¹ See Hawkins, J. D., The Negatives in Hieroglyphic Luwian, Anatolian Studies, XXV (1975), 119 ffGoogle Scholar. In the writing of this note I have been able to work in close collaboration with Mr. Hawkins, to whom I owe a great debt of gratitude. I am also grateful to Professors A. F. L. Beeston and T. Burrow and to Dr. G. L. Lewis who have helped me with the Arabic, the Dravidian and the Turkish evidence.

² Cf. Laroche, E., Dictionnaire de la langue louvite [DLL], Paris 1959, p. 74Google Scholar s.v. nauwa and p. 75 s.v. . See also Kammenhuber, A., Handbuch der Orientalistik, I.2, 2Google Scholar (Altkleinasiatische Sprachen) [Handbuch], Leiden/Köln 1969, 173Google Scholar.

³ Cf. Neumann, G. in Handbuch, op. cit., 394Google Scholar; Houwink Ten Cate, Ph. H. J., The Luwian population groups of Lycia and Cilicia Aspera during the Hellenistic Period, Leiden 1965, 82 fGoogle Scholar. The meaning of Lycian newe in which Meriggi, , WZKM 53 (1957), 203Google Scholar wants to see a negative is not yet clear. Lycian ni could well derive from nis, with the usual loss of final -s, but in view of the existence of Hier. Luw. ni, it is also possible to suggest that it goes back to a simple ni without the -s extension. ne certainly derives from an earlier na; nipe and nepe are strengthened forms of ni and ne.

⁴ The origin of Hitt. lē is disputed. It could be connected with IE *nē provided an alternation n/l, similar to that of Hitt. lāman and e.g. Lat. nōmen, were established: cf. for the evidence Kronasser, H., Etymologie der heth. Sprache, Lief. 1, Wiesbaden 1962, 68Google Scholar, but see the reservations of Kronasser, H., Vergl. Laut- und Formenlehre des Heth., Heidelberg 1956, 161 f.Google Scholar, Kammenhuber, A., Handbuch, 173Google Scholar. natta, the factual negative, is normally taken as derived from the negative na or n- (from IE *ne) + one or more enclitic particles (see e.g. Kronasser, H., Vergl. Laut- und Formenlehre, op. cit., 161Google Scholar, Friedrich, , Heth. Wb., 149Google Scholar and Holt, , Bibl. Or. 15 (1958), 156 note 158Google Scholar). In addition to these forms Hittite also has a frequently used negative nāwi (spelled na(-a)-ú-i) with the meaning “not yet”, and two rarer forms nūwān and nūmān which are phonetic doublets and seem to mean “never, not at all” (cf. Friedrich, Heth. Wb., s. vv.). In spite of the phonetic difficulties it seems likely that here too we are in presence of the negative na (or n-) joined to various enclitic particles. Finally it is possible that a negative is contained in the even rarer nik(k)u or neku: cf. for the evidence Hahn, E. A., Language, 12 (1936), 110 f. note 14CrossRefGoogle Scholar; Friedrich, J., JCS 1 (1947), 303 fGoogle Scholar. Friedrich, , Heth. Wb., 151 s.v.Google Scholarnik(k)u denies his earlier interpretation of the particle as a negative, but cf. more recently Josephson, F., RHA 79 (1967), 149Google Scholar, note 39 and Siegelová, J., Appu-Märchen und Hedammu-Mythus (Studien zu den Boğazköy-Texten, 14), Wiesbaden 1971, 8—9Google Scholar.

⁵ Cf. Kammenhuber, , Handbuch, 173Google Scholar. See for the evidence, Carruba, O., Das Palaische (Studien zu den Boğazköy-Texten, 10), Wiesbaden 1970, 65Google Scholar, and Carruba, O., Beiträge zum Palaischen, Istanbul, 1972, 42 and passimGoogle Scholar.

⁶ Cf. Gusmani, R., Lydisches Wörterbuch, Heidelberg 1964, 172—3Google Scholar; Heubeck, A., Handbuch, 415Google Scholar.

