¹ For any abbreviations not self-evident, as well as for bibliographical data in general, see Edgar H. Sturtevant, A Hittite (Philadelphia, 1936) and Supplement (Philadelphia, 1939).

² The Anittas goes back to the very early period of Hittite history, when Nesa not Hatti had the hegemony : cf. Götze, Hethiter, Churriter und Assyrer 52 f. and 52 fn. 5 (Oslo, 1936). For the dates of the other three more or less archaic documents, see Hahn, Lang. 20.91 and fn. 4. The documents from the so-called classical Hittite period I group together, since they do not, so far as I can see, reveal any variations in style: for instance, in the Madd., which probably dates from the very end of the Hittite dynasty (see Götze, Madd. 157-9), I can detect no marked differences from the others.

³ These examples are as follows (in each case I add after the citation a key-number in parentheses which I shall use throughout this paper for reference to the example in question; in these references I shall separate examples in the same document by commas, and those in different documents by semicolons).

Anittas: 22 (1), 35 (2), 37 (3), 49 (4), 58 (5).

Bil. :2.66 (6), 3.63 (7).

Tel.: 1.39 (8), 2.37 (9), 38 (10), 40 (11), 50 (12), 56 (13), 61 (14), 4.19 (15), 19 (16), 23 (17).

CH: 1.1.14 (18),59 (19), 61 (20), 63 (21), 67 (22), 68 (23), 2.15 (24), 70 (25), KBo. 6.2.2.58 (26), 58 (27), 59 (28), 1.3.3 (29), 7 (30), 4.53 (31), IA.4.8 (32), 16 (33), 32 (34), 38 (35), II.1.23 (36),39 (37),44 (38), 48 (39), 3.15 (40), 16 (41), 17 (42),42 (43).

AM: KBo. 2.5.1.11. (44), 15 (45), 17 (46), 2.4 (47), 17 (48), 4.7 (49), 3.4.1.8 (50), 10 (51), 12 (52), 14 (53), 20 (54), 23 (55), 32 (56), 2.10 (57), 33 (58), 34 (59), 41 (60), 43 (61), 67 (62), 3.18 (63), 19 (64), 32 (65), 34 (66), 39 (67), 43 (68), 48 (69), 52 (70), 54 (71), 58 (72), 61 (73), 77 (74), 90 (75), 95 (76), 4.41 (77), 42 (78), 46 (79), 48 (80), 4.4.1.7 (81), 2.3 (82), 46 (83), 63 (84), 68 (85), 4.6 (86), 26 (87), 29 (88), 35 (89), 48 (90), 5.8.1.1 (91), 18 (92), 2.14 (93), 16 (94), 19 (95), 29 (96), 32 (97), 35 (98), 3.24 (99), 37 (100), 38 (101), KUB 14.15.2.11 (102), 3.28 (103), 51 (104), 4.21 (105), 14.16.1.25 (106), 31 (107), 3.30 (108), 14.17.2.2 (109), 31 (110), 3.21 (111), 23 (112), 14.20.7 (113), 19 (114), 14.29.1.17 (115), 19.29.4.7 (116), 11 (117), 20 (118), 19.30.1.14 (119), 4.9 (120), 19.37.2.44 (121), 3.8 (122), 36 (123), 39 (124), 42 (125), 45 (126), 19.39.3.7 (127), 10 (128), 12 (129).

Ann.: 2.7 (130), 11 (131), 21 (132), 3.11 (133), 4.29 (134).

AU: Tav. 1.32 (135), 62 (136), 2.7 (137), 73 (138), 3.68 (139), 4.8 (140), 48 (141), 51 (142), Mil. Unterer Rand 2 (143), KBo. 2.11.2.14 (144), KUB 5.6.2.48 (145), 49 (146), 57 (147), 14.2.1.6 (148), 9 (149), 21.38.1.11 (150), 2.1 (151), 4 (152), 23.1.4.1 (153).

BoSt. 3: 2.1.5 (154), 2.32 (155), 50 (156), 51 (157), 54 (158), 3.2 (159), 18 (160), 30 (161), 33 (162), 35'(163), 4.7 (164), 28 (165), 3.3.20 (166).

