sturtevant - the gutturals in hittite and ie

7/28/2019 Sturtevant - The Gutturals in Hittite and IE

1/17

Linguistic Society of America

The Gutturals in Hittite and Indo-EuropeanAuthor(s): E. H. SturtevantReviewed work(s):Source: Language, Vol. 6, No. 3 (Sep., 1930), pp. 213-228Published by: Linguistic Society of AmericaStable URL: http://www.jstor.org/stable/409218 .

Accessed: 28/01/2012 01:32

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

of scholarship. For more information about JSTOR, please contact [email protected].

Linguistic Society of America is collaborating with JSTOR to digitize, preserve and extend access toLanguage.

http://www.jstor.org
http://www.jstor.org/action/showPublisher?publisherCode=lsahttp://www.jstor.org/stable/409218?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/409218?origin=JSTOR-pdfhttp://www.jstor.org/action/showPublisher?publisherCode=lsa


2/17

THE GUTTURALS IN HITTITE AND INDO-EUROPEANE. H. STURTEVANT

YALE UNIVERSITY[Hittite shows no trace of assibilation or palatalization of the gutturals,and so the development of the Indo-Europeanpalatals seems to have beensubsequent to the Indo-Hittite period. Hittite shows an independentphoneme (u or M)corresponding to the labial element of the IE labio-velars, and this state of affairs must be assumed for IH also. The lossof this labial sound had begun under certain circumstances in IH, and itcontinued in primitive IE and in the historic IE languages.]

In common with a few other scholars I have for several years assumedthat Hittite is related to the Indo-European languages only by virtueof the descent of primitive Indo-European and Hittite from a commonancestral language, and I have recently adduced proofs of the correct-ness of this view.' The new doctrine is of considerable importance,since, if it is accepted, it will enable us to reconstruct an earlier parentspeech, for which I have suggested the name Indo-Hittite, and especiallybecause it will provide an unhoped-for check upon numerous theoriesabout primitive IE and the Pre-Indo-European period as far back as IH.In the present paper I shall collect all the Hittite words I knowwhich are cognate with IE words containing gutturals, in the hope ofclearing up some of the vexed questions connected with the history ofthese sounds. We shall soon have much more material of this sort atour disposal, since our understanding of the Hittite vocabulary israpidly improving; but enough is already available to show the mainoutlines of the story. It will be convenient to group the material underthe familiar IE symbols as far as that is possible, and to use these sym-bols in reconstructing III words. We shall find, when the evidence isall in, that the IH system of gutturals was far simpler than that of IE(see p. 224); but for the present 'we shall for convenience assume the IEstate of affairsfor IH also.

1Transactions of the American Philological Association 60. 25-37.213


3/17

214 E. H. STURTEVANT

Hitt. kardiag (gen.), kardi (dat.) 'heart' is cognate with Gk. Kapala,Lat. cor, Lithuanian szirdis 'heart'.2Hitt. katta(n) 'along side, with; down; afterwards' is cognate with Gk.KaT", as I have tried to show,3 but my former treatment requires cor-rection at several points. It is now clear that the syllabic liquids andnasals yielded Hitt. ar, al, am, and an before consonants as well as beforevowels4 (arnuzzi 'cause to go' = Skt. prnotimove'; kardia' = Gk. Kapita,Kpa6LfM; alh- 'strike' = Gk. a3XX; -an [acc. sing. ending of consonant-stems] = Skt. -am, Gk. -a; anzag'us': Gothic uns, etc.). Consequentlyboth KaT"aand katta(n) represent an IH form which we may tentativelywrite **"fzt, while KaraT and katti represent **"ktai. Hitt. kuttag'side' (?) and kuttar 'neck' (?) cannot be connected (correct LANG.4.125). The reason for assuming an IE palatal rather than a pure velaris the connection with IE *de-/t 'ten' and kltom 'hundred' (see AJP48. 247 ff.).Hitt. -kan 'ov' I have connected (AJP 48. 254-7) with Latin cum,on the basis of IE *kom, which I take to be a shorter form of the stem*fomt, 1lpt, just discussed.

The deponent verb, kitta, kittari 'he lies', must be identical with Skt..eteand Gk. KE7Tra.5Hitt. wek- (firstsing. 'i-e-ik-mi, third sing. Ui-e-ik-zi) ask, demand, beg'has been convincingly identified with Skt. vadmi, vafti 'be eager,desire, command' and Gk. &Kivwilling' by Friedrich.6Hitt. takg- (third sing. tdg-ga-as-si, tdg-gi-es-zi, tdk-ki*-i*s-zi,dk-ki-iz-zi)'join, build, do' whence taksan 'together', takian 'joint, junction',taksatar 'union',7 tak"ul 'friendly; peace', taksula(i)- 'make peace',tak"ulatar'peace', is clearly the same word as Skt. taxti 'builds', Lat.texit 'weaves'. The word is interesting as indicating that the highlyspecialized meanings of Skt. taksd, Gk. TrEKrwvcarpenter' and of Lith.taszyti 'hew' are secondary, and also because it sides with the majorityof the IE languages against Gk. in showing a sibilant rather than a stop-sound as the root final.