⁷ Cf. Hawkins, J. D., Davies, A. Morpurgo, Neumann, G., Hittite Hieroglyphs and Luwian: New evidence for the connection [HHL], Nachr. Ak. Wiss. Göttingen, 1973 no. 6, 47 fGoogle Scholar.

⁸ If Holt's interpretation of Hitt. natta as derived from n+at+a (see above note 4) were correct, it could be possible to argue that Hier. na too derives from *n+at, with the normal dropping of final plosives in Luwian—but all this is far too speculative.

⁹ Cf. Hawkins, 123 and 128 ff., 142 ff.

¹⁰ Cf. Hawkins, 121 f. and 131 ff.

¹¹ A clear example is contained in the letter of the widow of the Pharaoh to Suppiluliuma (KBo. V 6 iii 14): ÌR-IA-ma-wa nu-u-wa-a-an pá-ra-a da-aḫ-ḫi nu-wa-ra-an-za-kán ^LUMU-TI-IA i-ia-mi, which Güterbock, , JCS 10 (1956), 94Google Scholar translates “Never shall I pick out a servant of mine and make him my husband!”.

¹² Here it should be pointed out that in theory it would be possible to interpret both Luw. na-u-wa and Hier. NEG-wa/i/ na-wa/i as nawi. Neither the cuneiform nor the hieroglyphic spelling exclude this transcription. If so, however, we would create a form in all respects identical to Hitt. nawi (see above note 4), although this has a meaning “not yet” which is not suitable for either the Cuneiform or the Hieroglyphic texts.

¹³ Friedrich is far too cautious to say anything of the sort, but from Elementarbuch, I, 146Google Scholar, one could receive the impression that the negatives can occur initially only in interrogative sentences. This is certainly wrong. To give just two examples, there must have been an initial negative in KBo XXII 2 Vs 19Google Scholar (cf. Otten, H., Eine altheth. Erzählung um die Stadt Zalpa, Studien zu den Boğazköy-Texten 17, Wiesbaden 1973, 6Google Scholar) and there certainly is an initial negative in the sentence na-at-ta-an ú-uk / t[(ar-na-a)]ḫ-ḫu-un LUGAL-ša-an SAL.LUGAL-ša tar-na-aš “Nicht ich habe ihn ausgelassen, (sondern) das Königspaar hat ihn losgelassen” (Otten, H. u. Souček, V., Ein althethitisches Ritual für das Königspaar, Studien zu den Boğazköy-Texten 8, Wiesbaden 1969, 30 f.Google Scholar, Rs III 4-5).

¹⁴ It is possible that in the sentences of KUB XXXV 24, 5 ffGoogle Scholar. = 25, 7 ff. = 43 III 7 ff. we must recognize an initial na-a-ú-wa.

¹⁵ Cf. Hawkins, , citations 37–45Google Scholar.

¹⁶ Cf. Friedrich, , Elementarbuch, I, 69 and 135Google Scholar.

¹⁷ Cf. Hawkins, f. and citations 29, 30.

¹⁸ See Friedrich, , Elementarbuch, I, 145Google Scholar.

¹⁹ Cf. Hawkins, , citation 34Google Scholar; for other examples, see ibid, citations 33 (iii), 34 (i), 46, 49, 51 (iii and iv).

²⁰ For other clear examples see TL 75,3; 88,3; 91,2; 93,2; 118,2; 134,2 (another sentence in addition to that quoted above), 149 a, 4; 150.

²¹ For a survey of the problems which concern the relative position of the verb and the preverb in the Indo-European languages see Watkins, C. in Proceedings of the Ninth International Congress of Linguists, The Hague 1964, 1035–42Google Scholar.

²² Jespersen, Otto, Negation in English and other languages, Copenhagen 1917, 64 ffGoogle Scholar.

²³ Cf. Friedrich, , Elementarbuch I, 146Google Scholar, and especially Sommer, F., Die Aḫḫijavā-Urkunden, München 1932, 106Google Scholar.

²⁴ Cf. for instance Pedersen, H., Hittitisch und die anderen indoeuropäischen Sprachen, Copenhagen 1938. 200Google Scholar.

²⁵ Die satzeinleitenden Partikeln in den idg. Sprachen Anatoliens, Rome 1969, 67 fGoogle Scholar.

