Müller, Friedrich. Lectures on the Science of Language. Second Series – T04

Lecture IV.
Phonetic Change.

From the investigations which I laid before you in
my last Lecture, you know the materials which were
at the disposal of the primitive architects of language.
They may seem small compared with the countless
vocables of the countless languages and dialects to
which they have given rise, nor would it have been
difficult to increase their number considerably, had
we assigned an independent name and position to
every slight variety of sound that can be uttered, or
may be discovered among the various tribes of the
globe. Yet small as is the number of the alphabetic
elements, there are but few languages that avail
themselves of all of them. Where we find very abundant
alphabets, as for instance in Hindustani and
English, different languages have been mixed, each
retaining, for a time, its own phonetic peculiarities.
It is because French is Latin as spoken not only by the
Roman provincials but by the German Franks, that we
find in its dictionary words beginning with h and with
gui. They are due to German throats; they belong to
the Teutonic, not to the Romance alphabet. Thus hair
is to hate; hameau, home; hâter, to haste; déguiser points
to wise, guile to wile, guichet to wicket. It is because
English is Saxon as spoken not only by Saxons, but
likewise by Normans, that we hear in it several sounds
160which do not occur in any other Teutonic dialects.
The sound of u as heard in pure is not a Teutonic
sound. It arose from an attempt to imitate the
French u in pure. *1 Most of the words in which this
sound is heard are of Roman origin, e. g. duke,
during (durer), beauty (beauté, bellitas), nuisance
(nocentia). This sound of w, however, being once
naturalized, found its way into Saxon words also; that
is to say, the Normans pronounced the A. S. eów and
eaw like yu; e. g. knew (cneów), few (feawa), dew
(deáw), hue (hiw). †2

The sounds of ch and j in English are Roman or
Norman rather than Teutonic sounds, though, once
admitted into English, they have infected many words
of Saxon descent. Thus cheer in good cheer is the
French chère, the Mediaeval Latin cara; ‡3 chamber,
chambre, camera) cherry, A. S. cirse, Fr. cerise, Lat.
cerasus; to preach, prêcher, prædicare; forge, fabricare.
Or j in joy, gaudium, judge, judex, &c. But the
same sounds found their way into Saxon words also,
such as choose (ceósan, German kiesen); chew (ceowan,
German kauen); particularly before e and i, but likewise
before other vowels; e. g. child, as early as Layamon,
instead of the older A. S. cild; cheap, A. S. ceap;
birch, finch, speech, much, &c.; thatch (theccan), watch
(weccan); in Scotch, theek and waik; or in bridge
(brycg, Brücke), edge (ecg, Ecke), ridge (hrycg, Rücken).

The soft sound of z in azure or of s in vision is likewise
a Roman importation.161

Words, on the contrary, in which th occurs are Saxon,
and had to be pronounced by the Normans as well as
they could. To judge from the spelling of MSS., they
would seem to have pronounced d instead of th. The
same applies to words containing wh, originally hv, or
ght, originally ht) as in who, which, or bought, light,
right. All these are truly Saxon, and the Scotch
dialect preserves the original guttural sound of h
before t.

The O Tyi-herero has neither l nor f, nor the sibilants
s r z. The pronunciation is lisping, in consequence of
the custom of the Va-herero of having their upper
front teeth partly filed off, and four lower teeth
knocked out. It is perhaps due to this that the O
Tyi-herero has two sounds similar to those of the hard
and soft th and dh in English (written s, z). *4

There are languages that throw away certain letters
which to us would seem almost indispensable, and there
are others in which even the normal distinctions between
guttural, dental, and labial contact are not yet
clearly perceived. We are so accustomed to look
upon pa and ma as the most natural articulations,
that we can hardly imagine a language without them.
We have been told over and over again that the names
for father and mother in all languages are derived
from the first cry of recognition which an infant can
articulate, and that it could at that early age articulate
none but those formed by the mere opening or closing
of the lips. It is a fact, nevertheless, that the Mohawks,
of whom I knew an interesting specimen at
Oxford, never, either as infants or as grown-up people,
articulate with their lips. They have no p, b, m, f, v,
w — no labials of any kind; and although their own
162name Mohawk would seem to bear witness against
this, that name is not a word of their own language,
but was given to them by their neighbours. Nor are
they the only people who always keep their mouths
open and abstain from articulating labials. *5 They
share this peculiarity with five other tribes, who
together form the so-called six nations, Mohawks,
Senekas, Onandagos, Oneidas, Cayugas, and Tuscaroras.
The Hurons likewise have no labials, and
there are other languages in America with a similar
deficiency. †6

The gutturals are seldom absent altogether; in
some, as in the Semitic family, they are most prominent,
and represented by a numerous array of letters.
Several languages do not distinguish between k and
g; some have only k, others g only. The sound of
g as in gone, of j as in jet, and of z as in zone, which
are often heard in Kafir, have no place in the Sechuana
alphabet. ‡7 There are a few dialects mentioned by
Bindseil as entirely destitute of gutturals, for instance,
that of the Society Islands. §8 It was unfortunate
163that one of the first English names which the
natives of these islands had to pronounce was that of
Captain Cook, whom they could only call Tute. Besides
the Tahitian, the Hawaian and Samoan *9 are likewise
said to be without gutturals. In these dialects, however,
the k is indicated by a hiatus or catching of the
breath, as ali'i for alihi, 'a'no for kakano. †10