2 See Hroznr, Journal of the Society for Oriental Research 6. 691; G6tze, Zeit-schriftfur Assyriologie 34. 183; Zimmern,Streitberg-Festgabe, 39.3 American Journal of Philology48. 249-53 (1927).* Cf. LANGUAGE 6. 30.5 See LANG. 6, 26.6IndogermanischeForschungen41. 369f. (1923);cf. Sommer,Boghazkii-Studien7. 55 and fn. 2.SSee Sommer,BoSt. 7. 352;Friedrich, ZA NF 1. 179;2. 49.


4/17

GUTTURALS N HITTITE AND INDO-EUROPEAN 215With somewhat less confidence I suggest that ni(n)k- 'rise, be exalted,

satisfy one's self (with drink)' and nini(n)k- 'raise, take'8 may beidentified with Skt. aBnoti 'obtains, attains', na'ati 'obtains', Gk.EVE7KELP 'carry', Lat. nanciscor'get'. If so the basic meanings sug-gested for this root by Walde-Pokorny9 require some revision.Possibly one should also connect Hitt. yuga" 'yearling' and the secondelement of dayugas 'two-year-old' with Skt. yuvasas 'young', Goth.juggs 'young' and Lat. iuvencus 'young bullock'. If so we must assumeloss of the nasal in the Hitt. word, although the combinations ank, enk,and ink certainly survived. The meaning of yuga' must originally havebeen similar to that of Lat. iuvencus; but the word came to be appliedexclusively to very young animals, and then dayugas was formed (nodoubt on the basis of some popular etymology that escapes us). Onemay compare also Lat. vetus 'old' beside Gk. &rosyear'.Hitt. kammari (dat.-loc.) 'bee-hive'"os certainly to be connected withGk. Ka~cipa'vault', and Goth. himins 'heaven'." I am inclined to seeanother meaning of the same Hitt. word in kammaras 'light, sunshine',which Ehelolf,'2 on the contrary, regards as an assimilated form ofkalmarag. If I am right, Goth. himins preserves the most primitivemeaning of the stem. There is no way of determining whether theinitial IE sound was k or q.Hitt. hark- 'have' may belong to the root of Gk. ApKEOprotect, wardoff' and Lat. arceo 'ward off, protect'. This root may have had eitherk or q in primitive IE.

Hitt. genu 'knee' was convincingly identified with Lat. genu, Av.iznu-and their cognates by Friedrich.'3I have connected (LANG. 3. 164 f.) Hitt. ak- (third sing. aki, third pl.pret. eker) 'die' with Lat. igit 'he has done' (Skt. aj-, Av. az-).Hitt. harkil 'white' has been plausibly connected by Kurylowiczwith Skt. arjunas 'white', rajatam 'silver', Gk. apybs 'white', ip'yvpos'silver', Lat. argentum 'silver', etc."4

8 See Gatze, Madduwatta' 115-8.9 VergleichendesWbrterbuch erIndogermanischenSprachen 1. 128.10HrozngS,CodeHittite 70; Zimmern,OrientalistischeLiteratur-Zeitung25. 299(1922).11See Boisacq, Dictionnaire Rtymologique de la Langue Grecque, s.V.Ka,&pa.2 Ap. Friedrich ZA NF 5. 80 f.13IF 41. 372-6. See also ZA NF 5. 37f.

14 J. Kurylowicz, Symbolae Grammaticaen Honorem oanis Rozwadowski101.


5/17

216 E. H. STURTEVANTHitt. mekkis' 'much, great', plural 'many', must be cognate eitherwith Gk. pAyas,Lat. magnus, and Goth. mikils, which represent IE

*meg-or *meg,or with Skt. maha, mahi, Av. maz- 'great', which representIE *meoh-. EhThe clearest Hittite evidence for IE h is presented by gimmanza'winter' = Skt. hemantas,Gk. xeTia, etc. (See LANG.3. 119.)I am inclined to see another instance in gimral 'field' (see Friedrich,ZA NF 1. 180 f.): Gk. xalal 'on the ground', Lat. humus 'ground'.Pedersen's'5 connection of Hitt. tekan, dagan 'ground' with the other

