Capella. We have seen that his exact words were: z vero idcirco Appius Claudius detestatur quod dentes mortui dum exprimitur imitatur. To this Mommsen (Römische Forschungen, I. p. 304) adds: "Appius kann dies wohl nur als Grund angegeben haben (oder haben sollen) für die Verbannung des aus Sprache und Schrift." And this natural inference of Mommsen's grows from book to book into the story that "Martianus Capella tells us that the letter was removed from the alphabet by Appius Claudius Caecus, the famous censor of 312 B.C., adding the curious reason that in pronouncing it the teeth assumed the appearance of the teeth of a grinning skull" (Lindsay, The Latin Language, p. 6). Jordan, in his Kritische Beiträge, p. 157, argues that it may have been this same Appius Claudius who invented the letter G, rather than the traditional Spurius Carvilius Ruga (cf. p. 24 above). And this theory is accepted by Stolz (Historische Grammatik der lateinischen Sprache, I. p. 84, § 71) as probable and is stated by Lindsay (Short Historical Latin Grammar, p. 5) as a fact; though it is really little more than guesswork.

The other evidence brought forward for such a Z is also (as has been shown by Harrington, Proceedings of the Ameri can Philological Association, July, 1898, p. xxxiv, and by others) pitifully meagre and uncertain. In three of the

medieval texts of Varro's quotation from the Carmen Saliare az is found in the group cozeulodorieso (indeed, we were formerly told there were two: cozevlodoizeso, Seelmann, p. 319), but, as I shall show later (see page 39), the rarer reading coceulodorieso is the correct one, and the more frequent ≈ is only a medieval spelling for c, both sounded ts. It is thus impossible that Velius Longus (circa A.D. 100) had this passage in mind when he wrote: mihi videtur nec aliena latino. sermoni fuisse (≈ littera), cum inveniatur in Carmine Saliari, p. 2217, Keil, VII. p. 51. In all probability Velius Longus found in the antique text of the Salian Hymn an angular >, or perhaps a retrograde, and saw in this a Z. Some have tried to find a Z on the Duenos bowl, but this is certainly a mistake. The letter is and so has some resemblance to

the old I, but none whatever to Z. It is evident that when the text was written this letter was omitted,1 for it is crowded in between the two adjacent letters. In spite of this, it is recognizable as a Vor V, and the word is retrograde DVENOI, corresponding to the DVENOS in the same inscription, as explained by Bréal, Pauli, Comparetti, and Conway. That their explanation of this letter is not only a happy one but also correct beyond all question, I shall show in detail in a forthcoming article on the Duenos inscription and the etymology of certain words in it. [Here I need say only that Conway's interpretation (AJP. X. p. 455) is most nearly correct, but that duenos and manom are not names but OldLatin forms of bonus and malum. manom became malom by dissimilation (Brugmann2, I. § 976 b), and is identical with pavós thin, slight, flaccid, scanty, few,' the weak form of μóvos 'single'; compare the development of the meaning of English slight, German schlecht, from 'simple, slight, etc.,' to 'worthless, bad.'] The only case of a good Z is on one or two coins of the Etruscan town Cosa: COZA(NO) and (CO)ZANO, Ritschl, PLM. I. vii. 40 a; ONAZOƆ 40b; COSA(NO) This Ritschl (Opusc. IV. 721 ft.) regards as Z, Jordan (Kritische Beiträge, p. 155) and others as only a form of angular or ≤ (cf. p. 37). In connection with this might be mentioned the Z used in the Oscan inscription in Latin letters on the Bantine Tablet, but to these two cases I shall return (p. 35 etc.).

41 a.

Not only is the theory of an early Latin Z ill founded and inconsistent, but there are also other serious objections to it. In the first place it takes for granted that Greek zeta had in early Latin the form Z. Stolz has the more correct form I, which, when appearing in Oscan and Umbrian, is referred to by Lindsay (The Latin Language, p. 6) as "the letter written in the Oscan alphabet like a capital I with top and bottom strokes prolonged, and in the Umbrian alphabet with the same strokes slanting instead of horizontal." When speaking of the early Latin zeta, he (The Latin Language, p. 5–6), like

1 It is not improbable that this was due to the fact that the du had already passed into a labialized d, on the way to b, Brugmann2, I. § 359.

Seelmann (Die Aussprache des Latein, p. 319), has only Z in mind, and on page 2 of his Short Historical Latin Grammar, actually gives Z not only as the early Latin, but also as the early Euboean form! On page 5 he tells us, as though a well established fact, that the genitive plural ending was originally written AZOM, and Cicero's words (Sed tum Papisii dicebamini. Post hunc XIII fuerunt sella curuli ante L. Papirium Crassum, qui primum Papisius est vocari desitus. Ep. Fam. IX. 21) reappear in Lindsay (Short Historical Latin Grammar, p. 5) as “L. Papirius Crassus, dictator in 339 B.C., was the first of his family to write his name ПAPIR- instead of PAPIZ-." Not stopping to comment on this strange substitution of Z for the S given by Cicero, Varro, etc., we know Z to be a form that arose in Greece at a comparatively late date, being not at all a true epigraphic form, but one that developed in writing and later passed from the cursive into the monumental hand. Like most cursive forms, it is due to the avoidance of raising the stylus, observe the forms shown in Müller's Handbuch, I. page 304): IIZZZ. We have, therefore, no reason even to look for such a form of zeta in early Latin.

