rni^m^mmmmmmmmm. M i COEHELL OTIVEESITY LIBRARY. This bael^is ngk'io be taken from the_,^;^a4ing Moor © m i fz WiTH, RETURN AT OWCE TO SHELf ^S- ^ fyxmll ^nivmH^ J '/xhxmi^ BOUGHT WITH THE INCO FROM THE SAGE ENDOWMENT THE GIFT OF 1891 ME FUND J\....H.3-lt«^S. CORNELL UNIVERSITY LIBRARY 3 1924 077 098 543 Cornell University Library The original of tiiis book is in tine Cornell University Library. There are no known copyright restrictions in the United States on the use of the text. http://www.archive.org/details/cu31924077098543 COMPARATIVE GRAMMAR OP THE INDO-GERMANIG LANGUAGES. Da muss aich manches Ratael loseu Doch roanches Rateel kniipft aich auch. Goethe. A COMPARATIVE GRAMMAR OF THE INDO-GERMANIC LANGUAGES. A CONCISE EXPOSITION OF THE HISTORY OF Sanskrit, Old Iranian (Avestic and Old Persian), Old Armenian, Greek, Latin, Umbro - Samnitic, Old Irish, Gothic, Old High German, Lithuanian and Old Church Slatonic BY KARL BRUGMANN, professor of comparative philology in the university of LEIPZIG. VOLUME III. MORPHOLOGY, PAETII: NUMERALS. INFLEXION OF NOUNS AND PRONOUNS. TRANSLATED FROM THE GERMAN R. SEYMOUR CONWAY, M. A. and W. H. D. ROUSE, M. A. FELI-OW OF GONVILLE AND CAIUS COLLEGE, FELLOW OF CHRIST*S COLLEGE, CAMBRIDGE, CAMBRIDGE, AUTHOR OF 'VERNER*S LAW IN ASSISTANT MASTER AT CHELTENHAM ITALY*. COLLEGE. NEW-YORK. B. WESTERMANN & CO., 812 BROADWAY. 1892. TO JOHN PEILE, DOCTOR or LETTERS, MASTER OP CHRIST'S COLLEGE, THE FOUNDER OP THE STUDY OF COMPARATIVE PHILOLOGY IN CAMBRIDGE THIS TRANSLATION IS DEDICATED IN TOKEN OF THEIR GRATITUDE AND AFFECTION BY HIS OLD PUPILS. TRANSLATOR'S PREFACE. In this volume I have used, though sparingly, the terms suggested in the preface to volume II — re-form ate (formate), ad-formate, trans for mate. These are applied to single words, as on p. 30, Kem. 1. When a word is modified by the analogy of another, it is said to be an ad-formate of it (p. 29, line 7 from the bottom, is an example). In its new shape it is transformed from the old , or a transformate of it (p. 44, footnote). Absolutely regarded, it is a re-formate (sometimes, where there can be no mistake , the simple word formate stands). Re-formation and transformation are used when not single words, but groups, come in question (as p. 90, line 6 from bottom); also when certain sound changes are exemplified by the words cited (as the z in sibunzo ahtozo, p. 40). These terms may by ugly, but they are so very convenient that their ugliness will, it is hoped, be forgiven. In such words as Pali, Prakrit, Gathic the quantity has not always been marked. It seemed needless to do so when this had been indicated often enough to ensure its being re- membered. The word polysyllable is used to include dissyllables, unless otherwise implied. VIII Translator's Preface. I had hoped to get out this volume by Christmas last. The delay is due partly to the waste of time in sending proofs to and fro from Germany, and partly to the almost ceaseless pressure of other duties. Mr. Conway's criticism and advice has been very useful all through, and I take the opportunity of thanking him for it. W. H. D. KousE. Cheltenham, May 30, 1892. CORRIGENDA TO VOLUME 11. page X line 1 for mascnline read masculine „ 23 „ 19 n Clasne ., Classes „ 360 „ 14 „ from below for -ieii- read -ieu-., with stop. „ 366 „ 5 „ for 116 read 116 — „ 395 „ 15 „ „ fiSdvr-Tf read fiSd-vT-a „ 434, footnote 2) „ live „ line „ 437 line 11 from below for noveji read nove-ji „ 443 „ 12 „ „ „ Sfl-dq „ flS-tOi „ 474 „ 13 „ „ „ ijelta- „ gelta „ 486 „ 9 „ „ „ *snu-di „ *suii-di CONTENTS OF VOLUME III. Page. PREFACE . . . . VII COEBIGENDA IX THE NUMERALS: General Remarks (§164) 1 Cardinals, Abstract Numerals, and Ordinals (§§ 165 — 181). One (§ 165) 4 Two (§ 166) 6 Three (§ 167) . 8 Four (§ 168) . . 9 Five (§ 169) 13 Six (§ 170) 15 Seven (§ 171) . . ... 17 Bight (§ 172) 19 Nine (§ 173) 21 Ten (§ 174) 22 Eleven to Nineteen (§ 175) 24 Twenty to Ninety-nine (§ 176—178) 29 Hundred (§ 179) 42 Two Hundred to Nine Hundred (§ 180) ... ... 44 Thousand (§ 181) 47 Multiplicatives and Distributives (§§ 182—183) ... 48 THE CASES OF NOUNS: (§§ 184-406) General Remarks (§ 184—189) 52 Singular: Nominative Masculine and Feminine (§§ 190-199) . 66 Vocative Masculine and Feminine (§§ 200-210) . . 81 Accusative Masculine and Feminine (§§ 211 — 221) . 88 Nominative and Accusative Neuter (§§ 222—227) . . 99 Genitive f-Ablative) (§§ 228—239) Ill Ablative (§§ 240-244) 133 Dative (§§ 245-255) 143 Locative (§§ 256-273) 156 Instrumental (§§ 274-282) 173 XII Contents. Dual: Nominative and Accusative Masculine and Feminine (§§ 283-291) 189 Nominative and Accusative Neuter (§§ 292—295) . 197 Dative, Ablative, and Instrumental (§§ 296—305) . . 200 Genitive and Locative (§§ 306—311) 205 Plural: Nominative Masculine and Feminine (§§ 312—324) . 211 Accusative Masculine and Feminine (§§ 325—336) . 224 Nominative and Accusative Neuter (§§ 837 — 343) . . 236 Genitive (§§ 344—355) 244 Locative (§§ 356-366) 256 Dative-Ablative (§§ 367-378) 266 Instrumental (§§ 879—389) 273 Metaplastic systems of Declension (§§ 390—406) . . 280 Tables ofNoun Declension 296 PRONOUISfS (§§ 407-459) 322 Pronouns with Gender (§ 408) 325 Etymology and Morphology (§§ 409—411) 327 Their Cases (§§ 412—459) 384 Singular: Nominative Masculine and Feminine (§§ 413—416) . 385 Nominative and Accusative Neuter (§ 417) .... 388 Genitive (§§ 418—420) 339 Instrumental (§§ 421—422) 344 Ablative, Dative and Locative (§§ 423-425) ... 346 Dual (§ 426) 352 Plural: Nominative Masculine and Feminine (§ 427) . . . 352 Nominative and Accusative Neuter (§ 428) . -tt- • 854 Genitive (§ 429) 355 . Loeatiye, Dative-Ablative, and Instrumental (§§ 430-432) 35 6 Declension . . . .'i". '.'~\. 360 Personal Pronoun^ (Pronouns without Gender), and ^ Ui-eir P.naa-eg_8ixe3„X§§-^S3--459)„ , — . — __^ . - , , 35R ^ Etymology; Formation of the stem (§§ 434—438) 364 Nominative (§§ 439—441) 372 Accusative (§§ 442 - 443) 375 Ablative (§ 444) .... .... 879 Dative (§§ 445-446) ... ... 380 Locative (§§ 447—448) 334 Instrumental (§ 449) 388 Genitive, and the Possessive Adjectives (§§ 450—456) . . . 388 The Bemaining Cases, and the Possessives (§§ 457 — 459) . . 396 THE NUMERALS.') § 164. In the original Indo-Germanic language, the numerals 1 to 999 were expressed in one of three ways. Some were simple words, as *tri- 'three' (Skr. tri-); some were compounds, as 1) Bopp, Yergl. Gr. II» 55 ff. Schleicher, Compend." p. 477 ff. Bopp, Tiber die Zahlworter im Sanskr., Griech., Lat., Litth., Goth, und Altslaw., Abh. der Berliner Akad. 1833 p. 163 ff. Lepsius, tJber den Ursprung und die Yerwandtschaft der Zahlworter in der Indogerm., Semit. und der Koptischen Sprache, in 'Zwei sprachvergleichende Abhandlungen', Berlin 1836, p. 81 ff. J. Grimm, Gesohiohte der deutsoh. Sprache* 167 ff. Pott, die quinare und vigesimale Zahlmethode bei Volkern aUer Welttheile , nebst ausfuhrlichen Bemerkungen fiber die Zahlworter Indo- germanisohen Stammes, Halle 1847. Id., Die Sprachverschiedenheit in Europa an den Zahlwortern nachgewiesen, sowie die quinare und vigesimale Zahlmethode, Halle 1868. E. Schrader, Tiber den -Tlrsprung und die Bedeutung der Zahlworter in der indoeurop. Sprache, Stendal 1854. Zehetmayr, Verbalbedeutung der Zahlworter, als Beitrag zur Be- -leuchtung des urspriingl. Verhaltnisses der indogerm. Sprachen zum semit. Sprachstamme, Leipz. 1854. W. "Wackernagel, tJber Zahl und Ziffern, Michaelis' Ztschr. ftir Stenogr. 1855. Bernloew, Recherches sur I'origine des noms de nombre japh^tiques et s^mitiques, Giessen 1861. Krause, TJber den Ursprung und die Bedeutung der Zahlworter, Ztschr. fiir osterr. Gymn. 1865 p. 867 f. J. Schmidt, TJber einige numeralia multiplicatira, Kuhn's Ztschr. XVI 430 ff. — Ed. Mil Her, Spraohver- gleichendes iiber die Numeralia, Fleckeisen's Jahrbuoher fur class. Phil. 97, p. 585 f. Ascoli, TJber eine Gruppe indogermanischer Endungen, Krit. Stud. 85 ff. Osthoff, Formassociation bei Zahlwortern, Morph. Tint. I 92 ff. 'J. Baunack, Formassociation bei den indogerm. Numeralieu mit besonderer Beruckaiohtigung der griechischen , Kuhn's Ztschr. XXV 225 ff. J. Wackernagel, Zum Zahlwort, jfii'd. 260 ff. The Author, Die Bildung der Zehner und der Hunderter in den idg. Sprachen, Morph- Unt. V 1 ff., 138 ff., 268. B rug-mann, Elements. III. 1 The Numerals. § 164. *dud-deki^ 'twelve' (Skr. dvd-daia) ; and some were expressed by phrases, as *treies qe tit^cmti qe 'twenty-three' (Skr. trdyak ca viSatU ca). Simple words existed only for the numbers 1 to 10, and 100. We find in the second stage, when the various branches of the language had begun to develope on their own lines, simple words for 1000, as Skr. sa-hdsra-m, Gr. Lesb. /£A./l-^ot; but it is uncertain whether a corresponding form *gheslo- existed in the proethnic period, or whether the phrase 'ten hundreds' (cp. Skr. daia-iatf f.) was the sole expression for this number. If the simple words for 1000 were not earlier than the second stage, the change was similar to one which took place in Greek, where Homer's dsud-yn'koi '10,000' was replaced later by /nvpioi. The word *defcm 'decem' played an important part in the Indo- Germanic decimal system. It is in the highest degree probable that the Indo-Germanic elements *-Jcmt- and *-1comt- which appear in the expressions for multiples of ten (Gr. Dor. fl-y.un and "Whitney, Sanskrit Grammar, p. 160 if. Spiegel, Gramm. d. altbaktr. Spr. p. 176 ff. — G. Meyer, Griech. Gr.^ p. 372 ff. The Author, Gr. Gr. (Iwan MuUer's Handb. II^) p. 135 ff. Ahrens, Ein Beitrag zur Etymologie der griech. Zahlworter, Kuhn's Ztschr. Till 329 ff. H. E b e r t , Quaestionum de rocabulorum cum numeralibus Graecis com- positorum formis ac signifioationibus specimen, Spandau 1858. — Stolz, Lat. Gr. (Iwan MuUer's Handb. II 2) p. 349 ff. Neue, Pormenlehre der lat. Spr. II 2 p. 144 ff. Merguet, Die Entwickelung der lat. Formen- bildung p. 132 ff. Aufrecht, Die lat. Zahladverbien auf iens^ Kuhn's Ztschr. I 121 ff. — Zeuss-Ebel, Gramm. Celt.^ p. 300 sqq. Stokes, Bezzenb. Beitr. XI 166 ff. — J. Grimm, Deutsche Gramm. Ill 226 ff., 634 ff. Ic?., liber die zusammengesetzten Zahlen, Germania I 18 ff. Holtzmann, tlber das deutsche Duodecimalsystem , Germania I 217 ff. Id., Das Grosshundert bei den Gothen , {bid. II 424 f. Rumpelt, Die deutsch. Zahlworter sprachvergleiohend dargestellt, Bresl. 1864. Id., Die deutschen Pronomina und Zahlworter, 1870. Scherer, Zur Gesch. der deutsch. Spr.^ 576 ff. Kluge, Zu den german. Numeralien , Paul-Braune's Beitr. VI 393 ff. Id., Paul's Grundriss I 402 ff. — Schleicher, Temy imenii cislitelinychil vu litvo-slavjanskomu i nemeokomu jazykachii (Prilozenie kti X. tomu zapisoku Imp. Ak. Nauku), St. Petersburg 1866. Id., Lit. Gr. p. 149 ff. Kurschat, Gr. der litt. Spr. p. 259 ff. Miklosich, Vergl. Gr. IV 51 ff. Leskien, Handb. der altbulg. Spr.^ p. 78 ff. — Reference may also be made to the discussion of the Lycian numerals by Deeoke, Bezzenberger's Beitr. XIV 181 ff. (see especially p. 240 ff.). § 164. The Numerals. 3 XQia-mvxa^ etc.), and the word *hyit6-m "centum", were connected with *de^ip,, and came from *-dfci^t- *-d^omt- and *dlci^td-m, syncope having taken place because the first syllable was un- accented (I § 310 p. 247); see Scherer Zur Gesch. der deutsch. Spr.2 579, Bugge Bezz. Beitr. XIV 72. We are in the dark as to the precise significance of the original Indo-Germanic words for 'two' and all the following numerals. Many conjectures have been put forward, some of them not at all amiss. It has been suggested, for example, that *ter- *tr-i- 'three' may have been a name for the middle finger, connected with Skr. tdr-man- 'top of the sacrificial pillar Gr. XEQ-d-Qo-v 'end, point, top'; *pe'>9qe 'five' has been compared with Goth, figgr-s 'finger' (cp. O.H.G. fust O.C.Sl. msti 'fist' for *pt3gsti-s, n § 101 p. 306), [and the slang phrase for a fist, bunch of fives] ; and *defcm 'ten' with Gr. SsKOjuac Ssxofiai I receive' '). But many others are certainly far from the mark, as Zehetmayr's, in the work cited on the first page. Our attention will be given first to Cardinal Numbers, to Abstract Numerals — so far only as they are used along with adjectival cardinal numbers in ordinary reckoning — and to the Ordinals. The Abstract Numerals are derived from the Cardinals by the suffixes -ti- and -t- (-d-) , which serve as secondary suffixes in other abstract forms besides these (see n § 99 p. 293, § 101 p. 306 f., § 123 p. 390). Some of them were used in the proethnic period along with ordinary numerals ; instead of 'ten men', for instance, the expression 'a ten of men' served equally well. Sometimes they even drove the cardinals out of the field altogether; in Balto-Slavonic *dekrp, is not represented, but only *deJ(im-t(i)-, which appears in Lith. as des2imt(i)-, in Slav, as desqt(0-. The Ordinal Numerals contain -to- and -JWO-, suffixes used in comparison; a conjecture as to the origin of these has been given in II § 72 Kem. p. 167 and § 81 Eem. 1 p. 242. 1) Soberer, op. cit. p. 578: "It therefore seema most natural to regard the word as an ancient expression for both hands held out to receiye something". 1* 4 Cardinals, Abstract Numerals, and Ordinals. § 165. CAEDINALS, ABSTRACT IHJMEEALS, AND ORDINALS. § 165. One. In the original language, one or more derivatives from a pronominal stem oi- served to express 'one : cp. Gr. Ital. Kelt. Germ. Balto-Slav. *oi-no-s (Skr. has an enclitic ena- with the meaning 'he') , Iran. Gr. *oi-uo-s, Skr. *oi-qo-. Aryan. Skr. e-ka-s. Avest. a^-va- oi-va-, O.Pers. ai-va-. Greek, ol-vb-c ol-vv 'ace, the number one on a die', olvi] napd xdlg "IwOi luovag (Pollux YII 204), olvltsiv' to /.iovdt,iiv yMvd ylmrrav and olvwvra' /.lovT^pi] (Hesych.). Then there is the Homeric olsri^g 'of the same age, contemporary', which Wacker- nagel (Kuhn's Ztschr. XXV 280) derives from *olfo-f£T?]s by syllabic dissimilation (I § 643 pp. 481 f.), whilst oto-g Cypr. ol-fo-g meant 'alone'. Italic. Lat. oi-no-s oeno-s Unu-s, Umbr. unu 'unum' Osc. uln[itu 'unita'. Remark. Many scholars connect e-ka-s with Lat. aequo-s and Gr. aiaa 'equal portion' (for *aiAa). If *oi- belongs to the stem o-, *ai- might be connected with the feminine stem a- (§ 409). But at the same time such forms as Lat. auri-s: Gr. oiiara suggest the possibility of a similar vowel variation here. — See Hiibschmann, Das idg. Vocalsystem pp. 190 f. Old Irish, oe-n. Germanic. Goth, di-n-s O.H.G. ei-n O.Icel. ei-nn. Balto - Slavonic. Lith. v-ena-s (cp. I § 666. 1 p. 526), Pruss. ace. ai-na-n. O.C.Sl. i-nU (cp. I § 84 p. 82, § 666. 3 p. 527) , which in composition means 'one', as ino-rogu 'one- horned animal, unicorn'; elsewhere it has the meaning 'alter, alius'. *sem- was another word for 'one' in the parent language. The idea originally conveyed by it was probably that of being together or united. This became the regular numeral in Ar- menian and Greek: Armen. mi (gen. mioj) for *sin-i (I § 560 p. 416); Gr. sv- instead of ""s^c, nom. masc. in the dialect of Gortyn 'iv-q Att. sIq, fem. fiia for *(}/.i-itt, f^wvv^ 'one-hoofed' for *a/u-wvv^ (II § 160 pp. 479 f.). Compare Skr. sa-kft 'once' Gr. a-7ia^ 'once' -dnXoo-g 'single, simple', Lat. sim-plex, sin-gull, § 165. Cardinals, Abstract Numei'als, and Ordinals. 5 semel (see § 182), sem-per 'in one unbroken sequence, always', G-oth. simle 'once, once upon a time'. Isolated forms, of doubtful origin, are Horn. Gort. Im Lesb. Thess. M (cp. Osthoff, Morpli. Unt. IV 186 f.), and ^\a.N. jedinu je(Mnu, the regular word for 'one' in that language. First. In all periods, from the proethnic onwards, ordinals for this number were formed from the v^^^e/--, which is seen in Gr. TTsgav ntg-vai ngo etc. (Fick, Worterb. I^ 140 ff.), by means of the suffixes -mo- (Ar., Gr., Slav.), -mo- (Ital., Germ., Bait.), -tinmo- (Ar.) and -isto- (West-Germ, and Norse). *pf-uo-, *pf-uo-. The former became O.C.Sl. pn-vU, the latter is seen in Skr. purv-iyd-s purv-iya-s (also purva-s 'situated before') ; Arest. pourviya- paoirya- ; Gr. ngoSro-g Dor. nparo-s for *7rpto/-aro-?, besides which we have forms without the ex- tension -nvo- (cp. TQir-nro-q § 167), Dor. nQav 'formerly' for *npa-fa-v, and (with the suffix -io-) ngw-tjv 'recently' for *ngm- -f-ia-v. Cp. I § 306 p. 242, II § 63 p. 138, § 64 p. 134. Apparently we must assume a form *pro-uo- for Goth, frduja 'lord', O.H.G. fro 'lord' frouwa (= Goth. *frdujo) 'lady'; this *pro-uo- will be related to *p^-uo- in the same way as *pro-mo- to *pi--mo-. *p^-mo-. Goth, fruma O.Sax. formo A.S. forma. Lith. pirma-s. Cp. O.Ir. rem- 'ante, prae' II § 72 p. 168 and Lat. prandiu-m, which Osthoff is probably right in explaining as *pram-(e)d-iio-m 'early food' (cp. Morph. Unt. Y p. III). *pro-mo- is seen in Umbr. prumum promom 'primum': cp. Gr. ngofAo-s 'front man, front warrior, leader, prince' Goth, fram- -aldrs 'advanced in age', -mo- also occurs in Lat. primus Pelign. prismu 'primo' or primum' (I § 570 p. 427, II § 72 p. 168), which, like pris-cu-s and pris-tinu-s, is derived from a comparative form connected with prius (II § 135 pp. 433 f.). -t'inmo-: Skr. pra-thamd- (for the th cp. II § 73 p. 178), Avest. fra-tema- O.Pers. fra-tama-. Cp. Avest. fra-tara- Gr. TiQO-rsQO-g 'former, earlier'. -isto-: O.K.G.furist A.S. fyrst O.lGel fijrst-r. Cp. O.H.G. furiro 'earlier, superior'. 6 Cardinals, Abstract Numerals, and Ordinals. §§ 165,166. Alban. i-pars 'first' belongs to the same root as these words. "Words derived from other roots : Skr. ddya-s from a-di-s 'beginning'; and later adi-md-s. Armen. nax and arajin, the latter from araj 'foremost side, front, origin, beginning' (cp. verj-in 'last' from verj 'end', ver-in 'highest' from ver 'above', and the like). O.Ir. cet-ne, in composition cet- , Mod.Cymr. kyntaf, Grail. Ointu- in proper names, such as Cintu-gnatu-s ('primigenitus') ; perhaps, as Thurneysen conjectures, this is connected with Goth. hindumist-s 'extreme, outermost' A.S. hindema 'last', since the ideas of 'first' and 'last' are easily interchanged. In Irish there is a further form oen-mad, used where larger numbers follow; here the termination -mad has come from sechtm-ad 'seventh' dechm-ad 'tenth', where m is part of the stem. § 166. Two.i) The stem is *#o- *duuo- (I § 117 p. 109); in composition and in some ordinary derivatives we have *dui-, a form which recals *tr-i- 'three' (cp. II § 13 p. 28) and *'i}-i- 'two' in fi-xari etc. (§177). I find it impossible to agree with Bartholomae (Ar. Forsch. Ill 39) , who conjecturally restores *duoi- *duoi- *dui- as the primitive base of this numeral. 8\.r. dvclii' dvd {dumu dtivd), fern. neut. dve (duve). Avest. dva, fem. duye (cp. Bartholomae, Handb. § 92 p. 40). Gr. dvm (used for both masc. and fem., like Lat. duo and Lith. dialectic dit, cp. also to) atTJXS § 426), 6vo (which was perhaps originally the neuter, see § 293); Soj-Ssy.a for *<5/(b-. The Dor. and Later Att. dval Lesb. dvsai or Svsaat Thess. Svug are re-formates, apparently caused by Svmv (Svoiv) ; the relation of tqwI: rgioiv made it seem natural to coin dvai as dative to Svav. Soioi cannot be derived from *duoio- (cp. I § 180 pp. 117 f.); possibly it came from *duoi-io-, cp. Skr. dve- -dhd 'double, twice' and §§ 297, 311. 1) Benfey, Das indogerm. Thema des fZahlworts 'zwei' ist DU, Getting. 1876. — Zander, De vocabuli S6o usu Homerico Hesiodeoque et Attico, I, II, Konigsberg 1834, 1845. — Meringer, Die Flexion der ZweizaM, Kuhn's Ztschr. XXVIII 234 ff. § 166. Cardinals, Abstract Numerals, and Ordinals. 7 Lat. duo, fem. duae (cp. §§ 285, 286), ace. duo, duos, duas, dat. duo-bus, duci-hus. In Umbrian the inflexion is plural throughout: nom. masc. dur, ace. masc. fem. tuf, nom. aec. neut. tuva (cp. Lat. dua beside duo). O.Ir. dau do, older dau, and also da (§ 285), fem. d%; O.Cymr. Mid.Bret. masc. dou. Groth. tvdi, neut. tva, fem. tvos; dat. tvdim, gen. tvaddje (see § 311). The Germanic dialects show various re-formates among the cases, as O.H.G. zwene, which are not yet satisfactorily explained; the latest discussion of them is by Kluge in Paul's Grundriss I 403. Lith. masc. du for *dvu, fem. dvl for *dve, see I § 184 p. 160, § 664. 3 p. 528. O.C.Sl. dva dma, fem. neut. dv6 duv6. *dui-. Examples of this base in composition are: Skr. dvi- -pdd- Gr. Si-novg Lat. hi-pes A. 8. twi-fete 'two-footed', O.H.Q. zwi-valt 'two-fold'. In derivatives : *dui-qo- : Skr. dvi-ka- 'consist- ing of two' Gr. dtoao-g Strro-g 'two-fold' for *SFi-y.-i.o-g, O.H.G. zweh'o 'doubt' A.S. ttoi'^ 'twig, branch' (II § 86 p. 257). In Italic we find cZm- (as well as *d%d-) : Lat. du-plu-s du-plex du-cenU (cp. § 180), Umbr. du-pursus 'bipedibus'; this was a re-formation, developed possibly with the aid of quadru- (II § 34 p. 61). Remark 1. Side by side with Lat. bi-, dm- is found twice in com- position, dui-dens and dui-census (Paul. Fest.). This, like O.Lat. duis = bis (§ 182), may have preserved an Idg. *duiii- (cp. Ved. duvis beside dvis), or it may simply be a modification of bi- on the analogy of duo (and of dvr-) ; I leave the matter undecided. In either case we may reject the view of Skutsch (De nominum Latinorum compositione , Nissae 1888, p. 35), who holds that bi- arose from the dui- which is preserved in these two compounds. On the other hand, I agree with this scholar in regarding di- in late compounds (as di-loris) as borrowed from the Greek S^ (p. 36). Armen. erJcu (gen. dat. erku-c), of uncertain origin. Bugge (Beitr. zur etym. Erlauterung der arm. Spr., 41 f.) derives the word from *ku- = Idg. *diid(u), with er- prefixed on the analogy of ereM 'three'; a most daring suggestion. Pr. Miiller would connect it with Suanian JSru Georgian ori 'two'. Second. For this numeral the different languages show very different forms. Cardinals, Abstract Numerals, and Ordinals. §§ 166,167. Skr. dvi-ttya-s , Avest. hi-tya- Gathic d'^hi-tya- O.Pers. I- (I § 159 p. 143). Armen. erkir and erkr-ord (for *erhir-ord). Remark 2. Perhaps erhir, like corir 'fourth', was formed on the analogy of eri-r 'third'. The termination -ord, found in erkr-ord and all the numerals which follow, is very common in other words besides numerals: e. g. hanapaz-ord 'daily' from hanapaz 'always', parap-ord 'otiosus' from parap 'otium', ors-ord 'hunter' from ors 'hunt', Ic-ord 'com- panion, avCvyog' from luc 'yoke'. Petermann (Grammatica Linguae Armen. p. 162) and Bopp (Vergl. Gr. 11^ 97 f.) offer very questionable conjectures as to its origin. Gr. SsvTspog properly means 'removed to a distance from something, at a distance from it, coming after it in time or position' {Ssvraroi; is also found). It is etymologically connected with dsvofiai and the Sanskrit adj. du-rd- 'far off, afar', and did not belong to the numerals until Greek had become a separate language. Its similarity in sound with dWw certainly had some- thing to do with this new use. See the Author, Kuhn's Ztschr. XXY 298 ff. Lat. secundu-s, connected with sequor; see II § 69 p. 161. O.Ir. aile (Mid. Cymr. Mod. Bret, eil) = Lat. alius, and tanise, which is connected with imthanad 'change', though nothing further is known of its etymology. Goth, an-par O.H.G. andar. Lith. antra-s O.C.Sl. vutoru. Cp. n § 75 p. 198. It possibly may be analysed vu-toru (op. Skr. u-bhaii), see § 285. § 167. Three. Idg. masc. neut. *tr-i-. The -i- was a suffix, as is proved by such forms as Skr. ff-ttya-s Gr. Lesb. rsp-To-g. Skr. trdy-as, loc. tri-su. Avest. pray-o, loc. pri-sva. Armen. ere-M , gen. dat. eri-c , instr. eri-vK , see I § 263 p. 213. Gr. TQsXq Gortyn. TQltq for ^xQsi^-sg, loc. rpi-al; Hipponax has TQioTai, which was suggested by rptmv (tqIo). The Boeot. TQs-nsSSa (beside rpi-novg etc.), taken in conjunction with Lat. tre-centf, Lith. tre-czia-s O.C.Sl. tre-tij% 'third' Lith. tre-ji 'three by three', seems to point to an old stem *tr-e- (*tr-o-). §§ 167,168. Cardinals, Abstract Numerals, and Ordinals. 