¹ The abbreviations employed here are in general listed in Edgar H. Sturtevant, A Comparative Grammar of the Hittite Language, Philadelphia, 1933; (with George Bechtel) A Hittite Chrestomathy, Philadelphia, 1935; A Hittite Glossary², Philadelphia, 1936; Supplement to a Hittite Glossary Second Edition, Philadelphia, 1939 (here abbreviated HG, Chr., Gl., and Gl. Sup., respectively). Deviations from these are: Bois. for Boisacq, Dict.; WP for Walde-Pokorny. Additional abbreviations used here are as follows: Berneker = Erich Berneker, Slavisches Etymologisches Wörterbuch, Heidelberg, 1908–13; BPW = (Berliner) Philologische Wochenschrift; Br.-Th. = Karl Brugmann, Griechische Grammatik⁴, revised by Albert Thumb, Munich, 1913; Brugmann, Dem. = Karl Brugmann, Die Demonstrativpronomina der Indogermanischen Sprachen, Leipzig, 1904 = Abhandlungen der Philologisch-Historischen Klasse der Königlich Sächsischen Gesellschaft der Wissenschaften 22.6; Brugmann, KVG = K. Brugmann, Abrégé de Grammaire Comparée des Langues Indo-Européennes, translated into French by J. Bloch, A. Cuny, and A. Ernout, under the direction of A. Meillet and R. Gauthiot, Paris, 1905; Brugmann, Total. = Karl Brugmann, Die Ausdrücke für den Begriff der Totalität in den Indogermanischen Sprachen, Leipzig (1894); Buck = Carl Darling Buck, Comparative Grammar of Greek and Latin, Chicago (1933); Cocchia, Saggi = Enrico Cocchia, Saggi Glottologici, Naples, 1924; CP = Classical Philology; EM = A. Ernout and A. Meillet, Dictionnaire Étymologique de la Langue Latine, Paris, 1932; Friedrich-, Kl. Beitr. = Johannes Friedrich, Kleine Beiträge zur Churritischen Grammatik, Leipzig, 1939 = MVAG 42.2; Güntert = Hermann Güntert, Indogermanische Ablautprobleme, Strassburg, 1916 = Untersuchungen zur Indogermanischen Sprach- und Kulturwissenschaft 6; Hahn, Coord. = E. Adelaide Hahn, Coordination of Non-Coordinate Elements in Vergil, Geneva, 1930; Harv. Stud. = Harvard Studies in Classical Philology; Heraeus, Kl. Schr. = Kleine Schriften von Wilhelm Heraeus, edited by J. B. Hofmann, Heidelberg, 1937; Hübschmann, Arm. Gr. = H. Hübschmann, Armenische Grammatik, Leipzig, 1897; Idg. Jb. = Indogermanisches Jahrbuch; Kieckers, Sprachw. Misc. 4 = E. Kieckers, Sprachwissenschaftliche Miscellen 4, Dorpat, 1926 = Acta et Commentationes Universitatis Tartuensis (Dorpatensis), B. Humaniora 10.2; Kluge = Friedrich Kluge, Etymologisches Wörterbuch der Deutschen Sprache, Berlin and Leipzig, 1921; Kretschmer, Einl. = Paul Kretschmer, Einleitung in die Geschichte der Griechischen Sprache, Göttingen, 1896; Lidén, Studien = Ewald Lidén, Studien zur Altindischen und Vergleichenden Sprachgeschichte, Upsala, 1897; Lindsay, Lat. Lang. = W. M. Lindsay, The Latin Language, Oxford, 1894; Lindsay, Synt. Plaut. = W. M. Lindsay, Syntax of Plautus, Oxford, 1907; Lodge, Lex. Pl. = Gonzalez Lodge, Lexicon Plautinum, 2 vols., 1904–33; Loth, Mots latins = J. Loth, Les Mots Latins dans les Langues Brittoniques, Paris, 1892; LS = Henry George Liddell and Robert Scott, A Greek-English Lexicon, revised by Henry Stuart Jones, 2 vols., Oxford, no date; Meyer, Alb. Wb. = Gustav Meyer, Etymologisches Wörterbuch der Albanischen Sprache, Strassburg, 1891; MV = A. Meillet and J. Vendryes, Traité de Grammaire Comparée des Langues Classiques², Paris, 1927; NED = A New English Dictionary on Historical Principles, edited by James A. H. Murray and others, 10 Parts and Supplement, Oxford, 1888–1933; NJrb. = Neue Jahrbücher für das Klassische Altertum, Geschichte und Deutsche Literatur; PBB = Beiträge zur Geschichte der Deutschen Sprache und Literatur; PW = BPW; RB = Revue Belge de Philologie et d'Histoire; REW = W. Meyer-Lübke, Romanisches Etymologisches Wörterbuch, Heidelberg, 1935; RFC = Rivista di Filología e d'Istruzione Classica; Saussure, Mém. = Ferdinand de Saussure, Mémoire sur le Système Primitif des Voyelles dans les Langues Indo-Européennes, Leipzig, 1879; Sch. = Eduard Schwyzer, Griechische Grammatik, Munich, 1934-; Schmidt, Hesychius = Hesychii Alexandrini Lexicon, edited by Mauricius Schmidt, 5 vols., 1858–66; Solmsen, Untersuchungen = Felix Solmsen, Untersuchungen zur Griechischen Laut- und Verslehre, Strassburg, 1901; Sommer, Hdb. = Ferdinand Sommer, Handbuch der Lateinischen Laut-und Formenlehre^2,3, Heidelberg, 1914; Thurneysen, Hdb. = Rudolf Thurneysen, Handbuch des Alt-Irischen: Grammatik, Texte, und Wörterbuch, Heidelberg, 1909; Wackernagel = Jacob Wackernagel, Altindische Grammatik, 3 vols., Göttingen, 1896–1930; Walde = Alois Walde, Lateinisches Etymologisches Wörterbuch², Heidelberg, 1910; WH = A. Walde, Lateinisches Etymologisches Wörterbuch³, revised by J. B. Hofmann, 2 vols., Heidelberg, 1938-; ZöG = Zeitschrift für die Österreichischen Gymnasien.

² He had formerly equated man with this word (HG 88, 132, 135).

³ Early attempts to identify mahhan with mehur ‘time’ (Sturtevant, RHA 1.78, Lang. 7.119–20) had to be abandoned when it became clear that intervocalic h and hh are different phonemes (Sturtevant, Lang. 12.185 fn. 17; cf. id., ib. 211, and Couvreur, Hett. Ḫ 202–3).

⁴ Thess. μά ‘δέ‘ and Attic-Ionic μά (in oaths) come into the picture likewise. Hittite has some forms without final n: ma = man, perhaps a mere orthographical variant (see Sommer, AU 383 fn. 1) and perhaps nasma (see below). For n movable in Hittite, consult HG 206; in Greek, Br.-Th. 299.

⁵ Cf. e.g. Br.-Th. 627.

⁶ KVG 658, Br.-Th. 627, Bois. 596, WP 2.685 (where the IE forms are given as sme, sm ā̆).

⁷ For parallels in use see especially Delbrück, Grund. 4.2.507–11.

