sound, so as to assign to any particular vowel its nearest representative. Still less had they any such definitions of vowels as Mr Bell's system affords, and his Visible Speech exhibits. It is quite possible that the same letter did not always represent the same quality of vowel; indeed, when we see one letter supplanted eventually by another, we may be sure (as I have said before) that the sound had been already supplanted, before the letter was changed.

But there seems no ground for doubting that a, u, i were what they are now in Italian, the distinction between the Germ, a and Ital. a being relatively unimportant. o and e are intermediate vowels, o being somewhere between a and u, e somewhere between a and i. Modern Italian which, as the local representative of Latin, has perhaps the right to maintain its identity, until a reason for inferring a difference is brought forward, has two sounds of each of these vowels. They will be found included in the list on p. 9, the close sounds being further from a and nearer respectively to i and u than the open sounds. Illustrations of the present use of these sounds in connection with their Latin original are given in Diez, Gram. Vol. I. ed. 3 (see also Schuchardt III. p. 161 sqq.). The most important facts seem to be these:

Open e arises (1) from short e, (2) from e before two consonants, (3) from æ:

close e arises (1) from short 1, (2) from i before two consonants, (3) from long e; and (4) is usually heard when e is final.

Open o arises (1) from short o, (2) from o before two consonants, (3) from au; and (4) is heard (without exception?) when o is final:

close o arises (1) from short u; (2) from u or y before two consonants; (3) from long o in the suffixes one, oso, ore, ojo (though this last is identical with orio which has open o).

From most of these rules there are more or fewer individual exceptions, especially (perhaps in accordance with the real length or shortness of the vowel) from the rules relating to the vowel before two consonants: and both e and o have the close sound frequently, when the former of the two consonants is n. Moreover it does not appear that there is always an agreement as to whether a particular word has the close or open vowel1.

1 I am not acquainted with Italian myself. My notion of the Italian sounds is mainly derived from Mr Ellis's book.

Two points here are noticeable. The first is that both e and o are often written in Italian where the Romans had i and u, and in this case the e and o have the close sound, i.e. a sound nearer to 1, u than the open sound is. If the cultivated Latin dialect had been the parent of the Italian, we should have had here a reversal of the early tendency by which o became u, and e became 1 (§§ 196, 213, 234). But, as the Italian has sprung not from the cultivated language, but from one or more rustic provincial dialects, the explanation is simpler,—the old sounds having been preserved, if the close sounds were original, or, more probably, having advanced only half, and not the whole, distance towards i and u. In either case we gain little if any light on the question, how o and e were pronounced in the cultivated language of, say, the Augustan age.

The second point is that long e and (though less decisively) long o in Latin generally receive in Italian the close sounds, short e and o receive the open sounds. The inference which may be drawn from this is confirmed, as Schuchardt1 maintains, in the case of e, by the fact that æ is often miswritten for ĕ, and i for ē; by the language of the grammarians, who describe ĕ as having the sound of a diphthong (apparently æ), ē as having the sound of i; and by the same difference in quality accompanying the difference in quantity in the e of the Greeks, Kelts, Germans,, English (Schuchardt, I. 461 sq.). In the case of the o sounds the miswriting is not so decisively one way. And though Marius Victorinus (p. 2454, ed. Putsche) says, 'O geminum vocis sonum pro con'dicione temporis promit...Igitur o qui correptum enunciat, nec ‘magno hiatu labra reserabit, et retrorsum actam linguam tenebit: 'longum autem productis labiis, rictu tereti, lingua arcu oris 'pendula, sonum tragicum dabit; cujus observationis et in e litera 'similis pene ratio est :' yet other grammarians (Sergius in Donat. IV. p. 520, ed. Keil; Pompeius v. p. 102, ed. Keil), probably copying from Donatus, speak of ŏ as being expressed at the extremity of the lips (primis labris exprimitur), and ō as sounding within the palate (intra palatum sonat), which apparently would make ō to be a

1 In reading Schuchardt it is well to remember that his distinction of 'clear' and 'dull' corresponds with 'open' and 'close' in the o sounds, with 'close' and 'open' in the e sounds. His use of accents in Italian words is different from Diez's (see II. p. 146 n., but also III. 213).

sound nearer a, and 8 to be nearer u. And the Greek w1 never became so completely u as ʼn became 1 (Schuchardt, II. p. 146), though the Germans and English, it may be added, give to their long o a sound nearer to u, and to their short o a sound nearer to a.

