Linguistic Society of America

On Indo-European Noun-Declension, Especially of -O- and--Stems Author(s): Louis H. Gray Reviewed work(s): Source: Language, Vol. 8, No. 3 (Sep., 1932), pp. 183-199 Published by: Linguistic Society of America Stable URL: http://www.jstor.org/stable/409649 . Accessed: 28/01/2012 05:21
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 of content 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 support@jstor.org.

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

http://www.jstor.org

ON INDO-EUROPEAN

NOUN-DECLENSION, -0- AND -A-STEMS

Louis H. GRAY UNIVERSITY COLUMBIA

ESPECIALLY OF

[See Summary] consonantal noun-declension shows, as is well known, Indo-European the following endings for the singular and plural (the dual is not discussed at this point): Nom. Sing. -s, -0 " Acc. -r1, -m
Gen. "
""

Plur. -es " -ir-s,

-/mS i

-m-s

Abl. Dat. Da t. Inst. Loc.

"

" "

-o/es, -s -o/e -o/ei -_5/1,-bhi, -mi -i, -0

e"

,
i2

-bhi-es, -bh-os, -m-os -bhf-s, -mI-s -s-u/i

" "

The -o-declension differs from this scheme in having the gen. sing. in -s(i)o or (in Italic, Ligurian, and Celtic) in -i; the abl. sing. in -o/et; and the dat. plur. in -o/eis. The amalgamation of these endings with the final vowel of the -o- and -d-stems is clearly implied by the accentual conditions found in Greek and Lithuanian:' Nom. Sing. abX6-s, aiila-s, td-s Acc. Gen. "
"

au'lq,tq,2gerqj aV'X6-,, asxoi < abXoto, *abXb6/-o-5, abX6b/-o6

Plur. (abXol, aulai, geri, gerieji) i " abkobs< *aV'xblE1/--s, germs, geriosius " avXcv < *aXkb/E-d/1y, auli?

1 For noun-accent in Lithuanian see, most recently, N. van Wijk, Die baltischen und slavischen Akzent- und Intonationssysteme 42-60, Amsterdam, 1923; T. Torbi6rnsson, Die litauischen Akzentverschiebungen und der litauische Verbalakzent 10-26, Heidelberg, 1924; G. Bonfante, in Studi baltici 1. 76 sqq. (1931). 2Apparently for *tam, with compensatory lengthening for loss of -m (K. Brugmann, Grundriss der vergleichenden Grammatik der indogermanischen Sprachen2 1. 389, Strasbourg, 1897-1916). 183

184

LOUIS H. GRAY

Abl. Sing.

< (cf. acXw, FOLKW, *aX6,b/E-r(6)3 OLat. Delph.

facilumad, meretad),

"
abhots5 <
*aabh1-s,

ajilo, t5
Dat. " < *aibX/e4-o/e, aiilui, abXCX6/E-c?/r

auldms,
" geriemsiems

Instr.

paskui, gerdmjam " aA C < (cf. *ab tuo, td, t4

9rLaXEpo, Il 'Iaprt?3), auli,

"
"

aulais < *aiA6bl-, abXots5 auluose (OLith. namasu)

aXotn5 < *aVbX61-at,

Loc.

"

aibXo

< *abX'6/e-L,

aule, namie, gerainjam(e), OLith. sosti, warduia, wardC,dariiq4

OKOL : OKE,, Osc.

As clearly appears from such alternations as Greek : rtLaXepw "Aapb-7; Lacon. Greek terei; 8VL : KX7PLKeos; o1KcP: ApLo70-LKOU Goth. dag? (< *-Jm?);6 OPruss. deiwas, OHGerm. tagas : OPruss. steises, Goth. dagis; Skt. vr'kd : uccd' (cf. also OCSlav. dobre7),the -ostems show an alternation o/e resting on a difference in qualitative accentuation, so that the earliest types would seem to have been *uoiko- : *aule-, thus accounting also for the voc. sing. type in -e (Greek ab X), which was analogically extended to all -o-stems, as OtKEinstead of *otKo
< "51K6."
SFor the accent-one would expect circumflexes, but caaprj is explicitly attested by Aristarchus-see O. A. Danielson, Grammatische und etymologische Studien 1. 13, Upsala, 1888; H. Hirt, Indogermanische Grammatik 5. 287, Heidelberg, 1929. 4 The termination is of uncertain origin (references in J. Endzelin, Lettische Grammatik 306, Heidelberg, 1923). It seems identical with that seen in OCSlav. kamen-e, sloves-e, etc., and perhaps with the -a of the Skt. dat. type of deviy-a (Hirt, Grammatik 3. 10-11, Heidelberg, 1927; cf. Brugmann, 2. 1. 185-6), possibly even with -a < *-o in Kanis. ?uppi-y-a, genu-w-a (Gray, in JA 218. 324 [1931]). Torbi6rnsson (Akzentverschiebungen 24-5; cf. van Wijk, Intonationssysteme 65), however, derives it from -e < -in (*namig-en > *namij-e, > *nami-e > name,). 6 For the formation of these three cases on the neut. plur. *toi, see below, p. 189. 6 For various explanations of this difficult form, see E. Kieckers, Handbuch der vergleichenden gotischen Grammatik 108, Munich, 1928; H. Hirt, Handbuch des Urgermanischen 2. 32, Heidelberg, 1932. A. Meillet, Le Slave commun 337, 354, Paris, 1924. 8 Cf. Hirt, Grammatik 2. 175-6, Heidelberg, 1921; 3. 42-4. Does this alternation likewise explain such formations as Greek XLeos, xpbeEos, < *XLOE-o-s, raXeos *xpvae-io-s, 7raXE-Fo-s(cf. also Ion. artX'os < *#acXL;F-Fo-s)?

ON INDO-EUROPEAN NOUN-DECLENSION

185

In the -a-declension, one finds: Nom. Sing. Oea',algd, ger6ji " Oea'-v, algq, t4, Ace. gerqjq
Gen.
"

Plur. (OEal), algos, tos " O~a'-s,algds, ger4sias

OE6'W

Abl. Dat. Instr. Loc.