²⁶ Carruba, loc. cit., denies altogether the existence of a particle šu in Hittite (hence the need to explain nassu as derived from na + assu) but his views have not been generally accepted: cf. e.g. H. Otten, Ein altheth. Erzählung, op. cit. (note 13), 44 note 8; Kammenhuber, A., Materialien zu einem heth. Thesaurus, Lf. 1, no. 1 (Heidelberg 1973), 2Google Scholar.

²⁷ Cf. e.g. Hahn, E. A., Language 12 (1936), 110 fCrossRefGoogle Scholar. and Carruba, op. cit. (note 25), 69 ff. It is difficult to separate this particle from the strengthening element which appears e.g. in immakku, , Ú.UL-akku etc.: cf. Friedrich, , JCS 1 (1947), 302 ffGoogle Scholar. For nikku see above note 4. The connection with Palaic -ku (for the references see Carruba, loc. cit. and cf. Kammenhuber, , Handbuch, 353Google Scholar) is clear, but does not help to explain the origin of the disjunctive (for a possible suggestion see below 4.5). It may be useful to point out here that the meaning of -(a)ku … -(a)ku is sometimes ambiguous: it could be expressed by “both … and …” or by “either … or …” (see also Hahn, loc. cit. note 15).

²⁸ Cf. Houwink Ten Cate, op. cit. (note 3), 76; Neumann, , Handbuch, 393Google Scholar.

²⁹ Cf. Gusmani, , Lyd. Wb., 86 fGoogle Scholar. s.v. buk(-) and for the comparison with IE *bhu- see Vetter, E. in Sitzber. Oesterr. Ak. Wiss (Phil. Hist. Kl.), 232 (1959), 3, p. 40 note 18Google Scholar.

³⁰ See Carruba, , Athenaeum, 38 (1960), 52 f. note 34Google Scholar.

³¹ This is not altogether surprising; in a linguistic group as closely knit as that formed by the Slavonic languages we find that the words for “or” are respectively ili in Russian, lub or albo in Polish and nebo in Czech.

³² See Pokorny, , Idg. Et. Wb., 75Google Scholar, but the connection which Pokorny postulates between IE *ṷe and the pronominal stem au-, *u- is far from certain.

³³ Holt, , Bibl. Or. 15 (1958), 156Google Scholar has suggested that IE *ṷe survived in the Luwian negative nawa (cf. Lat. neue).

³⁴ Obviously not all languages have coordinating elements in their surface structure and not all languages make a distinction between “and” and “or” conjunctions: cf. Bergsland, K., Norsk Tidsskrift for Sprogvidenskap, 15 (1949) 374 fGoogle Scholar. and the acute follow-up of Roman Jacobson, , Selected Writings, vol. II (The Hague-Paris, 1971), 263Google Scholar (a propos of the Samoyed particle for and/or); cf. also Hahn, , Language 12 (1936), 111 note 15CrossRefGoogle Scholar. It may be worthwhile to point out that even in languages which knew the distinction this may tend to be obliterated or at least neutralized in some contexts: in late Latin uel could be used with the meaning of et and vice-versa (see Löfstedt, E., Phil. Kommentar zur Peregrinatio Aetheriae, Oxford, Uppsala, Leipzig 1936, 197 ff.)Google Scholar; for the ambiguous use of Old French ne (from Latin nec) see below 4.7.2.

³⁵ See below 4.4. and note 46.—It is worth mentioning that in interrogative sentences a disjunction may be marked by elements which are not used in positive statements: in classical Latin utrum … an … (or the simple an) may serve as an example. In Hittite too there are special ways of marking double questions: if the questions are direct nu -ma occur in the second question; if the questions are indirect, man occurs both in the first and in the second question (cf. Friedrich, , Elementarbuch I, 147 and 167)Google Scholar.

