¹ 22.68-85. Note especially 70 fn. 6 and 81. A few emendations to the earlier paper that subsequent study has revealed as desirable are pointed out below; see fnn. 19, 41, 46, 67, 101.

² For the list of documents in which these examples occur, see 22.68; for the examples themselves, ib. 69 fn. 3. On abbreviations used see 68 fn. 1 (to be supplemented where necessary by Lg. 18.83 fn. 1); on manner of citing and of writing Hittite, 70 fn. 8. In addition it may be noted here that for convenience of cross-reference all examples of Hittite relative clauses cited in the text of the present paper have been given serial key-numbers, which are printed throughout in italics.

³ Cf. Sturtevant’s statement, Lang. Monogr. 7.148, that a large proportion of relative sentences in Hittite and early Italic are generalizing. This has been completely borne out by my investigations; cf. Lg. 22.81.

⁴ Thus the example taken as typical of the generalizing or restrictive relative sentences in my earlier paper (70), Pap. 1.8 Ú-NU-TE-MEŠ-ya-kán ku-e an-da-an na-at sa-ra-a da-a-i, is assumed to have shifted in meaning from ‘some utensils (are) there, and these he picks up’, to ‘the utensils which (are) there, these he picks up’. The corresponding example of the definite or non-restrictive type must have shifted from ‘the utensils—these are there—these he picks up’ to ‘the utensils, which are there, he picks up’. Another way of looking at it would be to say that in the first type the antecedent corresponds to a noun in the partitive sense—‘some (French des) utensils—(they) are there—he picks up’ (note some representing the original indefinite); and in the second type the antecedent corresponds to a noun in the definite sense, ‘the (French les) utensils—(they) are there—he picks up’ (note the definite article, both in English and in French, representing the original demonstrative). If we stress the difference between the two antecedents, as provided by the accompanying adjectival pronouns, the difference between the clauses disappears. There is, however, another distinction between the two types; in the first the antecedent belongs in the dependent clause, and we may expect the relative clause to precede the main clause; in the second it belongs in the main clause, and the relative clause, which is practically parenthetical, may indifferently precede or follow the main clause. This may have a bearing on the prevalent type of attraction; cf. fn. 17. In this connection we may note the difference pointed out by Goodwin, A Greek Grammar §1030, between ‘what he wanted, that he took’, and ‘he took these things, which he wanted’; the first is the restrictive type, where I think Indo-European had an indefinite relative, not preserved in Greek, and the second is the non-restrictive type, where I think Indo-European had a demonstrative relative, as Greek still has.

⁵ Thus in English, in such a sentence as ‘once he gets there all will be well’, once is a subordinating conjunction like Latin simul (its semantic parallel); and in such sentences as ‘everywhere I go I meet him’ and ‘any time you come you’ll be welcome’, everywhere is practically a synonym of wherever and any time of whenever (cf. Hahn, Lg. 18.113 and fn. 185).

⁶ The indefinites also generate interrogatives. Cf. Lg. 22.70.

⁷ Sturtevant has proved (Lang. Monogr. 7.141-9) that in Hittite and Latin the relative use was inherited, perhaps also in Greek (see especially 148-9); however, in Greek the use seems to have been lost and reintroduced later. In Germanic and Balto-Slavic the use was certainly a late development and therefore presumably an independent one. The cases of Armenian and Albanian are uncertain (on the latter cf. Pedersen, KZ 36.315). Of course the question whether in any given language group the relative use was inherited or independently developed is irrelevant; cf. fn. 16.

⁸ On the Hittite indefinite and relative, see Pedersen, Hitt. 67-72; Sommer, Bil. 164-6; Sturtevant, Lg. 14.241 fn. 11a. On the relation of the Hittite relative with the Indo-European indefinite, see Hahn, Lg. 18.83-116. Since the use of this stem as a relative (and also as an interrogative) is practically confined to Hittite, perhaps we may conclude that this development is purely Hittite or Anatolian; the few sporadic instances of relative use that may be cited for Indo-European languages, Greek μέχρι ‘until’, Latin simul ‘as soon as’, Irish má ‘if’, may well be independent and isolated developments (see Hahn, Lg. 18.112-3) like that cited for English in fn. 5.

