a German b is pronounced like Germ. w1. It will be seen that the instances from Dionysius are all thus disposed of. As regards Plutarch it is perhaps not inappropriate to remark that he expressly tells us he was not a good Latin scholar (Vit. Demosth. 2, p. 846), and secondly, that he was a Boeotian; and the relations of the Boeotian dialect to the digamma were such as to make it possible that his native pronunciation or habits may have had something to do with this peculiarity. But all the MSS. of these authors are, I suppose, posterior by many centuries to the time of confusion of v and b; and this fact, while not at all impairing their testimony when they represent v by ov, is strong against its trustworthiness when writing B. For there is no apparent reason why a copyist, if he found ẞ written, should have changed it to ou, while the change of ov (for consonantal v) into ß would be in accordance with the tendencies either of pronunciation itself or of its expression. A reference to Benseler's lexicon will show at once a number of words, written earlier with ov, which in Byzantine writers received a ß. Or look to the names of consuls, &c. given from various authorities side by side in the Corpus Inscript. Latin. 1. 483 sqq., and it will be seen how persistently the Chronicon Paschale of the 7th century writes ẞ where Dionysius or Diodorus or Dio has ov, and how often the v of the Inscriptions gives place in the Latin of the 4th century to b; e.g. Calvus to Calbus, &c.

Again, the MSS. of the New Testament, are, I believe, the earliest MSS. existing (except some papyri and the Herculaneum rolls), and the following facts may therefore be of use. The name Silvanus occurs four times (2 Cor. i. 19; 1 Thess. i. 1; 2 Thess. i. 1; 1 Pet. v. 12). In St Peter Vat. alone (against Sinait. Alex.) has Bavos. In St Paul Vat. like the rest (and Ephr. in 2 Cor., being lost in 1, 2 Thess.) has λovavos: two bilingual MSS. Clar. Boern. (cent. 6 and 9) with the transcripts Sang. Aug. and (once) the second hand of No. 67, are the only MSS. late or early, as Mr Hort

1 Schleicher (Deutsche Sprache, p. 212, ed. 2) says: 'b and g we 'write in accordance with the old language, but pronounce these sounds, when medial, between vowels, as w and [voiced] ch, consequently as 'spirants not as momentary sounds...e.g. graben, sagen, as grâwen, 'sachen...The b also in the combinations 1b, rb is pronounced as w; 'e. g. in gelber, farbe, but not when the 1 and b belong to different 'words, e.g. stulbein, harbeutel.'

informs me, which are known to spell the word with B. The Latin version of Clar. (though not of Boern.) has Silbanus. The solitary instance of EλBavos in the Vatican is probably (as Mr Hort suggests) only one of several indications of the Vatican scribe being familiar with Latin; the confusion of v and b being common in early as well as late Latin biblical MSS.; e.g. the Codex Vercellensis of the Gospels (middle of 4th cent.; i.e. same date as the Vatican).

What then was the value of B? Not I think that of the labiodental v. For the only argument that is brought for this value is that it has this value in modern Greek. I do not doubt that some Greek speakers give it this sound, but I am not disposed to admit that all those who think they hear this sound are right. The truth is there is a labial f and v, as well as a labio-dental f and v, and by those who are not familiar with the labial the sound is often taken for that of the labio-dental. Mr Ellis (p. 518) says of an eminent modern Greek, ‘The letters ß, & seem to be naturally pronounced by Prof. Valetta as a labial v and f, but when he became particularly emphatic he made them the labio-dental v and f.' Mr Geldart (Fourn. of Fhilology for 1869, II. p. 159) says, 'ß is pronounced in 'Greece not like our v but like the German w, only much more strongly and explosively, if one may use the word. It is not 'sounded by bringing together the lower lip and the upper teeth, 'but by compressing the two lips together. So too 4, and the con'sonantal sound of v, are pure lip-letters, and very different in 'point of formation from f or v.' (See also Appendix A. xviii.) It is obvious that a sound like this stands in at least as close a relation to the English w as to the English v.

Here then we meet with a solution of the difficulties presented by the confusion of Latin v with b, by the occasional representation of Latin v by ẞ, and by the historical substitution of the labiodental v in the Romance languages for the Latin v. The phonetic pedigree of the Romance v might be at once stated as: 1. u vowel; 2. French ou, pronounced as in oui; 3. English w; 4. Labial v; 5. Labio-dental v. But I do not assert that this represents an historical succession in a single line. It is very probable that the labial v existed dialectically in Italy (and probably in Greece) in classical times, and that this accounts for such instances of the tran

scription of Latin v by B, as may be really the writing of Polybius, Dionysius of Halicarnassus, and others1 (e.g. Béoßiov opos for Vesuvius), and such vacillation in names of places as may be really due to the ancient authors (e.g. Labici, Cic. Agr. 2. 35; so also Greek writers generally; but Lavici, Liv. 2, 39. 4, 45). In and after the 3rd century this sound encroached upon the domain of the w, and rendered e. g. verba indistinguishable from berba. But because the Greek ẞ may very possibly have had this sound, and may have been used for Latin v, it does not follow that Latin ▼ had this sound, but only that in the want of an exact representative ẞ came near enough to be used. I see no reason whatever for supposing that in classical times educated persons pronounced the letter v (u) (except in certain positions) otherwise than as the vowel oo, either with a pause after it, or running on to a succeeding vowel, (as in French oui,) or as English w. The first of these modes was the usual sound of v when called a vowel, the third when called a consonant. After q it may have been a mere sign of the labialisation of the guttural, an effect which most people would not distinguish from w. And possibly the same may be its purport sometimes after g, 1, r, s. (See §§ 89.94, 2, and Append. A. xx.—xxii.) With a short i following, qv made a sound which the Greeks represented by kʊ, i.e. κ followed by the 12th vowel (see below). The rise of b out of v in a few cases is noted in § 76, and this was probably negotiated by a labial v, which perished in the transaction.

