be the penultimate, but only in versification.

It is sometimes asked why we continue to write
what is no longer pronounced. Simply in order not
to take away all the rhythm from the poetry of
Corneille, Racine, La Fontaine, Andr^ Chenier, De
Musset, and Victor Hugo. This is to me an all-
suflScient reason, and renders me very sceptical about
the proposed reforms in spelling. There are others.
In many cases the e mute could not be removed with-



Digitized by VjOOQIC



Indo-European Phonetics. 363

out being replaced by an apostrophe, or how should
we distinguish charmant and charmante, or Sem, the
son of Noah, from sdme ? For the rest, French spelling,
which I am far from defending, be it observed, is
capable of defending itself by the mere force of inertia,
and if it allows an occasional useless letter to be
dropped, it will never, being the result of long use, of
ancient custom, yield all at once to the laws of a
narrow reason. Spelling reform is advocated with
excellent sense, but it is none the less an empty
hobby.

There remains one vowel to be considered, a doubt-
ful vowel, employed only by Sanscrit, but it is at
any rate one of the continued sounds which hold the
debatable land between the vowels and the consonants.
This is the liquid r, doubled with an I in Sanscrit,
which is also regarded as a vowel. Most philologists
consider these letters to be the contraction of a syllable
ar, m, ere ; they point out that in all Indo-European
languages r is replaced by one of these groups, and
that even in Sanscrit it generally disappears from
most of the cases of the nouns in tr : pitri, pitaras ;
mdtri, mdtaram ; and that in hihhrmas it takes the
place of ara : babharamas. Yet, as there are several
manners of pronouncing r, it is probable that this r
answered to something particular. iZ, says the master
of M. Jourdain, is pronounced by " carrying the tip of
the tongue to the roof of the palate ; so that, being
pressed by the air as it is emitted with force, it yields,
and returns always to the same place, producing a
species of trembling ?•, rrra" This r is the dental r
of singers, who learn to pronounce it by repeating
td, td. But there is also a guttural r, very difficult
to pronounce at the end of words, soir, glaire, art,



Digitized by VjOOQIC



364 The Indo-European Organism.

yiqueur^ monsieur (whence monsieu and piquetiai).
The r which figures as a vowel in the Sanscrit alphabet
appears to be closely related to a ; it is dropped from
the nominative of the nouns in tr : pita, mata, bhrafa,
and only reappears when a vowel follows it. It is
represented in Prakrit by an i, and not by the group
ar : hidaya, heart, for hrdaya.

In any case, the vowel r corresponds in Greek and
Latin to er, or, ar, ur, and ra, Bhrtas, carried, is the
Greek cfyepro^ (in af^eproi) ; drshtha, seen, Sepicrog (in
aSepKTo^); str-na-mi, (TTopvvjun, I extend] mr^os, death,
fipoTO^ for /uLoproi ; rksha, bear, apKTo^ ; jakrt, fprap,
liver; pitrshu, iraTpda-i, Latin shows fertis, fertilisy
sterno, mortuus, jecur ; vermis for krmisy cord for hrdj
7nordeo for mrd, to crush, stratus for strtas.

The terminations in tr, tar, which abound in Sans-
crit, play a great part in Greek and Latin. Greek
has them under the two forms rrjp, rcop, long and
short. SoTYip, giver ; ^orripy shepherd ; oivoTromip,
wine-drinker; pTjrij p a,nd prirwpi /uLvrjcrrrip, hTidegroom;
OTTTYip^ spy ; a(ppYiTwp, inhuman (not a brother) ;
/jirjTpoTraTcop, maternal grandfather ; icrrcop, witness
(whence historian) ; aXeKTwp, cock, and jjXe#cT<»^, the
sun; a(priT(ap, archer; aXd(TT(jop, avenger. Latin has
comparatively few words in ter: /rater, arbiter, magister;
but the number of words in tor, such as dator and
stator, monitor, hellator, pastor, qiuBstor, is considerable ;
from these proceed, as we have seen, the future par-
ticiple, faturus, daturus, natura, and the desideratives,
pasturire, esurire, emptuire, &c.