9 Lat. tres, dat. abl. tri-bus, Umbr. trif tref ace. 'tres' triia 'tria'. O.Ir. tn dat. tri-h. Goth, preis dat. pri-in, O.H.G. dn dat. dri-m. Lith. trys loc. tri-se. O.C.Sl. tnj-e trij-e loc. tn-chu. For the feminine there was a special form in the parent language , preserved only in Aryan and Keltic , *tiser- *tisr- : nom. Skr. tisr-ds (for the weak form of the stem see § 320) Avest. tisar-o, O.Ir. teoir O.Cymr. teir (cp. I § 576 p. 431). It is conjectured that *tisr- came from *tri-sr- by dissimilation, and that its second part is identical with the second part of *sue-sor- 'sister'; see Bugge, Bezzenb. Beitr. XIV 75 f. Skr. cdtasr-as is a similar formation (see § 168). Third. The Indo - Germanic languages have forms with -to-; those without the -i- of *tr-i- may be considered the oldest: Skr. tf-tiya-s, Gr. Lesb. rig-rog, Lat. ter-tiu-s Umbr. tertim 'tertium', Pruss. tir-ti-s ace. tir-tie-n {tir- = *tf-). The following have Hr-i-. A.rest. pri-tya- O.Pers. si-tiya-. Gr. Att. etc. rpi-TO-g, and the Homeric rglr-aTo-Q on the analogy of sivaro-i; ^syMXO-g, cp. npmTO-g for *7tpa}f-aro-g § 165 p. 5 and s^Sofi-aro-g § 171. Lat. trit-avo-s^ unless the true form of the word be strit-avo-s, see II § 81 p. 246. Mod.Cymr. trydydd for *tri-tiio- or for *t^-tiio-, we cannot tell which. With different suffixes: Armen. eri-r (and err-ord for *erir-ord, cp. § 166 Kem. 2) and O.Ir. tri-s, in composition tress- (see II § 81 p. 247). For tr-e- in Lith. treczia-s for Hretia-s and in O.C.Sl. trettfi, see last page. The last-named forms make it doubtful whether Goth, pridja O.H.G. dritto are derived from *tri-tio-, or from *tre-tio- (ac- cording to I § 67. 8 p. 57). §168. Four. The Idg. stem masc. nent *qetuer- *qetuor- had a variety of ablaut - forms ; this was because there were several distinct weak-grade forms of the second syllable : *q(e)tur- *q(e)tur- *q{e)tru- *q(e)ttff- *q(e)tyf-. Cp. I § 155 p. 140, and J. Schmidt, Kuhn's Ztschr. XXV 43 ff. and 138, Pluralbild. 191 f. ; Wackernagel, Kuhn's Ztschr. XXV 283 ff., XXVIII 136; 10 Cardinals, Abstract; Numerals, and Ordinals. § 168. G. Meyer, Gv. Gr.2 376 f . ; Osthoff, Phil. Rundsch. I 1592, Morph.Unt.IY 333; Kluge, Paul-Braune's Beitr. VIII 517 ff., Paul's Grundr. I 403 f. The nom. plur. masc. in the original language would be *qetuor-e8: Dor. tsxoqsi; Lat. quattuor Skr. catmr-as; neut. *qetudr *qet'!fdr-9 : Skr. catvari, Lat. quattuor^ Goth, fidvor, whose passed over into the masculine. Skr. catvdr-as ace. catiir-as; in composition catur- instead of pr. Ar. *catru-. Avest. capwar-o; catur-a- occurring four times'; in composition capru-, as capru-karana- 'four-cornered', and captoar^-, as capwar^-zawgra- 'four-footed*; the latter was probably suggested by capwar^-sat- '40' (see §§ 176, 178, and the Author, Morph. Unt. Y 80). For the ablaut in the first syllable cp. Avest. O-xtUirya- 'occurring four times' and the ordinals Skr. tur-ya- tur-tya- Avest. tuirya- for *ktur- (I § 646 p. 491). Armen. cor-K, gen. cor-i-c, from *qetmr- or *qetur-, see I § 455 p. 336. Also Mr-, in Uar-a-sun '40', probably from *(q)tyi-, see §§ 176, '178. Gr. Dor. tstoqss, Late Ion. TtnatQsg Arcad. Tsaatpd-novta, Lesb. mavQsc, Horn, niavgiq, Att. rsmxpig Hom. reaaapsg Boeot. nhrapeg, dat. Hom. TSTga-ai. Cp. I § 166 p. 147. n- in nsavQsg nicyvQeg nsTvagsg can hardly have been taken over from *nTQa- {xQo.Tii'Qa) and *nrQv- (rgv-ffjaXsia) , since these had dropped their n- in the proethnic Greek period. More pro- bably it came from niwB. Two other forms are tstqu- and *(n)TQa-. The former is seen in rsTga-at, vsTgd-xig 'four times', TSTga-To-g (beside rsrapTO-g) and in most compounds, as xetqu- -xvKlog 'four-wheeled; the latter in rgd-ns^a 'table' for *(Tr)rQd- -nsta (beside rap- in rdgmv = TfTrdgcov, a word used by the comedian Amphis, and in ruprTj^opiov = TsrapTi]/j.6piov, preserved by Hesychius ')■ tstqu- stands for *TSTfpa- Cqetiif-) and *(n)TQa- 1} Hesyohius' explanation, to TfiTrj/joQioy, seems to be corrupt; read TO rera^Tinudgior. It is not at all probable that this word has preserved a form r«pTo- belonging to Skr. triiya-s Pruss. Urti-s 'third'. §168. Cardinals, Abstract Numerals, and Ordinals. 11 for *(n)Tfga- i*qtu^-), f having been lost in proethnic Greek (the Author, Gr. Gr.2 pp. 43, 71). Tfrpio-y.ovra iov *qetyf-, see §§ 176, 178, 341. Tv^rmo-g, if Pott is right in adding this to the list (cp. TpiToio-g), should be compared with Skr. tiir-ya-; TVQ- instead of *-nTVQ- on the analogy of tqv-. With Homer's Tsaaapd-^oio-Q 'worth four oxen' compare another Homeric form, ifixoad-jjoio-g (beside Svwxauixoai'-TT tj^v-g). liAt. quattuor, quadru- (in composition), quadra-ginta instead of *quettuor^ *quedru-, *quedra-, being assimilated to quartu-s (see next page), quattuor, nom. pi. masc. and nom. ace. neut. (see last page), dropped its inflexions because the following numerals (quinque etc.) were not inflected, quadra- stands for *qett!f-, see §§ 176, 178. -d- has taken the place of -i-, reminding us of a similar change of the breathed to the voiced sound in angulu-s for *aidklo-, septin-genti septua-ginta (I § 499 p. 366, m § 177), Gr. s^do^w-g O.C.Sl. sedml (§ 171). Umbr. petur- -pursus 'quadrupedibus', Osc. petora 'quattuor' (Pest.) and petiro- -pert 'quater'. O.Ir. cethir (dat. ceihrib), O.Cymr. petguar. Gall. Petru- -coriu-s and petor-ritum 'four-wheeled vehicle'. Goth, fidvor for *fidvor-(i)z like stiur 'steer, ox' for *stiur(a)-z (I § 660. 6 p. 516, HI § 194), dsA. fidvori-m, see § 169; Jidur- -dogs 'lasting four days' (fidur- ? or fidur- instead of *fida'Ar- because the second syllable was unaccented?). The ^-sound which appears in Gothic is not found elsewhere, except in A.S. and O.Swed. compounds; e. g. A.S. fyder-fete 'four-footed' for pr. Germ. *fl^ur- (H § 19 p. 36). O.H.G. fior O.Sax. fiwar A.S. feower O.Icel. masc. fjorer neut. fjogor fjugur point to a form *ktfek^or- *kuekur- before the great Sound-shifting [Laut- verschiebung) in proethnic Germanic; for the change of -jm- to -u- see I § 444 c p. 330. I assume that in *kiietuor-, -tu- was assimilated to the initial guttural (cp. *pempe Goth. Jimf for *peKqe, § 169 p. 14); then *kuetur- followed suit, and became *kuekur-. In Gothic, on the other hand, fidur- held its ground, and fidmr (instead of *fi(^)vdr) has been assimilated to it. 12 Cardinals, Abstract Numerals, and Ordinals. § 168. Remark 1. I have discussed this - ntvrt-Tttlavro-g) we find vtvza- (e. g. ntvTU-xooiot, TTevrd-nrj/v-q)^ which is a re-formation following the model of rsvQa-, s-nm-, sva-, tvvsa-, Ssxa-. Remark 1. Two stems are found; nefinds Uke Skr. paficdt-, and a <8-stem with the same meaning, Skr. panhti-s O.Icel. fimt O.Sl. pqti. The 14 Cardinals, Abstract N^umerals, and Ordinals. § 169. first two words are ad-formates of Sexaf and dasat- respectively ; and considering how widely the suffix -aS- was used in Greek — /joviq, evdg, Svaq, Tgiag, TST^dc, fjac, i^So/iai (cp. ffiSo/uij-xoVTa) fTiTag, oySodq (cp. oySori- -xovTo) otrrag^ fvj'fa'e, zeTTaQfcxovrd gj exarovrdgj ^'-^idgj ^uv^tdg — it is extremely doubtful whether there is any immediate historical connexion between Trsfinig and pancdt-. For the -S- of -aS- see II § 123 p. 392 and III § 178. Lat. quinque quinque (for f, see Thurneysen in Kuhn's Ztschr. XXX 501 f.) , Umbr. pumperias Osc. pumperias, equivalent to 'quintiliae' or 'quincuriae', Osc. Pumpaiians 'Pompeianus'. Pr.Ital. *kuefdkue, see I § 336 p. 267. quincu-, in quincu-plex etc., through association with quadru-. O.Ir. coic, O.Cymr. pimp, Gall. nsf-msdovXa nsvrd(pvXlov' (Dioscor.), see I § 436 p. 324. Pr.Kelt. *kuei3kue, see I § 389 p. 269. Goth. Jimf O.H.G. fimf finf (the u of O.H.G. funf is dis- cussed below under the ordinal). Probably the second / is to be explained by supposing that *pei9kue became *pempe (cp. I § 444 Rem. 1 pp. 329 f.) as *hif,etuor- became *kuekuor- (III § 168 p. 11). The 2-inflexion, which we see in Goth, fimfim O.H.G. finfin, is discussed in § 168 Kem. 2 p. 12. Lith. penki and penk-erl, see § 168 p. 12. In Slavonic, the cardinals 5 to 10 inclusive were represented by the abstract formation: p^t% "fivefold character, the number five' (= Skr. pankti-s O.Icel. fimt) governing the gen. pi. of the thing. The old numerals were indeclinable, and this may have had something to do with their being dropped. Bemark 2. Be it observed in passing that the Albanian numerals 5 to 10 are based upon these same i«-abstracts : pese 'fire', §asU 'six', staU 'seven', teU 'eight', nende 'nine', SCeU diets 'ten'. See Or, Meyer, Albanes. Stud. II 50 ff. Fifth. Idg. *pfq-to- (which can be traced with certainty in Germanic, but nowhere else) ; and perhaps *pei9q-to- too is proethnic (cp. *pet9qe). Skr. paficamd-s (following saptamd-s etc.) and pafica-tha-s (cp. O.Ir. coiced). Avest. puxda- (for -^-, cp. uxcta- I § 475 p. 351), according to von Fierlinger (Kuhn's Ztschr. XXVII 194), comes from *p't9qto-; but why should it have u and not a? The u reminds us of Gr. nvy/Li'i] Lat. pugnu-s. Besides puxda- we find the further form Avest. patgtcmhe-m ace. 'one -fifth'. §§169,170. Cardinals, Abstract Numerals, and Ordinals. 15 Annen. hing-er-ord. Remarks. For -ord, see § 166 Rem. 2 p. 8. The -er- which precedes -ord in this and the succeeding numerals is still unexplained. Gr. ns/x7t-To-e, Gortyn. nswo-g (I § 427 a p. 312). Lat. quintu-s QuJncUu-s, Osc. Piintiis no/nnviss 'Quinctius'. The ground-form may be either *pei9qto- or *pKqto-. Bar- tholomae (Kuhn's Ztschr. XXIX 503) conjectures that Lat. quint- Osc. pont- were the regular forms , while the cardinals were responsible for -c- in quinct- and -p- in pompt-. O.Ir. coiced O.Cymr. pimphet, see II § 81 p. 247. Goth, fimfta in fimfta-taihunda 'fifteenth', O.H.G. fimfto finfto. A form *fu't9xta- = *p9qto- must be assumed for pr. Germ, to explain Mod.H.G. Swab, fuchze '15' fuft 'fifth', O.H.G. funfto funf, Mod.H.G. Rhine-Prank, fufzen fufzich etc. ; see Kauifmann, Paul-Braune's Beitr. XII 512. Compare too O.H.G. fmi (pr. Germ. *futdxsti-z) O.C.Sl. pqsU 'fist' common ground- form *pi9qsti-s, H § 101 p. 306 f. Lith. penkta-s. O.C.Sl. pqtu may stand for *pet9q-to- or *pf9q-to-. § 170. Six. Three forms may be restored with more or less probability. Iranian, Greek, and Keltic point to *sueTcs; Armenian and Baltic to *»«e^s ; Sanskrit, Latin, Germanic, Baltic, and Albanian to *se^s (Alban. gaste, see G. Meyer, Alban. Stud. II 56 fif.). *sueks and *ue'ics would be parallel forms like ■\^ suelq- and uelq- 'draw' (Gr. sIkco Lat. sulcus: Lith. vdkU) and other pairs of the same kind ; see I § 589. 3 pp. 445 f. *suehs and seks, again, recal such pairs as *sue- (Skr. svd- 'suus' etc.) and *se- (Avest. he hoi, Lat. se, Goth, si-k), *suesor- (Skr. svdsar- 'sister' etc.) and *sesor- (Lith. sesu O.C.Sl. sestra) and so forth; see I § 170 p. 150, § 184 p. 160 (and see II p. 441 footnote 2), § 187 p. 162. Both in the prehistoric parent language, and in the historic period of Aryan, Armenian, Greek, Italic, Keltic and Germanic, this word was indeclinable. Por example: Avest. xsvas satdis 'with six hundred', Gr. si ij^-dQa^, Lat. sex mensibus, Goth, afar dagans saihs 'after six days'. But it was sometimes inflected, 16 Cardinals, Abstract Numerals, and Ordinals. § 170. as 8kr. sadbhis, Armen. gen. vec-i-c, Gtr. s^aai or s^dai (in an inscr. of the fourth century a. d., C.I.G. no. 5128. 27 roTg s^ocm ^aaiXsiay.oig) like rsrpaai, Mid.Ir. gen. se m-bo 'sex vaccarum', O.H.Gr. dat. sehsin (only used when the subst. precedes). The Latin word, sex, was never declined ; the Lithuanian, szeszl, always. Skr. sds (Sdt, see I § 401 Rem. 2 p. 29,7) ; cp. sodaia *16' for *saMasa, like vodhum for *vaMhu-m (I § 404. 2 pp. 298 f.), and saSthd-s 'sixth'. Avest. xsvas, also xstva-, which latter is regarded as standing for *xvsta-. Apparently it should be assumed that there were two forms in proethnic Aryan, *suas and *sas, which became *suas and sas by assimilation of the sibilant These would become quite regularly Avest. xsvas (see Bartho- lomae, Ar. Porsch. Ill 20) and Skr. sas respectively. Armen. vec doubtless represents *ue]cs ; see I § 560 Rem. p. 417, § 589. 3 p. 446. Gr. /s'J f| for *S'iie]cs. For sy. nodwv, sy Saxxvhov, tK-nXsd-Qo-g £y.-/xt]vo-Q etc. see the Author's Gr. Gr.^ p. 71. s^a- in s^a-xoOtoi and other compounds follows the type of rerpa- snra- etc. Lat. sex. O.Ir. se (cp. mor-feser 'magnus seviratus') Mod.Cymr. chwech for *sueh, see I § 175 p. 154, § 517 p. 377, § 576 p. 432, § 657. 10 p. 510. The second s has been preserved in the Irish ses-ca 'sixty' and sess-ed 'sixth'. Goth, saihs O.H.G. sehs, ground-form *seh. Cp. p. 18 footnote 1. Lith. szesz-l (cp. ketur-l § 168 p. 12) doubtless represents *seszl, as szeszura-s represents *seszura-s (I § 587.2 p. 442). Pruss. wuscht-s uscht-s 'sixth'; probably we have a borrowed word in Lith. iisses beside szeszios pi. 'childbed'. Slavonic has the abstract, sesit: cp. Skr. sasti-s ('group of six tens, sixty') O.Icel. sett. sesti brings us to *c.hesti at the first step back- wards, and is doubtless one of the instances of ch- = s- (see I § 588 Rem. 3 p. 444); this change has not yet been satis- factorily explained. Sixth. The parent language may have had the word *sue^to-s {*se-, *ue-) : cp. Skr. sasthd-s Avest. xstva- (see above), §§170,171. Cardinals, Abstract Numerals, and Ordinals. 17 Gr. hro-g, O.H.G. sehto O.Icel. sette setti, Lith. szeszta-s Pruss. wuscht-s 0.C.81. sestu. And the -s- of Lat. sextu-s Umbr. sestentasiaru 'sextantariarum' Osc. ^sarsg 'Sextius', Goth. saihsta O.H.G. sehsto (beside sehto) may have come from the cardinal. But it is uncertain whether or not pr. Idg. *s'!ieMo-s grew out of *suelcs-to-s by a purely phonetic change. Who can tell whether the -s of *SMe& was not an inflexional suffix? If so, it would not at first be found in the ordinal any more than (say) the -e of *pet9qe 'five' in *peiaqto-s. Cp. I § 589 Rem. 2 p. 446. Armen. vec-er-ord. O.Ir. sessed Mod.Cymr. chweched. As to the supposed origin of this re-formation see Zimmer, Kuhn's Ztschr. XXX 214. § 171. Seven. Fv.Idg. *sept'^; the accentuation is inferred from Skr. saptd, Gr. sma and Alban. sta-ts (G. Meyer, Alban. Stud, n 65). Another form of the same stem, *septom-, is perhaps to be inferred from Gr. s^So/uijxovra s^Sofidg h^do/naxig O.Ir. secht- moga (cp. *-d^om-t- beside *-dlc')yi-t- § 164 pp. 2 f.), see § 178. The word was originally indeclinable : examples are Ved. saptd sindhusu 'in septem fluminibus' Avest. hapta satais 'with seven hundred', Goth, sibun hldihans ; and Gr. snrd^ Lat. septem, O.Ir. secht n- were never declined. Inflected forms are: Skr. dat. abl. saptd-bhyas, Armen. gen. evfan-c, O.H.G. sibin-in (only used where the subst. precedes). In Lith. only the word is always inflected, septynl. Skr. saptd, later sdpta, Avest. hapta; we may conjecture that the Skr. word was originally "saptdm, but followed the lead of ndva and ddka; see I § 226 p. 193, § 230 p. 196. Armen. evfn, see I § 380 p. 265, § 560 p. 416. The final -n must be explained in the same way as that of tasn (§ 174). Gr. sma, which we may conjecture should have been *snrdv, but took its present shape under the influence of ewea and dsy.a ; see I § 226 p. 193, § 235 p. 198. Lat. septem. In composition we find beside septem- the re-formation septu- septi-, like octu- octi-. O.Ir. secht n- Mid.Cymr. seith; see I § 339 p. 269. Brugmann, Elements. HI. 2 18 Cardinals, Abstract Numerals, and Ordinals. § 171. Goth. O.H.Gr. sibun. For the retention of -n (on the analogy of the ordinal Goth. *sibunda O.H.G. sibunto , unless indeed it came from an older form *sibun-i), see the Author, Morph. Tint. Y 55 '). As to the loss of -t- — we assume *septi^ to be the Idg. ground-form — we must certainly not ignore Ascoli's theory that the parent language possessed two forms, one with -t- (*sept'>p)^ and one without (cp. Skr. asiti-s 'eighty' beside Idg. *oJctdu) which was kept in Germanic (see Ascoli's Krit. Stud. 101). But it is more natural to assume that there were two forms in proethnic Germanic, *septm6- 'seventh* which became *sepm6- and then *sebm6-, and *septm, which became *seftum (this seems to be the form represented in the Salic Law by septun = seftun) and was then assimilated to *sebm6- and became *sebum; cp. Pruss. sepma-s beside septma-s 'seventh' and pr. Balto-Slav. *osmo- 'eighth' for *ostmu- (§ 172). Sievers (Paul-Braune's Beitr. V 119) and Osthoff (Morph. Unt. II 51 f.) think that the i^-form *septfri could become *sepm in pr. Germ, by a direct phonetic change ; cp. also Noreen , Ur- ^ermansk judl. p. 108. Lith. septyn-l, like devyn-i 'nine' in its ending, and similar to aszt&n-i 'eight'. It may be conjectured that these three forms once were *septin(-i) *devin(-l) — cp. the ordinals septin-ta-s ■devin-ta-s — and *asztu(n4), and that their present shape is due to mutual assimilation. The long u caused the lengthening of i to y; similarly the long vowel of trylika etc. caused the lengthening of the antepenult in venulika (p. 28), and that of Idg. Hrl- caused the lengthening in *qetuf- *pei9qe- (§ 178); many other examples might be found. *septin-l septin-ta-s in- stead of *septim4 *septim-ta-s owe their n to *devin-i *devin-ta-s. O.C.Sl. sedrm, an abstract noun, beside sedmu 'seventh', was shaped on the analogy of sestt : sestu etc. (II § 97 p. 290). The 1) If it is assumed that there were proethnic forms, *sibun-i *niun-i *tehun-i, ad-formates of *fimfi = Idg. *penqe, it follows that O.H.G. sehs, Tfrhich should have been *sihs, must be regarded as modelled upon the analogy of sehsto sehto. For on this assumption there must have been a pr. Germ. *se;(S-i, vfhich would then have become ^si^si. §§ 171,172. Cardinals, Abstract Numerals, and Ordinals. 19 pr. Idg. abstract would doubtless be *sept')p,-ti-s : Skr. saptati-^ ('seyenty'), O.Icel. sjaund. Seventh. Idg. *septm6- (perhaps *sepdm6- *sebdmd- may be inferred from Gr. s^d'o/no-g O.G.8\.sedtnu; see I § 469 p. 345) and *sept't^m6-. Possibly ^septip-to- may also be regarded as proethnic. Skr. saptamd-s. Also saptdiha-s Avast, haptapa-. Armen. evfn-er-ord. Gr. s^do/Lio-g Epidaur. s^SsjuaTo-g, cp. s^So/xt^Kovvot Heracl. Delph. e^SsfiriKovza ; Hom. s^So/Li-aro-g like npwTO-g (*npwf-aTO-g} and XQiT-uTo-g, see § 167 p. 9. The history of s^do^o- is obscure. There seem to have once been two parallel forms, *ejSJ^o- = O.C.Sl. sedmo- and *67ita/xo- = Skr. saptamd- ; more we cannot say with certaiaty. Cp. § 178 for s^SojuijxovTa, and the Author, Morph. Unt. V 36 ff. Lat. septimu-s. O.Ir. sechtmad Mid.Cymr. seithuet for *sept'^m-eto-s, see n § 72 p. 168. O.H.G. sibunto. Pruss. septma-s sepma-s. Lith. sekma-s (sbhma-sj , see I § 345 p. 271, § 877 p. 286; the ordinary word now is septinta-s (Lett. septUdis) instead of *septim-ta-s through assimilation to devin-ta-s. O.C.Sl. sedmu. § 172. Eight. Idg. *olct6 *oktoy,. -t- must have been something of the nature of a suffix, as ak-ti-s 'eighty' seems to shew; this word is unintelligible if regarded as a derivative from *olct6u (astd,4J. Remark. It can hardly be a mere coincidence that the ending agrees with that of the nom. ace. du. masc. of o-stems (§ 285). *oM too, which we see in atl-tl-s, may have been a dual, like *oql 'the two eyes' (§ 295). Perhaps the meaning may have been 'two sets of four' (op. Mid.Cymr. deu-natc 'eighteen', properly 'two nines', etc.) It must be ad- mitted that in that case the numeral 'two' might have been expected before *oktoii., as in Lat. n-ginfi 'two tens', du-centi, and so forth. Still this might have been dropped in course of time. Uninflected in pr. Idg. : e. g. Avest. aSta satais 'with eight hundred"; and Gr. oxtw, Lat. octo^ O.Ir. ocht n- are always 2* 20 Cardinals, Abstract Numerals, and Ordinals. § 172. indeclinable. Inflected forms are : Skr. instr. asta-bhis, Armen. uf-i-c, O.H.G. dat. ahtow-en (only used when the substantive precedes). Inflexion is regular only in Lithuanian, aszt^ni. Skr. astd astM^ Ayest. asta. In Skr. we find also astd, loo. astd-su etc., an ad-formate of saptd. Compounds with asta- (cp. Lat. octi-) had not a little to do with giving currency to this form. Armen. uf, for *uvf, and that for *optd(u), whose p came from the numeral seven (cp. El. onrw); see Bugge, Beitr. zur etym. Erl. der arm. Spr. 43. Gr. oitTca. The numeral seven gave its rough breathing to Heracl. oxtw, its n to El. onruj, and its a to oy.ra- in oxru-xoaiot (Lesb. oxTw-ndaiot) oxTu-novg (beside oxrcS-novg : Skr. asta-pad-). Boeot. oy-To is like 6vo, see §§ 166, 293. Lat. octo. In composition octo- and octi- odu-, cp. Skr. asta- Avest. asta-. Osc. Uhtavis 'Octavius'. O.Ir. ocht n- (see I § 517 p. 377) follows secht n-; for forms without the nasal see Stokes, Bezzenb. Beitr. XI 170. Mod.Cymr. wyth Mod.Bret. eiz for *oMi, older *oktu *okto. Goth, ahtdu; O.H.G. ahto, inflected dative ahtowen. See I § 659.3 p. 512, § 660.3 p. 515, § 661.3 p. 519. Lith. asztu-n-t, cp. § 171 p. 18. O.C.Sl. osim (ordinal osmu) follows sedrm, see § 171 p. 18. The original Idg. abstract numeral is represented by Skr. aii- -ti-s ('eighty'), cp. p. 19. Eighth. Idg. *o1ctdu-6- or some such form. The mo- forms follow the example of the numeral seven, as do Skr. navamd-s Umbr. nuvime (§ 173 p. 22). Skr. astamd-s, Avest. astema-. Armen. uf-er-ord. Gr. iiyiSoo-c for "oySofo-g (in Homer also oySo-aro-g, like s^S6,u-aT0-g).i cp. oySo-rj-/.ovra, oySo-dg. -yS- for -xr- follows the -jScS- of 'seven'. In all other points the history of oySo(f)o-g is obscure; see the Author, Morph. Unt. V 36 if., and below § 311. Lat. octav-o-s, Osc. Uhtavis 'Octavius'; the a is strange nor has it been satisfactorily explained even by the attempts §§ 172,173. Cardinals, Abstract Numerals, and Ordinals. 