Sk. pronominal-stem e-na- (this), a stem in -na- from demonstra- § 109. tivo-pronominal √i.

2. Indo-European original-language dua- or dva-; Sanskrit dva-; Greek dvo-=dva-; Latin duo-dvo- (duō-bus like ambō-bus is probably caused by the nom. dual duō, ambō, f.f. dvā, ambhā).

3. Indo-European original-language tri-; -i- is a suffix and tar, tra, the root (cf. the ordinal); tr-i- is therefore probably shortened from *tar-i-, or perhaps from *tra-i- by loss of the a; the assumption that tri- is an archaic weakening from tra- is, I think, less likely; Sanskrit tri-, the fem. having the stem -ti-sar-, in which Bopp conjectures a reduplication, and which he imagines to have arisen from *ti-tar-; Greek Tp-; Latin tri-.

τρι-;

4. Indo-European original-language katvar-; Sanskrit katvár-, shortened katúr-, fem. katasár- (clearly formed after the analogy of 3, according to Bopp compounded with it, in which case kaunum'); Greek TÉTTAρ-, TÉσσαρ-, for *TEτFαρ-, KEτFap-, Dôr. TÉTOρ- for *TETFOρ-, with 7=original k (§ 62, 1), Boiôt. πéτтаρ-, Hom. and Aiol. πíσup-, f.f. katur-, π=к (§ 62, 1), and σ before v=T, as in σú for Tú; Latin quatuor-, the best authenticated spelling quattuor is unsupported etymologically (inscriptions have also quattor, quator).

5. Indo-European original-language kankan-, an obviously reduplicated form; Sanskrit páńkan-, with p=k (§ 52, 1); Greek πévτе, Aiol. πéμñe, π and τ=k (§ 62, 1); Latin quinque.

6. Indo-European original-language. By combining the Zend khšvas on the one side with the Greek, Latin and Gothic form with final ks on the other, the resulting f.f. would be ksvaks for Indo-European (Ebel, Beitr. iii. 270; Zeitschr. xiv. 259 sqq.), which likewise seems to be reduplicated, perhaps original *ksva-ksva-; Sanskrit šaš, probably immediately from *kšaks for *ksa-ks (§ 55, 2); Greek and Latin with dissimila

§ 109. tion of the initial sound from a f.f. *svaks for *ksvaks, Gk. e, Dôr. Féğ, like Lat. sex, both therefore from *sveks (Leo Meyer, Zeitschr. ix. p. 432 sqq.; cf. Lat. se for *sve, etc.).

7. Indo-European original-language probably saptan-; Sanskrit saptán-, later sáptan-; Greek éπTά, i.e. *saptan(a=an § 19); Latin septem, i.e. *septim. Bopp supposes that the m of septe-m has worked its way in from the ordinal septi-mo-, because it would be unlikely that n should change to m; possibly however an otherwise unusual sound-change is found in it, and we should not cut the Latin numeral adrift from the Erse and the Greek.

8. Indo-European original-language. Stem aktu-; Sanskrit ašṭán-, later ášṭan-, probably after the analogy of 7 and 9, and asṭú- (the latter in the nom. acc. ašțáu, probably shortened from *aktáv-as or *aktáv-as, apparently a dual-form, as also in Gk. and Lat.); Greek OKTÓ, Latin octo, dual-forms arising just like Sk. ašṭáu through loss of the termination, in which respect the notion of 4+4 may have assisted; in öydoF-o-, octău-o-, the f.f. of the stem aktu- is unmistakable; it occurs also in Goth. and Lith.

9. Indo-European original-language navan-; Sanskrit návan-; Greek evvéa, i.e. *veFa(v) with e prefixed, and unoriginal doubling of initial consonant v; Latin nouem (on the m cf. 7).

10. Indo-European original-language dakan-; the conjecture that dakan stands for *dva-kan- (kan for kan-kan), i.e. 2×5, is not proved, but it is too tempting to be passed by; Greek déxa, i.e. *Sexav; Latin decem, i.e. *decim (cf. 7). $110. The numerals 11-19. They were formed by joining the number 10 to the units; in some languages we clearly see a mere addition.

Indo-European original-language. Originally the two words would probably exist separately, e.g. perhaps 12 dua dakan; 13, tray-as dakan, etc.

Sanskrit. 11, ékā-daçan-, with lengthening of final a of § 110. stem éka- (one); 12, dvá-daçan-, dva probably must be considered a dual; 13, tráyō-daçan-, later trayó-daçan-, n. pl. trayas+daçan-; 14, kátur-daçan-; 15, páńka-daçan-; 16, šố-ḍaçan-; 17, sápta-daçan; 18, ašṭá-daçan-, with dual-form aštá; 19, náva-daçan-.

Greek. 11, ev-deка; 12, Sά-Sexα; from 13 the words originally separate are merely joined together, e.g. in 7pio-Kaí-deka, τρις must be taken as a shorter form for τρεῖς ; 14, τεσσαρες-και-δεκα, etc.

Latin. un-decim for *uni-decim; 12, duo-decim; 13, trě-decim, perhaps with older stem-form tra-, cf. ter-tius, or else tre- is a shortening from trēs; 14, quatuor-decim, etc.

The numerals 20-90 (the intermediate numerals do not need § 111. discussion here; they are clear in all languages, and mostly quite uncompounded). In Aryan and South-European 20-90 were expressed by means of units compounded with a substantive formed from daka-, mostly shortened or otherwise altered. Whereas in the North-European division units and tens are separate words, at most joined together. The contrast between the two closely-related branches of the speech-stem—Aryan and Graeco-italo-keltic-and the Sclavo-teutonic is here clearly shown (cf. Introduction, IV.). It is scarcely to be supposed that in the original-language composition had already taken place; but the tens and units would still be separate words.