form of this stem, which is represented by Gk. x0cv, Skt. ksas, etc.,cannot be accepted. The Hitt. orthography might possibly indicatea pronunciation tkan, and such a metathesis of the two consonants inthe initial group is plausible enough; but, as we noticed above (p. 214),Hitt. shows ks'for Gk. Kr= Skt. ks (takG-: Skt. taksa, Gk. T'KT7oV).A different treatment of the aspirated group is improbable. For apossible etymology of Hitt. tekan, see below (p. 227).Hitt parkus' 'high'"6must be identified with the root of Skt. brhant'high'. The nt- suffix is widely extended in IE, but a number of wordslack it, e.g. Skt. barhas 'tail-feather' especially of a peacock, Av. barazi-,baraz- high', Arm. berj 'height'.Hitt. pankug 'widespread, all, whole' as a substantive 'multitude,totality, populace, army', has beside it a ro-stem (or ri-stem?), pangarit(instrumental) 'in force', whence the preterit middle pangariyattatti'became numerous'.17 Hitt. pankus corresponds precisely with Skt.bahus 'much, many' and Gk. raxis 'thick, large' on the basis of IE*bhn his.Hitt. kessar 'hand' = Gk. Xelp (see LANG. 3. 121) may contain eithergh or gh.Hitt. laki 'causes to fall', laganza 'prone', (middle) lagari, lagaittari,lagaru 'fall, lie',18and saliga 'avyyiylra'9 must be cognate with Gk.I formerly (LANG. . 119)comparedGoth. bairhts bright', etc., on the assumptionthat initial bh became Hitt. h. I expect to show soon that this assumption isfalse. No further account will be taken in this paper of the etymologies that Ihave based upon it.

15Le Groupementdes Dialectes Indo-Europdens41 f.18 See G6tze, KF 1. 109, fn. 1, and references.17See Friedrich, ZA NF 2. 279 and references;Gdtze, Madd. 114f.18 See Forrer, Forschungen1. 183ff.19 Hroznr, CH 146.51, 53; Sommer and Ehelolf, BoSt. 10. 66. I do not under-


6/17


7/17

218 E. H. STURTEVANT'approach' (G6tze, Hatt. 57 ff.). This u-stem adjective is obviouslycognate, except for the stem-vowel, with Skt. manak 'a little', Tocharianmenki 'less', Lith. menkas 'little', and Germ. mangeln. The formalsimilarity to our Hitt. stem inclines one to assign also Skt. manku?'weak, tottering' to this root, in spite of Walde-Pokorny, VWIS 2. 268.

gHitt. yukan (i-zi-kdn- 2 Boghazkdi-Textein Umschrift 14 2. 6) 'yoke'is obviously the same word as Skt. yugam, Gk. ?vyb,, etc. G6tze (IF42. 327 f.) has held that this is an Indic loan-word; but the reason whichhe alleges (final n in a neuter a-stem) is not valid."5Hitt. hatk- 'shut' (Sommer, BoSt. 7. 1-6) may contain the prefix ha-= IE *o- and the nil-grade of the root of Lat. tego 'cover'.

ghHitt. dalugaes (nom. pl.) 'long' and dalugaiti 'length' are cognate withSkt. dirghas, Church Slavonic dligi, Gk. o6XLx6slong'.26 Forrer andFriedrich regard the abstract noun as a loan word, but their reason (therarity of the suffix) is scarely conclusive, and there is no reason at allfor suspecting the adjective.Hitt. linkzi 'swears', lingais' 'oath', etc. goes well with Gk. AM'yXw'test, put to the proof'. If we connect also Skt. langhati 'leap, mount,escape, transgress', the final consonant of the root must have been gh;for the Hitt. and Gk. words Mhwould be equally possible.

q?Hitt. shows the relative-interrogative-indefinite most frequently asan i-stem (cf. Lat. quis, quid, Gk. r1s, etc.). We have the pronominal

forms kuil, kuit, kuin, (nom. pl.) kuel, which are directly traceable tothis stem, as well as a number of compounds like kuilki 'quisquam',kuissa 'quisque', kuiskuil 'quisquis', and the conjunctions kuit 'because,that, when' and kuitman 'while'. The neuter plural kue probably cor-responds to IE *qUai.27 The o-stem appears in kuwat [kwat] 'why' andkuwatta[kwata] 'how many' (?), which are probably cognate respectively

25 See AJP 48, 249f., LANG. 5, 140f. Onthe initial consonant, see LANG. 4. 231.28 See Hrozn?, SH 23; Forrer, Zeitschrift der Deutschen Morgenlandischen

Gesellschaft76. 262. Sturtevant, LANG.3. 121,4. 231; Friedrich, ZA NF 5. 35 andfn. 5.27See LANG.6. 25.