21 of Thurneysen and Meringer (Kuhn's Ztschr. XXVIII 154, 232). Cp. the vulgar Latin octua-ginta , for *octov-a-, which is like Grr. oydo(f)-rj-. See the Author, loc. cit. O.Ii". ocht-mad^ Mid. Cymr. wyth-uet. Groth. ahtti-da, O.H.G. ahto-do; O.Fris. ahtunda following sigunda niugunda. Lith. aszma-s (now growing obsolete) , Pruss. ace. asma-n, 0.C.81. osmu, pr. Balto-Slav. *os(t)-mo-s. The Lith. has another word asstunta-s, an ad-formate of septinta-s devinta-s. § 173. Nine. Idg. *neu'^ and *4nmj,, the latter in Armenian and Greek. Also *enuen-, which is preserved in Gr. evEv-TJxovra (§ 178). The final was -^ -n, not -in -m, as we see from Gr. ii'sv-tj-xovTa^ Lat. non-CL-gintd non-u-s and Lith. devin-ta-s (contrast deszim-ta-s 'tenth'). In Indo-Germanic , it was not inflected ; e. g. Avest. nava satais with nine hundred'; and in Greek, Latin, and Old Irish it is always indeclinable. Inflected forms: Skr. gen. navdndm, Goth. gen. niun-e, O.H.G. dat. niun-in (only when the sub- stantive precedes). It is always declined in Lithuanian, devyn-l. Skr. ndva^ Avest. nuva. Armen. iwM, pi. inun-U or innim-Jc (cp. Osthoff, Morph. Unt. I 122), see I § 232 p. 197. Gr. *svfni preserved in Ion. flvd-wxsg tlva-xoaioi uva-ro-g Att. sva-Koam sva-ro-g, Hom. fw-ij/nug like iwrj-yovTa (§ 178). Also h'Vfa, which should probably be explained with "Wacker- nagel (Kuhn's Ztschr. XXVIII 132 ff.) as *sv vsfa 'nine in all, a good nine', this original meaning having been subsequently weakened; Heracl. f cjt'k, like ojtrfu, following £7rra'. svtv-ii-Kovva 'ninety' preserves an original "enuen-, op. § 178. Lat. novem instead of *noven follows septem decern, -n is kept in non-a-ginta non-u-s. noun-dinu-m non-dinu-m, usually nUn-dinu-m. Umbr. nuvis 'novies'. O.Ir. noi n-, Mod.Gymr. Corn. naw. But whence came this a? Goth. O.H.G. niun for "niuun, I § 179 p. 156. O.Sax. nigun A.S. ni^on, where 3 is a transition-sound or glide (cp. JelUnek, Paul-Braune's Beitr. XIV 582). The ending -un is 22 Cardinals, Abstract Numerals, and Ordinals. §§ 173,174. to be explained in the same way as that of sibun, for which see § 171 p. 18. Lith. devyn-i O.C.Sl. dev^U instead of *navynl *nov^ti on the analogy of the initial de- of 'ten'; Pruss. nemnts ('ninth') has been influenced by the Grermanic form. For the termination of dmyn-l see § 171 p. 18. devqU is the Idg. abstract *neui^- -ti-s : cp. Skr. navati-s ("ninety'), Avest. navaiti-s ('nine' and 'ninety'), O.Icel. niund. Ninth. Idg. *neui}n-6- or *neuig,-t6- (^enu-^n-d- or *eny,')}-t6-)^ perhaps both. Skr. navamd-s Avest. naoma- = *navema- (as ker^naom = *ker"navemj I § 158 p. 141) O.Pers. navama- instead of *navand-, following (Skr.) saptamd- daiamd-, cp. Umbr. nuvime. Armen. inn-er-ord. Gr. Horn, siva-ro-g, Att. Hom. sva-ro-q for *ivfa-To-g. Lat. non-u-s for *noven-o-. If the dzenoine of the Duenos inscription means 'die noni', its oi makes some difficulty, although not for the reasons which Pauli suggests (Altital. Stud. I 32 ff.). Umbr. nuvime 'nonum', where m is not original, but is like that of Lat. novem and Skr. navamd-s. O.Ir. no-mad, Mid.Cymr. naw-uet, re-formates like ocht- -mad wyth-uet etc. Goth, niun-da O.H.G. niun-to-, pr. Germ. *niuun-dd-n-. Lith. devin-ta-s (Pruss. nemnts, see above), O.C.Sl. dev^-tu. §174. Ten. Idg. *(?^^^. Originally indeclinable, and still so in Yed. ddka hakslycLbhis 'with ten girdles', Gr. J«xa vavai, Lat. decern ndvium, Goth, taihun skattans, O.H.G. stat zehen hurgo 'Decapolis', and similar phrases. Inflected : Skr. instr. dakd- -bhis, Armen. instr. tasam-bk tasam-b, Gr. gen. dsxwv in a Chian inscription (a trace of Lesbian influence), dat. Goth, taihun-im O.H.G. zehin-in (in O.H.G. only found where the substantive pre- cedes). *rfe% has died out not only in Slavonic, but in Baltic too. Skr. ddka, Avest. dasa. Armen. tasn. If the ace. mard 'hominem' is a regular developement from *mxto-m, in which case original final -m was § 174, Cardinals, Abstract Numerals, and Ordinals. 23 dropped, tasn like evfn must be an ad-formate of in-n 'nine', cp. I § 202 p. 169, § 651.2 Eem. p. 497. But it is pre- ferable to regard the ending of tasn as coming quite regularly from *de%'r^^ and mard as being a nominative used for the ac- cusative (see § 212). Gr. Jsxa. Arcad. dvo-Ssm (Bullet, de corresp. hellen., IV 1889 p. 281) like Js'xoro-i,' (see p. 24). Lat. decern, -decim in un-decim etc. is due to the accentuation, see I § 65 p. 53. Re-formates are decu-plu-s dec-enni-s dec- -unx etc. beside decem-plex etc. Umbr. desen-duf 'duodecim' tekuries dequrier 'decuriis', Osc. dekmanniiils 'decumanis'. O.Ir. deich n- (indeclinable, since deich and deck are meaningless variations in the mode of writing the same sounds), O.Cymr. dec. Goth, taihun O.H.G. zelian. The final -n must be ex- plained in the same way as that of sibun, see § 171 p. 18. "We should not have expected the -a- which is found in O.H.G. zehan O.Sax. tehan; cp. O.H.G. zehanzo beside Goth. taihunte(-hund) § 179. Possibly in words like drt-zehan, *-tehun became *-teh^ and then -tehan, and the a passed thence into *tehun etc. (cp. O.H.G. Sigi-frid as contrasted with fridu). A different ex- planation is given by Noreen, Arkiv IH 26. In Balto-Slavonic the only forms left are the two Idg. ab- stracts : Lith. deszim-t- 0.C.81. des^-t- and Lith. deszim-ti- 0.C.81. des^-tt-: cp. Skr. daidt- Gr. Sexdg Goth. gen. pi. taihunte (in taihunte-hund '100', see § 179) and Skr. dakati-s ('tenfold character, group of ten', specialised to mean 'group of ten tens, hundred') O.Icel. tiund. In early Lithuanian deszimti- is still an inflected singular substantive and is followed by the genitive; but now the inflexion is gone, and we have deszimt (doubtless both ace. sing. = desqti and loc. sing. = des^te) and deszimts deszimts (doubtless nom. pi. = desete) '), although still 1) The history of the plural form dSszimts needs further inTBstigation. Has it been influenced by dvldeszimts 'twenty' trlsdeszimts 'thirty' etc.? Or is it merely due to an idiom of the language which we find in the old books, whereby the abstract noun is used like an adjective with the 24 Cardinals, Abstract Numerals, and Ordinals. §§ 174,175. gOTerning the genitive plural. 0.C.81. desqti is declined throughout as an ^-stem ; there is a parallel stem desqt-, e. g. in jedinu na desqte (loc. sing.) 'eleven' = 'one upon ten'. Tenth. Idg. *dehn-to- (*deh^-t-o-? see 11 § 81 Rem. 1 p. 242) and *dektpm-o-. Skr. daiamd-s^ Avest. dasema-. Armen. tasn-er-ord. Gr. Ssxavo-g. Lesb. Arcad. 6s>coTo-g (cp. Arcad. Svo-Ssxo), whose follows -y-ovra -xoaro-g, cp. §§ 176, 177. Lat. decimu-s, Osc. dekmanniuis 'decumanis'. O.Ir. dechni-ad, Mid.Cymr. decu-et. Groth. taihunda, O.H.G. zehanto (cp. p. 23). Lith. desdmta-s, 0.C.81. desqtu. § 175. Eleven to Nineteen. When the units were added to multiples of ten in the parent language, both units and tens of the resulting number were independent in the sentence. The copula 'and' may have been generally used with them, as in the phrases Yed. ekq ca viSati ca ace. '21', trdyai ca tr\^ac ca '33', Gr. dvo) y.al 7iivvrjy.ovTu '52', Lat. quattuor et viginti; but not always, as we infer from Ved. trj,idtq trtn ace. '33', Gr. nsvTrjuovra Svo, Lat. vigintl quattuor etc. But in the cardinal numbers 11 to 19 there was a closer combination between the unit and the numeral 'ten' which followed it (see II § 16 pp. 31 f.). In the numbers 11 to 14 the unit was inflected, in 15 to 19 it was not; hence 15 to 19 readily became true com- pounds, whilst 11 to 14 may not have become compounds so soon, since their ending had first to become stereotyped. Bern ark. There can be no doubt as to the reasons for this difference between the expressions for 11 to 19 and those for 21 — 29, 31 — 39 etc. The former group was more often used, for one thing ; but the chief reason was that the words for 20 and the other multiples of ten were themselves compounds, and therefore it was less convenient to compound them again with other words. name of the thing whose number is stated, and takes the case of it ; as loc. deszimtisa wientosu 'in decem urbibus' (cp. Bezzenberger, Beitr. zur Gesoh. der lit. Spr. 178 f.)? § 175. Cardinak, Abstract Numerals, and Ordinals. 25 But the numbers 11 to 19 were not expressed in the parent language only by pairs of words in juxta-position, like Skr. dvd- -daSa Gr. cFw'-Jsxa. We are justified in regarding as original expressions like Gr. dixa ^vo, Ssku rgeTg, Lat. decern duo, decern tres. We may also believe that phrases of subtraction were used for the numbers immediately preceding twenty as well as for those immediately preceding thirty, forty and so forth; such, for example, as we find when the languages had begun to follow their own separate lines — Skr. ekonaviiati-s, unav\§ati-s and ekan nd v\sati-s for 19, tryUnasasti-s for 57, Gr. evoc, Heowa fiY-ooiv hrj '19 years', f.iid<; Stovaai TSTTa^dy-orru vrjsQ 39 ships, Lat. un-de-vlginti duo-de-triginta, A.S. twa Ices twenti^ for 18, an Ices tiventi^ for 19, H.G. dial, ains-min-zwainzich ztiai-min- dreisich (Goth. 2 Cor. 11. 24 fidvor tiguns ainamma vanans to translate rsaaa^dxoi'Ta napa /.dav). Of the different modes of expressing the ordinals which we find, the Latin for 13th to 19th, tertius decmus etc., occurs in Armenian, Greek, and Germanic besides. We may therefore fairly regard this as original. Aryan. Cardinals. 11 Skr. eka-daSa; the first part of which crystallises the form of the instr. sing. masc. (Ved.) and nom. sing. fem. ; the form thus chosen was suggested by dva- -daSa, cp. Avest. a^van-dasa- etc. '11'"' below. 12 Skr. dva- -daSa dmd-daia Avest. dva-dasa. 13 Skr. trdyo-daia. 14 Skr. cdtur-daia, showing now the stem without inflexion, cp. A.\est. capru-dasa- '14:*^'. 15 Bkv. pdnca-daia, Avest. pawca- -dasa. 16 Skr. s6-daia. 17 Skr. saptd-daka. 18 Skr. astd- -dah. 19 Skr. ndva-daia , also eMnav{§ati-s {eka-una-v\iati- a score too little by one, a score less one'), or simply una- -viiati-s, and ekan (i. e. ekad) nd v\hti-s (cp. Delb., Altind. Synt. pp. 112, 543). Ordinals. In Sanskrit all the numbers have both -daid-s and -dasmna-s, cp. Lat. -decimic-s. 11"' Skr. ekadaM-s, Avest. a^van-dasa-, aeva-dasu-, aevo-dasa- ; aevu-dasa- may be like dva-dasa- = Skr. dva-dain-, or is it the bare stem instead of a case, as in pri-dasa- capru-dasa-? cp. II § 25 p. 41. 26 Cardinals, Abstract Numerals, and Ordinals. § 175. 12* Skr. dvadaid-s (dvadaiama-s like duodecimu-s) ^ Avest. dvadasa-. IS* Skr. trayodaid-s, ATest. pridasa- with the bare stem instead of a case. 14"' Skr. caturdaid-s, Avest. caprudasa-, cp. the cardinal. 15* Skr. pancadaid-s, Avest. pancadasa- and panoadasya-, the latter like tUirya- 'fourth' bitya- 'second'. 16* Skr. sodaSd-s, Avest. xsvasdasa-, etc. Side by side with Skr. navadasd-s (Avest. navadasa-) '19*' is found ekonav\§a-s, unaviSa-s and ekannavj^d-s, cp. the ordinal. Armenian. Ordinals. 11 me-tasan. 12 erko-tasan. 13 ereTi-tasan. 14 coreK-tasan. 15 hnge-tasan. 16 ves-tasan. These are all inflected as i-stems; e. g. gen. dat. metasanic, instr. metasanivlt (cp. Ksan '20', gen. dat. Esanic). The numbers from 17 onwards have ev 'and', and inflect sometimes both parts, sometimes only tasn (cp. air-ev-ji II § 28 p. 46). 17 evih-ev- -tasn. 18 uf-ev-tasn. 19 inn-ev-tasn. Ordinals. Two modes are used, tasn-erord ('tenth') may be followed by^ the ordinal of the unit, as tasnerord corrord 'decimus quartus'; or -er-ord may be simply added to the cardinal, as metasan-erord '11*' coreMasan-erord '14*'. Greek. Cardinals. 11 sv-d'sxa (ev- is nom. ace. neut.), Delph. 3sy.a sig. 12 da-dsxa (Hom. (^Dw-Jexa), Hom. 6vo-y.ai-Ssxa, and in Att. and Dor. Sit^a Svo as well. 13 tqeIc, y.ai Ssym and (with the nom. rgsTg crystallised) Tpsig-Kai-d'ey.a^), Att. Dor. J/xct TQsTg as well. 14 TSTvagsg xui dfxa, TSTTagfa-xaL-dtxa and Seku rsTTupsg, and so forth. As to the form of e| in h-xai-Ssxa beside Boeot. sa-xi]-6sxaTos see the Author's Greek Grammar^ § 59 p. 71. In Attic Sixa Svo, Sexa rpftg etc. were used when the substantive preceded ; e. g. Sgayuai Ssv.a rpsTg but rpsTg xai Ssxa SQa/jxai (cp. Wackernagel, Kuhn's Ztschr. XXV pp. 284 f. and Philol. Anzeiger 1886 pp. 78 f.; Meisterhans, Gr. der att. Inschr.^ pp. 126 f.). Ordinals. 11* EvStxazo-g. 12* d(t)StxaTo-g, epic SvaSfxaro-g, 1) If the by-form TQiaxaiSsxa is to be admitted (cp. Meisterhans, Gr. der att. Inschr.^ p. 126), it contains the ace. rfn = *rp«'s (Wheeler, Der gr. Nominalaooent 42) in a crystallised shape, or rqlg, the form it assumed in proethnic Greek before consonants (I § 204 p. 171). § 175. Cardina'ls, Abstract Numerals, and Ordinals. 27 Ion. SvoiSey-aro-s. From IS* onwards the usual mode of ex- pression in classical Attic, followed consistently in the in- scriptions, is XQiTog Kcci Sinarog, riruprog xal SexuTog etc. Homer has TptidxaiSsxaTo-g oxrwuttidfyMTo-g, and Herodotus TsanaQfayiatdsxaro-g nsVTSxadfxuTo-g , while we find say.tj^fyMTO-g in Boeotian. Italic. Cardinals. The following is the ordinary Latin inode of expressing these. 11 un-decim, the first part of which we may conjecture to have come from more than one case-form (cp. un-de-mginti), cp. I § 633 p. 474, II § 86 p. 62. 12 duo-decim. 13 tredecim for Hrez-dedm (I § 594 p. 450); also tredecim, perhaps following trecenti. 14 quattuor-decim. 15 quindecim for *quinqu(e)-decim, see I § 683 p. 474. 16 se- decim for *segz-decim, see I § 594 p. 450. 17 septen-decim, see I § 207 p. 174. 18 duo-de-viginti. 19 Un-de-vlgintt , cp. un- -decim. Other expressions are interchanged with these, such as tres et decern, odd et decern; decern duo, decern novem; decern et unus, decern et duo. Umhr. desen-duf ace. 'decem duo', see I § 207 p. 174. Lat. Ordinals. 11* Undecimu-s. 12* duodecimu-s. 13* tertius decimus, 14* quartus decimus etc., rarely decimus tertius etc. 18* duodevicesimu-s, rarely octavos decimus. 19*- undevt- cesimus, rarely nonus decimus. Old Irish. In the cardinals we have the form deac, dissyl- labic in the older language, in the later contracted into dec (Mod.Ir. dSag). 11 oen — deac, 12 dd — deac, 13 tri — deac etc., with the noun in between, as da cath deac '12 battles'. deac dSc has nothing to do with deich n- ; it was possibly a word meaning much the same as Skr. adhika- Goth. -Uf Lith. -lika. Ordinals. 11* oenmad — • deac, 12* aile — deac etc. Grermanic. Cardinals. 11 and 12 contain -lihi-. This is a noun stem connected with Goth, bi-leiban 'to remain' (v^ leip-, Skr. limpdmi 'I cleave, stick'), and it originally meant 'excess' or 'being inexcess' — elf would then be 'one in addition', i. e. to ten; cp. below Lith. -lika from v^leiq-, and Skr. adhika- 'being 28 Cardinals, Abstract Numerals, and Ordinals. § 175. in excess' as used where 10 and its multiples are coupled with units, e. g. astddhikanavati-s 'a ninety increased by eight' = '98'. Goth, din-lif O.H.G. ein-Uf, Goth, tva-lif O.H.G. zwe-Uf appear inflected under the same conditions and in the same way as the numbers 4 to 10, e. g. tvalibi-m, swelifin (O.Sax. elleban '11' nom., following tehan '12'). 13 O.H.G. drt-zehan, but also fone dien anderen drin zenin (Graff, Ahd. Spr. Y 628). 14 Goth, fidvor- -taihun, O.H.G. fior-zehan. 15 Goth, fimf-taihun O.H.G. finf- -zehan. 16 O.H.G. sehs-zehan. 18 O.H.G. ahto-zehan. 19 O.H.G. niun-zehan. Ordinals. 1 1* (fem.) O.H.G. einlif-to O.Icel. ellifte ellifti. 12* (fem.) O.H.G. zwelif-to O.Icel. tolfte tolfti. The following ordinals began by being phrases of the same type as Lat. tertius decimus; but their first member crystallised, it would seem, in proethnic Germanic, and they then conformed to the rules of stem-compounds. Goth. Luke 3. 1 in jera fimfta- taihundin 'sv srst nsvrsaKiSsy.ario . O.H.G. dritto-zehanto, fiordo- -zehanto etc., and also with -a- (later -e-) as the final of the first member. Another series, derived from the cardinal, was used in later O.H.G., as fierzen-do sehszen-do. Icelandic has a corresponding series, Jim(tn)tan-de sextan-de etc. Balto-Slavonic. Lithuanian. 11 venil-Uka^ 12 dvy- -Uka, 13 try-lika, 14 keturio-lika, 15 penkio-lika, IQ szeszio-Uka^ 17 septynio-Uka, 18 asztunio-Uka, 19 devt/nio-lika; 11* O.Lith. leka-s, 12* O.Lith. antras lekas, but the words now used have -likta-s, as 11* venulikta-s, 12* dvylikta-s. try-Uka, keturio- -lika etc. contain forms of the neut. pi. in both parts (§ 338), and accordingly O.Lith. has the dat. -likams and instr. -likais. When the neuter dropped out of use in Lithuanian (§ 408), -lika was treated as a nom. sing. fem. ; and then it was declined gen. -likos etc. This inflexion is seen in Old Lithuanian, and is still found in dialects of the language, -lika came from an adjective *lika-s 'remaining over, being in excess', a by-form of the O.Lith. leka-s just mentioned; and to this day leka-s is in regular use in the sense of 'remaining over singly, odd'. The root is leiq- (Lat. linquo Gr. Xh'ttw). Cp. Goth, din-lif above. In § 175. Cardinals, Abstract Numerals, and Ordinals. 29 ven4'-Uka and dvy-lika the final of the first member has assimilated itself to the numerals immediately following, and become long ; cp. Skr. eka-daia p. 25 and § 326. Cp. Bezzenberger, Beitr. zur Gesch. der lit. Spr. 179 ff.; Kurschat, Gram. p. 269; the Author, Lit. Volkslieder p. 309, and in Techmer's Internat. Ztschr. I 251 f.; Mahlow, Die langen Vocale 49 ; Joh. Schmidt, Pluralbild. 39, 42. — Slavonic. 11 Jedinu na desqte (loc. sing, of stem des(^t-) = 'one upon ten', 12 dtlva na des^te 'two upon ten' etc. Cp. Lett. mn-pa-dsmit '11' ^ 'one over ten' diw-pa-dsmit '12' etc., and Gr. Thess. ra sy.ra In IkolSi (Collitz, Samml. der Gr. Dialekt- inschr., no. 345. 10). There are two types of ordinals. Some- times only the unit takes the ordinal form, as osmyp na desqte 'the eighth upon ten' = 18* ; sometimes -tnu was used to make a derivative from the expression for the cardinal number. In the latter case the unit might either show the form of the nom. ace, as pqti-na-desqtmu 'IS""'; or be treated like the first member of a stem-compound, as pqto-na-des^tmu, cp. II § 47 p. 86. § 176. Twenty to Ninety-nine. The Indo-Germanic expressions for multiples of ten from 20 to 90 at first meant two, three, or the proper number of tens. Originally both parts of the phrase were inflected ; both the unit (except the uninflected units 5 to 9, see §§ 169 ff.) and the word for a ten — a neuter ^Homt- *Rtp,t- (for *d1comt- *dJcmt-^ see § 164 pp. 2 f.). '20' was a dual, %? (?) Tcmt-i; the others plural,^ as '30' *tri icomt-9. But these expressions for multiples of ten are not inflected in any language ; in all of them the nom. ace. has become stereotyped. Some at least of the units in these phrases were stereotyped in the parent language itself. This is proved by *qetuf-Jcomt9 (Gr. Tirpw-/.ovTa Lat. quadra-ginta Avest. capwar'-sat- Armen. fcar-a-sun) and *peidqe-hmt9 (Gr. nsvvrj-KovTa Skr. panca-idt-), which were ad-formates of *tn- -komtd. It is doubtful whether ^h^tti and *komt9 had also become crystallised so soon. These forms of the nom. ace. pi. (du.) neut. became in the Indo-Germanic period the foundation upon which were built singular abstract nouns (collectives) of the feminine gender, and 30 Cardinals, Abstract Numerals, and Ordinals. § 176. ordinals whose first member was the nom.-acc. form. The ab- stract nouns were ad-formates of the feminine *deh^t- 'group of ten' 1) : examples are Skr. tr\^dt- 'group of 30', Gr. Boeot. J^img Att. fixa'e, O.Ir. fiche 'group of 20' gen.Jichet. The ordinals were derived from these by the suffixes -to- and -t-rnmo-. When these two groups of abstract nouns and ordinals sprang up, the ex- pressions for the cardinal numerals, from which they were derived, had not yet fully become compounds. Thus HriHomt- *trih^t- \piayMs' and *trihrifto- *trlk,infti^mo- 'rptamOtog' may have stood related to *trT %omt9 'three tens' much in the same way as O.C.Sl. duvadesqtmu "20"'' to duva des^ti '20', and as Lat. quartadecumam to quarta decuma, Sacravienses to sacra via etc. (II § 3 p. 5). Remark 1. Perhaps the re-formates *qel'yf- and *penqe- first ap- peared only in collectives and ordinals, in which there was a closer con- nexion between the word for the unit and the word for the ten. This might explain certain pairs of forms, Tsrqw-xorTa and T^Taqa-xovTa in Oreek, panca-sat- and panca-sat- in Avestic; and the difference between Avest. capwar^-sat- and Skr. caivarj-sdt-. Gr. TirqiaxovTa would then be an ad-formate of TeT^uxoard-?; while reTTagaxoarot, on the contrary, would liave followed rsTtagaxovra. The dual *'k,trt,t-i once had the weak stem in all its cases; hence come Avest. vt-saiti Armen. U-san Gr. fi-xari Lat. vt- -gintl, hence also the collective with -Tomt- : Gr. fi-ndg sl-ndc O.Ir. fi-che (O.Cymr. u-ceint Corn, u-gans). But the nom. ace. pi. was *Jcomt-9, whence Armen. -sun Gr. -xowa and the col- lectives Avest. pri-sqs O.Ir. -cha -ga. In cardinals and col- lectives of the tens from 80 upwards "^hnt- is also found (Lat. -ginta and Skr. tr\-§dt- Avest. pri-sat- etc., Gr. rgia-xdg). Two possible causes may be assigned for this. (1) Beside *Uomt-d there may have once been weak cases with *lcvp,t-^ as loc. *h^t-su^ or (2) the corresponding forms for the number 20 may have set the type. The ordinals had all of them doubtless *hrit- to begin with, as Gr. Boeot. fi-xadro-g Skr. tr\-iattamd-s Lat. trl-cesimu-s. 1) Words were formed later on the same principle in Old Icelandic, in Lithuanian , and in Greek. Examples : O.Icel. tmteg-t 'elxdg priteg-t ^Tgiaxdf' etc.; Lith. dvideszimti-s 'eixag (e.g. yo dvideszimtes metu, in Bretken) from *dvl dSszimti '20'; Gr. hSsxtii SioSfxdg etc. § 176. Cardinals, Abstract Kumerals, and Ordinals. 