only is the oldest form of G thus far discovered identical with the characteristic Italian form, of seta, but it thus appears. that the letters are one and the same and that this character from the beginning to the end maintained its place as the seventh letter of the alphabet. The development of form is exactly parallel with that of gamma, simply somewhat slower in the early stage: – But one naturally asks: How did it come about that the letter zeta should stand for the sound g? The Greek dialect that gave the Italic peoples their alphabets still had I as the sign for the sounds de (Planta, Gram. der O-U. Dialekte, I. p. 73). The Oscans and Umbrians took the letter for their nearest correspondent, namely ts, and we may ask why the Latins did not do the same. The answer is very simple. Original ts became ss in primitive Italic, ss and s in Latin (Brugmann, I. § 753; Planta, Gram. der O-U. Dialekte, I. § 190). Later, new ts's arose: (1) by syncope, particularly in Oscan and Umbrian, for example, Oscan húrs, that is, hirts Latin hortus (Planta, § 109, etc., § 190); when ts arose in Latin in this way, it passed on to s(s), as original ts had done, for example, *parti-s > *parts > pars (Brugmann 2, I, § 763 c; § 753); (2) by the change of ns into nts in Oscan and Umbrian, but not in Latin (Brugmann2, I. § 415). There thus was a ts in Oscan and Umbrian to be represented by zeta, but none in Latin. The character was, therefore, in Latin an idle letter. In shape it resembled one form of kappa, as gamma (< ) resembled another. In order to make this clear, we must call to mind the early Italic forms of these letters: = gamma zeta < C K K K E F EK2 1 As, for example, 'since,' that is, sins has become sints in the English of the northern central States. Cf. also Whitney's Sanskrit Grammar, § 207, where, however, the process is incorrectly explained. 2 For such forms of K see IGA. Nos. 155, 341, 484, etc., the Duenos bowl, the That and k became confused, we all know; that the forms of zeta should also become confused with those of kappa was but natural. Indeed, it is more than likely that the form F, which is usually classed as a kappa and regarded as a corruption of K, is really a zeta, the further development of ‡, cf. the Etruscan . Similarly, the found in the Duenos inscription and elsewhere is more likely a development of such a zeta as is seen in the of the Caere alphabet (page 30), than a perversion of k. It might be said, and has been, that gamma and kappa became confused because of the similarity of their sounds. And there has been a great deal of talk to the effect that the distinction between Greek κ and y was greater than that between Latin c and g (Corssen1, p. 5, 16; Seelmann, p. 344; Stolz in Müller's Handbuch, II. p. 250); though the idea is not at all supported by the history of the language, and would probably never have been suggested but for the confusion observed in the letters. In the same way the confusion of and and of and 9 (see page 33) whereby Etruscan lost the means of distinguishing g from k and d from t and by analogy subsequently ceased to distinguish and in writing, has led to the inference that in Etruscan the voiced stops became voiceless (cf., for example, Conway, Italic Dialects, p. 464). That this confusion 1 was not due to the similarity of the sounds is shown by the archaic inscription given by Egbert, Introduction to the Study of Latin Inscriptions, p. 274, the inscription just found in the Forum (Stele, etc. Estratto dalle Notizie degli Scavi del mese di maggio, 1899; Berliner Philologische Wochenschrift, Aug. 5, 1899), and, best of all, Plate K, Annali dell' Inst. 1876. 1 The history of the Runes presents a similar case. As Greek was simplified to and H in Greek, Latin, Runic, etc., so Greek E were simplified to П, *, and *, the last by such inversion as changed Greek V to Runic A. These forms of e came in conflict with those of p, namely, and, by inversion, *. After a period of more or less confusion, a differentiation set in, which gave to e the form in the whole Germanic territory, and top (1) the form on the Continent, (2) * or in England, and (3) in Scandinavia. That is, in the North-Germanic countries the letter for e drove the similar letter for p entirely out of use, so that p and b were represented by the same, letter. In time the graphic distinction of td, kg, was also given up. Still, there was in the language no corresponding confusion of the voiced and voiceless stops; in fact, they were later again distinguished in writing. fact that Etruscan transmitted g and k correctly to Oscan and b and Пp to both Oscan and Umbrian,1 and that there was in Latin no confusion of d with t or of b with ; but particularly by the fact that confusion of gamma and kappa is characteristic of the Western Greek alphabets, in which gamma, by becoming <, approached k in form. Thus < was confused with k in Etruscan, Umbrian, Faliscan, and Runic (page 34), as well as in Latin; in all cases, except in Umbrian, to the disadvantage of k. But no such confusion took place in those Greek alphabets that retained the form r for gamma. It was, therefore, formal rather than phonological similarity that led to the confusion of gamma and kappa, and it was similar formal likeness that caused the confounding of kappa and zeta. There was thus a period of more or less confusion during which the sounds g and k were represented by ‹ or C, kor K, and or (cf., for example, 1 From this it is clear that the confusion of d and qr in