8/17

GUTTURALS IN HITTITE AND INDO-EUROPEAN 219with Lat. quod and quot.28 The same stem must also lie at the basis ofkuwali 'while' (?),29kuwapi 'where, when, usquam, unquam',30 kuwapit'where, whither',31and kuwaten'why' (?).32 We are not now concernedwith the final syllables of these words; but I should like to suggest inpassing the comparison of kuwali with Lat. qualis, of kuwapi with Lat.quippe, and of kuwatenwith Gk. 1rbOEso HroznjV).Another clear instance of qYbefore a vowel has been recognized ingakuwa (i.e., s'akwa) 'eyes', which is cognate with IE *oqU-'eye' (Gk.6ove, oculus, etc.) or with *seqU- 'see' (Goth. sailvan, etc.).33 Quitepossibly the two IE roots are related to each other in the same way asGk. arTiyosand Tr-os, etc. For the connection between sakuwa and?ak- 'know', see below, p. 226. If Friedrich34s right in understandingakkusa to mean 'pits' (for capturing wild animals), the word may repre-sent the same stem without the initial s. The meaning is close to thatof Gk. bnr 'hole', and the extension in s appears in Skt. aksi (gen.aki.zas) 'eye', and elsewhere.The verb eku- 'drink, give to drink' shows the following forms: e-ku-zi, e-ku-uz-zi 'he drinks', e-ku-ut-te-ni,e-ku-wa-te-ni (KUB 1. 16. 3. 34)'you drink', a-ku-wa-an-zi 'they drink', e-ku-ir, e-ku-i-e-er 'they drank',e-ku 'drink', e-ku-ut-te-en 'drink ye', a-ku-wa-an-du 'let them drink',a-ku-wa-an-na 'to drink', a-ku-wa-an-na-dy 'of drinking', ak-ku-ul-ki-iz-zi 'he keeps drinking', etc.35 There is also an agent noun a-ku-ut-ta-ra-as, e-ku-ut-ta-ra-ag'one who gives drink'.36 Possibly ekunas 'kalt' andekunimas 'Kailte'37belong to this root. Hrozny' connected Hitt. aku-,eku- with Lat. aqua, Goth. alva 'water, etc.; but several scholars weredoubtful about the equation, since they could not find any relatedverb-forms in the IE languages. Now that Pedersen38has found such

28 See AJP 50. 363 f.29 Delaporte, Grammairede la Langue Hittite 63 (?214).30 Sommerand Ehelolf, BoSt. 10. 54; Gotze, Hatt. 86, 124; Friedrich, Staatsver-trige 2. 141.31Sommer and Ehelolf, BoSt. 10. 54.32Hrozn4, SH 146 and fn. 5, 148.33Cf. LANG. 3. 163, AJP 50. 364 and fn. 15.34ZA NF 5. 39. f. (1929).35For citations and discussions of this verb, see Hrozn?, SH 42 f., 61 f.; Zim-mern, Streitberg-Festgabe439; Ehelolf, Kleinadiatische Forschungen1. 137-142;Friedrich, Staatsvertr 176.36 See Sommer, BoSt. 7. 60.3' See Gitze, KIF. 1. 186.38Le Groupementdes Dialectes Indo-Europ6ennes40.


9/17

220 E. H. STURTEVANTa verb in Tocharian joka 'he drinks', ma jokalle 'one should not drink'(cf. joko 'thirst'), the objection does not hold. OIsl. xger 'sea' requiresthe ablaut 8:a,39and so we must assume for IH, a conjugation **eqYti:aqXYnti. The Hitt. verb shows an analogical spread of the vocalism ofthe singular into the second (and presumably the first) person plural,while the third plural retains a (


10/17

GUTTURALS IN HITTITE AND INDO-EUROPEAN 221identification of the latter word with. Gk. palvw on the basis of IE*gU~jio- is less attractive than it was. Perhaps we should rather traceboth the Lat. and the Hitt. words to IH **gUemio-,while Gk. palvwmaycontain *gVai-, nil-grade of the root seen in rtqvand Skt. agam, asFick45has suggested.At any rate the root of Gk. ePYvDoric Epa~ and Skt. agamis to be seenin Hitt. wa-, we- 'go, come',46 which presents among others the followingforms: (first singular) i-wa-mi [wami], (third singular) U-iz-zi, i-e-iz-zi,t-wa-az-zi [wetsi], (third plural) %-wa-an'zi,first singular pret.) ui-wa-nu-un, (third singular) Ui-it, (third plural) i-e-ir. The verb had evi-dently gone over, at least in part, to the thematic conjugation.I have identified Hitt. walh- 'strike' with Gk. ahXXwhit, throw'.47The IH root was **g~Uelh-cf. Arcadian rzabXXovrEs). he Hitt. presentwalhzi (wa-la-ah-zi) comes from **gu1h-ti,and Gk. pdcXw rom **g'Ulh-io-.Hitt. wars-, war.iya- 'mulcEre, be or become propitious; reap (afield)'48 may be connected with the ablaut base **g erx- 'praise',whence Skt. gradti'he praises', gartas 'welcome', Lat. gratus, etc. Ifso the Hitt. word represents IH **gqY-s-with a formative s akin to theaorist sign of IE.I know of only one occurrence of the verb '-en-zi (KBo. 6. 26. 3. 33 =Hrozn?, CH 144 = Zimmern-Friedrich, Hethitische Gesetze 30 ?77);but it must mean 'he has sexual intercourse with, violates'.49 Perhapsthe word is cognate with Skt. jani, Gk. ~yvv',pava 'woman'.Hitt. nekumanza 'naked' must somehow be connected with Lat.n'dus and Goth. naqaps, which represent IE *nogtedhos. G6tze, Madd.120, has suggested that in this word kum is written for w, and I ratherhastily followed him (AJP 50, 363). The difficulty with the theory isthat in the above words Hitt. shows w corresponding to IE gUor gUhbe-fore a vowel. Before a consonant, however, we should expect ku, as inkunanzi 'they smite' (see below, p. 222). It is, therefore, better to seein nekumanza the suffix -men or -ment, which, as I have shown in AJP

46Glotta 2. 128-30. Cf. Walde, Lateinisches Etymologiches Wirterbuch2 .v.venio.16So Meillet, Memoires de la Socite' de Linguistique de Paris 23. 258 (1929).For the Hitt. word, see HroznV,SH 243f.; G6tze, Hatt. 84, Madd.603. For uwa =wa, see Sturtevant, AJP 50, 363-5.'' LANG. 3. 220; 4. 160.