Zend and Persian are also without the vowel r, and
present the same phenomenon as our two classical
languages. In the inscriptions of Persepolis, harta,
karta, correspond to hhrta and krta, Tarsno is trshna,



Digitized by VjOOQIC



Indo-European Phonetics, 365

thirst ; 'parst is prshtha^ the back. Zend prefers ere,
are, Svkrta, well made, becomes hto-kereta, hu-kareta ;
hhrta, hereta ; Vrtra, Verethra ; prastrta, extended,
fragtareta ; hrdaya, heart, zaredaya ; prthu, wide (Gr.
irXarv^), perethu, whence Superethti, ffufratu, the
Euphrates, the wide river.

The vowel / was invented by the grammarians of
India for symmetry ; it exists only in one Vedic word,
kip, to succeed (perfect tchaUpe), participle Mpta, well
made, constituted ; whence halpa, rule, order, one of
the great periods of the Brah manic universe. This
single root, AZp, will account for one of our most useful
and most frequently employed words.

Since the two liquids I and r were not clearly dis-
tinguished the one from the other before the separation
of the idioms, we have more chance of finding the
root kip under the form karp or kerep in the Indo-
European languages, and the Veda itself presents it
to us under this form : uttishtha daivyd krpa, " arise
in thy divine form." What is this 'p joined to kr ?
Probably the remains of a suiBx of causation, pa ; kr
(creare), to do ; krpa, to fashion ; whence the Zend
hU'kerep ta, well made, beautiful ; karefs or kerefs, the
body, genitive kehrpo, kehrpamy accusative kehrpem.
Old High German has hi^ef, Anglo-Saxon hrif^ the
body, the womb (cf. mid-rj/*). We have here the
original of the Latin corpus, an isolated word, which
has no allies in Greek, nor even in Latin, and which
has come down to us with its numerous derivatives,
corporeal, incorporate, coiporation, corpvscle, &c.

The Indo-European vowel system may be represented
by a triangle, of which a, short and long, occupies the
summit, i and ou the two lower angles ; beside the two
last are placed the two diphthongs ai and au, which



Digitized by VjOOQIC



366 The Indo- European Organism.

have given to Sanscrit' its long ^ and 6, The short a
has furnished to all the idioms, except Sanscrit and
Gothic, the short e and o ; from the long a arose the
long S and 6 of all the Western languages. Side by
side with the primitive i and ou, a secondary i and ou
were developed as weaker forms of the a, by the inter-
mediary of the short e and o; a sound u (upsilon)
served as link between ou and i. Two semi-vowels,
the labial t?, the dento-guttural y or j, issued from ou
and i; and perhaps a liquid r was the result of a
guttural pronunciation of the a. Finally, a naso-dental
closely connected with i, e, and ou, modified from the
beginning the timbre of the vowels. The few examples
I have been able to give will have shown what Greek,
Latin, and Zend owe to the various uses of these
resources.

Letters, says Moliere, are divided into vowels, so
called, because they express the voice ; and consonants,
because these sound with the vowels and serve to mark
the various articulations of the voice. But the science
of phonetics brings an amendment to this summary
definition ; it points out that the vowels, or pure con-
tinuous sounds, are connected with the explosives by
mixed sounds, which may be at once and by turns
continuous and explosive, the semi-vowels, liquids, and
spirants, which partake of both natures.

In dealing with the vowel i, we have seen that in
Greek and Latin, in the middle of a word, it frequently
represented the sound y (ie), which belongs to both
vowels and consonants. Distinctly retained at the
beginning of words in Zend, Latin, and Teutonic, this
y has undergone various alterations in Greek which
have changed it, now into an aspirate, now into a soft
sibilant, now into a dental. Finally it became con-



Digitized by VjOOQIC



hido'European Phonetics. 367

founded with a softened, palatal guttural, gemns, genre,
and took in French the value and the sound of/ These
changes are partly determined by the proximity of
certain soft consonants; but some can only be explained
by a defect of pronunciation proper to the Greeks.
Where Sanscrit has yuhsma, we, Greek has vcrjj.e^,
v/uL/uL€^ (classical v/meig) ; yas, yd, yat, the relative pro-
noun, becomes 09, 17, o; the root yog, to sacrifice, to
sanctify, whence the Persians have derived Yaxata
(Ized), the name of the divinities who form the escort
to Ahuramazda, furnishes to Sanscrit the verbal ad-
jective yagyaSy worthy to be sacrificed ; this same word
must be recognised in the Greek ayio^^ saint. Aspira-
tion is the great resource of the larynx when embarrassed.
Greek, in using it here and in many other circumstances,
as we shall see, does not differ from Spanish and
Florentine, of which the one replaces the y and the x
by the jota^ Quijote, Quexada, Jeres, or the / by an A,
hijo, humo, hahlar ; the other, initial c or 5^ by an A,
hasa, house, hrazie^ thank you.