The dentals seem to exist in every language. ‡11 The d,
however, is never used in Chinese, norm Mexican, Peruvian,
and several other American dialects, §12 and the n
is absent in the language of the Hurons ‖13 and of some
other American tribes. The s is absent in the Australian
dialects ¶14 and in several of the Polynesian languages,
where its place is taken by h. **15 Thus in Tongan
we find hahake for sasake) in the New Zealand dialect
heke for seke. In Rarotongan the s is entirely lost, as
in ae for sae. When the h stands for an original s, it has
a peculiar hissing sound which some have represented
by sh, others by zh, others by he or h; or simply e.
Thus the word hongi, from the Samoan song; meaning
to salute by pressing noses, has been spelt by different
164writers, shongi, ehongi, heongi, Kongi and zongi *16
But even keeping on more familiar ground, we find
that so perfect a language as Sanskrit has no f, no soft
sibilants, no short e and o; Greek has no y, no w, no f,
no soft sibilants; Latin likewise has no soft sibilants,
no ϑ, ϕ, χ English is deficient in guttural breathings
like the German ach and ich. High German
has no w like the English w in wind, no th, dh, ch, j.
While Sanskrit has no f, Arabic has no p. F is absent
not only in those dialects which have no labial articulation
at all, but we look for it in vain in Finnish
(despite of its name, which was given it by its neighbours †17),
in Lithuanian, ‡18 in the Gipsy languages, in
Tamil, Mongolian, some of the Tataric dialects, Burmese,
&c. §19

It is well known that r is felt to be a letter difficult
to pronounce not only by individuals but by whole
nations. No Chinese who speaks the classical language
of the empire ever pronounces that letter. They say
Ki li sse tu instead of Christ; Eulopa instead of
Europe) Ya me li ka instead of America, Hence
neither Mandarin nor Sericum can be Chinese words:
the former is the Sk. mantrin, counsellor; the latter
derived from Seres, a name given to the Chinese by
their neighbours. ‖20 It is likewise absent in the language
of the Hurons, the Mexicans, the Othomi, and
other American dialects; in the Kafir language, ¶21 and
165in several of the Polynesian *22 tongues. In the Polynesian
tongues the name of Christ is Kalaisi, but also
Karaita and Keriso. R frequently alternates with l,
but l again is a sound unknown in Zend, and in the
Cuneiform Inscriptions, †23 in Japanese (at least some
of its dialects) and in several American and African
tongues. ‡24

But what is even more curious than the absence or
presence of certain letters in certain languages or
families of languages, is the inability of some races to
distinguish, either in hearing or speaking, between
some of the most normal letters of our alphabet. No
two consonants would seem to be more distinct than
k and t. Nevertheless, in the language of the Sandwich
Islands these two sounds run into one, and it seems
impossible for a foreigner to say whether what he
hears is a guttural or a dental. The same word is
written by Protestant missionaries with k, by French
missionaries with t. It takes months of patient
labour to teach a Hawaian youth the difference between
k and t, g and d, l and r. The same word
167varies in Hawaian dialects as much as koki and hoi,
kela and tea. *28 In adopting the English word steel,
the Hawaians have rejected the s, because they never
pronounce two consonants together; they have added
a final a, because they never end a syllable with a
consonant, and they have changed t into k. †29 Thus
steel has become kila. Such a confusion between two
prominent consonants like k and t would destroy the
very life of a language like English. The distinction
between carry and tarry, car and tar, key and tea,
neck and net, would be lost. Yet the Hawaian language
struggles successfully against these disadvantages,
and has stood the test of being used for a
translation of the Bible, without being found wanting.
Physiologically we can only account for this confusion
by inefficient articulation, the tongue striking the
palate bluntly half-way between the k and the t points,
and thus producing sometimes more of a dental,
sometimes more of a palatal noise. But it is curious
to observe that, according to high authority, something
of the same kind is supposed to take place in English
and in French. ‡30 We are told by careful observers that
the lower classes in Canada habitually confound t
and k, and say mékier, moikié, for métier and moitié.
Webster goes so far as to maintain, in the Introduction
to his English Dictionary, that in English the
letters cl are pronounced as if written tl; clear, clean,
168he says are pronounced tlear, tlean; gl is pronounced
dl; glory is pronounced dlory. Now Webster is a
great authority on such matters, and although I doubt
whether anyone really says dlory instead of glory, his
remark shows, at all events, that even with a well-mastered
tongue and a well-disciplined ear there is
some difficulty in distinguishing between guttural
and dental contact.

