of many phrases of modern Italian, which involve the prepositions and the sentence-introducing conjunctions (see above, p. 151), it will be sufficient here to refer to Bährens, praef. Poet. Lat. min., I, p. XII, and to L. Müller, Res Metr.2, p. 579, for the treatment of 'in arca' (v. 5), ‘ut vere' (11) 'ego sum' (10: enclisis of the substantive verb) as single words of six letters each in the ingeniously constructed verses of the carmina duodecim sapientium (Bährens, IV, p. 120 f.). Two monosyllables are often similarly treated, as aes est (ibid., v. 5); for iamnunc and sivis similarly counted as single words, v. L. Müller, 1. 1., p. 581. The effect of the traditional word-order, in causing the combination of monosyllabic conjunction and pronoun to be felt as a single word, is further seen in the following: Corp. Gloss. Lat. IV, 22, 4 astilla verum illa femininum est; ib. IV 480, 18. Similar is the statement of the grammarians that a conjunction, like at (ad), may be 'prefixed' to any case of a noun or to any verb, while a preposition, like ad (at), can commonly be 'prefixed' to one case of a noun only or to a verb through composition (Prisc., K. III 25, 24f.; Audax VII 351, 17 f.; Suppl. LI). Thus we apparently have a play upon the two uses of the prefixed ad or at in Poe. 544:

Át trepidate sáltem: nam vos ddproperare haud póstulo,

i. e., at-saltem (like at—tamen) and adproperare. The editors commonly correct to attrepidate, which seems unnecessary.-Again the close connection between the monosyllable and the following word in pronunciation is indicated by the complaint of Consentius (Keil, V 395, 7) that in the pronunciation of some sic ludit was indistinguishable from si cludit. As is well known (Lindsay, L. L., pp. 122, 215), the monosyllables *cord, *terr, *ess, med, ted, hisc(e), hosc(e), etc., long retained their final consonant in early Latin before an initial vowel, while haud (Caper, K. VII 96, 4) and *hocc (Velius Longus, K. VII 54, 6) never lost the final consonant in this position, e. g. hocc erat, alma parens'; CIL. IX 60, 3 terminus hicc est. Cf. also Skutsch, Forsch., p. 60 f.

ELMIRA, N. Y.

R. S. RADFford.

III. STUDIES IN ETYMOLOGY, II.

In KZ. 36, 103 Pedersen presents a brief and interesting discussion of the phenomenon of Greek - Skr. y- and Latin j-. In place of the usual transcription for the primitive Aryan speech of spirant j and semivowel (the latter for cases where the spiritus asper stands in Greek, e. g. in rap: Skr. yákṛt-, Latin jecur) he favorably considers Havet's transcription by zz and i, respectively; and makes the suggestion, exempli causa, that is a reduction of gi-, noting that this initial group has not been found, though ki and ghi- are attested, the latter by Skr. hyás, Gr. xoés yesterday' (cf. Latin heri, without trace of the i). In the history of the Germanic dialects, on the other hand, -yy gave rise variously to -ddy- (Gothic) and -ggy- (Old Norse). In this paper I shall use y for i, and the Old English guttural spirant for j.

The examples for the phenomenon under discussion, extracted from Brugmann's Griechische Grammatik § 115, are, so far as Latin is concerned, the following: (vyóv, Skr. yugám, Lat. jugum 'yoke', with their verb systems; (úun 'leaven', Skr. yūşam 'soup', Lat. jūs. True, objection to the latter cognation has been raised of late by Bally (Mém. Soc. Ling. 12, 314), on semantic grounds, viz.: that Cúμn 'leaven' was alien in signification to the other words, which mean 'soup', but this objection seems to me irrelevant as long as Latin fermentum 'leaven' comes from fervet 'boils', while French bouillon 'soup' similarly comes from bouillir 'to boil':

=

The above examples show of course that j- in certain Latin words corresponds to - in Greek' and to y- in Sanskrit, but they are valid only for the equation of Latin ju- = Greek (v- Skr. yu-, and do not further prove that Latin je-, say, would be the normal correspondent for Gr. e- Skr. ya-.