31 Remark 2. Avest. visqstema- is an ad-formate of *prisqstema- (which we may infer from- ^nsq^s), which had itself taken the place oi *prisastema-. In Greek, and doubtless in its proethnio period, -xovra influenced the con- nected ordinals in "-xaaro-;. The change may hare taken place in either of two different ways; (1) *-xaaTo-g may have become "-xovaro-s and then -xoaro-i (ep. xeffTo'-s for *xffaTo-g, I § 204 p. 171), or (2) *-xaaTo-i may simply have taken over the o of -xovra. The o then spread backwards to 20 and 10 (Ion. Att. elxooTo-c elxoai, Arcad. Sixoro-g Svo-Sfxo), and forwards to 100 and its multiples (Arcad. Ixozor-jSota and Ion. Att. -xooiot). The old expressions for the cardinal numbers, consisting of an adjective with a substantive, remained in Armenian, Greek, and Latin, and in the Avestic word visaiti '20'. In Aryan and in Keltic these were displaced by the group of singular abstract nouns; the only Aryan forms which recal the old type are Avest. vtsaiti, and indirectly Skr. viiati-s (see § 177). But in Aryan these forms were themselves displaced in the numbers from 60 to 90 by a second group of abstracts, such as Skr. sasti-s (see § 178). In Germanic and Balto-Slavonic, both these expressions for the cardinal numerals and the singular abstracts had disappeared before the historical period begins. Their place was filled by other expressions which had really and truly the same etymological factors, and the same meaning, as the original Indo-Germanic expressions. Take for example 30, Goth, preis tigjus Lith. trys deszimtys O.C.Sl. tri des^ti, where the substantive was the Indo-Germanic word for a group of ten, *delc'>nt- *delcmU-, still used independently. i) It is probable that *hmt- *'kmt- became obscured quite early in Germanic and Balto-Slavonic, as in the other languages, and sank to the level of a suffix; and the new expressions served to refresh somewhat the original meaning of the words. But then the same thing happened again, and the new words 1) Grerm. He-gu- 'group of ten' must be derived from the following manner. In the instr. pi. *te^und-mi and in an instr. dual form containing an m-suffix (in the expression for 20), -undm- became -unm-, -umm-, and -urn- successively. Thus we have Hgum, which gave the type for a new set of cases, Goth, tigjus etc. See §§ 379, 386. What may be the relation of forms with u in the root-syllable (O.H.G. -ziig O.Icel. togr tugr) to *<«gM- still remains an open question. — For the masc. gender of the word compare O.C.Sl. duva desqti. 32 Cardinals, Abstract Numerals, and Ordinals. § 176. were themselves obscured and became suffixes. For example, in Mid.H.G. drt-^ic vier-zic Mod.H.G. drei-ssig vier-zig the final part was and is a mere suffix, no less than was that of Grr. TQitt-y.ovra or that of Lat. tri-gintcL at the beginning of the historical period of the classical languages. And in German [and English] these multiples of ten are used as adjectives agreeing in case with the substantive which follows, just as happened with the similar expressions which the Romans and the Greeks had in- herited from the parent language : Mod.H.G. in vierzig wochen 'in forty weeks' as contrasted with O.H.G. feorzug wehhono 'rsxruQa- xovrdg sji^ofid^wv and with Goth, dage fidvor tiguns 'ijfisgav tst- TuQug dsxciSag, just like Gr. rpiaxovta av^gsg instead of *rQia xovra uvSgwv, Lat. trigintcL virt instead of *tri conta virom. However, in Germanic these new expressions with *iejM- held their ground only from 20 to 60, while the three others of the series — 70, 80, 90 — were displaced in proethnic Germanic by a new group formed on the analogy of an old expression for 100, Goth. taihunte-hund '^fxd^wv Ssxdg. This change will be discussed in § 178. In the parent language there never was any very close connexion between the words for the various multiples of ten and any intermediate units which might be used with them (in numbers such as 21, 22, 31 and so forth). The unit al- ways remained an independent word. See § 175 p. 24. It was also independent in the differentiated idioms of the different languages. Sanskrit is the only noteworthy exception. Along with the old method of expresssion, Vedic itself contains femi- nine words like trdyas-tr\kat- '33' cdtus-triiat- '34', which follow the analogy of trdyo-daia '13' cdtur-daia '14'. Later, these compound forms became the rule; and for other numbers be- sides 24, 34 etc. the bare stem was used in them; e. g. eka- -v\kati-s '21' (but on the other hand ekadaia), dvi-tr\^at- '32'. Sanskrit always shows a marked preference for compound words (see II § 21 p. 37), and this new group only followed the general lines of the language. Words formed on the principle of subtraction have been §§ 176,177. Cardinals, Abstract Numerals, and Ordinals. 33 already discussed (§ 175 p. 25). Examples are ekonatr\ht- Unatrj^at- '29', panconq iatam '95', ekan nd iatdm '99'. § 177. Twenty. The cardinal ended in *-h^U (Avest. vi-saiti Grr. fi-xan)^ which was nom. ace. du. neut. of the stem -Rint- (§ 294). The first part was *uei- (Gr. Heracl. fii-xan ') si-xoai), *ui- (O.Ir. fi-che), *iiim- or *«^m- (Skr. vi-Sati-s)^ perhaps also *y,i- (Ayest. vt-saiti, Gr. fi-y.an with t?, Lat. vt-ginti, Armen. Jlsan for *gi-santi or *gT-santi). It would seem, then, that different case-forms were used; but we cannot get anything like a clear idea as to what the original method of expression was in Indo-Germanic. It seems certain that all these variations of *M«i- meant 'two', and it is natural to connect them with two particles — (1) Skr. vi 'apart' vi-su- vi-sva- 'on both sides, on different sides' (cp. Avest. pri-sva-) vi-tard-m 'further' Goth. vi-pra 'against, with- (in composition)' Lat. vi-tr-icu-s (II § 75 p. 191); and (2) u in Skr. u-hhau O.Q.Si.vu-toru and in the nom. ace. du. Skr. dva-il. Then *u-4- 'two' will be like *tr-i- 'three' and *dui- 'two' (§ 166 p. 6, § 311 Rem. 2). See the Author, Morph. Unt. V 23 ff., Bartholomae Stud, zur idg. Sprach- gesch. I 74, and below §§ 285, 296. The abstract ended in *-Tcmt-s (in the nom. sing.) : Gr. fi-xds l-xdg st-ncig O.Ir. fi-che. The ordinal ended in *-hnfto- ^-Umftmmo- : Gr. Boeot. ft-y.aato-c Lat. vi-cesimu-s. Aryan. Avest. visaiti. Skr. vUati-s is a singular ab- stract noun formed from the nom. ace. du. in *-iati, after the analogy of sasti-s '60', saptati-s '70* etc.: people said v\§atyd, hdrmanij and with the case of the latter word assimilated v{katyd hdribhis 'with 20 bay steeds' just as they said sastyd hdrmdm and sastyd hdribhis 'with 60 bay steeds'. The later v\kat- seems to be merely an ad-formate of the numbers 30 to 50 tr\idt- etc., 1) Danielsson (Epigraphioa, Upsala 1890, p. 33) would now regard Heracl. feixari as flxan influenced by the form of Att. fi'xooi, which he takes to stand for efixoai. The diphthong of /«- has no real support whatever in the other Indo-Qermanic languages; still I can see no valid reason for denying that it represents an original proethnio form. BrugmanD, Elements. III. 3 34 Cardinals, Abstract Numerals, and Ordinals. § 177. as on the other hand viiati-s was the type for tr\kati-s which was used in more recent times along with tr{-kdt-. Avest. vi- may come from v\--^ see the Author, Morph. Unt. V 27. 20"^. Ayest. visobstema- instead of *vi,sastema-, see § 176 Eem. 2. Skr. v\kati-tamd-s, derired from vpati- (cp. sasti-tamd-s § 178), and v\^d-s like ekadaM-s 'eleventh' etc. Armenian. Ksan probably for *gsan *gtsan(ti) with g- = M-, -s- = -^-, -an- = -tn-, see I § 232 p. 197. Icsan, like the multiples of ten that followed it, received inflexion once more (as an «-stem), e. g. gen. dat. Ksan-i-c; and later it was also declined in the singular. Ordinal Ksan-erord. Greek. Dor. Boeot. Flxmi, and with si Dor. funari Ion. Att. fly.oai Horn, sf/xorrt. 20"^ Boeot. /(xaerro-g kit. dy.oaroc,. Abstract: Boeot. fv/Mg Thess. i/a? Att. stxag. The quantity of i in the first syllable has not been ascertained, -o- in place of -a- was due to the following multiples of ten, its first source being the ending -xoira; see § 176 Rem. 2 p. 31. Hesychius has preserved another form Yxawiv (MS. Ixdvnv) with the -v- of -y.ovta. The v erpslxvanxov may have been first added to fixoai when it was used as a dative, cp. inscr. dvSodaiv svl xal sly.oaiv (Maassen, De littera v paragogica, 1881, p. 34). Italic, vi-ginti. vicesimu-s^ rarely vigesimu-s. Whether vt- represents Idg. *ui- or *uei- is uncertain; we find veiginti in C.I.L. I 1194, later than 105 B.C. It is also doubtful why the final -i of -ginti is long; was an original -i lengthened on the analogy of -ci, in triginta etc., or was -oi or -ej, the ending of the nom. ace. du. neut. of o-stems (see § 293), sub- stituted for it? The -g- Thurneysen holds to be correct pho- netically in septingentl nongenU (quadringentT octingenti) , and then to have extended itself by analogy into other numbers (I § 499 p. 366); in considering this question, we must not forget that a media g seems also to be indicated by the z- of Alban. -zet 'group of 20' (ns-z4i 'one score', dil-zet 'two score' etc.) — see G. Meyer, Abh. zu M. Hertz' 70. Geburtstag 1888, pp. 90 f., and compare the mediae in Lat. quadru- § 168 p. 11, Gr. £/?Vo-e O.C.Sl. sedmu § 171 p. 19. §§177,178. Cardinals, Abstract Numerals, and Ordinals. 35 Old Irish, fi-che (gen. fichet dat. fichit) for *-Jcip,t-s (I § 243 p. 201, § 620 p. 467, § 634 pp. 474 f., § 657. 6 p. 509, § 685 p. 552) ; possibly fi- took the place of *ut- or *uei- after the analogy of tricha. O.Cymr. u-ceint Corn, u-gans, the u of which has not been explained; cp. Thurneysen in Kuhn's Ztschr. XXYI 810 footnote 2. 20* Mid.Ir. fichet-mad (in- ferred from coicet-mad). Germanic. Goth, tvditigjus, dat. tvdimtigum. O.H.G. zwein-zug O.Sax. twen-tig, the first part being a crystallised dative. 20* O.H.G. zweinzug-osto. As to He^u- for *deJcmt- see p. 81 footnote 1. Balto-Slavonic. Lith. originally *dvi deszimti (dual), hence dvideszimt uninflected, and also a dialectic form dMeszimts through assimilation to trlsdeszimts etc.; 20* O.Lith. antra-s deszimta-s^ modern dvideszimta-s. O.C.Sl. duva des^ti (dual, so desqti is masc); 20"" duvades^tmu dvadesqtmu, and, following the fashion of stem-compounds, dvodes^tmu (ct^. p^to-na-des^tmu § 175 p. 29). § 178. Multiples of Ten, from Thirty to Ninety. Aryan. Traces of the old neuter plural phrases *trt Homtd etc. survive in the first components of Avest. capwar"-sat- and Skr. panca,-idt- Avest. panca-sat-, since these followed the analogy of *M- (see § 176 p. 29), and also in the first part of Skr. catvari-idt-, in which *catvari- changed to catvarjr- on the analogy of v{- and tr\-. In proethnic Aryan the feminine singular abstract nouns displaced the old plural phrases, and in the same period these same forms, in the numbers from 60 to 90, were themselves displaced by abstract nouns derived from the units, Skr. sasti-s Avest. xsvasti-s 'a group of six' {i. e. six tens), and so forth. In the second member of 80, 40, and 50 the weak stem *.^rp.t- == Skr. -Sat- Avest. -sat- has levelled out the others; Skr. tri-Mt- catvar{-kdt- pafica-idt- , Avest. pri-sat- capwar"- -sat- panca-sat- ; *-h)mt- is found only in Avest. pri-sqs, an indeclinable word, which corresponds to O.Ir. nom. sing, tri-cha (cp. Ascoli, Krit. Stud. 100). In Avestic the ace. in -sat-em was 3* 36 Cardinals, Abstract Numerals, and Ordinals. § 178. used for the nom. as well, doubtless under the influence of the neuter sate-m '100'; hence the word was declined as an o-stem, gen. pi. prisatanqm , panccLsata-gCiya- 'space of 50 paces'. In Sanskrit the analogy of vjiati-i gave rise to triiati-s, which was used along with tr\kdt-. "With regard to the first member, the following are directly descended from Indo - Germanic forms: Avest. capwar'- from *qetuf- (cp. I § 306 p. 242) and Skr. paflca- Avest. panca- from ^peKqe-, see § 176 p. 29. "Whether Skr. trj- represents pre- Aryan *trim-, or is an ad-formate of vjr-,- is a doubtful point; Skr. catvarj- instead of *catvari- must count as an ad-formate of this kind. Avest. capwar^-sat- was confused with compounds like Mar" -car ana-; hence alongside of capru-mCihya- adj. 'every four months, connected with four months', and the like, were coined such compounds as capwar'-zangra- 'four-footed'. This same capwar"-sat- may therefore have suggested pri-sat- in- stead of *pri-sat- or *prpat-, and panca-sat- (beside panccl- -sat-). But the pri- of the MSS. may be an incorrect mode of writing pr%- or pr\- (see the Author, Morph. Unt. V 27), and panca-sat- may be a genuine product of the time when *pet9qe Ttomtd was still spoken and had not yet been changed to *pe'>dqe Jcomta (see § 176 Rem. 1). 60 Skr. sasti-s Avest. xsvasti-s. 70 Skr. saptaii-s Avest. haptaiti-s — the Cl of the latter is due to that of astoLiti- ; a is retained in haptaipi-vant- 'seventy-fold'. 80 Skr. aiiti-s (cp. § 172 Rem. p. 19), Avest. astaiti-s. 90 Skr. navati-s Avest. navaiti-s. Skr. sasti-s and Avest. navaiti-s have not ceased to bear the more general sense of 'group of six', 'group of nine'. Remark. We may assign a reason for the use of sasti-s etc. in place of corresponding abstracts of the same kind as tr{-^dt-. Such a use suggests that in proethnio Aryan higher numbers could be expressed by a sexagesimal notation, in which the word scisti-s xsvasti-s 'threescore' held the most prominent place. Even in historical times the ancient Persians had a remarkable liking for the number sixty and its multiples, as the Romans had for sexaginta and sescentl (see Cantor, Mathemat. Beitr. zum Kulturleben der Volker, 1883, p. 361 f.) The original compound numeral for 60 was displaced by sasti-s in proethnic Aryan, and corresponding ex- pressions for the following tens established themselves later by analogy. § 178. Cardinals, Abstract Numerals, and Ordinals. " ' In the phrases Skr. tr\kdtci hdrlnclm, sastyd hdrlnciin 'with 30, 60 bay steeds' the cases were assimilated , giving tr\katcL hdribhis, sastyd hdribhij; and a further assimilation of the first word to the number of the second produced Skr. panccliadbhir variais 'with 50 arrows', the numeral being now regarded as an adjective. The Avestic prisatanqm bawrinqm 'triginta fibrorum' is similar (cp. Lith. deszimtisa mestosii p. 23 footnote 1, Lat. ducentl virT § 180). It is improbable that the latter construction is immediately connected with the original neut. pi. phrases *tn %omtd 'three tens' etc. Ordinals. Skr. trj^at-famd-s catvcLr{sat-tamd-s pancClsat- -tamd-s and tr\kd-s catvcir\sd-s pancaid-s like viSd-s (§ 177 p. 88). sasti-tamd-s saptati-tamd-s akti-tamd-s navati-tamd-s and sastd-s saptatd-s aSitd-s navatd-s (observe that t distin- guishes these from sasthd-s 'sixth' saptdtha-s 'seventh', which have th)\, the last three are to be compared with -Satd- '100*' (§ 179). In Avestic only prisata- '30"'' is actually found. Armenian, -sun came from the stem -Jcomt- (I § 79 p. 70), and is probably shortened for *-sonta. Sometimes we have -a-, the 'vowel of composition', (cp. II § 28 p. 45). 30 eresim for *eri- -a-sun. 40 /ccir-a-sun; fear- probably stands for *qtuf- (cp. arm-ukn 'elbow' = Skr. ir-md- Avest. ar"-ma-, I § 306 p. 241). 50 yi-sun, cp. I § 330 Rem. p. 265, and Bugge, Beitr. zur etym. Erlaut. der armen. Sprache, p. 10; whether *pe'idqe- or *peidqe- be the form contained in this word it is impossible to decide. 60 vat-sun^ cp. vec 'six' and ves-tasan '16'. 70 evfan-a- -sun. 80 uf-sun. 90 inn-sun. The numerals in -sun remained for a long time indeclinable, and afterwards, like Ksan '20' (§ 177 p. 34), became inflected ; they were declined as «-stems, e. g. gen. dat. eresn-i-c ; later they were declined in the singular as well. Ordinals: eresn-erord liarasn-erord etc. Greek. -Kovta. was indeclinable from proethnic Greek onwards. Occasional exceptions to this rule, such as ri-aasQay.6vTMv (inscr. of Chios), rpiiyxoVrwv (Hesiod), TQirjmvisaai (Anthol.) are re-formates of a late period, and so are ni/tmov (§ 169 p. 13) and Skioy (§ 174 p. 22). 38 Cardinals, Abstract Numerals, and Ordinals. § 178. 30. TQta-Mvta Ion. rgirj-xovta instead of *XQl->ioi)Ta. *r(ji- first gave place to *Tpia-, since all nom. ace. pi. neut. took the termination -a from consonant-stems (§§ 337 ff.) ; then a was lengthened on the analogy of rsTQa-xovra and ntvrij-Kovva. Similarly we have Tpia-xdg. 40. Dor. Ion. TSTgco-xovra = Idg. *qety^-. Att. rsTTagd- -y.ovta Ion. Arcad rf (Tfffpa'-xojTa Boeot. Tifrrapa-xovra. Cp. § 176. 50. Att. Dor. etc. nsvTtj-xovTa = Idg. *pei9qe-. Cp. § 176 p. 29. The ->]- of this word passed on to the following multiples of ten, as in Latin the a of quadrci-gintci passed on to quinqua- -gintoi sexa-gintcL and the rest. 60. Att. Dor. etc. s^-tj-xovra, Cret. fsSrjy.ovra. 70. Att. Ion. s^So/u-ij-xovra^ Heracl. Delph. s^SifxrjxovTa. It is not clear how s^So/li-tj- is to be explained (cp. s^do/.iug £^do/.idxig and s^io/uo-g). Perhaps it contains an Idg. *septom- (cp. *-dh)m-t- beside *-dZm-t-). Cp. § 171 p. 17, and the Author, Morph. Unt. Y 36 fP. e^3s/.iijxovrtt seems to be a modification of s^dofirjKOvra due to the influence of ivtvrixovra. Thus modified it gave rise itself to the Epidaurian e^ds^iaZog (p. 19). 80. Horn. oydoi-xovTa like Lat. octo-gintd. Att. Lesb. oydo- -TJ-xovTa Heracl. bySoijxovTa (cp. Heracl. oxrcJ § 172 p. 20), like vulgar Latin octua-ginta for *octovol-. 6ySo-rj- (cp. oydodq and oySoo-g) presents the same difficulties as £^So/.i-?j-. See § 172 p. 20 for oyi^oo-g, and the Author, Morph. Unt. Y 36 ff. 90. Hom. fi'vijxovra Oetean svrjy.oyra for *ivf-rj- like Armen. inn-sun^ c^.Hom. Ew-fj /.tag -^ in the Homeric age the words used seem to have been hfTJxovra evfij/uao (the Author, op. cit. 41 ff. and 45). Att. Hom. Ion. hsvrj/.ovra Heracl. svsi'fjy.ovra (cp. Heracl. evvsa § 173 p. 21) for *evfiv->-'-xoi'Ta, from which we should infer an Idg. stem *enuen- *neuen-, to which it is possible to refer Lat. non-a-ginta non-u-s and O.Ir. noicht-ech "of ninety years'. Cp. the Author, op. cit. 39 ff. In the ordinals of the tens from 30 to 90 *-y.affTo-g became -xoarii-g through assimilation to -xovra in proethnic Greek (§ 176 Rem. 2 pp. 30 f.). r()iaxo(}T6-g. rfrpwxoffro-? and rfTTagaxoaro-g. 7iEVT7jxoar6-c. a'irjxoazo-g. £(ido/u7]xoor6-c. oydotjxoOTo-g. evtvijxoovo-g. § 178. Cardinals, Abstract Numerals, anj Ordinals. 39 Italic. No evidence is forthcoming except in Latin, -gintd instead of *-gonta owed the change of its final to the analogy of o-stems (cp. § 342); -i- (earlier -e-) instead of -o- may be due either to the vowel gradation seen in different cases of the Idg. neut. pi. *]comt-3, or to the analogy of vi-ginti (§ 176 p. 30); as to -g- instead of -c-, the student may refer to § 177 p. 34. tri- is the old nom. ace. neut. quadra- is the Idg. *qetuf- ; its -a- passed on to the following tens, as the -rj- of TiEVTij-xoi'Ta did to sS>j-y.ovTa and the rest, quinqua- (instead of *(juinque-). sexcL-. septus,- has been assimilated to octua-. octud- for *octov-S,- (like Gr. 6'ySo-7J-y.ovTa) belongs to the popular language (cp. § 172 p. 21); the literary form is octo- (like Gr. nyd(u-xovTa). It is uncertain whether non-cL- stands for pre-Italic *neuen-, like Gr. sv(f)i!v-ri-y.nvTu, or for pre-Italic *MeM^-, like Gr. *svJ^-7]-Y.ovra, (Horn. ivvrjxovTa). Ordinals, trlcesimu-s (like vuesimu-s) and tn-gesimu-s. Only -gesimu-s occurs in the rest of the series, quadrdgesimu-s etc. Old Irish, -cha -ga and -ca (see I § 514 pp.375 f.) for *-Tcomt-s, gen. -chat dat. -chit -chait. 30 iri-cha with original short j, as Bret, tregont shows; Hrecha would be the regular form; the word may have taken its present shape under the influence partly of tri 'tria', which is used before substantives as an independent word, partly of fi-che '20'. tri- is the stem, in place of nom. ace. pi. neut. *tri-. 40 cethor-cha either for *cetiAra- (nom. ace. pi. neut., cp. Gr. nrrapd-KOvra) or for *cetru- (the stem, cp. Gall. Petru-corius and tri-cha); Mid.Ir. cethrarha, which doubtless follows cethri 'four'. 50 coica, perhaps by syllabic dissimilation (cp. Gall. Leucamulus for *Leuco-camulo-, I § 643 p. 488) ; is the contained unit *peKqe- or "petdqe-? see the Author, Morph. Unt. V 83. 60 ses-ca. 70 sechtmo-ga -go, which may stand for *sechtrp,m-u-cont- or for *sechtom-u-cont- (cp. cethorcha for *cetru-cont- [?] and O.Cymr. trimuceint '80'). Cp. the Author, Morph. Unt. V 38. 80 ochtmo- -ga is certainly an ad-formate of sechtmo-ga. 90 Mid.Ir. nocha or nocha , O.Ir. perhaps *noicha (cp. noicht-ech 'of ninety years') ; was *nd(i)ca the older form (see I § 212 pp. 178 f. and 40 Cardinala, Abstract Numerals, and Ordinah. § 178. § 513 p. 375) and did tri-cha cause the change from c to ch, or was it *no(i)ncha, where ch instead of c would shew that a vowel had dropped between n and cha? It remains a doubtful point whether the contained unit is *neu'^- or *neuen- (cp. Gr. ev(f)sv-TJ-i(ovTu). The Ordinals end in -mad, as 50"" coicet-mad coicat-mad. For expressions like tri deich '30', coic deich '50' (cp. the Germanic and Balto-Slavonic) and da fichit '40' tri fichit '60' (cp. Alban. du-zet, tre-zet) see Stokes, Bezzenb. Beitr. XI 167 f., and Pott, Die quin. und vig. Zahlm. 99 ff. Germanic. Goth. 30 preis-tigjus, ace. prins-tiguns, gen. prije-tigive. 40 fidvor-tigjus. 50 fimf-tigjus. 60 saihs-tigjus. O.H.G. dn-zug (the spirant 2 is due to the preceding vowel, see I § 533 p. 390; yet on the analogy of zwein-zug and the following tens the word came to have z = ts, as the spelling tncig etc. shews), fior-zug, finf-sug, sehzug sehszug (the latter a re-formate, cp. Lat. sescenti and sexcentT, § 180). As regards the origin of tigu- and -zug, see p. 31 footnote 1. For 70, 80, and 90 we have in Gothic sibunte-hund ahtdute- -hund niunte-hund, which are mostly indeclinable, though once we find a gen. in -is, niuntehundis ; in Old High German of the oldest period, sibunzo ahtozo {-z- instead of -g- is a re-formation) niunzo (not actually found, but this is a mere accident) ; in Old Saxon ant-sibunta ant-ahtoda; and in Anglo-Saxon hund- seo- fonti^ hund'Cahtati-^ hund-ni^onti^. These were all ad-formates of an original expression for 100, Goth, taihunte-hund O.H.G. zehanzo A.S. hund-teonfi^ , which will be explained in § 179. Probably the Indo-Germanic expressions for 70, 80, and 90 which answered to Goth, preis-tigjus etc. lost their original meanings in proethnic Germanic, and were then superseded by this new series which follows the analogy of taihunte-hund. Yet in West Germanic there fas a kind of reaction to the older type, and O.H.G. sibunzo ahtozo niunzo during the ninth and succeeding centuries were gradually made to conform to the type of the preceding tens, and transformed into sibunzug ahtozug niunzug ; and similarly, in Anglo-Saxon, *hund-seofonta became § 178. Cardinals, Abstract Numerals, and Ordinals. 