, " "
"

eas < *O"ea'-b/es, algs < ta *Oia'-o/e, algai,

<OE&Wi

<

*OEa'-c-W/Ing,

"

algi , algms / < *Oea/91-s,9 alg6ms Oe-s < *OEa'/al-s,9algomis, ger5msiomis 6EatZo < *O~a'/at-L,9 algose (OLith. dienosu), gerosiose

" "
"

tai'Ocak Dor. Kpvcp < *Kpvua'-&/fl

algd, gerdja bvXi < ?vXa'-l,algoj, ger~joje

A survey of the -o-declension seems to explain the difference in accentuation between the loc. sing. OiKOL< *5&K1-L and the nom. plur. OLKOL < < *aiX6b-,and nom. plur. abXol. Final 61KoI, loc. sing. abXo0 original diphthongs in Greek were not only monophthongised, but shortened (cf. the types AvOpwroL, -~pvpaL), whence OKOL = -(<.)"; whereas preconsonantally they remain long and are true diphthongs (e.g. abhot : abhols). On the other hand, secondary diphthongs retained their value of two morae, whence the type olKOL= - W, and, similarly, the verbal type XEciro < (cf. Skt. bhdret, Goth. habai, Lith. *XEhrb-L te-velie) = _".Wo The genitive singular of the -o-stems presents problems of much difficulty. Such forms as OHGerm. tagas, OPruss. deiwas, Goth. (Thess. BEXpatVL, dagis, Greek o0'Kov etc.) show that one Cypr. apyvvp--v, of the endings of this case was *-o/eso. This seems analysable as *-o/es-o, of which *-o/es strongly suggests the neuter (vague, generalised, collective) nouns in -o/es represented by the type of Greek 7y6os,-y~e(a)-os, while -o- appears to be the very frequent nominal formative. The whole type of *g no/es-o would then mean 'relating to the collectivity represented by *geno/es', and of *uoiiko/es-o(Greek o0Kov)'relating to the collectivity represented by *yoiko-es', and would date from a preinflexional period of Indo-European. It is even possible that the genitival termination *-o/es is itself identical with the neuter (vague,
I For the formation of these three cases on the nom. plur. fem. *tdi, *tai, see below, p. 189. 10

For the current explanations

of

oiIOL :

Griechische Grammatik4 178, Munich, 1913.

oIKOL,

etc.,

cf. Brugmann-Thumb,

186

LOUIS H. GRAY

generalised, collective) ending *-o/es of the neuter type *d-n-o/es-, whence genitival types like *#Mno/es-o/es(cf. Skt. jdnasas, Lat. generis) would be doubly generalised. With -u- infixed as the dual sign (*-o/eu-s), the same termination seems to recur in the gen. dual, and with -bhi- and -m-, in the abl.-dat. plur. The type found in Skt. devdsya, Hom. oIKOLO, Armen. getoy, Lat. for the final Dor. cuius (< *quo-sio-s; -s, cf. literary 'Afo0, CE's, kiAs, may, similarly, be analysed as *deiuos-io. 'perTr7E,mro3S, Locr. FEOS)11 taining to the collectivity represented by *deiyos', and would likewise belong to the pre-inflexional stage. This *-/o- may be regarded as the grade ZS[N] b of the formative base *4i6-ei, and may itself be the source of the Indo-European relative pronoun *io-s (Skt. yd-s, Phryg. Lo-s, Greek 6-s, etc.). This hypothesis suggests a clue to certain other difficult forms, including the Italic, Ligurian, and Celtic 'genitives' in -T (Lat. viri, Messap. hidazimaihi, Lepontine agkoneti, Gaul. Segomari, OIr. fir < *yiri), which Wackernagel'2 rightly connected with such Skt. types as akkhal1kr'tya, phali'etc. (cf. also the Lat. type uni-bam, Lith. vien mithuni'kar-, karana-, davau), explaining both as an 'indefinite case'. This ! would be the grade I4Z b of the same base (*i < ei); and the grade ZZ b, i, seems to solve the problem of the compound type represented by Lat. damnificus, agri-cola, lecti-sternium, longi-manus, belli-ger, etc., as contrasted with oinu-mama, locu-ples, etc., and by Gaulish Comati-mara : Comatu1 Cf. A. Meillet, Esquisse d'une grammaire compar6e de l'arm&nienclassique 48, Vienna, 1903; Stolz-Schmalz, Lateinische Grammatik5 289-90, Munich, 1928. With the type of *#"no/es-oone may compare such adjectival formations in -o- to

stems in -o/es as Skt. rajas-d-, tamas-d-, rabhas-d-, Greek < *srAaeX~pbs qyelbho/es-o- (cf. OHGerm. kilbir-a), K'1eos < *K7q3o/Ea-o-, ryeos < *rTeo/e-o-, Lat. creperum < *crepos-o- (cf. crepus-culum), OHGerm. egisa (cf. Goth. agis, Greek dxos), Lith. ialesa (cf. Lat. holus); and with the type of *deiuosj-o-, Skt. usas-yd-, Greek ipKEZOS< *ipKOo-jO-, 9ELOS < *FET700o-o,Lat. Venerius < *uenos-io-

(Brugmann 2. 2. 543-4, 192).

12

In Ml6anges de linguistique

offerts a ...

F. de Saussure 123-52, Paris, 1908.

The view of H. Ehrlich (Untersuchungen iiber die Natur der griechischen Betonung 66-79, Berlin, 1912) and Kent (in LANGUAGE6. 310-2 [1930]) that this i represents an old loc. in -ei seems less probable (cf. Stolz-Schmalz 268-9). J. Schrijnen ('De latijnsche genitief singularis van de o-stammen', in Mededeelingen der koninklijke akademie van wetenschappen te Amsterdam, Afd. Letterkunde, 67, series A, no. 4, Amsterdam, 1929) regards the Italic forms as borrowed from the pre-Indo-European peoples of Italy; and H. M6ller (KZ 49. 219-29 [1920]) seeks parallels between Indo-European and Semitic (cf. the Arab. 'genitive' maliki 'of a king'). For the types in -i, cf. also Hirt, Gram. 4. 305, Heidelberg, 1928.