How difficult it is to catch the exact sound of a
foreign language may be seen from the following
anecdote. An American gentleman, long resident in
Constantinople, writes: — ‘There is only one word in
all my letters which I am certain (however they may
be written) of not having spelt wrong, and that is
the word bactshtasch, which signifies a present. I
have heard it so often, and my ear is so accustomed
to the sound, and my tongue to the pronunciation, that
I am now certain I am not wrong the hundredth part
of a whisper or a lisp. There is no other word in the
Turkish so well impressed on my mind, and so well
remembered. Whatever else I have written, bactshtasch!
my earliest acquaintance in the Turkish language,
I shall never forget you.’ The word intended
is Bakhshish, *31

The Chinese word which French scholars spell eul,
is rendered by different writers öl, eulh, eull, r'l, r'll,
urh, rhl. These are all meant, I believe, to represent
the same sound, the sound of a word which at Canton
is pronounced i, in Annamitic ni, in Japanese ni. †32169

If we consider that r is in many languages a
guttural, and l a dental, we may place in the same
category of wavering pronunciation as k and t, the confusion
between these two letters, r and l, a confusion
remarked not only in the Polynesian, but likewise in
the African languages. Speaking of the Setshuana
dialects, Dr. Bleek remarks: ‘One is justified to consider
r in these dialects as a sort of floating letter,
and rather intermediate between l and r, than a
decided r sound.’ *33

Some faint traces of this confusion between r and l
may be discovered even in the classical languages,
though here they are the exception, not the rule.
There can be no doubt that the two Latin derivatives
aris and alis are one and the same. If we derive
Saturnalis from Saturnus, and secularis from seculum,
normalis from norma, regularis from regula, astralis
from astrum, stellaris from stella, it is clear that the
suffix in all is the same. Yet there is some kind of
rule which determines whether alis or aris is to be
preferred. If the body of the words contains an l, the
Roman preferred the termination aris; hence secularis,
regularis, stellaris, the only exceptions being that
l is preserved (1) when there is also an r in the body
of the word, and this r closer to the termination than
the l; hence pluralis, lateralis; (2) when the l forms
part of a compound consonant, as fluvialis, glacialis. †34

Occasional changes of l into r are to be found in
almost every language, e. g. lavender, i.e. lavendula;
colonel, pronounced curnel (Old French, coronel;
Spanish, coronel); rossignole = lusciniola; cœruleus
170from cœlum; kephalargía and lēthargía, but ōtalgía,
all from álgos, pain. The Wallachian dor, desire, is
supposed to be the same word as the Italian duolo,
pain. In apôtre, chapitre, esclandre, the same change
of l into r has taken place. *35

On the other hand r appears as l in Italian albero =
arbor; celebro = cerebrum; mercoledì, Mercurii dies;
pellegrino, pilgrim = pèregrinus; autel = altare. †36

In the Dravidian family of languages the change
of l into r, and more frequently of r into l, is very
common. ‡37

Instances of an utter inability to distinguish between
two articulate sounds are, however, of rare occurrence,
and they are but seldom found in languages
which have received a high amount of literary cultivation.
What I am speaking of here is not merely
change of consonants, one consonant being preferred
in one, another in another dialect, or one being fixed
in one noun, another in another. This is a subject we
shall have to consider presently. What I wished to
point out is more than that; it is a confusion between
two consonants in one and the same language, in one
and the same word. I can only explain it by comparing
it to that kind of colour-blindness when people
are unable to distinguish between blue and red, a
colour-blindness quite distinct from that which makes
blue to seem red, or yellow green. It frequently
happens that individuals are unable to pronounce
certain letters. Many persons cannot pronounce the
l, and say r or even n instead; grass and crouds instead
of glass and clouds; ritten instead of little.
171Others change r to d, dound instead of round; others
change l to d, dong instead of long. Children, too,
for some time substitute dentals for gutturals, speaking
of tat instead of cat, tiss instead of kiss. It is
difficult to say whether their tongue is more at fault
or their ear. In these cases, however, a real substitution
takes place; we who are listening hear one
letter instead of another, but we do not hear as it were
two letters at once, or something between the two.
The only analogy to this remarkable imperfection
peculiar to uncultivated dialects may be discovered in
languages where, as in Modern German, the soft and
hard consonants become almost, if not entirely, undistinguishable.
But there is still a great difference
between actually confounding the places of contact as
the Hawaians do in k and t, and merely confounding
the different efforts with which consonants, belonging
to the same organic class, ought to be uttered, a defect
very common in some parts of Germany and elsewhere.

This confusion between two consonants in the same
dialect is a characteristic, I believe, of the lower stages
of human speech, and reminds us of the absence of
articulation in the lower stages of the animal world.
Quite distinct from this is another process which is
going on in all languages, and in the more highly
developed even more than in the less developed, the
process of phonetic diversification, whether we call it
growth or decay. This process will form the principal
subject of our sixth Lecture, and we shall see
that, if properly defined and understood, it forms the
basis of all scientific etymology.

Wherever we look at language, we find that it
changes. But what makes language change? We
172are considering at present only the outside, the phonetic
body of language, and are not concerned with
the changes of meaning, which, as you know, are sometimes
very violent. At present we only ask, how is it
that one and the same word assumes different forms in
different dialects, and we intentionally apply the name
of dialect not only to Scotch as compared with English,
but to French as compared with Italian, to Latin
as compared with Greek, to Old Irish as compared with
Sanskrit. These are all dialects; they are all members
of the same family, varieties of the same type, and
each variety may, under favouring circumstances,
become a species. How then is it, we ask, that the
numeral four is four in English, quatuor in Latin,
cethir in Old Irish, chatvar in Sanskrit, keturi in
Lithuanian, tettares in Greek, pisyres in Æolic, fidvor
in Gothic, fior in Old High-German, quatre in French,
patru in Wallachian?

Are all these varieties due to accident, or are they
according to law; and, if according to law, how is that
law to be explained?

I shall waste no time, in order to show that these
changes are not the result of mere accident. This
has been proved so many times, that we may, I believe,
take it now for granted.

