p. 33) brings the use of Z on the Bantine Tablet into connection with this, to which Planta (I. p. 72, ft.) rightly objects that the text is too old for that, Mommsen himself placing it between 129 and 118 B.C., and others still earlier. But there

are other reasons for supposing that this Z was not derived directly from Greece. At first the Romans represented both Greeks and Greek ≈ by or S, just as we found that the Italians generally represented s and ≈ by > or S. When they later used Z in writing Greek names, it was not to transliterate the Greek letter Z, but to distinguish the sound z (whether written in Greek with a Z or a ≤) from the sound s: ZMVRNAE, CIL. VI. 3, No. 16030, etc. (for collections of such cases see Seelmann, p. 315, and Stolz, I., p. 85, § 73-74). This spelling surely does not reflect a Greek text; in ZMVRNAE we have not only Z for Greek ≤, but also the Latin spelling V rather than the Greek Y. If the Romans used the letter Z in this way, it is clear that it was to them the sign for the sound and not simply a transliteration of the Greek letter Z. The use is identical with that in the Oscan text written in Roman letters on the Bantine Tablet. While this cannot be derived from the Greek, neither can it be derived from the Oscan zeta; for the Oscan zeta was I not Z, and spelled the sounds is not, and the Oscans used retrograde >, that is and Z for both s and ≈. usage have arisen in Latin, for Latin no longer sound having passed into r. Let us examine the matter more closely. The Italic dialects represented both s and ≈ by S or 2. We saw that this was also true of early Latin and of the Latin treatment of Greek words up to the time of Augustus. Now, it would not have been strange, even without the special reasons that I shall state directly, had the diversity of the symbols (≤S≥2) used to represent the two sounds s and tempted writers here and there to differentiate and, while retaining S or for s, to use or 2 for z In exactly this way we find C used for k and for g in Praenestine (Conway, I. § 281, p. 313). That the rustic should become the monumental Z is just what was to be expected (compare the change of ELM into ELN). In fact, there

Nor can the had a ≈, this

already was a distinct approach to the form Z, especially in Oscan. Compare the ZZZ on Zvetaieff's Plate V. No. 1, and the in No. 5. On Plate II. the letter is in many cases more like Z than like., especially in B, line 23. Compare also the Z-like forms in XIX., particularly line 19, end, 22, 23. The Faliscan letter in his No. 345 Conway says "is rather square (Z)." We have also seen that on the coins of the Etruscan town Cosa the letter looks so much like Z that some authorities regard it as such, while others think it a retrograde ≤ (page 27, above). Both are right in a sense: the letter is in form Z (and so accidentally identical with Greek Z), but it is by development only a differentiated form of ≥2≤S (cf. below).

and Z,

But where can this differentiation have arisen? We saw that it did not arise in the native non-Latin alphabets and that it could not have arisen in Latin, where there was no z to be represented. The differentiation doubtless arose just where we first find it, namely, in one or more of the Italic dialects that had the sounds s and z but used the Roman alphabet. These conditions specially favored the differentiation Ss Z. The Umbrians recognized in Latin rounded S their own rounded 2, and, as they used the latter for both s and, so they used S for both sounds when they employed the Latin alphabet. The Oscans could, and to some extent did, do the same. But when the Oscans began to use the Latin alphabet, the established Oscan forms were and the established Latin form was S. To the Oscans, S was not simply a reversed Z, but a new letter. They learned it in Latin as the symbol for, the sound s, and for that only (as the sound ≈ did not exist in Latin at the time); but their native Z was to them the sign of both and s. What, then, was more natural than that they should, when writing Oscan with Latin letters, be tempted to use Latin S for the sound s, as it was used in Latin, but to employ the native Z to represent the native -sound, for which the Latin alphabet offered no symbol? So too in COZA(NO) and ONAZOƆ we find the native Z employed, regardless of the direction of the writing; while in COSA(NO) there is a complete yielding to the Latin, in form and direction.

A similar problem arose in writing Umbrian with the Latin alphabet, which had no sign for the Umbrian fricative usually spelled d. In the Iguvine Tables (Conway, p. 399, etc.) an S, usually with a diacritic, thus Ś, takes the place of the native sign. But in the Picene inscriptions in the Latin alphabet (Conway, p. 449) the sound is represented by 4, a form differing from the usual d (see page 34) fully as much as the Z of the Bantine Tablet differs from the usual Oscan >. Parallel with this introduction of native or Z and or into the Latin alphabet is the introduction of the native digamma [ into the Ionic alphabet when the latter was adopted by the Greeks of Tarentum (Conway, p. 461).