³⁶ Indeed, it is so superficial that it can hardly be called a classification. Descriptively it is at fault because it does not attempt to distinguish among the various syntactical properties of the disjunctive particles and among the various meanings (e.g. inclusive and exclusive “or” are not separated). Also, each particle has been considered in isolation and all its distributional or semantic connections with other elements of the language have been ignored. From a historical point of view, one of the shortcomings of this classification is that it only includes those “or” words whose origin is immediately apparent; yet, this very fact means that all too often we are dealing with linguistic elements which have not been entirely grammaticalized (see below 4.2 apropos of Italian vuoi … vuoi …) and therefore are not very representative. Yet, it is the assumption of this paper that the evidence adduced, however defective, is sufficient to answer the question asked in 3.1.

³⁷ Cf. e.g. Fromm, H., Sadeniemi, M., Finnish Elementarbuch I, Heidelberg 1956, 198Google Scholar, who compare Old Norse ella, but see also Hakkulinen, L., The structure and development of the Finnish Language, Bloomington, The Hague 1961, 55Google Scholar.

³⁸ See e.g. Brockelmann, C., Osttürkische Grammatik der islamischen Litteratursprachen Mittelasiens, Leiden 1954, 195Google Scholar. For the Turkish conjunctions in general see Temir, Ahmet, “Die Konjunctionen und Satzeinleitungen im Alt-Türkischen”, Oriens 9 (1956), 41-85, 233–80Google Scholar.

³⁹ Cf. Ernout-Meillet, , Dict. Etym. de la langue latine, Paris 1954Google Scholar, s.v. vel.

⁴⁰ Cf. Hakkulinen, loc. cit. (in note 37).

⁴¹ See Mirambel, A., La langue grecque moderne, Paris 1959, 179Google Scholar.

⁴² In all these cases the meaning may oscillate between “either … or …” and “both … and …”.

⁴³ Cf. Arden, A. H., A progressive grammar of common Tamil, Madras 1934, 129 (and 286)Google Scholar; Beythan, H., Praktische Grammatik der Tamilsprache, Leipzig 1943, 156Google Scholar.

⁴⁴ See Spencer, H., A Kanarese Grammar, Mysore 1914, 162 f.Google Scholar; Jensen, H., Grammatik der kanaresischen Schriftsprachen, Leipzig 1969, 142Google Scholar.

⁴⁵ Cf. Lewis, H., Pedersen, H., A concise comparative Celtic grammar, Göttingen 1937, 73 fGoogle Scholar.

⁴⁶ Kretschmer, P. in Scritti in onore di A. Trombetti, Milano 1938, 27–50Google Scholar, argued that in a number of languages interrogative sentences have arisen from the second part of disjunctive sentences and adduced the similarity of disjunctive and interrogative particles as an argument in favour of his hypothesis (ibid. 37 ff.). The suggestion was rejected by Ed. Hermann in the course of his lengthy analysis of the problems posed by interrogative sentences: cf. Ed. Hermann, , “Probleme der Frage”, Nachr. Ak. Wiss. Göttingen, 1942, 121–408Google Scholar (especially 369 ff.).

⁴⁷ Cf. Senn, A., Handbuch der lit. Sprache, I, Heidelberg 1966, 382 and 483Google Scholar; see also E. Fraenkel, Lit. Et. Wb. s.v. .

⁴⁸ See e.g. G. Nandriş, , Old Church Slavonic Grammar, London 1965, 209Google Scholar and cf. also Meillet, A., Le slave commun, Paris 1934 ², 477 and 484Google Scholar.

⁴⁹ Cf. Bright, W., The Karok Language, Berkeley and Los Angeles 1957, 131 and 136Google Scholar.

⁵⁰ See Milner, G. B., Samoan Dictionary, London 1966Google Scholar s.v. pē ¹ and s.v. po-, Marsack, C. C., Teach yourself Samoan, London 1962, 60 and 112Google Scholar.

⁵¹ Since in Indo-European languages at least most particles and subordinating or coordinating conjunctions arise from pronominal stems this statement does not exclude that “or” words may have originated in different conjunctions.

⁵² Cf. Arden, A. H., A Progressive Grammar of the Telugu Language, Madras 1927, 82Google Scholar.

