it was not a consonant, because in that case the e in CH. IX. equites must have been long; it was not a vowel, for that would have lengthened the second syllable by crasis with i. In other words, the U was merely a symbol, expressing further and somewhat unnecessarily the indistinct aftersound which made Q different from K. This sound, as I have said above, was liked by the Latins, and therefore they retained the koppa. The Greeks did not use the sound, and therefore soon dropped the symbol which they had taken from the Phoenician alphabet; it could never have been of use to them, for there is no trace of any period in Greek during which was passing into : the transition would seem to have been immediate. It is this transitional sound which the Latin Q represents; only the transition was never accomplished in Latium, though it was in the rest of Italy'.

Exactly analogous to the change from K to π and p is that from G to B and b; and here also we have the middle step denoted by the Latin gu. Here too the u is not parasitic in every case; thus in pinguis the u is a suffix, found in Tax-ú-s, and a new suffix has been added in the Latin; similarly in breuis for bregh-u-is (Bpax-v-s2), But it is parasitic in tinguo, the Greek Téyyw, in urguere, where the language has presented the simpler form urgere, &c. Rather frequently the v has forced out the preceding g, and thus given rise to an apparent strengthening; in reality there is a loss. Such cases are uiuere for guig-u-ere; compare the old Norse kvik, our old English "quick," and Sanskrit jîv"; bre(gh)uis, &c. mentioned above; in these the v is original. In fruor for frugu-or (frug-es) both g and v have fallen out. The Latin

1 In the few Doric inscriptions where is found, it is generally followed by o, see Ahrens, II. 88, and New Cratylus, p. 190. This seems to be an attempt, similar to the Latin, to express the after-sound more clearly.

2 Corssen, 1. 85.

3 Krit. Beitr. 65, &c.

4 I think that Corssen is right in thus explaining the word (Krit. Beitr. 72), as opposed to Curtius, who treats the g itself as parasitic (Gr. Et. 547).

Change of

G to B and b.

CH. IX.

Change of
GH to .

=

words so far have no Greek equivalent which shews any
corresponding change. But answering to uor-are for guor-
are (Indo-European GAR) is Greek Bop-ά, where the gut-
tural has passed into the labial: (g)uen-ire is in Greek
βαίνειν for βαν-ψειν, and the original guttural is kept in
Gothic kviman, our "come." The Latin, on the other
hand, shews no change in grauis, where the Greek has
Bapús: but the Sanskrit is guru, so the u may be original,
or at least there may have been a by-form guar; and
in the Greek itself we have the Boeotian Bavá by the
side of yvvý, shewing that the old form must have been
yvavā, our "quean'." Curtius is probably justified in
assuming an original g, where B is found in Greek, in the
word Babus with which he compares the Sanskrit gáh
(doubtless originally gâdh) to dive into: in Bía, compared
with Sanskrit ji, to conquer, and in Bíos, compared with
jyd, which may belong to this root: so also Baλ may be
Sanskrit (and original) √gal, found in German quelle, a
fountain. Perhaps the only undoubted case in which both
Greek and Latin have the labial is the certainly Graeco-
Italian bov-," a cow :" here all the other languages have
the guttural; the Sanskrit is gaus, the German kuh.
These examples are sufficient to shew that the v is less
frequently a mere Graeco-Italian sound after g than after
k, as might have been expected from g being a softer and
easier sound but it certainly was so in some cases; and,
whether original or parasitic, it equally had the power in
Greek of turning the guttural into a labial. The Italians
seemed to have stopped at gu, as the Latins did at
kv (qu).

The same cause may account for the rare change of the guttural aspirate in Greek. It becomes & in víþ-ew, from the original root SNIGH. Perhaps also eλap-pós may exhibit a weaker form of the base which we see in eλaxús: the v is there, which in Latin le(gh)u-is has • See p. 132,

See page 109,

been strong enough to eject the guttural altogether. I have already mentioned the not unfrequent change in Latin from gh to f in my account of the Latin Aspirates', and said that the same explanation is possible; it rests principally on analogy, there being no middle step preserved by the Latin, as in the case of the unaspirated gutturals. It is also possible that the change may be due to greater strength being given to the breath which is the second member of the compound: in this way the distinction between the initial momentary sounds would tend to become obscured: though this result was very rare in Greek, and not very common even in Latin. It is common enough, as has been already mentioned, in English at the end of a word, as laugh, but still more commonly the sound is lost altogether, as in though and (medial) in light, &c.

CH. IX.

