sounds, such as sneezing, for instance,
are by no means simple and monosyllabic; and
whatever may be the origin of language, it is
certain that on the phonetic side, the side, that is,
230of the non-mental physiological machinery, we
can draw no distinction between emotional cries
and articulate utterance. The clicks of the Hottentot
cannot be called either simple or easy; and
yet it is impossible to explain these as a later
accretion to the language; they go back to the
very roots of it, and may possibly be a relic of
what once characterised most of the other languages
of the world, but has since been lost
through the influence of phonetic decay. Phonetic
decay is but another name for laziness, for the
effort — for effort it is — to save trouble in speaking;
and it is the great principle of change in all
languages. We can no longer talk of the interchange
of letters, except loosely; sounds can only
pass into one another in accordance with strict
physiological laws, and the action of these is
determined by the endeavour to facilitate pronunciation.
K, the harder sound, may become h, but
the reverse cannot take place unless other laws
interfere. When, therefore, we find that an
English t answers to a Greek δ and a German z,
we cannot suppose that the more difficult t has
been adopted instead of the easier d; and yet, to
assume that the Gothic t has remained faithful to
the original sound, while the d of Sanskrit, Greek,
and Latin exhibits phonetic decay, would leave
the High German z altogether unexplained. The
231only interpretation of the facts which is allowable
is, that all these sounds have been independent
differentiations of one original obscure sound
which contained within itself the other clearer
consonants; just as the meaning of the root-word
has been gradually worked out, until the undeveloped
conceptions that lay implicit in it have
been severally marked off one from the other. 182
My friend Mr Sweet has come to the conclusion
that primitive man could only roughly distinguish
between sounds, just as he could only roughly
distinguish between ideas and the relations of
grammar. The belief is borne out by all the facts
which we have at our disposal. The musical ear
232is as much the creation of a high civilisation as the
eye of the painter, and the modern savage finds his
music only in the rudest and coarsest high-pitched
notes. It is naturally the same with phonetic
speech. The appreciation of the delicate distinctions
of sound which have resulted in poetry and
music on the one hand, and in languages like
Greek on the other, is unknown to the barbarian.
The Sandwich Islander cannot discover any difference
between c and t; and when we rise higher in
the scale of civilisation, we see the Chinese transforming
Christ into Ki-li-sse-tu. 183 The further
back we push our phonological researches, the
greater becomes the number of neutral sounds.
Ancient Egyptian made no difference between r
and l, and a comparison of roots would show that
the same was the case in the parent-Aryan.
Arguing from the alphabet, we should conclude
that Sanskrit was once unable to distinguish
between b and v, and Assyrian writes m and ν
with the same character. Finnish has but eleven
consonants, and no Polynesian language more
than ten; while some Australian dialects contain
233only eight, with three variations. 184 All this would
go far to show that the number of sounds possessed
by early language was extremely small,
and that these were mostly of a neutral, indistinct
character, and what we should consider difficult to
pronounce. A satisfactory explanation would thus
be afforded, to a certain extent, of the phenomenon
before alluded to, the existence, namely, of roots
in the Aryan family which differ in the final
consonant or consonants, but which cannot be
separated from one another, owing to their
similarity of meaning, and the identity of their
initial or characteristic sound. The same is yet
more conspicuously the case in the Semitic group,
where roots repeatedly occur which agree in signification,
but have different letters, though of the
same class. Thus ץצק, ססק, זזג, הזג, םזג, עזג, לזג,
רזג, דדח, דדג, דדק, הדג, זוג, ץצח, הצח, עצק, רצק, חסכ, םסכ, הצח, all seem to contain the idea of
“cutting,” and little distinction can be made
between דונ, דדנ, and הדנ. 285

The root-period, therefore, was characterised by
complexity, indistinctness, and vagueness in sound,
meaning, and grammar. It was but a reflection
of the hive-like community, in which the parts
were as yet undistinguished, and the several
factors of society lay undeveloped in a single
234embryonic germ. It was a life of the senses
rather than of the mind, in which the past and
the future were equally ignored, and language was
employed in the service of the bodily wants,
principally of hunger. Consequently we cannot
expect to find any traces of spiritual and intellectual
conceptions in this early stage of articulate
speech. The oldest roots are of the most purely
sensuous description, and the words which denote
the higher ideas of religion or mind are derived
from these by the help of metaphor, metaphor
itself having its basis in the objects of sense.
Thus in the Aryan family, deus, Ζεύς, was “the
bright” heaven, anima and spiritus are “the
wind,” and soul comes from “the heaving” of
the sea. The Semitic ruakh, “the breath” of life,
is simply the breeze, and el, “God,” is “the
strong” one. The numerals have been arrived at
in the same way: three was originally “that
which goes beyond” (root tar, trans, &c.), four
was “(one) and three” (cha-twar), nine was the
“new number” (navam). 186 Even the pronouns
themselves may have a similar sensuous origin.

This brings me to the last idolum connected
with the doctrine of roots to which I shall refer.
235It is generally known as the theory of Pronominal
Roots, and assumes that language at its first
starting possessed a large number of words which
had a demonstrative meaning only, and formed
a great part of the material of inflection. The
theory is another result of the attempt to analyse
flection by a comparison of the Aryan languages
alone. We meet with certain roots, such as ta,
sa, ya, which we cannot trace back to any other
signification than that of the demonstrative pronoun.
Because our data fail us, however, we are
not justified in asserting that the demonstrative
meaning was the original content of these roots.
Our ignorance does not allow us to do more than
affirm that these roots had a demonstrative signification
so far back as we can go. But to suppose
that such was their first and original force leads
us into great difficulties. We may pass over the
objection that the inventors of language would
not have found such words mutually intelligible,
as this might be explained by the instinctive
uniformity of understanding which pervaded the
beehive community; but how could the savage
elaborate them without any idea of contrast?
Here implies there, this implies that; but in the
root-period, in the beehive life, all was here and
all was this. This is the essential nature of words
with the chaotic vagueness of meaning which we
236have seen characterised the so-called roots, as well
as of a life of the senses, in which man is conscious
of the passing moment only. Moreover, what
need could there have been for such words, when
the root contained within itself all the signification
that could be expressed in speech, primarily
denoting the individual object, and secondarily —
since there was no idea of contrast, and so of
distinction between the individual and the general
— all individual objects? Any further specification
that was required could not be pointed out in
language; it called for the finger and the eye.
If language starts with sentences, it cannot start
with the demonstrative, which is not a sentence.
But observed facts in other families of languages
do not support the pronominal theory. In Japanese
the same word may stand for all three
persons; but this is not because it was primitively
a demonstrative, but because it was a substantive,
such as “servant,” “worshipper,” and
so forth. Chinese ki, “place,” has become the
relative, and the Semitic relative, whatever its derivation
might be, was properly the demonstrative.
A close similarity has been observed in many languages
between the demonstrative and the substantive
verb, and this again has in several instances
been traced back to a sensuous origin. In the same
direction points the formation of the demonstratives
237by a change of vowels, of which Mr Tylor
has collected so many instances, and to which
others might be added. Thus in Javanese iki is
“this,” ika “that,” iku “that there;” in Japanese,
ko is “here,” ka “there;” in Zomba, na,
is “this,” ni “that;” in Carib, ne is “thou,” ni
is “he;” in Brazilian Botocudo, ati is “I,” oti
is “thou;” 187 in (African) Tumali, ngi is “I,”
ngo “thou,” and ngu “he.” It is the same
with Malay and Siamese, which possess an extraordinary
number of pronouns of the first and
second persons, employed according to the rank
or age of the speaker, but which are really
so many substantives. Such a distinction by
phonetic means alone implies a late period of
linguistic development; one of the forms must
have preceded the other; and in this case there
would, have been no contrast, no this and that,
and consequently no possibility of expressing
the demonstrative. It is plain that substantives,
and not pronominal words, would first have been
differentiated in this way; and accordingly we
find the Carib baba, “father,” contrasted with bibi,
“mother;” the Mantschu chacha, “man,” and ama,
“father,” with cheche, “woman,” and eme,
“mother;” the Finnic ukko, “old man,” with akka,
“old woman;” and the African Eboe nna, “father,”
238with nne, “mother,” where a pretended pronominal
root makes its appearance. Similarly the distinction
between the primary numerals is denoted in
the same manner in many languages; thus in
Lushu, tizi is “one,” tazi “two;” “three” and
“four” are ngroka and ngraka in Koriak, niyokh
and niyakh in Kolyma, gnasog and gnasag in Karaga,
and tsúk and tsaak in Kamschatkan. But
the expression of grammatical relations by internal
phonetic change clearly cannot belong to a
period when the broadest differences of sound were
confused together, and the utilisation of delicate
vowel-distinctions to denote nuances of meaning
was utterly unknown; and accordingly we find
not only the Aryan languages employing vocalic
changes to represent verbal differences of signification
only gradually and tardily, but even Semitic,
in which internal vocalic change plays so large a
part, has developed the three case terminations -u,
-i, -a out of an original a, while the Bedouin even
now pronounces his vowels so indistinctly that it
is often impossible to say which vowel precisely
is represented. In fact, the pronominal root
theory is the product of the belief that the inflectional
stage of Aryan was preceded by an
agglutinative stage. Without the assumption of
pronominal roots, which might mean anything or
nothing, it was found impossible to explain many
239of the case-endings. But the matter seems but
little mended when we lay down that the nominative
and genitive singular as well as the plural
number are all formed by means of the same pronominal
suffix with the common signification of
“that.” 188

There is one point connected with this subject
of roots which must be touched upon before we
finish the present chapter. The several members
of the Aryan family, while agreeing in the main
body of their roots, yet exhibit others which
seem peculiar to each. Greek, Latin, Teutonic,
each appear to possess a certain number of radicals
240which cannot be attached to roots found in the
cognate languages without doing violence to all
the laws of the change and development of signification.
There are many words the etymology
of which can never be settled by Glottology, or,
to speak more accurately, which refuse to be compared
with allied words in other dialects. To
attempt to discover a derivation for every word in
the Greek Lexicon will only end in error and
discomfiture. We seem forced to conclude that
the different branches of our race have, beside their
common stock of roots, others of native and
peculiar origin and growth. The residuum of unconnected
roots which scientific philology leaves
in each Indo-European language is an evidence
that language is still living, is still the outward
expression of an active progressive society.
Literature and civilisation will do much to restrain
that unbounded license of striking out new
words which, distinguishes the idioms of savage
tribes; but our own age and country will still
produce such inventions as absquatulate and swoggle,
which cannot be reduced to any common
Aryan radical. They have come into the world
fully formed, however much they may contain
sounds similar to those in words of like meaning;
and this single fact is a striking commentary upon
the belief that our ancestors once spoke a language
241of roots. The root is the unconsciously conceived
mental block, as it were, out of which our words
are shaped; but to imagine that it was ever
consciously realised in speech by a race which was
afterwards to evolve inflection by some unexplained
means, is not only improbable, but opposed
to the data before us. As Professor Pott has said, 189
“There is no inward necessity why roots should
first have entered into the reality of language,
naked and formless; it suffices that, unpronounced,
they fluttered before the soul like small images,
continually clothed in the mouth, now with this,
now with that form, and surrendered to the air to
be drafted off in hundred-fold cases and combinations.”242

Chapter VII.
The metaphysics of language.

The term “Metaphysics of Language” has not
been very happily chosen. It can only be defended
upon the ground that pure being and pure thought
are identical, and that the generalisations which
sum up the several phenomena introduce a mental
element foreign to the phenomena themselves, and
may therefore be considered to partake of a metaphysical
character. From this point of view all
scientific laws will be more or less metaphysical;
and we can hardly refuse this title to such transcendental
conceptions as that οf force. A conception
like this has nothing answering to it in
material nature. We see certain phenomena happening
cotemporaneously or in succession, and
we imagine a bond or power of which these are
the result and manifestation, and to which we give
the name of force. Yet, after all, this power is
merely a mental product which we project into the
world of the senses. Similarly the fundamental
243postulates of mathematics pass beyond the reach
of direct experience. We know very well that, so
far as our experience has extended, when we place
two things by the side of two other things we
have four objects before us; but what that conception
of four is in itself is a matter of which the
senses alone cannot inform us. There are some
tribes who cannot count beyond three, or rather
are unable to generalise so far as four. What
numbers are in themselves, what they mean and
how they originate, or whether they are universally
true, are metaphysical questions. However much
their verification may belong to observation and
experiment, the radical ideas of number generally
and of the numbers specifically fall under the sphere
of metaphysics. The metaphysics of language,
accordingly, will be those general mental conceptions
which underlie the phenomena of articulate
speech, and to which an induction of the latter
will conduct us. Thus we shall have to place under
this head all inquiries into the origin and nature of
gender or of declension, the nature of these in an
historical science necessarily implying a knowledge
of their origin. Such inquiries are no new
thing. From the days of Plato's Kratylus downward
attempts have been made to solve the obvious
questions raised by a consideration of language.
The Greek disputed as to whether language
244originated by convention (νόμῳ) or by nature
(φύσει), and according to the system of philosophy
he adopted, ranged himself on either side.
The modern form of the discussion would be
whether or not the relations of grammar, along
with the words which expressed them, grew up
spontaneously and instinctively, or were settled by
an arbitrary compact among the first men? or in
other words, whether grammar is an invented art
or the necessary development of mind? I say
grammar, and not vocabulary, because although it
was the single word which at first sight seems to
have attracted Greek speculation, it was really the
relation of the word to the mind and the grammatical
fulness of meaning which was implicit in
it. The word was regarded from the side of its
content, and not of its outward form; and this
perhaps was inevitable when the native language
alone was known, and education was oral rather
than literary. The attention is not so likely to be
centred upon the external sound of words until
they are written down and analysed into syllables
and letters. Hence it is not surprising that the
early speculations into the character of articulate
speech did not result in a formal grammar until
the Greek language had been brought into collision
with the Latin, and the critical era of
Alexandria had succeeded to the old political life
245of Greece. A regular grammar begins with
Dionysius Thrax, who utilised the philological
lucubrations of Aristotle and the Alexandrian
critics for the sake of teaching Greek to the sons
of the aristocratic cotemporaries of Pompey at
Rome. Before his time the Sophists, notably
Prodikus, had made a rough classification of the
principal parts of speech for the purposes of oratorical
study; but without the contrast afforded by
another language these classifications could not
but remain confused with rhetoric and devoid of
all method and thorough-going arrangement. Indeed
it is hard to understand how any real analysis
of a language can be made unless the idea has
been suggested by the comparison of another: the
grammatical labours of the Assyrian scribes in the
time of Sardanapalus and of Chayyug and his
cotemporaries in the tenth century were due to a
necessary knowledge in the one case of Accadian,
and in the other of Arabic; and it is very
possible that the Sanskrit grammarians were
excited to their work by the native dialects, which
had been quickened into activity and raised to the
level of respectability by the spread of Buddhism.

The elaboration of a methodical grammar brought
about a more definite treatment of those speculations
into the nature of language which had
before been current. With a system of rules
246to which every one was obliged to conform, the
belief in the conventional origin of grammar
became more and more prevalent. Thus in the
noun the nominative was regarded as the typical,
fundamental case, from which the oblique cases
were so many “fallings,” casus (πτωσεῖς), so
that the whole internal relation of the inflected
noun became a declension. It had declined, fallen
off, from its primitive correct form and meaning.
In this way a systematic theory of the origin and
nature of the cases was tacitly assumed, which
fitted in well with the philosophic creed of the last
century, when society was explained by a social
contract and religion by interested artifice. It
was easy enough to furnish an answer to any
questions that might be asked regarding the
primary meaning of the relations of grammar:
the thoughts and feelings of the eighteenth century
were transferred to the first men, and ready explanations
were given in accordance with the
arbitrary philosophy of each “illuminated” savan.
This à priori mode of going to work, however, is
more easy than satisfactory. We have no more
reason for accepting the opinion of one thinker,
based upon a hasty review of certain selected
phenomena, than that of another; what we require
is the generalisation obtained from a conscientious
à posteriori induction in accordance with the slow
247critical comparative method of science. Our
generalisations, transcendental as they may be,
must be the final result of a careful survey of all
the phenomena which are at our disposal. If we
would get at a settlement of the various questions
raised by grammar, such as what is gender or
what is declension, we must set to work with our
available materials, first reducing the different
parts of grammar into their original form, so far
as is possible, and then by the help of comparison
determining what was the meaning implied by
these original forms.