ON INDO-EUROPEAN NOUN-DECLENSION

187

marus, Devi-gnata : Devo-gnata, Mari-talus : Maro-bodus, Orceti-rix : Orgeto-rix.13 Assumption of a grade RZ b (ai < ai) may also elucidate Greek infinitives of the types b8EZ-aL,Etr-ac, 86eV-acL, o~60F-at, and such adverbs as xaA-aL,irap-aL, which seem, contrary to the current view, to be neither datives nor locatives, but, rather, mere 'indefinite cases'.14 The accent of 6/uEvac, etc.-that of xa'al, etc., appears to be secondaryshows that -ai is here original, thus forbidding it to be a locative, where, as already seen, the diphthong is secondary. The importance of this formative base *-si6/i-, as of the closely parallel *-EV/6i, seems greater than has thus far been supposed. Its grades actually found may be arranged as follows :" N[P]I) a --ei > E07 Lat. fida Lat. fids < *-4i-s N[P]Z b Gk. abXc3P < -*b-o', S[N]Z b -o/ei RZ b IZ b ZS[N] b ZZ b -ai -ei> -io/e -i
*,-E-E " 86jEv-aC 6e'Eus<

-au -eu

Hom. ep?(F)-Es Gk. r~xcEts <

* .rlX-EF-Es

Lat. vir-I, Skt. dev-~i' -eu > f Skt. vadhf'-s Skt. yd-s, Phryg. to-s -uo/e " pu'r-va-s Z < *-6-1, Gk. acXo " -u sfnis.-u Lat. longi-manus

The similarity of Skt. devW' with Skt. mithuni'-kar-, Lat. viri, etc., together with the generally acknowledged identity of the 'fem. sing.' in d/9 with the neut. plur. in d/a, strongly suggests that devi', like mithun!', was originally no more 'fem.' than was Skt. kald'. At first, it also was probably a collective noun with neither gender nor number, but, through analogy with the type of kald', which chanced to coincide in form with sd', the true feminine of the demonstrative (I-E *sa, Dor. 'd, Goth. s5, etc.), it was itself regarded as feminine and, again like the type of kald', served as basis for a whole declension. In like manner, the 'fem.' type of Skt. vadhf'-s < *yedheu- was originally a collective of precisely similar formation. Finally, the sole essential difference
13 This interpretation appears more plausible than the current view (e.g. Stolz-Schmalz 202, 248), which considers i as weakened from u < o, though that u is probably found in the types locu-ples, Comatu-marus. For the Gaulish forms see A. Holder, Alt-Celtischer Sprachschatz 2. 2, Leipzig, 1896 sqq. 14 For such interpretations, see Brugmann-Thumb 265, 295, 461-2, 410-2; Hirt, Grammatik 3. 51-2. 1 Cf. L. H. Gray, in AJPh 51. 273-85 (1930).

188

LOUIS H. GRAY

?v-y

between nouns in -i- and -u- (really in -ei- and -eu-) and those in -i-, -a- is one of apophonic grade. Even the type of Lat. fides, res (cf. Skt. rd's < *r6i-s) appears to belong here, at least in part, as does that of Hornm. Skt. bhrf'-, OHGerm. brdwa).16 Lp(F)-Es (cf. also Greek b&ppvs, To the same source one may assign the neut. plur. of -i- and -u-stems of the types of *tri (Skt. tr!'[n.i],Lat. tri-ginta, OIr. tri, Lith. try[lika], OCSlav. tri) and *medhi (Skt. mddhi[ni]), which appear to be identical both in outward form and in development, thus forming a complete parallel with the neut.-fem. types of *iuqa : (Skt. yuga', Greek *alqt-hd : Returning to the 'genitive' as really an 'indeterminate case' dating from pre-inflexional times, one possibly finds a similar state of affairs in the Armenian gen. plur. in -c (get-oc, am-ac, etc.) < *-sko-, *-sfo-s, or *-s/ko-m.1' Which of these three is present here cannot be determined from this isolated formation, nor is the question of material importance in the present connexion. Here, too, belongs the type of OCSlav. togo < *to-gho, similar indetermination being shown in Skt. mdma < *mane (cf. Av. mana), tdva, OCSlav. mene, OLith. mane, Mid. Welsh teu, and possibly in OPruss. maisei, swaise, stesse(i), tennessei, etc., if they are based on the nom. sing. masc. of the possessive pronouns mais, swais, etc., plus the -e seen in *mane and the facultative deictic particle *i found in Greek obroo--',Umb. po-i, pur-i.s8 The hypothesis that a vague or collective neuter underlies the types of o'KOt < *yoikos-o, ol0KLO < *uoikos-jo may receive some support from the pronominal types of Ved. asmd'ka(m), yugmd'ka(m), Lat. nostrum, vestrum, which obviously coincide in form with the nom.-acc. neut. sing., Goth. meina, etc., being nom.-acc. plur. neut. At this point one must consider the general question of such genderdenoting pronouns as *so(s), *sd, *to-d, particularly as regards the types of Skt. tdcsya: tdsyas (but devdsya : kald'yds), tgadm: td'sdm (but Ved. cardthdm,Av. staorqm,Lat. deum). In BSLP 31. 41-2, I have reached the conclusion that only the nom. and acc. were independent cases, all
16Cf. Brugmann-Thumb 217-8.

ycpvpa, axpil, Lith. algd).

17Meillet, Esquisse 47. For the Kanisian types hkmand-al, ammbl, see Gray, in JA 218. 322-3. 18 Meillet, Slave 378-9; Endzelin, Grammatik 372-4; H. Pedersen, Vergleichende Grammatik der keltischen Sprachen 2. 168, G6ttingen, 1909-13. For other explanations of the OPruss. forms, see R. Trautmann, Altpreussische Denkmiiler 262, G6ttingen, 1910; N. van Wijk, Altpreussische Studien 58-9, The Hague, 1918. For Skt. mdma see Meillet, in Bulletin of the School of Oriental Studies 6. 435-7 (1931).