Hatt.: 1.73 (167), 2.26 (168), 39 (169), 41 (170), 46 (171), 63 (172), 70 (173), 72 (174), 3.6 (175), 46 (176), 48 (177), 57 (178), 70 (179), 4.20 (180), 27 (181), 50 (182), 53 (183), 55 (184), 56 (185), 59 (186), 62 (187), 68 (188), 69 (189), 71 (190), 75 (191), 81 (192), 86 (193), KBo. 4.12.1.28 (194), 6.29.1.15 (195), B.1.15 (196), 17 (197), 2.14 (198), 16 (199), 22 (200), 31 (201), A.3.6 (202), 9 (203), 19 (204), 32 (205), 34 (206), 41 (207), B.4.6 (208).

Madd.: 1.24 (209), 25 (210), 28 (211), 37 (212), 48 (213), 58 (214), 2.34 (215), 56 (216), 63 (217), 85 (218).

MS: 1.18 (219), 2.16 (220), 17 (221), 18 (222), 19 (223), 19 (224), 23 (225), 24 (226), 25 (227).

Pap.: 1.8 (228), 31 (229), 38 (230), 2.15 (231).

Pest.: 1.1.4 (232), 7 (233), 19 (234), 31 (235), 35 (236), 2.22 (237), 29 (238), 43 (239), II.1.2 (240), 5 (241), 3.1 (242), 4.10 (243), 5.3 (244), 8.1 (245), 4 (246), 6 (247), 9.8 (248), 10.4 (249), 5 (250), 7 (251), 7 (252), 7 (253), 7 (254), 11.2 (255), 3 (256), 5 (257), III.I.1.1 (258), III.II.1.12 (259), 19 (260), 33 (261), 33 (262), 39 (263), 40 (264), 41 (265), 42 (266), 4.19 (267), 20 (268). Rein.: 2.16 (269), 17 (270), 29 (271), 3.3 (272), 4 (273), 7 (274), 9 (275), 11 (276), 21 (277) Sold.: 1.15 (278), 30 (279), 34 (280), 2.10 (281), 34 (282), 44 (283), 46 (284), 3.5 (285), 14 (286), 19 (287), 26 (288), 33 (289), 39 (290), 4.7 (291).

Tern.: 1.14 (292), 17 (293), 19 (294), 48 (295), 59 (296), 64 (297), 2.14 (298), 27 (299), 30 (300), 32 (301), 34 (302), 48 (303), 53 (304), 53 (305), 55 (306), 59 (307), 63 (308), 80 (309), 3.13 (310), 16 (311), 19 (312), 21 (313), 41 (314), 46 (315), 49 (316), 49 (317), 51 (318), 55 (319), 4.25 (320), 34 (321), 47 (322), 68 (323).

Tunn.: 1.31 (324), 2.9 (325), 15 (326), 19 (327), 30 (328), 41 (329), 48 (330), 49 (331), 3.6 (332).

Vert.: Dupp. D.1.29 (333), 2.38 (334), 3.3 (335), 7 (336), 9 (337), Targ. 1.25 (338), 33 (339), 34 (340), 43 (341), 2.10 (342), 19 (343), 25 (344), 33 (345), 42 (346), 44 (347), 50 (348), 51 (349), Kup. 3.D.15 (350), 16 (351), 6.D.44 (352), 7.C.15 (353), 10.D.19 (354), 18.C.25 (355), 20 D 11 (356), 27.A.20 (357), 24 (358), 38 (359), 29.G.9 (360), 10 (361), Man. A.1.49 (362), 51 (363), B.3.6 (364), 7 (365), Al. B.1.5 (366), A.1.65 (367), 2.77 (368), 3.10 (369), 31 (370), 36 (371), 44 (372), 45 (373), 48 (374), 73 (375), 4.38 (376), Hukk. 1.9 (377), 12 (378), 27 (379), 2.6 (380), 22 (381), 56 (382), 3.25 (383), 30 (384), 64 (385), 66 (386), 69 (387), 4.32 (388), 50 (389), B.1.15 (390).

⁴ These are all examples of the simple relative kwis. I have excluded the gene'ralizing reduplicated form kwis kwis or kwis imma kwis, which would introduce complications into my study of word order; for my views as to its development from the indefinite, as well as for my belief that in many instances where it has been called a relative it actually is still an indefinite, see TAPA 68.388-402.

⁵ Language Monograph 7.141-9 (Baltimore, 1930).