⁹ Delbrück, who maintains (Gdr. 5.316) that the development of yos is from relative to demonstrative instead of vice versa, struggles in vain (as it seems to me) against the evidence: cf. inf., fnn. 10, 11, 13, and 15. He exercises an undue influence on Brugmann, who states (KVG §898) that all the Indo-European uses of the stem are either relative or developments of the relative, even though he, unlike Delbrück, admits that the original use of the stem was anaphoric. For the connection of yos with is, see e.g. Hirt, Idg. Gr. 6.162, 7.130; Walde-Pokorny 1.96-7. For the connection with the form of the stem seen in Hittite -, Sanskrit ayám, Latin eum, see Sapir, Lg. 14.272, and Sturtevant, The Indo-Hittite Laryn- geals 78. Presumably the Hittite conjunction -a -ya, parallel semantically to Greek - and Latin -gue from the kwo- stem, must somehow fit into the picture; cf. Sturtevant, Lang. Monogr. 7.148; otherwise Pedersen, AOr. 5.184.

¹⁰ The original demonstrative use still survives, it seems to me, in the Avestan article, which, despite Delbrück (Gdr. 5.303-4) and Brugmann (KVG §898), I believe is no more a development of the relative than the to- forms that we find in other languages. The Sanskrit usage described by Whitney (Skt. Gramm. §512B) as a ‘conversion of the subject or object of a verb by an added relative into a substantive clause’ seems just a step beyond this: where the noun is a nominative of any number or gender, a neuter of any number, a dual of any gender, or a feminine plural, the pronoun may be either a demonstrative or a relative; in the comparatively few instances where the noun is a masculine singular accusative, a feminine singular accusative, or a masculine plural accusative, the following nominative relative may be due to a misunderstanding of the construction that arose after the stem ceased to function as a demonstrative.

¹¹ In Homeric Greek, still clearly retains its demonstrative use, as in II. 1.405, Od. 1.286; see Monro, A Grammar of the Homeric Dialect 234-5, and Chantraine, Grammaire homérique 276. Even in 5th- and 4th-century Greek we have a few traces of the usage, as in Herodotus’ ‘this man and that’, Plato’s ‘said he’, and the occasional use of as an equivalent of , and of as an equivalent of rote . To explain this demonstrative as a by-form or variation of the relative as do many editors of Homer and the less scientific grammars (e.g. Seymour, II. Introduction §42p and Od. Introduction §45p; Goodwin, Greek Grammar §1023) may be acceptable descriptively but not historically. It is customary to ascribe demonstrative toso- instead of to yo- (see Brugmann-Thumb, Griechische Grammatik⁴ 282; Schwyzer, Griechische Grammatik 1.611, and the references there given), but this seems to me to be ruled out by the corresponding use of neuter (on which see Hirt, Idg. Gr. 7.131) and of . Windisch’s similar objections based on were brushed aside by Delbrück (Gdr. 5.311) on the ground that scholars of his day had more faith in analogy than those of Windisch’s, but perhaps we of a still later day are more guarded or at least more discerning; analogy seems nowhere to have wiped out the difference between the two forms in s- and the numerous ones in t-, and if it ever did work we might expect the t- group to prevail over the s- group.

¹² So Pedersen, KG 2.235-6.

¹³ Balto-Slavic presents a difficult problem. Delbrück admits (Gdr. 5.392 fn. 1) that in this field the stem yo- is anaphoric, but he has not thought through the problem of the demonstrative adjective.

¹⁴ The demonstrative pronouns could not have been earlier than Indo-European (cf. Sturtevant, HG 199-200 and Lg. 15.11-9), and the relative use (despite Hirt, Idg. Gr. 7.133) may not have been so early; certainty it seems later than the relative use of the other demonstrative yo-. That the Greek relative from so- to- retained its original demonstrative nature more clearly than is suggested by two facts (on which see Monro 231, Chantraine 278) : (1) Whereas os is often indistinguishable in use from the indefinite (Brugmann- Thumb 644), is used only with a definite antecedent, and according to Chantraine 278, it is not readily used epexegetically, by which he clearly means restrictively. A good passage to bring out the difference in meaning between them is II. 16.156-9