ON INDO-EUROPEAN NOUN-DECLENSION

189

the others being quasi-enclitics or quasi-proclitics ('dependent'). It is highly probable, furthermore, that only the nom. masc. and fern. (i.e. animate) could, properly speaking, be an active case, and that the ace. neut.--originally there could have been no nom. neut., since an inanimate thing cannot accurately be regarded as acting-was a passive case,'9 a statement which applies equally to all the 'dependent cases'. Thus we have *so(s), *sd, *so/ei, *sac/12Ofor the masc. and fem. nom. sing. and plur. (Greek 6, i, o0, al; Skt. sd, sd', t6 td's; Goth. sa, so, fai, bas; *to/ei, *ta/ai have replaced *so/ei, *sd/ai everywhere except in Greek, and even there West Greek has rol, ral); but for the neut. sing. *to-d, plur. *to/ei, *td, *tai (Skt. tdt, td'; Greek r6, rd; Lat. is-tud, is-ta; Goth. fat-a, OHGerm. dei)1. The reconstruction of all the nominatives of this base would, then, seem to be: Masc. Sing. Plur. Dual *so(s) *so/ei *s6/?(u) Fem. *sd *sd/ai *sai Neut. *to-d *to/ei, *td, *tai *tai

It is noteworthy, in considering this type of inflexion, that, in the singular, the fem. 'dependent cases' are built immediately on the base in Indo-Iranian, Gothic, Old Prussian, and Slavic (Skt. tdsyds, tdsydi, tdsydm, tdyd; Goth. Jizas, f]izai; OPruss. stessias, stessiei; OCSlav. tojq, toji); in the fern. plur., Indo-Iranian builds throughout on *td (doubtless acc. plur. neut. rather than the fem. stem-form *td: Skt. ta'sdm, td'su, td'bhyas, td'bhis), as does Greek in ra'ov, iiait (but on *td/ai in rats, ratat), Lith. in tosd, t6ms, tomis; in the masc. plur., IndoIranian constructs throughout on *to/ei, *tai < *t9i (ace. plur. neut., as in Lat. quae, Osc. pai, OHGerm. dei, Lith. tai, aulaf: Skt. tegam, thgu, tebhyas, tebhis), as does Greek in rots, roZtr; Gothic in faim (fem.
19 Cf. C. Uhlenbeck, 'Agens und Patiens im Kasussystem der indogermanischen Sprachen', in IF 12 (1901). 170-1 (cf. KZ 41. 400 [1907]), denied by Hirt, Grammatik 3. 39'. Meillet, in MSLP 20. 176-8 (1916), considers that the neuter type is itself a stem-form, and that -m is here not an inflexof Skt. yugdm, Greek rvuy6b, ional ending, as it is in the acc. masc. type of Skt. vr'kam, Greek Xbov, in which connexion it may be noteworthy that Kanisian shows the neut. sing. type irmala as contrasted with the acc. sing. masc. irmalan < *-om. Hirt, Grammatik 3. 86-94, regards the neut. -om as a determinant. 20 Cf. Meillet, in MSLP 18. 433-4 (1913). 21Cf. Meillet, in MSLP 22. 52 (1922).

190

LOUIS H. GRAY

also), OPruss. in stWison,steimans, and OCSlav. in techil, temil, timi (fem. also).2" In the -a-declension, on the other hand, the Indo-Iranian sing. is built directly on the stem (Skt. kald'yds, as contrasted with tdsyds), thus giving further evidence of the derivation of this entire declension from the neut. plur. (cf. Skt. ta'sam, etc.). In connexion with the neut. plur., one notes, also, that the alternation of the nom.-acc. neut. plur. termination -d/a (Ved. yugd', Osc. prtiftii, OIr. dliged(a), Pomesanian OPruss. slayo, OCSlav. iga, but Greek i43a, Lat. jugd, Goth. juka) reappears in the fem. (Skt. kald', Greek x&pa, Osc. touto, Goth. giba, Samlandian OPruss. mensa, Lith. algd, gerdji, OCSlav. rqka, but Greek y-7vpa, Lat. stelld). In the Skt. and OPers. fem. sing., not only the gen., but also the dat., instr., and loc. present an aberrant and isolated type (Skt. kald'yds, kala'ydi, kaldya-contrast Ved. jihva', GAv. daenda-kald'yam; cf. also Lith. algoje). This type coincides with the Avestan, except that the latter, like the Skt. types kaldyd, tdsyds (and for the same reason?), has a instead of d (daenayd, damnaydi, sasnayd, frasayd; cf. also OCSlav. instr. devojq). The origin of all these cases is currently explained23 as based on the loc., with -ayam < *-di-dm < *-di-d-(e)m (rather, < *di.-6/m, the latter element recurring in the abl.-dat.-instr. plur.); or by analogy with fem. stems in -i- and -T- (Skt. matyd's, matya'i, matyd', matyd'm; devyd's, devya'i, devyd', devyd'm). That Skt. kaldyd, Av. sasnayd are late analogical formations is shown by Ved. jihvd', GAv. danda; and it seems likely that all four forms were constructed on the plur. (i.e. collective) in *-d/ai (kala'yds, kald'ydi, kald'ydm < *qo/eladi-s, etc., with -d- by analogy with the normal type seen in Greek Oea8, OEq; but kaldya < *qo/elai-51/). Moreover, analogy and confusion between stems in -i-, -f-, and -d- seem to have played an important part, so that the inflexion of the Indo-Iranian -a-stems would appear to have been, relatively, a very late development.
22 Here, perhaps, belong Greek adjectives of the type of ro-o (the type of 68a-o is apparently built on the 'indefinite case' of the type of xaAali), despite see Brugmann-Thumb the genitival types okJoLo (for other explanations troo, der Laut- und Formenlehre2436, HeidelH. Handbuch

griechischen Hirt, 212-3; berg, 1912). Hermann ('AvrI6opov217-9), less probably, thinks that the plural -oi was carried through from the loc. sing. of the noun. For the vexed question of OHGerm. dei, cf. Brugmann, Die distributiven und kollektiven Numeralia der indogermanischen Sprachen 60, Leipzig, 1907. 23 So most recently by Thumb-Hirt, Handbuch des Sanskrit2 178, 495, Heidelberg, 1930; cf. Brugmann 2. 2. 152-3.

ON INDO-EUROPEAN NOUN-DECLENSION

191

In the instrumental, the -o-stems show *-5/ (Greek rLtoXepec, Forms in *-mi, like OCSlav. rodomi, recall the formative *-mi-aapri?). seen in *qlr-mi- 'worm' (Skt. kr'mi-, OIr. cruim)24,i.e. an agential type, this being the grade ZZ b of the formative base *mo/ei6/ei-, which recurs, with or without pluralising -s, in the instr. plur. (OHGerm. dagam < *-mi[s]), and, in the grade 4IZ b, in Lith. algomis, OCSlav. gre"chilmi. The dat.-abl. plur. also has the formatives *-m-o(s) (OLith. sunamus, Lith. auldms, OCSlav. rodomil) and *-bh-o(s) (Lat. dedbus, Gaul.
valavOclago).

