- Some scholars may perhaps doubt the existence of prepositions and preverbs for Indo-European. I cannot follow this opinion. While I certainly would deny it for Proto-Indo- European, I admit both prepositions and preverbs—altho to a very limited extent—for what we usually call “Indo-European,” that is, the last period of Indo-European unity. Compound verbs were then limited to the dependent clauses, as in Modern German, which has kept the ancient distribution (see on this subject my doctoral dissertation, published in AGlIt., 24, 1930, pp. 14 ff.). A verb like Latin pōnō (from *po-sinō), which is certainly very old, seems to confirm that Indo-European had compound verbs, at least in certain cases (note both the preverb po-, which is not found as a preposition in Latin, and the syncope in pōnō, posuī, -postus; cfr. Leumann, Gramm., p. 92).
- We therefore exclude from this study the “improper” Latin and Greek prepositions, that is, those that never appear as preverbs; they are all of (relatively) late origin, formed usually from nouns, and probably most of them do not go back to Indo-European. At least four Latin “improper” prepositions, however, might be rather old: they are uersus (uersum, uorsum, aduersus), which reappears in Old Irish frith, fri (with the accus.), Middle Welsh wrth, Tocharian B wrattsai ‘towards’ and Germ, -wärts, Engl, -ward (cfr. A. Sommerfelt, De en italo-celtique2, Kristiania, 1920, p. 21; Ernout-Meillet s.u., Leumann-Hofmann, p. 518); cis, Old Irish cen- ‘cis-’, cen ‘without’ (see also on this Wackernagel, Vorles.2, 2, p. 234); sine, Old Irish sain, Toch A sne B snai, Gr. ἄνις (perhaps, but less close, Gr. ἄτερ, O.H.G. suntar etc.); and secus, Irish sech (also with the accus., like Latin), Breton hep, Lettish sec, secen (also with accus.), Avestan hača, Old Pers. hačā, Vedic sáčā, sākám.
- The other Latin “improper” prepositions are erga, retrō, iuxtä, citrä, ultrā, conträ, penes, circum (circō circā circiter), dextrā, sinisterā, secundum, absque, cōram, palam, simul, fīnī, tenus, subter, subtus. To almost all these the norm of the “area sparita” can be applied (see Ernout-Meillet s.u.-, Hofmann Gramm.). Apud, clam, extrā, intrā, prope, propter, suprā and praeter (the latter is also used as a preverb, but probably late, Hofm., Gr., p. 516: praeter-mittō praeter-eō), altho also “improper,” might perhaps be old. For cis and uls see the text at the end. Some of these prepositions, as cōram, circō, circā, circum, circiter, penes, fīnī, tenus, secundum (not to speak of dextrā and sinisterā, which are certainly late) are of nominal origin, which is an added indication of their relatively recent date.
- Note that Latin cis-alpīnus is translated with cen-alpande in Irish.
- I cannot understand why Wackernagel writes that Latin has dē IN COMMON with Greek (like ab ante ex in per prae s-ub s-uper).
- But right in the following clause of the same sentence he adds: “obwohl es im ganzen schon in seinen ältesten phasen eine viel weiter entwickelte sprache ist als das griechische,” which is entirely wrong. It is interesting to observe that Wackernagel, when he dealt with the specific, concrete problem of the preposition, with the material under his hands, saw the situation correctly; whereas when he was confronted with the general problem of the respective “archaism” of Latin and Greek, he followed the usual prejudice of his time (and in great part, unfortunately, still of our own).
- See also, on all these and the following prepositions, the well-known etymological dictionaries of Berneker, Walde, Trautmann, Boisacq, Walde-Pokorny, Feist, Ernout-Meillet, Kluge, Pokorny and J. B. Hofmann.