K

Corssen appears to think such a sound as the Engl. w to be too weak for v generally, and points to its having expelled the preceding consonant in some words. But the words in which this took place, leaving evidence in historical Lạtin behind it, are very few2, viginti from duo (§ 76), nivis from nigv-is, shown by ninguit and nix, vixi compared with vivo (§ 129), possibly reduvia with ungvis. Others are evidenced only by comparison with Greek or Sanskrit stems. That these changes may have been produced by the mediation of a

1 Some few instances in inscriptions between the battle of Actium and the end of the 4th century p. Chr. are mentioned by Franz (Elem. Epigraph. Græc. p. 248). I have not the means now for further inquiry.

*2 Corssen does not mention such words as sēvoco, seviri; and they are only instances of the usual habit of sed; see §§ 93, 2. 113.

labial v is likely enough, but they seem to me to be part or remnants of the changes which constituted the separation of the Latin language from its common stock, and to prove nothing for the pronunciation of v in the days of Cicero and Quintilian, unless indeed guard (once, I suppose, pronounced gward) compared with ward, &c. shows that w is in English pronounced as v. That Corssen should also consider (Aussprache, 1. 315) the omission of v in such words as sos for suos, savium compared with svavium, &c., or the absorption of v in fautor for favitor, nuper for novum per, as proofs that v had not a 'weak vowel sound like the English w,' but a consonantal tone like the Germ. w1, is to me very surprising. I draw the precisely opposite inference. (See above, 5.)

On F.

On the sound of f I have already spoken (p. xxxv

The facts adduced in this first book and in §§ 766 of the third book would be almost enough to show that f was not a sound of the Indo-European original alphabet, but of a much later and more special source. The number of words, in which it occurs as initial, is not very large, but the number in which it occurs, as initial of a suffix or after a vowel, is exceedingly small-four or five only. (Of course compounds must for such a purpose be separated into their members, e.g. in æstifer f is initial). A few more are named by Corssen (Krit. Nachtr. p. 193 sqq. Aussprache, I. 140 sqq. Ed. 2), e.g. Alfius, Orfius, Ufens, Aufidus, but these are proper names and probably not Latin. Certainly such a rare occurrence of f in suffixes goes far to show that the sound did not exist in the time when these suffixes first assumed shape and use. It may well be that -bro is of the same stock as ferre to bear, but, if so, they are collateral relatives, and -bro is the earlier of the two. Similarly the verbal tense suffixes -bam, -bo, &c., the derivative noun suffixes -bulo, -bili, -bo, the case-suffix -bi in tibi, -bis in nobis, vobis, -bus

1 Corssen means by the Germ. w the labio-dental English v. The south Germ. w is, according to Mr Ellis, the labial v (see App. A. xviii.). But this is not known to all Germans, though Rumpelt (Deutsch. Gram. I. 322-327 note) seems groping for it. See also p. 319, where he argues for the old high German w or uu having had the sound of English w.

in nouns, may very possibly have correspondents in Latin (or Umbrian or Oscan 1) beginning with f, but I should be inclined to regard such words with f as in a collateral not a parental relation to those with b. And thus amavi would not be for ama-ful, but it may contain a suffix from the same root as fui,

On C before æ, e, i, &c.

That c before e or i was in Latin not pronounced as either Engl. ch, i.e. tsh (so in Italian), nor as s (so in French and English), nor as ts (so in German), nor in fact noticeably different from k, may be inferred from the following arguments.

I. Closely connected forms exhibit perpetual alterations of the letter following, without any sign of a variance in the sound of c when followed by e or i. Can Vergil in writing replictus, instead of the usual replicitus, have made so great a change as hardening s or sh or ch into k? If a final e be omitted, could the effect have been to harden these dentals or palatals into k? Yet dic, duc, sic, hunc stand for dice, dūce, sīce, hunce. Hosce is common, but is never abbreviated into hosc: that is to say, c is frequently added when it would, if a sibilant, be indistinguishable, it is not added, when its presence would have been audible! Can decem have been pronounced dechem or detsem or desem, and yet its derivative ordinal have been sounded dekumus, and then, at the same time with that, dechimus, &c. Kailius became Cælius: did the c change its sound when the diphthong ai was changed into the diphthong ae? or did it wait until the diphthong ae gave place to the single vowel e (§ 262)? Compare audacter (Quint. I. 6, § 17) with audaciter; difficulter and difficultas with difficile; capio, recipio, cepi, captum, receptum; cano, cecini; acer, acris; locus, loci, loco, loculus, locellus; lacus with its genitives laci and lacus, and dat. pl. lacubus and lacibus ; piscis, pisciculus, piscosus; qverqvetum with qvercetum; præqvoquis contracted into præcox, and præcox with its genitive præcocis ; fax with its old nom, faces. I am aware that the substitution of a

1 Is it certain that the signs in Umbrian, Oscan, &c., for which we write f, had the sound which we ascribe to the Latin f, and not rather a labial sound?