where the τ(ε) clause is restrictive and the clause is non-restrictive. (2) Whereas the clause, like the kwis clause in Hittite and the qui clause in Early Latin, usually precedes the main clause (cf. Meillet-Vendryes, Traité de grammaire comparée des langues classiques (574) the clause normally follows, suggesting a parenthetic paratactic origin (cf. fn. 4), as in II. 22.340-2

where we can translate ‘but take plenty of bronze and gold—these gifts my father and my lady mother will give you—and give my body back home’; cf. the very similar passages Od. 8.428, 21.13, 24.335. With these examples of the normal order we may contrast II. 1.125 and Od. 4.349 = 17.140 (said by Brugmann-Thumb 645 to be the only Homeric examples of a relative clause preceding the main clause).

¹⁵ Cf. Brugmann, KVG §495.2. Delbrück (cf. sup., fn. 9) has particular difficulty with this form. Though he is willing to admit that in Greek the demonstrative stem so- to- generated a relative—indeed, he says of the relative in Homer that it ‘entsteht so zu sagen vor unseren Augen aus dem anaphorischen^, (Gdr. 5.315), and in many cases it is hard to distinguish it from the demonstrative—he will not recognize the same possibility for Old Persian. He insists that the stem yo- still exists in Old Persian, which seems indubitable but not an argument against the development side-by-side with it of another relative from so- to- precisely as happened in Greek; and he finally decides that Old Persian probably shared with A vestan the use of ya-as an article (cf. fn. 10), that there was a conflict between the relative and the demonstrative stems in this particular use, and that when the demonstrative stem won out here it extended its victory into the field of the relative (5.314). This seems to me extremely circular reasoning based on a dubious premise.

¹⁶ The lost category might eventually be restored, as happened in Greek and Germanic (cf. fn. 7), so that ultimately the conditions of the Ursprache might be reproduced.

¹⁷ The Greek and Latin relatives show their difference in origin in certain marked differences in usage. (1) A demonstrative origin is perhaps suggested by the use after a relative of a demonstrative instead of a second relative coordinate with the first; this is regular in Greek, but comparatively rare in Latin. (2) The fact already pointed out (fn. 4), that with a relative stemming from an indefinite the antecedent belongs in the relative clause, and with a relative stemming from a demonstrative the antecedent belongs in the main clause, may well have a bearing on the prevalent type of attraction: in Latin the antecedent is more often attracted into the case of the relative, in Greek the relative is more often— indeed under some conditions regularly—attracted into the case of the antecedent.

¹⁸ So Sturtevant, Lang. Monogr. 7.149: ‘Hittite seems to show an early stage in the development of a relative referring to a definite antecedent out of a general relative.’

¹⁹ In the list of 75 (in which I classed 22 as certain) which I enumerated in Lg. 18.81 fn. 77, I would now make several changes, bringing up the total to 79 (cf. inf., fn. 123). I would remove one, 385 (Vert. Hukk. 3.64); see below, fn. 67. And I would add five, as follows: one, 75 (AM KBo 3.4.3.90), which is uncertain because of a difficulty in interpretation (kwit here may be the causal conjunction, but if it is a relative pronoun the clause is clearly non-restrictive), and four others all extremely dubious, but perhaps no more so than some others that I included in the original list, namely 135 (AU Tav. 1.32), 161 (BoSt 3.2.3.30), 369 (Vert. Al. A.3.10), 389 (Vert. Hukk. 4.50). (The key numbers for these five passages in the present paper are respectively 64, 33, 30, 12, 34.) Also, I would transfer from the certain to the doubtful category 93, 107, 176, 375, 383, 390 (key-numbers here 68, 65, 67, 35, 20, 66); and from the doubtful to the certain category 51, 229, 230, 231 (key-numbers here 84, 41, 43, 40)— thus reducing the total of passages deemed certain by two. It should likewise be noted that one highly doubtful example, 382 (Vert. Hukk. 2.56), which I included because of the presence of a demonstrative with the relative (ki-e ku-e), is too mutilated to be dealt with here at all.

²⁰ Anittas, Bil., Tel., and CH.

²¹ Some languages even in certain cases avoid one or the other type, as we shall see later.

²² As already pointed out, fn. 14. But, true to its demonstrative origin, even is not so completely generalizing as the Latin qui, as is proved by the fact that in general expressions the indefinite (Homeric) or is added to it.