The case-termination *-bhf is more interesting. It occurs as -bhi in Arm. mardov; plur.: Ved. the instr. of all numbers (sing.: Hom. Pin/3p,25 devebhis, OIr. feraib; dual: Skt. devebhy-dm, GAv. zastaiby-d, OIr. feraib), and in the abl.-dat. plur. (Skt. devebhy-as, GAv. daOiaeby-6); and as *-bh! in the instr. plur. (GAv. daenabi-s, hizbi-s', bavar bi-?). These represent respectively the grades ZZ b and 4tZ b of the base Gaul. *bho/ei6/i.26 The grade ZZ b occurs also in Skt. a-bhi, Greek au-<p1, am-bi, Goth. -bi, and OCSlav. o-bi. A grade *bhei- seems to be present in Skt. u-bhayd' beside one in *bho/ei- in Skt. u-bhdya-, Lith. a-beji, and OCSlav. o-boj (these three apparently late, since a grade NN *bho/eio/e- is scarcely possible as an original form). The primary force of *-bht- would seem to have been 'with', thus explaining its use predominantly in the instr. In the singular, the ablative has a special form only in Indo-Iranian, Kanisian, and Italic: in the first, solely in -o-stems; in the second and third, throughout (Skt. deva't, Kanis. irmalaz, tuzziyaz, genuwaz, meniyanaz, h mandaz, happaraz, nepidaz, OLat. poplic6d, facilumid, sententiad, loucarid, castud, coventionid, Osc. sakarakliid, amprufid, toutad, slaagid, praesentid; cf. also Delph. FOLKW). The I-E ending would seem to have been *-tos, which appears to survive in Skt. mukhatds beside mi"khat, hrttds beside hrdds, Greek 'r6s, ~KiTs, Lat. intus, funditus, caelitus. In Kanis., *-tos > *-ts > -z; in Indo-Iranian, -t, the preceding vowel being lengthened by compensation for the loss of -o- in *-tos (as in the normal conditions for the origin of the prolonged grade), whence Skt. mitkhat, OLat. poplicbd, etc. The choice between *-t and *-tos was probably governed originally by conditions of sandhi.27 Elsewhere the
24

Brugmann 2. 1. 253-4. 25For further details, see below, p. 192-3.

28 27

Cf. Brugmann 2. 2. 76-7, 80, 795, 820; E. W. Fay in AJPh 32. 408-9 (1911). Cf. Sturtevant, in LANGUAGE 8.1-10 (1932). Whether one may safely put
for one

Greek adverbs of the type of KaXws < *KaX6-r(6)s here seems doubtful,

192

LOUIS H. GRAY

abl. coincides with the gen. in the sing., with the dat. in the plur., and with the dat.-instr. in the dual; and Meillet28 regards the gen. sing. itself as primarily an abl., though this seems rather doubtful. Beside the long vowel preceding the ablatival -t, -z < *-t(o)s in the nominal declension, one finds a short in the Indo-Iranian personal pronouns- Skt. mdt, tvdt, dvdt, yuvdt, asmdt, yuamdt (cf. also mdt-krta-, asmdt-sakhi-, tvdt-pitar-, yugmad-devatyd-), GAv. mat, Owar (beside Ow&d), ahmat, yftimat, xgmat, OPers. ma-and in YAv. even in nouns (e.g. zrayavhat, apat, drujat, tbilyantat, dOrat,nar t, &asmanat,vardOraynat; cf. also yahmat, ahmat beside yahmdt, ahmdt). This suggests that the most primitive ablatival ending was *-t, to which, through confusion between abl. and gen., the genitival *-o/es was added, so that *-tos is really a composite termination, the original ending surviving only in Av. and in the Skt. personal pronouns.29 Besides the normal types, the instrumental and locative sing. show forms characterised by -m: instr.-OLith. wardV, darwii, naturoig, smertijq, Lith. gerafmjam(e),algd, gerdja, OCSlav. silq, devojq, kostnijq, , OPruss. tojq; loc.--Skt. kald'ydm (but GAv. frasayd, OPers. arb[a]irdyd) schisman.30 A similar element -i(n) is found in Skt. and Av. loc. types like a-sm-in, a-hm-i, a-hm-y-a, and in Homeric ablatives, instrumentals, and locatives (both sing. and plur. without distinction of form) in -(pt() < *-bh-i(n): abl. sing. LeXaOpb~e'?L, plur. bore6~w; instr. sing. Pi~7(pv, plur. O6eVt; loc.
-ro's The was more probably an instr. than a loc. owo is ambiguous, since it beside oturws most plausible explanation of the final -s here still seems to be that of Brugmann 2. 2. 700, 737-8. The Turfanian abl. wramis as distinguished from the gen. wramis is considered to contain the postposition -qu (W. Schulze, E. Sieg, and W. Siegling, Tocharische Grammatik 36, 294, G6ttingen, 1931). 28 'Quelques remarques sur le sens du g6nitif indo-europeen', in Festschrift Vilhelm Thomsen ... dargebracht 21-3, Leipzig, 1912, and MSLP 22. 50. 29 A case-ending -t seems to recur in the Kanis. instr. type irmalita (conventionally read irmalit), Greek KarT, OWelsh cant < *krI-ta, Greek lErA,Lat. ita, ali-uta (Gray, in JA 218. 324-5). 30 OPruss. mensan, krawian, etc., considered instr. by Trautmann 226-7, are regarded as acc. by van Wijk, Studien 81 sqq. (but cf. OCSlav. kri~vnijq). For OPruss. schisman, see Trautmann 264; for the OPruss. instr. maim, ib. 269-70 and van Wijk 92-3; for examples in OLith., B. Bezzenberger, Beitrage zur Geschichte der litauischen Sprache 133-4, Gottingen, 1877. Here also belongs the Phryg. dat. aoeovv (cf. J. Fraser, in Transactions of the Cambridge Philological Society 6. 2. [1913] 10. 12).

would then expect

*KaXo(b (cf.

