what it possesses, and therefore if the sounds are not the same (and the whole range never is the same), the representation of them can be but approximate. Hence the Latin u and the Greek o may sometimes be representative of one another. But generally Gr. o and o go together, and ov represents Latin u. That ov should in the Roman period represent u even exactly, is no obstacle to its having earlier represented the long sound of the Greek o. This change is analogous to that which has befallen which is now identical with long i. And both are but instances of the same law as that which we find to have prevailed in English. So e (at least before consonants, § 229) was in the Roman period a long i, but earlier a long e. Whether both e and ov had, at first, the slight diphthongal termination which we hear in our ordinary Southern English long a (= eɩ), ō (=ov), is not easy to say positively, but it looks probable enough on the mere face of it.

η,

On the diphthongs ai, ae, oi, oe, ui.

αι,

be

The Latin æ, the ordinary representative of the Greek came eventually hardly, if at all, distinguishable from e, just as at was confounded with e. Originally it was doubtless a diphthong. And this seems to have been the case in Varro's time; for he states (Lat. Ling. v. § 97, Müll.) that in the country edus was used, in the city ædus, 'with the addition of a as in many words;' (see also VII. § 96, Müll). Speaking, not writing, must be meant. Now a diphthong with so small an interval between its limiting vowel-positions easily passes into a single intermediate sound. It may be assumed that this sound, if it differed from e, lay on the side nearer a and not on the side nearer i. So that if Latin e be represented by the Italian open e, perhaps the English ǎ (13th vowel) may be taken (in quality) for æ. The sounds are quite near enough to be readily confused, and yet are in themselves distinct. A Saxon (says Mr Ellis, p. 58, 106) would pronounce the English words had, head, with the substitution of the Italian open e for the vowel in each. If the Latin e be represented by the English open e, we get a somewhat greater distinction (and that is desirable) between Latin æ (13th vowel) and e (15th vowel).

The sound of œ is somewhat perplexing. Mr Ellis has suggested (Trans. Phil. Soc. 1867, Supp. p. 65, and Early Eng. Pron. p. 529) that Greek oɩ was originally ui with the first element preponderating, Latin œ was originally ue with the second element preponderating. This seems possible enough for the Greek, as o had frequently that approximation to our w, which is here presumed (see App. A. xii.). But the Latin sound is much more doubtful. It is true that œ is the successor of Latin oi and the representative of Greek or, and that both of and ∞ passed frequently into u, e.g. coirare, cœrare, curare; mœrus, murus; monia, munia; pœna, punire, &c.; but I am not aware of any indication that Latin o had any such approximation to our w; and œ never alternated with ui. The passage of oi and of œ into u seems to imply that at that time the first, not the second, element, the o, not the i or e, was in the preponderance. In imperial times ∞ became confused with e and æ, and then the second element may have been preponderant. And this was the case also in the words which in very early times were spelt with œ, e.g. læbertas, oloes (cf. §§ 264, 363, 366), and afterwards were written with I. I am inclined to think that the diphthongal sound implied by the letters o and i, or o and e, (with their Latin sounds) is the safest conclusion, and that in the words which the ordinary language spells with ∞ (e.g. amœnus, cœpi, monia, fœdus, Pœni, pœna, obodio) the stress should be laid on the o rather than the e.

ui as a diphthong occurs (besides an interjection or two) only in huic, cui. In both of these words it represents an earlier oi, e. g. hoic, quoi. In Quintilian's time (1. 7. § 27) cui and qui appear to have been pronounced alike. Probably the sound was French oui. In the dative of -u stems, e.g. gradui, the vowels would probably be pronounced separately, when both written. But a diphthongal pronunciation may have led to the omission of the i. On the pronunciation of -aius, -oius, -eius, see § 138.

On a supposed sound like ü.

There are three cases in which it appears more or less probable that the Latins had a modified sound of a short vowel similar

to that of French u or Germ. ü, especially when it inclines, as it does in some parts of Germany, more to i than to u.

(1) The first case is in the combination qui-, which is generally represented in Greek by κu, though sometimes by kuɩ or kol. Cf. § 90. 2. And in some Latin words qui- is descended from cu- or co-; e.g. Quirinus from cures, esquiliæ from æsculus, inquilinus from incola, sterquilinium (§ 934) from stercus. So Tarquinius from the Etruscan Tarchun (Schuchardt, II. 277). The labialisation of the guttural, which is expressed by qu, affected the following vowel, and the result was a pronunciation like kü instead of kwi.

(2) The second case is that of I after v, which is noticed by several of the grammarians in almost the same words. Priscian (Part. § 24, 25, III. p. 465, Keil) lays down generally, that words beginning with vĩ followed by d, t, m, r or x appear to have the sound of the Greek v, and instances video, vim, vis, virgo, virtus, vitium, vix, and says most people gave the same sound to fi. But I see no other authority for such a statement, the only examples quoted by Diomedes, Servius, Sergius and Cledonius being vir, to which Velius Longus adds virtus, and the Appendix to Probus (IV. p. 198, Keil) adds virgo and virga. (Cf. Schuchardt, 11. 219. Schneider, 1. 19 sq). I notice this because vir and its probable kin are almost the only words in which short i occurs before r, and some peculiarity of the sound of Ĭ in these words is therefore not unnatural (cf. § 184, 3).