I shall only quote one passage from the Rev. J. W.
Appleyard's excellent work, ‘The Kafir Language,’
in order to show that even in the changes of languages
sometimes called barbarous and illiterate, law and
order prevail (p. 50) : —

‘The chief difference between Kafir and Sechuana
roots consists in the consonantal changes which they
have undergone, according to the habit or taste of the
respective tribes. None of these changes, however,
173appear to be arbitrary, but, on the contrary, are regulated
by a uniform system of variation. The vowels
are also subject to the same kind of change; and, in
some instances, roots have undergone abbreviation by
the omission of a letter or syllable.’ Then follows a
table of vowel and consonantal changes in Kafir and
Sechuana, after which the author continues : ‘By
comparing the above consonantal changes with § 42, it
will be seen that many of them are between letters of
the same organ, the Kafir preferring the flat sounds
(b, d, g, v, z), and the Sechuana, the sharp ones (p, t,
k, f, s). It will be observed, also, that when the
former are preceded by the nasal m or n, these are
dropped before the latter. There is sometimes, again,
an interchange between dentals and linguals; and
there are, occasionally, other changes which cannot be
so easily accounted for, unless we suppose that intermediate
changes may be found in other dialects….
It will thus be seen that roots which appear totally
different the one from the other, are in fact the very
same, or rather, of the same origin. Thus no one, at
first sight, would imagine that the Sechuana reka and
the Kafir tonga, or the Kafir pila and the Sechuana
tsera, were mere variations of the same root. Yet a
knowledge of the manner in which consonants and
vowels change between the two languages shows that
such is the case. As corroborative of this, it may be
further observed, that one of the consonants in the
above and other Sechuana words sometimes returns
in the process of derivation to the original one, as it
is found in the Kafir root. For example, the reflective
form of reka is iteka, and not ireka; whilst the noun,
which is derived from the verb tsera is botselo, and
not botsero.’174

The change of th into f, is by many people considered
a very violent change, so much so that Burnouf's
ingenious identification

of Thraêtona with
Feridún, of which more
hereafter, was objected to
on that ground. But we
have only to look at the
diagrams of th and f, to
convince ourselves that the
slightest movement of the
lower lip towards the upper
teeth would change the
sound of th into f, *38 so that
in English, ‘nothing’ as pronounced vulgarly, sounds
sometimes like ‘nuffing.’

Few people, if any, would doubt any longer that
the changes of letters take place according to certain
phonetic laws, though scholars may differ as to the
exact application of these laws. But what has not
yet been fully explained is the nature of these phonetic
laws which regulate the changes of words. Why
should letters change? Why should we, in modern
English, say lord instead of hlâford, lady instead of
hlœfdige? Why should the French say père and mère,
instead of pater and mater? I believe the laws
which regulate these changes are entirely based on
physiological grounds, and admit of no other explanation
whatsoever. It is not sufficient to say that l and
r, or d and r, or s and r, or k and t, are interchangeable.
We want to know why they are interchangeable,
175or rather, to use more exact language, we want to
know why the same word, which a Hindu pronounces
with an initial d, is pronounced by a Roman with an
initial l, and so on. It must be possible to explain
this physiologically, and to show, by means of diagrams,
what takes place, when, instead of a d an l,
instead of an f a th is heard.

There are, however, some words in English which,
if compared with their originals in Anglo-Saxon, seem
to have added to their bulk, and thus to violate the
general principle of simplification. Thus A. S. thunor
is in English thunder. Yet here, too, the change is
due to laziness. It requires more exertion to withdraw
the tongue from the teeth without allowing the
opening of the dental contact to be heard than to slur
from n on to d, and then only to the following vowel.
The same expedient was found out by other languages.
Thus, the Greek said ándres, instead of áneres; ambrosia,
instead of amrosia *41 The French genre is more
difficult to pronounce than gendre; hence the English
gender, with its anomalous d. Similar instances in
English are, to slumber = A. S. slumerian; embers =
A. S. œmyrie; cinders = cineres; humble = humilis.

It was the custom of grammarians to ascribe these and
similar changes to euphony, or a desire to make words
agreeable to the ear. Greek, for instance, it was said,
abhors two aspirates at the beginning of two successive
syllables, because the repeated aspiration would offend
delicate ears. If a verb in Greek, beginning with an
aspirate, has to be reduplicated, the first syllable takes
the tenuis instead of the aspirate. Thus thē in Greek
178forms títhēmi, as dhâ in Sanskrit dadhâmi. If this was
done for the sake of euphony, it would be difficult to
account for many words in Greek far more inharmonious
than thíthēmi. Such words as Χθών, chthṓn earth,
φθόγγος, phthóggos vowel, beginning with two aspirates,
were surely more objectionable than thíthēmi would
have been. There is nothing to offend our ears in the
Latin fefelli, *42 from fallo, or in the Gothic reduplicated
perfect haihald, from haldan, which in English is
contracted into held, the A. S. being heóld, instead of
hehold; or even in the Gothic faifahum, we caught,
from fahan, to catch. †43 There is nothing fearful in
the sound of fearful, though both syllables begin with
an f. But if it be objected that all these letters in
Latin and Gothic are mere breaths, while the Greek
χ, ϑ, ϕ are real aspirates, we have in German such
words as Pfropfenzieher, which to German ears is
anything but an unpleasant sound. I believe the
secret of this so-called abhorrence in Greek is nothing
but laziness. An aspirate requires great effort, though
we are hardly aware of it, beginning from the abdominal
179muscles and ending in the muscles that open the
glottis to its widest extent. It was in order to economize
this muscular energy that the tenuis was
substituted for the aspirate, though, of course, in cases
only where it could be done without destroying the
significancy of language. Euphony is a very vague
and unscientific term. Each nation considers its own
language, each tribe its own dialect, euphonic; and
there are but few languages which please our ear when
heard for the first time. To my ear knight does not
sound better than Knecht, though it may do so to an
English ear, but there can be no doubt that it requires
less effort to pronounce the English knight than the
German Knecht.