²³ This seems somewhat surprising, since one would expect that to behave like der rather than like wer; but the explanation lies in the rather complicated history of the English relative. Since Germanic did not inherit the relative use of the kwi- kwo- stem (cf. fn. 7), the only English relative originally in use was the one from the to- stem, which thus served for the restrictive as well as the non-restrictive use of the relative (cf. 349); but later the kwi- kwo- stem reacquired its lost relative use (cf. fn. 16), and the usage spread to such a degree that it ousted the to- stem not only to a great extent from its acquired restrictive use but altogether from its natural non-restrictive use.

²⁴ As I pointed out years ago in an article on the ab urbe condita type of expression, Cl. Journ. 23.266-74 (1928); see for this example 272 and fn. 14.

²⁵ On this cf. Lane, A Lat. Gram §1044; on the avoidance of a genitive or an ablative of quality with a proper noun, Hofmann, Stolz-Schmalz Lateinische Grammatik⁵ 398.

²⁶ Horace, Serm. 1.6.54-5.

²⁷ Those who contributed to the Cl. Weekly under the editorship of the late Charles Knapp will doubtless remember that the style-sheet of the paper, with what may have seemed an excessive insistence upon logic, rigidly demanded the use of commas with the non-restrictive adverb, as well as with the non-restrictive adverbial phrase. A few excerpts from Knapp’s own writings are here quoted, not as models of punctuation, but as clear examples of the recognition of the non-restrictive use of words or phrases : ‘To such teachers of Latin proposals to eliminate the writing of Latin would, naturally, be welcome’ (CW 28.139); ‘No Classical Association is as well off, to-day, in members or in finances, as it was five years ago’ (ib. 185); ‘Some time ago I was asked, by a correspondent, to state’ (ib. 29.1).

²⁸ Thus there is no need to assume two different origins for the definite and indefinite relatives as is done for Latin by Leumann (Stolz-Schmalz 288), who traces the definite relative qui to the indefinite and the indefinite relative quis to the interrogative. Hofmann (ib. 706) seems to agree with him at least to some extent, but is less ready to commit himself in regard to either the generalizing or the definite relative; and as a matter of fact the examples that he himself adduces (ib.) of the interchange of quis and qui as a relative surely militate against any such hard and fast distinction as that made by his colleague.

²⁹ In Latin the difference between the purely factual and the circumstantial is frequently marked in relative clauses (both with qui and with cum) by a difference in mood. Cf. fn. 39.

³⁰ Parallels in the same document are 2.59, 3.55, and 4.34. Cf. too the similar passages 1.48 nu-us-ma-as su-me-es LÚ.MEŠ É DINGIR-LIM ‘you temple-officials^, , which lacks the relative kwis, and 4.25 an-da-ma ŠA KISLAḪ GUD.APIN.LAL.ḪI.A ku-i-e-es KIN-te-ni ‘(you) who care for the granary plow-oxen’, which lacks the personal pronoun sumes (or sumas).

³¹ Parallels in the same document are 3.3, 3.9-11.

³² An anacoluthon follows (on which see Friedrich, Vert. 2.71 fn. 3 and 97; and Hahn, Lg. 22.78 fn. 54), but later comes the main thought, 43-4 nu 1-as 1 -e-da-ni wa-ar-ri sar-di- ya-as ŠU.BULUG-as-sa e-es-du nu 1-as 1-an pa-ah-sa-ru ‘let one be a help and a rescuer against force to another, and let one protect another’.

³³ The corresponding numeral would in English probably appear in the main clause, ‘you 3 who are noble men’, ‘you 4 who are the kings in Arzawa’. On this cf. inf., fn. 43 on 14. Or is the meaning rather ‘you who are the 3 noble men’, ‘you who are the 4 kings of Arzawa’?

³⁴ A causal force might also be felt in the first group cited in illustration of the type ‘you who are priests’—i.e. ‘you who are temple-officials’, ‘you who are water-carriers’. The idea may be ‘you who are such-and-such persons’ are to do such-and-such a thing precisely because you are so-and-so. Cf. further below, fn. 86 (on 57), and fn. 103 (on 78).