41 hund-seofonti^, and the others of this set were changed in like manner. Cp. the Author, Morph. Unt. Y 49 f. The forms in -zug and -zo were still regarded as substan- tives in O.H.G., since they governed a genitive case; as feorzug wehhono, sibunzo wehhono; the present type is in vierzig wochen 'in forty weeks', like Skv. pancaiadbhir vdnais (§ 178 pp. 36 f). No ordinals are found in Gothic. In O.H.G. we have dn^ugosto etc. like zweinzugosto. Balto-Slavonic. Lithuanian. 30 trps deszimtys (stem deszimti-) and deszimts (stem deszimt-), like O.C.Sl. cetyri des^ti beside ceUjre des^te. Each word of the expression was declined independently (with the gen. pi. of the word whose number was to be expressed), as ace. trh deszimtis, gen. triju deszimtu. Similarly 40 keturios deszimtys [deszimts) , ace. ketures deszimtis, etc. These expressions are found in Old Lithuanian, and still survive as dialectic variants; but as a rule they became com- pounds, the unit coalescing with the ten. The accusative became the regular form in the first part, and in the second, -deszimts was crystallised in some dialects, as trls-deszimts ketures- -deszimts etc.; whilst elsewhere (in the literary language) dvi- -deszimt '20' set the type for the final member, and its -deszimt passed on to the rest of the series, as trh-deszimt etc. Other kinds of change in the older language are discussed by Bezzen- berger, Beitr. zur Gesch. der lit. Spr., 181 f. — Ordinals. Old Lithuanian has such phrases as penkta-s deszimta-s (cp. antra-s deszimta-s '20"''), and such compounds as penkta-deszimta-s, 'SO*'. The forms now used, trlsdeszimta-s keturesdeszimta-s etc., have been modified by association with the cardinal. Forms with the 'vowel of composition', like keturid-deszimta-s '40*' septynid- -deszimta-s '70*''', are also said to occur. See Bezzenberger, op. cit. 185 f.; Schleicher, Lit. Gr. 151 f. Slavonic. 30 tri desqti. 40 cetyri desqti, and masc. cetyre desqte. 50 p^fi desqtu = nsi'Toig dsxddcoy, 60 sesti desqtu etc. — The Ordinals end in -mu, trides^tmu, cetyrides^tmu p^ttdes^tmu 42 Cardinals, Abstract Numerals, and Ordinals. § 179. etc. ; sometimes they contain the Vowel of composition' -o-, as sedmodes^tmu beside sedrmdes^tmu '70"''. § 179. Hundred. The Idg. cardinal was *h^t6-m for *dhrkt6-m (§ 164 pp. 2 f.) group of ten {sc. tens)', a neuter subst. governing the gen. pi. In this word 'tens' is understood, as it is in Skr. dasati-s, which means both 'decas' and 'centum'. But in Goth, taihunte-hund 'diy.u^wv dsxdg' the original expression seems to have been kept without abbreviation. Skr. iatd-m. Yarious constructions are found, — iatena hdrinam, iatena hdribhis and iatd hdribhis 'with 100 bay steeds'; and the Yeda has katd puras as well as iatq puras '100 cities', sata- in composition, as iatd-patra-s 'having 100 wings', but also satdm-uti-s 'offering a hundred helps, giving help an hundred-fold', Avest. safe-m. Armen. hariur, of doubtful origin (cp. Ascoli, Kuhn-Schl. Beitr. V 212 f.). Gr. sy.arov, which has become indeclinable and is used always as an adjective, as harov avdgdai. This is the sole form found in composition, Ixaro- having entirely disappeared; examples are exaTo/u-ftf; f^aToy-xngo-g (ey.ar6oTOf.io-<; may be derived from *sxaTovarofto-g, as laid down in I § 204 p. 171); — we even find such compounds as snarovTa-xd^rjVo-c (cp. sy.arovrus saavov- Totyig) , following rpiay.ovrd-Cvyo-c and the like. Arcad. saorov- -^oia like Ion. Att. -xnaioi, cp. § 176 Rem. 2 p. 31. s-xarov is probably a confusion of two modes of expression, *u-xar6-v (cp. Skr. sa-hdsra-m 'one thousand') and *£v xarov (cp. Alban. ns Hint O.H.G. ein hunt 'one hundred'), which were used inter- changeably like Skr. dvi-iatd-m and dm iate '200'. Lat. centum, like ty.aTni>, is crystallised and used as an ad- jective; but centi- (centu-) is found in compounds, as centi- -manu-s, though we also have centum-pondiu-m centum-peda etc. (cp. Skutsch, De nominum Lat. compositione, p. 37). O.Ir. ce/., declined as a neuter o-stem. Also coic fichit. Goth, hund O.H.G. liunt n. only in 200 and the following hundreds: Goth, tva hiiiida O.H.G. zwei hunt etc., whence ein hunt, but only in late O.H.G. The word for hundred in § 179. Cardinals, Abstract Numerals, and Ordinals. 43 Gothic was tai-hunte-hund (sometimes distorted into taihuntai- -hund), in Anglo-Saxon hund-teovti^ instead of *-teonta, in the earliest Old High German zehamo {hunt being dropped), lit. 'dsxdSmv dixug; taihunt- = dsxuS-, common ground-form *de^- i^d-, cp. Wheeler, Der griech. Nominalaccent p. 38, and in this work vol. I p. 199 footnote 1 and § 469. 7 p. 346. I re- gard this, as I have already said, as being the oldest Indo- Germanic mode of designating a hundred, and I consider the old Germanic expressions for 70, 80, and 90 to be re-formates following the analogy of the number 100, Goth, sibunte- being equivalent to Gr. sTjrddmv, and niunte- to Gr. swiddiuv. See § 178 p. 40, and the Author, Morph. Unt. V 11 ff., 139 fP., and 268. O.H G. zehamo with a like zehan, see § 174 p. 23. With regard to O.Sax. ant- in ant-sibunta, which is a distorted form of hund-, see the Author, op. cit. p. 142, and what is said in § 352 of this volume on Norse Himic pri-taunta. Lith. szimta-s (which has become masculine , see § 403) and O.C.Sl. suto are in living use as substantives. In suto the M is strange; perhaps the word was borrowed (cp. the Author, Techmer's Internat. Ztschr. I 251; G. Meyer, Alban. Stud. II 13 f.); we should expect *sqto., which seems to be represented in tysqsta for *ty-s^t-ia; see § 181. For the Ordinal, the original proethnic expression has not been clearly determined. Only two branches of the language agree in a formation which could be regarded as proethnic : Skr. -Satd- Lith. szimta-s. Skr. sata-tamd-s Avest. sato-tema- (for the -6- cp. II § 73 p. 178). Sanskrit has also katd- in composition, as ekaiatd-s 'lOP''. Armen. hariur-ord., hariur-erord. Gr. exar-oavo-g following XQimoOro-s etc., cp. also smtovtu- -YMQrjvn-g on the last page. Lat. cent-esimu-s following trtceshnu-s etc. O.Ir. cet-mad. O.H.G. zehanzug-osto. Lith. szimta-s (szimtds-is) ; it is certainly wrong to assume 44 Cardinals, Abstract Numerals, and Ordinals. §§ 179,180. that this stands for *S2imtn-ta-s, as Bezzenberger does, or for *ssimt-ta-s, with Pott and Schleicher. O.C.Sl. sut-inu. § 180. Two Hundred to Nine Hundred. Cardinal and Abstract Series. The parent language had two methods of expressing these multiples of a hundred. The unit might be prefixed to '*^itit6-m, both being in the same case and in the dual or plural number; as *duoi Tcmtoi du. '200', *tr% Hf^tci pi. '300' and so forth. This usage is found in Aryan, Irish, Germanic, and Balto-Slavonic. The other mode was to make a singular compound, whose first part was the stem of the unit; as *dui-1crnt6-m 'the state of being 200', tri-Jcint6-m, etc. This appears in Aryan, Greek, and Latin. Aryan. Skr. 200 dve sate and dvi-satd-m , and later a re-formate dvisati f.i), 300 trini katdni and tri-satd-m triSatt etc. Avest. 200 diiye suite (for duye see Bartholomae, Handb. § 92), 500 panca sata, 900 nova sata. Armenian. 200 erku hariur and erkeriur, 300 erelc-hariur, 400 coreJc-hariur etc. Greek. A group of compounds formed with -xano- (so Dor. and Boeot., -xaaio- Arcad., -aoato- Ion. Att., as to the first o of which see § 176 Rem. 2 p. 31 and § 179 p. 42) was derived from the neuter abstract series by adding -iio-. For example, tstqu- xor-jo- 'connected with the state of being 400, consisting of 400' is deriyed from *rfrp«xaro-)' 'the state of being 400'; cp. Skr. -iat-t/a-, as sastriSacchatya-s 'consisting of 136' satin- (Ved.) 'forming a group of 100, hundredfold' (where -in- stands for from -io- -en-, see II § 115 pp. 857 f.); and cp. also Goth. pusundi O.C.Sl. tys^sta '1000' (§ 181), which is probably to be derived from *tus-1crri,t-io- 'containing many hundreds', and the same suffix -iio- in x'i'^-">- Skr. sahasr-iya- (§ 181). Hence the use of the singular, for example, in Thucydides I 62 rrjv Sia- •/.oalttv "nnnv 'cavalry consisting of a group of 200' and Xenophon Cyr. lY 6 2 "nnov syo dia/jliai' roia/o'v'av. This series of 1) Kluge holds that these compounds in -tatl are original forms, of which the Greek and Latin words in -eenti and -xanoi are trans-formates (Paul's Grdrss. I 408). This view is untenable. § 180. CarJinals, Abstract Numerals, and Ordinals. 45 derivatives in -m- then superseded constructions corresponding to Skr. dv^ Sate and dviiatd-m, which must have once existed in Greek, precisely as x^^ioi has ousted */f (jAo-v, which answered to the Skr. sahdsra-m. 200 iSiaxoaioi; Ion. dirjy.ooioi instead of *J«-xo'aio« by assimilation to 300 rgia-xoawt Ion. rfjitixoaioi. This latter form itself may have arisen from a blending of *Tpt- -Koaioi with *Tpia y.ard (Skr. trtni iatdni), the u being length- ened after the analogy of T(jid->iovra (cp. § 178 pp. 37 f.) ; or, as seems to me more probable, it was transformed from *rp(- jto'ffioj after the analogy of roia-xovra, as the Homeric nivrt]- y.oawt undoubtedly has been assimilated to vevTij-y.ot'Tn. 400 TfTpaxnrrtoi. 500 Tifvtay.oatni instead of *nsi'TS-y.6atoi, like nsfta- -nr)yyc etc., see § 169 p. 13. 600 s^amaioi. like s^d-nohg etc., see § 170 p. 16. 700 s7TTay.6cioi. 800 oxrny.naioi, like o/.Tci- -77otv etc., see § 172 p. 20. 900 tmy.oaioi. Cp. the Author, Morph. Unt. V 7 ff. Italic. Only Latin has any examples. The neuter abstract series is represented by O.Lat. forms with -centum -gentum, du- -centum 'a group of 200', etc. Their original character is seen most clearly when they govern the genitive case, in descriptions of weight and measure with aeris, auri, frumentt and the like, as argenti sescentum (Lucihus). And in one instance nongentum is used as a crystallised adjective, precisely as centum is, C. I. L. rV" 1136 locantur balneum Venerium et nongentum tahernae per- gulae cenacula. ducentum became the plural adjective ducentt in very much the same way as Gr. *SsKaxti'>^ov avdpuv becomes iSmdxiiXoi avigsQ (Horn.), and Skr. pancaiata vmanam becomes pancasadbhir vamiS, etc. (§ 178 pp. 36 f.). nongentu-s "be- longing to 900' (Plin. XXXIII 2 § 81) is an instructive form; it is related to nongentu-m as tri-viu-s 'connected with three ways' to tri-viu-m 'place where three ways meet', du-centi hke du-plex etc., § 166 p. 7. tre-centl, cp. § 167 p. 8. quadrin- -genti instead of *quadru-, following septin-gentl. quin-genti (quincentum Fest.) for *quinque-cento-. sescenti Uke misceo for *mic-sceo (I § 503 p. 369), and, once more assimilated to sex, sexcenti, cp. 0. B..G. sehs-eug '60' instead of sehzug, which is 46 Cardinals, Abstract Numerals, and Ordinals. § 180. also found (§ 178 p. 40). septin-genti. octin-genti instead of *octi- or *oct6- following septin-gentT. non-gentt, and in Co- lumella non-in-gentt following septin-gentl. The -g- and -c- haTe been discussed in § 177 p. 34, where we concluded that the sound represented by g is probably Idg. g ; and that if the voiced character of the consonant is really so old, these Latin numerals are based upon proethnic stem-compounds, *d^i-hp,t6-m and so forth. Cp. the Author, op. cit. 3 ff. Old Irish. 200 da cet, dat. dib cetaib, 300 tri cet etc. Germanic. 200 Goth, tva hunda (dat. tvdim hundam) O.H.G. swei hunt, 300 Goth. pri)a hunda O.H.G. thriu hunt etc. O.H.G. also has such phrases as zwiro zehamug 'twice 100', finf stunt zehanzug 'five times 100', cp. Gr. dia-%tXiot. Lith. 200 du szimtu or duszimtu, 300 trys szimtal or try{s)szimtai etc. In Bretken we find szimtas crystallised in the singular form : du szimts vyru '200 men', szeszi szimtas vyru '600 men' etc. O.C.Sl. 200 duvi sut6, 300 tri suta and so forth. Ordinals. Sanskrit. Here the words are associated with the neuter abstracts : 200"" dviiatd-s and dviiatatamd-s, 300* triiatd-s and triiatatamdrs etc. Armen. 200* erkeriiir-erord etc. Gr. Siaxoai-onTO-g, rpiay.oOi-oOro-g etc. are re-formates like sAaT-oard-c, see § 179 p. 43. Lat. ducent-esimu-s trecent-esimu-s and so forth (besides nongentesimu-s Priscian vouches for noningentesimu-s , which is like noningenti, for which see above). Eemark. Priscian has preserved certain forms which do not occur elsewhere, namely ducesimus trecesimus quadrigesimus guingesimus sescesi- mus septigesimus octigesimus vSngesimus. These cannot be really an old series, simply for the reason that -cesimo- must represent *-cent-\-tumo-, and -cent- (instead of -cento-) cannot have been really an old expression for 100. They look as though the names for the multiples of ten, vi- -cesimu-s and the rest, had been altered by the stem being substituted for the old ease or quasi-case, the meaning of so many hundreds being given to the new word. O.Ir. and O.Germ. No forms preserved. §§ 180,181. Cardinals, Abstract Numerals, and Ordinals. 47 Lith. 200* duszimths-is etc. 0.C.81. 200"' dvosutlnu (where the Vowel of composition' has foand its way into the word), SOO**" trisutmu, and so forth. § 181. Thousand. The different languages do not agree in their modes of expressing a thousand; hence we cannot be sure how it was expressed in the parent language. See § 164 p. 2. *gMslo- is the form indicated by Skr. sa-hdsra-m Avest. ha-zafdre-m, Gr. Lesb. x^^^'"" Dor. xv^'oi Ion. /slXtoi (I § 565 p. 423); Att. /thoi. may come from Idg. *ghsl6-, see Thurneysen, Kuhn's Ztschr. XXX 353. Skr. sa-hdsram 'one thousand' like Gr. s-y.ar6v 'one hundred', see § 179 p. 42; sahdsram fstnam and sahdsram fsayas '1000 bards', like katdm fsmdm and katdm fsayas; and, with the number assimilated, sahdsrWny ddhirathani '1000 waggonloads' like ^atd ptiras. In Greek, *ivvsaxHXo-v 'a group of 9000' and *dsiiax£(Xo-v 'a gTOup of 10,000' became plural adjectiTes: Hom. swsaxsdoi, dsmxfdot (cp. § 180 p. 45), Idg. *ghesl-iio- 'consisting of 1000' : Skr. sa-hasr-iya- 'consisting of 1000, thousandfold', e. g. sahasriyo hhagds^a. share consisting of a thousand, thousandfold share', Gr. ;ft^{o- ^'S'^'o-? ^^^ 'Innog (J/tf/tA/a (see p. 44), and further /tAtoi avSpi-g like tptSHoaioi avJpfff (see p. 45). Ordinals: Skr. sahasra-tamd-s , Gr. Att. XlXi-ociTo-g. Armen. hazar is borrowed from the Iranian. Lat. mille milia (meilia in Lucihus); it is often connected with Gr. f.ivpioi (see L. Havet, Mem. de la Soc. de ling., Ill 415, and Thurneysen, Kuhn's Ztschr. XXX 353); but I do not consider that this connexion has, been proved, mill-esimu-s like cent- -esimu-s. O.Ir. mile (I do not know whether the ordinal mtl-mad has been found) was probably borrowed from the Latin. Goth, pusundi f., in one place neuter, (in tva pusundja '2000'; but this form may be regarded as a nom. du. fem. in Idg. *-ai, see § 286), O.H.G. dusunt thusunt f. and n. Lith. tukstanti-s gen. -czio (Lett. tMst&t-s), ordinal tukstantys-is; O.C.Sl. tysqsta tysqSta f. for *-entia *-ontia, ordinal tys^Stmu. On the strength of the Frankish thus-chunde thius-chunde (from 48 Multiplioatives and Distributives. §§ 181,182. the dialect of the Salii, one of the three great branches of the Franks) O.Icel. pushundraS and "West-Goth, thyu-phadus 'chiUarch, leader of a thousand' (cp. pusundi-faps in "Wulfila), it has been prettily suggested that this word, common to Germanic and Balto-Slavonic, is a compound of an adjective *tus- and the word for 100. *tus- would be a word connected with Skr. tavds- 'strong, strength' tuvis-tama- 'strongest', showing the weak form of the stem, cp. Skr. instr. bhis-d from hhiyds- 'fear'. The meaning of this compound would be 'a group of many hundreds' ; see Scherer, Zur Gesch. der deutsch. Spr.^ 590, Bugge in Paul- Braune's Beitr. XIII 327, and Kluge, Paul's Grundr. I 406. The -n- of the Lithuanian and Lettic words (cp. Pruss. tusimto-ns ace. pi.) is enough to shew that some analogical transformation must have affected them ; perhaps they were associated with the participle of the present (Lett.) tukstu 'I swell'. Cp. the Author, Morph. Unt. V 10 f. MULTIPLIOATIVES AND DISTRIBUTIVES. § 182. Multiplioatives. 1. Numeral Adverbs and Adverbial expressions. The parent language had adverbs ending in -s for twice, thrice, and four times. 'Twice' *dui-s (*duui-s), cp. *dui- in composition and used independently § 166 p. 7. Skr. dvis^ Ved. duvis, Avest. his. Gr. Si'c. Lat. &js, O.Lat. duis also, see § 166 Rem. 1 p. 7. Goth. tvis- 'apart'. Mid.H.G. zwis^ O.H.G. zioir-or swir-o, O.Icel. tvis-var 'twice', and further O.H.G. zwis-k zwis-ki adj. 'twofold', O.Icel. tvis-t-r 'divided into two parts' Engl, twis-t, i. e. a cord or thread of two strands. 'Thrice' *tri-s. Skr. tris, Avest. pris. Gr. rglg. Lat. ter perhaps for Hers and this for *tris (I § 33 pp. 33 f.); beside which we find trmu-s for *tris-no- (§ 183). O.Ir. tress- 'third. doubtless for *tris-to- (II § 81 p. 247). O.H.G. drir-or O.Icel' pris-var 'thrice', cp. zwir-or tvis-var above mentioned; O.H.G. dris-k dris-ki 'ternus'. 'Pour times'. Skr. cat'Ar for *catur§ (I § 647. 7 pp. 493 f.) § 182. Multiplicatives and Distributives. 49 Avest. ca-Prus, cp. Skr. catur-daSa as contrasted with Avest. capru-dasa etc. discussed by Wackernagel in Kuhn's Ztschr. XXV 283 f. Lat. quater, the ending transformed by association with ter : cp. quaternu-s : ternu-s. There seems to have been another mode of expression in the parent language, to which are due the following : Skr. sa-kft 'once, one time' paflca kftvas 'five times', Lith. venq kafqt 'once' dii Jcartu 'twice' trls hartiis 'three times' O.C.Sl. duva kraty 'twice' p^tl kratu 'five times'. Uses peculiar to single languages: Skr. eka-varam 'once' tri-vOram 'thrice' from vara- 'the right moment for something, one's turn'. Avest. hiz-vap 'twice' priz- -vap 'thrice', neuters of forms with the suffix -uent- (cp. below, under 2) ; prisat-a-pwem "30 times' (suffix -tuo-). Gr. a-7iai 'once' ; the second part is connected with nriyvv/xi 'I make fast, strengthen' -naoaalo-^ 'peg', and probably had at first much the same meaning as another word belonging to the same root, namely O.H.G. fah 'part, portion' A.S. foec 'space of time, time' (cp. Mid.H.G. zwi-vach, manec-vach.) The adverbs from 'four times' onwards end in -m or -y.ig (Dor. -/.tv) : TSTgdni, nsvTcixi etc. The same -y.i occurs in ov-xl nokkd-y.i; it was doubtless a nom. ace. sing. neut. with the meaning 'hoc' (cp. Lith. ssl-s O.C.Sl. SI 'hie', § 409) ; cp. OsthofF, Morph. Unt. IV 241 f., and the Author, Gr. Gr.2 p. 131.1) Lat. semel; Wackernagel, Kuhn's Ztschr. XXX 316, con- jectures that this word comes from *si^-meli and is connected with Goth, mel 'time' (Mod.H.G. -mal). The adverbs from "five times' onwards end in -iens -ies: quinquies sexies etc. (and toties quoties), Umbr. nuvis 'novies'. Many conjectures have been made as to the origin of this ending; the most likely of them is Pott's, connecting it with Skr. My ant- 'how great? how much? how manifold?' iyant 'so great, etc.* (cp. the Author, Morph. 1) If -xi were the interrogative pronoun *gi-, as is assumed by "Wackernagel (Kuhn's Ztschr. XXV 286 f.) and J. Schmidt (Pluralb. 352), all the dialects but Thessahan must have had -ti in place of it. I there- fore oppose this view. For Thess. x(-g, see my Gr. Gr.^ pp. 54 f. Brugmann, Elements. III. 4 50 Multiplicatives and Distributives. § 182. Unt. Y 14). Thurneysen's view (Arch, fiir lat. Lexicogr., Y 275 f.) as to -iens is probably to be accepted; he regards it as a pro- ethnic Italic transformation of *-ient, so that My at, the neuter of the Sanskrit form, would exactly correspond to it. Osc. petiro- -pert 'quater', cp. Lat. semper. O.Ir. oen-fecht oenecht once' (fecht 'way , course") , fo di 'twice', fo thn 'thrice' etc. Groth. dinamma sinpa 'once' tvdim sinpam 'twice' etc. O.H.G. eines (gen. sing.) 'once', drio-stunt fior-stunt etc. (stunta 'section of time') ; also expressions with warb , as sibun warb {hwarba 'a turning round'); with spurt 'stadium', as drim spurtim; and with mal 'point of time', as z'einemo male 'one time, once', su drin malen 'thrice'. And see further: J. Grrimm, D. Grr. Ill 231 ff., and Rumpelt, Die deutsch. Pron. und Zahlw. 167 IF. Lithuanian has also a set of phrases with syki-s 'blow, stroke', as penkis sykius 'five times', cp. the Upper German schlag 'blow' = mal. O.C.Sl. has phrases with -s(t)di -Mi (from sid- 'to go') : dvasdi 'twice' trisdi 'thrice' etc. (cp. Leskien, Handbuch p. 95). 2. Adjectives. With -y,ent- (II § 127 p. 404): Avest. visaiti-vant- 'twenty- fold' prisap-wanf- 'thirtyfold' xsvasti-vant- 'sixtyfold', Gr. vsTQag -ai'Tog, a coin worth four xakmX, for *rsTQa-fsvT; of which rgiag is an ad-formate. The following are etymologically connected: Gr. ci-nko-g d't-Til6-c etc., dl-nakTo-g TQi-nuXTO-g and di-nXdaio-g rpi-nkdnio-g etc., Lat. sim-plu-s du-plu-s tri-plu-s etc., Goth, din-falp-s fidur- -falp-s O.H.G. ein-falt zwi-falt dri-falt etc. These are related to Goth, faltan 'to fold' Skr. puta-s puta-m 'a fold' (cp. I p. 209 footnote 1), as Lat. sim-plex du-plex etc. to plectere plicare (J. Schmidt, in Kuhn's Ztschr. XYI 430, gives an explanation which may, I believe, be reconciled with this, although at first sight it seems to be different); but the second r of Umbr. tri- -brisine 'triplicitate' (-&>•- for -pr- quite regular, I § 499 p. 366) as compared with tri-pler 'triplis' du-pla 'duplas', has not been explained. Gr. -nXno-g in d-nXoo-g etc. we may conjecture to §§182,183. Multiplicatives and Distributiyes. 