But from this, the most important class of phonetic
changes, we must distinguish others which arise from
a less intelligible source. When we find that, instead
of Latin pater, the Gothic tribes pronounced fadar, it
would be unfair to charge the Goths with want of
muscular energy. On the contrary, the aspirated f
requires more effort than the mere tenuis; and the d,
which between two vowels was most likely sounded
like the soft th in English, was by no means less
troublesome than the t. Again, if we find in Sanskrit
gharma, heat, with the guttural aspirate, in Greek
thermós with the dental aspirate, in Latin formus,
adj., *44 with the labial aspirate, we cannot charge any
one of these three dialects with effeminacy, but we
must look for another cause that could have produced
these changes. That cause I call Dialectic Growth;
and I feel strongly inclined to ascribe the phonetic
diversity which we observe between Sanskrit, Greek,
180and Latin, to a previous state of language, in which, as
in the Polynesian dialects, the two or three principal
points of consonantal contact were not yet felt as
definitely separated from each other. There is nothing
to show that in thermós, Greek ever had a guttural
initial, and to say that Sanskrit gh becomes Greek th is
in reality saying very little. No letter ever becomes.
People pronounce letters, and they either pronounce
them properly or improperly. If the Greek pronounced
th in thermós properly, without any intention of pronouncing
gh, then the th, instead of gh, requires another
explanation, and I cannot find a better one than the
one just suggested. When we find three dialects, like
Sanskrit, Greek, and Latin, exhibiting the same word
with guttural, dental, and labial initials, we gain but
little if we say that Greek is a modification of Sanskrit,
or Latin of Greek. No Greek ever took the Sanskrit
word and modified it; but all three received it from
a common source, in which its articulation was as yet
so vague as to lend itself to these various interpretations.
Though we do not find in Greek the same
confusion between guttural and dental contact which
exists in the Hawaian language, it is by no means
uncommon to find one Greek dialect preferring the
dental *45 when another prefers the guttural; nor do I see
how this fact could be explained unless we assume that
in an earlier state of the Greek dialects the pronunciation
fluctuated or hesitated between k and t. ‘No Polynesian
dialect,’ says Mr. Hale, ‘makes any distinction
between the sounds of b and p, d and t, g and k, l and
r, or v and w. The l, moreover, is frequently sounded
181like d, and t like k.’ *46 If colonies started to-morrow
from the Hawaian Islands, the same which took place
thousands of years ago, when the Hindus, Greeks, and
Romans left their common home, would take place
again. One colony would elaborate the indistinct, half-guttural,
half-dental articulation of their ancestors into
a pure guttural; another into a pure dental; a third
into a labial. The Romans who settled in Dacia,
where their language still lives in the modern Wallachian,
are said to have changed every qu, if followed
by a, into p. They pronounce aqua as apa; equa as
epa. †47 Are we to suppose that the Italian colonists
of Dacia said aqua as long as they stayed on Italian
soil, and changed aqua into apa as soon as they
reached the Danube? Or may we not rather appeal
to the fragments of the ancient dialects of Italy, as
preserved in the Oscan and Umbrian inscriptions,
which show that in different parts of Italy certain
words were from the beginning fixed differently, thus
justifying the assumption that the legions which
settled in Dacia came from localities in which these
Latin qu's had always been pronounced as p's? ‡48 It
will sound to classical scholars almost like blasphemy
to explain the phenomena in the language of Homer
and Horace, by supposing for both a background like
that of the Polynesian dialects of the present day.
Comparative philologists, too, will rather admit what
182is called a degeneracy of gutturals sinking down to
dentals and labials, than look for analogies to the
Sandwich Islands. Yet the most important point is,
that we should have clear conceptions of the words
we are using, and I confess that, without certain attenuating
circumstances, I cannot conceive of a real k
degenerating into a t or p. I can conceive different
definite sounds arising out of one indefinite sound; and
those who have visited the Polynesian islands describe
the fact as taking place at the present day. What then
takes place to-day can have taken place thousands of
years ago; and if we see the same word beginning in
Sanskrit, Greek, and Latin, with k, t, or p, it would
be sheer timidity to shrink from the conclusion that
there was a time in which that word was pronounced
less distinctly; in short, in the same manner as the k
and t in Hawaian.