One point, however, we must not overlook.
The analysis of the material is not the same as
the analysis of the mental. All that we can do is
to penetrate to the earliest marks of thought, the
most primitive utterances of society, and infer
from these outward symbols the view of the world
and the condition of the mind which so expressed
itself. It is not the symbol that we want to discover;
it is what that symbol stands for. To
mistake the symbol for the symbolised is the
error of those who would develop the inward out
of the mechanical, and find a ready explanation
for the various relations of grammar in the accidents
of phonetic decay. But between the two
there is a gulf which cannot be passed. The conception
of the dative case, for instance, was intellectual,
248not formative, in its origin. It was
evolved out of the developing thought, not out of
an accidental difference of sounds. All that the
outward symbol can do is to assist developing
thought by means of association. The symbol
recalls to the mind a certain idea, and the likeness
between two symbols will suggest a likeness
between the two ideas which they severally represent.
The Latin sestertiûm was originally the
contracted genitive plural of sestertius; but the termination
-um called up the idea of a neuter nominative
of the second declension, and hence arose
the new substantive sestertium, sestertii. But no
previously unknown idea was struck out by this;
the conception which answered to the termination
um already existed, and by the very nature of the
case necessarily existed. A rightly-conducted
investigation into the metaphysics of language
can only lead us back to the oldest symbols of
thought; the thought which lies behind these
must be reached by an application of the general
principle of the uniformity of intellectual action
at all times and in all places.

We may take, by way of illustration, the
question of gender. What, we may ask, was the
source and primary signification of the sexual relation
of nouns? It cannot have been a primitive
necessity of speech, since there are many languages
249which altogether want it; and some of these, like
the Chinese or the Accadian, belonged to races
that have taken high rank in the history of
civilisation. The theory, therefore, that would
account for gender, by assuming that our first
ancestors so far confused subject and object as to
impose the conditions of the former on the latter,
fails to satisfy all the facts. Besides, this confusion
lay not so much upon the side of the
subject as upon that of the object; the primitive
savage was overpowered by outward nature, and
immersed, as it were, in nature, not the converse.
The objective case of the personal pronoun is
older than the subjective; indeed, the subjective
element in human consciousness and speech is
only slowly and gradually evolved. Even in fetichism,
the object retains all its characteristics,
the subject merely imparting to it the vaguest
possession of power; and the worship of dead
ancestors is far from being a step in advance.
Gender could only originate, according to the
theory, in the transference of the characteristics
of the subject to the object, and this implies at
once awakened consciousness and quick imagination.
In this case, however, we should expect
to find the existence of genders rather among the
pioneers of Asiatic civilisation than among the
rude forefathers of the Slavonic tribes. The theory
250fares the usual fate of à priori attempts at explanation;
and Grimm's suggestion, that gender was
a kind of delicate insight into the distinction
between things, has no better fortune. In actual
fact, we do not find any delicate insight into
nature in the modern barbarian; and the endeavour
to explain the phenomena of language as
the results of a spontaneous growth and instinctive
apprehension is nothing more than to state
the problem in new words. All such unverified
hypotheses are shipwrecked at once as soon as we
consider that whereas there are three genders in
the Aryan group, and eight in the Nama Hottentot
dialect, Semitic and old Egyptian have but
two, while what Bleek calls the prefix-pronominal
languages of South Africa possess a large number
of genders, in one instance as many as eighteen.
This curious circumstance gives us the clue to the
origin of gender, and Bleek has accordingly put
forward a theory which is based upon an inductive
comparison of phenomena, and fully accounts for
all the known facts. 190 He believes that the nouns,
when combined with pronominal suffixes, which
were originally nothing more than explanatory
substantives, could be replaced by their corresponding
251pronouns, and these determined what we
call the gender. Thus, masculine, feminine, and
neuter were primarily only so many different pronouns,
each of which appropriated a class of substantives
that custom had amalgamated with the
same, or allied, pronominal suffixes. The prefix-pronominal
languages of Africa admitted a larger
number of combined and separate pronouns than
the Aryan group, and consequently the number of
genders possessed by them is larger than is the
case with our European dialects, Kafir having no
less than thirteen classes of nouns, and one dialect
as many as eighteen. In the Semitic verbs, a
difference of gender is plainly expressed by a
difference in the constitutive pronouns, as may be
illustrated by such examples as the Ethiopic
gabar-ca, gabar-ci, “thou art strong,” masculine
and feminine, or the Hebrew k'dhal-tem, k'dhal-ten,
“ye are killing;” and the absence of gender in
the agglutinative and isolating languages, which
do not make use of formative pronominal suffixes,
may be accounted for by the want of these derivative
elements. Indeed, the exceptions to this which
have been detected in a few of these languages by
Castrén and Schott unmistakably confirm such a
view. A feminine ending in -a occurs among the
Kottes, and another in -m among the Yenisei-Ostiaks
(among whom also fun, “daughter,”
252stands by the side of fup, “son”). Now this -a
or -m is simply am, “mother,” just as in Accadian
“daughter” was denoted by sal-tur, literally
“woman-son.” So in Tibetan the masculine termination
-pa, -po, -pho, -bo, is the word which
means “father,” and the feminine suffix -ma or
-mo is “mother.” 191 In these cases the primitive
substantives have not yet become mere pronominal
suffixes. Such, however, must have been the
origin of all these suffixes; for it must be remembered
that the pronominal theory in the
Aryan family rests on a foundation of sand; and
upon the hypothesis, as Bleek puts it forward,
two cases formed with different pronominal
elements like the nominative and accusative
would require to be assigned to two different
genders. Moreover, we should expect the Aryan
verbs, as well as the Semitic, to exhibit a distinction
of gender, and the Turanian idioms ought
to distinguish to some extent between the personal
pronouns, however genderless their substantives
may be. Man and woman, for instance, or animate
and inanimate, ought not to be represented by one
and the same personal pronoun, any more than
the first personal pronoun in Semitic by the same
253form. 192 This is all the more requisite, in so far as
these pronouns are old forgotten nouns. Bleek's
theory, therefore, must be modified: satisfactory
as it is in its main features, I should prefer to
state it in the following way: — Out of the endless
variety of words that might have been set apart to
denote the personal and demonstrative pronouns,
common use selected a certain number; each of
these, through habit, euphony, or affinity of sense
or sound, was associated with an ever-increasingly
specified class of nouns, and where the pronouns
continued different, the classes of substantives
connected with them continued different also.
Thus in Zulu the pronominal bu has ceased to
have any meaning of its own; but it is employed
to form abstracts such as u-bu-kosi, “a kingdom,”
and may be used alone like a pronoun to represent
these, just as though we were to use dom to
represent the whole class of words with which
dom (e. g., kingdom) is compounded, saying, for
instance, “the dom of England.” The classes of
nouns so created perpetually tended to become
more defined and numerous. The Aryan languages
rarely show us that uncertain wavering between
254two genders, that is, the substitution, of two different
pronouns, which we so often find in Semitic;
and where the majority of words with a common
termination were of a certain gender, all other
words with the same ending were referred to the
same gender. We see the process arrested in an
early stage of growth in such idioms as the Moxa
and Abiponian, in which a large number of
common words have an inseparable prefix, nu, na,
“my,” not unlike the Hebrew use of ינדא, or the
(Taic) Kuki numeral affix ka and prefix pa. Indeed,
these numeral suffixes can be shown to have
the same origin and intention as the pronominal
suffixes of South Africa, although the final result
of creating classes of nouns distinguished by what
we call gender has not been so perfectly attained.
Thus, in Burmese, the numeral termination changes
according to the object numbered, “two men”
being lu nhit-yauk, “two fowls” kyet nhit-gaung,
“two pagodas” tsadi nhit-chu; in Mikir, bang
is prefixed when individuals are enumerated, jon
when inferior animals, hong and pap when inanimate
objects; and in Malay, êkor, “tail,” has to be
added to the numeral whenever cattle are spoken
of, as sa-êkor kerra, instead of sa kerra, “one
monkey.” Further advanced on the road to
gender is the phenomenon that meets us in the
Tshetsh language in the Caucasus, where adjectives
255and the substantive verb change their initial
letter after certain substantives; e. g., hatxleen wa
means “the prophet is,” hatxleen ba “the prophets
are,” waso wa “the brother is,” wasar ba “the
brothers are.” The change here must be ascribed
to the attempt to substitute for class-grouping by
the help of independent suffixed words, class-grouping
by means of phonetic distinction only;
sound rather than sense has been the principle
at work. We find the same mode of procedure
in the Wolof article, the initial of which has to be
altered so as to correspond with whatever is the
first consonant of its noun. Possibly, the way to
this was led by the use as articles of various
separate substantives which began with different
letters; and when once the ear had become
accustomed to a consonantal harmony between
the article and the majority of nouns to which it
was joined, and the original independent meaning
of the words employed for it had been forgotten,
nothing would have been easier than to extend
the harmony to all instances, and establish the
general rule that the article and its noun must
commence with the same consonant. Such, at
least, was the case on a small scale in old Egyptian.
Here the sign of the feminine was the
affixed t, the universal Semitic feminine ending.
When the definite article in the singular was used
256with this, it required the form ta — plainly a repetition
of itself — in place of the masculine form
pa. This change of form is what we call gender;
whereas it was really an attempt to mark out the
substantive more definitely by guarding it, as it
were, with the same suffixed noun set at the
beginning and at the end. It was thus separated
from the rest of the sentence, and proved the yet
living consciousness of the origin and force of the
feminine termination. Gender, consequently, is
by no means engrained in the nature of things.
It is a secondary accident of speech, ornamental,
perhaps, from an æsthetic point of view, but
practically highly detrimental; and it is curious
that modern English has, in this, as in so much
else, gone back to the simple beginnings of the
sexual relations, and distinguishes gender only by
means of the corresponding pronouns. It is true
that the return is but apparent; we can never get
rid of our intervening history; and whereas gender
started from transferring the differences between
the pronouns to the substantives associated with
them, we now transfer the inherited differences of
meaning in the substantives to their representative
pronouns.

An examination of the available data of Glottology
has thus led us by the à posteriori road to
the original conception which lies at the bottom
257of gender. It is meagre enough, and very unlike
the magnificent poetic insight which à priori
theories have attributed to our remote forefathers.

Let us now see whether we can ascertain by a
similar method of procedure what was the germinal
notion that has resulted in the formation of a
plural number. Nothing seems to us more
natural, nay, more necessary, than the existence of
the plural; we might suppose that its roots go
deep down into the very beginnings of language;
and yet there are two facts which militate most
clearly and decisively against such an opinion.
The first fact is the extended employment of a
dual. All over the globe, in Aryan, in Semitic,
in Turanian, in Hottentot, in Australian, we meet
with a dual both in the substantives and in the
verbs, though the dual becomes more and more
disused with the progress of culture and the increased
use of the plural. Now, it is plain that
there must have been a very good reason for this
dual, which seems to us so utterly superfluous,
and it is also evident that there was a time when
the idea of plurality did not comprehend the idea
of duality as well; and yet, “two” is the first
plural conception to which we can attain. The
second fact to which I have alluded is the later
formations of the numbers after “two,” in so
many languages. In our own Aryan group, three,
258tres, tri, has the same root as the Latin trans,
our through, Sanskrit tar-âmi, and simply means
“going beyond.” Our earliest predecessors, accordingly,
must have exhausted their power of
definite numeration at “two,” and have regarded
all beyond that as a vague, indefinite, and therefore
unintelligible series. Observation of actually
existing savage races affords abundant illustration
of this. The aborigines of Victoria, according to
Mr Stanbridge, “have no name for numerals
above two;” 193 the Puris of South America call
“three” prica or “many;” and “the New Hollanders,”
says Mr Oldfield (of the western tribes),
“have no names for numbers beyond two.” 294
Some of these, it is true, can now count on their
fingers as high as “five,” or even higher; but the
acquisition of this power has been too recent to
have impressed itself as yet upon the language.
All this goes to show that the conception of plurality
was not part of the primary stock-in-trade
of mankind, and that the plural was preceded by
the dual. Other facts may be added in support of
this. The group of African languages which are
termed Khamitic by M. d'Abbadie want a plural
in the substantives altogether; and the Amara can
259only say fŭrŭsn ayŭhu, “I have seen horse,” leaving
it to a future question to be decided whether
horse is one or many. 195 In Accadian, again, the
pronoun bi is indifferently “he” and “they;” and
as the formative affixes are appended to the whole
series of words to which they refer, the plural sign
is attached to the adjective only when an adjective
is conjoined with a substantive, as in dimir galgal-ene,
“the great gods,” dimirri-ene being “gods”
when used alone. 296 In the case of the Khamitic
idioms, it is difficult to ascribe the want of a
plural to phonetic decay, as in our own “sheep,”
since the defect extends throughout the nouns;
much less to the influence of Semitic neighbours
who had substituted collectives, or broken plurals
as they are commonly called, for the original plural
forms. A slight advance upon this utter powerlessness
of passing beyond the singular in thought
is the formation of the plural of the personal pronouns
in the Tumali of Africa. Here the pronouns
ngi, “I,” ngo, “thou,” and ngu, “he,” which are
260distinguished from one another only by a modification
of the vowel, are changed into plurals by the
addition of the postposition da, “with.” Hence we
get ngi-n-de, “we,” ngo-n-da, “ye,” and nge-n-da,
“they.” It will be noticed that phenomena which
approach inflection are met with here, in the insertion
of the fulcrum nasal and the vocalic mutation
in the first and third persons; but there is still no
clear consciousness of anything except the singular
number: the second factor, which ought to be
coupled by the postposition, is left a mere blank,
reminding us of those savage tribes who can only
denote the relations of the verb by accompanying
a word with significant gestures. But not only do
we meet with languages which do not possess any
plural forms, we also find many others in which
the formal expression of plurality has never passed
beyond that of dualism. In the language of the
Bushmen, the plurals are throughout formed by
reduplication; and this is but one way of saying
that the doubling of a thing is the furthest
point of multiplicity to which the mind of the
speaker can attain. To repeat a word in order to
express the idea of more than one, is to identify
plurality with duality, and to imply the priority of
the latter. And nearly all our evidence makes for
the belief that the formation of the plural by this
means is one of the oldest contrivances of language.
261Thus the Accadian was still able to form
plurals in this way, as in khar-khar by the side of
kharrine, “hollows,” though he preferred to do so
by the help of the postfixes mes (“many”) and ene.
Canarese even now makes use of reduplication to
create collectives, and the Basque plural preposition
zaz shows traces of the same process; so in Malay
raja-raja is “princes,” and orang-orang, “people.” 197
The idea of the superlative, as an intensification and
increase of the visible individual qualities, cannot be
separated from that of plurality; and superlatives
are made by reduplication from the Mandingo ding-ding,
“a very little child,” to the Accadian galgal,
“very great.” It does not appear, however, that
the specific conception of duality was the one most
prominent in this primitive expedient of speech.
When we consider how often reduplication is used
simply to intensify the imitation of natural sounds,
262and to denote their continuousness, as in the
Dayak kaká-kaka, “to go on laughing loud,” or
the Tamil muru-muru, “to murmur,” or to express
the length and continuity of an action, as in the
reduplicated Aryan perfect, we are inclined to
believe that the contrivance of reduplication was
adopted by language before it had arrived at a
clear idea of duality, and while it was still struggling
to pass from the single individual to a more
general concept. The most obvious means of
expressing this vague endeavour was the repetition
of sounds; and when once thought had thus
made itself objective in articulate speech, it was
comparatively easy to acquire a clear and distinct
conception of duality and separation. Before this,
all beyond one would have presented itself as a
misty and indefinite repetition of one. In this
case reduplicated plurals would once have represented,
not merely an indistinct amplification of
the individual, but a definite idea of two individuals,
and the further extension of this to denote
the plural only shows the poverty of invention
among those races who have retained the primitive
dual form to express the plural.