⁶ In the present paper I am confining myself to a consideration of the kwi- kwo- stem. This investigation also has implications in regard to the relative stem yo-. Sturtevant suggests, op.cit. (fn. 5) 148, that the parent tongue had two relatives kwi- kwo- and yo-, the former used as a generalizing and the latter as a definite relative. I thoroughly agree with this suggestion, and hope to give evidence for it elsewhere. The existence of two relative stems side by side need cause us no surprise : we have a parallel in the German use of the quo- and the to- stems specialized as the generalizing and the definite relative respectively (the English use of that as a restrictive relative reverses the German usage) ; and Hittite evidence has now been recognized for still a third relative stem (also indefinite and interrogative), Hittite ma-, Indo-Hittite ino-, which I identify with the Indo-European indefinite stem sem- som- sm- (see Hahn, Lang. 18.83-116, and, on mo-, the references there cited, 83-5; and note especially 85 fn. 8 on the relationship of the relative to the indefinite).

⁷ See e.g. Hirt, Indogermanische Grammatik 6.159; Meillet and Vendryes, Traité de Grammaire Comparée des Langues Classiques 549.

⁸ This citation is complete; but in general in quoting Hittite passages I shall include only the pertinent parts, with no indication of omissions. Neither shall I indicate plausible and irrelevant suppletions. In the transliteration of all Hittite words cited, h is to be interpreted as . and s as š.

⁹ Literally, ‘this‘—a neuter singular though and ku-e are plural. This inexactness is typical of Hittite (cf. Sturtevant, HG 164-5), whose system of agreement greatly needs a special investigation.

¹⁰ As in Vergil, Aen. 1.573 urbem quam statuo vestra est, and Horace, Serm. 1.1.1-3 ut nemo, quam sibi sortem seu ratio dederit seufors obiecerit, illa contentus vivat (examples respectively of the types which I shall later designate D and G). For the Hittite construction cf. Hrozný on KBo. 3.4.2.41 (BoSt.3.188 fn.5) and Götze on Hatt. 2.63 (Hatt.p.86). The truth of their statements is particularly well demonstrated by the following passages: KBo. 4.4.2.68-70 nu-kán ¹Na-nα-LÚ-in ku~in DUMU.LUGAL ne-eh-hu-un nu u-un-ni-es-ta ‘N. (accusative case) the king's son (after the relative), whom I had sent, came’ (cf. KBo. 3.4.2.67, KUB 14.29.1.17); Dupp. D.2.38-40 ‘the Nuhassian captives (accusative, before the relative) and the Kinzan captives (accusative, after the relative) whom my father had brought, if one of these captives (genitive) flees’ (cf. Hatt. 3.70); KBo. 3.4.2.10-2 ERUM.MEŠ-YA-wa-at-ták-kán ku-i-e-es an-da ú-e-ir nu-wa-ra-as-mu EGIR-po Ú-UL pa-is-ta ‘my slaves who came to you (enclitic ‘to you’ added to the word for ‘slaves') these you have not given back to me’ (cf. Tav. 1.32); Tem. 1.18-20 -ma-as-kân ku-e-da-as an-da-an e-es-sa-an-zi na-at-kán sa-an-ha-an e-es-du ‘in the bakeries in which they prepare them, let it (i.e. the bakery or bakeries) be swept’. But, by a process of combining the type in which the antecedent of the relative is not expressed in either clause, and the type in which the antecedent is expressed in the relative clause and then in the same or a different form is repeated in the main clause, we find a type in which the antecedent, or something very like it, occurs only in the main clause, as Kup. 7.C.15 nu A-BU-ŠU ku-e-el wa-as-ta-i kat-ta-ma DUMU-ŠU Ú-UL wa-as-du-la-as-be ‘whose father is guilty, his (i.e. the father's) son (antecedent of ‘whose') is not correspondingly guilty’ (were the two instances of the possessive -SU omitted, the sentence would seem more logical).

¹¹ I hesitate to say ‘enclitic’, since postposition does not necessarily imply enclisis. Hittite has a number of little words that are undoubtedly enclitics : particles (mainly conjunctions and adverbs) and personal pronouns, which are appended to the first word of the clause (two of them, -ma and -a, may also appear elsewhere), and which, significantly, are written by the Hittite scribes as part of that word. The indefinites and relatives receive quite different treatment.

¹² Greek Latin alius ... alius, are less close to the Hittite, since these locutions involve demonstrative stems instead of indefinite.

¹³ The regular idiom seems to be nu kwis ‘one’ in the first clause (on which cf. fn. 16) and kwisma ‘another’ thereafter; see the examples on pp.83-4. (For this reason in AM KUB 14.29.1.7-8 I would suggest as a suppletion nu-wa ku-it instead of Götze's ku-it-ma-wa.)