There is, no doubt, this other point to be considered,
that each man has his phonetic idiosyncrasies, and that
what holds good of individuals, holds good of families,
tribes, and nations. We saw that individuals and
whole nations are destitute of certain consonants, and
this defect is generally made up on the other hand by a
decided predilection for some other class of consonants.
The West Africans, being poor in dentals and labials,
are rich in gutturals. Now if an individual, or a family,
or a tribe cannot pronounce a certain letter, nothing
remains but to substitute some other letter as nearly
allied to it as possible. The Romans were destitute
of a dental aspirate like the th of the Greeks, or the dh
of the Hindus. Hence, where that letter existed in
the language of their common ancestors, the Romans
had either to give up the aspiration and pronounce d,
or to take the nearest consonantal contact and pronounce
183f. Hence fumus instead of Sk. dhûma,
Greek thýmos. It is exactly the same as what took
place in English. The modern English pronunciation,
owing, no doubt, to Norman influences, lost the guttural
ch, as heard in the German lachen. The Saxons
had it, and wrote and pronounced hleahtor. It is now
replaced by the corresponding labial letter, namely, f,
thus giving us laughter for hleahtor, enough for genug,
&c. If we find one tribe pronounce r, the other l, *49 we
can hardly accuse either of effeminacy, but must appeal
to some phonetic idiosyncrasy, something in fact
corresponding to what is called colour-blindness in
another organ of sense. These idiosyncrasies have to
be carefully studied, for each language has its own,
and it would by no means follow that because a Latin
f or even b corresponds to a Sanskrit dh, therefore
every dh in every language may lapse into f and b.
Greek has a strong objection to words ending in consonants;
in fact, it allows but three consonants, and all
of them semi-vowels, to be heard as finals. We only find
n, r, and s, seldom k, ending Greek words. The Roman
had no such scruples. His words end with a guttural
tenuis, such as hic, nunc; with a dental tenuis, such
as sunt, est; and he only avoids a final labial tenuis
which certainly is not melodious. We can hardly
imagine Virgil, in his hexameters, uttering such
words as lump, trump, or stump. Such tendencies or
dispositions, peculiar to each nation, must exercise
considerable influence on the phonetic structure of a
language, particularly if we consider that in the Aryan
family the grammatical life-blood throbs chiefly in the
final letters.

Words beginning with more than one consonant
are most liable to phonetic corruption. It certainly
requires an effort to pronounce distinctly two or three
consonants at the beginning without intervening
vowels, and we could easily understand that one of
these consonants should be slurred over and be
allowed to drop. But if it is the tendency of
language to facilitate pronunciation, we must not
shirk the question how it came to pass that such
troublesome forms were ever framed and sanctioned.
186Strange as it may seem, I believe that these troublesome
words, with their consonantal exuberances, are
likewise the result of phonetic corruption, i. e. of
muscular relaxation. Most of them owe their origin
to contraction, that is to say, to an attempt to pronounce
two syllables as one, and thus to save time
and breath, though not without paying for it by an
increased consonantal effort.

It has been argued, with some plausibility, that
language in its original state, of which, unfortunately,
we know next to nothing, eschewed the contact of
two or more consonants. There are languages still in
existence in which each syllable consists either of a
vowel or of a vowel preceded by one consonant only,
and in which no syllable ever ends in a consonant.
This is the case, for instance, in the Polynesian languages.
A Hawaian finds it almost impossible to
pronounce two consonants together, and in learning
English he has the greatest difficulty in pronouncing
cab, or any other word ending in a consonant. Cab,
as pronounced by a Hawaian, becomes caba. Mr.
Hale, in his excellent ‘Polynesian Grammar,’ *51 says,
‘In all the Polynesian dialects every syllable must terminate
in a vowel; and two consonants are never
heard without a vowel between them. This rule
admits of no exception whatever, and it is chiefly to
this peculiarity that the softness of these languages
is to be attributed. The longest syllables have only
three letters, a consonant and a diphthong, and many
syllables consist of a single vowel.’

There are other languages besides the Polynesian
which never admit closed syllables, i.e. syllables ending
187in consonants. All syllables in Chinese are open or
nasal, *52 yet it is by no means certain whether the final
consonants which have been pointed out in the vulgar
dialects of China are to be considered as later
additions, or whether they do not represent a more
primitive state of the Chinese language.

In South Africa all the members of the great family
of speech, called by Dr. Bleek the Ba-ntu family,
agree in general with regard to the simplicity of their
syllables. Their syllables can begin with only one consonant
(including, however, consonantal diphthongs,
nasalised consonants, and combinations of clicks with
other consonants reckoned for this purpose as substantially
simple). The semivowel w, too, may intervene
between a consonant and a following vowel.
No syllable, as a general rule, in these South African
languages, which extend north beyond the Equator,
can end in a consonant, but only in vowels, whether
pure or nasal. †53 The exceptions serve but to prove
the rule, for they are confined to cases where by
the falling off of the generally extremely short and
almost indistinct terminal vowel, an approach has
been made to consonantal endings. ‡54

In the other family of South African speech, the
Hottentot, compound consonants are equally eschewed
at the beginning of words. It is clear, too, that all
radical words ended there originally in vowels, and that
the final consonants are entirely due to grammatical
terminations, such as p, s, ts, and r. By the frequent
188use of these suffixes the final vowel disappeared, but
that it was there originally has been proved with
sufficient evidence. *55