< *rbr-s < *rb8-s); and the type of

ON INDO-EUROPEAN NOUN-DECLENSION

193

Here, too, one must place Dor. E1Z'v, sing. 'aXappLtv, plur. CKpLw~pV.31

ri'v, Boeot. dw < *Jergv, Lesb. aiuy(v), byt(v), Attic i)Ztv, i5bZv. The relation of Lesb. aEtL, bVLyL to a/4ylv, b ytY is precisely like that of Skt. asmin to Av. ahmi, Prdk. tammi, Pdli tamhi, or of Ved. mdhyam, titbhyam to mdhya, tibhya (cf. also Prdk. mamammi, tumammi with -ammi < *-asmi). The origin of the long vowel in Attic-Ionic ytv, is rather problematicl1, but it may have sVA, Dor. eilu'v,rt'v, 'ap~'v, It Y'v for example, arisen by adding the particle *-i(n) to the forms in -i, )?Thy, being for *n-sm-i-in.32 This *-i(n) seems to recur in the type of the Skt. instr. sing. devind (but Ved. yajih-', Pdli dhamma beside dhammena), found only in Indian, and pointing to *deiuo-in-d,33as well as in the Greek dual type of Twrottv,~wrov. In the plural, generally speaking, the characteristic is -s (nom. Skt. devd's, kald's; Osc. Niivlaniis, scriftas; Goth. wulfos, gibas); and this ca: devdm, Cret. appears in most of the cases, as acc. Skt. devadTh Goth. wulfans : wulf (feminines like Cret. rtyavs, iXEvOEpos XEvEpov, :? are late Lith. rankdcs, developments; originally, significantly enough, the acc. and nom. plur. fem. coincide, as Skt. kal&'s, Goth. gibis); : asjw < *aVX6-o1; instr. Ved. devebhis: dat. Greek abXots< : ahXot < *abXb-'. The genitival plur. -s- found abXoZWt< *ab6Xb-a-i in Homrn. < B~a'w *0Ea'--A, Lat. dearum, equdrum, Osc. egmazum, Kanis. pronominal sumenzan < *-nt-s-5m (?) is probably a late analogical introduction of the same pluralising force; and in the Goth. genitival types wulf?, anstj, fadr , etc., it seems, on the whole, that *-jm alternated
31 It

*ab51-s OCSlav. vllkomi, Greek acXots < *abX61-s : abX&< *aXb-d; loc. Greek

but

vEXKVTLK6V', late -t, development (cf. Brugmann-Thumb 168-9). Outside Attic-Ionic, this -v is found only in the dat. plur. in Thessalian and Heraclean (C. D. Buck, Greek Dialects 78, rev. ed., Boston [1928]). 32 For other explanations, see Brugmann-Thumb 291; Hirt, Formenlehre 423; J. Wackernagel, Sprachliche Untersuchungen zu Homer 138-9, Gottingen, 1916. For the distribution of the forms, see F. Bechtel, Griechische Dialekte 1. 74; 2. 124, 189, 254, 403, 584, 605, 736, Berlin, 1921-24. 33 F. G. Aleksandrov, 'On the Declension of -o-Stems in Sanskrit' (Russian), in U6. zapiski v 'koly g. Odessy 2. 108-13 (1922) (known to me only from Indogermanisches Jahrbuch 11. 138 [1927]), considers the pre-form to have been *deiuoi-na, on the pattern of the plur. Whether such Arm. local adverbs as astfn, aststin < *-in, *-en belong here seems uncertain (cf. Meillet, Esquisse 63, 69); they suggest, rather, the postpositional -en of Osc. exaisc-en, Umb. arven <

before vowels ('vi

is generally recognised that the rule whereby -t appears before consonants, but is a
e.g. EbvptPv'O3vao-0s,
VoptL pqeptvos),

*arve-en.

194

LOUIS H. GRAY

with *-6m.34 In Skt. te~im, Goth. blindaizi, OPruss. st ison, OCSlav. t1chil, the ending is added to the neut. plur. stem + pluralising -s-, thus giving the pre-type *o/ei-s-6/1m. Besides the forms already cited, the ablative-dative and instrumental show others characterised by -bh- in Indo-Iranian, Armenian, Greek, Italic, and Celtic, and by -m- in Teutonic, Baltic, and Slavic.35 For the abl.-dat., these were *-bhi-o/es, *-bhos, *-m-os (Skt. devebhyas, Greek Lat. dedbus, Gaul. vaiavtLKc#o : Lith. auldms, OCSlav. vllbrPE6~bw, koml), and for the instr. *-bht-s, *-mi-s (Ved. devebhis,Prik. devehi(rih), Pali devehi, Arm. getov-k', Greek GIeI6p, OIr. feraib: Goth. wulfam, Lith. algomis, OCSlav. devami). Here *-bhi-o/es as the termination of the abl.-dat. plur., when compared with *-o/es as that of the gen.-abl. sing., clearly suggests that the original ending was the latter; the -s of *-bhf-s, *-mf-s was doubtless the plural mark. The dat. plur., had it been formed on the normal pattern, would have been either *deiyoi-oi-s, Skt. *devdyes,or *deiyoi-bhf-oi, Skt. *devebhye. The origin of the Indo-Iranian instr. plur. of the type of Skt. devd'is, Av. davadic,is problematical, since the long diphthong *3i here present can scarcely be reconciled with the short *oi of Greek abXos < *abXb1-S, Lith. aulais, for which one would expect Skt. *deves. It may be suggested that we have here a contamination of the normal *deiyoi-s with *deiyoi-5-s into *deiuyi-s (Skt. *deves- *devdyds> devd'is). The only other explanations for the form would seem to be either from *deiy5i-s as based on a nom. plur. *deiubi, or from *deiyo-&i-sas the base *deiyo+ an instr. sign -6i- + pluralising -s; but evidence either for such a nom. plur. or for such an instr. sign is thus far lacking."6 In the locative, the only point to be noted here is the alternation OCSlav. rodchui, -u/i (Skt. devegu, kald'su, GAv. yao0anaalz, gajOdhfi, devachTi:Greek abiXo-a, Obpaac). This -u/i appears to be the grade and *-e?6/1i- respectively.37 ZZ b of the formative bases *-u6/dei3 For a summary of the various explanations, see Kieckers 108; Hirt, Urgermanisch 2.32. Hirt, Grammatik 3. 60, 86-97, considers the element as < *o/em, which appears without case-value at least in Skt. ah-dm, mahy-dm, ay-dm. He denies (ib. 54-6) that -s in *-bhi-s originally had a pluralising force. 3 On the problem of bh/m see A. Meillet, Les Dialectes indo-europ6ens 119-23, Paris, 1908. 36For explanations of the form, see Wackernagel-Debrunner, Altindische Grammatik 3. 107, G6ttingen, 1931. Whether the type of OCSlav. rody belongs here, as Meillet (Slave 356) holds, seems not wholly certain. 3 Cf. above, p. 187. For Lith. auluosh, algosh beside OLith. namasu, dienosu, see Endzelin 300.