48 See Friedrich, ZA NF 3. 202;G6tze, KlF 1. 192-7.9 See Friedrich, ZA NF 5. 572.


11/17

222 E. H. STURTEVANT50. 360-9, arose in IH from earlier -uen and -yent in case they werepreceded by u or a. Although nekumanza is an nt-stem (e.g., 2 BoTU.9. 4. 8 = KBo. 3. 23. 2. 8: ne-ku-ma-an-ti), a transfer to this declensionis not uncommon in Hitt. (e.g., gimmanza 'winter': Skt. hemantas).That the dental of the Germanic and Lat. words is formative is shownby Skt. nagnas, Lith. ngas, and Ch. Si. nagii. Hitt. nekumanzamay,therefore, come from an IH stem **negumn-,50which probably alter-nated with **ngumen-.A transfer of the latter to the o-declension might well yield IE *fgum-nd-, and that, in turn, would become Gk. *,yv/v6s. Now we haveevidence in Hesychius for a Gk. synonym: d'roXbyiaros-rroyVbwIV"-KbrpLoL;XvYvbsov6vbs. No doubt -roXb-yiaross genitive of *d?bXvypa,and Xvpvbss *Xvybs influenced by -yvv'bseither at the hands of somescribe or in the dialect from which the word is cited. Now *&rb6XvyIaand *Xvybs may confidently be connected with Xvyil? 'bend, twist,turn'.5'1 The prefix iro6must originally have been essential to the mean-ing 'strip', but the adjective Xv/vbs hows that it came in time to be feltas a mere intensive which might be omitted (as in A&rXXvpL,aro0vf)KwO,a7OKX~rlrw,tc.). This situation, I would suggest, may have led to thefeeling that initial a of *dyvvbs was also an intensive, as in arXbos,a8p6os, 7ev7is,52nd then yvivu6was substituted for a*yv/pbsto match*Xvy~6sbeside *d-roXvy/os.

g hHitt. war- 'burn' (intransitive), warnu- 'cause to burn'53contains

the root **g~her'hot, be hot, heat, fire', which appears in Skt. gharmas'heat', haras 'heat, anger', Gk. OEpAbswarm', OBpossummer', Lat. formus'warm', Ch. S1.gor ti 'burn', etc.Hitt. kuen-, kun- 'strike, kill' (e.g., third singular ku-en-zi, third pluralku-na-an-zi, third singular pret. ku-en-ta, infinitive ku-na-an-na) isidentified by all scholars with Skt. hanti 'he strikes, kills', plural ghnanti,

etc. As is shown by war- 'burn' from **gYhr-,wemiya- 'come upon'from **gVemio-,wenzi 'has intercourse with' from **gVnti, etc., IH60 For gu instead of gy, see below p. 224.5' For cognates in other languages, see Walde-Pokorny, VWIS 2. 413 f.62Possibly this processwas assisted by phrases in which *-rvpv6s was preceded

by a word ending in a; cf. Thucydides 3. 23 rd -yvuC'&,iodorus 1. 76 Tr irp&ypara53See Sommer and Ehelolf, BoSt. 10. 10; Friedrich, ZA NF 2. 48 f.; Sommer,KlF 1. 120-4.


12/17

GUTTURALS IN HITTITE AND INDO-EUROPEAN 223**gYhentishould yield Hitt. *wenzi, but the corresponding third pluralkunanzi from **gYhnentiquite regularly shows ku before a consonant; cf.ekuzzi, nekuz, nekumanza. It is, therefore, clear that the ku of kuenziis analogical.Hitt. danku?'dark'54corresponds with Gk. vbr6osgloom' and loavepri?'violet-dark' except that it has nil-grade of the root and no suffix. Justas kuw corresponds to 7r in kuwat r6brEpossee above, p. 218), so kucorrespondsto phere.

Since Hitt. is related to the IE languages only through IH, the evi-dence just cited brings complete certainty on certain features of IEwhich have hitherto been in doubt.

(1) Since Hitt. shows k for IE k, ^, and ^h,there can be no furtherdoubt that these were stops in IE. Otherwise we shall have to assumeeither that 6 became k independently in Hitt. and in the centum-groupof IE, or that IH k > IE 6 > Lat. k. Either hypothesis is too improb-able for serious consideration.65(2) More important is the proof that both IE and IH knew a distinc-tion between pure velars and labio-velars. For it is incredible that thedistinction could have been developed independently by Hitt. and theIE centum-languages. We must assume, therefore, that the satam-languages lost the labialization either independently or in a satam-dialect of IE. Since even the centum-languages tend to lose the labiali-zation, there is no difficulty with the former assumption. (Cf., forexample, Gk. K~sbeside ris, and the total loss of labialization in ModernFrench).