51 be akin to nXov-ro-g, and to have been early associated in the popular mind with -ttAo-^.') There remain a large number of other formations answering to Modern German adjectives in -fach -faltig '-fold'. Of these a few examples may be given. Skr. cdtur-vaya- 'fourfold', ddia-gva- dasa-gvin- 'tenfold', tri-vdrtu-s tri-vft- 'threefold', Gr. TQi-qiaTo-g TQi-(pda(o-g "threefold"; Lith. dvi-linka-s 'twofold' (linki-s, gen. linkio, 'a bending'), 0.C.81. dvo-guU -gubmu Lith. dvl-guba-s Pruss. dwi-guhhu-s 'twofold' (0.C.81. gU(h)nqti 'to bend, incline, fold, move', Lett, gub-stu 'I crouch, bow' Lith. guha 'stack, rick"). § 183. Distributives. The oldest mode of expressing distributives was to repeat the numeral, as Skr. pdflca-panca 'five each' (Eig-Veda III 55 18), eka-eka-s (ekaika-s) 'one each, one at a time', purvas-pUrva-s pUrva-purva-s 'the first on each occasion' (cp. 11 § 53 p. 99), dvan-dvd-m 'two at a time, a pair', Armen. mi mi 'singuli' tasn tasn 'deni', Aesch. Pers. 981 /^vgla /.ivgia nsjunaaTav = xara /.ivgid^ag nsjund^ovTa. Cp. Pott, Ztschr. der deutsch. morg. Ges., XII 458 fP., Doppelung pp. 156 ff.; Lobeck, Pathol. I 184; "Winer, Gr. des neutest. Sprachidioms '' p. 234; Wolfflin, Zur distributiven Gemination, Archiv fiir lat. Lex. n 323. Adverbs : Skr. -sds, as eka-ids 'singly , one after another' dvi-Ms 'by twos, in pairs' sata-ids 'by hundreds': cp. Gr. s-xdg 'by itself, apart, afar' avSga-mg 'man by man'. Adjectives with the suffix -no-. Lat. binu-s for *bis-no-, tnnu-s for *tris-no- and ter-nu-s, quater-nu-s, senu-s for *sexno-, etc. (cp. J. Baunack, Kuhn's Ztschr. XXV 257 if.). O.Icel. has corresponding forms, tvenner prenner ferner pi. 'two, three, four at a time' for pr. Germ. 'Huiz-na- etc., compare Mid.H.G. swim m. 'thread of two strands' O.H.G. swirnSn zwirnon 'to twist'. Goth, tvei-h-ndi 'two at a time, two each', perhaps by a fusing of two suffixes -qo- and -no-. Lith. dvynii du. 'twins'. 1) In the derivation of -nXo-g from ~nXoo-q by 'hyphaeresis' I have no belief whatever. Cp. the Author, Ber. der sachs. G-es. der "Wiss. 1889, pp. 51 and 52. 4* 52 The Cases. §§ 183,184. Lith., witlt the suffix -io-, tre-ji 'three at a time' ketver-l 'four at a time', and by ad-formation penh-erl szesz-eri etc. Cp. Avest. tuirya- 'fourth' a-xtuirya- 'to be spoken four times" (used of a certain prayer). THE CASES.i) General Remarks. § 184. A noun or pronoun can express Case, Number, and Gender. 1) On tlie Indo-Germanic cases in general: Bopp, Vergl. ©r.I' §112ff. p. 245ff. Schleicher, Compendium* p. 497 ff. Fr. Muller, &rdr. der Sprachw. Ill 529 ff. Bopp, tjber das Demonstrativum und den Ursprung der Casus (Abhandl. der Berl. Akad. der Wiss., 1826). Scherer, Zur Gesch. der deutsoh. Spr. ^ 382 ff. Dfintzet, Die Declination der Idg.. Sprachen nach Bedeutung und Form entwickelt, 1839. Schleicher, t5l)er Binschiebungen vor den Casusendungen im Indogermanischen , Kuhn's Ztsohr. IV 54 ff. Grassmann, tJber die Casusbildung im Indogerm., ibid. XII 241 ff. Ludwig, tJber den vocalischen Ausgang der Bildungs- sufflxe, ibid. XV 443 ff. Stenzler, tJber die Terschiedenen Conjugationen und Declinationen in den Idg. Sprachen, bes. im Lat., Abhandlungen der Schlesischen Gesellsch. fiir vaterland. Cultur, Philosoph.-hist. Abtheil. 1864,. Heft I. Hilbsohmann, Zur Casuslehre, 1875. Bergaigne, Du role de la derivation dans la deolinaison indo-europeenne , M^m. de la Soc. de ling. II 358 sqq. Breal, Sur le noinbre des cas de la deolinaison indo- europeenne, ibid. IH 322 sqq. Penka, Die Bntstehung der syn- cretistiachen Casus im Lat., Gr. und Deutschen, 1874. Id., Die Nominalflexion der idg. Sprachen, 1878. Whitney, General Consi- derations on the Indo-European Case System, Trans, of the Am. Phil. Assoc, Xin 88 ff. De la Grass erie, i,t. de gramm. comp. : Des relations grammatioales considereea dans leur concept et dans leur expres- sion ou de la categorie des cas, Paris 1890. Leskien, Die Partikel -am in der Declination, ein Beitrag zur Analyse der idg. Casusendungen, Ber. der sachs. Gesellsch. der "Wiss., 1884, p. 94 ff. "We nek, Zur idg. Casus- bildung, Borna 1884. The Author, Zur Geschichte der stammabstufenden Declinationen, Curtius' Stud. IX 361 ff. Osthoff, Zur Frage des Ursprunga der germ. M-Declination, nebst einer Theorie iiber die ursprungliche Unter- scheidung starker und schwacher Casus im Idg., Paul und Braune's Bei- trage Illlff. Hillebrandt, Zur Lehre von den starken und schwachen Casus, Bezz. Beitr. II 305 ff. Regnaud, Examen du mouvement vocalique dans la deolinaison des themes indo-europeens en «, /, r et questions oon- nexes, 1883. Collitz, Die Flexion der Nomina mit dreifaoher Stamm- abstufung im Altind. und im Griech., Bezz. Beitr. X 1 ff. Straohan, § 184. The Cases. 53 The Cases. The original language had seven cases: Nominative, Accusative, Genitive, Ablative, Dative, Abstufung in Case-endings, ibid. XIV 173 ff. L. H a t e t , Le renforcement dans la d^clinaison en A, Mim. de la Soc. de ling. II 9 sqq. Id., Sur la dficlinaison des thfemes f^minins en A, ibid. II 387 sqq. Aryan. Bartholomae, Zur ar. Flexion der StSmme auf -r, -n, -m, -j, -■», Arisohe Forsohungen I 25 ff. Id., Die ar. Flexion der Ad- jeetiva und Partizipia auf tit-, Kuhn's Zeitsohr. XXIX 487 ff. "Whitney, Sanskrit Grammar p. 80ff. Boehtlingk, Die skr. Deolinat., St. Petersb. 1844. Lanman, On Noun-Inflection in the Veda (Journ. Am. Or. Soc. X.), 1880. Dutens , Essai sur I'origine des exposants casuels en Sanscrit, Paris 1883. H anus z , tjber das allmahliohe Umsiohgreifen der M-Decl. im Altind., 1885. F. Or. P. Storck, Casuum in lingua Palica forraatio compar. cum Sanscritae linguae ratione, Monast. 1862. Bartholomae, Handbueh der altiran. Dialekte, p. 65 ff. Osthoff, Das determinierende a bei Casus- formen im Altiranischen, Morph. Unt. II 76 ff. Horn, Die Nominalflexion im Avesta und den altpers. Keilinsohriften, I: Die Stamme auf Spiranten 1865. Bartholomae, Die gathische Flexion der ii-Stamme, Bezz. Beitr. Xin 89 f. Armenian. Fr. Miiller, Beitr. zur Declination des armenischen Isomens, 1864. Greek and Latin. Henry, Precis de grammaire oomparee du greo et du latin, ^ p. 192 sqq. Hartung, IJber die Casus, ihre Bildung und Bedeutung in der gr. und lat. Sprache, 1831. Grotefend, Data ad Hartungium de principiis ac significationibus casuum epistula, Gott. 1835. Schmidt-Stettin, tJber die Anordnung der Declination der Nominen im Qriech. und Lat., Hofer's Ztschr. fur die Wiss. der Spr., Ill 310 ff. Leo Meyer, Gedrangte Vergleichung der griech. und lat. Declination, 1862. Ebel, Starke und schwache Formen griechischer und lateinischer Nomina, Kuhn's Ztschr. I 289 ff. Leo Meyer, Die einsilbigen Nomina im Qriech. und Lat., Kuhn's Ztschr. V 366 ff. Sohwarzmann, Uber Ursprung und Bedeutung der griech. und lat. Flexionsendungen, Ehingen 1865. Duntzer, Die urspr. Casus im Gr. und Lat., Kuhn's Ztschr. XVII 33 ff. Wegener, De casuum nonnullorum Graecorum et Latinorum historia, 1871. Bornhak, D'ber die Casuslehre der gr. und lat. Sprache, Ztschr. fiird. Gymn. 1872, p. 307ff. Chaignet, Th^orie de la declinaison des noms en greo et en latin d'apr^s les principes de la philologie com- pares, Paris 1879. Petroni, Dei oasi nelle lingue olassiche e particolar- mente del locative, Naples 1878. Greek. Kiihner, Ausfuhrl.Gr. der griech. Spr., Pp. 280 ff. G.Meyer, Gr. Gr.* p. 299 ff. The Author, Gr. Gr. (J. MuUer's Handb. der Klass. Alter- tumsw.n^) p. 116ff. Pezzi, La lingua greoa antica,p. 178 sqq. F. C. Serrius, Wissensohaftliche Entwickelung fiber Ursprung und Bedeutung der griech. Casus, 1839. Stolz, Beitr. zur Decl. der griech. Nomina, 1880. Moisset, iltude de la declinaison grecque par I'accent, Par. 1882. Gatto, Morpho- 54 The Cases. § 184. Locative, and Instrumental. It has always been the custom to define and arrange the whole mass of recorded forms under logia greoa: Osservazioni sulla deolinazione dei nomi con tema in a, Torino 1882. B. J. Haupt, De nominum in -svs exeuntium flexione Homerioa, 1883. A. Torp, Den greeske Nominalflexion, Christiania 1890 (published after this work had gone to press, and so not available for use). Italic. Lindsay, The Early Latin Declension, Class. Rev. II 129 ff. and 273 fp. Kiihner, Ausfiihrl. ttr. der lat. Spr., I p. 172 ff. Stolz, Lat. Gr. (J. Mailer's Handb. des Klass. Alt. II ^), p. 332 ff. F. N e u e , Pormenlehre dor lat. Sprache, I^ und 11^ 1 ff. K. L. Struve, tJber die lat. Declination und Conjugation, 1823. Bk, De formis oasuum Latinorum, Gotoburgi 1839. P. Biioheler, Grdrss. der lat. Deol. (1866), new edition by Windekilde, 1879 ; French translation (Precis de la decl. lat.) by L. Ha vet, with additions by the author and the translator, Par. 1875. Stoesser, Lat. Decl. der Substantiva und Adjectiva auf Grand der Brgebnisse der vergleich. Spraoh- forschung, 1872. Merguet, Die Bntwickelung der lat. Pormenbildung mit bestandiger Berucksiohtigung der vergl. Sprachforschung, 1870, p. 7 ff. Pumi, Note glottologiche, I: Contributi alia storia comparata della deoli- nazione latina, Palermo 1882. "Walter, Zur Declination der !<-Stamme im Lateinischen, Kuhn's Ztschr. IX 370 ff. Stolz, Zur lat. Decl., Wiener Stud. VI 136 ff. Aug. Milller, De priscis verborum formis Varronianis, 1877, p. 22 sqq. Schuchardt, Lateinische und Romanische Declination, Kuhn's Ztschr. XXII 153 ff. H. d'Arbois de Jubainville, La dSclinaison latine en Gaule k I'epoque merovingienne. Par. 1872. "W. Meyer, Die Schioksale des lat. Neutrums im Romanischen, 1883. B. Appel, De genere neutro intereunte in lingua Latina, 1883. Suchier, Der Untergang der geschlechtlosen Substantivform, Arch, fiir lat. Lex. und Gr., Ill 161 ff. G. Koffmane, Lexicon lateinischer Wortformen, 1874. Georges, Lexikon der lat. 'Wortformen, 1889 (not yet completed). — Z e y s s , De sub- stantivorum Umbricorum declinatione, Tilsit 1846—1847. C. Stephany, De nominum Oscorum declinatione cum Latinis comparata, Rostock 1874. Keltic. Zeuss-Bbel, Gr. Celt. p. 220 sqq. Stokes, Celtic Declension, Bezz. Beitr. XI 64 ff. "Windisch, Die irischen Auslauta- gesetze, Paul und Braune's Beitr. IV 204 ff. Stokes, Bemerkungen iiber die ir. Declinationen , Kuhn and Schleicher's Beitr. I 333 ff. and 448 ff. Ebel, Celtische Studien: Die Declination, ibid. I 155 ff., II 67 ff. Idem, Neutra auf -as im Altir., ibid. VI 222 ff. C. A. Serrure, Essai de gram- maire gauloise: Les d^olinaisons, in Le Museon VI 489 ff. and 511 ff. Germanic and Balto-Slavonic. Leskien, Die Declination im Slavisch-Litauischen und Germanischen, 1876. Germanic. Grimm, D. Gr. I^ (1870)]p. 508 ff. Delbruck, Die Decl. der Subst. im Germanischen, insonderheit 'im Gotischen, Ztschr. fiir deutsohe Phil., II 381 ff. S o h e r e r , Zur Gesoh. der deutschen Spr. ^ 546 ff. Wilken, Zur deutschen Declination, Germania XIX 18 ff. Kluge, Noreen, Behaghel, Paul's Grdrss. der germ. Phil. I 384 ff., 490 ff., S 184. The Cases. 55 these seven heads. But since meaning, and not form, is the basis of this classification, it often happens that forms etymo- logically distinct are grouped together, as in the Lat. gen. sing. equi and (O.Lat.) equas; whilst others which are really con- nected are separated, as in Skr. me dat. and gen., or the bh- suffixes, which have one part, and that the most important, in common. Details of case-usage will be found in the Syntax; this is the place only for a few general remarks. The Nominative implied that the noun idea was the central point of the action ex- pressed by the verb. The Accusative brought the noun into some dependent relation to the verb, the exact relation being deter- mined by the sense of the verb and noun in any given instance and 612 ff. Burghauser, Germ. IToininalflexion, 1888. Kahle, Zur Ent- wiokelung der eonsonantischen Declination im Gterm., 1887. Braune, Got. Gr. ' p. 37 ff. Ebel, Bemerkungen zur got. Deol., Kuhn's Ztschr. IV 138 ff. Treitz, tJber die Dec), der starken Substantiva im Gotisoben, Kuhn's Ztschr. XVI 344 ff. Braune, Althochd. Gr. p. 148 ff. Dietrich, Historia declinationis theotiscae primariae e fontibus describitur, Marburg 1859. Primer, On the Consonant Declension in Old Norse, Am. Journ. Phil. II 30 ff. and 181 ff. Balto-Slavonio. C.G.Smith, De locis quibusdam grammaticae linguarum Balticarum et Slavonicarum, II: De nominum declinatione, Havniae 1857. Leskien, Spureu der stammabstufenden Declination im Slav, und Lith., Arch, fiir slav. Phil. Ill 108 ff. Schleicher, Lit. Gr. p. 170 ff. Kurschat, Gr. der littau. Spr. p. 229 ff. Bezzen b erger, Beitr. zur Gesch. der lit. Spr., p. 120 ff. Bruckner, Zur Lehre von den sprachl. Neubildungen im Lit. (fiber Decl.), Arch, fur slav. Phil. Ill 238 ff. Pauli^ Preussische Studien, IT: Pormenlehre , Kuhnund Schleicher's Beitr. VII 515 ff. Bezzenb erger, Zur lettischen Declination, in his Beitr. XV 294 ff. Miklosich, Vergl. Gr. der slav. Spr. III^ 1 ff. Leskien, Handb. der altbulg. Spr. ^ p. 53 ff. Scholvin, Die Declination in den pannonisch- slovenischen Denkmalern des Altkirohenslav., 1877. Th. Vetter, Zur Gesch. der nominalen DecL im Russisohen, 1888. Przyborowski, Ve- tustisslma adjectivorum linguae Polonae deolinatio, Posen 1861. Baudouin de Courtenay, Binige Palle der "Wirkung der Analogic in der Polnisohen Declination, Kuhu und Schleicher's Beitr. VI 19 ff. Stephan, Smal, Stockij, tJber die "Wirkungen der Analogic in der Declination des Klein- russischen. Arch, fiir slav. Phil. VIII 194 ff., 409 ff. und IX 58 ff. Oblak, Zur Gesch. der nominalen Declination im Slovenischen , ibid. XI 895 ff., 523 ff. and XII 1 ff., 358 ff. "Works and Essays treating of single oases vfill be cited below. 56 The Cases. § 184. (accusative of object, of result, and so forth). The Genitive expressed some relation between noun and noun, this also being determined by their sense (genitive of origin, of object, and so forth) ; it also attached a noun to a verb in such a way that only a part (greater or less) and not the whole of it was affected or mastered by the action of the verb ; and thirdly, it formed adverbs of time and place. The Ablative denoted that the noun was the source from which the verbal action came. The Dative denoted that the noun was that for which the action of the verb held good, or to which it was directed. The Locative gave the sphere in which something was or some action took place, the goal of motion and the place where a moving thing comes to rest. Lastly, the Instrumental expressed that with which something was (accompaniment), or with which something was done (means). The Yocative is traditionally classed with these as an eighth case. But this was merely a method of address, or call, standing outside the sentence as far as syntax was concerned, and there- fore not properly a case at all. Numbers. There were three numbers. Singular, Plural, and Dual. The Singular expressed unity, and this number served for both single and collective ideas. The Plural denoted a number of similar things, and was also used where the same thing had a variety of forms or phases (as Skr. mftydvas Gr. d-avuTot 'kinds of death') ; it further denoted anything complicated, anything which consisted of parts or sections (e. g. Skr. Yed. dhUmdsas Lith. diimai 'smoke'). The Dual was used of two complementary things, commonly where by nature or convention they formed a pair. Further discussion of these points will be found in the Syntax. Genders. Lastly, there were three Genders in the parent language. Masculine, Feminine, and Neuter. The gender depended not so much on what we call case-suffixes as on the stem of the word, and it has accordingly been discussed under Formative Suffixes; see especially 11 § 57 Eem. p. 106, § 145 p. 458. More will be said of this in theSyntax. § 185. The Cases. 57 § 185. The manner in which particular relations of case, number and gender were expressed was in most instances the same: the stem received an accretion (sound^K) which brought with it some one of these meanings. But we cannot always tell exactly what the accretion was. In a certain number of forms, the point where the new part joined on to the old is quite clear, so that we may use a hyphen to divide the ending from the stem; e. g. in the nom. sing. *wa^-s (= Skr. nolti-s Gt. vttv-g) and *e^uo-s {= Skr. dsva-s Lat. equo-s), in the nom. sing. neut. *jugo-m (= Skr. yugd-m Lat. jugu-m)^ and in the loc. pi. *nau-su {== Skr. nau-sii, cp. Grr. vav-ai). In others it is a moot point how far we are justified in making a division: examples are the dat. sing. *eTcuoi (= Gr. "nnw, cp. A.vest. haomai dat. 'the soma plant') and in the nom. pi. *efi,y,ds (= Skr. asms, cp. Goth, vulfos wolves'). These forms might, it is true, be analysed *ekuo-i and "^ekuo-s, -o- being explained as a lengthening of -0- (ablaut), and as a matter of fact the -i and the -s were no doubt regarded in the unreflecting consciousness of the speaker as exponents of the relation in which these words stood to their sentence, even though the length of the stem-vowel served as a further mark to distinguish these cases from others, as from the nom. and ace. sing. Still, there is no reason why they should not have come from *ektioai and *ekuoes by vowel-contraction (cp. I § 115 p. 107). If so, -ai and -es would have been the proper case-endings at a period earlier than that which came just before the gradual dissolution of the parent language.') But these accretions to the stem were not indispensable; case, gender, or number could be expressed by the stem alone. For example, *ekua (== Skr. dsvoi Lat. equa) was nom. sing., cp. loc. pi. Skr. dsva-su; *me (= Gr. /xs Goth, mi-k) was ace, cp. Lat. mi-hi Skr. md-hyam; *dhemen and *dhdmen (= Skr. 1) Johansson (Bezz. Beitr. XIV 156) refuses to allow the rules for vowel contraction in the proethnic language which were given in the pas- sage of the first volume of this work cited above. I freely admit that if that paragraph were to be written over again, I should state most of these contractions with more reserve. 58 The Cases. §185. Gr. infin. &sf.isv) were loc. sing., cp. dat. sing. Skr. dhdman-e Gr. inf. d-s/^fv-ut; *pllu (== Skr. _pM»"t< O.Pris. fulu, cp. Gr. Tiolv Goth, filu) and *dusmenes (= Skr. durmanas Gr. dvajLisveg) were nom. ace. neut., cp. nom. sing. masc. Skr. pur'A-s Gr. noXv-g and ace. sing. masc. Skr. durmanas-am Gr. dvC^evi-a. There was another mode of expressing these three relations. This was by giving to the ending of the stem a particular grade of Ablaut. For example *p9ter (= Gr. TraTjfp Lat. pater) and *dusmenes {= Skr. durmanas Gr. Sva/usvijo) were singular nominatives, marked as such partly by the e {-ter- -es- being one grade of the formative suffix) which is wanting to the other cases; *§enos {= Skr. jdnas Gr. ysvog) was distiu- guished as nom. ace. neut. partly by its o (the os-grade of the formative suffix), which at the same time showed that the form was a substantive, cp. Gr. tptv^og as contrasted with ipevMq. In this instance, as in so many others, a difference which arose naturally from the working of what we call the Laws of Sound has been turned to account in distinguishing varieties of usage. The same remark holds good for some of the words whose form was distinguished by a special inflexional suffix. To . take an example: in *pdter-m and *p3ter-es (== Skr. pitdr-am pitdr-as Gr. narsQ-a narsQ-K;)^ the ace. sing, and the nom. pi. were expressed partly by the inflexional suffix and partly by the ablaut-grade of the formative suffix -ter-^ which distin- guished them from other cases with -tr- -t^- -ter-. In root- nouns, in the same way, this or that case was marked partly by ablaut-differences in the root-syllable (see II § 160 pp. 479 ff.). Cp. II § 7 pp. 15 f. Thus it becomes clear that in treating of declension, case- suffixes are by no means all we have to do with. We must also take account the difi'erent shapes of the stem. The chief relations of ablaut within the stems of words, so far as they affect declension — this we may call Case-Ablaut — have already been considered under the head of Stem- Formation. To this part of the Grammar we shall often have occasion to refer in what follows. §§ 185,186. The CaBes. 59 Oae special point must be mentioned. Forms which show strong-grade vocalism have been called Strong, and those with vocahsm of the weak grade have been called Weak. On the same principle, -we speak of Strong and Weak Cases; the Strong including the nom. ace. and voc. of all numbers (excepting the nom. and ace. sing, and du. neut., and perchance the ace. pi. masc. and fern, as well; see § 325), and the loc. sing.; while all the other cases are Weak. But two cautions should be given. 1. This classification of the cases holds only for consonant- stems ; e. g. Skr. uksdn- Goth, auhsan- 'ox', Skr. pitdr- Gr. narsQ- 'father'. It does not hold for stems in u^ such as Skr. sunu- 'son'. 2. It holds good primarily only for the proethnic stage of Indo-Germanic. Sanskrit has kept these old distinctions between the cases fairly well ; but in the other languages form-association and re-formation have changed and effaced them to a great extent; compare, for example, Greek narsQuov for the older Remark. In Sanskrit grammar, the Weak cases are subdivided into Middle and "Weakest, according as the case-sufflx begins with a con- sonant or a sonant ; e. g. instr. pi. uksd-hhis pitf-bhis and instr. sing, uhm-a pitr-d. Cp. I § 308 p. 245, §§ 311 f. pp. 247 k Gender will of course be discussed in the following pages only in so far as it is expressed by peculiarities in the case- endings. § 186. One difficult question must not be entirely passed over in this place. How did the case-endings, as we are able to restore them for the end of the proethnic period, come to have the meaning which they had? From the principles laid down in the first paragraphs of Volume II, we must assume that forms with a case-suffix, such as *e]cuo-s *elcuo-m, are compounds which once were phrases. What the final of each word of this kind actually was, before it became the sign of a case and the type after which new words could be formed at will, we have not the means of discovering by etymological research; the forms which have been trans- 60 The Cases. § 186. mitted from the parent language as fully developed cases do not give enough evidence. Conjectures there are in plenty, not a few of them reasonable enough to deserve mention here; prin- ciples which can be seen in action during later times often throw light upon what must have happened in times gone by. In those cases which expressed some relation in space, the inflexion may have been generally a demonstrative with some local meaning. "With regard to the -m of the ace. sing. {*eTcuo-m), we must remember that neuter forms which have it (as *juQO-m) serve for the nominative as well. Thus -m can hardly have had a proper accusative meaning to begin with. We may conjecture that -m was first used with o-stems only; that where an o-stem could have a form in -s (such as *efcuo-s), the m-form came to be contrasted with this in some vague indeterminate way, its meaning being narrowed to that of an accusative case ; and that afterwards -m was regarded as an accusative-suffix proper, and used as such with other classes of stems. It is tempting to identify this with the particle -m which appears in so many Cases, especially in pronomiaal forms (as Skr. ahdm 'ego' mam 'me'). See Gaedicke, Der Ace. in Yeda, 17 '); Leskien, Ber. der sachs. Gres. der Wiss. 1884, p. 101; Torp, Beitr. zur Lehre von den geschlechtl. Pron. 1888, p. 23. Remark. In a simUar way, the nominative -s became a sign of the masculine. It certainly had nothing to do originally with the contrasting of masculine and feminine, but was used indifferently with either ; then in the class of o-stems it was brought into contrast with the feminine, because words of that class had corresponding feminines without s ending in -a and -I (*e%oo-s 'horse': *eS?