⁸ The fact that it is indefinite only, never interrogative (with one possible exception to be taken up later), in IE, seems to me a point in favor of those who hold that the relative developed from the indefinite rather than from the interrogative—a group to which I have long belonged.

⁹ Thus in WP we find Latin semper in the first category but Av. hama- ‘omnis’ in the second, Lat. similis in the first but Skt. samás ‘like’ and Gr. ὁμός in the third, Lat. simul in the first but Gr. ἅμα in the third. The truth is that the meanings are inextricably interwoven, and the same word may have meanings belonging in more than one category. Undoubtedly some of the overlapping in meanings is due to the fact that the editors are trying to keep two classifications going at once: one for meaning, the other for form (on the various ablaut grades, see in particular Güntert 16; Brugmann, Grund. 2.2.896–7; Schmidt, KZ 32.372–3).

¹⁰ WP errs in printing samá- (489); see Wackernagel 3.577–8.

¹¹ Sommer, IF 10.219, Bois. 1060, Wackernagel 3.371, Buck 232; disputed by WP 1.633; denied by Brugmann, IF 21.12.

¹² Brugmann, IF 21.8; Schwyzer, Glotta 5.196; WP 2.489. Kent's counter-proposal, TAPA 42.81–2, based on his doubts concerning sahásram and mīlle (cf. fnn. 11 and 13), seems to me unconvincing.

¹³ Sommer, IF 10.216–20; accepted by KVG 388, Bois. 231 and 1060, Walde 484; questioned by Brugmann, Grund. 2.2.48 and IF 21.12–3; doubted by WP 2.488, and definitely denied 1.833; ignored by EM 583 and MV 474. Since sm- remained in OIr. (Brugmann, Grund. 1.770), it is assumed that OIr. mīle if from smi- must be a loanword from Latin (ib. 2.2.47; see also Loth, Mots latins 188; Lindsay, Lat. Lang. 420; Sommer, IF 10.216, 219); yet the possibility that this is an instance of the disappearance of s- movable should not be wholly lost sight of (cf. fn. 182).

¹⁴ Kretschmer, Einl. 12, believes the word for ‘one’ developed later than the other numerals, and that when the need for it arose, it was simply borrowed from words already existing for ‘alone’, ‘together’, ‘like’, etc. Hirt's answer (IF 22.89–91) that we often find some language groups using several different roots where others use only one, seems hardly applicable to numerals. But Kretschmer's surely erroneous assumption that there are four different stems for ‘one’ in IE (cf. fn. 17) vitiates his conclusion.

¹⁵ Brugmann, KVG 383, explains that the meaning was originally ‘precisely this one and nothing else’, as opposed to ‘both, all three, etc.‘ Solmsen (BPW 26.182) questions this explanation, I think with justice; he himself apparently believes that even in the oino-group the numeral idea came first, but in that case it is hard to see the force of the various suffixes.

¹⁶ A group of suffixes seem particularly likely to be associated with numerals-and allied terms: they include -no-, -qo-, and -wo-, cited here, also -mo- and -to- (to list only those referred to elsewhere in this paper). The original ordinals seem to have included forms ending in all but -qo-: e.g. -mos in ‘seventh’ and ‘tenth’, -nos in ‘ninth’, -wos in ‘eighth’, -tos in a number of other cases. In most of these the consonant is a part of the stem; but -mos and -tos became generalized as suffixes, for superlatives as well as for ordinals (cf., too, Gr. πόστος, Lat. quotus, which ask the question to be answered by an ordinal). Distributives usually show -no- (Lat. bīnī etc.); but Lat. singulī has -go- (which may be compared with -qo-, KVG 393) + -lo-. These suffixes are also used to form adjectives with meanings more or less related to the idea of number or size, as follows: (1) 'Sole'—Skt. ekakas; Gr. μόνος, οος; Lat: ūnicus, prīvus; Goth. ainaha, OHG einig. (2) Part ('right' or 'left')—Gr. σκαιός, λαιός; Lat. scaevus, laevus. (3) 'Whole'—Skt. sarvas, višvas; Gr. ὅλος, σάος; Lat. tōtus, salvus (cf. omnis with -ni-); Germ. alle (<al-no), ganz (Brugmann, Total. 59); cf. Gr. πα̑ς with -nt- suffix. (4) 'Large' or 'small'—Skt. tanikas; Av. kamnas; Gf. στ∊νός, ταναός; Lat. magnus, tantus, quantus, parvus according to Buck 319 (but this seems strange); cf. Skt. iyant-, kiyant- with -nt- suffix. (5) 'Many' or 'few'—Gr. μανός; Lat. paucī, multī. (6) 'Full' or 'empty'—Skt. pūrnas; Gr. κ∊νός; Lat. plēnus. (7) 'Equal'—Gr. ἴσος. These suffixes often alternate with each other: cf. Skt. p ū́rvas, OCS prъυъ: Gr. πρω̑τος: Lat. prīmus, Lith. pìrmas: Lat. prīscus; also Skt. tanukas: Gr. τανααός.

¹⁷ Kretschmer (Einl. 10) takes the three groups just enumerated as from three separate stems, disconnected from one another as they are from sem- (cf. fn. 14). Br.-Th. 249 omits mention of ékas but joins the other two, and Buck 229 and KVG 383 join them all. KVG adds to the group ἰός, which is classed with sem- by Kretschmer (Einl. 10 fn. 3) and listed separately by Br.-Th. and Buck, also by Berneker 432; on this see fn. 18.

¹⁸ As long ago as 1859 Meyer (KZ 8.139–40) correctly explained this word as connected with Lat. is and ita, but there have been numerous attempts to derive it from sm-. See Schmidt, KZ 36.391–9 for a résumé and a refutation of these early efforts, as well as of other theories concerning the etymology of the word in question, followed by an attempt of his own to prove the same thing in a new way. His reason for believing that ἴα must be a derivative of sem- is the fact that every language has only one word for each number (301). Why we must posit for the cardinal ‘one’ something that is assuredly not true of the ordinal ‘first’ (cf. inf. 93) is not clear; nor is the statement strictly accurate, since later μόνος became a synonym for ∊ς. His argument in favor of a declension μία ία̑ς ἰ μίαν is highly ingenious and free of the unsound methodology of his predecessors; but Brugmann (KVG 383 fn. 1) seems to dispose of it effectually by pointing out that the substitution in early hexameters of forms of a different pronoun such as ἴα for *μμιη̑ς and *μμιη̑ was originally a matter of metrical necessity. Meillet (MSL 15.197–8) expresses at least partial agreement with Schmidt, and Sch. (1.588) does not decide between his theory and the usually accepted one of a connection with is, except to say that the latter is more commonly held; but most authorities (e.g. Bois. 232, Br.-Th. 249) definitely pronounce him wrong.

¹⁹ Cf. Buck, CP 1.409–11. Meillet (MSL 15.197) to me seems wrong in questioning the existence of any relationship between the pronoun and the numeral.

²⁰ Cretan ἴττον ‘one’ (Hesych.) is probably a formation from this stem after the analogy of the multiplicatives διττός ‘double’, τριττός ‘triple’, etc. Cf. Solmsen, BB 17.335 fn. 1.

²¹ Dem. 103, Grund. 2.2.344, 355, KVG 317; Br.-Th. 283.