(3) Reichelt, IF. 40. 57-61 (1922), has shown that IE labiovelars arepreserved before consonants only in Gk., where they appear as 7r,0 or ?p. Elsewhere the labilization is lost, and so Reichelt is inclined tothink that, in Gk., forms like r~rrs6 and riqw (instead of *7EKrT6 nd*7re'w)are analogical. In Hitt., on the contrary, we have found ku foran original labio-velar before consonants in ekuzzi, ekutteni, ekutten,akkugkizzi, akuttara?, ekuttaras, nekuz, nekumanza, kunanzi, kunanna,and danku?, and perhaps in akkusa, ekunas, and ekunima?. The ques-tion arises how these words were pronounced. Our conventional

54 See G6tze, Madd., 82 and references.55This proof is not really needed since the demonstration by Hermann, KuhnsZeitschrift iir VergleichendeSprachforschung41.38-46 (1907)andBloomfield,AJP32. 36-57 (1911),that primitive Indic and primitive Indo-Iranian had stop-soundsin this series.


13/17

224 E. H. STURTEVANTtreatment of the IE labio-velars may dispose us to pronounce ekuzzi intwo syllables, although the history of the cuneiform writing demandsthree syllables. If the etymology suggested above (p. 221) for neku-manza 'naked' is correct, it seems to furnish an answer to the question;for a change of **negY-Meno **negU-mens scarcely possible.Having thus decided upon the trisyllable pronunciation of nekumanza[ne-ku-mants], I was surprised to discover the resulting harmonybetween this word and Gk. -yvuEbs,hich I have already shown in detail.A similar instance would probably appear if we should find the pluralof wenzi, which should be *kunanzi < **gunenti corresponding to Gk.yvwi. One may thus suppose that yvv7i rom *gund and Boeotian parafrom *gYtdshow equally normal forms of the nil-grade in the first sylla-ble. Parallel examples from the IE languages are Gk. aKbvXa'youngdog': Norwegian skvaldra 'bark continuously', Lith. skalikas 'barkingdog', etc.;56 Gk. aKVba'lust', aKVEVbe in heat': sanhrETaKvUt. 'AxaLol(Hesych.); Gk. KbKXOS:Skt. cakras 'wheel'; Gk. UAaxbssmall': kXacppbs'light';57Gk. ppaxbs 'short': Lat. brevis;Skt. kutra, kuha 'where', Lat.si-cubi 'if anywhere', Cretan 6irvt whither', Syracusan 7Ir3'whither' (con-tamination of stems *Kv-and ro-): *qYo-,*qUe-;58 at. gurdus 'dolt': Gk.Opabs 'slow'; Lat. gurges'abyss, whirlpool':vorare devour', Gk.ofp&OaKco,etc. (cf. Specht, KZ 51.11; Leumann, Glotta18. 274).(4) If we must thus assume for IH an ablaut qye: qu, it seems tofollow that the IE labio-velars were originally nothing but k-soundsfollowed by consonantal y. This is the conclusion at which Hirt, Idg.Gramn., 1. 231 f., has arrived from quite other premises; the Hitt.evidence confirms his reasoning completely. It is probable, however,that in primitive IE qY, gU, and gUhwere independent phonemes; andso I propose to retain these traditional orthographies while using in-stead ky, gy, and ghy for IH. Accordingly the term labio-velar properlyapplies to IE, but not to IH. To avoid misunderstanding I haveseveral times written IH reconstructed forms with the symbols gy etc.Henceforth I shall restrict such symbols to IE, and write ku or ky, etc.,in IH.The comparative rarity in the IE languages of Kv,cu, etc., for a labio-

566 ee Persson, Bezzenberger's Beitrdge zur Kunde der IndogermanischenSprachen 19. 275 f.5 See Osthoff,MorphologischeUntersuchungen6. 1-69.58 Several less certain examples are cited by Hirt, IndogermanischeGrammatik1. 232. Cf. also Reichelt, IF 40. 62 ff.