1 I cannot doubt but that this - is in some cases the product of DY-, e. g., in (vyóv yoke' and worós 'belted', both ultimately cognate, I take it, with déel binds, Skr. dy-dti, from a root DE(Y). With a Sanskrit pair sydti 'binds': syħi-ma 'band', we might infer beside dy-áti · binds', a * dyū-ma, cognate with Skr. yauti binds', with lost initial d. Projecting this conclusion back on the primitive period we get a base (D)YĒW-.

For all we know, primitive ge- may have had a different history in Latin from gu-, just as in Old English the normal representation of primitive zu- and yu- is iu-, but primitive ge- and ye- yield ge(cf. Sievers-Cook, Old English Grammar, § 175).

These considerations lead me to propose the following etymologies, in which Gr. ge- and Skr. ya- will correspond to Latin ge-.

(1) Lat. gemini 'twins', Skr. yamás 'coupled', Old Irish emuin 'gemini', from a Celtic stem *jemno-s.

(2) Lat. gestit 'desires eagerly, burns' [cf. fervet 'boils, desires eagerly'; furit 'boils (Aeneid, i. 107), desires eagerly' (Horace, Carm. i. 15, 27)], Skr. yásyati 'becomes hot, (boils), exerts oneself, strives'. For the semantic chain, cf. also Old Bulg. kypěti 'to seethe', Skr. kúpyati 'is angry', Latin cupit 'desires'. If these etymologies are to be rejected, and gemini: yamás, in spite of Weber's advocacy, has long been rejected, it will not be because the words in question show any incompatibility on the side of their signification.

If the proposition that Latin ge- represents primitive ze- should arouse in our minds the hostility of surprise, this hostility may perhaps be dispelled by noting, in addition to the Old English analogies mentioned above, that the reduction of gye- (I speak now in the terms of Pedersen's hypothesis) to Latin ge- may be compared with the equivalence of Latin he- (in heri) with Skr. hya- in hyás (cf. Gr. xôés: primitive *ghyes, *gyhes).

Difficulties still remain, however, for solution. Latin aemulus 'rival' and imitatur ' rivals, imitates' have been paired with Skr. yamás; aemulus being by some, e. g. Uhlenbeck, ai. Woert., s. v. yamás (cf. Hirt, Ablaut No. 654), explained from a base ayem-: while Thurneysen (KZ. 32, 566), who has meantime omitted his explanation from the Thesaurus, derived it from *ad-yemolos. But the cognation of aemulus and imitor with Skr. yamás is so far from simple that there would be little room to hesitate about preferring the cognation I am defending between gemini and yamás, if this were all. On the other hand, gemini has been compared with Skr. vi-jáman- 'related, corresponding', jāmis 'leiblich verschwistert', and Thurneysen (1. c., footnote) would find in gemini a contamination of both these derivations, that is, make it akin to both yamás and jāmis. Inasmuch as English kin is so apt a translation for both vi-jaman- and jāmis I cannot bring myself to separate this pair of words, -ja-man- and jā-mís

(with suffixal m), from the root jan and from Avestan zāmi-, 'posterity, children' (cf. German kinder, English kin, both cognates of this root). This consideration leaves the preference for Old Ir. emuin Latin gemini: yamás, always provided that Latin ge- be proved the equivalent of Skr. ya-, Gr. (e-. The second etymology suggested, viz. gestit, Skr. yásyati, Gr. ¿éeɩ ('boils, boils with passion'), though giving room for no valid objection on the semantic side, calls for an account on the morphological side why -- has been added to ges-. In view of the narrow range of the -te- suffix in Latin verb inflexion (but cf. Feist's explanation of sentire in his Gotische Etym. No. 495; and now Brugmann in I. F., 15, 76), we should perhaps explain the -t- of gestire as left over from an iterative inflection, gestare, attracted to the flexional type of the synonymous verb cupire; cf. the double inflexion of the verb concretely synonymous with gestire, viz. bullare, bullire, 'to bubble, boil'.