²² E.g. Bois. 731, Buck 224, Sch. 611; cf. MV 604 on hic.

²³ From our root words meaning ‘self’ include OIr. -som, OCS samъ, and, in my opinion, -met, treated below. In late Lat. ipse ‘self’ and īdem ‘same’ are interchangeable (SS 479–80). Ipse itself may belong to the -se- stem; for if there is really, as is suggested by EM 781, a connection between -pse as seen in ipse and -pte as seen in τίπτ∊ (cf. EM 172 and 709), we probably must split them into p + se and p + te respectively, the p being the particle seen separately in quispiam, quippe, etc.

²⁴ Seen also in ὁμαλός, similis, etc.

²⁵ Johansson, IF 2.7; Lidén, Studien 55 and fn. 1.

²⁶ Cf. EM 1084, where it is said that ūnus supplanted sem- ‘one’ and was itself eliminated by sōlus ‘sole’.

²⁷ Cf. fn. 9.

²⁸ Cf. Junker, KZ 43.342. Wackernagel, KZ 25.280, thinks we have οἰο- in the sense of ‘same’ in Homeric οἰέτ∊ας (or οἰ∊τέας as he writes it) ‘eiusdem aetatis’ (Iliad 2.765), which he explains as for οἰ ο ∊τέας. (However, the explanation of Schmidt, KZ 36.397–8, that οἰ-here is a lengthened form of ỏ- metri causa has been generally accepted—e.g. by Bois. 689.)

²⁹ Similarly Brugmann, Total. 48–9, tries to connect sōlus with solidus and sollus, saying, ‘Die Wörter sollus salvos sárva-s u.s.w. gehen auf ein solches Beisammen, bei dem nichts fehlt, dagegen sōlus auf ein solches, bei dem nichts hinzukommt.‘ This explanation may well be applied to overlapping of the notions of solitariness and totality, even though in regard to the etymology of sōlus it probably is not acceptable and certainly has not been accepted.

³⁰ Such combinations or contaminations of two equivalent suffixes are by no means uncommon. Cf. Skt. višvas ‘whole’, from wi- + -ko- + -wo- (Brugmann, Total. 75), and Gr. ἐλαίν∊ος ‘of olive wood’, Lat. eburneus ‘of ivory’, from -eyo- as seen in Gr. χρύσ∊ος and Lat. aureus ‘golden’, and -no- as seen in Gr. λίθινος ‘stone’ and Lat. fāginus ‘beech’ (Buck 318).

³¹ On this see Wackernagel, KZ 25.262; Kretschmer, KZ 31.444; Solmsen, Untersuchungen 180–6, 302–9; Brugmann, RhM 62.635. Meyer seemed on the right track much earlier (KZ 8.143).

³² Brugmann's argument based on this etymology is invalidated if we accept the very plausible suggestion made by Schmidt in his edition of Hesychius ad loc. (3.69), that μαν∊ύ∊ται is a corruption of μαρύ∊ται (presumably Doric for Attic μηρύ∊ται, which suits the sense admirably).

³³ Cf. Kretschmer, Einl. 236 fn. 3, and Glotta 2.318; Fraenkel, Glotta 2.37; Br.-Th. 157.

³⁴ The stem presents some problems (cf. EM 587). The derivation of both Gr. μανός and Arm. manr from men- seems to be generally accepted (Bois. 608, WP 2.266, EM 586). Arm. manr is associated with Lat. minuō by Meillet, MSL 8.164, and with Lat. minor by Bugge, KZ 32.18. But WP trace minuō and minor to mei- mi-neu- (242), and definitely disassociate the former from men- (267), while EM similarly (despite Meillet in MSL) assign minuō to minu-, but take minor as the result of contamination of menu- and minuō (586).

³⁵ ‘Few’ or ‘rare’ is not the same as ‘isolated’, offered by KVG (132) for μανός. Similarly in WP 2.266, the secondary meaning ‘vereinzelt’ offered for men- seems added simply to fit Brugmann's theory of a common origin for μανός, μόνος, and βáναυσος, which, strangely, WP appear willing to accept (266–7), as also (at least so far as μανός and μόνος go) EM 586.

³⁶ Bonfante declines to accept the etymology sṃ- > sa- because, since IH ṇ becomes Hitt. an (cf. HG 104–5), he believes IH ṃ must become Hitt. am (RB 18.389). There is force in his argument (ib. fn. 2) that the development of the two syllabic nasals is parallel in all the IE languages; yet so good a linguist as Pedersen deems a lack of parallelism not impossible (Hitt. 90), and it is probably dangerous to demand an invariable uniformity for vocalic m and n that certainly does not exist for consonantal m and n. Of Sturtevant's other examples (HG 105) for the passage of ṃ to a in Hitt., one, taswanz ‘blind’: Skt. tamras, Lat. tenebrae, is not particularly compelling, but the other, katta: Gr. κατά, seems to me completely so; and as to Bonfante's attempt (RB 389 fn. 4) to do away with a syllabic nasal in the original form of these words, I can say only that I find this etymology, as he himself does Sturtevant's for taswanz, ‘fort peu attractive’. Bonfante (390) equates Hitt. sa- with Gr. ὀ- seen in two sets of words: (1) several in Hesychius, where ὀ- is equated with ὁμο-; (2) several in Homer, e.g. ὄπατρος, ο(ἰ)έτης (cf. sup., fn. 28). This is quite in line with the view of Brugmann, who (Total. 49) equated the prefix ὀ- with s - s - (incidentally the element which, as already mentioned, he considers present in our sem-). But there seems to me a serious difficulty in the fact that all the ὀ- words are cited with a smooth breathing (Bonfante attributes the lack of aspiration in the Homeric forms to the Aeolian dialect of the first redaction, but ignores the question as applying to the glosses of Hesychius). I think it more likely, then, that the IE prefix was not so- but o-, as it is given by Bois. (681) and WP (1.95–6); and this is plausibly equated not with Hitt. sa- but with Hitt. ha-, as is done by Sturtevant (Lang. 15.147, 150–4; note especially his identification, 150–1, of Hitt. hasdwer with Gr. ὄζος, regularly cited as an example of Gr. ὀ-).

³⁷ On this verb see further below.

³⁸ Friedrich (Kl. Beitr. 33 fn. 2, and Idg. Jb. 21.398) and Sommer (Bil. 77 fn. 3) decline to accept Götze's thesis, but do not state what in their opinion is wrong with it; Goetze's rejoinder (Tunn. 127) seems to me cogent. Sturtevant (Gl. 132) and Couvreur (Hett. Ḫ 178) accept his view, the latter commenting that the derivation of sa- from sṃ- and of sanas from sṃno- is adequately established by the combined evidence of the examples, though these are uncertain when viewed separately.

³⁹ He there spells it sannas, but in Tunn. 127 adopts the spelling sanas (advocated by Sturtevant, Lang. 12.119 fn. 76).

⁴⁰ It is thus an even closer parallel than Götze realizes for kwapi, which he offers as a comparable formation (189). Cf. further below.

⁴¹ In the case of indefinites from the kwi- kwo- stem, forms ending in -ki (e.g. kwapiki) are commoner in conditions than reduplicated forms; but the latter are occasionally found (see Hahn, TAPA 68.394–5; also 64.34–5 on the interchangeability of indefinites in general).