³⁵ In the parallel passage Vert. Targ. 2.2-3 ka-a-as-ma ŠA(G)-BI KUR-77-FA 3 LÚ.- MEŠ EL-LU-TIM zi-ik ^ITar-ga-as-na-al-li-is ^IMas-hu-i-lu-wa-as^IMa-na-pa-^DU-sa nu 1-as 1 -e-da-ni li-e i-da-la-a-u-e-es-zi ‘behold in my country there (are) 3 noble men, you T., M., and M.; let not one do harm to another’, we find a main clause conveying information which corresponds to that in a subordinate clause in Kup. and Al. Cf. fnn. 30 and 36.

³⁶ The same idea is conveyed in a main clause just above, 16 zi-ik-ma-wa-za DUMU-as nu-wa Ú-UL ku-it-ki sa-ak-li ‘ “you (are) a child and you don’t know anything” ’, also in the parallel text KBo 3.4.1.14-5 ki-nu-un-nia-wa-za-kán ku-is A-NA ^GIŠGU-ZA A-BI-ŠU e-sa-at nu-wa-ra-as DUMU-Zo-as nu-wa KUR Hat-ti ZAG.HI.A KUR Ha-at-ii-ya-wa Ú-UL TI-nu-zi ‘ “he who has now taken his place on his father’s throne, he (is) a child and will not cause Hatti and the bounds of Hatti to live” ’. Cf. sup. fn. 35 and inf. fn. 53.

³⁷ We shall find Mursilis manifesting the same shrewdness in his similarly emotional appeal for relief from pestilence (examples 28, 24, 25). Hattusilis also has a passage of the sort, 3.70-1 (38). For Biblical parallels, see fnn. 39 and 75.

³⁸ For the emotional flavor, cf. fn. 37.

³⁹ The Vulgate in the corresponding passage, 15.30, has the indicative, qui devoravit; but classical Latin would almost assuredly use the subjunctive (cf. fn. 29).

⁴⁰ It must be noted that the form ‘he who did this, namely John’, or its variant (to be treated below) ‘he who, namely John, did this’ is from the nature of the case bound to be the Hittite approach to the development of the completely non-restrictive clause. This is due to the fact that in Hittite the antecedent, just as much when it precedes the relative as when it follows it, is a part of the relative clause (cf. Hahn, Lg. 22.71 and fn. 10). This peculiarity is probably both cause and effect in reference to the circumstance that early Hittite relative clauses are all restrictive. Hittite says not ‘the man who did this is my frined’ but ‘which man did this, he is my friend’. For ‘John, who did this, is my friend’ there is no correspondent that accords with Hittite idiom; ‘which John did this, he is my friend’ is restrictive, implying that of a number of Johns one particular one is involved. In other words it is a basic characteristic of the fully-developed non-restrictive relative clause that the antecedent should stand not only before, but definitely outside, the clause; cf. examples 40-42, 73-82, 84. Thus only in cases where the antecedent and all its modifiers precede the relative pronoun can we say that we have a full-fledged non-restrictive clause; on the other hand the mere fact that antecedent, or modifier, or both, precede the relative pronoun does not warrant us in assuming that they are outside the relative clause. Actually, either the antecedent or some modifier of it, if not both, comes before the relative oftener than after, because of the tendency of the latter to be postpositive (cf. Hahn, Lg. 22.73-4). This tendency probably accounts for variation in the position of a numeral modifying a relative; the numeral precedes the relative in Hatt. 3.70 (38), Ann. 2.21 (39), Pap. 2.14 (40), and AM KBo 4.4.4.26 (77), and follows it in Vert. Kup §27.A.20 (3), Al. A.3.31 (4), Al. A.4.38 (14), and AM KBo 5.8.3.24 (29), where there is a personal pronoun or a demonstrative preceding the relative, but it is equally a part of the relative clause in both cases. (As a matter of fact, the clearest example of a restrictive clause among these eight passages occurs in one in which the numeral precedes the relative, 39.)

⁴¹ In my earlier paper, I did not recognize this (cf. sup., fn. 19). In other words, I took it as an example of the type ‘the men who do this’, not realizing it was rather ‘the men who do this, namely Tom, Dick, and Harry’.