14/17

GUTTURALS IN HITTITE AND INDO-EUROPEAN 225velar before a consonant and the frequent substitution of somethingelse (e.g., Gk. rerr'bs, Lat. coctus), make it likely that at an early periodanalogy tended to get rid of the syllabic u of ku beside kye. Possiblythe syllabic inequality of such verb-forms as IH **ghuyntiand **ghuninti(Hitt. kuenzi, kunanzi) provoked re-formation, and it is probable thatthe resulting third plural would be from the start **ghnenti (cf. Skt.ghndnti). At any rate some factor or factors tended to eliminate Yafter k, g, and gh even in IH. For that appears to be the explanation ofthe loss of labialization in Hitt. as well as in IE in the following cases:Brugmann, Grund.21. 596 f., wrote: "Schwer zu beurteilen ist eineAnzahl von Formen mit k, die zu den Pronominastaimmen *qYo-,*qYi-geh6ren: ion. KOS6-KOST67pOSKp; iol. 0-KaL;dor. -Kain 6-Ka; hess. KlIKWd-KE ite-Kl; allgemeingr. 7rOXXa-KLKLS 'oft' = ai. ved. purfi cid und Kal'und' = lit. kai aksl. ce'."59 In Hittite the labialization of this stem islost in the particle -ki or -ka, or kuiski, kuiska 'quisquam'. The variantkuigku proves, if proof were needed, that the particle belongs to therelative-interrogative-indefinite stem."6 In view of the compound pro-noun kuis'kuis'6 one can scarcely ascribe kuifki to dissimilation.Hitt. kuer-, kur- 'cut, shave' (third singular ku-ir-zi, third pluralku-ra-an-zi, third singular pret. ku-e-ir-ta, third plural imper. ku-ra-an-du)62evidently belongs with Gk. KELPocut, shave', Lat. caro 'meat',corium 'hide', Umbr. karu 'pars', Skt. sarman 'hide', etc. The Hitt.word is clear evidence for an initial ku, but all trace of the labializationhas vanished from the IE languages. On the other hand, the extensionof this root with t (Skt. kyntati, Lith. kertu, kirsti, etc.) appears in Hitt.without labialization in kars- 'cut off, keep away, banish'.6" The final Sof kars'-regularly developed from t before t of the third singular endings(KBo. 6.11.1.24: kar-dg-zi, KUB 8.51.2.8: kar-ds4-ta). I have pointedout g of this origin in ses- 'sleep' = IE. *sed- 'sit',6 and I shall soonhave occasion to cite further examples of the change.As I have shown (LANG.3. 162 ff.), there is no doubt that Hitt. sak-,lek- 'know' is connected both with sakuwa 'eyes' (see above, p. 219)and with Goth. saiIvan, sakv,s8Ivum'see'. I was mistaken, however, in

5 Cf. Brugmann-Thumb, GriechischeGrammatik4 37 f.6s These three particles are combined with the IE enclitic *ne in Skt. -cana,Av. -cina, OHG -gin; and Goth. -hun. It is safe to say that the last-named hada pure velar in IE. (*qu-ne), and probably the others had also (*qe-ne, *qi-ne).61See Hrozn?, SH 147.62See Forrer, Forschungen1. 196.63 See Friedrich, ZA NF 2. 46 f., G6tze, Madd.45 f., 62.64See LANG.6. 27.


15/17

226 E. H. STURTEVANTthinking that the loss of the labialization in Hitt. was regular beforeconsonants. In view of the instances listed above of Hitt. ku before aconsonant in roots which show a labio-velar in IE, it is more probablethat the root was inherited by Hitt. in both forms (saky-, sak-). Theformer is to be recognized not only in the substantive but also in theparticiple, if Sayce (Journal of the Royal Asiatic Society 1930. 305) isright in translating sakuwan estu (KUB 17. 10. 2. 16) by 'let it becomevisible'. The passage is too fragmentary for complete certainty, but noother possible interpretation occurs to me. In the IE languages, onthe other hand, only the form with labio-velar is usually recognized.Aside from the meaning 'see' of Goth. saihvan,etc.,

there is a more wide-spread meaning 'say' in Gk. ~Aivr, ~eicre^v, Lat. inseque, Lith. seku,sekti, etc. In spite of Walde-Pokorny, VWIS 2. 479, it seems probablethat the latter meaning along with the delabialized form of the root isto be recognized in OE insiht, Goth. insahts 'account, narrative', sakan'quarrel, scold', insakan 'remind'.Hitt. kistari 'is quenched', kestanu- 'quench'65must be identified withGoth. qistjan 'destroy', Lith. gestu, gesti 'be quenched', and Gk.p0vVy , 077P,3Pea, a 'quench'. A vocabulary fragment (KBo.1. 42. 4. 4) preserves the word kuftuwantawwar ruin', which seems to bea denominative verb from an adjective *kut1uwanza having destruction,ruined'. Apparently the u in the first syllable is a trace of the originallabial which survives in the IE languages (cf. Hitt. kun-: kuen-).Hitt. huek-, huk- 'conjure' (see Friedrich, ZA NF 4. 187 f. fn. 3) is

evidently cognate with Skt. vac- 'speak', Gk. hros 'word', Lat. vox'voice', etc. (for the initial h, see LANG. 4. 159-70). The original labialappears in KUB 7. 53. 2. 6: hu-uk-ku-is-ma hu-uk-ki-is-ki-iz-zi, 'theconjurer (?) conjures' or 'he chants incantations (?)'.

Hitt. watku- has been recognized in the senses 'leap' and 'flee';66but I am inclined to see a more fundamental force in the passages quotedabove (p. 220), in which MUL watkuzi means 'the star sets'. Probablywatku- contains a prefix wa-, cognate with IE uu 'down', a by-form ofau.67 The verbal root I would identify with that of Lith. denkti 'cover'and Gk. 6tarrw bury', namely IE *dheng~Uh-cover, bury'."8 With this

65See Friedrich, ZA NF 1. 16.11See Friedrich, ZA NF 2. 54.67See Walde-Pokorny, VWIS 1. 14 f.68 See Zupitza, Die GermanischenGutturale99. Lid6n, Armenische Studien 41ff., followed by Boisacq, s.v., separates Oarrcarom this root and connects it withArm. damban'tomb'.