⁴² Cf. Sturtevant, Lang. 10.269, on the spelling ap-pí-iz-zi-iš for apezis.

⁴³ Cf. WP 2.55–6; Hirt, Ind. Gr. 3.319–20.

⁴⁴ We find sanas itself written 1-iš (i.e. sanis) in KBo. 5.2.3.41, KUB 30.15.1.34 (I am indebted to Professor Sturtevant for these references). Cf. also *karuwilas beside *karuwilis (Gl. 74), hantesiyas -an beside hantezis -in (HG 179–80), etc.

⁴⁵ Cf. Sturtevant, Lang. 10.273.

⁴⁶ This meaning it seems to me fits into the passage where Sommer, rejecting Götze's view, calls sannapilis ‘unklar’ (Bil. 77 fn. 3): Bo. 2314.2.15 ff. ‘Kessi roves around 3 years in the mountains, and (remaining) all alone, he never goes back to the city’ (sannapilis might stand with somewhat better logic in the first clause, but is not seriously out of place in the second).

⁴⁷ Friedrich (Idg. Jb. 21.398), who does not accept Götze's explanation of sanas as meaning ‘one’, seems to take ‘empty’ as the prime meaning of sannapilis. Certainly the word has this meaning in the passage cited by him (KUB 12.11.4.1 ff.), and also in 31.68.40 (a passage brought to my attention by Professor Goetze, as was also the following one). The verb sannapilah- means ‘empty, make empty’ in KUB 31.71.4.11, and sannapiles- might be similarly taken in its second occurrence ('let his house be emptied of men, cattle, sheep'), though certainly not in its first, in the simile from the military oaths (KBo. 6.34.3.30–5) discussed by Götze (Lang. 11.188). The semantic development is not quite clear: the shift from 'lonely, left alone' to ‘left empty, empty’ suggested by Götze (ib. 187) may fit the passage to which he applies it (KUB 5.7.2.16), where it is used of an altar, and also 31.68.40, where it is used of a barn, but not so well the passage cited by Friedrich, where it is used of a goblet. Perhaps, however, we are to regard the root meaning not so much as ‘lonely, deserted’ but rather as ‘sole, single, simplex’; to associate this with ‘empty’ may be as natural as to associate ‘many’ with ‘full’ (WP 2.63–5).

⁴⁸ The parallelism is not perfect, to be sure, for the Gr. and Lat. verbs mean ‘divide in two’, the Hitt. ‘divide in one's’. But various shades of meaning may of course be combined (note that διπλόω and duplicō may also mean ‘multiply by two’; so, too, Eng. double, double up); and the meaning ‘tear apart, tear’ is cited for duplicō.

⁴⁹ Götze's parallel (188 fn. 9) of prīvus and prīvāre does not seem to me very apt, since ‘first’, especially in the sense of ‘foremost’ (cf. EM 773), hardly behaves in the same way as ‘one’.

⁵⁰ Taken by Sturtevant (HG 205) as a dative; Hrozný (BoSt. 3.202 fn. 2) and Götze (AM 222) call saneta an instrumental.

⁵¹ Sturtevant there spells the verb sanai, and for the sake of uniformity I adopt this orthography throughout, though he has san(n)- in his Glossary (133) and Götze uses sann-. In connection with the derivation of the verb, it may be noted that we also find saneta nai-as a special phrase (AM 222), but this need have no bearing on the case.

⁵² If Lat. taceō is really akin to Gr. πτήσσω ‘crouch’, as Saussure has suggested (Mém. 285), that might furnish a parallel of a sort; but such a connection is extremely doubtful (cf. EM 970, WP 1.703).

⁵³ Cf. Friedrich, Vert. 1.43.

⁵⁴ See WP 2.494–5. A connection of sōlus with se- has also been suggested (ib. 458).

⁵⁵ So Sturtevant, Lang. 10.269, 271 (despite HG 155).

⁵⁶ Cf. fn. 54.

⁵⁷ He is questioning Ehelolf's translation ‘süsse’ in Bo. 706.2.25 (OLZ 36.3–4). I think he is right: here sanezis is used of sleep or dreams, and means ‘delectable, precious’, with ‘sweet’ only in a metaphorical sense. The word seems to be particularly often used of foods (like Eng. delicious, luscious). Some kinds of food when excellent may suitably be described as ‘sweet’, and this is a particularly natural rendering in KUB 27.29.2.18, where a comparison of words and honey is involved (cf. Ehelolf's translation, ‘süss wie Honig’, OLZ 36.5, followed by Zuntz, Ortsadverbien 62). But the word by no means applies in every case; cf. in particular KUB 13.4.4.67 and 71, where the reference is clearly to meat.

⁵⁸ Cf. the distinction in meaning between e.g. Lat. suprēmus and summus, postrēmus and postumus.

⁵⁹ This is apparently a distinction unknown to Hitt., which has no comparative degree and few if any signs of a dual. Cf. the discussion by Bartholomae, IF 22.95–116, in regard to comparable conditions in early Indo-Iranian.

⁶⁰ Cf. Lat. secundus and alter, Fr. deuxième and second, for ‘second’.

⁶¹ ∊ς, μόνος, ἰός—cf. sup.

⁶² Like Toch. A sas, B ṣe.

⁶³ Cf. sup., fn. 16.

⁶⁴ See Berneker 262.

⁶⁵ Discussed above.

⁶⁶ Brugmann, IF 21.8.

⁶⁷ Av. mat is ambiguous (cf. Brugmann, Grund. 1.2.738). On the connection of our stem with π∊τά and mit, as well as with Lat. -met, see below.

⁶⁸ Probably to be explained as due to sandhi (cf. HG 140). Or is the real explanation that the s- movable that our stem unquestionably had in IE was generated by a quite different cause, and did not as yet exist in Hitt.? In other words was Brugmann right after all in his view (IF 37.160–1) already referred to, that IE sem- is a combination of the roots se- + -mo-? Were they still separate in IH (in which case Pedersen and Sturtevant are right after all in positing a Hitt. indefinite stem ma-)? On this see further below, Excursus.

⁶⁹ E.g. Goth. dat. þamma; Lith. dat. támui, loc. tamè, OCS loc. tomъ (Dal, however, would separate these, NT 9.192 fn. 1). On the whole subject, see Brugmann, Grund. 2.2.308, 362–3; on Skt. in particular, Wackernagel 3.434, 499–501; on Gr., Sch. 1.601, 609–10, and Dal, NT 9.199–200.

⁷⁰ Brugmann, Grund. 2.2.362.

⁷¹ Many of the views expressed in this article (which dates back to 1927) he tells me he no longer holds; but some of them seem to me to call merely for modification rather than for total rejection.

⁷² So earlier Möller, KZ 49.231. Dal agrees with Hirt in both respects, NT 9.202–4.

⁷³ Or -me as it may still have been (cf. sup., fn. 68, and inf., Excursus).

⁷⁴ Friedrich, Vert. 2.92 fn. 2; Götze, NBr. 35–6; Sturtevant, Lang. 10.272–3; Sommer. Bil. 164–6; Pedersen, Hitt. 67–8.