Z is another sound which imitates a certain number
of allied sounds ; half way between the gutturals and
the liquid dentals, 2? is as near as possible to the semi-
vowel y. In Zend, Yama, the god of death, has
become Djem-schid from Tima-Kshaeta, as if a d had
placed itself before the inconvenient syllable. This is
pretty much what has happened in Greek in the case
of l^euywjuLi, l^vyov, e^i/yi/i/, compared with yugam.jugum,
jungere ; with ^ew, I boil, corresponding to the Sanscrit
yasdmi ; with barley, from yava. The Greek trans-
cription of the Latin words Julia^ Jesics, Maia, conjux,
was very often ZofX/a, Zr/av, Ma^a, koI^ov^,

What was the exact pronunciation ot this z? Slightly
dental no doubt, more like dz than z sibilant. This is



Digitized by VjOOQIC



368 The Indo-European Organism.

indicated by the change of DyatLS into Zeus, and of all
the verbal terminations in dyd into ^w. A slight
effort of the tongue applied to the palate produces a
prosthetic d; from the confusion of the two sounds
arose dz ; the reverse order appears to have the same
effect : when the d is preceded by an s, the same z
replaces the group sd : aupiS^w for arvpla-Sta, Perhaps
the group sd represented the simple z, itself the result
of very various origins. In any case, the primitive
sound y (the agreement of Zend, Latin, and the
Teutonic languages witnesses to its antiquity) is not
less related to the consonant d, perhaps to the con-
sonant g, than to the vowel t.

F", which is rightly called consonantal u, offers even
more striking examples of these transitions between
vowel and consonantal sounds. It presents itself
either as the substitute of u before a vowel : divas,
genitive of dyu; or as the link between u and a
termination: gi^avas, dya, vas ; dhenva, with a cow;
or in a group beginning with a sibilant: sva^ a
guttural ; gva^ kva, or a dental ; tva, dva ; or finally
in the pure state: va enclitic, or va suffix (sai-va,
viqva, all ; nava^ now), vant, termination of Sanscrit
adjectives and participles ; vd^ to blow ; whence vdta,
the wind ; vid, to know, to see ; vagh, to drive ; vam,
to vomit ; vak, to speak ; vart, to turn ; am, sheep.

Classic Greek has no character to represent v, but
it had existed. It was the Phoenician vav, which
remains, for the rest, in Greek numeration, where it
represents six. Grammarians have called this letter
digamma, because its form resembles a double gamma ;
it may still be read on a few inscriptions : poi^a, foiKO^
for oiSa, ouco^ (Latin veicus, vicus, Sans. vi^a). The
Beotian, Laconian, and ^oHan dialects long wrote



Digitized by VjOOQIC



Indo-European Phoftetics, 369

and pronounced it: fipyov (work, root varg). The
digamma has disappeared from the Homeric texts as
they were compiled at the time of Pisistratus, but it
certainly existed in the older songs from which the
editors of the sixth century composed the Iliad and
the Odyssey, Many verses would bo incorrect if it
were not supplied. Here are two : —

Tov Se ap viroSpa iSa)v iroSag oDKvg *A')(i\\€vg,

/JLvQilDV T€ ptp'rip €/JL€V<U TTpfJKTnpa T€ epyWV,

To restore the rhythm we must read fiScov, fp^rnp,
fcpywv. Philologists have noted hundreds of similar
passages.