Etruscan preceded the confusion of 》g and ;|k, and that both preceded the loss of in Etruscan. The various stages of Etruscan and the relation of each to Oscan and Umbrian may be seen from the following: -- (1) 9r, Od Tt; )g k; 867.p. A confusion of 9 and arises (cf. page 32): — (2) ¶r and d, Ɑd and r, Tt; )g ak; db 7p. The alphabet passes to the Oscans, among whom the byformЯ prevails and the signs for and d are differentiated, whereby we get the usual Oscan: ", Яd Tt; gk; bПр. In Etruscan and too become confused: (3) ¶r and d, Ɑd and r, Тt; )g and k, 》 g and k; gb 7p. The alphabet passes to the Umbrians, who differentiate the letters for and d as the Oscans did and let | drive out ), whereby we get (a) primitive Umbrian : Or, ¶d Tt; &g and k; Ɛb 1p. Intervocalic Umbrian ♂ becomes ✈ and, taking the symbol q, leaves other d's to be represented by T. Thus we get (b) the Umbrian alphabet as we know it: 9; Td and t; g and k; 8b 1p. In Etruscan, on the other hand, drives out, and drives out 9, and then scribes give up the anomaly of distinguishing the labial stops b and pin writing, and we get the last stage of Etruscan: (4) Or; Td and t; )g and k; 1b and p. 2 This is a changed to a, not the reverse, as stated by Egbert (p. 27), Conway (p. 331), and others. on the Duenos bowl; the c in Praenestine, Conway, No. 297; and aciptvm comvivia hvc gondecorant volgani gonlegivm aged(ae) Garrucci, Syll. No. 557, as quoted by Seelmann, p. 344; I have no access to Garrucci). And, for that matter, koppa sometimes entered the competition (cf. 1 E, etc., that is, eqo K = ego Kaiso, Egbert, p. 274, and Seelmann, P. 344, etc.). In the course of time there developed out of the chaos more or less order. Thus the complicated characters and or Q became restricted to special and limited use (to which had a tendency from the start) and the letters C and G, which could easily be scratched without raising the stylus, were most generally employed and became differentiated into < C = k and G = g. The corresponding process in the Runic alphabet is so similar that it must not be passed without a word. In this originally Western Greek alphabet (see Journal of Germanic Philology, II. p. 370), ‹ k X passed through a period of confusion (corresponding to that of < k【 in Latin), which resulted in the loss of k and the shifting of < to k as in Latin, whereby X got the value of g, as did in Latin. — In a similar way, as we have seen (page 33, ft.), the likeness of form in d and 9 (later Я) r led to their confusion in Etruscan, and thus in Oscan and Umbrian. In Oscan a differentiation set in whereby the values of the two letters were just reversed. In Etruscan and Umbrian both the letters became lost to d, which was therefore expressed by the sign for the corresponding voiceless stop t In Umbrian there developed out of and ¶ the byform or d, and the three were ultimately differentiated for the three similar sounds r,, . The three-stroke letter. for n, or N, was similar to the four-stroke Ms, and therefore sometimes confounded with it. So we find W as the spelling Was for both and s in a Tarentine inscription (Roberts, No. 268), and in the Caere alphabet and inscription we find that, after such a period of confusion, a differentiation set in, whereby the values of the two letters were reversed, that is, is s and M is n (see IGA. No. 534 and Roberts, p. 17). We must remember that after a period of confusion, nobody knows that one of the sounds had an original claim on one of the letters, and thus a new differentiation may result in the absolute exchange of values. = [While reading the proofs of this paper, it occurred to me to look up the treatment of gamma and kappa in the Celtiberian alphabet. As gamma there had the form <, I was not surprised to find that it had been confused with kappa and, as in Latin, had become one of the signs for the sound k. But I was not prepared to find that, exactly as in Latin, sayin or seta (in the form 1 Z etc.) had assumed the old value of gamma, namely g, and that it had even adopted the name guimel gamma. It is evident that the old I assumed in Celtiberian the forms: (1) λ (cf. the Italian forms); (2); (3) ↑ etc. (cf. the Phrygian form, p. 30 ft. 2). The first type brought it into conflict with opened A, that is A, and the second with <, which in turn was confounded with K. In this way, all these forms became signs of k and g; but by a later differentiation ♪ Z etc. were restricted to the representation of the sound g, the others continuing to represent k. Cf. Berger, Histoire de l'écriture2, p. 336.] = There was, thus, no loss of old zeta and no invention of G. And now that we look back upon it, we cannot but wonder that we never found it strange that a new letter should not have been placed at the end of the alphabet, as Y and Z were, or next to the letter out of which it was supposed to be evolved, as J and U were in modern times but that a good snug place was reserved for it all those centuries in the middle of the alphabet by the accommodating old seta. III. It remains for me to say a few words as to the Z found in the Oscan text written in Roman characters on the Bantine Tablet, as well as the Z used in Latin in the spelling of Greek names (page 27, above). We have seen that this Z cannot possibly go back to an early Latin zeta. It is generally assumed that the use of Z in writing Greek names (a practice that began in the time of Augustus) was derived directly from the Greek of that time, and Mommsen (Unteritalische Dialekte |