⁷⁵ Sturtevant, Lang. 10.272–3, questions the form masiyas.

⁷⁶ He is a little dubious about masiwanz, but decides in its favor (Hitt. 67).

⁷⁷ His spelling is of course ja, which I am changing in the interest of uniformity.

⁷⁸ Questioned, however, by Sturtevant; cf. fn. 75.

⁷⁹ Cf. (with Sturtevant, HG 161 fn. 54) Skt. iyant-; also Hitt. humanz, Gr. πς (see fn. 16).

⁸⁰ Cf. (again with Sturtevant, op. cit.) Skt. tāvant-. Götze (NBr. 35) classes it with -wa- formations, but see Sturtevant, HG 160 fn. 53.

⁸¹ The existence in Hitt. of -si- as well as -sa- is not surprising; see fnn. 44 and 45.

⁸² Cf. fn. 16.

⁸³ It should be pointed out, however, that in Hitt., which differentiates subordinate and coordinate clauses far less sharply than do the younger IE languages, there is not such a clear-cut distinction between indefinite and relative as we think we observe.

⁸⁴ Cf. Hahn, TAPA 64.33–8, 68.394–6. We have already had occasion to note the use of the indefinite sannapi sannapi in what is probably iterative reduplication.

⁸⁵ Because there is no direct connection between the mases clause and the clause following. However, from just such stuff as this, whether the indefinite was masis or the commoner kwis, grew the quasi-relative ‘as to the fact’ clause.

⁸⁶ He seems tacitly to imply, as I firmly believe (cf. sup., fn. 8), that the fully-developed relative use presupposes, for masis as for kwis, a preliminary parallel indefinite; yet cf. Bil. 166 fn. 2.

⁸⁷ Unless daitta is supposed to hark back to the very outset of Hattusilis's career, long before the period designated by tittanut (66), which seems most unlikely.

⁸⁸ Another parallel (62–4) is discussed in my next paragraph.

⁸⁹ Cf. fnn. 75 and 76 on the doubts expressed by Sturtevant and Pedersen as to some of the forms that have been cited.

⁹⁰ Götze comments on this parallelism, NBr. 36.

⁹¹ This interpretation is based on my opinion in regard to the meaning of UL manka, on which see below.

⁹² Cf. the use of quanto without a balancing tanto, Plautus, Rud. 1301.

⁹³ The text here has masiyante, which Götze (NBr. 36; cf. fn. 1), Sommer (Bil. 165 fn. 1), and Pedersen (Hitt. 67) all alter to masiyanki; Sturtevant, however, (Lang. 10.272) keeps masiyante, explaining it as a neuter plural, and translating in a manner which, despite Pedersen (Hitt. 67), seems to me to make perfectly acceptable sense. According to Sturtevant, they drink ‘as much as’ seems good to them (quanta); according to the others, ‘as often’ (quotiēs). The combination of the neuter plural masiyante and the neuter singular assu is not impossible; cf. HG 164–5.

⁹⁴ Baunack (loc. cit.) thought the α arose in numeral adverbs such as ἑπτάκι(ς) and δ∊κάκι(ς) (237), and spread thence to πολλάκι(ς) (239) and quantitative adverbs in general (241); but this would not account for the occurrence of κ instead of τ.

⁹⁵ I hope to treat all of these more extensively elsewhere: man and mahhan in the course of a monograph, manka in a separate article.

⁹⁶ Asseverative μά regularly follows νή or οὐ except in Attic (LS 1.1070).

⁹⁷ On the relation of -ma to Skt. sma and also to Gr. μέν, cf. Sturtevant, JAOS 47.180. We may add to the parallels adduced by him the fact that both -ma and μέν are practically sentence-connectives. Cf. further on this fn. 112.

⁹⁸ See Sommer, BoSt. 10.6.

⁹⁹ The Celtic conjunction (though Pedersen himself gives a quite different explanation of it, KG 1.441) is probably, like Skt. ná-vā, Av. nava, from the negative stem ne- seen in Lat. neque (WP 2.319) plus the particle u, we seen in Lat. nēve (ib. 1.189). In that case the negative idea has faded out (cf. inf., fn. 107); such loss is well attested in Irish (see Thurneysen, Hdb. 294 and 500), but not in Hitt. Granting its possibility, however, we might explain the first element of nassu as derived from IH nъ precisely like that of natta (HG 99, 132); the final element is readily enough taken as the particle u, which certainly exists in Hitt. (HG 101, 110, 113–4, 214). But this still leaves the -ss- unaccounted for: Pedersen (loc. cit.) simply ignores this element. A real Hitt. parallel for the Celtic conjunction would, it seems to me, be not nassu but nu (for the variation in meaning between ‘and’ and ‘or’, cf. fn. 102). No one so far as I know has proposed breaking up this stem, but if it were to be done I should rather regard the n- as from the pronominal element no- (WP 2.336–7) than from the negative (cf. my suggestion in regard to su, inf. fn. 195).

¹⁰⁰ Here as elsewhere the na- is presumably to be accounted for as the result of assimilation, or, perhaps more probably, of analogy (possibly the influence of false division of nasta).

¹⁰¹ In Code §5 we actually have man and nasma as variants (Hrozný, CH 4 fn. 19).

¹⁰² Like Gr. δέ, nu may mean ‘but’ as well as ‘and’. As for the variation between ‘and’ and ‘or’, cf. Hahn, Lang. 12.111 fn. 15.

¹⁰³ Cf. fn. 68, and Excursus.

¹⁰⁴ See Sturtevant, Gl. 57, and the references there given.

¹⁰⁵ Accepted by EM 455 and WH 1.682,

¹⁰⁶ Of course he has no thought of the initial (s), which would in any case have been lost in Lat.

¹⁰⁷ To this -mo he also attributes Skt. mā, Gr. μή, on the ground that both affirmation and negation may come from the same asseverative particle. This is by no means impossible; the two ideas are not so remote as may at first thought appear. The notion of smallness may develop into that of negation: cf. inf. fn. 132, and observe how Eng. few, little, Fr. peu, lean toward the negative, while a few, a little, un peu lean toward the affirmative. So, too, the notion of badness: cf. Lat. male (almost = non) as in Horace, Carm. 1.9.24, and its derivatives Fr. mal, Eng. mal-. A particle may either lose the idea of negation (as Ir. no, on which cf. fn. 99) or acquire it (as Fr. pas, point, rien, etc.); the latter is particularly likely to happen to a particle which is in the habit of following a negative to reinforce it, and which ultimately comes to be viewed as a substitute for it (in this connection we may note the occurrence of -man after le in Hitt., KUB 1.16.3.65; see Sommer, Bil. 189). The development may also be through the interrogative, since a questioned negative becomes an affirmative and vice versa: cf. Gr. μή (on this see fn. 180), Lat. nōnne and also perhaps -ne itself, which it is possible to assign either to the negative stem seen in Lat. ne- nē, or to the affirmative stem seen in Skt. ná, Gr. vή and ναί (WP 2.336–7, EM 627–8). Indeed, it seems to me not impossible that all these words are to be ultimately combined, in other words that there may have been really only a single ne- no- instead of the two recorded by WP: the negative ne, nē, 2.319–20, and the pronominal no-, 2.336–7 (whence possibly Hitt. nu; cf. fn. 99), to which it seems customary to refer practically all the non-negative n- derivatives except the Celtic ones (cf. again sup. fn. 99, and also the comment in WP 2.319).