The letter at last fell into desuetude, and the sound
with it. It is very singular, w)ien we recollect that
Greek pronunciation had no repugnance to the sound
v; far from it, the b and u have very often that
sound : facrtXefy. One would expect at least to find the
primitive v represented by one or other of these
characters ; but this is rare : we have fiouXo/mai, jSovXii,
compared to the Sanscrit vol, and with volo and
voluntas, and a few Laconian forms such as ^epyov,
fiiSeiv, iSepyiXiog. In yovvo^ from yow, knee ; Sovpo^y
from Sopvy lance ; vevpog, iravpo^ (compared with
parvus and nervus), u seems to be the equivalent of v.
F' is 2^ in 8vOy in /8oi/9, and in virvo^y for svapna.

It is probable that the digamma was abandoned
because J and u were commonly pronounced like it ;
but its loss has none the less materially altered the
physiognomy of the language, and separated Greek
from its congeners, Latin and Sanscrit. Sometimes
the soft breathing, as in e/iw (emetic) for vam, e^S,
for vagh ; the hard breathing in epvufxi, kcrdri^ (vestis,



Digitized by VjOOQIC



370 The Indo-European Organism.

vestire), in urrtap, witness, historian, in ecrirepo^ (vesper,
the evening), takes the place of the v ; sometimes it
disappears completely : SdScKa, duodecim ; SnrXoik,
duplus ; ^069, hovis ; vtio^y navis; vio9, noviis ; ivvea,
novem ; 019, ovis; /oeo), TrXeco, i/e<», eiSov, ^Xeo9, and
the suffix K\fj9 (^rava), instead of KXcfo^, e^iSov, /ocfo),
TrXe^, vefOD. It is true that in the case of^ the three
last the forms irXcvw (pluo, pluviics), peiw, vevoo, present
equivalents. But who would recognise in 0X09 for
croXfo^ the Latin salvus, in XaioV the Latin Icevus, in
cTKaio^, scceviis, in ^Su^, suavis, in uXiy for avXfvi, silvay
in 09 for avo9t the theme s^?a, the Latin suvs ? Pro-
bably the V was not precisely either J or w; it was
the w, as is shown by the transcription OvepyiXiog
(Virgilius) ; and this sound was not familiar to Greek.

Latin is here much more faithful to the original
type. It has vinumj vicus, verio, vigor, valerCy volere^
videre, veJiere, vir, vis (Greek ??), nervus, parvus^ salvus,
novuSy curviis, eqvus, boves, novem, divus, Jovis {Aios),
arvum, alvus, that which is ploughed, from arare, that
which nourishes (cf. alu-mmcs). Yet two special
causes have often in Latin brought about the modifica*
tion or the loss of the v, Tlie first is the confusion
with the sound and the letter u; the second the
tendency to contraction.

Latin had attributed to the spirant / the vav of
the Phoenician alphabet, and therefore had but one
character to represent the on vowel and the ov, con-
sonant. Claudian, who had a few valuable ideas, had
proposed to represent the pure i; by a reversed /,
which is found on a few inscriptions of his time ; but
custom was too strong for his reform, and we do not
know whether quatuor, vaeutcs, reliquus, fatuus, equtis,
were pronounced quatvor, reliqvos, fatvos, eqvos, or



Digitized by VjOOQIC



Indo-European Phonetics. 37 1

hatouTy vacons, reliqous, &c. In dico and dioplex, the v
was a true ou, as in houm, buhus, for hovum and hovilms.
In ^e, 56, siJi, dio, diu, dies, the v had simply dis-
appeared; a species of friction had obliterated it.
Sometimes it is hidden in a contracted syllable : nolo
for non volo, nuper for novumpeVj nauta for navita,
upilio, shepherd, for ovipilio (organic, avipalayan),
prudenSy for jyrovidens, (the prudent man is he who
foresees); seorsum, seversum ; rursum, anew, for re-
versum ; elsewhere it leaves no trace ; compare retror-
sum with retroversum ; surstim with svhversum ; coiicio
with conventio; ditior, junior, ?iomis with dives, juvenis,
novem; malo, covimorunt, petii, prohai, with magvolo
(which is partly recovered in mavvlt), commoveruniy
petivi, probavi, A primitive aiva-Sy time, from which
the Greek derives cuwv, am, gives in Latin cevum; the
addition of the suflSxes tas, ter, nus, lis, cus, lengthens
it into cevitas, cevitermts, ceviternalis, ceviternitas, cevita-
ticurriy of which we know the Latin and French forms,
cetas, ceternus, ceternalisy Stemel, 4ternit4y Mage, 4age,
aage, and finally dge. Thus age and eternity are
without a doubt derived from the same root. In the
second of these words the initial e still represents the
first syllable of o'.tas ; but in dge, the circumflex accent
alone reveals the strong contraction which has devoured
the significant parts of the word.