*a, and *?*|g'o-s 'wolf: *^dql). It is clear that to the Greeks s denoted the masculine, because they added -; to old feminine nominatives, such as *vean.s ('youth') when they were used to designate male persons, as vesrlsi; (§ 190 p. 67). Here -s came to denote the masculine gender, as we are supposing that -m came to denote the accu- sative case, and the masc. vsavCag bore the same relation to the 'feminine' noQTi-g noQi-q vav-q as the acc. 'Inno-v to the 'nominative' fwyo'-y. The -i of the nom. pi. Ho-i (= Gr. xo-i Skr. tt) cannot 1) ftaedicke's suggestion for the origin of this -»» (p. 18) seems to me improbable. The Cases. 61 be separated from the -i- which precedes so many plural case- suffixes (Gr. To-t-ai Skr. i, which served alike for singular and plural, both then and later; see §§ 274, 281, 379. The nom. ace. neut. too, in the proethnic stage, seems often to have had the same formation for singular and plural; see §§ 223, 337, 340, 342. § 189. The subject of Case Formation is not confined to cases proper, but includes adverbs as well. The history of Adverbs in their special uses will be set forth in the Syntax. We are here concerned with their form; and we shall discuss them after the following fashion. There are two classes of adverbial words. One consists of words which once were ordinary cases, but became isolated and thus crystallised; as GT.'Ad-r,vt]ai, oixoi, a^ia, vol, Jjat merito, modo, bene, facile, multum. Sometimes these are the sole evidence for a case-formation in some language or dialect; thus in Greek the old ablative in -od only survives in crystallised adverbial forms (§ 241). Then, but not otherwise, do they concern us here. The second class embraces words which never belonged to a regular paradigm; they were isolated words, used in such phrases as their meaning suited, but having no more than one or two other words at most connected with them closely enough to form such a grammatical group as we call a Paradigm. Most of them were built up on some pronominal Bnigmann, Elements. HI. 5 66 Nominative Singular Masculine and Feminine. §§ 189,190. stem, as Gr. sv-rog sn-rog, Lat. in-tus, Skr. kii-tas. However, these were often associated in meaning with the cases of certain complete systems, and raised to the rank of true cases; thus this same *-tos became a widely used abl.-gen. suffix in Sanskrit, Armenian, and Greek (§ 244). The suffixes of adverbs of this second class are accordingly included in the discussion which here follows, so far as they were in this way attached to any case-system. This part of our subject also includes Infinitives. We shall see in the Syntax how these forms, originally living cases, came to be used as they are. Here Infinitives belonging to any of the separate languages must be cited at least when they represent cases which have dropped out of living use in that particular language, as Gr. do/^sv-ui (§§ 245, 251) and So/.iiv (§§ 256, 257). THE CASES.') Nominative Singular Masculine and Feminine.-) § 190. I. Stems without any Case-suffix used as Nom. Sing. Masc. and Fern. 1. a-stems. Pr.Idg. *e^?^a 'mare'. Skr. ai»a 'mare' ; Avest. ha^na OJPevs. haina 'hostile host*. Gr. /w'pa 'land'. Lat. eg'wa; Umbr. muta mutu 'multa', Osc. tovto 'civitas' (I § 105 pp. 98 f.. 1) One or two kinds of Indo-G-ermanic inflexion — e. g. that re- presented by Skr. (dhiyq-)dhas^ dat. -dh-e — are themselves rare, and teach us nothing of the case-suffixes which cannot be learnt from the others. To avoid excessive detail, I have either passed these over entirely, or only just touched upon them by the way. In order to present before the student a complete paradigm of the oases of a given word, it has often been necessary to fill up gaps in the tradition by making certain forms after the analogy of other words. In a work like the present, I hold this to be not merely allowable but necessary. 2) C. Ma as 8, Tocales in stirpium terminationibus positae nominum Ital. Graec., imprimis vero Q-erm. post quas potissimum consonantes in sing, nominative perierint, Breslau 1873. The Author, Erstarrte Nomi- native, Curtius' Stud. IX 257 ff. J. Schmidt, Heteroklitische Nominative Sing, auf -as in den ar. Sprachen, Euhn's Ztschr. XXVI 401 if. / d., Idg. o aus oi in der Nominalflexion , with an Excursus: Zur Bildung des Nom. Sing., ibid. XXVII 369 ff. Osthoff, Der got. Nom. Sing, der mannlichen -j'a-Stamme, ibid. XXIII 89 f. § 190. Nominative Singular Masculine and Feminine. 67 and § 655. 2 p. 502). O.Ir. tuath 'folk' (I § 657. 2 p. 507). Ogam inscr. inigina girl, daughter' = O.Ir. ingen ; Gall. Deva. Goth, giha 'gift', cp. so 'that (f.), she' dino-hun 'any one (f.)', (beside dina) ; O.H.G. buoz 'improvement' (cp. below), A.8. ^iefu 'gift'. Lith. rankd, O.C.Sl. rqka 'hand'; cp. Lith. gerd-ji beside gera 'bona' (I § 664. 3 p. 523). Avestic. Forms in -e, as kainike 'girl, virgin' ber^xSe 'blessed' and Prussian forms in -ai, such as mensai 'flesh, meat' (Lith. mesa) show a pronominal ending; see § 414. Compare also § 202. Greek. Masculines like veavla-q 'young man' ysvsvrj-Q'he- getter' were originally feminine, and received their -g through being assimilated to such nominatives as i^io-g (11 § 79 pp. 229 f., § 80 pp. 239 f., § 157 p. 472) ; cp. the corresponding re-formation in the gen. sing., § 229. But the form without -g remained in use as a vocative, as 'Ep/ns/a, alvagsrij, cp. O.Ir. voc. pi. Jiru = *uiros beside nom. pi. fir = *uiroi (§ 314). This suggests the simplest mode of explaining masculine nominatives in -cl like Boeot. KaXXla, oXv/xmovtxa and Leucad. ^tkoxXslScc (cp. Megar. rov 'Aguiag^ § 229); these may be called vocatives used as no- minatives. Cp. the Author, Gr. Gr.^ p. 117, and J. Schmidt, Pluralbildung 354. As regards such feminines as roX/xa 'daring' fA.sQiftva 'care', see the Author op. cit. p. 102. Masculine forms like innora = tnnoTTjg 'charioteer' will be explained in § 202. Some curious masculine nominatives have been preserved in Italic. These are Latin 'hosticapas' ('hostium captor') and 'paricidas beside scnha agricola etc., and Oscan Katag 'Caha' Maglag 'Mara' Tanas 'Tana' beside Santia 'Xanthia, Savd-iag. The record is too scanty to enable us to decide whether these were imitations of the Greek forms in -S-g or independent of them. Old High German shows traces of a few, but only a few of these nominatives in Idg. -a: e. g. buo$, hwll 'while'; most of them, however, are abstracts in -ungo-, as samanunc 'assembly, gathering', cp. also siu 'this, that (f.), she' like Skr. siyd syd. See I § 661. 1 and 2, pp. 516 ff. The common 5* 68 Nominative Singular Masculine and Feminine. §§ 190,191. forms in -a, as geha 'gift' sipp(e)a 'kinsman', are accusatives used as nominatives (§ 213). § 191. 2. t- ie-stems'). Pr.Idg. ^bh^gh'^t-1, fem. of Hhfghont- 'projecting, exalted, high'. Skr. b^hatt, Avest. harenti 'ferens'. O.Ir. Brigit 'exalted lady' (= Skr. Ifhatt)^ inis 'island', I § 657. 2 p. 507 , op. also s-% "ea = O.H.G. s-i. Goth, frijondi 'friend (f.)'. A.S. thiwi thiu 'maid' = Groth. pivi, O.H.Gr. herzohin 'duchess' wirtun wirtin 'hostess' (11 § 110 p. 339), cp. also O.H.Gr. s-i s-i 'ea' (the latter, Hke Groth. si, shortened by being used in a position where it lost its accent). Lith. vemnti 'vehens' for *vesantt (cp. dial, geresny-ji 'better (f.)'); O.C.Sl. vezqsti 'vehens' instead of *vezqti (II § 110 p. 337). Aryan. In Sanskrit, these stems occasionally followed the analogy of stems in -«- -ii- (II § 109 p. 334) : v^M-s 'she-wolf naptt-s 'grand-daughter, daughter. Similarly in O.Pers. we have harauvatis (i. e. -t-s) 'Arachosia' as compared with Avest. haraxwaiti Skr. sdrasvatl. Grreek has lost the forms in -f. Those which actually occur have -(m -la, as cpiQovaa 'ferens' for *(psQovr-iM, noxv-ia 'lady', dXijd-sia 'truth' for *aXad-fa-ia; these I hold to be re- formates following the accusatives in -^av -lav § 216). Remark. J. Schmidt (in Kuhn's Ztschr. XXVII 291 and 309, Pluralb. 59 f.) sees in tlie Greek -ta -m the original ending of the nom. sing, of 1) The strong-grade form of the suffix of which -i- was the weak grade in the Indo-Germanic declension is hard to determine, as I have already said (II § 109 p. 333), adding that -ie- seemed to me the most pro- bable. But in numerous instances -ia- forms are found amongst the cases, — in Greek, Italic, Germanic, Balto-Slavonic, and possibly Keltic (gen. sing, inse, § 230) ; hence it is perhaps more correct to place the variation between -ie- and -ia- in the proethnic period. There is, however, another possibility, -id- as well as -ie- might become -i- in unaccented syllables, 80 that perhaps our I- ie-class is to be split up into two original classes. Then the i-: ia-class would be parallel to the ?-: io-class (Lith. mSdis gen. m^dzio). But in that case we should have to postulate two distinct declensions in -ia-, one varying between -ia- and -i-, the other having invariably -id-. I leave others to investigate these difficult questions more closely. Johansson has tried, but comes to no certain conclusion whatever (Kuhn's Ztschr. XXX 398 ff.). §§ 191,192. Nominative Singular Masculine and Feminine. 69 these stems, and believes that the -i of the other languages arose from a contraction of *-ia-. Against this view, see my Gr. Gr.^ p. 102, and Morph. Unt. V 58 f. Such words as these were often associated with stems in -ia, {aocpia), and -{.a -la gave way to -j^a -ia\ e. g. iraiQa 'com- panion instead of sraiga for *STaQ-i.a^ fem. of sraQo-g, Att. aXrjd^sIa (Ion. aXrjd^slij) beside al^d-fta, and so on. Cp. in O.H.Gr. herzohinna instead of the older herzohin (see below). Latin, too, has lost the -i. In fades pauperie-s, -ie- has spread from the other cases, and -s has been added through assimilation to re-quie-s dies res. avia (from avo-s) is perhaps like mcLter-ia beside mater-ies (cp. II § 109 p. 333, and the footnote on the last page). As regards the fem. sudvis (Skr. svadv-t) see 11 § 109 p. 334 : was sucLvis the result of a form *suamm for *sMac?ttTm (cp. § 216)? Germanic. OJI.G. herzohinna insteeiA oi herzohin, O.Sax. thiwa instead of thiwi, and similar forms were produced by an assimilation to the nominative of i^-stems (as O.H.Gr. sijop(e)a = Goth, sibja 'kin', see § 190 p. 68). Lith. Seme Lett, feme Pruss. semme 0. C. 81. zemlja 'earth' pr. Balt.-Slav. *sem-ie arose by the intrusion of -vS- from the other cases. § 192. 3. All polysyllabic n- and r-stems show the forma- tive suffix in the 3''* or 4*'' (strong) grade, which contrasted the nominative vnth the other cases, and so connected the difference in ablaut with a difference in case; e. g. Gr. notfiTjv 7iaT?jp as against noifiev-a narsQ-a, and so on. "We have to postulate even for the proethnic language pairs of forms, some ending in -en -on and -er -or., others simply in -e -o in both classes of stems. The cause assigned has been the varying conditions of sentence position, -n and -r being supposed to disappear before certain consonants. But this theory is far from certain, in spite of Meringer's arguments (Zeitschr. fiir osterr. Gymn. 1888, p. 137), especially in view of Johansson's new theory of the origin of w-stems and some of those in -r (Bezz. '0 Nominative Singular Masculine and Feminine. § 192. Beitr. XIY 163 ff.) which has been mentioned already on page 63 of this Tolume. a. w-stems. Forms in Idg. -n. Pr.Idg. *Jc(u)uon 'dog'. Armen. sun 'dog' akn (gen. aJcan) 'eye' anj'n (gen. anjin for *-en-os) 'soul' I § 651. 1 p. 497). Gr. xdW 'dog' noip.}]v 'herdsman'. It is a question whether Lat. pecien lien flcimen are old w-nominatives or not; see 11 § 114 p. 352. O.H.G. gomo A.S. -^uma m. 'man, and doubtless Norse Run. Haringa for *-ow; Goth, tug go O.Icel. tunga f. 'tongue' for *-on (Goth, rapjo = Lat. ratio), O.H.G. zunga A.S. turtle f. for *-en (I § 659. 5 p. 513, § 661. 4 p. 519, and II § 115 pp. 361 f.).i) Lith. dial, s^^n (beside szi) 'dog' O.C.Sl. kamy 'stone (I § 92 pp. 86 f., § 663. 1 p. 521, and § 665. 2 p. 524). Forms without -n. Pr.Idg. *Jc(u)uo 'dog'. Skr. ha Mva^) Avest. spa 'dog', O.Pers. xsayarsa 'Xerxes' [xsaya- 'ruler' and arsan- 'mas'). Dubious relics of this kind are seen in Gr. dyM arjdw beside slumv 'image' atjdaiv 'nightingale', and so forth; the genitives belonging to these nominatiyes, slaovg drjdovg etc., would then be ad-formates of the class ATjrm ArjTovi; (G. Meyer, Gr. Gr.2 pp. 315 f.). Lat. homo homo, Umbr. karu 'pars' = Lat. caro, tribrisu i. e. *tri-prikid 'triplicitas' (abl. trihrisine)?) O.Ir. 1) Kluge (Paul's Grundr. I 366) equates O.H.G. -a A.S. -e with pr. Germ. *-0M, and O.H.G. -o A.S. -a with pr. Germ. *-en, admitting at the same time that the phonetics of this are 'strange'. Possibly he was driven to postulate these changes by the aoc. O.H.G. geha A.S. gj'c/e ; for Osthoffs hypothesis that -is- had become -ie- in proethnic Germanic — a hypothesis which offered a possible explanation of these forms {sunt-ia for *-%p-n and hence gebd) — has too slight a foundation to build upon. I hope to settle the question of geha '^iefe in a different way (§ 213), and so I am content with the equation O.H.G. -a A.S. -e = pr. Germ. *-en, O.H.G. -o A.S. -a = pr. Germ. *-on. 2) The accentuation of the Yedio Mva du. suvanau (the texts have svd svdnau with the udattu) is to be restored not only on the authority of Gr. xiiiay, but from the accent of sim-as sun-e etc., tun- being related to suvan- as yuv- to yAvan-. In both words the accent, which in the weak cases fell upon the suffix (orig. *8un-ds like Gr. xw-Se, and *yun-ds) was changed on the analogy of the strong cases. 3) See also the Author, Ber. der sachs. Ges. der "Wiss., 1890, p. 207. §§ 192. Nominative Singular Masculine and Feminine. 71 cu Mod.Cymr. ci 'dog' (for a disyllabic *kuo through the inter- mediate stage of *kui()^ O.Ir. esc-ung eel' air-mitiu 'honour' = Lat. mentio (I § 657. 2 and 3, p. 507); Gall. Frontu Alingu. Goth, guma man' should doubtless be derived from *gume (not *gumo) on account of O.Icel. gume gumi. Lith. szu 'dog' akm'li, 'stone'. In the following words we have re-formates in place of original *-io(n) *-ie(n), the suffix having been levelled down to the weak form of it: Skr. arct (stem arcin- 'beaming') Avest. kaini (stem kainin- 'girl') , Gr. M(fiv (stem dslfpiv- 'bellyfish, dolphin'), Goth, managei (stem managein- 'crowd'). See 11 § 115 pp. 358 ff. In several languages there were re-formations following the analogy of nominatives in -s. Avestic: e. g. ver"pra-jA 'victorious' beside -ja = Skr. vftra-hd (note that an old nom. *-Qh§-s would necessarily have become *-yd}; cp. Bartholomae, Ar. Forsch. I 31, Handbuch § 215 Anm. 2, § 220. Greek: e. g. Lac. a^/Crjg = agarjv 'male', dtl(ptq beside diXcptv (cp. II § 115 p. 359). Lat. sanguis for *sanguins 'instead of sanguen. Osc. liittiuf 'usio, usus' and statif 'statio, statua', for *-ions and *ins according to II § 115 pp. 359 f. O.Ir. cLru 'kidney' doubtless for "^-ons, menme 'mens' for *-ens\ see 11 § 114 pp. 352 f., § 117 pp. 373 f. Similar re-formations of r-stems are described on the next page. b. r-stems. Forms in Idg. -r. Pr. Idg. *mclter 'mother', *d6tor 'giver'. Armen. mair 'mother'. Gr. i.i7]rrjQ, Smtcoq. Lat. mater, soror dator; TJmbr. lu-pater 'Juppiter' af-fertur 'infertor, flamen', Osc. censtur 'censor'. O.Ir. mathir 'mother', siur 'sister* (I § 657. 6 p. 509). In Germanic, with *er-, O.Icel. moder modir, and probably O.H.G. muoter; also Goth, fadar 'father' if Streitberg is right in holding that pr. Germ, -er became -ar in Gothic, as -ei -eyj became -ai -au (cp. § 263 Rem.). ^) The explanation 1) Streitberg, D. germ. Compar. auf -oz-, pp. 22 f. This law would enable us to explain Goth. adv. ^ar as compared with O.H.G. dar^ un- accented der^ as follows. Starting from pr. Germ. *pe>-, we should have 72 Nominative Singular Masculine and Feminine. § 192. of Grotli. svistar A.S. sweostor 'soror , Goth, bropar A.S. hroohr O.H.G. bruadar 'brother' (Gr. rp^armp) is doubtful. These may have come from -or, or perhaps they were accusative or voca- tive forms; cp. II § 122 pp. 381 f. Remark 1. In any case, A.S. modor dohtor O.Swed. /of wr mofor are re-formates. I take this opportunity to call attention to a question which appears to me to need more thorough investigation. How far did Idg. -er- in unaccented final syllables become -ar- ; and where -ar- seems to correspond to Idg. -er-, ought we not sometimes to assume that it came from -or- (or -ar-)'i See the Author, Curt. Stud. IX 374 and 378; Paul in his Beitr. VI 246 f. and 253 f.; J. Schmidt, Pluralb. 197 f . ; Kluge, Paul's Grundr. I 861. Forms without -r. Pr. Idg. *mate, *d6td. Skr. mata Avest. rndta 'mother', Skr. bhrdtd O.Pers. bratcl, 'brother', Skr. data data Avest. data 'giver', Skr. hantd 'murderer' O.Pers. ja(n)ta 'slayer, foe'. Lith. mote mote 'woman, wife' sesu 'sister', 0.C.81. mati 'mother'. Remark 2. Joh. Schmidt and other scholars assume that -r was dropped in Balto-Slavonic (Schmidt, in Kuhn's Ztsohr. XXV 22, Pluralbild. 193 f.). I am still unable to regard this as proved. Cp. I § 663 Rem. pp. 521 f. Re-formation in the separate languages gave rise to Avest. atar-s 'fire' (II § 122 p. 383) , Gr. /udy-aQ-g instead of fidxap 'blessed'. There were similar re-formations in the w-stems, for which see last page. § 193. 4. Polysyllabic s- stems show in the formative suffix the same case-ablaut as do stems in n and r; but it would appear that in the proethnic speech the -s of the formative suffix was never missing. The ending in s-stems will then be -es as opposed to -en -e, er e in the others. Pr. Idg. *dus-menes 'ill-disposed'. Skr. durmanas 'de- jected, troubled', Avest. dusmand 'thinking evil', O.Pers. aspacana (doubtless connected with Skr. cdnas- n. 'pleasure'), Skr. yaids 'glorious' (I § 649. 7 p. 496). Gr. dvn/^svrig 'ill- Goth, par (cp. also jdinar aljar) and O.H.G. der as equivalents. The form par would then have driven the accented *per from the field in Gothic. O.H.G. gen. unser follows jener, § 455. §§ 193,194. Nominative Singular Masculine and Feminine. 73 disposed, hostile' jptv^Tjg 'false, deceitful'. Lat. pUbes puber, de-gener (with -r for -s from the other cases). Pr. Idg. *aiisos 'dawn'. Skr. M|as. Gr. Horn, ^wq Att. cw?. Lat. honos honor (with -r from the other cases). Pr. Idg. comp. *dTc(i)ids 'quicker'. Avest. asyd. Lat. ocior {-r from the other cases). O.Ir. siniu 'older' mao mo 'larger'. For Skr. dsiycis diiyan instead of *dkyas and O.C.Sl. slasdyl 'sweeter' instead of *slad^ja^ see II § 135 p. 430. Pr. Idg. part. perf. act. *uSiduos or *uidues 'knowing'. Avest. vidvd. Gr. dSwc;. By re-formation, Skr. vid-vds vid-vdn instead of *-vds, Lith. da-v^s 'having given' mir-^s 'dead' instead of -*ves *-es and O.C.Sl. da-vu rmr-u instead of *-va *-a (or *-v6 *-vi, *-6 *-i). Along with these the parent language seems to have had a nominative in *-us : Skr. Ved. vid4s Avest. vtdus, with which may be classed Osc. sipus 'sciens' and O.C.Sl. triiru. See n § 136 pp. 439 if. Kemark. It remains doubtful whether the proethnic language had nominatives without s belonging to stems in dental explosives, as well as the above. See § 198 p. 79. § 194. n. Forms with -s as the sign of the Nomin- ative. 1. o-stems. I'r. lig. *ulqo-s 'wolf. Skr. vfka-s, Avest. vehrko (vehrkas-ca 'lupusque'), O.Pers. kara 'people, host' (see I § 556. 3 pp. 411 f., § 558. 4 p. 415, § 646. 3 pp. 490 f., and Bartholomae in Kuhn's Ztschr. XXIX 572 f.). Armen. gail, mard 'homo' = Skr. mftd-s (I § 651 p. 497). Gr. Xvxtrg. Lat. lupus -^ equo-s, vir for *mr(o)-s , satur for *sa- tur(o)-s^ morti-fer and -feru-s (the latter being a re-formate), ager for *agr(o)-s; Umbr. pihaz 'piatus' Ikuvins 'Iguvinus', ager 'ager' katel 'catulus', Osc. hiirz 'hortus' Piimpaiians 'Pompeianus', famel 'famulus'; see I § 655. 5 and 9, pp. 505 and 508 (the conditions of syncopation in Latin have not yet been properly made out, cp. mors for *mort(i)-s and the like, § 195). O.Ir. fer 'man' for *uiro-s, aile 'alius' for *aK(i)o-s (I § 34 p. 34, § 189 Rem. p. 125, § 657.3, 5 and 10, pp. 508 and 509 f.), Gall, tarvo-s 'bull', Andecamulo-S. 74 Nominative Singular Masculine and Feminine. § 194. Goth, vulf-s, vair 'man' for *uir(a)-z^ O.H.G. wolf, acchar 'tilled land' = Goth, akr-s (I § 660.6 p. 516 0, § 661.2 and 5, pp. 517 and 519), cp. also Goth, hva-s O.H.G. hwe-r we-r 'who ?' ; in the Salic Law focla = *fogla-{z) 'bird', Norse Eun. daga-R 'day' = O.Icel. dag-r Goth, dags; Goth, harji-s 'host' for *haria-2, which became *hari-s and took j afresh from the obUque cases (I § 660 Eem. 3 p. 515; Kauffmann, Paul-Braune's Beitr. XII 539; Sti-eitberg ibid. XIV 181). Lith. vilka-s; for the loss of -a- in the last syllable, see I § 664. 