⁴² I assume that these entities really are the gods in question; that the ‘posts’ at least are treated as divinities is indicated by the prefixing to ILLAT.MEŠ of the determinative D.

⁴³ This is a mixed type. DINGIR.MEŠ ‘gods’, together with its modifying numeral LI-IM ‘a thousand’ (on which see below, fn. 61), is interpolated within the clause after the manner of the antecedents with their modifiers in the next group of examples (16-29); but the true correspondent with John in our typical example, a group of specific names in apposition with the general term ‘gods’, has the position of John.

⁴⁴ Cf. fn. 40.

⁴⁵ Cf. CH 4.53 (11) and Tel. 2.61-3 (10) cited above.

⁴⁶ Somewhat like this passage, but peculiar and difficult, is AM KUB 14.15.3.28-9 NAM. RA.MEŠ-wa-mu-kán ku-i-e-es pí-ra-an ar-ha pár-se-ir NAM.RA ^URUHur[-sa-na-as-sa-wa- kân NAM.RA ^URUSu-] ru-ta Ú NAM.RA.MES ^URUAt-ta-ri-?na an-da ú-e-ir. I am following Götze’s suppletion, but am inserting -wa before kán, following 33, where it is present, rather than 31, where—strangely—it is absent. Despite his suppletion kán with NAM.RA ^URUHur- sa-na-as-sa, Götze seems to make a new clause begin not there but with an-da ú-e-ir; at least so it would appear from his translation with a new subject, ‘die kamen (hier) herein’. Accepting his interpretation, I classed this passage (Lg. 22.S2 fn. 80) as an example of the omission of a sentence connective, but such omission in a passage like the present one is so strange that I have come to believe that NAM.RA ^VRVHur[-sa-na-as-sa NAM.RA ^URUSu-] ru-ta Ú NAM.RA .MEŠ ^URU At-ta-ri-ma must be the subject of an-da ú-e-ir. In that case the passage appears to mean ‘ “the captives who had fled from me (literally, which captives had fled from me), the captives from H., the captives from S., and the captives from A. came here’ We then have another variant in order; the phrase corresponding to my John is no longer a part of the relative clause but the subject of the main clause, as if we were to say ‘he who did this, John came’ for ‘John, who did this, came’ (cf. Sturtevant, Chr. 96 on Hatt. 3.78). Or else we have an example of partitive apposition, like Tem. 3.13-4 (on which see Hahn, Lg. 22.77; and also cf. inf. fn. 51). Or are we perhaps to supply kwes somewhere in the an-da ú-e-ir clause, and assume that the main clause starts with nu-wa-ra-at?

⁴⁷ We may perhaps compare the corresponding, supposedly pleonastic, use of a demonstrative following a relative in Latin. This is pronounced late by Hofmann, Stolz-Schmalz 707, but we find some traces of it early, as Plautus, Cist. 691 quae ilia and Trin. 1023 quorum eorum.

⁴⁸ This is true whether the pronoun is anaphoric or deictic—a distinction which for purposes of the present paper seems to have no significance.

⁴⁹ Cf. once more fn. 40.

⁵⁰ In this particular instance the clause is perhaps restrictive, since the speaker appears to have other sisters if we may judge by 4.2 SAL+KU.MEŠ-YA ‘my sisters’; unfortunately the passage is so badly mutilated that we can form no idea of the context. The king’s sister herself possesses two varieties of sisters, sisters ŠA MÁŠ-ŠU and sisters ŠA NUMUN-ŠU (3.26); and if for those of the former category we accept Friedrich’s tentative translation (based on a suggestion of Sommer’s) ‘Vollschwestern (?)’ (Vert. 2.125; see also ib. 148 and fn. 1), these must of course be the king’s sisters also.

⁵¹ On the loose agreement, involving a shift from plural to singular, cf. sup. fn. 46 (close), and inf. fn. 88, also Lg. 22.77 fn. 46; on the reverse type, inf. fn. 109.

⁵² Cf. 1.2.35-40, III.1.2.7-11, III.II. 1.25-6.

⁵³ The relative clause is important enough to be a main clause: ‘only a few still survive; let them (at least) not die’. Cf. fn. 36.

⁵⁴ Cf. fn. 37.