Zend uses v largely ; and though it often changes
this letter into u (dceum for dcevam, the demon), it
sometimes retains it where Sanscrit alters it : vavatcha,
"he speaks," is nearer the primitive form than the Sans-
crit uvatcha; vaz (vagh), viz (vid), vig (inhabitant),
haurva (sarva) havami, vayu {vayou), the wind, are, as
far as the retention of the v is concerned, faithful to
the organic form ; w in the accusative thivam, thee,



Digitized by VjOOQIC



372 The Indo-European Organism.

in rathwoy rathtue, of the master, to the master, is only
a reinforcement for the sake of euphony.

There are other metamorphoses which we shall en-
counter here and there in Greek and Latin, dependent
on the consonant with which v is associated in a sort
of indissoluble group. Kva, ffva, sva, dva, are usually
considered as a species of national caprice, a pheno-
menon attributable to a more or less tardy preference,
and presenting interesting but entirely secondary varia-
tions, as, for instance, in Wilhelm, William, Guillaunie,
Grilles. I am rather disposed to recognise in these
groups very ancient transitions between the semi-vowels
and the consonants.

If we compare the relatives kas (Sanscrit), ir^ or
kS>9 (Greek), with the Umbrian po-ei, with the Latin
qui and quis, with the German kva, " who," we shall see
that they only differ by the abandonment or the reten-
tion of the semi-vowel v. The original theme appears
to have been kva. The same observation applies to
the forms kvwv, canis, gvan, Zend gpan. The root ki^it,
qvit, to shine, retained in the Sanscrit ^vetas, Gothic
hveits, German weiss, English white, gives in Zend
^paeto, brilliant. Whence are these u, v, p, if the v,
vocalised into u or hardened into p, be not paii; of the
original root ? The g of the Zend and Sanscrit words
gvan and ^an always represents a primitive k, as in
o^, rapid, Greek com, agva, Latin eqavs. The Zend
agpa, vi^a, all, helps us to understand linrog. The
Greeks received the horse from Asia, where it was
already called aspas or ispas (Ispahan) ; the s, a weak
representative of the primitive k, became assimilated
to the p, in which the Greeks no longer distinguished
the original v; certain dialects, which confused the
sounds p and k, adopted the form icKog, which corre-



Digitized by VjOOQIC



Indo-European Phonetics. 373

Bponds only in appearance with equiis. The two words
are in truth derived from the same root, but tbey
have a different descent. But the point of departure
of these changes is the presence of a «? in the root. So
with qzicUuor, tchutvar, Ticrcrapy Trlcrvp, patur, Jidvor,
the type is kvcUrar.

The soft guttural g, often palatalised into the Sans-
crit dj, frequently corresponds in Greek and Latin to a J
or to a «?; gans is jSoi;?, bos; ga (Sans. gam\to go, becomes
in Greek fialvoo, in Latin va-do ; djiv, to live, becomes
/S/09, for fit f 09, vivercy vita, Zend has zbayemi, I in-
voke, for ghvaydmiy Sans, hvaydmi (whence hotar,
the priest, in Zend zaotar) ; in the same way the
differences between our classical languages reveal to
us the primitive forms gvaus, gvam, gviv. The first
has preferred the guttural, the others the v pure, or
hardened into h.

The example of the groups tv, dv, lends support to
this explanation. We will not recur to the changes
of tvam into te, into tibi (for tve, tvhi), except to show
that vos (Sans. vas\ probably the remains of a plural
tvas, is connected with it. But dv has a special in-
terest ; the Zend dbis^ to hate, for dvish, bae for dvS,
show us the evolution of v. into b ; much more so
Latin: compare dvx)y dnelhim-y dtumtis, duplex, with
bis, bellum, bonus (Ital. buA)no, Span, bueno), bidens,
Greek, rejecting the u, has kept the dental : SwSeKUy
Sh, for SvoScKa, Sui'g. Thus these are forms of the
same root, not because d changes into b, but because,
when two sounds are combined to make one group, the
different languages have made a different use of them.