In some of the North American languages we
may actually see the process going on, whereby the
conception of duality, when once clearly defined,
extended itself to that of plurality. In Cherokee,
263the dual of the first person is divided into two,
the first of which is used when one of two persons
speaks to the other; the second, when the one speaks
of the other to a third. Thus inaluiha is “we two
(i. e., thou and I) are tying it;” awstaluiha, “we
two (i. e., he and I) are tying it.” Here the idea
of the limitation of the dual on the side of
plurality has been distinctly attained. The process
is to be observed still more plainly in the Papuan
dialects, in which the personal pronouns possess
not only a trinal form, but also exclusive and
inclusive forms. In Annatom, for instance, ainyak
is “I;” akaijan, “you two + I;” ajumrau, “you
two — I;” akataij, “you three + I,” aijumtaij,
“you three — I;” akaija, “you + I;” aijama,
“you — I.” So in Mallicollo, inau is “I;” khai-im
“you,” and na-ü, “he,” while na-mühl is “we
two, exclusive of others;” drivan, “we two,
inclusive of others;” kha-mühl, “you two;” na-tarsi,
“you three;” dra-tin, “we three;” and the
specification of number actually rises as high as
four, na-tavatz being “you four,” and dra-tovatz,
“we four.” It is difficult to understand how a
people could have reached the point of setting apart
a special form to denote the number four, and
should yet have not made what seems so short
a step in advance, and attained the notion of plurality.
The abstracting and generalising faculty
264was wanting, and the speaker was still unable to
get beyond the individual object of sense. It is
evident, however, that tovatz or tavatz must be
merely the numeral “four,” which is tacked on to
the singular personal pronoun, just as in the Taic
languages a plural numeral is attached to a singular
noun; the Burmese lu nhit-yauk, “two men,” for
example, meaning literally “man two.” Perhaps
we may compare our own “ten foot,” “ten stone,”
like the Hebrew use of the tens from 20 to 90
with the singular, as in 'esrim 'ir, “twenty cities;”
or the employment of collectives, which may be
regarded, from one point of view, as a survival of
the inability of primitive man to conceive the
plural. The collective sums up under a single
head the idea of plurality, and thus embodies the
last result of generalisation and classification;
whereas the primeval noun, like the primeval sentence,
was unable to reach the simplest classification,
and so was obliged to enumerate each
separate individual, although, owing to this very
incapacity to generalise, the universal lay implicit
in the noun, waiting to be developed out of it when
the time came. We cannot correctly call it a
singular, because there was no plural: no singular
existed until the idea of a dual was struck
out.

We may even call in the aid of à priori arguments,
265whatever these may be worth, in support of
the view that the dual is older than the plural. So
long as men lived in the primeval beehive community,
there was no need of any clear expression of
multiplicity. As the individual, however, emerged
from this early state, he would arrive at more definite
ideas of number; one necessarily implies two,
and the immediate wants of a savage life would often
require the employment of language. But these
wants were circumscribed, and the primitive barbarian,
like modern savages, would have been
extremely chary in his use of words. His simple
necessities would easily be satisfied by a single
neighbour; and time would elapse before the isolated
nomad came to mix freely with a large circle
of human beings. Primarily, therefore, his requests
would be addressed to one other person only,
and the dual accordingly would suffice for all his
wants. Consequently we are not astonished at
finding that an analysis of the pronouns teaches
us that the Aryan plural asma is compounded of
ma + sma, “I and he,” and not “I and they,”
tusma (whence Sanskrit yushmâ, with the insertion
of the semi-vowel and the subsequent loss of the
dental) being similarly “thou and he.” In this
way, moreover, we can alone account for the
existence and persistency of a dual, which seems
so superfluous by the side of a plural: with the
266latter already in use, it is hard to understand the
elaboration of the former.

The priority of the dual, however, is contrary to
the opinion which makes the dual in Aryan and
Semitic merely a lengthened form of the plural.
The Aryan plural is formed by a postfixed s, which
has been compared with the preposition sam, sahâ,
and the s of the singular nominative and genitive,
as if there were any compatibility between these,
or no difference between a preposition and a post-position.
Now, it is no doubt tempting to regard
the dual as an amplification of the plural forms;
but a few words will show clearly how improbable
this really is. In the first place, the assumption
of an uniform plural in s in the parent speech
cannot be sustained by the side of the second
declension in Greek and Latin, or of neuter bases
in i and u in Sanskrit, where the nominatives do
not exhibit any vestiges of an original sibilant.
Then, secondly, however easy it may be to get the
dual sâs out of the plural sas, it is absolutely
impossible to get at once sâs and aus out of ams,
the old accusative plural, and sâms, the genitive
plural. Moreover, we may ask what warrant we
have for postulating the change of m into v and u
in the parent Aryan? So far as our data go, it
is unheard of. And if we grant the possibility of
a transformation of sâms into aus, how comes
267swâs, the conjectural pattern-form of the plural
locative, also to become aus? This, indeed, is to
presuppose the desperate expedient of a metathesis,
which is contra-indicated by the usual loss of the
final syllable in the Sanskrit -su. But the last
difficulty is the greatest of all. The dative and
ablative plural in -bhyams may readily become
bhyâms in the dual; but unfortunately the instrumental
dual has exactly the same form, while the
instrumental plural, though derived from the same
postposition, bhi, is not bhyams, but bhis. The
most stout-hearted philologist will find it hard to
extract the same phonetic result out of a lengthening
of bhyams and bhis. The fact, however,
suggests another explanation. It is undeniable
that bhyâms and bhyams, whence come the Sanskrit
-bhyas, the Greek -σι, the Latin -bus, the Gothic -m,
and the old Norse -um, are closely connected with
one another; but both, as has been said in a
former chapter, are derived from the post-preposition
bhi, 198 and must have been applied to their
present purpose during the period which falls within
the province of Glottology; consequently they do
not belong to the original flection of the Aryan
noun. Bhis is also taken from the same independent
root, and it is very probable that both bhyams
and bhis existed as separate plurals, the first as
268an accusative, and the second as a locative (for
bhins), before they were attached to other vocables.
We are here dealing with an instance that is altogether
different from that of flection proper, where
the inflections cannot be separated from the noun
in which they inhere, and show no signs of having
ever been independent roots. Now if bhyâms
forms the dative, ablative, and instrumental in the
dual, while bhyams performs this office only for the
dative and the ablative in the plural, the instrumental
being denoted by bhis, the simplest mode
of explaining the relation of the two is to assume
the prior existence of the dual, the plural not
coming into general use before a further differentiation
of cases had taken place. When the plural of
these cases first became fixed, the instrumental had
already been separated off from the dative and the
ablative. Why the vowel of the dual should be
longer than that of the plural may perhaps be learned
from a consideration of the Semitic languages. In
these, while the Hebrew plural was -īm (from -am),
the Aramaic -īn, and the Arabic -ūna, the dual in
these dialects was respectively -áim, -ain, and -ánί or
-aini. So, too, in Assyrian, the dual ended in -ā,
the usual masculine plural being in -i. Now a
comparison of the Semitic languages leads us to the
conclusion that the plural primarily terminated in
-āmū, so that the original dual was probably -a'amu,
269which expressed the reduplication of the object by
the long-continued repetition of the pure primary
vowel. A close analogy to this may be found in the
idiom of the Aponegricans, in which “six” is ita-wuna,
and “seven” itawu-ú-una: the same principle
is at work in the extension of ouatou, “a stream,”
among the Botocudos of Brazil, into ouatou-ou-ou-ou,
“ocean,” or the Madagascar lengthening of
ra-a-atchi, “very bad,” from ratchi “bad.” If
the repetition of the primary vowel in Semitic,
therefore, with the mimmation sounded after it
(as in the singular), was intended to represent the
double character of the object, the dual would have
been formed upon the singular, not upon the plural,
and the latter would rather be a contraction of
it, the vowel being contracted in so far as the idea
expressed by the plural was less definite than that
expressed by the dual. The final case-ending -u
would have been copied from the singular.

From the numbers we naturally pass on to the
cases. These, as their name implies, are regarded
as so many fallings-off from the casus rectus,
or nominative, which is held to be the typical
form of the noun. This view, however, which is
really based on the logical analysis of a developed
grammar, is not borne out by scientific investigation.
The “naming” case of the noun, whose
title to the name of case was itself disputed, seems
270after all to be a later addition to nominal declension.
Everything seems to point to the accusative
or objective case as the most primitive form
of the noun. This is clearly patent in Semitic,
where the so-called case-ending in -a has been
retained in Ethiopic, Arabic, Assyrian, and apparently
Hebrew, to mark the accusative, the later
modifications of this original sound having been
appropriated to create the nominative in -u and
the genitive in -i. So, again, in Aryan the objective
mâ, “me,” is still found as accusative in
Sanskrit, while its priority is shown, not only by
the verbal termination in -mi, but yet more by
the compounded form of the nominative Sanskrit
aham, Greek ἐγὼν, Latin ego, Gothic ik. Whether
or not this is made up of ma, which has first become
va (as in the dual and plural of Sanskrit
and Teutonic), and then been dropped altogether,
and ga, an emphatic enclitic which has given birth
to the Vedic gha and the Greek γε, at all events
ἐγὼν is a less simple and ancient form than με.
It has been aptly remarked, that this is only in
accordance with the ordinary facts of infantile life.
The child says “Charlie does this or that,” before
he learns to say “I do this or that.” The existence
of neuters, the nominatives of which end in
-m, points in the same direction. Here the idea of
life, and therefore of subjectivity, is put out of sight,
271and consequently the conception of objectivity has
been so fixed in them, that when other classes of
things came to be conceived as capable of originating
actions, and were therefore assigned a particular
flection when regarded in this way, the neuters
were relegated to a class by themselves, and
preserved the old common termination for what
now became divided into nominative and accusative.
The outward form kept up a recollection of
that primitive state of things in which man still
regarded himself, and all about him, as objects,
and had not yet realised that he was a subject,
and the originator of action, still less had projected
this power into the objects about him.
The agglutinative languages made no distinction
between the nominative and accusative, thus
reflecting, as in so much else, the early condition
of human intelligence and speech.

Next to these cases, the most important part of
the noun-declension is the genitive. But the
relation which we express by this must originally
have been but imperfectly, if at all, comprehended,
if we are to judge from the grammatical phenomena
of the agglutinative tongues. Thus in
Accadian the relation of genitive and governing
noun was primarily denoted merely by placing
the former after the latter, as is still the case
with Taic and Malay; and it was only gradually
272that this simple method came to be supplanted by
the suffixing of words like lal, “filling,” and ga,
“making,” to the second noun. Here, then, the
relation would seem to be nothing more than what
we term “apposition,” that is, where two individual
notions are placed side by side without any
further effort being made by the mind to determine
their exact relations beyond the mere fact
that one precedes the other, and is therefore
thought of first. Hence we may say that there
was a time when the genitive, as such, did not
exist, and we have to discover, as far as is possible,
how it came into being. Now we are all
well acquainted with the distinction between what
is called the objective genitive, where the governed
word is the object of the other (as in amor Socratis,
“love felt for Socrates”), and the subjective genitive,
where the reverse takes place (as in Socratis
amor, “love felt by Socrates”). The distinction
corresponds to the difference made in formal logic
between predication and inhesion in a proposition,
the attribute being included in the subject in the
one, and including the subject in the other. The
genitive relation can be looked at under either one
of these two aspects, and consequently we ought
not to expect to find the grammatical relics of all
languages pointing to one and the same process.
This race preferred to conceive the relation, when
273it had once arrived at it, under the one point of
view, that race under another. The meaning of
the relation itself, however, was not that of simple
dependency, which it has since grown into. The
Semite centred his attention upon the governed
word, in agreement with that synthetising tendency
which has displayed itself in his language,
his literature, and his religion. The governing
noun was placed first, and its accent and importance
transferred to the following genitive, so that
the whole became a kind of compound pronounced
in one breath, in which the latter part alone had
prominence assigned to it. The so-called genitive
termination in i, which the second substantive
takes in Assyrian, is but a modification of the
accusatival -a, and consequently goes back to a
time when the nominative did not exist. The
periphrastic genitive, which placed the relative
(or rather originally the demonstrative) pronoun
between the two nouns, analysing the genitive
relation into “love that (is) Sokrates,” and so
equalising the two ideas, must be referred to a
later period. The Aryan procedure was the exact
converse of the Semitic, and would suffice of itself
to demonstrate the separate origin of the two
groups of languages. Here the mind fixed all its
attention upon the governing noun, suitably to the
genius of a race which was eminently practical
274and by its close observation of objects has been
the originator of inductive science. It was the
governed noun the dependency of which was
marked out by suffixes, and which naturally came
first in pronunciation, thus directing the attention
to the more important governing word, which was
last heard. The mind was turned towards the
object, not towards the source or end of that
object. These, on the contrary, were conceived as
so many attributes, which accidentally adhered to
the principal object of thought. It is the same in
the pronominal-prefix idioms of South Africa.
The Bâ-ntu genitive agrees with the gender of the
governing noun, just as much as the genitival
δημό-σιο-ς of the Greek must agree with its substantive; 199
thus in Zulu, i-SI-tya S-O-M-fazi, “the
dish of the woman.” The last instance, taken in
conjunction with what has been said above upon
the origin of gender, will throw much light upon
the primitive signification of the genitive relation.
The same pronominal word which has been attached
to one substantive is attached to another when
the idea expressed by the latter is sought to be
brought into connection with the idea expressed
275by the first. If we assume that the primary meaning
of si was “mass,” the words i-SI-tya S-O-M-fazi
would properly be read “mass-dish mass-woman.”
It is but a new application of the old
law of the syllogism in logic, or of the principle
which Mr H. Spencer has shown to be the ground
of all science. Two things are brought into connection
and equivalence one with the other by means
of a third. In the present case, two ideas were
first set over against one another, and expressed
in language in such a way that one of them came
to be always associated with the other, and with
the ideas cognate with the latter, until it was
reduced at last to a mere formative, constituting a
class; and then by the help of this pronominal
formative other ideas, not cognate with the idea
originally set over against the decayed prefix, were
united with it in thought. In this manner the
genitive would have grown out of apposition.
Equipollent conceptions could be placed side by
side in apposition, and one of these, after being
crystallised into a grammatical form, became the
medium of combining new conceptions with the
conception with which it was united. This, however,
could only be the case where the objective
genitive was the type of the relation. Languages
like the Semitic, in which the subjective-genitive
was the type, never rose beyond an apposition
276wherein the first factor was subordinated to the
second, and consequently never possessed a true
genitive, any more than the Malay and the Taic
languages generally. The insertion of the relative
pronoun between the two factors, which may be
made in Chinese by tchi, a word originally signifying
“a place,” is nothing else than an analysis
of the apposition. The agglutinative plan of affixing
a word of independent meaning to the governed
noun is equally little a genitive; it is really a
verbal clause; and the Accadian enu Huru-lal may
just as well be translated “the lord fills Ur,” as
“lord of Ur” (“Ur-filling”).