- Greek ὠρoμαι is very doubtful; cfr. Walde-Pokorny, 1, p. 96; 2, 349; Boisacq s.u. It is an ablaut form of Sanskrit rauti, ruváti, ravati, according to Boisacq, and contains (like ὠρūϒή, cfr. Latin rūgiō) the interjection ὤ according to Kretschmer, KZ. 38, 1905, p. 135: “Perssons Vermutung (Wurzelerweiterung s. 243) dass das ὠ- von ὠρύoμαι mit diesem ὀ- oder ἐ- [of ἐρυ;γóννα, ὀρυμαγδός etc.] ablaute und dem skr.-präfix ä-entsprede, ist nicht überzeugend, weil nur das Griechische vor dem r- einen vokal aufweist, und die kürze daher am besten als die im Griechischen übliche vokal prothese vor anlautendem r- (ἐρ υ ϑρóς u.s.w.) aufgefasst wird. […] Da das verbum gerade in den ältesten belegen (Herodot […] Pind. […] Plat.[…]) von menschen gebraucht wird, so liegt es nahe, in ὠ- die interjektion ὤ zu sehen: dann ist also ὠρὺoμαι aus ὢ ρúoμαι entstanden; vgl. etwa deutsch wehklagen.” In the ἐ- of ἐαγαγoν of the inscription cited by Brugmann, Grundr.2, 2,2, p. 997 I can see (contrary to Brugmann) only the augment.
- The Indo-European preposition *ni (Vedic ní, Avest. ni, Old Pers. niy-, Armenian ni-st'sit!; cfr. also OHG. nidar, German nieder, Old Bulg. nizŭ, nicǐ) is very probably contained in Latin nīdus, from *ni-zd-o-s (*ni and the root *sed- ‘to sit’). The word nīdus is attested (altho sometimes in strange forms) in Latin, Celtic, Germanic (Engl. Germ, nest), Baltic, Slavic, Armenian, Indo-Aryan: a perfect “horseshoe area” (which we may limit to “lateral areas”: Latin: Arm. Indo-Ar.) with respect to Greek, where the word does not appear (nīdus has also in its favor the norms of the “area isolata” and the “area maggiore”). Latin has therefore at least a trace of the Indo-European preposition * ni, which Greek has lost completely (Armenian, which is always very close to Greek, has *ni only in one or perhaps two verbs; see Brugmann, Grundriss2, 2, 2, p. 861).
- The dogmatic assertion of Sturtevant, Grammar, p. 66: “the conclusion is inevitable that the Akkadian distinction between K, Q, and G, T and D, P and B did not exist in Hittite” has been contradicted by Pedersen, Hittitisch und die anderen indoeuropäischen sprachen (Copenhagen, 1938) pp. 9 f.: “Und man hat längst (!) gesehen, dass diese relativ feste orthographie [as e.g. pí-e-da-an with da for the word ‘place,’ Gr. πἐδoν] in so vielen fällen mit der etymologie stimmt, dass der gedanken nicht ohne weiteres abzulehnen ist, es könnte sich um eine historische orthographie handeln”. It would be necessary to investigate whether the Indo-European aspirates (bh, dh, gh) are expressed usually by b, d, g or p, t, h in Hittite. From Sturtevant's grammar it seems that b, d, g are frequent in this case.
- Strangely, in Language 6, 1930, p. 27 Sturtevant admitted a prefix *pe- in Hittite in the verb pi-his (*pe + *bheid-, cfr. Latin findō, Vedic bhinádmi). In his Hittite Grammar (1933) he does not mention this etymology, probably because he now has (quite rightly) rejected his previous theory of Hitt. h- < Indo-Eur. *bh.—Walde-Hofmann and Ernout-Meillet admit a Cun. Hitt. pa-.
- Cfr. Thurneysen, A Grammar of Old Irish, Dublin, 1946, p. 521; Walde-Hofmann. On Latin endo see (besides Ernout-Meillet, Walde-Pokorny) the grammar of LeumannHofmann, p. 536.
- The interpretation of trans as a nom. sing, of a present participle of (in)-trāre (like uersus aduersus), altho attractive at first sight, is not certain.—trāns is certainly somehow connected with Old Ir. tar (constructed with the accusative, like trāns), Welsh trwy, traws, English through (German durch, Goth, pairh etc.), Vedic tirás, Avest. tarō (also with accus.); it is identical in form with Umbrian traf (trahaf), tra (traha); cfr. Ernout-Meillet s.u. trāns and Wackernagel, Vorles.2, 2, p. 159: “Bei trans, zu dem eine griechische entsprechung fehlt, war nach ausweis des sanskrit nicht bloss die konstruktion mit dem akkusativ, sondern auch ein kleiner ansatz zur verbindung mit einem verbum wohl schon ererbt”; cfr. also Leumann-Hofmann, p. 519: “Sowohl die verwendung als praeverb (vgl. umbr. trahvorfi ‘transuerse’) wie als praep. mit dem akk. ist ererbt […].”