ON INDO-EUROPEAN NOUN-DECLENSION

195

In the dual, the presence, side by side in the masc., of Skt. deva'(u), GAv. spada, avarend, YAv. gav3, OPers. uba, Arm. erku, Greek oKCW, bbw, OIr. ddu, dd, dd,38 OIcel. tottogo < *tu5-te3ii, ASax. tu', nosu, Lith. algai, gerz'oju, OCSlav. roda, but Greek bo, rbsc, Arm. erko- in erkotasan, clearly indicates that the earliest forms of the nom.-acc. were in *-4/1(u) (the Indo-Iranian types may be referred either to *-5[u] or to *-[u], the first being far more likely; and the YAv. type barazanta may stand for any of the grades *-./5, *-o being the most probable because of GAv. anaocaphd). The fem. and neut. both show *-ai, the former in Skt. kale, GAv. ubj, OIr. tuaith, Run. tuaiR, Goth. twai, ASax. twd (all three as masc. from fem. or neut.), Lith. algi, gerieji, OCSlav. dive; and the latter in Skt. yuge, YAv. saite, OCSlav. sell. The nom.-acc. neut. of stems in -i- and in consonants show the reduced grade *-ei > *-i (cf., in the plur., *iuqd/a, *menesa> Ved. yugd', mdndrhsi, Greek pbea), as Ved. 'ica, dhd'mani, brhati',Av. vi-saiti, West Greek rt,d, FL-KarL, Lat. vi-ginti. The genitive was in *-o/e-u-s, added either directly to the base, as in Skt. padds, OIr. fer, tuath, OCSlav. rodu (Lith. tiidviejf, better written ti4dviejq,is taken directly from the plur.39), or to the neut. dual *-ai, as Skt. devdyos, GAv. qsayd, OCSlav. rodu; and this *-o/eu-s, when compared with the *-o/es of the sing. and the *-bhi(>)-o/es, *-m-os of the plur., seems to show infixation of the dual characteristic -u-. The ablative, like the dative and instrumental, apparently ended in *-5-bhi-?(m)and *-ai-bhi-8(m), constructed on the nom. of the masc. and fem.-neut. dual respectively, as Skt. devd'bhydm,GAv. zastaibyd, YAv. zastasibya, dailrdbya, OIr. feraib, tuathaib (nasalising, at least after non-nominal forms, as i n-dib t-itarib deac 'in twelve hours'); or in *-5-mb, *-ai-m6, as Goth. twaim, Lith. geriem, OCSlav. tima, rodoma (the latter, like Lith. auldm, alg6m, OCSlav. devama, Skt. kald'bhydm, affected by the vocalism of the nom. sing.). 4 The element *-/0m seems to recur in the gen. plur., as well as in the loc. sing. of -a-stems in Skt. and in the instr. sing. of the same class in OCSlav. To this general category apparently belongs the Greek 'dependent
38 For the whole problem of the Celt. forms, see F. Sommer, 'Das keltische Dual', in Miscellany Presented to Kuno Meyer 129-41, Halle, 1912, and for the Teut., most recently, Kieckers 174. Greek &Xya'is formed by analogy with atbXc. J. Plakis, Lei'u valodas rokas grdmata 60, Riga, 1926. 40Cf. R. Thurneysen, Handbuch des Alt-Irischen 148, Heidelberg, 1909. The difference in accentuation between the Lith. dat. dual auldm, algdmand the instr. aulafm,algam is scarcely original (Brugmann, 2. 2. 202-3).

196

LOUIS H. GRAY

Att. o1KOW, case' represented by Hom. 'Irotv, 6opaXo"tev, GEatv, robo~olv, OAtt. Irabot, Argol. FavaKOL. The pre-form was evidently of the type *iwro-v < *i7~rat-tv i.e. the neut. dual (with -ot- for -at- by analogy with the plur.) + the termination -t(v), which seems to be found in the Ved. instr. sing. type of devrnd < *deiuo-in-6, in the Skt. loc. sing. asmin: Av. ahmi, in Greek datives like Dor. EAL'v, Lesb. aML(v), and in the type of the Hom. abl.-instr.-loc. type of EViVPL(v). Parallel with this are Arcad. forms in -v-: lrpeo(o)vvrots ~u~vUowVv and Eo-aKoe6v with The v < of this is far from *-at-v(v). Kpavatvv, origin rots -owvv certain, though it may be explained either as similar to the alternation i/u in the loc. plur. or (more plausibly) as the sole surviving Greek relic of the dual sign *u.41 The locative has a clearly distinctive form only in the type of Av. zastay5, OCSlav. rodu, devu, mezdu, and such Lith. adverbs as pusiaui, dviejau(s), all of which imply a termination *-o/eus. Comparison with the types of the loc. sing. *deiuo-i, *aldha-i and the loc. plur. *deiuois-u/i, *alhdh/ai-s-u/i, however, leads one to infer pre-forms of the types *deiyai-o/e-u-s-u/i, etc., i.e. with infixed dual characteristic -u-, with o/e carried over from the abl. plur. in *-bhi-o/es (cf. the gen. sing. in *-o/es), but without *-u/i as in the sing. of -s-stems like Av. av5, Greek alis, Lat. penes. SUMMARY. A. In the light of the foregoing discussion, the earliest Indo-European case-endings in singular, plural, and dual for nouns in -o-, -d-, and consonants would seem to have been as follows: -o-stems -d-stems Consonantal stems

Sing. nom.
" "
41

-o-s
-o-m -o-m -o/es-o, -o/es-

-0
-d-m

-s, -0
-m -0 -o/es,-s
an inherited u con-

acc. nom.acc. neut. gen.

-d-o/es

io
Buck 316-7 thinks that AL6avoLvv 'appears to contain

nected with the u-diphthong attested by the Sanskrit and Slavic dual forms (Skt.
tdyos, ChSl. toju). The -awv is analogical, like -awv after -ow'. Derivation of the type of irirorv from *ekuoi-u-i- (A. Cuny, Le Nombre duel en grec 36-7, Paris,

1906, La Cat6gorie du duel dans les langues indo-europbennes et chamito-s6mitiques 47, 60-1, Brussels, 1930), or from a pre-form in *-omawv (cf. BrugmannThumb 272; Hirt, Grammatik 3. 67) seems less probable than the view suggested in the text.