To judge from the age of the Bantine Tablet, we may estimate the rise of the differentiation Ss Zz at about 140 B.C., that is, fully a hundred years before the Romans ceased to write Greek names with S for Greek Z. That this use of the Latin alphabet in spelling Italic dialects should, in the course of time, extend to the spelling of Greek names in Latin was but natural, especially when the form of the letter used to represent the sound ≈ chanced to coincide with that most frequently employed to represent z in Greek. Nor should it surprise us that Latin scholars came to look upon this Z as the Greek seta and, on the model of it, introduced also Greek Y. Thus the older ZMVRNA was displaced by ZMYRNA (CIL. VI. 3989-90) with, however, the interesting retention of the Italian Z.

To recapitulate:

(1) As Latin did not possess the affricate d≈ or ts, the Greek seta was an idle letter in the Latin alphabet. As gamma, in the Western form, became confounded with kkappa, so too did zeta, in the Italian form. After a period of confusion, a differentiation took place, whereby the use of kappa was much restricted, or C became the sign for the k-sound, and or G the sign for g.

(2) The letter Z appears in Italy first in the writing of Italic dialects in the Latin alphabet. It is a natural development of the native and was used to represent the native z-sound, while Latin S was employed, as in Latin, for the

s-sound only. Later the use of Z extended to the spelling of the sound in Greek names in Latin, whether spelled Z or < in Greek.

IV. APPENDIX.

THE coceulod oricso OF THE SALIAN HYMN.

Among the many puzzles presented by the fragments of the Salian Hymns none seems to have tempted so many and baffled so many as the group of letters usually given as cozenlodorieso. The chief solutions offered (mostly taken from Maurenbrecher, Carminum Saliarium Reliquiae) are as

These attempts are certainly anything but satisfactory : they all contain in themselves their own condemnation. In fact, we cannot but imagine the god Zeul-Zaul-Zol, who has thus been conjured up, as enjoying the joke as much as any of us.

Spengel gives the evidence of the manuscripts as follows, ignoring spacing:

We need concern ourselves with the first five readings only.1 And here it is clear that the only real diversity lies in the

1 The reading of Laetus is evidently based on b and B, or their kin.

third letter. The problem might have been approached from this point, but I shall present the matter in the way I actually proceeded, and shall return to this phase of it later.

It is apparent that the group cosculodorieso, to take the usual reading, is made up of more than one word. Most scholars, misled by the aural suggestion of adōro, have put the d with the following letters. Considering the fruitlessness of the attempt, it occurred to me that the d might belong to the preceding o and be the ending of an ablative, and so I divided the group into cozeulod orieso. Now, if od is the ablative ending, zeul must be the stem; but if zeul is the stem, the only likely explanation of co is that it is a reduplicated syllable, for the attempt to make of it the prefix co(n)has proved unsuccessful. But, if it is a reduplicated syllable, we must look for the identity of c and z, and one of the two must be wrong. As z in early Latin would be an anomaly (see page 24 etc.), I decided for c and concluded that we should read with the Basel manuscript coceulod orieso.

It then appeared that the whole difficulty was solved; for coceulod orieso is perfect early Latin and corresponds exactly to classical cuculo oriere. The subject of the development of weak o before the stress has not yet, so far as I know, been cleared up (Brugmann 2, I. § 243, 3, and middle of p. 974), but, on the analogy of weak o> u after the stress (Brugmann 2, I. $244, 2), we should expect it to become u. With coceulod compare also κόκκυ ‘the cry of the cuckoo and κόκκυξ KÓKKŪyos ‘a cuckoo.' The change of eu to ū is normal (Brugmann, I. § 218). The loss of d (Stolz, p. 343, § 363) and the change of s>z>r in orieso> orière (Stolz, p. 276, § 274) are well-known matters; in fact, it was the latter point that Varro was illustrating by the quotation. But the form orieso brings us very welcome information. It has been customary to identify Latin -re with original so, whereby Latin sequere <sequeso would be identical with Greek ËTOV <ËTEO < *éπeσо (Brugmann1, I. § 81; II. § 1047, 2; Henry, Comparative Grammar of Greek and Latin, § 34, A8, 260, 2, 267; Stolz, Historische Grammatik der lateinischen Sprache, p. 119-120, 352, 11; Lindsay, Latin Language,