Before closing our list of illustrations of what is
meant by the Metaphysics of Language, it would
be well to take an example from the verbs. I
have already tried to point out in a former lecture
how a comparative study of languages leads us to
the conclusion that the aorist is the oldest tense.
Let us now see what we can learn about the person-endings,
the chief characteristics of the verb
so far as form is concerned. In Chinese, position
alone decides whether a word is used as a verb, a
substantive, an adjective, an adverb, or a preposition.
Place ngó, “I,” before a root, and it becomes
the first person of a verb, just like “I ride”
in English. The form of the language has scarcely
advanced beyond the rudimentary stage in which
277the distinctions of the several parts of speech were
all unknown, and lay undeveloped within the
embryo of a single monosyllable. The agglutinative
languages show further progress. Accadian
can not only say mu-ac, “I made,” and mu-nin-ac,
“I made it,” like the Chinese ngò wêï and ngò
wêï tschi, but has proceeded to create a present by
extending the last syllable of the radical, and so
appropriating to it a special verbal form, just as in
Tibetan we get nga jyed-do, “I do,” from jyed, “to
do.” Immense is the advance from this early
stage to such broken-down forms as the Basque
duzu, “thou hast him” (compounded of d, “him,”
au, “have,” and zu, “thou ”), or the Ostiak conjugation,
where the three persons of the singular of
the first two tenses of the indicative respectively are
madâdm, madân, madâ, and madâu, madâr, madâda.
In all cases, however, we find that the
forms resolve themselves into a combination of the
root with the personal pronouns, these being
sometimes affixed and sometimes prefixed. In
Accadian, as in Basque, both processes could take
place; but as a general rule the Turanian idioms
of Asia have remained true to their instinct of
postfixing the determinative words. It is the same
in old Egyptian, and with the Aryan verb, though
a difficulty meets us here. Every one can see that
ad-mi, at-si, “I eat,” “thou eatest,” go back to
278the two first personal pronouns, in spite of the
change of the dental of the second person into a
sibilant, and the dual and plural forms -vas, -thas,
and -mas, -tha make this indubitable. But the
third person is not so easy to explain, and Bleek
has even ventured to derive it from a conjectural
ti = “do,” which has made the perfect of the
Teutonic languages. The singular -ti might be
discovered in the demonstrative, which has helped
in the declension of the Sanskrit third personal
pronoun, but the plural, -nti, which cannot be
separated from it, still remains unaccounted for.
The nasal cannot have been a mere phonetic insertion,
nor is it likely that its derivation is to be
sought in an assumed demonstrative pronoun an.
Whatever may be the difficulties, however, connected
with the third person, the first and second
persons of the verb are unmistakably to be traced
back to the original objective forms of the personal
pronouns. But this implies a time when such a
combination did not exist, a time when the personal
pronouns were not yet fossilised out of their
earlier general significations, and when a verbal
force must have been given to the root in a different
manner. It is noticeable that in Accadian
enu-mu meant at once “my lord” and “I am
lord,” and this vagueness of meaning implies a
very faint realisation of the distinction between the
279two principal parts of speech; while, on the other
hand, the Japanese personal pronouns, true to their
substantival origin, may be used to denote all three
persons alike. Now here, as elsewhere, the dialects
of savage tribes let us into the secrets of early
language, and we find that the Grebo of West
Africa can distinguish between “I” and “thou,”
“we” and “you,” solely by the intonation of the
voice, mâ di being equally “I eat” and “thou
eatest,” a di, “you” and “we eat.” Nay, more
than this; according to the Rev. J. L. Wilson,
even these pronouns are but rarely employed in
conversation, it being left to gesture to determine
in what person a verb is to be taken; ni ne, for
instance, being “I do it” or “you do it,” according
to the significant gestures of the speaker, just
as in Mpongwe tŏnda means “to love,” tōnda
“not to love.” 1100 Spix and Martius describe a
similar condition of speech among certain Brazilian
tribes, with whom the projection of the mouth in
the direction intended serves to make the words
“wood-go” signify “I will go into the wood.”
Such a state of things is indeed hard to realise,
with no pronouns and no verbs; and yet out of it
grew first the conception of action in relation to
the person, and then in relation to time. Men
were slow in arriving at a distinction between
280one's-self and another; the three personal pronouns
could not have come into existence until
after the genesis of a plural, and the idea of a
subject-pronoun was evolved last of all. The verb
would seem to have been at first not unlike the
genitive. Primarily the rough-hewn chaotic word,
with its undeveloped potentiality of meaning, was
accompanied by visible action in order to impart to
it the signification of agency or intention; afterwards
a substantive was brought into juxtaposition
to it, the sense of the compound being settled by
outward action or by the circumstances of the
case; and finally, these substantives, worn down
to personal pronouns, became differentiated, and,
joined in apposition with the roots, formed a kind
of compound in which something — eating, doing,
or the like — was attributed to the pronoun. 1101 As
in the instance quoted above from the Accadian,
or as in so many Turanian languages, Magyar, for
example (where it is only by using different pronominal
words that kés-em, “my knife,” can be
distinguished from vár-ok, “I wait,” the aorist
vár-t-am being actually identical), the verbal form
was simply a genitive, and has to be explained like
281all other genitives. Were we to represent it symbolically,
we might say that “expecting = me”
was the source both of “my expectation” and of
“I expect.” The position of the pronoun in
Aryan is alone to be noticed: it follows instead of
preceding its governing noun; and this reversal of
the usual order of words implies not only that the
personal pronouns had been fixed before the verbal
forms became crystallised, but also that the feeling
that these pronouns were different from all other
substantives, and that the power of the individual
over action was omnipotent, was from the very
earliest times ever present to the Aryan mind. It
still required one step further, however, to ascend
from these merely personal relations to that conception
of time which with us lies at the very
foundation of the verb. It is a conception that is
still unknown to many races of men, and which is
conspicuous for its absence among the polysynthetic
languages of North America. The New
Caledonian, with whom “yesterday” and “to- morrow”
are unknown terms, or the member of
the beehive communities of the old world, had no
need, and no occasion, to mark the lapse of time
in their monotonous and vegetable existence. The
category of space historically precedes the category
of time.

Further illustrations of the Metaphysics of Language
282are, I think, unnecessary. Enough has
been said to show what is meant by the phrase,
and the way in which this part of Glottology can
be worked out. A comparative analysis of words
leads us to the earliest linguistic contrivances for
expressing the relations of grammar. They are
but the fossilised embodiment of the thought which
they clothed; and we are thus enabled to penetrate
to the germ and starting-point of those
conceptions which are summed up in an ordinary
grammar. They are the mental forms which we
finally reach, and which have developed into all
the elaborate grammatical machinery of modern
speech. We get back, as it were, into that very
thought in its most original form which has been
reflected in spoken language. We enter the world
of ideas, and, like the physicist with his doctrine
of force, find ourselves dealing with metaphysical
facts.283

Chapter VIII.
Comparative mythology and the science of religion.

Language, we have said, is the mirror of society,
because it is the embodiment of thought. Every
word has a history, and that history is really a history
of the mind. The two correlatives cannot be
separated from one another: thought is but the
internal, language the external. Form and content,
creator and created — these are other ways of
expressing the same thing: the statue does not
represent more truly the artistic imagination of its
sculptor than does the word the mind that shaped
it. And just as the statue will react on the artist,
and produce, as in Egypt, a conventional conception
of beauty and proportion, so in a greater
degree will the plastic word react on the mind of
man. The two sides of the prism, the inward and
the outward, act and react one upon the other; and
where the sense of objectivity is strong, or the absolute
nothingness of the mere empty husk of the
284word is forgotten, words are likely to become our
masters, and to dictate to us the meaning of things.
If the Greek with his autonomous individualism could
speak of the ἀξίωσις λόγου — the appreciation which
he set upon the utterances of his own lips — the
law-loving, abstraction-worshipping Roman, on the
other hand, knew only of the vis verbi, a fitting
echo of military martinetism. Language is a
natural growth as well as an artificial production.
It has developed along with the awakening consciousness,
and much of it will be at best but
semi-conscious. At the beginning there was no
clear distinction between the parts of speech or
the objects which were denoted; all lay chaotic
and undeveloped in each embryonic combination
of sounds, and these inevitably called up erroneous
ideas, and laid the foundations of a fetichism
which confounded together the agent and the
patient. But more than this; language, like
the rocks, is strewn with the fossilised wrecks of
former conditions of society. Words which were
once pregnant with meaning may either put on
new significations in consequence of social changes;
or long use and acquaintance may deprive them
of their sense, so that the sole meaning they
possess is their mere sound; or, again, their
original force may be forgotten, and they may
survive as proper names or in connection with
285obsolete ceremonies; or, lastly, they may be confused
with other better-known words, and so bring
about a confusion of ideas. Who now connects
the same conceptions with, such terms as “democracy”
or “church” as they conveyed to our
ancestors? “Shall” and “will” have become
auxiliaries, unmeaning by themselves; “Jove”
and “Yule” no longer remind us of the bright
vault of heaven or the burning wheel (old Norse
hjul) that symbolised the circle of the year, while
“beefeater” and “Brasenose College” show little
trace of the waiter at the side-table (buffetier) or
of the brewing-house (Brasen-huis) from which
they sprung. The words in which one period of
society struggles to express its knowledge and
meaning may become the misunderstood shams of
a later generation, and the explanation of them
which is demanded by the mind serves only to
perpetuate the delusion and stereotype an imaginary
world. Indeed, the first act of the young
consciousness is to ask what is the reason of that
which it sees about it? The formation of a language
itself implies a desire to know objects by
naming them, and so distinguishing them one from
the other. Every name that is given is the summing-up
of all attainable knowledge concerning a
thing; it contains within itself the answer which
man attempts to make to that ever-recurring question
286“why?” and all the knowledge and experience
which he can bring to bear upon it. But the
knowledge and the answer of the first men must
have been very different from that of a more cultivated
era of humanity. The Athenian of the
age of Periklês would view the world with eyes
very unlike those with which the primitive Aryan
gazed upon it. The old name would not express
the new meaning; and if it had not expanded
with the growing knowledge of the speakers, it
would of necessity cramp and confine the signification
within the limits originally assigned to it,
and cease to reflect the living knowledge of the
day, and to be anything more than an antiquated
symbol. Words have a life, because the society
which produces them has a life; and just as the
old forms of society become dead and misleading,
so also do the words which shadow them forth.
They no longer answer truly to objects, and therefore
objects must be made to answer to them; and
thus a dark cloudland is built up upon these worn-out
husks, hiding nature and reality from the
mind and the belief.

Now this is mythology exactly. Its creations
move, like the ghosts of Homer, in an unreal fairy-land,
and their sole basis is the names which are
given to them; for these names are the heirlooms
of a traditional past — the heritage which has come
287down from the giants of old time; this is their
only title to existence and respect. The traditional
past, therefore, which has given them their existence,
must furnish the key which shall unlock
them. We must track the names back historically,
until we reach the age when they were living and
full of significance. Mythology is founded upon
words, and the history of words, therefore, must
explain it. 1102 But we must not forget that, after
288all, words will only explain the external side of
mythology. It is true that this is its chief and
most important side; but, without an inward and
sustaining spirit, mythology could not have lasted
so long and so persistently as it has done, and
have blinded the eyes to its manifold absurdities.
There must have been an element in it which
appealed to the heart of man, and preserved it
from being relegated to the nursery like the fairy
tales which yet claim the same origin as the
gorgeous mythology of the Greek poets. This
element was the religious instinct. Behind the
outward veil of the myth was enshrined the belief
in God and the soul, more and more concealed
and over-encrusted, it may be, in the course of
generations; but still there it abided almost unconsciously,
and kept the old mythology from
premature death. It is clear that we are here
dealing with a similar case to that which we
described in the last chapter. As we get at the
original conceptions which underlie the several
relations of grammar by a comparison of the forms
which denote them, so in mythology we must discover
289the spirit that has given it birth by an
inductive comparison of the various forms with
which it has clothed itself. These are words and
phrases, and consequently Comparative Mythology
is but a branch of the Science of Language.

But the religious idea can make use of other
means of expression besides mythology. What we
call a religion differs from mythology in the same
way that a civilised state differs from a savage
tribe. The one is organised and artificial, the other
is spontaneous and natural. There is no longer a
sort of dim half-consciousness of spiritual being;
the individual has awakened to a consciousness of
himself and his relations to others. In a beehive
community morality is impossible, much less a
worship of one God; it is only when the conception
of the individual has been reached that the
idea of responsibility begins, and with it both
morality and the endeavour to obtain a personal
salvation. The barbarian knows nothing of all
this; sin and moral impurity are words which he
would not understand; his only idea of happiness
consists in abundance of food; the only evils
from which he prays to be delivered are material
discomforts. A religion must be organised and
individual; and this implies tradition and literature
on the one hand, and on the other hand a
hierarchic aristocracy, in so far as individualism
290presupposes distinction and superiority. To call
fetichism a religion, therefore, is a misuse of
terms. Where every man is his own priest, there
is no system in which one man knows the will of
the gods better than another. Rome had no
religion until the days of the Empire, for its
organised cultus was political; and religion in
Greece was confined to Delphi or the Orphic
hierophants. The individual character of a religion
is universally recognised; where history
can present us with no founder like Buddha or
Confucius or Christ, later legends delight to
trace back its ceremonies and organisation to
some single Numa Pompilius.

But the founder must have materials to work
upon. There must be the religious instinct, without
which all religion is impossible; there must
be a hallowed stock of traditional beliefs and
rites; and above all, there must be a willingness
on the part of the people to accept the system that
is formed out of them. The founder of a creed
generally comes forward as the reformer of a past
unorganised cultus, and if he would succeed, he
must strike a chord in harmony with the wants
and wishes of his age. Buddha preached a gospel
of freedom from the intolerable yoke of castes and
Brahmanical despotism; Mohammed broke up the
aristocracy of Arab traders, and proclaimed equality
291before one God and one Prophet to the sons of the
desert; and Joseph Smith flattered the sensuousness
of American enthusiasts and the millenarian
dreams of uneducated Protestantism. Mythology
necessarily precedes a religion. It may be extirpated
by its successor, or it may be taken up and
absorbed into it, or it may linger on side by side
with the new creed, sometimes in alliance, sometimes
in antagonism. It never follows it, however,
for the myths which so often gather round the
person of the real or imaginary legislator are borrowed
from older legends, and do but find a new
hero to whom to attach the venerated stories of
the timeworn folklore. The saints of Christendom
have taken the place of the gods and demigods of
pagan antiquity, and the deities of the Veda
became the evil spirits of Zoroastrianism. Tritâ,
the Hindu power of night, and Ahi, the serpent of
darkness, change, in the Avesta, into the human
Thraêtaona, the son of the first man, and Azhi
dahâka, “the biting snake,” which he destroys,
and the transformation is completed when the
religion can no longer assimilate the old mythology
even thus far, and Thraêtaona and the serpent
become the Feridun and Zohak of Firdusi — the
Kyrus and Astyages of the Greeks. Assimilation
of pre-existing beliefs must necessarily be the work
of a new religion; the beliefs will be modified and
292arranged; but if the religion is to make its way,
it cannot afford to ignore the current superstitions
and practices of the country. Indeed, these will
colour it the further it spreads and the more it
appeals to the uneducated portion of society; and
it is no strange thing for a religion which begins
with a protest against the popular idolatry to end
by becoming inextricably mixed up with it. Even
if this does not happen, however, it is plain that,
in order to understand a religion rightly, we must
know the meaning of the mythological elements
which it incorporates and rests upon, and of the
terms which are its own watchwords. These
change with the change of knowledge and circumstances
and generations; and a church will
often be found fighting over the signification of a
word which originally bore an import quite other
than any dreamed of by the combatants. The
interminable wranglings and divisions that have
been carried on in modern Europe over the questions
of the Eucharist and the ministerial orders
would have been unintelligible to the first Christians.
The battle is one of words, but the insertion
of an iota was once sufficient to deluge Alexandria
with blood. Here, then, Glottology, with its calm,
scientific dispassionateness and its rules of sound
comparison, is needed in order that we may comprehend
the origin and growth of religious ideas,
293and of the dogmas which endeavour to express
them. In so far as the science of religions consists
in comparing words with words, dogmas with
dogmas, and in tracing the development of the
one out of the other, in so far it is, like mythology,
a branch of the science of language, and this,
too, apart from its embodiment of mythological
elements, which, as we have seen, demand the
key of Glottology.

But there is another reason why the comparative
study of religions calls for the glottologist. The
oldest and the most interesting are locked up in
the recesses of dead languages, and it is only the
scientific method which can accurately explain
much that is most important in the language of
the Rig-Veda, and still more of the Zend-Avesta.
The traditional renderings of Sanskrit pundits are
often grotesque, often the result of modern misconception;
and some of the most valuable disclosures
of the old Hindu hymns, which have
helped to explain the problem of mythology, would
never have been made without the application of
glottological laws. Even the Old Testament
cannot afford to dispense with this assistance:
whether or not Samson is the Melkarth of Tyre
and the Herakles of Greece can only be decided
by Comparative Philology. The same holds good
of the Science of Religions if we regard it from
294another point of view. Every system of religion
consists of a certain number of doctrines which
circle round some central one, and the meaning of
this is all-important if we would understand the
system. But doctrines alter, although the words
in which they are formulated do not; and to discover
their original import is to discover the
original sense attached to the words. A good
example of this is the Nirvana, the point about
which the whole system of Buddhism revolves;
and until we have accurately settled the primary
signification of this word, and the historical modifications
which it has undergone in various ages
and among various races, we shall never properly
know what Buddhism is. Religion is the most
spiritual, and therefore the deepest and most enduring,
expression of society; and if the history
of society is to be sought in language, yet more
emphatically must the history of religion be.