- Wackernagel, Vorles.2, 2, pp. 157 f. calls ἄϰρι, μέϰρι, μέσφα, μέστε, ἄνεν, ἔνεκα, ἔκητι,ἄμα and δε “improper” prepositions, because they are never found as preverbs. We exclude them therefore from this study.
- μέϰρι (which cannot well be separated from μέτα, μέσφα, μέστε ←,ἄϰρι) is perhaps connected with arm. merj ‘near to’ (from *mejr), cfr. Boisacq s.u.; it is then, in any case, a recent innovation of a central zone. Both Greek and Armenian are strongly innovative languages, and have many innovations in common.—μέϰρι is Eastern Greek, ἄϰρι Western (Günther, IF. 20, 1906–1907, pp. 18 f.).
- μέϰρι and ἄϰρι were anyhow originally conjunctions, only later used as prepositions; cfr. Wackernagel, Vorles.2, 2, p. 165.
- On Latin ambi-, am see Wackernagel, Vöries.2, 2, pp. 160 f.; Hofmann, Lat. gramm., pp. 514 f., where it can be seen clearly as an instance of the “fase sparita” (it is replaced by the innovation circurm); see the text at the end.
- It is interesting to observe that an- (Green ἄνε etc.), which is very weak in Latin, is weak in Celtic too, whereas it is very well represented in Germanic and Greek and in all the languages to the East of these two, except the most distant ones, Tocharian and perhaps Indo-Aryan. It looks like a “Central” Indo-European innovation.
- Latin post(e) has no exact correspondent in Greek (Greek TTOS is merely the exact Arcado-Cyprian semantic equivalent of Dorian πóτι Attic πρτς, and never means ‘after’); it is on the other hand very closely connected with Osco-Umbrian post-, Lithuanian pãskui, paskut, Old Bulgarian po (cfr. poz-dě!), Alb. pas (mbas), Arm. ∂st, Tocharian B pesṯ(a), pasṯ(ä) ‘after this’, (om-)posṯ(-am) (= post-e¯, cfr. MSL., 18, p. 417), Old Persian pas¯, Avestan paskā, pasčaṯ, Sanskrit paśčt (and paśč), all of them prepositions meaning both ‘behind’ (local) and ‘after’ (temporal). We have the figure of the lateral areas, very near in this case to the horseshoe area (besides the isolated area, of course). It remains, however, very doubtful whether post really was used as a preverb in Indo-European: this use is rare in Latin (postpōne, posthabeō) and outside of this language appears only in one verb in Armenian (cfr. Meillet, MSL., 16, p. 127: ∂st-gtanem ‘I blame’, καταγιγνὠσκω Cfr. Wackernagel, Vorles.2, 2, 159; 211; 215; Walde-Pokorny, 2, pp. 78 f. Note also that posthabeō (as opposed to perhibeō, diribeō, inhibeō) seems to be a late compound, posterior to the action of the prehistoric initial accent (as Prof-A. Martinet kindly remarked to me).
- διá is probably a Greek innovation, in form as well as in function: Indo-European *dis- (< *dwis-?) is a preverb, not a preposition, in the other languages where it appears (Latin, Germanic, Albanian); cfr. Wackernagel, Vorles.2, 2, pp. 168, 175; Brugmann, Grundr.2, 2, 2, p. 814 and the etymological dictionaries of Feist, Boisacq, Ernout-Meillet, Kluge, Walde-Hofmann and Walde-Pokorny.
- From the point of view of form, διá is a transformation of *dis- on the analogy of μετὰ κκτὰ, πεδὰ, ἀνὰ.
- In πέδα, μέτα, ἄμα, μíγα (and also probably in πάρα, ἄνα, κάτα, διά, *Fέκα) an old instrumental ending -*∂ is perhaps preserved (cfr. Schwyzer, Griech. gramm., pp. 622 f.). Note that all these words are relatively recent (at least in form: διά and πάρα are exclusively Greek, cfr. Latin dis, Vedic pur, Gothic faura, Old H. Germ, fora with once long *-ā; only Armenian at CAN be identical with Greek πάρα but it might also be = Old High German furi).
- Notice that Greek μέτα and πέδα are usually mutually exclusive; when they are found in the same dialect, the one is obviously in the process of eliminating the other; they should therefore really be counted as one. μετα is of course a crossing of both. See especially R. Günther, IF., 20, 1906–1907, pp. 21 ff.