⁵³ See Vertés, E., Die Ostjakischen Pronomina, Bloomington and The Hague 1967, 143 f.Google Scholar, 148, 155. Similarly in Acoma, an Indian language of New Mexico, zíi is the interrogative-indefinite pronoun but also means “or” and (when repeated) “either … or …”: cf. Miller, W. R., Acoma Grammar and Texts, Berkeley and Los Angeles 1965, 178 and 183Google Scholar.

⁵⁴ See above para. 3 and note 28.

⁵⁵ If so, i.e. if -ku is related to the interrogative pronoun, there is no need to postulate that it derives directly from IE *k^we (see above para. 3 and note 27). The connection between the relative-interrogative pronoun of IE (*k^wis) and IE k^we has been differently explained (see e.g. O. Szemerényi, Einführyng in die vgl. Sprachwissenschaft, Darmstadt 1970, 194 f.).

⁵⁶ For the use of Latin siue cf. Hofmann, J. B. and Szantyr, A., Lateinische Syntax und Stylistik, München 1965, 503 fGoogle Scholar. For late Latin see Norberg, D., Beiträge zur spätlateinischen Syntax Uppsala 1964, 96 ffGoogle Scholar.

⁵⁷ See A. Mirambel, loc. cit. (in note 41).

⁵⁸ For classical Arabic see e.g. Fischer, W., Grammatik des klassischen Arabisch, Wiesbaden 1972, 148Google Scholar ('illā < 'in-lā) and 204 (wa'illā). For the colloquial language cf., for instance, Cowell, M. W., A reference grammar of Syrian Arabic, Washington 1964, 395Google Scholar: s-sǝne fīha … šah^ǝr waḥed ?ǝlo bass tmāna w-^tǝšrīn wǝlla tǝs ^ta w-^tǝšrīn yōm “There is one month in the year which has only twenty-eight or else twenty-nine days”. For Egyptian Arabic see also Mitchell, T. F., An Introduction to Egyptian Colloquial Arabic, London 1956, 49 ffGoogle Scholar. from where I quote the following sentence: húwwa-lli šuftu- walla lá? “Is he the one I/you saw yesterday or not?”

⁵⁹ See e.g. Weil, G., Grammatik der osmanisch-türkischen Sprache, Berlin 1917, 207Google Scholar and cf. the examples (here quoted in modernized spelling): vaktinde gel yoksa yalnız giderim “komm pünktlich, wenn nicht, gehe ich allein”, and ben yoksa biraderim “ich oder mein Bruder”. Cf. also Kissling, H. J., Osmanisch-türkische Grammatik, Berlin 1960, 203Google Scholar.

⁶⁰ Cf. Eckmann, János, Chagatay Manual, The Hague 1966, 185 and 201Google Scholar. For a similar form in Uzbek see von Gabain, A., Oezbekische Grammatik, Leipzig und Wien 1945, 137Google Scholar.

⁶¹ Cf. A. H. Arden, op. cit. (in note 43), 130 (and 295); H. Beythan, op. cit. (ibid.), 156.

⁶² See for the examples quoted Tryon, D. T., Conversational Tahitian. An Introduction to the Tahitian Language of French Polynesia, Canberra 1970, 107Google Scholar. See also Jaussen, Tepano, Grammaire et dictionnaire de la langue tahitienne, Paris 1969 ⁵, p. 33Google Scholar s.v. ou and p. 74 s.v.aore.

⁶³ Cf. Tryon, op. cit., 64 (ra) and 46 ('aore). For a general account of the Tahitian language see Sauvageot, A., Structure d'une langue polynésienne: le tahitien, in Conférences de l'Inst. de ling, de l'Univ. de Paris, 10 (1950–1951), 83–99Google Scholar.

⁶⁴ These two examples are taken from Moignet, G., Grammaire de l'ancien français, Paris 1973, 332 ff.Google Scholar, who offers a purely descriptive (i.e. non historical) analysis of the language in general and of this phenomenon in particular. See also Gamillscheg, E., Hist, französische Syntax, Tübingen 1957, 577Google Scholar and especially Antoine, G., La coordination en français, vol. II, Paris 1962, 1041 ffGoogle Scholar. and especially 1077 ff. and 1094 ff.