Before, however, we can venture to compare
religions together, we must establish the scientific
study of mythology upon a firm and satisfactory
foundation. As a branch of Glottology, it must
be investigated upon the same principles and in
the same way. We must never forget that it is
a dependent science, and is, therefore, not to be
treated as though the higher science did not exist.
To draw conclusions from a comparison of myths,
295which are not supported by etymological evidence,
is altogether unwarrantable. If Comparative Philology
can show that Paris is the Pans of the
Veda, the robbers of the bright cow-clouds of the
dawn; that Helen is Saramâ, the dawn goddess;
and that Akhilles, who dies at the western gate of
Troy, is Aharyus, the sun, from the Sanskrit ahar,
“day,” then the burden of the Iliad may well be
the old fight between the night and the morning,
the old story of the victory and death of the solar
hero around the walls and battlements of the sky. 1103
296But to resolve Orestes into the sun and Semiramis
into the morning is to step beyond the limits
allowed to us, and to assert what cannot be proved.
In comparing our myths we must never lose sight
of the etymological part of the subject, since it is
this which gives security to our conclusions.
Unless the features of a myth unmistakeably resemble
those of another, more especially in the
smaller details, we should be very cautious in setting
it by the side of another, where the proper
names are not transparent. There is no doubt
as to the meaning of the names of Phœbus and
Hyperion, and we may therefore class them with
other solar myths without hesitation, even supposing
that the outlines of the stories told about them
were vague and general; but to discover the sun
on the horizon of the sea in the frog-prince of
the fairy-tale is to transgress the boundaries of
scientific evidence, and incur the charge of riding
a hobby too hard. Besides the care which must
thus be taken to make language the ultimate
ground of our comparisons, we must be on our
guard against that hankering after unity which
has been so fatal to glottological progress. The
general laws of Comparative Mythology, like the
297general laws of Comparative Philology, must be
obtained by the widest possible induction of instances;
we must collect our myths from every
race and climate under the sun, and we shall often
find that some low and despised tribe of savages
can furnish us with a clue to the laws we are
seeking. Mythology, like language, is a reflection
of the human mind; it belongs more especially
to what we may call the natural era of mankind; 1104
and since the framework of the mind, and
the circumstances which surround the life of the
savage, are much the same everywhere, we shall
expect to meet with a common similarity and
obedience to general laws in the myths of all nations.
But we must not go further than this, and,
in disregard of all linguistic testimony, derive the
stories of Aryans, and Finns, and Kafirs, which,
resemble one another, from one and the same
298source. Where language demonstrates identity
of origin, there will there be identity of origin
among the myths, but not otherwise. To imagine
that the coincidence of legends among two races
unallied in language means anything more than
the common uniformity of intellectual action in
the mythopœic age, is to repeat the mistake of
bygone writers, who believed that the story of a
flood among different peoples bore witness to the
Biblical deluge. With them the belief was excusable,
for they had been taught the existence of a
single primeval language, and the transformation
of the heroes of Genesis into the personages of
heathen mythology. But where there is no disposition
to see Noah in Kronos, and his three sons
in Zeus, Poseidon, and Aides, the indiscriminate
lumping of myths together, without any heed to
the requirements of Glottology, is altogether indefensible. 1105299

There is yet another fault of which we must
beware. Mythology has a setting in geography
and history. Myths move in an unreal world of
their own, a dead reflection of this world, distorted
by the childlike ignorance of primitive man.
Hence there is a mythical geography, a mythical
history, and a mythical philosophy. When the
original physical reference of the myth had faded
away from the memory, it was necessary for the
story-teller to hang his tale upon some fact, or
person, or place. When this was once found, and
the needful local colouring imparted, the myth
continued to circle around it, and to attract fresh
elements until a change of conditions transferred
the circle of myths so formed to a new local centre.
To look for any traces of history here is obviously
out of the question. Even granting that the
mythical element has been grafted upon a real
person and a real fact, the latter were but the
framework, which was wholly swallowed up in the
animating mass of mythic matter. Not history
but folklore was what was wanted; and nothing
perishes so quickly as names which have no meaning,
which are merely the proper names of actual
men, and not the crystallised reflections of a
popular tale. The memory of the past dwells but
little in the mind of the uneducated; the battle of
Minden in 1759, little more than a hundred years
300ago, is utterly forgotten in the neighbourhood,
and, according to Hahn, all that Skanderbeg's
countrymen remember of him is a marvellous
escape which never took place, 1106 while the oldest
Albanian genealogy cannot mount beyond eleven
ancestors. The “Niebelungen Lied” is a most instructive
example of the relation between myth
and history. The Sigurd of the Edda, who gains
possession of the bright treasure of the Niflungs or
clouds by slaying Fafnir, the serpent of winter,
and after delivering Bynhild from her magic sleep,
is made by Gunnar to forget his betrothed and
marry her daughter, Gudrun or Grimhild, — a
crime to be avenged by his murder at the hands
of Gudrun's brothers, again to be avenged, after
Brynhild has burnt herself on Sigurd's pyre, like
Herakles on Mount Œta, by Atli, Brynhild's brother,
— this Sigurd of the Edda re-appears in the old
Saxon tale of “Dietrich of Bern.” Dietrich or
Theodoric rules at Bonn, the earlier name of which
was Bern, and Etzel, the Atli of the Scandinavian
version, is the younger son of Osid, the Frisian
king, who conquers Saxony from King Melias,
and lives in Susat, the present Soest in Westphalia,
while the Nibelungs or cloud-children dwell at
Worms. But the story, as we have it in the
great German epic of the twelfth century, has
301undergone yet another change. Bern has become
Verona, Dietrich Theodoric, the famous Gothic
conqueror of Italy, and Etzel, Attila the Hun.
The Jörmunrek of the Icelandic myth, who slays
Swanhild, Sigurd's posthumous son, is now Hermanric
the Gothic king at Rome, and Sigurd or
Siegfrid himself, with Brynhild and Gunnar
(Gunther), are identified with Gundicar, the Burgundian
victim of Attila, and the Austrasian Siegbert,
who reigned from 561 to 575, married Brunehault,
defeated the Huns, and was murdered by
his brother's mistress, Fredegond. But in spite
of these coincidences, and the historical colouring
that the later versions of a literary age have
given to the old Teutonic myth of the waxing and
waning of summer, we know that neither history
nor even historical names are to be sought for in
the legend. The Attila of history died two years
(453) before the birth of the historical Theodoric;
and Jornandes, who wrote at least twenty years
before the death of the Austrasian Siegbert, was
already acquainted with Swanhild, the child born
after Sigurd's death. If more were needed, the
Icelandic and Saxon versions of the story would
prove the mythic antiquity of the names of the
heroes. Similarity of name or local celebrity may
cause a myth to entwine itself about some personage
or event of actual history, but the latter
302thus far cease to belong to history, and, unless
supported by cotemporaneous evidence, must be
relegated to the ideal land of poetry. The life of
Mohammed is full of mythic elements; fragments
of old Arab folklore have fastened themselves upon
it; and were there no other record of the prophet's
existence, we should have to assign him to
the same category as the Rishis of Brahmanism.
The Charlemagne who has taken the place of
Wodin, as in the group of stars which we still
call Charles's Wain, belongs to myth, and not to
history. Myth has accidentally attached itself to
an actual personage, but it is not the myth which
tells us this. To seek for facts of ethnology and
tribal migration in the mythology of Greece is
but to modernise Euhemerus, who found a Kretan
king in Zeus, and a Pankhæan conqueror in
Uranus. To prop up conclusions so derived by
an appeal to local names is to argue in a circle.
We know that nothing is more liable to corruption
than the names of places and tribes; and the
attempt to explain their new forms will either
itself originate the myth, like the arrow that Little
John “shot over” Shotover Hill (Château Vert), or
occasion the old folklore to localise itself among
them. 1107 The architectural remains of the Peloponnesus
303bear witness to a powerful dynasty such as
that which the Homeric poems represent in the
Akhæan princes; but unless coeval monuments
be discovered to corroborate the legendary picture,
we must not look for further historical facts in the
Iliad and Odyssey. 1108 And even in this case we
304should learn the facts, not from the epic, but from
far different sources. All that the poems can do
is to reflect the manners and beliefs of the age in
which they grew up, and, however much modernized
they may be in their present form, to set
before our eyes the society of a period out of
which was to spring the glorious culture of Athens.
The scientific student of mythology must always
remember that he is dealing with the mythic element
only; historical facts may be imbedded in
it, — upon this point he cannot decide, — but unless
these facts are discovered by historical means, no
amount of ingenuity and conjecture can extract
them from the myth.

As in language, we must be careful to distinguish
in mythology between what is native and
what is borrowed. It would be worse than a
mistake to treat as a pure and original myth the
hybrid conception which resulted from the amalgamation
of Herculus, the old Italian god of
enclosures (from arceo), with the Greek sun-god
Herakles; or of Saturnus, the patron of sowing
and agriculture, with Kronos, who owed his existence
to his son Kronion, “the ancient of days”
305(χρόνος). Nothing but confusion would come out
of such a comparison. In this matter we have to
refer to history wherever this is possible, and, as
in the case of the later Roman mythology, discover
what elements have been imported from abroad:
where it is impossible to do this, language is our
only guide. Glottology alone can warrant us in
tracing myths to the same origin, and Glottology
also must inform us which of them come from a
foreign source. In no other way, for instance,
could the story of Melikertes, the Tyrian Melkarth,
be traced to a Semitic derivation; or, on the other
hand, could Minos be referred to the Aryan man
and manu, instead of being coupled with the Egyptian
Menes, “the founder” of the state. It may
sometimes be difficult to detect the presence of an
alien myth: like borrowed words that assume
native inflections, the borrowed legend may clothe
itself in a familiar form. But, until the two elements
are separated, the comparative mythologist
is not certain of his primary facts.

Again, we must distinguish from the myth a
good deal that is often confounded with it. The
myth is the spontaneous and necessary outcome of
the young mind, which takes its own subjective
fancies as the true objective answers to the questions
inspired by the world around it. Very
different are the conscious and deliberate allegory
306and fable, which generally have a moral intention,
and therefore belong to the period of religion. In
the one the material is lifted up to the spiritual —
it is an effort to express the higher yearnings of
the soul by the known and visible things of sense;
in the other, the spiritual is brought down to be
veiled in the material. The allegory is the product
of individual invention, designed either to conceal
the higher knowledge of the initiated from the
profane gaze of the unlearned, or to explain and
bring it home to them by the aid of metaphor.
It differs from the fable in not making the brute
animals the mouthpiece of its meaning. The
beast-fable seems to be one of the earliest creations
of the awakening consciousness. It was known to
the Egyptians at least as early as the reign of
Ramses III.; and “Renard the Fox” has its analogue
among the Kafirs. Mr Mahaffy conjectures that
Africa, the land of animal-worship, was its original
home; and he mentions, in corroboration of this view,
that the first essays in composition made by the
Vai-Negroes, after Doalu's invention of a syllabary,
were fables about beasts. 1109 At any rate, beast-fables
307were peculiarly appropriate to Egypt, where
“Oppida tota canem venerantur, nemo Dianam.”
Animals live and move like ourselves, and yet
between us and them lies a great gulf, which we
cannot cross to discover what their thoughts and
feelings are. The primitive races of men, accordingly,
regarded them with awe and wonder;
sometimes they were the sole companions of the
hunter and the herdsman, sometimes they were the
organs of departed spirits or divine beings — the
true root of totemism which has made the Malayans
look upon the orang-otang, or “man of the woods,”
as the possessor of superhuman wisdom. 1110308

Besides allegory and fable, another kind of
fiction has to be distinguished from myth. It is
not necessary to mention the imaginary chronicles
of medieval monks, whose seclusion from the
work-a-day world and morbid dwelling upon self
brought about an inability to separate truth from
falsehood, or the interested inventions of patriots
or ciceroni. But writers, more especially among
the Greeks, have in all good faith ascribed eponymous
ancestors to tribes and races, in the belief
that gentile names must have thus originated, and
that consequently the existence of populations
called Hellenes and Assyrians was a sufficient
proof of a Hellen and an Asshur. The notion has
the same foundation ultimately as the myth which
arises from the attempt to explain the signification
309of a forgotten word; and when once it has become
popular, and has been encrusted with the floating
mythology of the people, it passes into a genuine
myth.

Such, then, is the method, and such are the
dangers, of our new science. Already have conclusions
been arrived at which clear up this obscure
province of human history, and enable us to trace
the development and perversion of the religious
spirit. In these researches, Comparative Mythology,
as a branch of Glottology, cannot dispense
with the help of other sciences, more particularly
of Ethnology. The latter has allowed us to penetrate
back into the very roots of the old Theogonies.
We learn that the religious instinct first
exhibits itself in the worship of dead ancestors.
Society begins with a hive-like community, the
members of which are not individually marked
out, but together form one whole. In other words,
the community, and not the individual, lives and
acts. But the community does not comprise the
living only; the dead equally form part of it; and
their presence, it is believed, can alone account for
the dreams of the savage or the pains and illnesses
to which he is subject. In this way the conception
of a spiritual world takes its rise. The spiritual,
however, is recognised only in the sensuous. It
is a sensible image or a sensible feeling which
310convinces the barbarian of the existence of the
supernatural. The spirits are but part and parcel
of the community to which he himself belongs.
There is no difficulty in embodying them in the
objects around him. In his dreams they appear
to him in corporeal shape, and when his tooth
aches he thinks that he feels the gnawing of the
malignant ghost. Hence they are supposed to
take up their habitation in animals and material
things. The Hurons believe that the souls of the
departed turn into turtle-doves; and the Zulus
consider certain green and brown harmless snakes
to be their ancestors, and accordingly offer them
sacrifices. In fact, all serpent-worship has had
this origin: the serpent that crawled along the
ground, and was thought to eat dust, seeming
peculiarly fitted to be the representative of the
buried corpse. “Serpens Libavitque dapes, rursusque
innoxius imo Successit tumulo, et depasta
altaria liquit.” The Pythagorean saying that the
human marrow after death was changed into a
serpent, is but a later form of the old idea; and
the Accadian god of the house, as well as of cities
and wisdom, who was symbolised by the snake, was
primarily the earth, reminding us of the answer of
the Telmessians to Krœsus, ὄφιν εἶναι γῆς παῖδα.
The ascription of spiritual existence to material
objects was from the first inevitable among those
311who had not yet attained individual and subjective
consciousness. Objects equally with persons appeared
in dreams, and it was the ghost of the food
that was offered, and the ghost of the flint-weapon
that was buried, which delighted the dead and
supported him in the spirit-land. As yet there
was no distinction between the form and its content.
Now the cause of the worship paid to the
spirit, and, in short, of any recollection of him at
all, was fear or the desire of food. Terrified by
dreams, or tormented by disease, the savage would
try to appease the angry ghost, while the sole
source of a continuous cult was the appetite. It
was to obtain the needful supply of food that the
daily sacrifice was made and the daily prayer addressed.
It was the animal wants of early man
that kept the light of the religious instinct unextinguished.
When, therefore, the conception of the
spiritual had passed from mere ancestor-worship,
mere adoration of one's own bodily feelings, to
the second stage of object-worship, those objects
which directly influenced the acquisition of food
would receive the principal homage. Fetichism,
by localising the spiritual, instead of leaving the
remembrance of it to the chance of a dream or an
illness, first made it possible to select the objects
which were to be accounted divine, and to remind
the worshipper of his religious duties by having
312his gods perpetually before his eyes. But the
religious ground and kernel of fetichism is the cult
of the deceased forefathers of the community.

With fetichism, the germs of a mythology make
their appearance. The objects worshipped are, as
I have said, those upon which the satisfaction of
hunger mainly depends. The arrow, the spear,
the harpoon, the fruit-tree, such are the gods of the
lower races. Their investiture with independent
life shows that man is still in the infantile stage
in which the object and the subject are confounded
together. Human action is attributed to the inanimate,
and the work of the hands is described in
language as effecting all those results which we
now predicate of nature.

When once, however, human action has been
transferred to an inanimate object, a number of
phrases has been stereotyped in language which will
survive into an advanced condition of knowledge.
Ceasing to represent the knowledge of the day, they
will create an ideal world, illuminated by traditional
reverence and the halo of divinity; and thus
the foundations of a mythology are laid. So the
marvellous Sampo of the Finnic Kalewala is the
last relic of a time when the quern was invested
with the attributes of religious sanctity. Unquestionably,
however, myths which go back to the
period of fetichism are rare. It rather survives in
313the symbols which are attached to different divinities,
in the wand of Hermes and the arrows of
Apollo, or in the refined conceptions of Agnis,
“the fire,” and Hestia, “the hearth.” The period
of fetichism was not one in which the capabilities
of language were much tried; the savage was still
chary of his words, and unconcerned at the loss of
old ones, while the verbal idea of action was still
struggling to express itself. But out of fetichism
came a higher order of things. Through the medium
of conceptions like that of “fire,” primitive
man transferred his religious associations from the
objects which his own fingers had wrought, or
which lay immediately about him, to those whose
nature he could not explain, whose working he
could not influence, and whose power he himself
had felt. The bright vault of heaven, the toiling
sun, the raging thunderstorm, these were now his
gods. The old motive that drove him to select his
deities was still strong; the divine beings that he
honoured were those that seemed to give him his
daily food or to withhold it when they were angry.
The feelings of terror once inspired by the appearance
of the departed in sleep were now confined to
the gods of night, whose subterranean abodes well
agreed with the sepulchres of the dead. It was only
in dreams that these could afflict him; they could
not bring the prey or nourish the plants on which
314he lived; and consequently the worship that he
paid them was forced and scanty. It was the
brightness of the day and the sun, and more especially
of the dawn, when man goeth forth to his
labour and his search for food, that absorbed
almost all his religious care. As Von Hahn has
acutely remarked, 1111 the small part played by the
moon in mythology is in great measure due to the
little share it has in providing for human necessities.
To the sun, on the contrary, the mainstay of
life, the altar smoked and the hymn ascended.
Man was content not to look for his gods beyond
the atmosphere, beyond the space between the earth
and the sky, since here alone were to be found the
powers which enabled him to live and be conscious
of a higher existence.