- According to Brugmann, Grundr.*, 2, 2, p. 863, πέδα is more recent than μέτα, which seems also reasonable for geographic reasons, since πέδα is an Eastern form (Greek, Arm., Lettish, Slavic) and μέτα a (relatively) Western one (Greek, Germ.). Cfr. also Günther, IF., 20, pp.21 ff.; 126 ff.
- πέδα May perhaps be preserved in Latin pedisequus, cfr. Brugmann-Thumb, Griech. Gramm., p. 510; Grundriss2, 2, 1, p. 131; 2, p. 863; Persson, IF., 26, 1910, p. 66; Schulze, Quaest. epicae, p. 497, n. 6. But it is very doubtful.
- Also perhaps Latin ūsque: but the adverbial usage seems to have developed into the prepositional one in Latin itself; cfr. Leumann and Hofmann, Lat. gramm., p. 498. See here notes 2, 11 and 15. Ernout-Meillet does not mention the connection of ūsque with Vedic úd etc., and proposes another one instead (Latin ut etc.). Cfr. also Walde-Pokorny, 1, p. 190, who separates Old Irish o-, u-, ūa-, ō Thurneysen in the first volume of the German edition of his Old Irish Grammar, p. 476, admitted (for Old Irish ūa-, ō-, o-, u-) a crossing of úd and au (Lat. au-ferō etc., see above); but in his recent English edition (Dublin, 1946), p. 524, he sticks exclusively to the last hypothesis and thinks that the -d of the forms ūad, ūad-, ūadi, ūa'dib “may have been suggested by and, indib (§842).”—See also here n. 22.
- Greek ἐς, εìς is a Greek innovation: cfr. Wackernagel, Vorles.2, 2, p. 156 and Boisacq s.u.; Meillet, Introduction7, p. 352; see also Leumann-Hofmann, p. 129, §113 zus.; Meillet- Vendryes, Traité2 (1948), pp. 521, 523. Several Greek dialects, among them all the “Achaean” ones (inclusive of the archaic Arcadian and Cyprian dialects), use ἐν with the accusative, like Latin, Irish, Osco-Umbrian, Baltic, Slavic, Armenian and Germanic, instead of ἐς or εìς. Cfr. Brugmann, Grundr.2, 2, 2, p. 831: “Im griechischen galt einmal allgemein ἐν mit akk. In einem teil des sprachgebiets, im Böot., Nordwestgr. usw. (Günther, IF., 20, 6 ff.), blieb dieser gebrauch lebendig, während im Ion-Att. usw. (Günther a.a.O. 4 ff.; 10 f.) die neubildung ἐνς (εìς, ἐς, ìς) im lebendigen gebrauch durchdrang.” For the accretion of the -s cfr. Iranian patiš, Greek ἀμφίς, ἄνευς, ἄϰρι, Latin abs, sups, obs, cis, uls and see Standerwick, Language, 7, 1931, pp. 174 f.; Meillet, Introduction7, p. 352; Leumann-Hofmann, Gramm., p. 129; R. Günther, IF., 20, 1906–1907, p. 54.
- It is also remarkable that Greek shares three prepositions (μέτα,ἄνευ, ἄτερ) only, or almost only, with Germanic (German mit, ohne, be-sonder-s, Engl, asunder). This is probably, as are most Graeco-Germanic isoglosses, a relatively late innovation; and the two last of these prepositions are certainly late, being “improper.” On the Graeco-Germanic isoglosses see my Dialetti indoeuropei, Naples, 1931, pp. 172 if. (Annali del R. Istituto orientale di Napoli, vol. 4), and also my article “Quelques isoglosses gréco-germaniques” in Annuaire de l'institut de philologie et d'histoire orientales et slaves, 7, 1939–44, pp. 369 ff.—If Slavic Sŭ is from *kom, it builds lateral areas with Celtic kom (Goth, miþ etc.).
- The remaining one, ὐ-, ὐ-, whose geographic extension is wider, is very rare in Greek (ὅ-βρις, ὅϰηρoς) and really vestigial, with the sole exception of the very archaic Cypriote dialect (ὐ τυϰᾱ. ὐ fηρι), where it replaces ἔπι. I must however openly admit that I consider its connection with Vedic úd, Old Persian (doubtful) ud-, Gothic út, Engl, out, Old Bulg. vy, Lithuanian už-, Irish u- o- (with gemination of the following consonant) not quite above suspicion; cfr. also Walde-P. 1, 190. It could perhaps represent the zero degree of au (see above in the text). Cfr. also Günther, IF., 20,1906–1907, pp. 152 f.
