the other darsanas derived from the speculative character of their contents; its scope is limited and the nature of the investigations in which it is engaged leaves no room for high flights of the imagination. But it possesses counterbalancing advantages. Its subject matter is of a positive nature, its method is sound, and its reasoning in most case convincing.'

• This, I think, is needlessly exaggerated in its depreciation of the speculative side of the system. Thibaut was presumably thinking of the work he was dealing with, the Arthasaṁgraha, which like the Mimāñsā Nyāya Prakāśa (see note 2) does not deal with that side.

THE IE ROOT *meik-: *meiĝ- AND AVESTAN mīzān

MARIA WILKINS SMITH

TEMPLE UNIVERSITY

Certain forms of the Greek word for 'mix' which have y instead of K as the final of their root-syllable, e.g., the present μlyvoμ and the 2nd aorist éμlyny, appear to be anomalous when compared not only with the other forms of the same Greek verb but also with its cognates in other Indo-European languages, where all forms point to k as the primitive IE root-final. Boisacq1 derives μlyvvμι from pIE *meik- and states, without further comment, that 'le grec a la gutturale douce'. Walde,2 s. v. misceo, gives the Greek cognate as 'mit Media', but makes no remark on the phenomenon except to reject Wackernagel's theory of an independent origin for ulyvvμt. Apparently, then, Boisacq and Walde explain the disagreement in root-finals by the theory of the socalled variation of the root-determinative.1

Another possible explanation is that some word-contamination affected certain of the Greek forms, but did not reach or, at any rate, did not affect the rest. This solution of a given linguistic riddle is easier to assume than to prove, and none such has yet been found to throw light on the present problem.

A third possibility is that the pIE root was originally not *meik- at all, but *meiĝ-, weak grade *miĝ-, and that the few Greek forms are, with an exception to be mentioned presently, the only ones to show the original root-final unchanged. With this as a working hypothesis, we need not depend on the theory of a varying root-determinative to explain incongruities in Greek and other languages, but we may simply apply the ordinary principles of sound-change. Most verbal suffixes begin with either s or t, both of which would cause the preceding voiced

1 E. Boisacq, Dictionnaire étymologique de la langue grecque 637-8 (Paris, 1916). 2 A. Walde, Lateinisches etymologisches Wörterbuch2 488 (Heidelberg, 1910). 3 J. Wackernagel in Kuhn's Zeitschrift für vergleichende Sprachforschung 33. 39. 4 P. Persson, Studien zur Lehre von der Wurzelerweiterung und Wurzelvariation, Upsala, 1891.

THE IE ROOT *meik-: *meiĝ- AND AVESTAN mizān

5

179

root-final to become voiceless. Thus, for example, *miĝ- followed by the suffix of an s-present, an s-aorist, a sko-present, a to-participle, would become *miks-, *miks-, *mi(k)sko-, *mikto-, respectively. The comparatively few forms in which the change from voiced to voiceless root-final cannot be explained by the initial sound of the suffix may fairly be credited to the influence of analogy. Greek is the only language in which the present stem-formation offers definite resistance instead of assistance to such a change, and the Greek preservation of the original root-final may be thus explained, in μίγνυμι and in Homeric μίσγω.

It was said above that only one word outside of Greek had preserved the voiced root-final unchanged. That word is Avestan mīzān, found in Yasna 44. 20. The text of the stanza is as follows:

nōit him mizon ašā vāstrǝm frādainhē

Bartholomae translates thus: 'Sind denn wohl die Daeva's gute Herrscher gewesen? Ich will die danach fragen, die sehen, wie ihretwegen der Karpan und der Usig das Rind der Raserei preisgeben, und wie der Kavay es unaufhörlich jammern macht, statt dass sie es hegen, um durch Aša die Landwirtschaft zu fördern'.

9

The verb mīzōn is generally connected with maez- 'urinate', cognate3 with Sanskrit mehati, Latin mingere, Greek oμixeîv, from *meiĝh-, original ĝh becoming Avestan z. But the aspirate and non-aspirate palatal stops fell together in the primitive Iranian period, so that Avestan z in mīzān may equally well represent an original ĝ. Since phonology does not serve to identify mizen with its pIE source, let us examine the semantic possibilities. Here there has been much discussion but little agreement among scholars. Söderblom10 has listed in some detail

5 For this important step in my argument I am indebted to Prof. Roland G. Kent of the University of Pennsylvania.

• Transliteration as in Chr. Bartholomae, Altiranisches Wörterbuch, Strassburg, 1904.