But the instincts that underlay fetichism were
only transferred to less coarse and unintelligent
objects. There was a worship of nature instead of
stocks and stones. The old confusion between
object and subject was still present, the old childish
ignorance that had fixed its religious intuition in
lifeless things. The new gods, therefore, were
endowed with human action; and when men came
to be more self-conscious and informed, they found
their language teeming with expressions which
could only be explained by remembering that the
315phenomena of the atmosphere had once been divine
beings whose actions were the actions of men.
But this had been forgotten; and so there grew up
an ever-increasing mythology. As the old names
and phrases became more and more obscure, popular
etymologies were invented to account for them,
and Promêtheus, the pramanthas or fire-chark of
the ancient Aryan, crystallised into the wise representative
of forethought, who stole the fire of
heaven for suffering but finally victorious humanity.
Mythology, however, had no past, just as it had no
future. It came down from a period when the verb
had not yet realised the idea of time, and when
the substantives which denoted the individual objects
still served to express also both action and
will. The labours of the sun were the same day
after day; there was no tense to describe them
except the aorist.

It is obvious that what we have called the Epithetic
Stage of language would have been the most
fruitful soil for the birth of mythology. An epithet
is necessarily a metaphor, implying action;
and when we call the moon “the measurer,” we
at once personify it, that is, ascribe to it the
action of a man. But not only was there thus
from the first a mythic element introduced; the
epithet, being equally applicable to a variety of
objects, would tend to confuse their qualities together,
316and when one special application of it was
preserved through religious sanctity as a mere
name, all the original meaning and reference being
lost, a number of incongruous attributes, derived
from other applications, would be associated with
it. In fact, in proportion as a community has
advanced towards the epithetic stage, the mythological
wealth of its language is large. Myths are
the traditional relics of the way in which primitive
man confounded his own subjective sense of power
with the objects which animal needs had led him
to consecrate as gods, as well as of the attempts
made to explain them when the state of society
and knowledge which had produced them was
changed. They rested upon the religious instinct,
and it was this that saved them from perishing.

The results of Comparative Mythology have not
escaped misconception and objection. It is no
doubt hard for those who have been brought up to
regard the myth as a corruption of revelation or a
perversion of an historical fact or a sacerdotal
allegory to disabuse themselves of their belief. It
is harder for those who have been accustomed to
hunt for fragments of history in the mythology of
a nation, under the guidance of a special divination,
to acquiesce in the decisions of a study which declares
that all such labour is in vain, that myth is
to uncultivated man what history is to us, and that
317any historical references that may be imbedded in
it can only be discovered from ordinary historical
sources. The method, however — that of comparison —
by which these conclusions are obtained is
the method of science; and, if properly carried out,
can alone lead us to scientific truth; but it must
not be held responsible for the rash statements of
over-hasty disciples, who are not contented with
the restrictions imposed by our evidence. Just as
we shall never be able to give the derivation of
every word in the dictionary, so we shall never be
able to explain every individual myth, and the
endeavour to do so necessarily brings discredit
upon the conclusions arrived at on sufficient data.
We must be content with general rules and the
explanation of the larger number of myths. The
two chief objections, however, raised against the
results are, on the one side, that they presuppose
in primitive man too high an imagination, and,
on the other side, that they ascribe to him too
feeble an imagination. We might leave these
mutually destructive statements to neutralise one
another, but it is better to clear up the misunderstandings
upon which they are based. We are
told, then, on the one hand, that to believe that
our barbarian ancestors were always busied in
describing the wonders of the dawn and the daily
progress of the sun through the sky in richly
318poetical metaphors is simply absurd. The country
boor is blind to the beauties of nature, and the
savage cares only for his selfish animal lusts.
But it is precisely the latter fact which solves the
difficulty. It was just because the dawn and the
sun and the fire seemed to provide him with the
food which he needed, that primitive man regarded
them as his gods, and invested them with human
power. The poetical dress which has been thrown
over them is a necessity of language. Poetry consists
in metaphor, personification, and terseness,
and all these were the inevitable characteristics of
early speech, when the spiritual could only be
understood through the sensuous, and when object
and subject were inextricably blended together.
It is scientific language that is furthest removed
from poetry; the savage still talks in poetic metaphor,
and the earliest compositions are in verse.
The rhythm that underlies the myth is the lyric
rhythm of speech — the most exquisite of all music;
and the deep insight that pervades it is the
naive simplicity of childlike humanity, and the
religious conviction which it would express. As
for the contrary objection, that our forefathers
could not have had such a poverty of ideas as to
confine all their attention to the phenomena of
the atmosphere, it is answered by the same consideration,
that the choice of the objects of mythology
319was dictated by the circumstances in which
the first men were placed. We do not find that
the range of ideas possessed by the modern savage
is very great, and the very growth of mythology
implies that the imagination increased with changing
conditions. That the elements were only
modified, enlarged, and combined, but not added
to, is due to the religious core to which mythology
owed its preservation. Indeed, without the religious
instinct, mythology would have had no
existence at all; it originated not in the imagination
of the poet, but in the requirements of
worship. As a matter of fact, however, the assertions
of the comparative mythologist, whether
likely or unlikely, are no subjective theory, but the
plain reading of the evidence before us. In many
cases, at least, the Rig-Veda, our earliest Aryan
monument, does show that a Greek legend had a
solar origin; and so long as we keep to our data,
we can find nothing to support us in tracing back
our European mythology to anything else than
atmospheric phenomena. If there were any primitive
myths of a different derivation, we have no
means left of detecting them. Nor is it in the
Aryan family alone that the same conclusion is
necessitated, although the inflective character of
the language and the extensive development of the
epithetic stage would lead us to expect to meet
320with more mythology here than anywhere else.
The myths of other races, wherever their meaning
is transparent enough, wherever the proper names
are capable of analysis, are all atmospheric and
celestial. Thus the Eskimaux have a legend
about the moon, how he met a girl in a dark hut
at a festive gathering, and declared his love by
shaking her shoulders. She smeared her hand
with soot and marked him; but when a light was
brought, she found it was her brother, and fled
ever pursued by him through the sky, where the
moon is always chasing the sun with a dark spot
upon his blackened cheek. The Assyrians, again,
borrowing apparently from their Accadian predecessors,
told how Allat or Astarte, “queen of
heaven, with crescent horns,” descended from the
sky through the seven gates of Hades, leaving at
each some one of her adornments — her earrings,
her necklace, her girdle, her anklets — so that at
last she reached the land of the dead, where the
sun of winter was sleeping, stripped and empty;
to return again, however, and receive back at each
gate the ornaments she had left behind. No one
can fail to see here the waning and waxing moon,
any more than to understand how the sun-god can
be addressed in an old Babylonian hymn as the
opener of the bright locks of heaven. 1112321

The last two arguments urged against the scientific
interpretation of mythology are, firstly, the
elastic limits and vague and general characteristics
assigned to the myth; and, secondly, the narrow
local restrictions to which it is often subject. It
is said that any story of life and death and marriage,
any tale in which the hero migrates from
322east to west, ought, upon the theory, to be admitted
into the circle of solar myths. In fact, so general
are the features which are attributed to the myth,
that it is possible to transmute any individual
whatsoever into an image of the sun, just as Arch-bishop
Whateley banished the great Napoleon to
the realm of fable. But all this proceeds upon
the mistaken assumption that it is only necessary
to compare two legends together to determine
their character. On the contrary, a scientific comparison
must conform to all the rules of the special
science; and since Comparative Mythology is
but a branch of Glottology, we must not advance
one step without the safeguard of language.
Hêraklês is the sun, not only because his life and
labours are those of other solar heroes, but also
because his own name discloses his origin from
swara, “the splendour of heaven,” like the names
of those with whom he comes in contact — Augeias,
Deianeira, Iole — in his struggles and in his death.
The second objection is even less plausible. When
it is asked why the story, for instance, of Kephalos
and Prokris, the rising sun and the dew-drop,
should have been so local in character that
no allusion to it appears before the time of Apollodorus
and Ovid, we can only reply, why is it
that so many old words are utterly obliterated in
the language of the country, and yet crop up in
323these latter days of linguistic research in obscure
provincial dialects? Our good old English laik,
“to play,” only lurks now in the corners of the
northern counties, just as many a myth of pre-Homeric
Greece survived in the mouths of illiterate
peasants, to be discovered and recorded in
the days of court dilettanti and antiquarian bookmakers.

When once the question of mythology has been
settled, we can proceed to the comparative science
of religions, or, if we might coin a word, of Dogmatology.
What we have to do here is to compare
and classify the various religious systems
that have prevailed in the world, and to trace
their connection, origin, and development. It is,
of course, only the external form and shell with
which we are concerned; the religious spirit which
inspires them must be left, as in mythology, to
other students. We have nothing to do with the
truth or falsehood of particular religions; that is
a point which must be handed over to the theologian.
Nor is it our business to ascertain the
history of a special creed, and the unfolding of its
dogmas; the quarrels of Catholics and Arians,
the disputes of Nestorius and St Cyril, are of
little consequence to us; what we want are the
general results, just as Glottology makes use of
materials provided by the specialists in each language.
324Still less have we to deal with the biographies
of religious founders or reformers; for
all that Dogmatology requires, they may be mythic
personages. It is the ideas, or rather the forms
of the ideas, which they utilised and arranged,
and the way in which these were afterward modified
and added to, that have an interest for us.
As in most cases we can arrive at these only by
the aid of language, the science of religions will
need the control of Glottology as much as does
Comparative Mythology. As yet the science of
religions has made but little progress. We are
still engaged in collecting materials, in learning
to read the sacred books of the East, and to
ascertain what it is they have to tell us. Nevertheless
certain general outlines, within which the
conclusions of the new science will have to be
comprehended, have already been sketched. We
have come to see that religious systems and their
development are obedient to general laws like
everything else, and that each race of men has
shaped its system in a manner of its own. The
isolating Chinese differs in his form of creed from
the inflection-using Aryan and Semite, and these,
again, carry out their religious ideas in a different
way; but a general likeness is to be observed
between the latter. As Buddhism and Zoroastrianism
have come forth from the bosom of
325Brahmanism, so have Christianity and Mohammedanism
from Judaism; and just as Buddha
preached the equality of men, in contradistinction
to the aristocratic creed of Manu, so the
exclusiveness of the Jew has given place to the
universality of Christianity; while the prophet of
the Avesta was not less clear in his sharply-cut
dualism, out of which Monotheism was to spring
by the absorption of good into evil, than was the
prophet of the Korân in his doctrine of one God.
Indeed, Buddhism and Christianity present closer
analogies than that of mere derivation. Just as
Sakya Muni appeared about 600 years before the
birth of Christ, so did Mohammed about 600
years after that event: 300 years after its institution,
Buddhism was made the state religion by
the powerful monarch Asoka, by whose orders a
general council was convened to settle matters of
faith and discipline, just as Constantine was converted,
and the Council of Nikæa assembled by
him, a. d. 325. The monasteries of Christendom
find their parallel in the monasteries of Buddhism,
the Pope of Rome in the Lama of Tibet; and the
image-worship, the proxy-prayers, and the elaborate
ceremonial of the medieval Church are not
more unlike the divine morality of the Sermon
on the Mount than are the adoration of relics,
the praying machines, and the rites of the Buddhist
326hierarchy unlike the simple code of morals
and life which their founder bequeathed to them. 1113

Much that is now dark in dogma may be cleared
up with the advance of our comparative researches.
Here, as elsewhere, it will be found that we inherit
the forgotten beliefs of our forefathers. The
words, the phrases, the practices have descended
to us from the past, but we have put into them a
new spirit and a new meaning. The founder of
a religion, however great he may be, however
much, as his disciples believe, a prophet of God,
or God himself, has yet to deal with men. He
must work upon the ideas current in his age; and
though he may give them a fresh direction, still
their comprehension and carrying out will be
limited by the intellectual knowledge of the recipients.
And as this will vary from generation to
generation, so will the ideas themselves vary, and
catch the colour of each succeeding century.327

Chapter IX.
The influence of analogy in language.

The phenomena of phonetic decay are among the
first to attract the notice of the student of language.
They show themselves, as it were, upon the surface
of speech; they force themselves upon the attention;
and the slow and gradual change that goes
on in language, seems, at first sight, to be due to
them alone. The wear and tear of words and their
meanings which is continually taking place, however
little perceived by the passing generation, is
like the washing of the rocks by air, and water,
and ice, that, through the long series of geological
ages, has piled up the crust of the earth, scooping
out the valleys and moulding the everlasting hills.
And just as this constant process of destruction
has blotted out myriads of intermediate links
between successive forms of life, bringing about
the so-called imperfection of the geologic record,
so, in language, the action of phonetic decay has
left us but waifs and strays of former states of
speech, and obliterated words and forms which
328alone can explain the origin of what is left, or
affiliate languages one to the other. It is not
wonderful, therefore, that phonetic decay has assumed
an exaggerated importance in the eyes of
philologists, or even been brought forward to the
exclusion of all other principles which affect linguistic
growth. Its very name, however, shows
that it cannot be a principle of universal application.
Phonology is but a subordinate part of
Comparative Philology, dealing at most with its
material, not with its content. And though, in
philology, material and content can only be arbitarily
separated one from the other for the purposes
of scientific analysis, inasmuch as language
is but the outward expression of thought, yet a
principle which primarily deals with the external
alone must be of limited and not general range.
In fact, when we come to look closely into the
matter, we shall find that phonetic decay is largely
influenced by another and wider principle, that of
analogy. This is a main element of change in
the signification as well as in the outward form of
words; and just as phonetic decay wastes and
destroys, so analogy repairs and reconstructs. The
one is the agent of destruction, the other of construction,
though they both spring from the same
root of human laziness.

One of the most important of the functions of
329analogy is the production of a new grammar.
Grammar is not only the skeleton of a language,
but the very life-blood of it as well, and the
changes that take place in it are in large measure
occasioned by the agency of analogy. But, analogy
may be either false or true; indeed, in the
history of speech we shall see that false analogy
has as often been at work as true. A large number
of feminine nouns in French, like étude and
voile, have arisen from the mistaken comparison
of the plural neuter ending in -a with the similar
termination of the singular of the first Latin declension.
And so, as the majority of the substantives
belonging to this declension were feminine,
mind and ear came to associate the idea of
the feminine so closely with, the termination -a as
to assign that gender to all words whatsoever which
ended with this particular vowel. The instinct
here led to a false conclusion; that is to say, it
ignored the true history and significance of certain
linguistic forms; and this confusion and violation
of the regular historical development of speech is
all that is meant by the philologist when he speaks
of the false in language.

But analogy does not act upon forms alone.
Both matter and form are alike subject to its
influence. While, on the one hand, the relations
of grammar, the rules of syntax, and the content
330and meaning of words grow changed and altered
by its subtle operation, the external shape and
character of the vocabulary, on the other hand,
also becomes insensibly transformed. No doubt
the two processes of change go on, for the most
part, side by side, since we cannot, except on paper,
separate the inner essence of a word from the
material in which it is expressed; but there is no
more fatal error than to assume that a new conception
or a new grammatical relation can arise
out of mere phonetic change. They are due to
analogy, not to phonetic decay. It was not the
neo-Latin pronunciation and external form of
voile that caused it to be feminine, but the fact
that a particular external form had already been
appropriated to the feminine gender in a preponderant
number of instances. The inward in
language, as in other things, cannot be originated
by the outward, however much the outward may be
originated by the inward. It was to be a vehicle
for internal thought that language first came into
existence; and the popular etymologies, which
modify the outward form of a word in order to
harmonise it with an intelligible idea, still bear
testimony to this.