¹⁰⁸ Brugmann, Grund. 2.3.810–11. See especially for Gr. Br.-Th. 620–1, for Lat. SS 575.

¹⁰⁹ See Friedrich, Vert. 1.163–4.

¹¹⁰ This will be discussed below.

¹¹¹ Despite Friedrich, KIF 1.286 ff, and Götze-Pedersen, MS 59, I believe with Sturtevant, HG 88, 132, 135, that the particle and the conjunction are the same word, and shall try to prove this in my monograph (cf. fn. 95). Here it suffices to say that Pedersen's refusal to associate Hitt. ma (which he takes, MS 59, as the basis of the particle man) with μέν because of a difference in use (57–8), or Hitt. man the conjunction with μήν because of a difference in meaning (59), seems to me unfounded: cf. the many divergent uses and meanings of particles derived from the kwo- stem.

¹¹² Probably its original function was simply to emphasize the meaning of the word to which it applied, precisely as did its synonym ἄν (MV 241–2; cf. 185–6, 187) and its cognates Skt. sma and Gr. μέν (Brugmann, Grund. 2.3.2.1008; Delbrück, ib. 4.2.507). Later this man may, I believe, have developed as μέν did into a sentence connective (Brugmann, Grund. 2.3.2.1008–9; Delbrück, ib. 4.2.510). Friedrich (KIF 1.293) states that asyndeton is a peculiarity of the particle man, and explains this (295) as a remainder of the earlier asyndetic style, but he does not account for its persistence with man in particular, and the true explanation for the absence of any sentence connective beside man may be that it fulfilled this function itself, as -ma often did (cf. Götze-Pedersen, MS 59).

¹¹³ Cf. Götze apud Sturtevant, Gl. 87. Often kwatka accompanies man the conjunction (cf. Lat. sīquid), and sometimes when standing alone seems almost synonymous with it (cf. fn. 107).

¹¹⁴ Pedersen (MS 59) connects the particle man and ἄν etymologically, taking man as from ma + a particle an denoting unreality. But the idea of potentiality, hence in the past of unreality, lies rather in the element ma- (cf. fnn. 112 and 113).

¹¹⁵ Cf. Götze-Pedersen, MS 56, on kwitman.

¹¹⁶ Cf. the correlative use of sī ... sīc in Lat. (SS 772).

¹¹⁷ We also often have two conditional man's used side by side, like Lat. sīve ... sīve, to be translated by 'if ... or if', or by ‘whether ... or’ (note the use in Eng. of the kwo- stem in whether as a correspondent for man from the sem- stem).

¹¹⁸ As man corresponds to cum in formation, so does manka to the indefinite particle cumque (Horace, Carm. 1.32.15). I shall revert to this below.

¹¹⁹ However, unlike cum, man does not seem to have meant ‘since’. The Hitt. causal conjunction is kwit ‘because’, corresponding in use to Lat. quod.

¹²⁰ I think this shift probably began in references to the future, the period in which we cannot distinguish between the temporal and the merely conditional (cf. the double use of Germ. wenn for this period), and that we can see it in process of taking place in the early Bilingue; this I hope to discuss elsewhere.

¹²¹ However, the one example from classical Lat. that Hofmann cites, Cicero, De Off. 3.24.93, seems to me doubtful, for there I think cum is explicative rather than conditional; but the example from the Itala (Matth. 5.23) that he offers is indubitable. Conversely, si has been thought to be used for cum, and sin for sed cum, in Vergil (Aen. 5.64 and Georg. 3.504 respectively); but see Hahn, Coord. 189.

¹²² At least to judge by the Code, where the only temporal conjunction is man, and mahhan is comparative only. But Sommer seems uncertain (Bil. 166 fn. 1).

¹²³ Cf. fn. 117.

¹²⁴ Sturtevant's postulation of an interrogative man (Lang. 14.241 fn. 11a) is due to a misreading of Sommer.

¹²⁵ A synonym (probably later according to Sommer, Bil. 42 and fn. 2) is iwar. This looks like a verbal noun (cf. HG 152–4, 267), and so Lat. īnstar (with which Sommer compares it, BoSt. 7.22) is indeed an admirable parallel, being, according to WH 1.705, itself an old infinitive īnstār(e).

¹²⁶ However, compare the comparable development of quōmodo in Romance, giving It. come, Fr. comme, etc. (EM 591, REW 6972). Greek and Latin do possess quasi-prepositions with the same meaning in δίκην and īnstar (on the latter cf. fn. 125), but these are from a noun and a verb respectively, not from a pronominal stem.

¹²⁷ I know of no sure instance of this use of mahhan; but doubtless in such a case as ŠAḪ GIM-an in Hatt. 4.26, in view of the fact that almost without question the conjunction GIM-an everywhere in the period when this was written stands for mahhan, we may safely assume that the postposition does so also.

¹²⁸ Cf. fn. 120.

¹²⁹ Just as when Lat. dum ‘while’ acquired the meaning ‘until’, dōnec and quoad ‘until’ soon acquired the meaning ‘while’ (SS 744, 754, 768).

¹³⁰ See Sturtevant, Gl. Sup. 30, and the references there cited.

¹³¹ This element, like -quam in Lat., is particularly likely to be added to indefinites in negative and interrogative sentences. This is quite in line with the fact that in all the occurrences known to us manka follows UL.

¹³² Possibly for the idea of ‘little’ as reinforcing a negative we might compare Eng. not a bit, not in the least, Fr. pas, point, etc.; yet we must remember that words denoting smallness are themselves likely to be negatives (Lat. minus, minimē), not additions to negatives.

¹³³ The -n in man is difficult, as is the -n in mahhan too—for Sturtevant's assumption (Lang. 14.241 fn. 11a) that mahhan is the accusative of a noun from the ma- stem + -hh-does not provide a wholly satisfactory solution of the problem. It certainly does apply in such instances of feminine accusative adverbs and prepositions as ἀρχήν and perhaps δήν (Br.-Th. 294), δίκην (cf. fn. 126) and χάριν (Br.-Th. 524); but nouns from pronominal stems are harder to find. Assuredly the last word has not yet been said about such forms as Lat. quam, tam, -dam, and nam, or, for that matter, such forms as cum, tum, dum, and num. But any explanation for Hitt. man that does not at the same time take cognizance of Gr. μέν and Lat. cum is bound to seem inadequate.

¹³⁴ Corresponding words in other languages may show a demonstrative stem (dum, dōnec), but also a relative one: yo-, as Skt. y vat, Gr. ἕως and ὄφρα; kwo-, as Lat. quoad and usque, Eng. while; or our own sem-, as Gr. μέχρι and possibly ἄχρι (cf. fn. 178).

¹³⁵ An interesting variant is the reduplicated KASKAL-ši KASKAL-ši-be after masiyanki (or masiyante; see fn. 93) in KBo. 3.5.2.14, corresponding to totiēs after quotiēs.

¹³⁶ I think there can be no doubt (despite Bois. 168) that δέ is from de- do- (cf. WP 1.769–70).