If we look to the Finnish and the whole Uralic
class of the Northern Turanian languages, we meet
with the same disinclination to admit double consonants
at the beginning, or any consonants whatever
at the end of words. The German Glas is written
lasi in Finnish. The Swedish smak is changed into
189maku, stor into suuri, strand into ranta. No genuine
Finnish word begins with a double consonant, for the
assibilated and softened consonants, which are spelt as
double letters, were originally simple sounds. This
applies equally to the languages of the Esths, Ostiaks,
Hungarians, and Sirianes, though, through their
intercourse with Aryan nations, these tribes, and even
the Finns, succeeded in mastering such difficult
groups as pr, sp, st, str, &c. The Lapp, the Mordvinian,
and Tcheremissian dialects show, even in words
which are of native growth, though absent in the
cognate dialects, initial consonantal groups such as
kr, ps, st, &c.; but such groups are always the result
of secondary formation, as has been fully proved by
Professor Boiler. *58 The same careful scholar has
shown that the Finnish, though preferring syllables
ending in vowels, has admitted n, s, l, r, and even t, as
final consonants. The Esthonian, Lapp, Mordvinian,
Ostiakian, and Hungarian, by dropping or weakening
their final and unaccented vowels, have acquired a
large number of words ending in simple and double
consonants; but throughout the Uralic class, wherever
we can trace the radical elements of language, we
always find simple consonants and final vowels.

We arrive at the same result, if we examine the
syllabic structure of the Dravidian class of the South
Turanian languages, the Tamil, Telugu, Canarese,
Malayalam, &c. The Rev. R. Caldwell, in his excellent
work, the ‘Dravidian Comparative Grammar,’ has
190treated this subject with the same care as Professor
Boiler in his. Essay on the Finnish languages, and we
have only to place these accounts by the side of each
other, in order to perceive the extraordinary coincidences.

‘The chief peculiarity of Dravidian syllabation is
its extreme simplicity and dislike of compound or
concurrent consonants; and this peculiarity characterizes
the Tamil, the most early cultivated member
of the family, in a more marked degree than any other
Dravidian language.

In Telugu, Canarese, and Malayalam, the great
majority of Dravidian words, i.e. words which have
not been derived from Sanskrit, or altered through
Sanskrit influences, and in Tamil all words without
exception, including even Sanskrit derivatives, are
divided into syllables on the following plan. Double
or treble consonants at the beginning of syllables, like
“str,” in “strength,” are altogether inadmissible. At
the beginning not only of the first syllable of every
word, but also of every succeeding syllable, only one
consonant is allowed. If, in the middle of a word of
several syllables, one syllable ends with a consonant
and the succeeding one commences with another consonant,
the concurrent consonants must be euphonically
assimilated, or else a vowel must be inserted
between them. At the conclusion of a word, double
and treble consonants, like “gth,” in “strength,” are
as inadmissible as at the beginning; and every word
must terminate in Telugu and Canarese in a vowel;
in Tamil, either in a vowel or in a single semivowel,
as “l,” or “r,” or in a single nasal, as “n,” or “m.” It
is obvious that this plan of syllabation is extremely
unlike that of the Sanskrit.191

Generally, “i” is the vowel which is used for the
purpose of separating inadmissible consonants, as
appears from the manner in which Sanskrit derivatives
are Tamilized. Sometimes “u” is employed instead
of “i.” Thus the Sanskrit preposition “pra” is
changed into “pira” in the compound derivatives,
which have been borrowed by the Tamil; whilst
“Krishna” becomes “Kiruttina-n” (“ṭṭ,” instead of
“sh,”), or even “Kiṭṭina-n.” Even such soft conjunctions
of consonants as the Sanskrit “dya,” “dva,”
“gya,” &c., are separated in Tamil into “diya,”
“diva,” and “giya.”’ *59

It is hardly to be wondered at that evidence of this
kind, which might be considerably increased, should
have induced speculative scholars to look upon the
original elements of language as necessarily consisting
of open syllables, of one consonant followed by one
vowel, or of a single vowel. The fact that languages
exist, in which this simple structure has been preserved,
is certainly important, nor can it be denied,
that out of such simple elements languages have been
formed, gradually advancing, by a suppression of
vowels, to a state of strong consonantal harshness.
The Tcheremissian 'sma, mouth, if derived from a root
śu, to speak, must originally have been śuma.

In the Aryan languages, the same process can easily
be observed as producing the same effect, viz., double
consonants, either at, the beginning or at the end of
words. It was in order to expedite the pronunciation
of words that vowels were dropt, and consonants
brought together: it was to facilitate the pronunciation
of such words that one of the consonants was
192afterwards left out, and new vowels were added
to render the pronunciation easier once more.

Thus, to know points back to Sk. jnâ, but this jnâ,
the Lat. gnô in gnôvi, or gnō in Gr. égnōn, again points
back to janâ, contracted to jnâ. Many roots are
formed by the same process, and they generally
express a derivative idea. Thus jan, which means to
create, to produce, and which we find in Sk. janas, Gr.
génos, genus, kin, is raised to jnâ, in order to express the
idea of being able to produce. If I am able to produce
music, I know music; if I am able to produce ploughing,
I know how to plough, I can plough; and hence
the frequent running together of the two conceptions,
I can and I know, Ich kann and Ich kenne. *60 As from
jan we have jnâ, so from man, to think (Sk. manas,
Gr. ménos, mens, mind), we have mnâ, to learn by
heart, Greek mémnēmai, I remember, mimnḗskō. In
modern pronunciation the m is dropt, and we pronounce
m-nemonics. Again, we have in Sanskrit a root mlai,
which means to fade; from it mlâna, faded, mlâni,
fading. The Teutonic nations, avoiding the complete
labial contact that is required for m, were satisfied with
the labial approach which produces w, and thus pronounced
ml like vl. Hence A. S. wlœc, tired, wlacian,
to be tired, to flag. The Latin has flaccus, withered,
flabby, where we should expect blaccus, Germ. welk. In
German we have flau, †61 weak, and what seems to be
merely a dialectic Low German variety, lau, in the
sense of luke-warm, i.e. water that is but weakly
193boiling. Now, whence this initial double consonant
ml, which in German meets with the usual fate of
most double initial consonants, and from ml sinks to l?
The Sanskrit root mlai or mlâ is formed like jnâ and
mnâ, from a simpler root mal or mar, which means to
wear out, to decay. As jan became jnâ, so mar, mrâ.
This mar is a very prolific root, of which more hereafter,
and was chiefly used in the sense of decaying
or dying, morior, άμ(β)ρόσια, Old Slav, mrĕti, to die,
Lith. mirti, to die.