⁵⁵ The Latin idiom is similar: Latin would certainly say qui pauci supersunl. Cf. e.g. Cicero, De Sen. 14.46 tempestivis quoque convivile delector, nec cum aequalibus solum, qui pauci admodum restant, sed cum vestra etiam aetate. Reid’s note ad loc., in which he calls this ‘the substitution of the nominative of the relative for the dependent genitive (quorum)’, seems quite wrong; the partitive genitive quorum would be in order if Cato were referring to all the contemporaries whom he once had, but only qui is suitable in a reference to the (few) contemporaries who now survive. If Reid were right, then Vergil’s pauci laeta arva tenemus (Aen. 6.744) would have to be viewed as equivalent to pauci nostrum laeta arva tenent, which of course it is not.

⁵⁶ The adverbially used ablative tepawaz ‘in small numbers’ (concerning which see Götze, Hatt. 81 on 2.21) is here equivalent to the adjective tepawes in the first example.

⁵⁷ We may compare ib. 1.1-4 nu-mu ERÍN.MEŠ UKUŠ ku-is te-e-pa-wa-az kat-ta-an e-es-ta ERÍN.MEŠ NA-RA-A-RI-ya-za ni-ni-in-ku-un nu IŠ-TU ERÍN.MEŠ UKUŠ Ù IŠ-TU ERÍN.MEŠ N A-RA-A-RI I-NA KUR ^URUTág-ga-as-ta pa-a-un; but there, I think, in view of the non-subordinate clause that follows it (which would have fitted in much more neatly had a kwin been inserted after ERÍN.MEŠ NA-RA-A-RI-ya-za to balance the kwis after ERÍN.MEŠ UKUŠ), we had better view the kwis clause too as nonsubordinate and render ‘I had some heavy-armed troops—a few of them—down there, and I gathered auxiliary forces, and with the heavy-armed troops and with the auxiliary forces I went to T.’ In other words the passage probably belongs with those where we can avoid an anacoluthon if we handle kwis as an indefinite rather than as a relative (see Lg. 22.77-9).

⁵⁸ We may contrast with this AM KBo 4.4.4.25-6 (77), discussed below. At first sight 3 LI-IM NAM.RA I-NA É LUGAL ku-in ú-wa-te-nu-un may look not unlike ku-in 9 LI-IM ERÍN.MEŠ ú-wa-te-it, and it might be that the passage meant ‘the captives from D.— (that is) the 3 thousand captives that I took to the palace’. But we have not heard before of these particular 3000 captives. Cf. further on this below, fn. 102.

⁵⁹ In 3.3 we hear of [L]I-IM ERÍN.MEŠ ‘-thousand troops’, for which Götze on the strength of our present passage proposes [9L]I-IM ‘[9] thousand troops’; but unfortunately there is a lacuna both before and after it, and we do not know their number or their leader. But when we first meet P. in 3.15 he is spoken of in a manner suggesting that he had been introduced earlier.

⁶⁰ Cf. fn. 43.

⁶¹ Cf. the reference to them in Targ. 2.55 and Hukk. 1.39, in each instance followed by an enumeration which, though mutilated, certainly in the former instance may come nearer, and in the latter does come nearer, to the specified figure than the passage here under consideration. Furthermore, note the preceding lengthy list in Al. 4.1-30.

⁶² Cf. fn. 33.

⁶³ See fn. 40.

⁶⁴ There is a similar variability of the relations when a noun has two adjectives modifying it; a feeble old woman is an old woman who is feeble, but an old, feeble woman is a woman who is old and feeble. Of course both in these two cases and in those cited above for relatives, the phrasing, pitch, and stress vary to suit the different cases; but so long as the words may remain the same, an accidental departure from the normal manner of speaking might well lead to a new type. That such departures occur can be easily verified by any one who listens to radio announcers and commentators with their odd tricks of emphasis, e.g. stressing an auxiliary such as is or will where the sense does not demand it.

⁶⁵ Like the ‘thousand gods’ in 14 and perhaps the 3 noble men and the 4 kings in S and 4. But numerals do not usually restrict; cf. the discussion of 29.

⁶⁶ Cf. what was said of demonstratives earlier, in connection with 20, 21, 22.