(3) The third case is that of the vowel in the penultimate of superlatives and ordinal numbers, which was u in the earlier language, and i usually in the subsequent language. Jul. Cæsar is said to have first written i. The variation in spelling remained for long. Quintilian (1. 7. 21) expressly says that the sound of i in optimus was intermediate between i and u, and this view is confirmed by the later grammarians.

But on the other hand there are difficulties. (1) It appears likely that, if the sound of this vowel had been that of the French u, the Latin y, which was the Greek u, would have been freely used to represent it. But from Schuchardt's collections (II. 224, 225), it appears that it is rarely found in this termination. Indeed it is more

common in gyla, Sylla (Schuchardt, II. 198, 205). Its rarity, however, may be accounted for by the natural shrinking of the Romans from writing their own words with a foreign letter. (2) The Greek transcription of these words is, so far as I am aware, uniformly by i, not by v. (3) Quintilian dwells on the beauty of two Greek sounds, and v, and expressly says the Romans have not got them (xii. 10, § 27). (4) The later grammarians, except Marius Victorinus, do not suggest the y sound for this vowel, though Priscian does almost in the same sentence suggest it for 1 after v.

I do not see much likelihood or possibility of u changing to 1, without some such intermediate step. But yet it may be, that the vowel was not specifically u or specifically i, but simply an unaccented vowel in a suffix, which for a time was, under the influence of the following labial, retained at the stage of ŭ, but afterwards was carried away by the general drift and became 1. In this case the precise quality of the vowel need never have been very sharply defined, and the representation of it by one of the five vowel signs was approximate only. Or, indeed, the relation of the two sounds in this and in many other cases may be more analogous to the correspondence of sounds in different languages. U may have belonged to one dialect and i to another, and the eventual substitution of i may have been mainly the triumph of the second dialect. Thus Mr Ellis (p. 473, n) speaks of the Peak in Derbyshire having two distinct pronunciations of e. g. sheep, and one of these is a sound which one Southerner might interpret one way and another another. Thus sheep might be sounded with the 16th vowel, or the 18th or the diphthong 3 to 18. We shall not be far wrong, if we print proxumus or proximus according to the best evidence we may have respecting the particular author in whose text it occurs, or the period at which each author wrote, and then pronounce accordingly either u or 1 lightly. But our English sound-the 3rd vowel-is, I suppose, entirely out of the question, though I expect many English speakers often utter it in these as in many other unaccented syllables.

Miscellaneous: chiefly on vowel pronunciation.

There are one or two other points respecting the pronunciation of vowels which may here be mentioned.

The length of the vowel should be preserved, as much before two consonants, as before one or more. In the cases of ns, nf a vowel originally short was lengthened by position (§ 167). Mr Munro takes Priscian's statement1 (II. 63), that the vowel before gn was always long, as meaning that the gn makes it long by nature: but I cannot agree to this. Priscian could on his principles come to no other conclusion; for he held that gn begun the final syllable (II. 8) and that gn made a preceding syllable common, i.e. allowed a short vowel to remain short (1. 11; II. 12). Hence, finding all words which ended in gnus had the penultimate long, he concluded the vowel must be long. But, I believe, gn did not belong to the last syllable; the g belonged to the penultimate. And, as in Greek such a syllable with a short vowel (e.g. ëyvwv) is always long by position, although at one time it was supposed that occasionally it remained short, there seems no reason for assuming in general the vowel to be lengthened. In ignavus, &c. where the n is omitted, the i may be lengthened in compensation. Many words no doubt had, or were supposed to have, a naturally long vowel, e.g. regnum from rēg-, rex, instead of from règ-ĕre; but tignum, signum, magnus, &c. (comp. tigillum, sigillum, măgis) probably have a short vowel. The Latin words Egnatia, Egnatius occur not unfrequently in Greek with e. (See Benseler's Lexicon).

Of course a short vowel before two consonants (unless lengthened as above) should be pronounced with its usual short quantity.

In English we are in the habit of changing, or pronouncing obscurely, short vowels in unaccented syllables, e.g. in the first syllable of appear, together, &c., and in the final syllable of mention, goodness, cabbage, futile, honour, &c. In Latin the pronunciation may be presumed to have been, as in Italian, more distinct; and though changes of the vowels occur, we shall be safest in following the spelling, which represents, though no doubt sometimes laggardly, the pronunciation.

1 Priscian is, I think, unsupported in this statement.

2 See § 272. The Verona palimpsest of Livy, which was probably written in the 4th cent. p. Chr., and consequently before Priscian's time, always divides words with gn occurring at the end of a line between the g and n, so as to give the g and n to separate syllables (Mommsen, Cod. Liv. Ver. p. 164).