¹³⁷ The efforts of Pott and Benfey to explain μέν and δέ as originally meaning ‘one’ and ‘two’, i.e. ‘first’ and ‘second’, respectively, are refuted by Meyer, KZ 8.144–7; and I know of no other attempt to account on etymological grounds for this repeated collocation.

¹³⁸ So μέν surely. This is another bit of evidence that the relative comes from the indefinite, not from the interrogative. Cf. fnn. 8 and 86.

¹³⁹ We do find in Gr. ὁτὲ μέν ... ὁτέ δέ beside τοτέ μέν . . . τοτὲ δέ, but here the stem, as in ὁ μέν . . . ὁ δέ, is supposed to be demonstrative, not relative (v. Br.-Th. 282).

¹⁴⁰ Once more cf. Hahn, TAPA 64.28–40.

¹⁴¹ E.g. in KUB 10.52.1.8; see Sommer, Bil. 165.

¹⁴² V. Sch. 1.617 and fn. 1.

¹⁴³ Actual reduplication is just one phase of a more widespread phenomenon. This may help justify my view (TAPA 68.401–2) that the two kwis‘s in Hitt. kwis kwis, or the two quis‘s in Lat. quisquis, are not necessarily identical with each other.

¹⁴⁴ See Hahn, TAPA 64.38 and fnn. 77 and 78.

¹⁴⁵ Eng. ‘one’ also figures in many colloquial and tautological phrases in combination not only with sem- but with other indefinites and quasi-indefinites as well—as all alone, all sole alone, all by one's lonesome, all by oneself, one and only, one and the same, etc.

¹⁴⁶ I regard -dem rather than -em as the suffix, but this has been much disputed; see e.g. EM 801.

¹⁴⁷ We may compare forms that show a doubling of demonstrative stems, as ὄδ∊, tandem, or even a tripling of them, as οτος.

¹⁴⁸ See Hahn, TAPA 68.390 and fn. 6.

¹⁴⁹ Verbs, too, though they fuse with pronominal stems less readily than substantives, do occasionally appear, as in Lat. nesciōquis, quīvīs and quīlibet, Fr. je ne sais quoi, n'importe quoi, Eng. anything you like, whatever you please. Cf. Rogge's suggestion, discussed above, that the first element in imm is ī.

¹⁵⁰ See Kluge 215, 218.

¹⁵¹ Is this an argument in favor of Brugmann's view? (Cf. fn. 68.)

¹⁵² Cf. NED 9, Part 2, 138; 10, Part 2. Wh 89, 100–1.

¹⁵³ On the -ku suffix see Hahn, Lang. 12.110 fn. 14, TAPA 68.392 fn. 19.

¹⁵⁴ See Götze, ZA NF 2.268; Sommer, AU 265.

¹⁵⁵ KUB 5.1.1.60, 79, 88. Cf. Sommer, Bil. 165.

¹⁵⁶ See the interesting and illuminating treatment by Delbrück, Grund. 4.2.507–11 (already referred to above, fnn. 7 and 112).

¹⁵⁷ Meyer did so take it, KZ 8.146, but no one else so far as I know.

¹⁵⁸ To be sure, they do not occur so early as the first and second person pronouns, as they appear first in Sallust and Catullus respectively (Sommer, Hdb. 449); but ipse, which really provides a sort of nominative for sē, corresponding to it as does ego to mē, is found with -met as early as Plautus (ib., also SS 479; cf. inf., fn. 168).

¹⁵⁹ Cf. Rogge's view, referred to above, that the -m- of immō was lengthened to avoid confusion with the adjective īmō.

¹⁶⁰ This may be determined at times by Volksetymologie or by grammatical purism rather than by actual historical conditions. Compare the debate that has been waged over the first element in equidem, which involves not only the question whether it was or was not ego, but also the question whether the Romans used the word as if it was, and why.

¹⁶¹ E.g., Brugmann, Grund. 2.2.382 (with reservations); Walde 481; Kieckers, Sprachw. Misc. 4.45.

¹⁶² Meillet, who considers the assumption of egom too doubtful (MSL 20.176), agrees that the second element of -met may be et(i), and posits a particle m- as the first element (ib.); but admittedly (cf. EM 580) this is far from certain.

¹⁶³ Lindsay, Synt. Plaut. 40; Nazari, RFC 44.110–1; Cocchia, Saggi 94–5; Sturtevant, Lang. 8.8–9.

¹⁶⁴ Lindsay, indeed, simply ignores them; he says merely that -mět and -tě are repetitions of the first and second person pronoun stems respectively.

¹⁶⁵ Note the objections of Hartmann (Glotta 10.256) against Nazari, and of Leumann (SS 282) and WH (2.80) against both Nazari and Cocchia. Even Sturtevant, though his theory takes care of the final t of -met (cf. WH, loc. cit.), is troubled by the short vowel of -met as he is by that of Skt. mát and of Goth. mis. Lat. mēd of course fits into the picture perfectly; and if we remove Lat. -met therefrom as I think we certainly should, perhaps Skt. mát with which Sturtevant identifies it may go too. This would be a point in favor of Petersen's view (Lang. 6.170) that Skt. mát (so too Av. maṯ) is not a real ablative at all but a mere stem, as seen in the possessive adjectives madyas etc. (Wackernagel, 3.442, 460, states that the resemblance of the two is accidental; but he does not indicate how he would account for the quantity of the vowel in the ablative.)

¹⁶⁶ CIL 1.2.364 (cited SS 283).

¹⁶⁷ However, there is doubt as to the -te of iste too. It is usually taken as from the demonstrative stem (Walde 394, SS 286, WH 1.721); but this is not universally accepted (note Buck 226). Brugmann (Dem. 81) prefers to connect it with the -te of Gr. τότ∊, ἔπ∊ιτ∊ (ἔπ∊ιτι), etc.—a view not commonly shared. Since I consider that this suffix is certainly represented by the -t of -met (cf. inf., especially fn. 174), Brugmann's theory offers a possibility of bringing -met and -te into contact after all.

¹⁶⁸ Lodge, Lex. Pl. 2.46–7.

¹⁶⁹ Lodge, Lex. Pl. 2.483.

¹⁷⁰ Cf. fn. 23.

¹⁷¹ REW 5551. Heraeus (Kl. Schr. 79 fn. 2) suggests that metipsimum arose in mēmet ipsum divided mē metipsum; but in view of the occurrence of ipsemet in Plautus already referred to (fn. 158), this seems hardly necessary.

¹⁷² Cf. sup., fn. 165.

¹⁷³ See Wackernagel 3.467–70, and cf. sup. fn. 69.

¹⁷⁴ This is the suffix that Brugmann thinks occurs in tūte and iste; cf. sup. fn. 167.

¹⁷⁵ I believe that Walde (481), Sommer (Hdb. 449), and WH (2.80) are justified in rejecting Stowasser's attempt (ZöG 52.865–8) to prove -met identical in meaning and use with μ∊τά and mit, for his article seems to me totally erroneous; but that does not preclude the possibility that the words are related in origin.

¹⁷⁶ E.g. Bois. 629, Br.-Th. 508 fn. 2. They are connected by Meyer (KZ 8.138–9), who anticipated so many of the ideas expressed in this paper.