¹⁴ Tern. 3.14 nu-za ku-is ŠA KÁ É DINGIR-LIM e-es-du. Here kwis, which is usually found in negative or conditional expressions like Latin quis, is employed more in the manner of Latin aliquis; this may account for its early position.

¹⁵ AU Mil. 2.2, where the text is fragmentary and uncertain; and BoSt. 3 5.4.10-1 nu-za ku-is ku-it ar-ku-wa-ar , where the structure is not clear. Perhaps kwis and kwit are both indefinite, in which case the clause in which they occur must be viewed as an implied conditional ‘ (if) any one makes any request’; if one of them is a relative, I think it is kwit, in view of the following at,‘whatever request any one makes, I hear it’ (otherwise Friedrich, AO 24.3.21, ‘und wer irgendeine Bitte tut, die will ich hören'). We have a similar problem in interpreting Tern. 3.21 ^URUHa-at-tu-si-ma-kán ku-e-da-ni ku-is sa-ak-la-a-is se-ir-; Sturtevant took kwedani as relative in JAOS 54.383, ‘for whatever Hittite (man) a rite (is performed)', but one might equally well reverse this and translate ‘whatever rite (is performed) for a Hittite (man)'; in Chrest. 159 he takes both kwedani and kwis as indefinite, ‘if there is a rite in behalf of anyone in Hattusas’ (if should be enclosed in parentheses).

¹⁶ For the significance of this order, noted already in the group cited in fn. 13, cf. further the next paragraph.

¹⁷ I am considering here only indubitable instances of the indefinite. Many of the pronouns that I am assuming to be relatives might be classed as indefinites. Indeed in some cases where the main clause contains no mention of the relative's antecedent, such a classification might avoid what amounts almost to an anacoluthon (cf. p. 78 f. on 316, and fn. 55 on 325, 326,328, and 332).

¹⁸ There are two possible instances of kwis in the sense of Latin aliquis (cf. fn. 14) : Tem. 1.21, a difficult passage (see Sturtevant, Chrest. ad loc.), and 1.25, where kwit, translated by Sturtevant as ‘when’ in JAOS 54.367 and as ‘something’ in Chrest. 149, may, I think, mean rather ‘because’. And there are 11 certain instances of kwis in the sense of kwiski or Latin quis (in some cases quisquam): 2 with the negative natta, Tel. 2.44 and Madd. 1.69; and 9 in a conditional clause with man 'if, AU Tav. 2.64, Tem. 2.6, 3.64, 68, 74, 4.41, Vert. Targ. 2.4, Kup. 7.C.14, Hukk. 1.27; to which we doubtless may add on the basis of highly plausible suppletions 5 other examples, Madd. 2.45, Pest. II.10.7 (line 1), Tern. 1.34, 2.12, Vert. Man. A.1.53. There are also 3 probable cases in what may be a paratactic conditional clause, no man being expressed, BoSt. 3 2.1.30, 42, 43 (all these three instances may be taken instead as generalizing relatives, kwis being used like kwis kwis). Precisely the same problem is presented by the two instances of double kwis discussed in fn. 15, in each of which we may have an implied conditional clause with two indefinites, or a relative clause with one indefinite and one relative; jointly these two clauses present at least one more postpositive indefinite, either kwedani or kwis in Tern. 3.21, and possibly as many as three, kwit in BoSt. 3 5.4.10, and both kwedani and kwis in Tem. 3.21.

¹⁹ I am not offering statistics about the more specialized indefinites, kwissa and kwiski, more or less equivalent in use to Latin quisque and quisquam respectively. Kwissa seems about evenly divided between initial and postpositive place, but its occurrences are at best too few for statistical value, and it is hard to distinguish from the form that must have given rise to it, simple kwis followed by an emphatic or connective -a. Kwiski is extremely common and normally postpositive ; but as its use is mainly confined to combinations with man, which regularly stands first in its clause, or with natta, which like Latin non regularly precedes the word that it modifies, the usual postposition of kwiski results inevitably and may not involve any peculiarity of the word itself.

²⁰ Cf. fnn. 13 and 16.

²¹ Cf. fn. 11.

²² We have already noted in fn. 13 that nu kwis (B) may correspond with kwisma (A). In the strictly formulaic Soldateneid, A's and B's succeed one another apparently with no distinction, 4 A's (280, 282, 288, 290) and 7 B's (278, 281, 284, 285, 287, 289, 291); on this see further fn. 23.