2. Dentalism.

K to T.

This change from K to T is much less frequent in Change of Greek. In Latin it does not seem to occur, except in the late transition of -cio into -tio, &c.2, which is caused by the i being really a semi-vowel when another vowel follows; in these cases it is of course part of the suffix. So also was the to (yo) in Greek, which we saw produced so much change among the Greek verbs; as, for example, πρακ-ψω became πρατ-ψω and πράττω. These examples are quite enough to shew that y really has the power of turning a guttural into a dental and justify us in assuming a parasitic y in cases where the change has happened in Greek without any suffix to explain it : especially when traces of the same action are discernible in the cognate languages. An undoubted middle step is given by the Sanskrit palatal ch, which is pronounced

1 See p. 337. More examples (not all very certain) are to be found in Corssen, Krit. Beitr. 203-226.

2 See p. 375.

CH. IX.

half-way between the guttural and the dental, and was probably caused by an attempt to sound k without bringing the tongue far enough back, so that the organs are partly in the position for sounding k, partly in that which produces the palatal breath y, which therefore slips out involuntarily after the imperfect k, and the whole result is ch or, perhaps more commonly, tsh, where we pronounce the t very quickly. In the examples which I am about to give from the Greek, the y does not seem to have been very fully heard, though it had the power to change the k to T and then fell out: so that the order of sound was K, Ky, Ty, T'. In the verbs and nouns mentioned under the head of Assimilation, where the y was part of the suffix, it left a permanent trace of itself in the doubling of the consonant. This difference of result in the cases where the y was radical, and where it was only parasitic, is, I think, no more than we should expect.

The certain examples in Greek are not very numerous, and have indeed been mostly mentioned before. Thus Téσσapes, and Sanskrit chatvaras, are instances of Dentalism, though we saw that the Italian dialects gave us the labial in the same word. These numerals were of course peculiarly liable to corruption: they are almost the commonest currency of language: from their being necessarily used in barter, they are liable to foreign influence more than any other words: a fact which may be the key to the perplexing agreement of numerals in totally distinct languages, and to the strangely-altered forms of some of the Sanskrit numerals. This numeral, katvar, of the IndoEuropean had apparently two separate indistinctly pronounced forms before its separation, kyatvar, whence TéTFapes, and kvatvar, whence quattuor: unless we rather believe that these weakenings took place after the ultimate separation, and so the agreement of réσoapes and chatvaras would be accidental: if this be so, as is on the

1 See p. 14. If the Norse fjord, fjeld, &c. are examples of the same principle, it would appear that the y-sound can slip in, even after labials.

whole probable, we must still assign to the Graeco-Italian the double form katvar and kvatvar: from the second we have the Latin quattuor, and an old Greek TéτFaρes, whence the Boeotian πέτταρες and the dubious πίσυρες, with which the Umbrian petur and Oscan petora also agree: the first form does not appear pure in either language; but the Greek alone dentalised the x, and arrived through verfapes at the Attic τέτταρες οι τέσσαρες and the Doric TéTapes, where the a has been dropped and the F vocalised: the Latin shews no t. Just the same variety of the Greek and Latin forms is seen in rís and quis, where the Sanskrit has the original k in kis; parallel however to Te and que, which are probably from the same base, the Sanskrit has cha, corresponding again to the Greek. Lastly, Ti-w corresponds generally to Sanskrit √chi, so that here also we have probably an instance of dentalism: no Latin word can be connected with these: for that timeo' belongs to the same family seems unlikely both from its meaning and from the t, for there is no evidence of the t occurring for K in Latin: both Tíw and chi have many meanings, but the radical idea seems to be to search," and then "tell over," "count;" and so in Greek "to estimate," "honour;" in Sanskrit to "collect." These forms are all which are given by Curtius as certain2: and he observes that in all of them the original k was followed by either or the cognate e; a fact which would very much assist the slipping-in of the parasitic sound.

CH. IX.

The change from G to 8 is exceedingly rare and uncer- Change of tain, occurring mostly in isolated dialectical forms. Cur- G to 8. tius, however, explains, though somewhat doubtfully, by this process the verb (aw, as being for Staw by the regular process of Greek assimilation. This & he would connect with GI, the simpler form of the old root which appears lengthened, but also dentalised in Sanskrit, as √jiv, and probably labialised in Greek, as BF in Bi(F)os, &c. If

1 Benfey, Sk. Lex. s. v. chi. 2 Gr. Et. 442, &c. 3 See p. 367.