ON INDO-EUROPEAN NOUN-DECLENSION

197

-o-stems Sing. abl. " dat. " instr. " Plur. " " loc. nom. acc. nom.acc. neut. gen. a abl. dat. -o/e-t-os -o/e-o/ei -o/e-5/8 -o/e-i -o-es -o-m-s -o/ei, -d/ -o/e-6/1m -oi-s

-d-stems -d-t-os -d-o/ei -d-5/j -d-i

Consonantal stems -t-os -o/ei -bhi, -5/1, -m -mi, -i, -0 (-m, -n) -ns -, -o - /Am

f--

-s-es

"
"

"

-d-6/'m -d/ai-s

f-bhi-o/es,
-bh-os, -m-os

" instr. " loc. Dual nom.

ace. Dualnom.ac nom.ace. neut. gen.

-oi-s-u/i
-6/(u) -ai -oi-o/e-u-s

-/a/i-s-u/i
-ai

-bhf(s),-mf(s) -s-u/i
-e -i < -ei

-ai-o/e-u-s -ai-mo

-o/e-u-s

abl.
" instr. loc.

-i-bh-e(m) -ai-bhi-e(m)
-imon -oi-o/e-u-s-u/i -ai-o/e-u-s-u/i -o/e-u-s-u/i

B. There seems to have been essentially but a single scheme for the inflexion of nouns and of non-personal pronouns. The sole material difference between such pronouns and stems in -o- and -d-, on the one hand, and consonantal stems, on the other, is that in the latter, the case-endings are attached directly to the base in the dual and plural, while in the former, all cases except the nom., ace., and voc. (i.e. the three original cases) are built upon the dual and plural neuter in *-ai and *-o/ei, *-d/9 respectively. Justification for the sharp current distinction between 'nominal' and 'pronominal' declension seems not so certain as is commonly supposed. C. The gen. sing. appears to have been constructed on the neut. as a vague, collective, and generalising nominal form42without special case42 Cf. the rendering of (7rreorst) YCvucK7 by (holev) serakan 'generic' in the Arm. version of the grammar of Dionysios Thrax (ed. Cierbed 34, 52, Paris, 1830; cf.

ib. 50). ser = yfivov,

198

LOUIS H. GRAY

force. Augmentation of this 'indefinite case' or of the noun-stem by various grades of the formative *-e6-/?i- (again without specific caseendings-cf. consonantal neuters ending in zero, e.g. *menes, Skt. mdnas, Greek pjos, and perhaps Kanis. neuters of the type of irmala) has led to creation of Italo-Celtic and Ligurian 'genitives' in -i, of Latin-Gaulish compounds in -i, of the Greek types 6/Eueat, Xaialc, of bases in *-e'i (Greek &6pts, Lat. fidas), and of feminines in *4 (type: Skt. devi'), a similar rble being played by the formative *- yu/i- (Skt. vadhiv'-,sinit-, Greek rnxvs, patrXebs). The dat. and loc. sing. and the loc. dual and plur. seem to derive from the same pair; instrumentals in *-mf-, *-bht-, appear as grades 4IZ b and ZZ b of the formative bases *mo/ei6/ei- and *bho/ei6lei'- respectively. The whole case-system, except in the nom., acc., and voc., was apparently made by adjectival formation rather than by agglutination of 'particles' of unknown meaning; and 'oblique cases' may more properly be termed 'dependent' or 'relational'. These case-endings, mere grades of bases without inflexion, imply, it would seem, a pre-inflexional stage of Indo-European. D. The Indo-European 'fem.' in *-d/9 is apparently derived from a chance resemblance of the neut. plur. in *--/a to the nom. sing. of the one type of words which seems in Indo-European originally to have distinguished between male and female: the non-personal pronoun of the type of *so-s, *to-m (animate male), *sd, *td-m (animate female), *to-d (inanimate). The feminine stems in -I-,-I-, of later origin, were then modelled on the feminines in -d-. E. The current view that the plur. was formed subsequently to the sing. receives additional support. The chronological position of the dual is less clear. If one may argue from its loc. *-o/e-u-s-u/i, evidently formed by prefixing the dual -u- to the plur. *-s-u/i, it would appear to be later than the plur. It is obviously younger than the sing. F. To the distinction between thematic and athematic verbs corresponds, it would seem, that between thematic and athematic nouns (i.e. stems in vowels and consonants respectively).43 It may also be noteworthy that thematic nouns and verbs are precisely those which show fixed accent as contrasted with the varying accent of the athematic types. G. The elements o/e, a in the vocalic stems are clearly no part of the original base, but are mere determinatives, so that we should, properly speaking, write *aul-o-, not *aulo-, etc. This is evident not only from
43 Cf. Meillet, Introduction l1'6tudecomparative des langues indo-europ6ennes5 151-3, 217-21, Paris, 1922.

ON INDO-EUROPEAN NOUN-DECLENSION

199

such pairs as Skt. padd- 'footstep', Greek Arbov'ground': pdd-, a ros 'foot'; Skt. tdna- 'posterity': tdn- 'duration',4 but more especially from the apophonic conditions, since padd-, for instance, if original, would involve the impossible sequence NN, *pedd- (and even PN, as in Skt. 'auka- 'purity': d6ka-'flame').45 It would accordingly seem that -o-stems (thematic) were developed secondarily, only consonant-stems (athematic) being known in the most primitive stages of Indo-European, and -a-stems being still later than those in -o-. The fixed accent also favours the hypothesis here advanced. In like manner, the thematic verb would appear to be later than the athematic because of the fixed accent of the former type,46 but, unlike verbs, nouns with disyllabic bases seem originally to have had only the zero-grade in the second syllable.
44Cf. Brugmann 2. 1. 156-66. '4 One cannot assume *pedo-, for the types of Greek atX6s,Lith. aiflas, Lat. servus,Gaul. Andecamulos,OCSlav. rabf~presuppose *-os, not *-os, 'Swasecundum'

giving a in Greek, Italic, and Celtic, i in Lithuanian, and I in Slavic. 46 Contrast, for example, Skt. bhdrami, bhdramaswith dddami, dadmds.