¹⁷⁷ Perhaps also with that of μέσος, medius (WP 2.236).

¹⁷⁸ Cf. fn. 134. Of kindred formation, meaning, and use is ἄχρι. Fick (BB 5.168) traces both words to the same root, μ∊-: ἀ-, as in μέγα: ἄγαν. This connection I believe may really exist, only we must now posit (σ)μ∊-: ἁ-; it is also quite possible that ἄχρι is a parallel formation from the yo- stem like ἄτ∊ (though its comparatively late and incomplete development as a conjunction seems against this—cf. Br.-Th. 637). Whether it is from sṃ- or from yo-, the loss of initial aspiration is of course due to dissimilation.

¹⁷⁹ Cf. fn. 134.

¹⁸⁰ For the possible connection of μή with our stem, see fn. 107. For its use as an interrogative, see Br.-Th. 610; also the extended discussion by Babbitt (Harv. Stud. 12.307–17), who concludes that μή and μν are pure interrogative particles not necessarily expecting a negative answer.

¹⁸¹ Arm. mi and Alb. mos are grouped with Gr. μή by WP (2.236–7) as derivatives of IE mē. On their use see respectively Hübschmann (Arm. Gr. 1.474), who classes mi as negative only, and Meyer (Alb. Wb. 287), who classes mos as both negative and interrogative. Brugmann (Grund. 2.3.827–8 and 974–5) and WP (2.236–7) recognize only the negative use for all three words and for their stem. This stem they give as mē, but it may equally well be smē, since sm- becomes m- in both Arm. and Alb. (see Brugmann, Grund. 1.740–1, 757–8).

¹⁸² He equates this conjunction with Hitt. man ‘if’, MS 58–9. It is highly interesting to note that Corn. and Brit. ma, cognate with Ir. má, may mean not only ‘if’ (Pedersen, KG 2.331), but also ‘where’ and ‘that’ like Lat. ubi and ut from the kwo- stem (ib. 230). If the word is to be assigned, as seems extremely likely, to sm-, we must assume the loss of s movable, since sm- remains in Ir. though not in Brit. (Brugmann, Grund. 1.770, and Pedersen, KG 2.86–7; cf. sup., fn. 13).

¹⁸³ It was originally a preposition only (cf. Solmsen, IF 31.449), and may have acquired its relative use, so rare in words from our stem, only by an interchange of meaning with ἄχρι (see fn. 178) or some other word of kindred meaning (cf. fn. 129).

¹⁸⁴ Cf. Sturtevant, Lang. Mon. 7.141–9.

¹⁸⁵ Possibly some of our Eng. indefinites are closer to relatives than we realize. The use of once as a conjunction has just been cited. Probably most people would class wherever as a relative and everywhere and anywhere as indefinites, yet there is little difference between them. To be sure, we cannot say ‘He is welcome wherever’ as we do ‘He is welcome everywhere’; but we do say ‘Everywhere he goes he is welcome’ just as we say ‘Wherever he goes he is welcome’. And lest it be objected that in such sentences it is where, not every or any, that provides the quasi-relative value, let it be observed that we can say just as readily ‘Every time he comes he is welcome’. It might be argued that we are to supply that just as in ‘Grant him every favor he asks’, whereas in the wherever sentence we cannot (though I have actually heard a that interpolated in sentences of that type in careless speech); this is an indication that the two kinds of expression are not identical. Yet the assumption of an ellipsis does not seem to me a thoroughly satisfactory way to handle problems of the sort.

¹⁸⁶ See fnn. 68, 73, and 103; cf. fn. 151.

¹⁸⁷ Cf. fn. 71. This form, incidentally, would not really involve the existence in Hitt. of -sme; it assumes only -me for that language (cf. inf. fn. 188).

¹⁸⁸ Sturtevant himself so took it once, and proposed that the s- in Skt. sma is due to false division (JAOS 47.180; cf. again fn. 71). Brugmann's explanation of the s- in words from sem- sm-, among which I would include Skt. sma, is quite different; yet perhaps in the welding together of the elements se- and mo- into a single complex which he postulates for IE, the possibility of false analysis of the pronominal forms in -sm-, which were certainly IE, may have played a part. The s- of the not dissimilar pronominal element -sy- seen beside -sm- in Skt. may enter into the picture (cf. Wackernagel 3.501, with references); if so, we may note that the following -y- both in the case ending -sya (ib. 96) and in the demonstrative pronoun sya- (ib. 550) has been identified with the relative stem yo-, which is an interesting parallel for Hitt. ma-.

¹⁸⁹ On this see further below.

¹⁹⁰ Cf. sup. fn. 9.

¹⁹¹ Cf. sup. fn. 68.

¹⁹² For the phonetic difficulties involved in assuming their descent from sem- sṃ-, see above, 94 and fn. 36 respectively. Both the explanation proposed by Professor Sturtevant for sanas, and my alternative suggestion, are undeniably clumsy. As regards sa-, there is certainly, as I have said (fn. 36), force in Bonfante's objection to Sturtevant's view, even though his own counterproposal of an equation with Gr. ὀ- is seemingly invalidated by Sturtevant's plausible equation of the latter with Hitt. ha-.

¹⁹³ Perhaps it is not irrelevant to point out that -no- too may be ultimately connected with a pronominal stem. Cf. fn. 107; also, on the suffixes, fn. 16.

¹⁹⁴ Sommer, Bil. 78; Sturtevant, Lang. 15.13–9.

¹⁹⁵ Cf. sup. fn. 99. Sturtevant (Lang. 15.16) believes that the IH and IE forms were so, but that Hitt. *sa became su under the influence of nu. It would, I think, be much simpler to assume that Hitt. possessed both *sa and *na side-by-side with the composites su and nu; this would clear up the phonetic difficulty involved in the forms sas and nas (Petersen, AJP 58.313, had something of the sort in mind). I hope to treat this point at greater length elsewhere.

¹⁹⁶ KVG 426, 650; WP 2.509. The Gr. seems indubitable; but on the Indo-Iranian v. Wackernagel 3.529.

¹⁹⁷ Sturtevant, HG 199–200, Lang. 15.13–9.

¹⁹⁸ Götze, Madd. 137; Sturtevant, HG 207; Sommer, Bil. 78. The position in the sentence of setani in Madd. 2.36 precludes any direct connection with su. This seems to me an indication that Sturtevant is wrong in maintaining, as he does at least in regard to ta, that we are not justified in recognizing the demonstrative stem as already existing in the IH and Hitt. conglomerates; I have elsewhere (Lang. 12.108–9) given other reasons for dissenting from his view.

¹⁹⁹ See Brugmann, KVG 432–3; Buck 220.

²⁰⁰ As may also be that of -za in Hitt.; no IE cognates for it have been recognized. On the other hand if the stem of -za was inherited from IH but lost in IE, the s- stem, starting from reflexive use in the third person, may have come to be used in place of -za as a universal reflexive. A good example of the demonstrative se- acquiring reflexive use in any person and number is seen in Eng. -self, -selves.

²⁰¹ Brugmann, Grund. 2.2.319–20, 395, KVG 433; Br.-Th. 478; Sch. 1.607 fn. 6; WP 2.438. Cf. Götze-Pedersen, MS 67.