⁶⁵ The two examples are taken from Yvon, H., Les expressions négatives dans la Conqueste de Constantinople de Villehardouin, in Romania 81 (1960), 296–307Google Scholar (at p. 299). Cf. also the other two articles by the same author on the negative expressions in the Queste del Saint Gral and in the Vie de Saint Louis ibid., 80 (1959), 63-78 and 81 (1960), 99-111.

⁶⁶ Cf. Wagenaar, K., Étude sur la négation en ancien espagnol jusqu'au XV^e siècle, Groningen-The Hague 1930, 115Google Scholar.

⁶⁷ Cf. D. Norberg, op. cit. (in note 56), 105 f., but it is important to point out that in Old French ne in most cases still has its full negative value: neither a second negative nor a strengthening particle (pas etc.) are necessary.

⁶⁸ Obviously the problem is extremely complicated; of the authors quoted perhaps only Antoine (see note 64) makes it clear how complicated. However, in most cases he tries to operate with the suggestion that ne tends to join negative sentences; in a number of cases in which this is not so he speaks of an idée négative implicite (ibid., 1041 ff.), but even this assumption is not easy to uphold (cf. ibid., 1094 ff.). Very much worth reading are Antoine's observations about the alternation between et and ou or rather about the frequent absence of semantic distinction between these two particles (cf. ibid., pp. 1059 ff.).

⁶⁹ See Pedersen, loc. cit. (in note 24).

⁷⁰ I have also not mentioned, since it could have seemed irrelevant, the connection which exists in a number of languages between negatives and “or” particles in comparative clauses. In a construction of the type “better than …” Ancient Greek uses ἤ “or”, where English has “than”. Similarly in Old Church Slavonic we find li “or” and in Gothic pau “or” after the comparative. Yet, in OCS “than” can also be expressed by niže “and not”; in early Welsh the comparative is followed by a particle no which must include a negative, in English dialects nor may be used instead of than, and in a number of Finno-Ugric languages a negative is an integral part of the particles used after the comparative (cf. Raun, A., in American Studies in Uralic Linguistics, The Hague 1960, 219 ffGoogle Scholar.). In Sanskrit “than” is expressed by na, which is identical with the negative; this etymology has been disputed, but see for the evidence Vendryes, J., BSL 46 (1950) 9 ffGoogle Scholar. (with references to the earlier literature). Once more we gain the impression that there is some sort of overlap between “or” and the negatives. Manu Leumann (Kleine Schriften, Zürich 1959, 202 ff.Google Scholar) has wondered whether the Greek use of ἤ after the comparative may be due to the use of ἤ in disjunctive questions such as ἆρα ὁ θεος ὁ κρείττων ἤ ὁ ἂνθρωπος; This is a convincing hypothesis and one which gains in plausibility because of the connections between “or” particles and interrogative particles or pronouns mentioned above (see 4.4. and 4.5.). However—and in a more speculative mood—we may suggest that, if there is a sense in which the disjunctive particle may include a negative (or may conceal an underlying negative), Leumann's hypothesis is no longer necessary: both surface evidence, such as that just quoted, and deeper syntactical analysis induce us to recognize in the second part of a comparative statement a hidden negative (cf. Seuren, P. A. M., The Comparative, in Kiefer, and Ruwet, [eds.], Generative Grammar in Europe, Dordrecht 1973, 528 ff.CrossRefGoogle Scholar).

⁷¹ In formal logic P v Q is equivalent to —Q→P, so that, provided we assume that the connectives of natural language behave in a similar manner to those of logic, there is no difficulty in recognizing the connection between “P or Q” and “P if not Q”. From this point of view the development of Turkish and Arabic is clear. The problems posed by the other languages remain. If we do not want to postulate that “not” can be treated in a similar way to “if not”, as indicated above, we can still point out that in formal logic P v Q is equivalent to—(—P & —Q). Provided we can accept the same assumption as before, it seems plausible to conclude that whatever can be expressed in terms of “or” can also be expressed in terms of “and” and “not”. However, can we then argue that “P and not Q” and “P not Q” in some way owe their origin to an alteration or a mispresentation or a “rewriting” of “not (not P and not Q)”?