Secondly, the theory puts Latin at variance with the other Italic dialects, not only in the form of the letter, but also in its sound, the latter of which points is evident to Lindsay (Latin Language, II. § 121, p. 105). In Oscan and Umbrian zeta represents ts, while, like s, is represented by retrograde Z2 (Planta, Grammatik der Oskisch-umbrischen Dialekte, I. § 26). As we have not a particle of evidence that the Latin intervocalic s that became z and later r, was ever written Z (even COZA would not be a case in point, for it was not a Latin name, and appears in Vergil as COSA, Aen. 10, 168, not CORA 1), and plenty of evidence that it was written S or 2 before it became and was represented by R; we have no reason whatever for supposing that Latin ≈ was ever written otherwise than S or ≥2, or in any other way than in the remaining Italic dialects.

1 With the Etruscan town Cosa must not be confounded Cora in Latium, one of whose coins is given by Ritschl, Pl. VII. 39, though the names of the two towns may ultimately be equivalent.

...

II.

Turning now to the origin of G, we have seen that Terentius Scaurus (pro ea [C littera] nota adiecta a Spurio Carvilio novam formam Glitterae positam, De Orthographia, Keil VII. p. 15) and Plutarch (καὶ γὰρ τὸ κ πρὸς τὸ γ συγγένειαν ἔχει παρ' αὐτοῖς· ὀψὲ γὰρ ἐχρήσαντο τῷ γάμμα Καρβελίου Σπορίου προσεξευρόντος· . . ὀψὲ δ' ἤρξαντο μισθοῦ διδάσκειν, καὶ πρῶτος ἀνέῳξε γραμματοδιδασκαλεῖον Σπόριος Καρβίλιος, àπελεúbeроs Kapßixíov. . . Quaest. Rom. 54, 59) ascribe its invention to Spurius Carvilius Ruga (circa 231 B.C.). Mommsen has, however, shown (Unteritalische Dialekte, p. 32) that this cannot be correct, inasmuch as the letter was in use before the time of Carvilius. Corssen (Über Aussprache, etc., first edition, p. 7) is doubtless right in supposing that Carvilius did not invent the letter, but taught and advocated the use of C for k and G for g. His own name (Carvilius Ruga) would tempt him to observe the distinction. We were formerly told that G was made out of C by the addition of a horizontal bar; later, that G was really earlier than G, and that the diacritic consisted in a perpendicular stroke or beard; and now our attention is called to the fact that even G is not the earliest form of the letter, but that an older form was G, according to which the diacritic consisted in an upward stroke. It is evident that those who have assured us of the contrivance of G out of C really possessed very little positive knowledge on the subject, and that it is incumbent on us to learn more about the early forms of G and about the forms assumed in Italy by the Greck seta before we venture to draw conclusions.

Somewhere about the seventh century B.C. the Greek alphabet, in its Western form, was brought to Italy by Greek colonists, and soon after was introduced among the native Italic tribes. In this alphabet zeta had the old Greek forms I, etc., but a modification of this letter appears to have arisen among the Greeks in Italy. This modification consisted in the shrinkage and ultimate disappearance of the crossbars on one side of the shaft. As this modification

is found in all the Italic dialects except the Oscan, it was probably common among the Greeks from whom the Italians got the alphabet; but this is hard to verify, as the letter is rare in inscriptions. The I of the Caere alphabet shows a decided shrinkage of the bars at the left (IGA. No. 534, Roberts, An Introduction to Greck Epigraphy, p. 17; Kirchhoff's reproduction, Studien zur Geschichte des griechischen Alphabets, fourth edition, p. 135, is quite wrong). In the Italic dialects zeta3 appears (turned to the right) as:

It is therefore incumbent upon us to look for some such letter in early inscriptions employing the Latin alphabet. This we find on the Rapino bronze in the forms C10, C3, 7, 48 (Zvetaieff, IIMD. II. 2. In his Italic Dialects Conway generalizes or levels the forms under the character, table, ← p. 254). In line 10 it has exactly the form that we should. expect the old seta to have assumed in Latin, if it did not remain I as in Oscan. In the other cases the character tips more or less, just as the upright gamma 【 became in the Western Greek alphabets. Now, it is remarkable that in all these cases this character has the value of g. That is, not

1 The development in Oscan may have been checked by the fact that retrograde F there had the form ]], etc. In Umbrian the two were differentiated by the direction of the bars: = F, J = I.

2 The zeta reported with uncertainty as in the Colle alphabet (IGA. No. 535, Roberts, p. 18), the alphabet of Cepello (IGA. No. 546, Roberts, No. 268), and even in an alphabet from Amorgos (IGA. No. 390 b, Roberts, No. 1596) may be misread for such a zeta, or rather represent the absolute shrinkage of the strokes, thus avoiding the [ = F. Compare the East Italic [ for I and ↑ for T. In the Phrygian alphabet (Journal of Hellenic Studies, III. p. 1, IX. p. 380) one stroke shrinks on one side and the other on the other, thus 1.

shows similar forms: Campano-Etruscan, Faliscan †,

+ Conway is wrong in giving, cf. the facsimiles in Zvetaieff and Bréal.