²³ Whether there is normally any difference in meaning may well be doubted. In Sold. along with the A's and B's cited in fn. 22 there are 3 C's (279, 283, 286). I see no difference in meaning between CH 1.67 (22) pa-ap-ri-iz-zi ku-is (C) and 1.68 (23) ku-is pa-ap-ri-iz-zi (A); or between BoSt. 3 2.3.33-4 (162) ku-is IK-RI-BU sar-ni-in-ku-wa-as na-an sar-ni-in-kán-zi (A) and 4.7-8 (164) (C). On the other hand in MS 2.16-20 (where we find not only kwis but, parallel to it, kwissa, used not in its usual sense but rather in that of kivis kwis) there may be a real reason for the shift from a series of C's, IŠ-TU ^GIŠBANŠUR-ma-za-kán ku-e-iz-za, IŠ-TU GAL-ya-kán ku-e-iz-za, sa-as-ti-ya-az-za-kán ku-e-da-ni, IŠ-TU ^URUDDU₁₀×A-ya-za-kán ku-e-iz-za, to the still more indefinite, and therefore in this case perhaps more emphatic, A that sums them all up, ku-it-ta-ya im-ma Ú-NU-TU ₄. But in Pest. III.II.4.19-20 the shift from ku-e-da-as A-NA at the end of its clause to at the beginning of its clause is probably due merely to the desire not to begin the clause with the relative, in other words to provide a second C instead of an A.

²⁴ 26 and 30.

²⁵ Of these, 3 are hopeless: 97; 236, 245. Of the other 4, 249 is probably either A or B; 46 and 128 either B or C, more probably C; 142 probably C.

²⁶ 1, 2, 4; 12, 15, 17; 18, 19, 20, 23, 25, 27, 28, 35, 36; 59; 148, 149; 162; 192, 193, 207, 208; 210, 211, 217; 224; 241, 261, 262, 263, 264, 266; 270, 271, 274; 280, 282, 288, 290; 293, 296, 297, 298, 302, 304, 306, 311, 314; 325, 328; 384, 387.

²⁷ 9, 10, 16; 32, 33, 34; 58, 94, 105, 112, 119; 133; 144; 155; 202; 234, 257; 269; 278, 281, 284, 285, 287, 289, 291; 303, 310, 312, 316, 317, 323; 332; 343, 350, 378, 381, 388.

²⁸ The conjunction kwit ‘that, because’ (formed like Latin quid and used like Latin quod) tends even more than the relative pronoun to stand late : out of 75 examples that I have examined I have found only one where it stands first, and in one it is actually at the end of its clause.

²⁹ I say apparent since in some cases there are gaps where it may have occurred.

³⁰ I include here the ‘personal’ pronoun -as, which, as I have said, really is demonstrative; and also possessive adjectives which are equivalent to the genitive of the demonstrative or the personal pronoun.

³¹ In all my J's the relative clause follows the main clause instead of preceding it. Cf. fn. 43.

³² 6 examples: 12; 18; 59; 263, 264; 325.

³³ 4 examples: 58; 112; 316; 332.

³⁴ 58 examples: 21, 22, 24, 39, 40, 41, 42, 43; 61, 66, 71, 84, 85, 96, 102, 103, 104, 118; 131 ; 137, 153; 156, 161; 170, 186, 201; 214; 231; 247, 250, 251, 252, 253, 254, 260, 267, 268; 272, 273, 275, 276; 279, 283; 299, 301, 307, 309, 319, 320, 321, 322; 326; 346, 370, 372, 376, 379, 385.

³⁵ Only 1 example: 317, the one cited in the text.

³⁶ 10 examples: 49, 110; 154; 179, 195; 233, 246; 324; 341, 386.

³⁷ 18 examples: 50, 55, 60, 64, 65, 67, 70, 77, 111; 134; 140; 187; 209; 219, 226, 227; 334, 373.

³⁸ All the instances in this paper which are not otherwise classified fall into this category.

³⁹ 5 examples: 224; 241, 261, 262; 328.

⁴⁰ 2 examples: 310 and 323.

⁴¹ 17 examples: 13; 54; 160, 166; 172, 173, 175, 190, 199; 220, 221, 222, 223; 300, 352, 353, 369.

⁴² for A: 1; 207. 1 for B: 234. 31 for C: 29; 44, 48, 62, 76, 81, 83, 91, 106, 108, 113, 114, 115, 116, 127, 129; 135, 139, 143; 218; 337, 342, 345, 348, 349, 357, 359, 364, 366, 382, 390. 3 for examples not classified because of a preceding lacuna: 128; 236, 245. However, the other 4 passages with a preceding lacuna, namely, 46, 97, 142, 249, (cf. fn. 25) are complete after the relative, and are to be classified as G. The 2 unclassifiable, 26 and 30, (cf. fn. 24) are G and D respectively.

