was then transferred to positions within the sentence, where before voiced stops a voiced ḥ resulted which was lost with compensatory lengthening. Thus *-os became *-oḥ and then -ō before voiced stops. The further extension of this final -ō to other positions in the sentence and its substitution for *-ē2 from *-es has been treated in masterly fashion by Bloomfield (Amer. Journ. Philol., iii. 32-39), and needs no comment here.

It is important, however, that IE. -ōs became Skt. -ās and appears before voiced initials as -a, not as -ō, while IE. -Ŏs, though appearing in Skt. as -us, has the form -ō before sonants. Similarly, in Avestan, IE. final -ās, -ēs, -ōs appear as -å, while IE. final -ŏs, etc. appear as -ō. This difference of treatment can be explained in only one way: The IE. long ō, è became a in primitive Aryan before the short o, ě became ă. In the period intervening between the change of the long vowels and the change of the short vowels, the loss of h before voiced stops produced -ō, -ē from the short final -os, -ěs, and a from the long vowel followed by s. These new long vowels -ō and -e were not subject to a further change to -a, since the law converting IE. ō, e to Aryan ā had already ceased operating.

As to the lengthening resulting from the loss of zor ž before d or dh in Skt., it has already been remarked that this change does not appear in Avestan; thus Skt. nédīyas-, Av. nazdyah-; Skt. miyedha-, Av. myazda- etc.3 We have here ample proof that the quality of IE. ĕ was maintained distinct down to the time of this specifically Sanskrit change.4

1 Bartholomae, KZ. xxix. 572 ff.; Brugmann, Grundr.2, i. 886. A similar suggestion by Brockhaus (1842): "Sollte nicht vielleicht auch der Nominativ auf -as wie -os ausgesprochen worden sein, dessen dumpfes s vor allen tönenden Buchstaben verloren gehen mußte, aber wie fast immer im Sanskrit, wenn ein Consonant abfällt, dieser durch die Verlängerung des vorhergehenden Vocals ersetzt wird, auf diese Weise aus -os die Form -ô wurde", (Zt. Kunde d. Morg. iv. 85).

2 On the debatable question whether the Aryan possessed such an -ē (from -es) see Wackernagel, Altind. Gr., i. 338; Brugmann, Grundr.2, i. 886, § 1005, 5, note.

3 Cf. the examples in Wackernagel, Altind. Gr., i. p. 37, § 34; p. 274, § 237, b, ß; Brugmann, Grundr.2, i. 735; KVG., p. 545, § 710, 2; Joh. Schmidt, KZ., xxv. 60 ff.; Bloomfield, AJP. iii. 27 ff.

4 Seeming exceptions like Skt. sáḍhar- for *sēḍhar- (from *seŷh-tor-) have a by analogy of the ǎ in other forms and derivatives. Cf. Bloomfield, AJP, iii. 30; Wackernagel, Altind. Gr. i. p. 38 (middle), § 34; p. 44, § 40.

That the quality of ŏ was similarly maintained can hardly be proved, since the final -ō received a great extension, even in the middle of words. But words with an original long vowel show consistently -ād(h)-, never -ēd(h)- nor -ōd(h)-: thus, śādhi, śaśādhi, cakādhi; ádhvam, śádhvam, árādhvam; ādaghná- (for *ōz-do, cf. Lat ōs).2

The conclusions are therefore:

1. That IE. ē, ō became a in the primitive Aryan period before the loss of final -h before voiced stops.

2. That IE. ě, o kept their quality until after the loss of final -h before voiced initial stops.

3. That IE. ě certainly, and IE. Ŏ presumably, kept their quality until after the Indian loss of z and with compensatory lengthening before voiced dental and cerebral stops.

For convenience the following chronological table of the changes is presented:

Primitive Aryan Changes.

1. Palatalization of gutturals by following palatal vowels and i.

2. Change of IE. ē, ō to a, becoming identical with IE. a. 3. Loss of final -h before initial voiced consonants, giving new ĕ and ō.

1 The ō of so-dasa "sixteen", sō-dhá "sixfold" (Wackernagel, Altind. Gr. i. p. 38, § 34, b) may be the result of such an extension. The post-vedic sodhum, sõḍhar- from the root sah is certainly an imitation of the Vedic vóḍhum from vah (Bloomfield, AJP., iii. 30). Vedic vódhum etc. (Skt. váhati, Av. vazati, Lat. vehit) cannot have IE. Ŏ and must be due to some analogy; there are some parallel forms of this root with zero grade showing uḍh- from *užḍh- : ūḍhvam (by the side of voḍhvam), udhá- (by the side of voḍha-), ūḍhvā und uḍhi-; a comparison with certain forms of ruh such as rūḍhá- by the side of rodhum, rūḍhvá by the side of rodhar- suggests a starting point for an analogy-formation.