Sanskrit. The tens were originally expressed by daça-ti-, daça-ta, the units being prefixed. Of daça-ti, however, -çatiand even -ti- only remains, of daça-ta- only -çat-; so strong is the tendency to diminution in words so much used.

20, vi-çáti- for *dvi-daçati-, the nasal of vi dvi and the corresponding nasal of the two following numerals is obscure; perhaps we may detect in it the remains of a case-ending. It is shown by the Zend to be a late formation. 30, trî-çát-, probably for *trīni daçatā, whence probably the nasal arose

§ 111. which worked its way into 20 and 40 by analogy; 40, katvārī-çát-; 50, pankā-çát-; 60, šaš-ți; 70, sapta-ti-; 80, açi-ti-, a form widely differing from the original; 90, nava-tí-, all with -ti- for *daçati.

Greek. Except in 20, where likewise a form daka-ti appears, -κоvтa appears as the second part of the compound, probably a neut. pl., f.f. dakan-tā, from a sg. *dakan-ta-m. 20, El-KOσI=*El-KOTI (§ 68, 1, c), Hom. eíxoot, earliest form Dôr. FeiKATI, FIKATI; FikaтI stands for *dvī-daka-ti, the length of i may have its origin in an earlier case-ending, whence may come also ei; čeíkoσi, i.e. ẻ-Felkoσ, with the frequent vowelprefix before consonantal beginning (§ 29, 2). 30, Tρiá-кOVтa, i.e. *triā-dakan-tā, similarly with the following; 40, teσoapá-κοντα ; 50, πεντήκοντα; 60, ἑξήκοντα ; '70, ἑβδομήκοντα, formed with the ordinal, like 80, ỏydоń-коνтα, and probably also 90, evevý-KOVтα, Hom. also evvý-Kovтα, which we should accordingly take for a shortening of ἐνενήκοντα ; ἐνενήκοντα= Lat. nōnā-ginta; the ordinal *ẻve-vo- from *ẻ-veFa-vo-, it is true, presupposes a strong shortening; the suffix -vo-, as in Lat. -no-no- (v. Ordinals), we must treat as having arisen from -μothrough assimilation to the initial sound.

Latin. Except -gin-ti in 20, -gin-ta generally appears, probably a neut. pl.; -gin-ti and -gin-ta stand for *degin-ti, *degin-ta, and these for *decen-ti, *decen-ta, the f.f. of these stems is dakan-ti-, dakan-ta-; c has here become g, even as uicēsimus from a form *uicenti has stood its ground beside the unoriginal uīgēsimus. Accordingly 20, vi-ginti from *dvi-decin-i (cf. Gk.; on the form, which seems to be a neut. dual, cf. Corssen, Krit. Nachtr. p. 96 sqq.); tri-gin-ta=*triā decintā, *triā becoming *trie, and this becoming tri, like s-yā-t, siēt, sit (3 sg. opt., √es); 40, quadrā-gin-ta, with softening of t to d, for *quatuoră decintă; 50, quinquã-gin-ta; 60, sexā-gin-ta; 70, septua-ginta from a stem septuo-, which does not appear elsewhere; 80, octō-ginta; 90, nōnā-ginta, from the ordinal, cf. the

Greek, with which the Latin essentially coincides in these for- §111.

mations.

Numerals 100-1000.

100. Indo-European original-language. Probably stem kan-ta- neutr., acc. n. sg. kan-ta-m, a shortening of *dakan-dakan-ta-, i.e. dakan-+ subst. dakan-ta- which forms the tens; we found it already shortened to kan-ta- (Zend -çata-, Greek -KOVтa, Latin -ginta). (dakan-da) kanta- thus means *ten-ty, *Sekńkovтa, *decāginta. The n of kan-ta is kept in Lat., Kelt.,· *δεκήκοντα, Lith. and Goth., but is lost elsewhere.

Sanskrit çáta-; Greek e-Kaтó-, é can be nothing else than a diminished form of ev- (one); Latin cento-.

200-900. Originally expressed in two words. Sanskrit by two words, or by ordinary composition (e.g. dvě çatë or dviçata- neut.); Greek from stem κато- Oг Kото-, f.f. ka(n)ta-, cf. Sk. çata-, was formed a derivative in -ya-, before which according to rule (§ 89) the stem-termination is lost, thus *-KATIO-, *-KOTIO-, f.f. *-kat-ya-; in Dôr. -κaтIO- remains unchanged, whereas elsewhere the *-κотIO- passes regularly (§ 68, 1, c) into -κоσo-. So from a hypothetical *τριᾶ-κατά οι -κοτᾶ, f.f. triä kantā 300, arose Dôric Tрiα-κáтIо-, Attic Tρlā-Kóσ-, these forms being used as adjs. and in the pl.

The Latin proceeds in a way similar to the Greek, employing as it does the stem cento- as the second member of the compound adjectivally in the plural, e.g. 200, du-cento-, du shortened from duo; 300, tre-cento-; 500, quin-gento- for *quinc-cento-, with softening of c to g after n, as in 400, 700, 800, 900 (cf. -ginta); 600, sex-cento-; 900, nōn-gento-, from ordinal-stem nōno-. The numerals 400, quadringento-; 700, septin-gento-; 800, octin-gento-, show an analogy, which perhaps may be traced to septin-genti; septin-, f.f. saptan-; octin- also corresponds to f.f. aktan-, cf. Sanskrit aštan-, and so also a stem quadrin- has arisen from quadro-, shortened from quatuor (cf. quadra-ginta). Pott (Zählmethode, p. 149) conjectures

§ 112.