theory of a 'Dactylic paroxytone law' there is pointed out the behavior of the accent in active oxytone compounds with trochaic endings like κυνηγός, μελοποιός, etc., and recessive compounds like ἀντίφωνος, etc. Here we find that the alternative is strict, either oxytone or proparoxytone-not even properispomenon is resorted to except in certain categories which doubtless have their own explanation. There was also drawn into connection with this the most striking deviation in Greek from the so-called Three Morae Law.' In such a sentence, for example, as οἱ δ ̓ ἄνθρωποι έμπίπτουσιν ἔξωθεν, there occurs in three of the five words this skipping of the long penult in violation of the 'three morae' principle. Whatever may be the ultimate explanation of all these phenomena, it does not seem to me that we can as yet go back of a simple repugnance of the language to accenting the penult in words ending with this trochaic cadence.2

In the words in question we find : —

that is, where the genitive already gave a properispomenon, no change was made; but, as changing the nom. and dat. into properispomena was evidently out of the question, the accent was retracted and we find the familiar type of proparoxytone with trochaic cadence. The accusative éμéye, again, gave no difficulty; the tribrach endings, as I have shown in the article cited, fare alike with the dactyls as far as paroxytonesis is

1 I venture still to call it a 'theory,' although Hirt, following Brugmann's weighty authority, speaks of it as 'eine Entdeckung.' But inasmuch as Hirt (p. 28) says that Wheeler has no phonetic explanation to offer further than die Neigung der Sprache . . . dem Hochton eine lange Silbe immer vorauszuschicken,' it would seem safer, however enticing this law may appear, still to reserve final assent. I have shown (1.c.) that Brugmann's essentially modified statement (i.e. bloss an die Kürze der vorletzten Silbe sich anknüpfend'; see Wheeler, p. 61, note) is consistent with facts about tribrach and other endings; these, on the other hand, so obstinately contradict Wheeler's 'Dactylic Law' as to leave, in my judgment, the burden of proof still upon the shoulders of its adherents.

2 Another explanation, which reduces the penult to one and one-half morae, is advanced by Hirt (p. 37), but not with any very great confidence.

concerned.

pyrrhics.

The point in common is that they are both

Hirt assumes an *eyo and *epot as preserved in the enclitic combination, and that eye and époí were accented anew after the analogy of ἐμέ.

The lack of agreement between such types as Grk, eidos and Skr. védas (Bloomfield, A. J. P. IX. p. 25, which is quoted with approval by Hirt) shows, it might also be urged, how inexorable was the tendency to avoid paroxytonesis in trochaic endings, and this much is plain whether it be attributed with Bloomfield to the recessive tendency or otherwise explained with Hirt. À propos of μnrnp-or Hirt's theoretical *μητήρ - μητρός may be mentioned the curious accent of the compound Δημήτηρ, Δήμητρος, where the accent refuses to remain oxytone or to follow the position of that of the nominative and to become properispomenon *Anunтpos in the genitive.1

Doubtless it would be more satisfactory if we could establish a survival of a more ancient *eyw and *μo, than it is to fall back thus upon this unexplained tendency to favor a certain cadence. There are other trochaic combinations of enclitics, indeed, that apparently conquer this repugnance and show paroxytone accent, — καίτοι, ἤτοι, τοιγάρτοι, τοίνυν, καίπερ, ὥσπερ, ἤπερ, ὥστε, εἴθε, and εἴτε These forms are difficult to explain. It may be claimed that the momentum of a common word like κaí is such as to forbid the change of accent in the new and temporary combination with the enclitic, and perhaps this is a sufficient explanation. In none of them, moreover, except Touɣápтoi, would recession beyond the long penult be possible. It is a temptation, however, to try to explain some, at least, of these on other grounds.

The recent but now generally accepted-explanation of the accent of (properispomenon) oiko (nom.) as compared with (paroxytone) oйkoɩ (loc.) and with the long quantity of the ultima in the opt. mood, by calling in the undertone, drawled (schleifender) accent from the Lithuanian, suggests

1 Bopp, Accentuationssystem, p. 20.

a possible explanation for the -Toɩ form, and Wackernagel's new explanation of a neglected tradition of the Greek grammarians would tempt one to explain in a similar way those compounded with -vvv and -Tep, if the derivation and other considerations would permit.