(3) gerit 'raises, bears'. An obvious objection would be felt to the separation of gestit from gerit. But after all, does our explanation of gestit separate it from gerit any more than the two already lie apart in point of meaning? A satisfactory etymology of gerit still halts (not Osthoff's [a]g-es-, at any rate).

I cannot satisfy myself with the derivation of gestit,-which exhibits but two senses (1) 'desires, cupit' (2) 'is eager, fervet',— from the noun gestus,' with the signification of 'gesticulates'.' It is just as impossible, starting with gestit 'cupit, fervet' to account for gerit with the signification of 'bears, carries', etc. But it is open to us, by mediating between the two, to try and derive gerit and gestit from a common source.3

I take it that fee 'boils' exhibits, not a primary, but a secondary sense of the root -zes-. The Germanic cognates (most con

1 This difficulty touches the meaning, not the form, as artire' to joint' stands beside artus a joint'. Otto (in I. F. 15, 25 sq.) supposes that derivatives in -tire and -tāre stood freely beside one another, setting up, e. g. a type *captire/ captare [? contaminated in low Latin *captiare, whence French chasser], for which he cites only artire, later artāre, as an actual authentication: a better pair were the coeval gestire/gestare, both kept alive because of their different meaning.

2 Pace Sittl, die Gebärden der Griechen und Römer, p. 10. In a passage like Plautus, Bacch. 596, ita dentifrangibula haec meis manibus gestiunt, 'heaving' is an interpretation as apt as 'gesticulating'.

3 Cf. the approximate semantic parallel in Lith. grëbiù 'rapio': Lett. gribēt 'velle'.

veniently examined in Kluge's Etym. Woert. s. v. gären) mean 'to ferment, foam'; among these are German gischt, English yeast, cf., for the signification, English 'leaven', colloquial 'rising', and German 'die hefen', all of which suggest that the bubbling of boiling and fermentation may have got its name from a verb meaning to raise' (trans.), 'rise' (intrans.). With a primary notion of heat in this verb we have nothing to do, any more than in Latin bullit 'bubbles'. Traces of the sense 'raises, lifts' are not altogether absent in derivatives of gerit, for instance in agger 'mound', congestus 'pile', suggestus 'platform'; perhaps also the ancient phrase re bene gesta (Persa, 754, and often) meant originally much the same thing as praedam tollere 'to lift plunder'. It is very easy, if we start with 'lifts' as the primary notion of gerit, to account for its subsequent development as a synonym of fert 'bears', for the perfect of fert is tulit, which means precisely 'lifts'.

It happens very appositely that the development of significations here assumed is of record in the Germanic languages in the history of the verb heben—,and here I will follow closely Paul's Deutsches Woerterbuch, s. v. heben-which seems originally to have signified 'to grasp, seize' but already in primitive Germanic exhibited the general sense of (1) 'lifts, raises'; specialized in (2) English heaves, used of 'the bubbling and boiling of the swollen sea'; and in (3) German die Hefen, a word for 'leaven, rising'. A corresponding classification of the progeny of primitive zes- yields (1) Lat. gerit 'lifts, raises, bears'; (2) Gr. (ée 'heaves, boils, bubbles', corresponding, in little, to the usage cited for English heaves,-and so, in the figurative use only, does Lat. gestit fervet, cupit'; (3) English yeast 'die Hefen'. Old High German jesan 'to ferment, foam' and yeasty in the phrase "yeasty wave" show the close relation of (3) and (2), while yeasty in "yeasty spirit" suggests Latin gestit.1

(4) gemma, germen, 'bud, sprout, button (of a plant'); gisma 'annulus'. The derivation of this pair of words from primitive

'I permit myself a passing suggestion to the effect that the g- of gären and gischt may constitute for German also a record of a difference between the history of primitive Ze- and ye-, if we might assume that the j- which is recorded in the older Germanic forms was an inexact orthography for a sound that was neither a guttural spirant (7) nor a semivowel (y) precisely, but a tertium quid whose earlier orthography with j- was quasi rectified by another approximation with g-.