Now the principle of analogy may be ultimately
traced partly to the desire of saving trouble, partly
to the natural instinct of imitation. It is easier
331for the vocal organs to repeat the same sound than
to attempt a new one, while the repetition of the
same idea, or the expression of an analogous one,
involves less exertion on the part of the mind.
Habit is a ruling power in life, and sounds or
ideas to which we are accustomed rise uncalled-for
to the intelligence and the lips. Every one
must have experienced the difficulty of pronouncing
some sound in a foreign language to which there
is nothing similar in his own; and in proportion
to the strangeness of the sound will be the vocal
effort to produce it. The more regular the inflectional
system of a language, the more readily do
we learn it; and the ease with which a knowledge
of Italian, as compared with German, may be
acquired, results to a great extent from the superior
regularity of its inflections. The tendency of
all linguistic progress is to reduce the number of
anomalous forms, and bring them all, whatever may
have been their origin, under one and the same
type. Thus, in modern Greek, certain declensions
have become the prevailing models in accordance
with which substantives are declined, and
words like φύλαξ have long ago become φύλακος.
This process of assimilation of sounds and grammatical
terminations is accelerated by contact with
another language. Exceptions and irregularities,
which seem quite natural to a native habituated to
332their use, are always hateful to the foreigner, inasmuch
as they require a greater effort of memory
and attention. To say nothing of the general loss
of inflections, many of which had grown otiose,
the preponderance of the plural in -s in English is
due to the Norman invasion, as well as the softening
or dropping of the guttural aspirate in words
like enough and though. 1114 Assimilation, so frequent
a cause of phonetic change, is wholly occasioned
by the attempt to avoid pronouncing a fresh sound,
and by allowing analogy to operate upon an
adjacent letter; and other phonetic changes are
extended and stereotyped in speech by means of
the same principle. But it is not phonetic change
alone that is influenced in this way by the wish to
save trouble. Nothing is harder than to think out
a new thought or to grasp a totally new idea;
hence the conceptions applicable to one set of
phenomena are transferred to an entirely different
and perhaps even contradictory set; and similarities
are ingeniously detected between the most
dissimilar things. It was long before the several
relations of space, time, and manner came to be
333distinguished from one another, and we may even
now hear expressions used of time which can
strictly be employed of space alone. So, again,
the wealth of meanings contained in our dictionary
is the result of the endeavour to get at new ideas
without having to take the trouble of inventing
them. The words post and arm may be selected
as illustrations of the roundabout way in which the
human mind goes to work to increase its store of
conceptions. From the simple idea of a thing
placed or set we have a myriad derivatives, from
the stake fixed in the ground to the medium of
modern correspondence. Similarly, analogy has
extended the signification of arm to the weapon a
man carries, a channel of the sea, or the power of
law. If Curtius is right, no better instance can
be found of the extraordinary transformations of
meaning undergone by words than the Homeric
adjective φοξὸς, which, derived from the root bhaj,
“to bake,” originally signified a vessel of baked
clay, and finally came to be applied to the head of
Thersites, the peaked shape of which resembled the
household amphora with its pointed bottom for
sticking in the ground. 1115 When we consider the
manner in which the lexicon is enlarged and altered
by the action of analogy, and the unlikely cases to
334which we find it applied, we may well be cautious
in assuming the primitive independence of two
roots which agree in sound but differ in meaning. 1116
It is not more difficult to understand how the
name of the guinea-pig could have been given, and,
what is more, accepted, among people acquainted
with the swine, than to comprehend how the South
Sea Islander could call the dog a pig, or the
Kuriaks the ox “the Russian elk” (Ruski olehn). 2117

We have now to examine the influence of analogy
in language as affecting its matter and its
form. And first, as regards its matter. Here we
find it bringing about changes in accent, in quantity,
and in pronunciation generally. False rather
than true analogy is the guiding principle; that
is to say, the historical reasons for a certain pronunciation
are forgotten, and a word is made to
335conform to what, from some cause or other, has
become the favourite and most common type.
Thus the general tendency of our own language is
to throw back the accent as far as possible; and
accordingly words like balcony and illustrated,
which fifty years ago were pronounced with the
accent on the penultima, are now usually sounded
bálcony and illústrated. Contemplate and blasphemous,
in which Tennyson and Milton preserve
the penultimate accent, are now almost always
accented on the first syllable; 1118 and revenue has
long followed their example. Sooner or later
analogy is pretty sure to force all exceptional
cases into harmony with what has become the
prevailing rule of pronunciation. In no other
way, again, can we explain how it is that whereas
Irish and Bohemian accent all their words on the
first syllable, Welsh and Polish accent all theirs
on the penultima. 2119 Welsh and Irish on the one
side, Bohemian and Polish on the other, are dialects
too closely related not to oblige us to believe
336that there was a time when the common idiom
out of which they have severally developed was
pronounced in the same way; but circumstances
caused a particular mode of accentuation to become
fashionable in each of the separated dialects,
and the whole stock of words in each was there-upon
gradually brought under the dominant type.
It must have been much the same with Latin and
the Æolic dialect in Greece. We now know that
the regular throwing back of the accent as far as
possible was the late product of the action of analogy,
and not the survival of a primitive practice.
The normal Greek accentuation agrees with that
of Vedaic Sanskrit, even in such seemingly arbitrary
cases as the different position of the accent
in the numerals pánchan, πέντε, and sáptan, ἑπτά;
and Doric Greek has more truly preserved the
paroxyton of the third person plural in the second
aorist, which primarily ended with a long syllable
(ἐτύποντ), than has the classical dialect, where
the ancient form of the word has been forgotten,
and the tendency that became a dominant law in
Æolic and Latin has been followed. 1120 Intimately
connected with accent is quantity; and in this
also the modifying influence of analogy has been
active. A good example is afforded by the Latin
rule, which allows a vowel before a mute followed
337by a liquid to be either long or short. Thus we
have latēbræ or latĕbræ from latēre, scatēbra or
scatĕbra from scatēre; and Horace addresses the
fountain of Bandusia as “splendidior vĭtro.” Now,
by all the laws of prosody, the first syllable of
vitrum ought to be as long as the middle syllable
of latebræ and scatebra, since the word stands for
vis-trum, that is, vid-trum, from the root vid. But
in numerous instances the vowel before the double
consonant was short by nature, and since this
could be lengthened when necessary, the belief
grew up that any vowel before a mute and a liquid
might be either long or short, and so a naturally
long vowel came to be used as a short one
wherever the exigencies of the metre demanded it.
Much the same has happened with words terminating
in d or t. The majority of vocables which
ended in a dental had the vowel of the last syllable
properly short; and accordingly all such syllables
came to be regarded as short, whatever may have
been the original length of the vowel. Hence it
is that we find words like sed, the old ablative of
the third personal pronoun, or sit, the contracted
form of the optative siet with long e, 1121 employed
with, a short vowel. So again, in Greek, Hartel
has shown that the -ι of the dative plural and the
338-α of plural neuters were primitively long — a fact
traces of which may still be observed in Homer. 1122
The short final syllable, however, of the third person
plural of the verb (λέγουσι, for λέγοντι), and the
short accusative termination of nouns of the third
declension (ποδὰ, for ποδὰμ), prepared the way for
shortening every terminal -ι and -α; and when once
the ear and tongue had become accustomed to the
shortened form of these terminations, every fresh
case that occurred had to conform to the general
analogy.

But not accent and quantity only, the pronunciation
of syllables and letters also falls under the
same principle of change. How much, indeed, a
change in the latter depends upon a change of
accent may be seen from the so-called guna, in
which the modification of the vowel is entirely
occasioned by the stress laid upon it. But whatever
may be the cause of change, when once the
new pronunciation has taken a firm hold upon the
speech, it gradually extends itself to the whole
vocabulary, so that it may happen that a sound
formerly familiar to a language dies out so utterly
that the speakers find themselves unable to pronounce
it when met with in another dialect. A
striking example of this is to be found in the
339history of the guttural aspirate in English. A
similar occurrence seems to have taken place in
Assyrian in the case of the letter 'ayin. This had
been thinned into a modified i vowel, so that
when they wanted to express the name of the Palestinian
city Gaza (הזע), they could find no better
representative of the old guttural sound of 'ayin,
as preserved in Western Semitic, than the ordinary
guttural aspirate kheth, and so הזע ('azzah)
was written khazitu. 1123 The action of analogy upon
pronunciation, however, is nowhere exemplified
more clearly than in the adoption of foreign words.
A Frenchman drops the final consonants of the
names and terms which he borrows or uses, and
the Englishman speaks of Marsails and Paris, of
lieutenant and passport. Indeed, our own language
is a most interesting monument of the profound
and universal change in pronunciation that may
be brought about by the influence of analogy when
special circumstances give a particular mode of
pronunciation a superiority in the linguistic
struggle for existence. Our vowels are no longer
what they were three centuries ago. The a and
the i have become diphthongs, and the e has taken
the place of the i. Sir Christopher Wren's cathedral
is St Paul's, not St Powl's, and the final e
340has long ceased to be sounded. But a change is
still going on. Just as it has been remarked
that French is becoming more and more nasalised,
so also has it been noted that the vowels in English
are continually growing more and more
thinned. The broad a in words like mast or bad
is a mark of Cockneyism, and the diphthongal
sound of u is extending itself on all sides. The
same preference for diphthongal sounds is making
itself apparent in words like either and neither,
the first syllable of which is beginning to be pronounced
as though it were German, although the
only other word in English by which such a pronunciation
could be supported is the misspelt
height from high. It has been pointed out 1124 that
the reason why we pronounce the three first vowels
of the alphabet in a way essentially different from
that in which they are sounded in the majority of
our words is because they are so pronounced in
the three vocables in most common use, a, me,
and I; and it is a curious instance of the power
of analogy that, although the numerical majority
of words is against it, yet the frequency with which
this peculiar pronunciation is heard in the three
words just mentioned, aided by the values assigned
to the three vowels as letters of the alphabet,
341induces us to give this pronunciation to them
wherever they occur in foreign terms with the true
pronunciation of which we are unacquainted. 1125 But
the influence of analogy will, of course, be proportionally
greater where we have to do, not with
the spelling, but with sound exclusively; and the
power exerted by the principle over written words
will enable us to understand how largely it will
affect spoken words, more especially in an illiterate
society. This, I think, will afford an explanation
of the phenomena of Grimm's law. Accident,
so to speak, may have made a particular pronunciation
of some letter predominant in one of the
branches of the Aryan family; but when this pronunciation
had once fixed itself in the most commonly
used words, or in the majority of them, or had
approved itself to the popular taste by its greater
easiness of utterance, or by some other reason, it
was extended to every case found in the vocabulary.
We may thus account for the remarkable
uniformity and regularity in the shifting of sounds
which is observed in the several members of the
342Indo-European group. In this instance the
action of analogy would have been natural; but
it may also be produced artificially. A good
illustration of this is to be found in the Homeric
dialect. The Iliad, and, to a far less extent, the
Odyssey, 1126 are the growth of generations, old epic
343formulæ and verses being handed down traditionally
from rhapsode to rhapsode, and imitated and
incorporated in a slowly increasing body of poetry.
But the imitations very frequently were based
upon a false analogy. The old pronunciation had
been lost; and what had really a sound philological
origin was supposed to be due to metric
license, and so came to be applied to totally
different cases. Thus, as Mangold has pointed
out, the analogy of the diektasis or so-called
resolution of the vowels in verbs in -αω, where it
had a philological reason, led to a similar resolution
of the syllable in verbs in -οω, in the conjunctives
of verbs in -μι, and elsewhere, where it
was wholly unjustifiable. 1127

Before leaving this part of the subject, I must
not forget to notice how, even in the material of
language, analogy shows itself as a creative and
reconstructing principle. English has somewhat
doubtfully enriched itself with several anomalous
plurals by means of it. The distinction of vowel
between man and men, foot and feet, was, in Anglo-Saxon,
purely euphonic. The dative singular was
men, just as much as the genitive and dative
plural were manna and mannum. But the thinner
344form with e occurred more frequently in the plural
than in the singular, and so, when the cases of
the old language disappeared, the unmeaning
difference became significant. A absorbed the
singular and e the plural, instead of being merely
predominant. The distinction between the present
and perfect of verbs like lead, led, is of the same
nature. The Anglo-Saxon imperfect was marked,
not by the change of vowel, but by the flection;
when this was dropped, however, the greater frequency
of the obscure vowel in the past tense,
owing to the inflection (ledde), caused it to be
assigned to all the persons, and to become a characteristic
of the tense. So in Greek, the distinction
between verbs in -άω, -έω, and -όω was at first
purely phonetic, each of the vowels being only a
modification of the same original termination
which we have in Sanskrit; but in course of
time, in consequence of the accidental fact that a
considerable number of verbs in -οω were active,
and of verbs in -εω neuter, the form in -οω came
to be more and more set apart to denote a transitive,
and that in -εω an intransitive notion, while
the form in -αω floated between the two senses.
It is true that the distinction in meaning was
never exhaustively carried out, but we can hardly
doubt that it would have been had Greek lasted
long enough and never become a literary language.
345It is highly probable that the significant vowel of
Arabic verbs has the same history. Ιn Arabic, u
(and i, for the most part) marks a passive sense,
a generally an active one. Now traces of this
distinction are to be found in Hebrew and Aramaic,
as well as in Assyrian; but what has become the
rule in Arabic is at most nothing more than a
tendency in the other Semitic idioms. 1128

The change of the euphonic into the significant
vowel would be much easier in a Semitic than in
an Aryan tongue, since analogy would be all in
its favour — Semitic grammar preferring to effect
by internal vowel-change what is left to external
flection in the Indo-European group; but the
instances given above show that the process is not
unknown to our own family of languages wherever
association of ideas and sounds may favour it. In
contrasting Aryan vocalism with Semitic consonantalism
it is impossible to draw any sharply-defined
lines of distinction. Here, as everywhere else, however
true our classification may be, yet the several
classes pass insensibly one into the other, and we
cannot precisely determine their boundaries. It
is to this fact that the idolum of the three stages
in the growth of a language mainly owes its origin.346

We must next consider the manner in which
analogy has acted upon the form and content of
speech. This is the side upon which it has been
most influential, and where its consequences have
been most important. Some form suddenly gets
into vogue and replaces older ones, or leaves but a
few of them, which henceforth are regarded as
abnormal exceptions; or, again, a new grammatical
relation is elaborated in some particular case, and
then extended to others more or less similar.
Thus the English perfect in -ed has become predominant
in the language. Originally dide, the
reduplicated past tense of do, it was affixed to
verb after verb, until only a few were left which
still follow the primitive method of conjugation,
and every new verb taken into use has to form its
perfect by means of it. 1129 The Latin perfect in -vi
347or ui, and future in -bo, grew up, in the same way,
by postfixing fuo, fui, in a few instances, which
continually tended to become more and more
numerous. In French, every fresh verb has to
belong to the first conjugation. There is no reason
in the nature of things why the language should
not employ words like électrisoir, photographir,
but the mysterious influence of analogy has ruled
that only électrifier, photographier, should be admitted
into the domain of speech. 1130 This is the
form which, in the struggle for existence, has
established itself to the exclusion of every other.
Ear and mind had grown accustomed to the
association of a special sound with a special sense,
and could allow of no other. One of the most
striking examples of the way in which this sometimes
results in the creation of wholly new flections
is the distinction of gender in the nominative
singular of Latin comparatives. This arose in the
historic period, and we can accordingly trace its
genesis. The termination of the nominative was
indifferently -ior or -ios (-ius), like the Greek -ίων
and Sanskrit -yan, from an original -yans, r in
Latin commonly standing for s, — e. g., in arbos
and arbor, generis (Sansk. janasyas), from genus,
&c. In Valerius Antias (apud Priscian., vii. 345),
348we still find prior used for the neuter (“senatus
consultum prior”), and the title of the Fourth
Annal of Cassius Hemina was “Bellum Punicum
posterior;” but the connection of the idea of the
neuter with the ending -us in such words as opus,
genus, and the like, and of the masculine with the
ending -or in such words as honor or arbor brought
about the specialisation of form which we meet
with in the classical age. 1131 The remarkable regularity
which we find in the Assyrian conjugations
has been produced in much the same way.
Its artificiality is shown by a comparison of the
other Semitic idioms. Kal, niphal, and shaphel,
the active, passive, and causative, were taken as
the three primary voices; and not only was kal
provided with the intensive pael, but a niphael
and shaphael were formed as well; while the
secondary conjugations in t and tan were attached
to each of the principal voices, including pael. 2132
349Analogy, however, will sometimes bring about
far more wide-reaching effects than the alteration
or production of certain grammatical forms
and relations. It may change the whole character
of a grammar, the whole structure of a language;
provided, that is, that the fundamental principles
upon which it is based, the mental view of the
people to which it belongs, be not violated.
Thus Coptic, which was formerly an affix-language
like old Egyptian and the Semitic tongues, has
become a prefix-language, resembling in this
350respect the Berber, the Haussa, and other sub-Semitic
dialects of North Africa. When we
remember the formal relationship between these
and the Semitic idioms, the conclusion seems
forced upon us that they also have undergone
the same change as Coptic, and assumed their
present appearance within a comparatively recent
epoch.