If oko (loc.), though scanned as a trochee, was accented as a spondee by virtue of the 'drawled' (schleifender) tone (cf. the perispomenon 'Iσ0μoî), why may not "To — certainly the accent predominating over τo (cf. Wackernagel, p. 21)

have been at least reinforced for the same reason? It is, at least, remarkable that the circumflexed with -To becomes paroxytone. The distinction made between To and тo by tradition is none too clear. Although -70 be identified with ooí, it may nevertheless be entitled to a 'drawled' tone. Wackernagel (1.c. note to p. 20), in speaking of the circumflex on of, says: "Im Grunde ist nicht sein (i.e. of of) Zirkumflex, sondern der Akut der beiden anderen Formen (i.e. σoí, μoí) auffällig, da die Dativ-Locative auf -o sonst zirkumflektiert werden."

If this hypothesis were true for ro, then for Kairo and TоLуáρTOι also. In the latter, as has been said above, a posτοιγάρτοι sible proparoxytonesis is avoided.

For τοίνυν, καίπερ, etc., an explanation might seem to be opened up by Wackernagel's theory to explain the accent of enclitic combinations like ἔνθά ποτε. This double accent of évoa, etc., is handed down by the grammarians (see Chandler, § 965), although modern editors generally (but cf. Wheeler, p. 128) have ignored it. Wackernagel, appealing for a parallel to the Litauischer Zircumflex' where a circumflex is found with one foot, so to speak, resting on a vowel and the other on a following liquid or nasal, claims that combinations like ἔνθά ποτε, φύλλα τε, λάμπέ τε, θάρσός μοι, really had a circumflex on the first syllable, though it was graphically unprovided for in the Greek signs. If this were admitted as a working principle on Greek ground, one might be tempted thus to explain the paroxytone Toívy and even to extend it to кaíтeр, etc., were it not for the probable history of -Tep as an enclitic form of I.E. péri. In addition to this objection Professor

Bloomfield would not, as I infer from an informal conversation, accept as proved on Greek ground this Baltic-Slavonic circumflex on a vowel + liquid or nasal, with which Wackernagel operates to explain 0ά Tоre, etc. Hirt, however (p. 38), accepts Wackernagel's contention with enthusiasm, and not only says, "wir müssten eigentlich evdóv Te schreiben, wobei alles klar wird," but he actually finds it necessary to account for the properispomenon olkov co-existing with the 'drawled' tone on the ultima, and he says (1.c.), "als notwendige Folgerung ist nun aufzustellen, dass die einfachen langen Vokale mit Stosston a, w für den Akzent kurz waren, es müsste eben so gut x@pa, wie oiko geschrieben werden." Certainly this is going still further than the suggestions here offered with all due hesitation.

There would still remain ὥστε, εἴτε, εἴθε, and the strange vaix, etc., though wσre and ere may perhaps be sufficiently explained as composed of the union of proclitics and enclitics.

One other contention made by Wackernagel in his Beiträge raises an objection that he does not seem to have noticed. He contends (Beitr. p. 21 ff.) that the Greeks really pronounced every accented ultima before an enclitic with the acute accent even where the circumflex was written thus ἀγαθοῦ τινος was pronounced ἀγαθού τινος. This would avoid the violation of the 'three morae' law which occurs in aya@oû TIVOS, where the circumflex really brings the accent four morae from the end of the combination. He says (p. 21): “Da ein wirklicher Zirkumflex nicht zwei unbetonte Silben hinter sich haben kann, muss in solchen Fällen ein Akut gesprochen und der Zirkumflex bloss darum geschrieben worden sein, weil er der betr. Form auch sonst eigen war." But in the case of ὧντινων, even if one were to assume *ὧντινων for the pronunciation, we should still have an equally inadmissible proparoxytone with a long ultima.

His statement of the accentuation of the last syllable before an enclitic must therefore be made more sweeping or this explanation must yield to another.

Finally, to return to the terminations with trochaic cadence, Wackernagel's convincing application of the enclisis

of the finite verb, upon which is built up his whole theory of the 'recessive accent,' deals notably with a case of the skipping over the trochee. The whole paradigm of eiμí exemplifies it, and the forms onμí and onoí- certainly the most prominent of that paradigm are of the same measure: in their case it was oxytone1 or nothing, unless under certain exceptional conditions.

So, too, among the enclitic pronouns the only trochaic forms, σowé and opwív, are oxytone when accented at all.

1 Cf. eg. the anomalous ri pnμí; Soph. O. T. 1471.