These instances must suffice in order to show
that in Sanskrit, too, and in the Aryan languages
in general, the initial double consonants owe their
existence to the same tendency which afterwards
leads to their extinction. It was phonetic economy
that reduced marâ to mrâ; it was phonetic economy
that reduced mrâ to râ and lâ.

The double consonants being once there, the
simplest process would seem to drop one of the
two. This happens frequently, but by no means
always. We see this process in English words like
knight, (h)ring, &c.; we likewise observe it in Latin
natus instead of gnatus, nodus instead of gnodus, English
knot. We know that the old Latin form of locus
was stlocus *62 thus pointing to root stâ, whence the
German Stelle; we know that instead of lis, litis, quarrel,
litigation, the ancient Romans pronounced, stlis, which
points to German streit. In all these cases the first
consonant or consonants were simply dropt. But it
also happens that the double consonant, which was
tolerated at first, only because it was the saving of a
syllable, is lengthened again into two syllables, the
194two syllables seeming to require less effort than the
double consonant. The Semitic languages are quite free
from words beginning with two consonants without an
intermediate vowel or shewa. This is, in fact, considered
by Ewald as one of the prominent characters of
the Semitic family; *63 and if foreign words like Plato
have to be naturalized in Arabic, the p has to be
changed to f, for Arabic, as we saw, has no p, and an
initial vowel must be added, thus changing Plato into
Iflatún. We saw that the Hawaians, in adopting a
word like steel, had to give up the initial s before the
t, pronouncing tila or kila. We saw that the West
African languages met the same difficulty by making
two syllables instead of one, and saying suku instead
of school. The Chinese, in order to pronounce Christ,
have to change that name into Ki-li-sse-tu †64 four syllables
instead of one. There are analogous cases nearer
home. Many words in Latin begin with sc, st, sp.
Some of these are found in Latin inscriptions of the
fourth century after Christ spelt with an initial i: e. g.
in istatuam (Orelli, 1, 120, a. d. 375); Ispiritus (Mai,
Coll. Vat., t. v. p. 446, 8). ‡65 It seems that the Celtic
nations were unable to pronounce an initial s before
a consonant, or at least that they disliked it. §66 The
195Spaniards in Peru, even when reading Latin, pronounce
estudium for studium, eschola for schola *67
Hence the constant addition of the initial vowel in
the Western or chiefly Celtic branch of the Romance
family; French escabeau, instead of Latin scabellum;
estame (étaim), Latin stamen; espérer, instead of Latin
sperare. Then again, as it were to revenge itself for
the additional trouble caused by the initial double
consonant, the French language throws away the s,
which had occasioned the addition of the initial e,
but keeps the vowel which, after the loss of the s,
would no longer be wanted. Thus spada became espée,
lastly épée; scala became eschelle, lastly échelle. Stabilire
became establir, lastly établir, to stablish. †68

Now it must be clear that all these changes rest on
principles totally distinct from those which made the
Romans pronounce the same word as quatuor which
we pronounce four. The transition from Gothic fidvor
to English four may properly be ascribed to phonetic
corruption, but quatuor and fidvor together can only
be explained as the result of dialectic variation. If
we compare quatuor, téssares, písyres, and fidvor, we
find a change of guttural, dental, and labial contact in
one and the same word. There is nothing to show that
the Greek changed the guttural into the dental contact,
or that the Teutonic nations considered the labial contact
less difficult than the guttural and dental. We
196cannot show that in Greece the guttural dwindles down
to a dental, or that in German the labial is later, in
chronological order, than the guttural. We must look
upon guttural, dental, and labial as three different
phonetic expressions of the same general conception,
not as corruptions of one definite original type. The
guttural tenuis once fixed in any language or dialect
does not in that dialect slowly dwindle down to a
dental tenuis; a dental tenuis once clearly pronounced
as a dental does not in the mouth of the same speaker
glide into a labial tenuis. That which is not yet
individualized may grow and break forth in many
different forms; that which has become individual
and definite loses its capability of unbounded development,
and its changes assume a downward tendency
and must be considered as decay. To say where
growth ends and decay begins is as difficult in living
languages as in living bodies; but we have in the
science of language this test, that changes produced
by phonetic decay must admit of a simple physiological
explanation — they must be referable to a
relaxation of muscular energy in the organs of
speech. Not so the dialectic varieties. Their causes,
if they can be traced at all, are special, not general,
and in many cases they baffle all attempts at physiological
elucidation.197