⁴³ 12 examples: 8; 95; 141 ; 157, 158, 159, 165; 240, 258, 259; 305; 330. On these see my sixth point below.

⁴⁴ At times this may involve scarcely any inconcinnity, as when an expression of place is followed by an adverb meaning ‘ there’, or an expression of time by an adverb meaning ‘then’ or ‘thereafter’: e.g. in Hatt. 3.6 É-ir kwit ‘the house which’ is balanced by anda ‘therein’, and in Tern. 2.30 and 4.68 kwedani mehuni ‘at whatever time’ is balanced by namma arha and nasta ‘after that’ respectively. Such examples might almost have been classed with those in which the second clause contains a demonstrative correlative with the relative (G). On the other hand, in Vert. Al. A.3.10, where the adverb apiya ‘there’ refers back not to an expression denoting certain lands but to an enumeration of the kings of certain lands, we have practically an anacoluthon.

⁴⁵ Cf. sup., fn. 9. In neuters there is particularly little distinction between singular and plural, and so I have listed my typical example, Pap. 1.8, with G rather than with H.

⁴⁶ We have a shift from plural to singular in AM KBo. 3.4.1.20, Hatt. 2.70, 4.71, KBo. 6.29.2.16, Tunn. 2.30 (cf. Tem. 3.52-3 with kwes kwes instead of simple kwes) ; and a shift from singular to plural in Tel. 2.56, Vert. Kup. 6.D.44, Pest. III.II.1.33 (the last-named involves a shift in person too).

⁴⁷ This particular example is rather complicated by the presence of a demonstrative beside the relative.

⁴⁸ 23 in all. They are enumerated in fnn. 50 and 54.

⁴⁹ Conjunctions—i.e. Hittite kwit and Latin quod—can have the same treatment. I believe, and hope to demonstrate elsewhere, that an ‘as to the fact that’ clause is simply a ‘that’ clause used with anacoluthon.

⁵⁰ 10 examples: 201 ; 247, 260, 263, 264, 267, 268; 299, 301, 322.

⁵¹ Sturtevant (Chrest. 153) combines the kwis clause with what precedes, but this rare order seems to me less probable than an anacoluthon. Sommer also apparently disagrees (apud Sturtevant, Chrest. 170).

⁵² I accept Sommer's analysis of the passage as a whole (Bil. 124 fn. 4) rather than Sturtevant's (Chrest. 153, 155) because the latter's interpretation of NU.GÁL (34 and 35) as ‘there is none (for you)', ‘there is nothing (which you possess)', does not seem to me warranted. But one can explain -za and -KUNU in 33 more simply than Sommer does: I think the notion is that the temple-officials must ‘watch their step’ carefully in the way they handle these precious articles whether in or out of the temple, the underlying idea being much the same as in the oft-repeated ‘fear the gods greatly’ (used almost directly thereafter, 2.36, and passim).

⁵³ With kwedani mehuni ‘at whatever time, whenever’ here and also in another example of D, Hatt. 6.29.2.31, cf. its use in two examples of H, Tern. 2.30 and 4.68, where it is followed by an adverb (see fn. 44). There is really not much difference between the D and the H type.

⁵⁴ 13 possible examples: 137, 153; 251-252-253-254; 316; 325, 326, 332; 346, 370, 372. 153 (AU 23.1.4.1-13) and 370 (Vert. Al. A.3.31-3) are particularly striking examples.

⁵⁵ Typical examples of these are three in Tunn., 2.9 (325), 2.15 (326), and 3.6 (332); but on the other hand in the parallel passage 2.19 (327, to be classed as G) it is natural (though not necessary) to view the pronoun as relative, because a personal pronoun (-at) follows. Midway between these two types is ib. 2.30 (328, already classed as H in fn. 39), where kwiski suggests the earlier pronoun without relating to it precisely.

⁵⁶ Bechtel's interpretation seems to me very unlikely, since it involves making two relative clauses (the kwis clause and the kwedani clause) parallel with each other (‘he who ... and to whom‘), which I think is contrary to Hittite idiom (as it also is to Greek, and to some extent to Latin).

⁵⁷ Some of the interpretations make the passage an example rather of H than of D.

⁵⁸ The man who finally does the extinguishing may or may not be the sinner who neglected this task originally.