7 Die Gatha's des Awesta, Strassburg, 1905.

8 Boisacq 700.

'K. Brugmann, Grundriss der vergleichenden Grammatik der indogermanischen Sprachen 1. §§612-3 (Strassburg, 1897).

10 N. Söderblom, 'Note sur l'agriculture dans l'Avesta' in Revue de l'histoire des religions 59. 333–7.

the previous interpretations. All but those of the Pahlavi tradition and of Bartholomae understand a reference to irrigation by rain, or by urine either of gods or of cattle. The traditional interpretation, connecting mīzān with mizd- 'reward', is impossible on etymological grounds. Bartholomae11 sets up a new root 'maēz- 'hegen', whose nearest cognate is Germanic schmeichen 'schön tun'. Söderblom takes it literally as 'urinate', citing the use of urine as a fertilizer in certain parts of Switzerland, and quoting, from Fargard 3 of the Vendidad, verse 6 which states that one of the best and most agreeable spots on earth is 'wo am meisten harnen Kleinvieh und Grossvieh'.12

Now since, as was said above, mīzān can be derived from *meiĝ-: *miĝ- 'mix', I suggest setting up an Avestan root maēz- 'mix', of which mizin is the 3rd plural imperfect active used with transitive meaning. Its Greek cognate μívvvμ means, in the passive, 'to have intercourse with, to be united to'.13 Thus mizan acquires the meaning given below in my translation:14

'How, O wisdom, have the daeva-adherents been good rulers? But this I ask (of those) who see how for them, indeed, the Karapan and the Usig have given the herd to passion, and how the Kavi has caused(it-)to-mourn continually; nor have they mated it for the furthering-of pasturage in accordance with justice'.

The sense of the last line agrees with what precedes, and with the well-known fact that Zoroaster preached the importance and the value of herding as the economic feature of his religious reform.15 Furthermore, verse 5 of the Vendidad Fargard already cited names as another of the earth's most agreeable spots that where, to quote Wolff's translation,16 'man am meisten Kleinvieh und Grossvieh züchtet'.

11 AiW 1108.

12 F. Wolff, Avesta.....übersetzt 326 (Strassburg, 1910).

13 Liddell & Scott, Greek-English Lexicon 966, med.

14 This is my own translation, shortly to appear as part of my doctoral dissertation, and differs in certain points from that of Bartholomae given above. 15 Cf. R. G. Kent, 'Cattle-tending and Agriculture in the Avesta' in Journal of the American Oriental Society 39. 329-33.

16 Av. übersetzt, 1. c.

LACHMANN'S LAW OF VOWEL LENGTHENING

ROLAND G. KENT

UNIVERSITY OF PENNSYLVANIA

Apparently without guiding principle, the perfect participles passive of certain verbs in Latin have a short radical vowel, while others have a long vowel: thus factus, spectus, scissus, to facio, specio, scindō; but āctus, rēctus, vīsus, to agō, regō, video. An excellent summary of the discussion which has grown up around this point was given by Sommer1 in 1914; but for the proper understanding of the problem it will be necessary briefly to repeat the main points.

Lachmann in 1850 set forth the formulation that the vowel of the root is lengthened in the participle if the root ends in a voiced consonant, but is not lengthened if the root ends in a voiceless consonant; he supported this in part by reference to Gellius 9.6 and 12.3, and by other data for the quantities.

Osthoff3 in 1884 pointed out the instances which were in opposition to Lachmann's formulation, and sought to explain the lengthenings as due to the analogy of the perfect indicative active, where in many verbs the long vowel was inherited from the original speech, or by other analogies; but he did not deal with all the examples.

Holgar Pedersen in 1896 tried to make out of Lachmann's rule of thumb a law of phonetic development: that when in the participle a voiced stop became voiceless a preceding vowel was lengthened in Italic. He restricted the phenomenon to roots in which there was an original non-aspirate stop; but even then he had three notable exceptions, fissus, scissus, sessus to findō, scindō, sedeō, which he explained as remade for participles in -no-, cf. the Sanskrit cognates bhinna-, chinna-, sanna-.6

1 Handbuch der lateinischen Laut- und Formenlehre2 122-3.

2 In his commentary to Lucretius 1.805 (fourth ed., 1882).

3 Zur Geschichte des Perfects im Indogermanischen 110–7.

♦ Nordisk Tidskrift for Filologi, 3d series, 5. 32–8.

'The curious chronological inconsistency of this, in comparison with the results in other Indo-European branches, seems not to have occurred to him.

6 But in fact Sanksrit has sattá- alongside sanná-.