- Meillet (in Ernout-Meillet s.u. ex) and Bàrtoli (Saggi di linguistica spaziale, Turin, 1945, pp. 11 f.) consider ex and úd as mutually exclusive: but Meillet thinks that ex is older, while Bàrtoli thinks the opposite. This opinion of Bàrtoli's seems very strange indeed, for the antiquity of ex with relation to úd is assured, as it seems, precisely by Bàrtoli's norm of the isolated area (Latin and Celtic, in our case). Moreover, surprisingly enough, all the facts seem not to have been carefully marshalled by Bàrtoli in this case, contrary to his habit: it seems clear that *ud, *ūd is found not only in Indo-Aryan, Iranian and Germanic, as he states, but also in the intermediary Baltic (Lith. uz, Lett, uz) and Slavic (vy, vŭzŭ, vŭz, vŭně, vŭnu, vŭnŭ), cfr. Brugmann, Grundr.2 2, 2, pp. 902 ff.; whereas ex is attested with certainty only for Latin, Osco-Umbrian, Celtic, Greek and Old Pruss. (cfr. Ernout-Meillet, Brugmann l.l. pp. 823 ff.; Walde-Hofmann, Walde-Pokorny s.uu.); so that the “figure” would be as follows (ex is the broken area):
- This central area has also produced, in the field of prepositions, the innovation represented by Brugmann, Grundr.*, 2, 2, pp. 846 ff., with *ghō (*o-gh-, *u-gh-, *oi-gh-), that is, in terms of historical languages, Old Bulg. za, vŭ-zŭ, Lith. a-žu, u-žu, u-žuo, u-ž-, Lettish u-z, ŭ-z, ai-z, Arm. z-, ∂-z-, s- (before voiceless stops) and perhaps Gothic ga-, German ge-. See especially Feist, Vergl, wb. der got. spr.3, s.u. Cfr. also what is said on μέϰρι here in note 12.
- “According to J. Wackernagel, Symbolae Danielsson, pp. 383 ff., Leumann, Schwyzer and Ernout-Meillet,2 p. 474, an ancient Indo-European preposition *énu, represented by Vedic ánu (and Gothic inu, German ohne), is preserved in Latin īgnōscō ‘I pardon’ (cfr. Sanskrit anu-jñā- ‘to pardon’), īnsequor, īnstō, īnstita etc. (J. B. Hofmann, Etym. wb., s.u., p. 677 hesitates). If this is so, we have another archaism of Latin, strengthened by the norm of the lateral areas (there is no *énu in Greek). Cfr. J. B. Hofmann, Wb., s.u. īgnōscō. Other scholars, however, connect *énu with Greek ἄνα,ἄνευ,ἄτερ, Vedic sanutár, OHG. suntar. To me the semantic connection of Vedic ánu, Avest. ana, Gr. ἄνα with German ohne, Gothic inu, Gr. ἄνευ seems very doubtful.
- We have regularly used in this article, as the reader can easily see, the areal norms of the area isolata, aree laterali and area maggiore, and moreover the norm of the fase sparita. The main principles of these norms are to be found, of course, in M. Bàrtoli, Introduzione alla neolinguistica, Geneva and Florence, 1925. I have added of my own the “horseshoe area,” which is rather obvious. I shall devote another article to it.
- On the exact value of the terms “archaic” and “conservative” see my note in Studi baltici, 5, 1935–36, pp. 30 ff.
- There is no doubt, of course, that sēd (sē) is of Indo-European origin: it is nothing but the ablative of the reflexive pronoun *sē (cfr. Leumann-Hofmann, pp. 102, 283, 530 f.; Ernout-Meillet, s.u.; the question here discussed is whether its use as a preposition is a Latin innovation or not. The same is true for uls, connected with ollus, olle, ille, ultrā, Old Irish oll (= Lat. ollus), ind-oll, Old Bulgarian lani etc. (see Walde-Hofmann, s.u. ollus, which is itself a disappearing archaism).
- Likewise, dē, which is found only in Latin and Celtic, has every likelihood of being very old.
- Choose new content alerts to be informed about new research of interest to you
- Easy remote access to your institution's subscriptions on any device, from any location
- Save your searches and schedule alerts to send you new results
- Export your search results into a .csv file to support your research