The group sva has an equal number of variants.
It remains pure in Sanscrit : svan, to resound ; svar,
to shine, the heavens ; svagura^ father-in-law ; svasar,
25



Digitized by VjOOQIC



374 "^^ Indo-European Organism.

sister ; sm, he ; svadu, sweet ; svapna, sleep. Latin
has rarely retained it (stcadvis, suadere, suits) ; it is
generally contracted into so: sonus, sonare (note the
Italian supno) ; sol, socer, socrus, soror (for sosor), somnus.
Greek has made an effort to imitate the sound in (r(^6^y
(T(f)€T€po^, but it has abandoned the v, and never likes
the initial s, which it changes into the hard breathing ;
consequently this sound is usually represented by e, 7,
V, 6 ; ^Xio^, eWfiVfj, for svarya, surya, svarana, the
brilliant ; cKvpog, for svagura, socer ; i/ttj/o?, for svwpna ;
^Su^, for svddu.

But it is in Zend that the difference is most consi-
derable. The sibilant is rejected at the beginning,
and even in the middle of words, and is supplied by
a very marked aspirate, sometimes accompanied by a
nasal ; hasanhrem, for instance, represents the Sanscrit
sahasram, a thousand. Hva, hvare, correspond to sva,
svar. The initial aspirate is often so strong that it is
equivalent to a hard guttural ; in that case all trace
of the V disappears, and the sibilant becomes q, Qa
represents sva. Qddata, which has been modified into
Koda, a god, Qddata^ the eternal, he who finds in
himself his law, his basis, is the Sanscrit sva-dhata,
and might be found in Greek under the form a-cfyeOero^y
avToOero^, Qafna is svapna, vttvo^, somnus; qap, sopire ;
qar, to shine, and^ to eat, from svar^ svorare^ vorare ;
qan, to resound (cf. the Latin canere, canorus) ; qanhaVy
svasaVj soror ; Haraqaiti, land of rivers, Sarasvati. The
Greek transcription, Arachosia, allows us to surmise
that a 1; or an (? was still heard after the aspirate.

After these various examples, no one would venture
to assert that the labials h and p ar^ not a hardening
of the -y, which itself is issued from the vowel ou^
and that the gutturals and their aspirates are not



Digitized by VjOOQIC



Indo-European Phonetics, 375

the result of the effort made to vary or to utter the
sound t?; this effort would have produced at first a
confused sound, gradually growing more precise as the
friction of agglutination obliterated the semi-vowel.
This is perhaps more than a hypothesis. In any case,
these considerations will not appear out of place in a
philosophical study of language and the formation of
sounds. They seem to grasp those physiological causes,
so long ignored, of the deviations which concealed from
the Latins, Persians, and Teutons the fundamental
identity of their vocabularies.

The pure sibilant, s, which has been so differently
treated by all peoples according to its position, whether
initial, final, or in the body of a word, and according
to the letters, vowels, or consonants with which it
found itself in contact, may be ranked among the first
elements of language ; it is common to man and many
animals, such as the goose and the snake ; it is even
one of the sounds of inanimate nature. Human breath
produces it in the same manner as the breath of the
wind. S is sometimes, and rightly, classed among the
dentals ; through the a; it is allied to d, pure or aspi-
rated ; it constantly permutes with t ; but ch connects
it also with ^, with the palatal tch^ and through these
with the gutturals. Lastly, with the Persians and the
Greeks it is confounded with the. aspirate ; with the
Latins, with the liquid r. These manifold affinities
render its history most complicated.

We will be content with a few rapid notes. In
Sanscrit the primitive s is sometimes replaced by the
palatal q : gushhay Latin siccus, Zend hushha ; gvaguray
Latin socer ; by sh: vsh, to bum, Latin cestvs, tcstus
(aurora, aurum, from urere) ; tarsh, to dry, Greek ripa-a),
Latin tersa, terra, the dry land. Often before certain



Digitized by VjOOQIC



376 The Indo European Organism.

consonants, and according to euphonic laws proper to