2 pp. 522 f. On O.C.Sl. vluku, see below. Stems in -io- sometimes made their nom. sing, ia -is -Is {-i- -%- are weak-grade forms of -io-^, and the corresponding ace. sing. masc. and neut. in -i-m (§§ 212, 227). O.Lat. alts Cornelis beside alius Cornelius, Osc. Pakis 'Pacius'. Goth. un-nuts 'useless' for *-nuti-s, hairdeis 'herdsman'; A.S. sec^ 'man' instead of pr. Germ. *sa^i-3. Lith. zodis 'word' mojis 'sign gaidys 'cock' beside vejas 'wind' sveczias 'guest' and the like. We should doubtless class here O.C.Sl. krajt 'rim, edge', and konji 'horse' instead of *kom, the n having been softened (palatalised) on the analogy of the genitive and other cases. Cp. II § 63 p. 122, and Streitberg, as cited, 166 ff. Remark 1. The student will observe that in Slavonic there is nothing to represent the nominative in *-io-8 (as Lith. vija-s Lat. aliu-s). This ending would regularly become *-ie, which has the look of a vocative ; and this is perhaps the reason for its absence. Cp. § 201 Rem. 2. Remark 2. Perhaps such V-stems as Skr. sarathi-s 'charioteer' Avest. mazdayasni-s 'belonging to the worshippers of Mazda' Lat. deeemjugi-s (II § 93 p. 284) were originally io-stems. O.C.Sl. vluku is an accusative form, which took the place of *vluko. The nom. and ace. in -io-, -i- , and -w-stems 1) Braune (Goth. Gr.' § 78 Anm. 2) has a different theory of the phonetic law affecting Goth. vair. This view has recently received the support of Wl Schulze (Kuhn's Ztsehr. XXIX 271), who explains stiur as a dissyllable. But this explanation is unsupported by the evidence (see Osthoff Paul-Braune's Beitr. XIII 454 f.), and furthermore it is opposed by the form fidvor for *fldvor-COz (§ 320), whilst akrs, which Braune himself holds to be dissyllabic (§ 27), should not have been brought in evidence at all. I therefore keep to may own explanation, as above cited- §§ 194,195. Nominative Singular Masculine and Feminine. 75 came eventually to be the same, *-i-s and *-i-m becoming -?, and *-u-s and *-u-m becoming -m; and this appears to have caused the substitution of vluku for *vluko. There may have been another factor in the change. If the -o (standing doubtless for *-o-d) which we find as' the ending of the nom. ace. neut. of adjectival stems in -o- appeared in this language .before the nom. sing. masc. *-o(s) had given place to -u, the nom. masc. and the nom. ace. neut. must both have come to end in -o; and the wish to keep the two genders distinct may have been an additional reason for substituting -u for -o in the nom. masc. ; cp. § 227. Another explanation of -u, by no means convincing, is given by Kozlovskij in the Archiv fiir slav. Phil. X 657. § 195. 2. «-stems. Pr. Idg. *om«-s 'sheep'. Skr. dvi-s; Avest. azi-s 'snake, dragon', O.Pers. siydti-s 'place of pleasure, dwelling-place' (^ Lat. quies^ gen. quietis). Armen. sirt 'heart' = Lith. szirdl-a, is 'viper' = Skr. dU-s (I § 651 p. 497). Gr. oifi-g 'snake, dragon'. Lat. ovi-s^ turri-s; deer for *cl,cri-s, and by the side of it a re-formate acri-s, mors = Skr. mfti-s 'death' — this syncopation of -i- is common in the final syllables of ii-stems (its conditions have not been fully made out, nor have those of syncope in o-stems, as has been pointed out in § 194, p. 73); Umbr. ^acer 'pacatus, propitius', Osc. cem 'civis' aidil 'aedilis', see I § 33 pp. 33 f., § 638 pp. 472 f. , § 655. 5 and 9, p. 503 and pp. 504 f. O.Ir. faith 'vates' for *uati-s (I § 657. 5 and 10, pp. 508 ff.); Gall, rati-s 'fern = Ir. raith. Goth, anst-s 'favour' haur 'son' for *hur{i)-z (hke vair § 194 p. 78), O.H.G. anst 'favour' chumi 'approach, coming' = Goth. qum-s, OJE.G. wini 'friend', but Fridu-win Lioh-win (I § 660. 1 p. 514, § 661.2 and 5, pp.517 and 519); Norse 'R\m. gasti-it 'guest' = O.Icel. gest-r Goth, gast-s. Lith. nakti-s O.C.Sl. nost^ 'night' (I § 665. 4 p. 525). Observe Skr. »e-s 'bird' beside vi-s Lat. avis, and these root-nouns of corresponding structure — Avest. yao-s 'leagued, confederate, friendly, allied' gao-s 'crying aloud'. And perhaps we should add Lat. ei-s i-s eis-dem beside *-s = Goth, i-s; see § 416. 76 Nominative Singular Masculine and Feminine. §§ 196,197. § 196. 3. M-stems. Pr. Idg. *sunu-s 'son'. Skr. sunii-s; Avest. hazu-s 'arm', O.Pers. kuru-s (read kurus) 'Cyrus'. Armen. zard 'ornament', mark 'death' (II § 105 p. 319), see I § 651 p. 497. Grr. n-fjxv-c, 'lower arm' Tj'JJ-g 'sweet'. Lat. manu-s. O.Ir. hith 'world' (I § 657.5 and 10, pp. 508 ff.); Gall. Esu-s (cp. Esu-nertus). Goth, sunu-s, O.H.G. sunu suno, situ sito 'custom' (= Goth, sidu-s), fridu frido 'peace', without -u or -o Sigi-frid, hand 'hand' (= Goth, handu-s), cp. the /-stems § 195. Lith. sunu-s 'son' sdldii-s 'sweet', 0.C.8I. synu 'son (I § 665. 4 p. 525). In Iranian there are by-forms in -au-s (with corresponding ace. sing, -avam and nom. ace. pi. -Clvas), such as Avest. bdzau-s, O.Pers. dahyciu-s 'neighbourhood'; these we may conjecture to be re-formates containing the loc. sing, in -ciu; see § 261. For Avast, per^nclyu beside per'ncLyu-s and the like, see Bartholomae Ar. Porsch. I 36 and J. Schmidt Pluralbild. 76 ff. § 197. 4. u- uu- and t- ii- stems (cp. 11 § 109 p. 334). Pr. Idg. *bhru-s 'eyebrow' *sue1cru-s 'socrus'. Skr. bhru-s, Svairu-s, tanu-s 'body', dht-s 'thought', nadi-S 'river; Avest. her^zai-di-s (ace. -di-m) 'having great insight'. Gr. otpgv-s, vExv-i; 'corpse', xt-(; 'weevil', n6li-g 'city'. Lat. su-s, vl-s-^ socru-s has become a w-stem, because stems in -u~ and those in -u- had the same endings in the ace. gen. and dat. singular (§§ 217, 233, 254). O.Icel. sij-r O.H.G. su 'sow' (I § 661. 5 p. 519); polysyllables are treated as w-stems, O.H.G. swigar 'socrus' (also swiger following muoter)^ Goth, asilu-qairnu-s f. 'fj-vlog ovmog, millstone' O.H.G. quirn as contrasted with O.C.Sl. Mny f. 'mill'. O.Pol, kry Mod.Slov. kri 'blood' = O.C.Sl. *kry (whose place was taken by kruv-i) , O.C.Sl. svekry (I § 665. 4 p. 525). Nominatives formed in the same way from stems in -^-• -i^n-, -J- -U-, and -f- -rr- (I § 312 pp. 250 f., n § 160 pp. 485 f.). Skr. jd-s 'being' for *g§-s, go-sa-s 'winning cattle', Avest. xd 'spring, source' (cp. ace. Ved. khdm § 217), Skr. pur 'stronghold' for *jo|-s, gtr 'praise' for *gf-s. No doubt Gr. (Lfio-fiQu'tQ, /pcJg are further examples. § 198. Kominatiye Singular Masculine and Feminine. 77 § 198, 5. Stems whose suffix ends in an explosive. a. The Suffixes -t- -tat- -tut-. Skr. vikm-ji-t 'gain- ing everything by victory' sarvd-tat 'completeness' (I § 647. 7 pp. 493 f.), Avest. haurva-tas 'wholeness, a being in good con- dition (I § 473. 2 p. 349). Gr. »rjg 'hired labourer for *i)-}]-T-g, vi'-l 'night' for *vvH-r-g, oXo-rijc; 'wholeness, completeness'. Lat. com-es (gen. com-i-t-is) nox (gen. noc-t-is), novi-tas, juven-tUs. O.Ir. cing (gen. cinged) 'hero, warrior = Gall. *Cinges (stem Cinget-), O.Ir. ditiu 'youth' for *(i)ovetU(s) = Lat. juventUs. In Germanic, such forms as Goth, naht-s 'night' menops 'month' O.H.G. naht mcLnod are re-formates, siace -ts became -ss (s) in proethnic Germanic (I § 527 p. 382). They may have been due to an attempt to restore the stem, which had been preserved in the other cases ; cp. Goth, instr. pi. frijond-am beside tigum, § 379. As regards nominatives hke O.H.G. nefo for *nefd(d), see p. 79. The Suffix -nt-. Fr.ldg. -nt-s, -^i-s'), as Hh^ghont-s, possibly *bhrghent-s (see 11 § 125 Rem. 2 pp. 395 f.) 'prominent high'. Skr. brhdn Avest. ber'sqs (I § 647.7 pp. 493 f.; Bartholomae, Kuhn's Ztschr. XXIX 501 ff. and 517); Ski-. dddat 'giving' for *dadat-s Avest. stavas 'praising' for *sta- vat-s. Gr. asiq 'blowing' = Skr. van, common ground-form *uent-s, odovg 'tooth'; as to -mv in (pspcov and the like, see below, hat. ferens, dens, stans = Gr. tfrd?; Umbr. zeref serse 'sedens' (I § 655. 9 p. 504). O.Ir. care cara 'friend' (gen. carat), cp. tri-cha 'group of 30' (gen. tri-chat) = Avest. pri-sqs (I § 657. 10 pp. 509 f.). Lith. vezqs 'vehens' dial. vezans vesus, Pruss. sldans stdons 'sitting'; 0.C.81. vezy 'vehens', in the first instance for *vezuns, according to I § 84 p. 80, § 92 pp. 86 f., § 219 pp. 186 f. In Sanskrit, iyan 'tantus' Myan 'quantus' are re-formates which followed certain words of kindred meaning, such as tvd- 1) An error must be corrected in II § 125, p. 395. In that place, following the traditional view, I wrongly, allowed myself to regard *-on as an original nominative ending as well as *-o-nts. This correction I have already made in my Greek Grammar' p. 109. Nominative Singular Masculine and Feminine. § 198. 'one who is as thou art' (see below). In Avestic par- ticiples, besides -as (*-ants) and -as {*-ats), we find -o = Ar. *-as, which is the commonest ending of such participles as concern us here ; e. g. per"sd 'asking' histo 'standing'. Bartho- lomae (Kuhn's Ztschr. XXIX 557 ff.) considers this formation in the light of Vedic forms like pra-mpid-s beside pra-mfndn 'destroying' viivam-invd-s penetrating everything' beside invan, and assumes that a certain number of adjectival compounds in Idg. *-6-s, used like participles, were brought into close relation with the corresponding verbs, the result being that true parti- ciples in -nt- took the ending of these adjectives through asso- ciation with them. The analogy seems to have gone further; and, in Avestic, nominatives in -vo were formed even from stems in -vant-, as par^na-vo 'furnished with a feather' (see Geldner, Kuhn's Ztschr. XXX 515). Greek has forms like cpsgcov 'bearing' iJwV 'seeing', besides those in -ou'g (for *-ovr-g). Now -cov can come neither from *-ont-s or *-ont-s nor from *-ont, and for *-dn as an Idg. ending in w^-stems there is no further evidence that can be trusted. i) I therefore offer the following conjecture as to its origin. I suggest that there were two influences at work. (1) The relation of the masc. iS/.i(ov Jitwr, and similar forms, to the neuter in -ov, Xd^ov niov, caused a masc. -wv to spring up in connexion with -ov (for *-ovt), beginning with participles used strictly as nouns, e. g. ^bXImv 'future' sy.mv 'willing'. (2) The relation of the vocative to the nominative in xvwv SaifiMv and the like, voc. kvov Sat/xov, gave rise to substantival nominatives like yspuv (voc. yspov for *-ovr). In Germanic, forms like Goth, frijond-s O.H.G. friunt 'friend' are re-formates of the same kind as Goth. menSps O.H.G. mOnod; see last page. Similarly, Pruss. dilant-s 'working' and Lett. dug'Sbt-s 'growing', for *-ant(i)-s. The Suffix -ijsnt-. Skr. tvd-vqs -van Avest. ptcO-vqs 'one like thee', Skr. dma-uqs -van 'pressing on mightily, powerful' for 1) Lith. sedun and Lett, sedu (J. Schmidt, Kuhn's Ztschr. XXVII 392) cannot be quoted as proving that Idg. had this ending. § 198. Nominative Singular Masculine and Feminine. 7f) *-vans, but Avest. ama-vS, for *-vCLs. Pr. Ar. *-vdns in *tvcL- -vans, we may conjecture, took the place of *-vas under the influence of -vant-am -vant-as; but pr. Ar. *-mSs, which was preserved in Avest. ama-vd, belongs to the suffix -ues- -y,os-. Cp. II § 127 p. 405, § 136 p. 441, § 208. Gr. 6rovn-f,Q 'wailing, lamentable' for *-fsvt-<;. Remark 1. Bartholomae (Kuhn's Ztschr. XXIX 499 ff. , 518 f.) postulates Idg. *-y,ent-s; in which I cannot follow him. It may be remarked in passing that, granted pr. Ar. nom. *-vas, the change of -vani- to -van- stems (Skr. fk-vant- and fk-van- 'singing' Avest. ama-vant- and ama-van-, and so forth) is more easily explained than it is on Bai'tholomae's theory, pp. 540 f. : -vas had another form -vd, its sentence doublet, and this resembled the ending of stems in -van- (§ 391). Cp. the reverse process in Ved. varimai-a from the nom. varima (stem varitnan-) 'width, distance'. Suffixes in -d, pr. Idg. nominative ending -ts. Skr. Sardt 'autumn', stem iarad-. Gr. (pvydc 'fugitive', stem (pvyaS-, danig 'shield', stem damS-. Lat. lapis, stem lapid-, palUs stem palUd-. All these examples ended in pr. Idg. -ts. But we have also certain forms, especially in Germanic, which seem unquestionably to point to a proethnic nominative singular without s. Such are Goth, mena O.H.G. mano, and doubtless Lith. menu for *mendt beside Goth, meno'p-s O.H.G. manod (II § 123 pp. 393 f.), O.H.G. nefo for pr. Germ. *nefdS beside Lat. nepos (see loc. cit.), A.S. hcele for pr. Germ. */alep beside hmled (loc. cit.), O.H.G. san 'tooth' for pr. Germ. *tanp (in I § 527 p. 382 erroneously traced back to pr. Germ. *tan(t)-s) beside Skr. ddn Gr. oSovg. Other examples are given by Kluge, Paul's Grundr. I 390 f. Here there are two possibilities between which I do not feel able to decide at present. There may have been double forms from the very first, one with s and one without; this view may be supported by the ablaut in *xalep (in consideration of this, Kluge op. cit. p. 385 even postulates an Idg. nom. *pod beside ace. *pod-iri). Or s may have disappeared when the words were used in this or that environment in the sentence (cp. the disappearance of s in such sound-groups as st-, I § 589. 3 pp. 445 f., § 645 p. 490). And compare Bartholomae, Stud, zur idg. Sprachgeschichte, I 65. 30 Kominative Singular Masculine and Feminine. §§ 198,199. Remark 2. If -m in O.H.G. hwemu is the direct and regular representative of pr. Germ. *-ot = Idg. *-od (§ 241), the -o of mano nefo must have arisen by an assimilation of these words to the nominative ending of M-stems, such as gomo. b. Suffixes in -h and -g. Skr. uMk- (stem utij-) 'de- manding', Avest. usixs (stem usij-) a kind of demon; for Skr. spat 'spy and the like see I § 401 Kem. 2 p. 297, § 404 Rem. 3 p. 299. Gr. /ueTgu^ (stem laiQun-) 'girl' (fdXuy^ (stem rpaXayy-) 'phalanx'. Lat. senex (stem senec-) hibcix (stem hibdc-). O.Ir. aire 'princeps' for *ariak-s^ gen. airech, ail 'rock, stone' for *alek-s, gen. ailech, nathir 'water-snake' gen. nathrach; Q-all. esox = Mid.Ir. ew (gen. iach) 'salmon' (I § 657. 10 pp. 509 f.). § 199. 6. Perhaps all Root-Nouns had -s in the pro- ethnic language (cp. § 197). Examples: Pr. Idg. *ncl,u-s 'navis': Skr. nUii-s Gv. vaU-g. *d(i)ieu-s 'heaven, daylight': Skr. d(i)ya,ii,-s , Gr. Zsv-g, O.H.G. Zio for *t(i)eti(z) (Streitberg, Die germ. Comp. auf -02-^ 18). ^go'if-s 'ox, cow': Skr. gaii-s Avest. gau-s, Gr. (iov-g, O.Ir. bo (I § 657. 10 pp. 509 f.). I leave it an open question whether Gr. Zi]c, Lat. dies and Gr. Dor. ^c5g, Lat. bos O.H.G. kuo O.Sax. ko ') were framed on the model of the ace. sing. (§ 221), or whether they represent proethnic sentence doublets *d(i)ies and *gos. See II § 160 p. 481 f., and Streitberg op. cit. 12. In com- position we have Skr. -gu-s., as su-gii-s 'having fine cattle', in- flected as a M-stem, e. g. nom. pi. su-gdvas du. su-gu. *uoq-s 'voice, speech': Skr. vdk Avest. vcl>x-s, Gr. 01//, Lat. vox. ^reU-s (v^reg-) 'ruler, king': Skr. rat (like spat § 198), Lat. rex O.Ir. n (gen. rig, Gall. nom. Dumno-rtx); Goth, reik-s (nom. pi. reik-s) instead of *reihs (I § 527 p. 381) is doubtless borrowed from the Keltic. Avest. bar^s 'height, high' (gen. bar^z-o her^z-o) , O.Ir. br% (gen. breg) 'mountain', Goth, baiirg-s "stronghold, fort, town' (gen. baurg-s) instead of the regular *baurhs, from y^bhergh-. Skr. dpat9 Avest. apqs 1) A.S. cu O.Icel. hyr must be added to this list, if in these u stands for 110. But cp. II § 160 p. 482. §§ 199,200. Vocative Singular Masculine and Feminine. 81 for pr. Ar. *apaKk-s "turned backwards' (I § 647. 7 pp. 493 f. ; Bartholomae, Kuhn's Ztschr. XXIX 501 ff. und 517 f.). Avest. af-s (stem ap-) 'water', Skr. stup (stem stuhh-) 'roaring . We may doubtless add *mUs-s 'mouse', which became *mas in the proethnic stage (§ 356 Eem.): Gr. i.iv(; Lat. mUs (II § 160 p. 485). The Greek ;(f^wV 'earth' (cp. Skr. ksd-s, II § 160 p. 482) may be an ad-formate of XQvyav and the like; /aj/ /t/v 'goose' and i.irjv 'month' are undoubtedly re-formates, taking the place of *x(ivg "xaq and Ion. etc. /.isiq for *n7ivg respectively (n § 132 p. 415, § 160 p. 485); similarly (pwg 'thief xw 'hedgehog' on the analogy of Swthoq dorrjQ etc. For the ground-form of Skr. pat Gr. Dor. nwc, noc, (Hom. rpi-nog) Att. novg Lat. pes 'foot', see § 198 p. 79 : were there once parallel forms *pdd {*ped) and *pot-s (*pet-s)? ov in novc; has not yet been satisfactorily explained. Vocative Singular Masculine and Feminine}) § 200. No special vocative forms are found in the Indo- Germanic languages except for singular nouns masculine or feminine. Prom the proethnic period onwards, the nominative has served for the vocative in the plural and dual, whilst in the neuter gender the form used for nominative and accusative has been used for the vocative in all three numbers. Genuine singular vocatives naturally enough had no case- sign at all; see § 184 p. 56. In the proethnic language the accent was thrown back to the first syllable of the word, as *mdter 'mother' = Skr. matar Gr. fifJTiQ ; this remains a general rule in Sanskrit, but in no other language. But the forms had a word accent of their own only when they stood first in a clause. In any other position it is probable that they were often enclitic, which is the rule in Sanskrit; e. g. 1) Benfey, tjber die Entstehung des indogerm. Vocatlva, Abhandl. der Ges. der Wlss. zu Gott. XVII (1872) pp. 3fF. Bezzenberger, Zur lett. Declination: Binige Vocatlvformen ; in his Beltrage, XV, 296 ff. Bru^mann, Elements. III. 6 82 Vocative Singular Masculine and Feminine. § 200. iddm indra s^nuhi 'Hear this, Indra !' See I § 669 p. 534, and § 672 p. 538. In all other branches of the language but the Aryan this practice of accenting the first syllable underwent many changes. Sometimes it was overborne by special rules in special languages. Thus in Greek and Latin certain changes were necessary in order that words should conform to the trisyllabic law; hence Grr. 'Ayd/ice/in'ov instead of *"Aya,us/iivov, Lat. aliimne, amplissume instead of *dlumne, dtnpUssmne (I § 676 p. 541, § 681 p. 548). Or the accent followed that of other cases from the same stem ; thus Gr. Sdiq)Qov instead of *ddt(pgov follows 6di(f,QMv (intelligent') datfpgoi'og etc., avToy-gdzoo instead of ^avTOXQaroo follows dvxo- xgdrtog ('having unlimited power') avroKQarogoi; etc., dloysvsg instead of *dto'y£)'f c follows dloy&vtjg ('born of a god') Sioysvsog etc. Elsewhere other factors less easy to detect may have been at work, as in the accentuation of the Lithuanian vocative — e. g. vilM ('wolf'), naktS (night'). i) But even in the singular the parent language would seem to have sometimes used the nominative form as a mode of ad- dress: compare, for example, Skr. Yed. (Rig-V. I. 2. 5), vayav indras ca cetaihah 'Vayu and Indra, ye take care', Gr. Horn. (J" 276) Zsv nuTsp . . . ' HiXwc ts . ., v/xsTg ixdgTVQol sare. And in most languages the forms of the nominative usurped more and more the place of the vocative; sometimes the proethnic vocative form belonging to some class of stems died out completely before the date of the oldest extant specimens of a given language. This happened in Latin to the vocative of S-stems. The genuine vocative forms are most faithfully pre- served in Sanskrit. Yet even there in certain monosyllabic stems the vocative was regularly expressed by the nominative form, although accented as a vocative would be ; e. g. diyclu-s, written dyaus (nom. diya,u-s),^) as contrasted with Gr. Zkv ; bhu-s ("earth'), 1) Bezzenberger's conjectures given in the essay cited in the footnote ■on the last page seem to me highly uncertain. 2) For this accentuation, see Bartholomae, Stud, zur idg. Sprachg., I 82 f. §§ 200,201. Vocative Singular Masculine and Feminine. 83 but Gr. i/9v ('fish'). Perhaps we may follow Collitz (Bezz. Beitr. X 32) in recognising the Idg. vocative of gdi'i-s in the Toc. -go, only found in composition (e. g. hhuri-go). Remark. In Sanskrit, the rules regulating the accent of the vocative singular held good for plural and dual nominative forms when these were used as vocatives (cp. the sing. dit/Sii-s just cited); e. g. p/taras (nom. pitdras 'fathers'). There is no reason why this should not be regarded as a genuine proethnic tradition, although it is true that no such practice can he proved for any European language: in Attic m nazs^e;, for example, might have been expected, since we have m nareq. § 201. 1. o-stems. Pr. Idg. *ulqe. Skr. vfka; Avest. vehrka, O.Pers. martiyd 'homo' (I § 649.1 p. 495). Gr. Xvy.s, aSslfpi beside aJsXqto-c 'brother', Suifiovit 'wonderful one'. Lat. lupe puere, flUe and fiU from filiii-s (cp. below) ; Umbr. Tefre, Fisovie. O.Ir. fir for *uire, maicc 'son' for *makue, cell 'com- rade' doubtless for *ceKe (I § 657. 8 p. 500). Goth, vulf, hairdi 'herdsman , O.H.G. wolf. Lith. vilkh, sodi {godi-s 'word') gaidy {gaidy-s 'cock'), cp. below; O.C.Sl. vluce. Remark 1. As regards -a instead of -ain Yedio, as wsaJM 'bull', see Lanman, Noun-Inflection p. 339, Oldenberg, Die Hymnen des Rigveda, I 393 ff., Wackernagel, Das Dehnungsgesetz der grieoh. Compp. (Basel 1889) pp. 12 f., Bezzenberger in his Beitrage XV 296 f. It cannot be proved that in the Latin vocatives from io- stems -% is contracted from -ie. Probably we have here the weak-grade -i-, as we certainly have in Lith. voc. gaidy and in the Italic nominatives in -is -i-s. Cp. II § 63 p. 122, m § 194 p. 74, and Streitberg, Paul-Braune's Beitr. XIY 201. In Lithuanian the ending -ai is also found. This occurs in names of men, as tevai 'father' Jonai 'Johannes' (cp. Bezzen- berger in his Beitrage, XV 299). Can it be that -i is the same particle as we see in pronominal nominatives in -oi (§ 414)? Avest. voc. haene: nom. pwoi (§ 202) points to this conclusion. Briickner (Archiv fiir slav. Phil. Ill 276) compares the em- phasising -ai in tas-dl toks-ai grasus-ai, and the Uke. For O.C.Sl. jumce beside nom. ^'mwct 'young bull*, etc., see I § 147 p. 184. ^o-stems whose nominative did not end in -c% 6* 84 Vocative Singular Masculine and Feminine. §§ 201,202. -z\ had the ending of w-stems in the vocatiTe (§ 203), as Tcraju (nom. krafi 'border') mqm (nom. mq^ 'man'). Remark 2. The following may be suggested as a conceivable reason for the latter change of inflexion. Nominatives such as hraji honji are parallel to the Lith. moji-s kodi-s and to the Lat. ali-s. Can there have been nominatives in *-ie = Idg. *-io-s in O.C.Sl., corresponding to Lithuanian nominatives like vija-s ('wind'} svSczia-s ('guest'), and to alius and the like in Latin ? Then the vocative in *-ie will have been transformed in order to avoid confusion with the nominative which had the same ending, while this nominative afterwards took the ending of that class of nouns whose nominative ended in -(jji-s. Cp. § 194 p. 74. § 202. 2. S-stems. Pr. Idg. *e^?ta, cp. 11 § 59 pp. 108 f. Skr. dmha 'mother' doubtless belongs here; for the Ar. voc. in *-ai (Skr. -e Avest. -e) see below. Gr. Horn, vvfirpa 'nymph'; -« is more commonly kept in masculine words, as Sionora 'master' av^wra 'swineherd'. O.C.Sl. rqko. And probably we must place in this class Lith. ranka, and with -a dropped, motyn from nom. mdtyna 'mother', Mdriuk from nom. Mariukct, and the like; see I § 664 p. 522, and J. Schmidt, Kuhn's Ztschr. XXYII 382. In Aryan the common ending was *-ai : Skr. dive Avest. ha^ne. The origin of this ending is imcertain. I think it most probable that a deictic particle has attached itself to this case, — the same deictic -i which is found in the nom. sing, in -ai, Avest. J5m>o? her'^xSe Pruss. s