2 Whitney, Skt Gr.3, § 166; Wackernagel, Altind. Gr., i. p. 273, § 237, a, a; Brugmann, Grundr.2, i. 735, § 830, a. According to the rule (allowed by Panini and required by the Prātiśākhyas) that the first consonant of a group be doubled (Whitney, Skt. Gr.3 § 229; Wackernagel, Altind. Gr. i. p. 112, § 98, a) the spelling addhvam etc. is frequent.

The Peshitta Text of Gen. 32,25.-By RICHARD GOTTHEIL, Professor in Columbia University, New York City.

In preparing the final part of my edition of the second half of the glosses of Bar Ali, I stumbled over the following:

i. e. "The flank became torpid or benumbed". The dictionaries try to explain the word Aia, which evidently comes from the passage Gen. 32, 25, as from the root, to change, to be altered-adding "for the worse", to be displaced, dislocated. So Mrs. Margoliouth in her Compendious Syriac Dictionary, p. 586. Audo, Dictionnaire de la langue Chaldéenne, II, 584: 2 — bilo x lia :qşoj \\glɔ qui ai Vsamo sol

i. e. to become slack; to lose the power of speech through laughter; his leg or his hand lost its cunning.

But these meanings are all derived from the passage in Genesis in its faulty tradition. And yet that tradition is quite old. Not only is the reading found in the Mss. used by Le Jay for the Paris Polyglott and by Walton for the London Polyglott, and taken over by Lee in his reprint for the London Bible Society. It is also to be found in the Urmia Edition printed by the American missionaries. That these are based upon good Ms. authority, may be seen from the fact that the Jacobite seventh century Ms. in the Ambrosian Library at Milan reads. Aia, as does the excellent Nestorian Massoretic Ms. in the British Museum (Add. 12, 138 fol. 15b) of the year 899 A. D., which the Rev. G. Margoliouth has been kind enough to look up for me:

Bar Ebhrāyā, in his scholia to the passage, is careful to punc

,See Uhry) ܘܩܪܒ ܠܚܪܘܬܗ ܘܫܢܳܬ ܒܙܩܦ ܢܘܢ ;tuate the word

Die Scholien des ... Barhebraeus zur Genesis Capitel 21-50, Strassburg 1898, p. 12); and in the sixteenth century Abhdisho

264 Richard Gottheil, The Peshitta Text of Gen. 32,25. [1913.

of Jazarta wrote in his heptasyllabic poem dealing with Syriac homonyms: Aia soos? aloin ollo (Hoffmann, Opuscula Nestoriana, p. 68, 8).

Of course, the correct reading is Aajo, Vas become weak, feeble, debilitated, torpid; and this reading was not unknown in certain parts of the Syriac Church. The scholarly grammarian and writer Jacob of Edessa (seventh century) reads correctly? aloj Aaĵo (see von Lengerke, Commentatio de Ephraemo Syro, p. 20)—a reading which has been preserved in another place by Bar Ali himself (ed. Gottheil p. 108, 6) and by Bar Bahlül (ed. Duval, col. 1283, falsely punctuated A). I might add that Payne-Smith in his Thesaurus, though seemingly suspecting the reading (see cols. 1360 s. v. Lo and 2471 s. v. ), has listed the passage also s. v. ha col. 4233).

The accepted reading in the texts has not failed to lead scholars astray; which is a signal proof how necessary a correct edition of the Peshitta is. Ball, in his edition of the Hebrew text for the Polychrome Bible (1896) p. 91, has the following note: ó éváρкnσev which became dumb. 6 thus renders

$ הנשה .cfr הַשְׁנֶה and pointed השנה v. 25. Here f evidently read ותקע

S Niao

=

.25 .v ותקע

In this connection I might mention the curious translation. in the Polyglotts of the Syriac rendering for a TA lia. They translate "nervum muliebrum" which they have gotten from a false punctuation of the Arabic rendering

"the sciatic nerve"- a common expression in Arabic medical works. They punctuates!!-which PayneSmith has also incorporated in his Thesaurus.