⁵⁹ Presumably this refers, as many of the penalties in CH do, to what happened in one particular instance. That would be the reason for bringing in the seemingly irrelevant detail that the fire was confined to the temple. Or is there a real point in this? Is the wholesale destruction within the temple perhaps the result of the fire itself, not of a punishment meted out later; and in that case is it thought of as coming from the hands of the gods themselves? The guilty man died, and all those with him in the temple shared his doom (as almost happened to Jonah's fellow-sailors; the idea that a single sinner contaminates his shipmates is also common in classical antiquity). But the reference to the children, who presumably were not in the temple, may militate against this.

⁶⁰ 46 in all (enumerated in fnn. 63 and 64). These and the 23 mentioned in fn. 48 constitute the entire collection of D's.

⁶¹ Cf. sup., 71.

⁶² Cited above, 70.

⁶³ It would be nominative in 30 instances (12; 22, 30, 39, 40, 41, 42; 58, 59, 102, 103, 112, 118; 131; 156,161; 170,186; 272, 273, 275, 276; 279,283; 307, 309, 319, 320, 321; 376), accusative in 5 (24, 43; 214; 250; 385), dative with a preposition in 1 (379).

⁶⁴ A nominative twice after a genitive (18, 21) and three times after an accusative (84, 85, 96) ; and an accusative five times after a nominative (61, 66, 71, 104; 231).

⁶⁵ Except in cases where it is definitely replaced by some other equivalent element. Thus note that the -at of Hatt. 6.29.2.15 with reference to kwedas in the preceding line is replaced by human in 17 (a shift from G to H).

⁶⁶ Cf. once again Sturtevant, op.cit. (fn. 5) 146-7. I agree with him that Plautus, As. 621 patronus qui vobis fuit futurus perdidistis (which he cites on 145) is a parallel for Hittite (cf. sup., fn. 10); but such an expression as Plautus, Poen. 644-5 hunc chlamydatum quem vides ei Mars iratust (which he cites on 147) represents so far as Hittite goes a later and rarer manner of speaking.

⁶⁷ With kas kwis 12 (47; 135; 203, 206; 239, 240, 248, 258; 375, 376, 382, 389) ; asi kwis 2 (161; 244) ; uni kwis 2 (99, 113).

⁶⁸ With kas . . . kwis 2 (8; 138); apas . . . kwis 2 (141; 385).

⁶⁹ With kwis kas 5 (172, 173; 355, 368, 383).

⁷⁰ 161; 376, 385.

⁷¹ 47, 99; 138; 203, 206; 239, 244, 248; 355, 368, 375, 383, 389.

⁷² 172, 173.

⁷³ 113; 135; 382.

⁷⁴ 8; 141; 240, 258.

⁷⁵ Cf. sup., 76 and 80.

⁷⁶ The presence of a demonstrative in 4 examples of J is particularly striking inasmuch as this constitutes a third of the total number of J's (cf. fn. 43, where they are listed).

⁷⁷ 49, 62, 82, 85, 87, 88, 93, 95, 102, 107, 110, 111, 115, 117; 138, 141; 176; 355, 368, 375, 383, 390. The other 53 are 44, 47, 51, 52, 54, 55, 56, 67, 68, 69, 72, 73, 83, 91, 96, 98, 99, 103, 108, 113, 118, 119, 123, 125; 172, 173, 179, 181, 187; 213, 214, 216, 218; 229, 230, 231; 232, 234, 239, 242, 244, 246, 248, 259; 324; 357, 364, 365, 370, 376, 382, 385, 386.

⁷⁸ It might be urged that such a connective need not be treated as a definitely coordinating conjunction, but can be viewed simply as a rough connective merely leading the thought forward, like nun in German or then in English. But I think the fact that a word of this kind is regularly used to introduce every clause without regard to the relation of that clause to the one before it, would seem to indicate that the language at its earlier stages made no formal distinction between coordination and subordination.

⁷⁹ Cf. sup., 74.

⁸⁰ 2, 3; 6; 15; 18, 21, 22, 26, 27, 28, 29, 30, 39, 40, 41, 42, 43; 82, 95, 103; 153; 156, 161; 188, 189, 201 ; 211 ; 232, 251, 252, 253, 254, 259; 333, 336, 337, 341, 363, 368, 370, 379. In a few of these instances, the asyndeton is quite regular. I am excluding from consideration the 12 instances (listed in fn. 43) where it is practically inevitable because the main clause precedes. There are also 45 passages where a lacuna makes it impossible to judge whether or not a sentence-connective was present.