16/17

GUTTURALS IN HITTITE AND INDO-EUROPEAN 227root belong several Germanic nouns such as Eng. dung and MiddleLow German tunc 'subterranean room', which may well be cognate withHitt. tekan 'earth'. The latter word has lost the labial, although all theIE words may be traced to the root *dhengY-.The Hitt. particle of direct quotation, wa, is appended to the firstword of a quoted sentence, thus: hatrait, 'uk-wawalhwanzi URU Daluwapaimi', 'he wrote: "I am going to attack Daluwas"'. If it is followedby a form of the enclitic pronoun al 'he', the particle of direct quotationis war, thus: numu memir: 'paiweni-war-an-kan kuennummeni', 'andthey said to me: "we will accordingly go and kill him"'. I suggestthat this particle be identified with Gk. 4, and be. It has been notedthat Ch. S1. ie combines the uses of both Gk. particles and that it mayrepresent either IE ge (whence Gk. ,yE)or IE *glUe whence Gk. b6);evidently *gYewas the original form, but Gk. T, reflects the loss of thelabialization. The vocalism of Hitt. wa appears in Doric y,, and theextended form war may be identified with Gk. -ydp. On the source of Gk.a (and the same holds of Hitt. a), see H. Giintert, Indo-germanischeAblautsprobleme77.The gradual elimination of the labial element of the labio-velarscontinued throughout the IE period and later until it was completelylost in the satdm-languages. Even among the centum-languages thereare to-day but few that still retain the ancient articulation, and theyonly in part (e.g., Italian quale, Spanish cual, Eng. quick). PrimitiveIE falls midway in the development, and it is not strange that we areforced to recognize some inconsistences at that period.In pre-Greekwe must assume a contrary tendency, according to whichlabio-velars spread by analogy to the anti-consonantal position, andthere developed into labials (e.g. 5potat, Aacpp6s). That such formswere really analogical is indicated by the fact that Gk. preserves moretraces of the regular development (yvvw,,Xaxbs, etc.) than any otherIE language.

(5) Hitt. does not make any distinction between the IE pure velarsand palatals, and apparently primitive IH did not either, although theavailable evidence is slight.I have suggested (LANG. 6. 27-8) a connection between Hitt. tekkugami'I point out', tekkus's~ta, tekkus'set he showed', tekkus(?a)nu- 'prove',etc., and IE *deikl-'show' on the assumption that the Hitt. words allcontain the nil-grade of the perfect active participial stem. Thisassumption, however, is as difficult as the etymology itself is self-evident.


17/17

228 E. H. STURTEVANTHitt. tekkus- rather represents IH **deikus-, i.e. IE *deik, plus theelement u/M (which elsewhere united with IH k to form IE qY)plus theformative s, which gave rise to the IE present and aorist suffixes in s.Similarly Hitt. parkui' 'pure, guiltless', parkue'zi 'is acquitted',parkunu- 'purify, forgive' are to be connected with Skt. bhrd&ate, nunquoted variant for bhrajate shines', Gk. pOopKObv'EUK6V,-OXLt6v,tc., onthe basis of IE bherek-"9 nd IH bhereku-. The Germanic languagesretain the labial element in Goth. bralv- (in bralva augins 'E,miil6OcaXtoi3'), E brw, brWaw,rag'eyelid'. The Hitt. words must be con-nected with IE bherek-rather than with the equivalent and much morefrequent *bhereg-, ince IH gy would yield Hitt. w (see above, p. 221).Hirt70has made it extremely probable that the differentiation of thepalatals from the velars arose in Pre-Indo-European times by palataliza-tion before front vowels. If so we may assume root forms **diku.u-, tc.,to account for the palatals in these groups of words. But, since thereis no evidence for a vowel after the k or g and before the labial, it issafer to assume that the loss of the labial element V in PIE preceded thedevelopment of the palatals.For our IH reconstructions we need assume, then, only one series ofk-sounds: k, kh, g, gh. These sounds were sometimes followed by aconsonantal t, which became syllabic before another consonant; and inIE the k-sounds combined with such a following V to form the labio-velars. From the earliest times, however, V in this position tended tobe lost under conditions that cannot now be determined. After the IEperiod this process was carried to its logical conclusion in the satam-languages, and, at a much later date, in some of the centum-languages.In IE k-sounds in certain positions (before front vowels?) were palatal-ized, and analogy generalized the palatalized sounds in some groups ofrelated words or forms. This process affected both k-sounds which hadnever been followed by u and those which had lost a following u.

69 See Walde-Pokorny, VWIS 2. 169.7oIdg. Gramm.1. 232-41.

sturtevant - the gutturals in hittite and ie

Documents