Analogy is equally active in the province of
syntax. The analytical character of the modern
European languages, of which English is the most
extreme example, is largely due to its influence.
The substitution of prepositions and syntactical
contrivances for inflection has gradually become
the rule instead of the exception. The contact of
the Teutonic and Romanic nations brought about
the consciousness and analysis of the relations of
grammar, which is out of the question so long as
the native dialect alone is known; and the tendency
to replace flection by analysis and accidence
by syntax extended itself with an ever-increasing
rapidity. The few remains of flection, the last
relics of an uncultured age, which still exist in
English, may be expected to disappear in time,
even supposing that pigeon-English does not become
the universal language, as a recent writer
prophesies. 1133 Already the inflected genitive in -s is
351more and more disused, and confined to poetry or
a highflown style, the general receptacle of antiquated
forms; and it may not be long before this
fossil-like survivor of nominal inflection becomes
as totally extinct as it is in modern Persian, where
the genitive is denoted by the short vowel placed
between it and the preceding nominative. The
manner in which we express the relation of the
genitive must follow the common analogy, and
be no exception to the analytic character of our
speech. Another example of the effect of analogy
upon syntax may be found in the history of the
relative sentence, which has been so ably investigated
by Jolly and Windisch. 1134 Comparative syntax
teaches us that the relative sentence was primarily
expressed by being immediately subordinated to
the principal clause without the addition of any
explanatory word, just as it may be in Hebrew or
Assyrian poetry, and in such English phrases as
“This is the man I saw.” For the sake of clearness
and emphasis, however, the object of the
antecedent clause was repeated in the consequent
by some demonstrative term signifying locality,
and the attention was thus drawn to the idea
intended to be signalised. Thus in Chinese, the
352relative so properly means “place;” 1135 and Philippi
has shown that the relative pronoun in Semitic
was originally a demonstrative. 2136 So it was also
in our Aryan family. But after a time, this pronoun,
this representative of the object denoted,
came to be used in all cases, and not merely where
peculiar stress was wished to be laid upon it; and
when analogy had thus uniformly extended this
particular employment of the word, it ceased to
convey any longer a purely demonstrative sense,
and assumed a relative signification, which was
then applied by the further operation of analogy to
instances in which the demonstrative could hardly
have been employed.

The last illustration that need be taken is the
position of the verb and objective noun in the
sentence. It is remarkable that whereas the normal
place of the verb in Latin is at the end of
the clause, and that the same rule may be said to
hold good in German and Dutch, the Romance
languages, which have grown up through the
contact of Teutonic and Latin populations, place
the verb before the objective case. English follows
the same order, although poetry or a poetical
style are still allowed to adopt a contrary arrangement
353without fear of unintelligibility. Now it
would seem to us, who are accustomed to such an
usage, that the verbal action ought naturally to
come between its subject and the object upon
which it is directed; and the fact that this is the
order of ideas observed in those dialects which
have arisen through the attempt of two races to
render themselves mutually intelligible would
appear to support this view. On the other hand,
this order is found only in the analytic stage of
Aryan speech, that is, in its latest and most
modern form, while the arrangement which sets
the verb at the end becomes more and more universal
the older the language with which we are
dealing. Which arrangement, then, is the most
natural? Not the most simple, for the two terms
are by no means synonymous; and philology is
continually reminding us that what is logically
prior is seldom historically so, but that simplicity
and clearness are only reached by a slow and
laborious process. The answer to our present
question is furnished by observation of the deaf
and dumb. Deaf mutes enable us to a certain
extent to make the experiment which Psammitikhus
is said to have attempted, and to see
what kind of language the uneducated mind would
form for itself when deprived of the power of
learning one of those spoken idioms which have
354been elaborated by preceding generations. Now
it is found that the deaf mute invariably places
the verb at the end of the sentence, the subject and
object, to which his thought is chiefly directed,
being the first to occur to his mind. 1137 The alteration,
therefore, which has been brought about in
English and the Neo-Latin dialects in this natural
order of ideas must be due to the action of some
influential principle like analogy. A speaker who
is imperfectly acquainted with the language of
another has to ransack his memory for the names
of objects and conceptions in the foreign tongue;
in order to gain time for this, he defers mentioning
the object of action as long as possible, and
interpolates any other words he can between the
subject and the objective noun, the verb being, of
course, one of the first. So convenient an arrangement
of the sentence would be more and
more extended by means of analogy, until it
finally became the characteristic of the language.

If analogy, however, has done so much for the
accidence and syntax of a grammar, it has done
everything for the meaning of words. Professor
Whitney reduces the changes of signification which
are perpetually going on in the lexicon to two
processes — one the specialisation of the general,
355the other the generalisation of the special. 1138 But
the agent of change is analogy. A general term
is applied to some particular object, or a special
term to a less special instance through some
likeness supposed to exist between them; new
likenesses are then detected; the terms are used
of fresh cases; and so the process of the analogical
expansion or contraction of signification
goes on indefinitely. To make our meaning plain
to another, it is necessary that we should employ
words which he understands, and we can only
convey a new idea to him by comparing and
likening it to one with which he is already
familiar. Indeed, it is not only in the instruction
of others, but just as much in the development
of our own knowledge, that the same contrivance
is required. One idea is best remembered
by being connected with another idea, no matter
how fanciful the connection may be; and it would
be quite impossible to recollect a large mass of
isolated ideas. Knowledge is one vast chain of
associations, and analogy is the principal forger
of its several links. A new fact falls within our
experience, a new object is discovered, or a new
notion is struck out, and we at once seek to bring
it within the circle of our previous knowledge, and
to connect it with something with which we are
356previously acquainted. The name assigned to it
accordingly expresses the resemblance believed to
exist between the new subject of thought and the
existing furniture of the mind. It has long been
recognised that all the terms which denote the
spiritual and abstract are derived from the physical
and concrete. Spirit is primarily “the
breath,” soul, “the heaving sea,” Deus, “the
bright heaven.” Language is the treasure-house
of worn-out metaphors. As Carlyle has said,
“They are its muscles, and tissues, and living integuments.” 1139

But the metaphors belong to thought, not to
the mere outward form. Language is the expression
of thought, and consequently, though thought
and its expression can scarcely be separated in practice,
yet, like the seal and its impression, they are
really distinct, or rather, while thought can exist
without expression, expression cannot exist without
thought. The outward presupposes the inward, the
impression presupposes the seal. The influence of
analogy, therefore, under which words change their
significations, is exerted upon the ideas they
denote; the mind first discovers similarities between
its conceptions, and these are then reflected
in speech. The dried grape is called a plum when
put into a cake or pudding, because it looks like
357the fruit of that name, and not for any linguistic
reason. But although it is the content that properly
modifies the form, the form may react upon
the content. In this case, analogy becomes a
truly creative philological principle. The simple
sound of the word itself, its mere outer husk, as it
were, calls up associations which create new sounds,
new ideas, and therefore new words. There arises
an imaginary world, answering to nothing real and
substantial, which stands solely upon the basis of
uttered speech. It is the creation of the external
side of language, and the demiurge is analogy.
There will be an unreal world either of content or
of form. In the first instance, the mind will be
deluded by those false notions, those Baconian
idola, which have done so much to impede progress,
and which are called popular etymologies. In the
second instance, it is only the expression of
thought that is made unreal, only the outward
form of language that is forced and artificial,
and we term it poetry. The two creations spring
from one and the same source, but they represent
two different stages in the growth of the mind.
The mythopœic age is the period of primitive unconscious
childhood and barbarism, and wherever
it still exists it bears witness to a naive unthinking
attitude of mind. Poetry, on the other hand,
is the cultured and conscious expression of thought;
358its artificiality is recognised, and it can accordingly
affect only the outward form. It is the spiritualising
of the material, which it therefore moulds
according to its will; while the etymological
myth is the materialising of the spiritual, which
thus becomes distorted and untrue. I am not now
going to discuss the origin of mythology in general,
and the cause of its continuance; I shall merely
confine myself to those portions of it which are due
to the action of false analogy. A large part of
our Aryan mythology, as has been eloquently
pointed out by Professor Max Müller, is derived
from homonymes and synonymes, from phonetic
decay, and the attempt to explain forgotten words.
Daphne, “the laurel,” and daphne, “the dawn,”
came both from the root dah, “to burn;” what
more natural than that the dawn should be changed
into a laurel in her flight from Apollo, the sungod?
Pramanthas, “the fire-chark” of the Hindu,
became the Προμηθεὺς of the Greek; and the
simple contrivance of the savage for the production
of fire passed into the wise benefactor of man,
with his brother Epimetheus or “afterthought.”
The various tribes with names, sometimes explicable,
sometimes obscure, had to be provided with
eponymous heroes; and the manifold appellations
assigned to the same object of worship were transformed
into as many separate deities. Led by a
359false analogy, men argued that what was different
or alike in name, must also be different or alike
in reality; and so a whole fairyland was built up
upon the mere sound of words. But the influence
of a false analogy went even further back than
this. Before the primitive man had learned to
distinguish between the subject and the object, the
actions and passions of the thinker were transferred
to the inanimate. So the sun was compared
to a charioteer or a one-eyed monster, and the
thunder was the voice of God. The similes in all
these cases, however, were still between ideas, not
words — still belonged to thought, not to its expression;
but when they had once been enshrined
in language, they tended to grow and multiply,
and the starting-point was no longer the inability
of the child to distinguish between himself and the
object, but the mere verbal metaphors themselves.
As the primitive state of mind passed away, the
original meaning intended to be conveyed in the
traditional phrases was forgotten; the myth became
purely etymological, and the solar charioteer was
transformed into Phaethon, and the one-eyed monster
into the Kyklops. The simple words, divested
of their real signification, were associated with,
others which represented intelligible notions to the
users of them; and out of these false analogies
grew up the fantastic shapes of many a legend
360and myth. The human mind cannot be satisfied
unless it can assign some reason for the existence
of a thing, unless it can believe that it understands
it. So long as an explanation is not forthcoming,
it feels itself in the presence of something mysterious
and supernatural, and this causes all the
discomforts of fear and uncertainty. The explanation
may be very far indeed from the truth; but
so long as any can be given, the man is content.
Now in order to explain we must compare; it is
only by bringing a phenomenon within the limits
of the known that we take it out of the region of
the inexplicable. Hence come all those popular
etymologies which interpret unknown terms by
words of the same or similar sound. The eponymous
heroes, already alluded to, who have been
manufactured out of the names of tribes and
places, are a case in point; here the attempt to
assign definite conceptions to the words themselves
has been despaired of, and they have accordingly
been explained by what seemed the analogous
instances of words without any signification of
their own, but which served to denote individuals.
Proper names have naturally been the special subject
of popular etymologising; there is nothing
else in language which so quickly and thoroughly
changes its form; and yet, since everything must
have a reason, the assumption is irresistible that
361they once had a meaning. Thus, as my friend
the Rev. J. Earle tells me, there are two neighbouring
places in Somersetshire called Saltford
and Freshford. The first was originally Sal-ford
(Sallow-ford), the “Willow-ford;” but when the
Saxon Salh (Salig) died out of use, a slight change
of pronunciation altered the unintelligible Salford
into the intelligible Saltford, and this in its turn
created the corresponding Freshford. In modern
Greece, again, we see the same process taking
place. Thus Athens is Ἀνθῆναι, “the flowery,”
in the mouths of the common people; Krisa is
Χρυσό, “the golden;” and a legend of a quarrel
between two brothers has fastened itself upon
Delphi. 1140 Not very unlike is the superstitious
feeling which has transmuted the forms of words
of ill-omened sound. Because an unfortunate
event was called malum, and a fortunate one bonum,
it was thought that the words themselves brought
good and evil; and so Maleventum was changed
into Beneventum, just as the Erinyes were addressed
as the Eumenides, and the left hand as “the
better;” or as, in modern times, the Cape of
Storms has become the Cape of Good Hope, and
the custom of Ta-pu is perpetually transforming
the languages of the Pacific islanders. But other
words besides proper names lose their true form
362and meaning through the influence of imaginary
analogies. Thus the German sündfluth, “great
flood,” has had the first syllable assimilated to
sünde, “sin,” through its application to the Biblical
Deluge; and even philosophy has been deceived by
the outward resemblance of the logical copula to
the substantive verb, while Bacon believed in the
existence of sensible qualities answering to the
abstract nouns derived from attributive adjectives.

Thus far this popular etymologising is unconscious
and instinctive. But it becomes more or
less artificial in cases like those which meet us in
the Homeric poems. Thus the old epic adjective
ἐπηέτανος, “long-lasting,” from ἐπὶ, ἄει, and
τείνω, came to be thought to be derived from ἔτος
(Ϝετος, Sansk. vatsas), “a year,” and accordingly
is used in the sense of “lasting for a year” more
than once in the Odyssey; while πλέες, “full,” was
fancied to be a contracted form of πλείονες, “more,”
through the false analogy of the Ionic πλεῦν, πλεῖν,
for πλέον (that is, πλεῖον), and hence we find the
monstrous solecism οἰωνὸι … πλέες ἠέ γυναίκες,
“more birds than women,” in Il. xi. 395. So,
again, τέλσος, “tilth,” was imagined to be identical
with τέλος, “end” (as in Il. xiii. 707, xviii.
544), and the aorist infinitives χραισμεῖν, Ϝιδεῖν,
handed down in various formulæ and stereotyped
verses, were believed to be presents, and accordingly
363provided with the futures χραισμήσω and
ἰδήσω. The Odyssey even goes a step further
than mere unconscious misunderstanding of the
traditional language of the past; and the affectation
of archaism observable in it, which ignores,
for instance, the existence of writing, even to the
extent of making a Phæakian supercargo commit
his freight to memory, renders Spenser's “Fairy
Queen” its best analogue, and results in a list of
similar etymological mistakes. 1141 A good parallel
to this appearance of popular etymologising in
literature may be found in some of the eccentricities
of modern English spelling. Further, the
comparative of forth, has thus been spelt and pronounced
farther, under the impression that it was
derived from far, the th being euphonic or something
of that kind; and this erroneous etymology
has reacted upon the meaning of the word. Whole,
again, the by-form of hale (the Greek καλός), has
been garnished with a w, through the supposed
analogy of wheel and whale, 2142 and could, from can,
has received an l because should, from shall, has
364one; and all this in defiance of pronunciation.
The authority of Webster's Dictionary has induced
our American cousins to get rid of the unnecessary
u in words like honor, favor; but instead of confining
themselves to vocables of Latin origin, they
have extended the practice to totally different
cases, like harbour and neighbour. It is only the
philologist who knows the deceitfulness of the
analogy.

But the influence of analogy upon the written
form of words leads us naturally to the consideration
of the way in which it has acted upon what is
pre-eminently the cultured expression of thought.
Poetry is artificial language; it is the endeavour
to express the best ideas in the best manner possible,
and, after the form has thus been elaborated,
to secure it by artificial means from being forgotten.
Hence come all the various contrivances of
metrical feet, of parallelism, of alliteration, and of
rhyme. Metrical feet, whether quantitative or
qualitative (that is accentual), is the satisfaction
of the striving after analogical harmony, of the
desire of mind and ear and lips for the like, which
lies at the bottom of speech itself. In the parallelism
of Semitic poetry material analogy passes
into conceptual analogy; clause answers to clause,
stanza to stanza. But the oldest device invented
to imprint upon speech the poetical form, and to
365enable the memory to retain it, is alliteration.
Here the natural tendency to repeat the same
sound or combination of sounds comes into fall
play; and so alliteration is the essential characteristic
of barbarian poetry all over the world, from
the Kalewala of the Finns to the songs of the
North American Indians. It has been an especial
favourite with the Teutonic race; all our old English
poetry is alliterative, and to this day the
musicalness which we recognise in the verses of
certain poets is due to this cause. Alliteration
may be said to belong to the beginnings of words,
and so to correspond with rhyme, which affects the
ends of words. Rhyme is the fulfilment of the
expectation of discovering an analogy between final
sounds. It may be detected here and there in the
poetry of most nations; examples of it, for instance,
are to be met with in the Old Testament,
and the charm of the Latin pentameter is enhanced
by the rhyming of the last syllables of the two
penthemimers. But it has attained its fullest
development in modern European poetry. According
to Nigra its origin is Keltic; but however
that may be, the Romance languages, such as Provençal
and Italian, with their words terminating
in the same sounds — the worn remnants of Latin
flection — seemed created for the application of
rhyme. No doubt the Latin poems of the Middle
366Ages, in which a jingle had to take the place of
forgotten quantity, helped considerably towards
the same end. The child of the South, rhyme was
soon transplanted to the North, and became a
successful rival of alliteration in Teutonic poetry.
But it could never win the same influence in languages
which abounded in monosyllabic words as
it had in the Neo-Latin dialects, for the simple
reason that such an ornamental help to the memory
ought rather to affect unmeaning and merely
euphonic final syllables than words every letter of
which is instinct with life and signification. Hence
the strong hold that alliteration still has upon our
taste; hence the fact that our greatest poems have
been written in blank verse; hence, too, our preference
for double rhymes, our dislike to a perpetual
rhyming of monosyllables.

With poetry, the highest effort of the human
mind consciously to shape and fashion language,
we must close our review of the influence of analogy.
I have tried to show how immense is its
power throughout the whole domain of speech, and
how it is present everywhere as a creative and
reconstructing principle. Whether the analogy
be true or false, whether it act upon the matter or
the form, is of little consequence. Phonology,
accidence, syntax, and signification are all equally
affected by it; while the poetry of the people,
367which is based upon unreal ideas — ideas, that is,
which have nothing actual and objective answering
to them — is not less the product of its ceaseless
activity than the poetry of literature, where
the form alone is unreal and artificial, a language
never spoken in the work-a-day world.368