It is not easy to draw with much confidence any argument from this to the pronunciation of the Romans in the classical period. For (1) Italian is (as has been already remarked) not the child of classical Latin, but of one or more unsubdued dialects. (2) The inference from misspellings is by no means clear in the case of o, and is not very weighty in the case of e. For is frequently miswritten for long e, and i for short e; and many instances of a for short e are probably due to mistaken etymology (e.g. præces, sæcundum, quæstus for questus). (3) The grammarians quoted (Schuchardt, III. 151, 212) are none of them earlier than the 4th century 2; and three centuries are a long interval, when delicate distinctions of sound have to be caught. (4) The analogy of other languages is proof only of what was possible, not of what was actual, still less of what was actual at a particular time. And long e and long o, even if they changed at all, may yet very well have been open e and open o in the mouths of Cicero and Quintilian. Mr Ellis's investigations into English pronunciation show a similar direction and at least as great an extent of change within the period from the 16th to the 19th century. The whole section of Mr Ellis's book (chap. 111. § 6) is very illustrative of the question, but some of his words describing the change may be quoted. 'The long vowels have altered more than the short vowels. The ' voice being sustained, there was more time for the vowel-sound to 'be considered, and hence the fancy of the speaker may have come 'more into play. This has generally given rise to a refining process, 'consisting in diminishing the lingual or the labial aperture. The lingual aperture is materially diminished in the passage from a long 'Italian a (2nd vowel) successively to Somersetshire a (13th vowel),

1 Mr Ellis says (p. 523), that Prof. Valetta (Greek) pronounced Greek (o and w) and English with a clear 7th vowel (Ital. open o), and did not seem to be aware of the 8th vowel at all.

2 Terentianus, quoted by Pompeius (Keil. v. p. 102), does not bear out the quotation, at least if the poem of Terentianus Maurus is

meant.

'to open e (15th vowel), to close e (16th vowel); and again in the passage from open e to Ital. 1. The change of long open o (7th ' vowel) to long Ital. u (10th vowel) was a similar refinement, con'sisting first in the elevation of the tongue, and corresponding 'narrowing of the labial passage, producing long 9th vowel, and 'secondly in the narrowing of the pharynx. The change from open 'o to close o consisted simply in narrowing the pharyngeal cavity.' (Ellis, p. 232.)

This tendency of long vowels is a tendency working through long periods of time, and is not at all inconsistent with Mr Bell's assertion, 'that the tendency of all vowels is to open in prolongation' (Principles, p. 34, comp. 122). This latter physiological tendency accords with the following line of argument, which seems to me to furnish us with some evidence as to the quality of o and e in Latin. It has two premises; (1) the representation of Latin vowels in Greek, and of Greek vowels in Latin; (2) the components which under crasis, contraction, &c., gave rise to W and

or a.

η

The details of the representation1 will be found under that head in Book I. Chap. 9 (viz. o in §§ 208, 218, 219; e in §§ 229, 239). The facts of Greek contraction, &c., may be found in Greek Grammars (e.g. Kühner's Ausführliche Gram. ed. 2, §§ 50, 51. Curtius, Gr. §§ 36-38. Comp. also ib. § 42).

Now the very introduction of the new symbols w and W η probably implied a sound different in quality as well as in quantity from o and respectively. And this is confirmed by the fact that the

e was el,

not ni

in other words

name of o was ou, not w, and of that, as the voice dwelt on the sound of o, it naturally uttered ov, and as it dwelt on it uttered el. E, In the same way, when the vowels and o were lengthened in compensation, as it is called (see below, § 273. 4), for an omitted consonant, they become ɛɩ and ov. But when ao and oa are contracted, we get ∞ in Attic: when ae is contracted, we get a; from ea, usually in Attic. From these facts I infer that ய and differed in quality from o and e by being

nearer a,

1

and not by being nearer the u and 1 sounds; i.e. and ய

η

Correspondence, i. e. Etymological representation (see p. 24, n.) is not here concerned. And to this head belong the suffixes of inflexion, e.g. Hecuba, 'Ekáßŋ.

were opener, not closer than o and e. But n was perhaps nearer to

e than @ was to o.

But Latin ō represents Greek @, and w represents Latin ō. Both Latin Ŏ and Latin à represent Greek o; Greek o represents Latin o; and both 0 and ov represent Latin ǎ (as well as Latin v, cf. §§ 90, 91). This seems to imply that Greek o was between Latin Ŏ and Latin u. Again Latin ĕ represents Greek and η, presents Latin ē. Latin ě represents Greek €, and Greek € represents both Latin ĕ and often Latin I.

← was between Latin ĕ and Latin I; but

η re

This seems to imply that

possibly, considering the

sphere of , e was nearer to Latin ĕ, than o was to Latin o. In the stricter Doric oo gives w, ao and oa give a1; ee, ae and ea give n. This is probably to be accounted for by supposing o and to have been opener in Doric than in Attic or Ionic, and perhaps a to have inclined more to the o sound than it did in Attic. But the language with which we compare Latin is the language of Polybius, Dionysius, Diodorus, &c., and this is an Attic dialect, though a late one.

Now, without professing to be able to assign any absolute quality to the ancient vowels, I may, if this argument be sound, express their relative qualities by a tabular arrangement. I take a, aw, Fr. au, u to represent four regions of labial vowel sound, and a, ê, é, i to represent four regions of lingual vowel sound. Then we may arrange Attic, Doric, Latin somewhat as follows:

It should always be borne in mind, in comparing the transcription of a word in different languages, that each can supply only

1 The Doric substitution of a for w is reproduced in the Cumberland quarter, and Somersetshire cord with 2nd vowel instead of 6th; the Ionic substitution of ŋ for a by the Somersetshire Bath with the 13th vowel instead of the 2nd. (Cf. Ellis, p. 67.)