⁸¹ 81.

⁸² Cf. Vergil, Ecl. 9.6 quod nec vertat bene.

⁸³ Friedrich originally took Vert. Kup. 18.C.25 (355) as an example of this (so too Sturtevant, Lang. Mon. 7.143); but in his handling of the parallel passage, Al. A.2.77 (368), accepted Götze's correction (Madd. 141 fn. 2). Sturtevant takes Tern. 2.27 (299) as an example; I have already (fn. 51) expressed my disagreement with this.

⁸⁴ Listed above, fn. 43.

⁸⁵ The examples in BoSt. 3 (157, 158, 159, 165) are troublesome and doubtful (cf. fn. 88). Also difficult are 8 (Tel. 1.39), 240 (Pest. II.1.2), and 258 (ib. III.I.l.l), all of a special type starting ki kwit (we meet the same collocation in two other places also, 279 and 283, Sold. 1.30 and 2.44; but these seem different). 240 begins ki-i-wa ku-it i-ya-at-ten ‘this (is) what you have done’, with a nu clause following. This might be explained as meaning ‘this thing, which you have done, (is) (as follows)‘, with the clause serving as predicate nominative; we would then have instances of D not J. But a substantive clause of this sort would probably not have a sentence-connective.

⁸⁶ Contrast with this the normal order in the not dissimilar passage 1.31 above.

⁸⁷ I am following Goetze very closely in my translation.

⁸⁸ I have already discussed the difficulty of classifying passages in both of these : Tunn. 2.9, 15, 30, and 3.6 (see fnn. 17 and 55); BoSt. 3 2.1.30, 42, 43 (see fn. 18) and 2.51, 54, 3.2, 4.28 (see fn. 85). In the latter document, I have classed as a relative 1.5, which some might prefer to call an indefinite; and a number of instances which some might prefer to call relatives, I have decided to classify otherwise, as instances of the indefinite (1.30, 42, and 43, just referred to), the interrogative (2.21), or the conjunction kwit (1.21, 2.29, 3.19, 26, 4.2, 22, 23).

⁸⁹ In the two cases of double pronouns cited in fn. 15, I have been particularly troubled. I have decided to class both the pronouns in BoSt. 3 5.4.10-1 as indefinite, but only one of those in Tern. 3.21. See also my discussion of kwis in Tem. 3.49 (78-9).

⁹⁰ See Friedrich, Vert. 2.39, 109, 140. The interpretation of kwedani as a relative, suggested by Sommer, is perhaps militated against by the absence of a sentence-connective (cf. fn. 80).

⁹¹ Sup., 72.

⁹² The connection is clearer if, following English idiom, we reverse the order—‘ I destroyed them all, both (those) who died of wounds and (those) who died on the appointed day’.

⁹³ The Hittite interrogative, though it does not usually have initial position in its clause, does regularly precede the word which it refers to. In the documents that I have examined in connection with this paper, I have found 7 instances of interrogative kwis: as an adjective it is placed directly before its noun in BoSt. 3 5.3.10, Vert. Hukk. 3.53 (it follows its noun once, in BoSt. 3 2.2.21, where, however, I am not absolutely sure that it is an interrogative; cf. fn. 88) ; as a direct object, in Pap. 1.16, and as an ablative used adverbially, in Pest. III.II.4.5 and 16 (cf. BoSt. 3 5.3.10), it is placed directly before the verb. (The seventh example, Bil. 2.9, is unclassifiable, since nu kwit constitutes the entire clause.) Since examples of the interrogative adjective and pronoun kwis are so few, I have examined the use of the fairly common interrogative adverb kwat ‘why’. It stands first in its clause twice: Pest. III.I.2.11 and Vert. Hukk. 3.71. But usually (12 instances) it comes directly before the verb, the latter standing at the end of the clause: AU Tav. 1.65, 2.2 (apparently), 4.41 (apparently); BoSt. 3 5.3.4, 24; Hatt. 3.75, 76; Madd. 2.56, 72, 82, 90; Vert. Hukk. 3.56.

⁹⁴ 36 out of 46. See fnn. 60, 63, and 64.

⁹⁵ It is natural to say ‘some utensils are therein and are taken up’ or ‘some utensils he sees therein and takes up’; but after the interrogation, ‘what utensils are therein?’ or ‘what utensils does he see?', the pronoun in the second clause cannot be spared, regardless of whether it is or is not in the same case as the antecedent.