⁶⁷ These demonstratives must be sharply differentiated from the quasi-personals -as and its more emphatic synonym apas, which appear in the main clause after the relative clause, corresponding to the English demonstrative in such a sentence as T went against those states which had fought against me’ (in Hittite idiom ‘which states had fought against me, against those I went’). Friedrich treats apas like kas in Vert. Hukk. 3.62-5 SA KUR ^URUAz-zi-ya SAL nam-ma A-NA DAM-UT-TIM li-e da-at-ti a-pu-u-un-na-za a[r-X-]X-X ka-ru-ú-za ku-in har-si ^SALNAP-TER-TA-ma-at-ta a-a-ra e-es-kán-zi DAM-an-ma-an-za li-e i-ya-si, which he translates (Vert. 2.129) ‘Auch sollst du keine Frau vom Lande Azzimehr zur Gattin nehmen. Und jene……, die du schon besitzest, die mag man mit Fug und Recht als deine Nebenfrau behandeln, zur Gattin aber mache sie nicht.’ But that he is wrong is proved by the fact that apun and kuin are in different clauses, as is shown by the two za’s (the absence of a sentence connective for the clause beginning karuza must be due to the parallelism between the two orders: do not henceforth—namma ‘again’—take a woman from Azzi as a wife, do not make even the one you have now and have had for a long time—karu—a wife, though it’s all right to have her as a concubine). I think a[r]- must begin a verb embodying an order parallel with le datti: ‘do not again take a woman from Azzi in wedlock, and treat (or ‘do not treat’) her (in some specified way); the one that you have had for a long time it is all right that they (should) treat her as your concubine, however, but do not make her your wife’. The only example that I have found of apas kwis, AU Tav. 4.47-8 (83), does not bear Friedrich out, for here if I am right kwis is almost a connecting relative and apas is thus in a clause which, while used parenthetically, is not truly subordinate; cf. fn. 111. Therefore I now believe I was wrong in classing apas … kwis with kas … kwis, Lg. 22.80 fn. 68.

⁶⁸ This passage was classed above with passages in which a numeral is incorporated within the relative clause, but might equally well have been placed here with the passages in which a demonstrative precedes the relative. There is a demonstrative present also in Vert. Al. A.4.38 04), and in Pest. II.9.8 (23) and one of the similar passages, 1.2.43 (24), all passages of the intermediate type between restrictive and non-restrictive; but it is not the demonstrative that puts them there.

⁶⁹ There is a comparable use of the demonstrative in the mutilated but clearly similar passage ib. 34 nu-us-si ki-i ku-it É-zr ka-ru-X-da-an. That in both cases the demonstratives make very little difference is indicated by the fact that the reference to ‘this son’ in 9 is followed by a parallel one to ‘the house’ in 10, and the reference to ‘this house’ in 34 is preceded by a parallel one to ‘the son’ in 32, both without demonstratives. For ‘the house’ in 10, cf. also 19; and on ‘the son’ in 32 (46), see further below.

⁷⁰ It might be suggested that Pest. II.1.2 ki-i-wa ku-it i-ya-at-ten, and the almost identical Pest. III.I.1.1 ki-i ku-it D[U-at-ten], are of the same type, meaning ‘this thing(,) which you have done(,) (is) (as follows)’; but the use of nu in each case in the following clause seems to me to rule out this possibility, and force us rather to the interpretation ‘this (is) what you have done; and’ etc. So Götze (K1F 1.207 and 237). See Lg. 22.82 fn. 85 on these and some other difficult examples of ki kwit: the similar Tel. 1.39, and two of a different type, Sold. 1.30 and 2.44, where indeed kwit may be interrogative not relative.

⁷¹ There may have been other Kaskan states not involved, since there were a number of them; cf. examples 57 and 58, and what is said with reference to these passages concerning the Kaskan states. Hostilities with the Kaskans are referred to passim (e.g. AM 19.39.2.1).

⁷² The ambiguity is precisely like that noted earlier in reference to Vert. Hukk. 3.25 (20).

⁷³ Hostilities with the Sunupassans are referred to in KBo 2.5.1.1-3. Cf. fn. 71.

⁷⁴ Cf. fn. 65.

⁷⁵ See also fn. 37. The tone is the same as that of Sam. 2.12.3 ‘the poor man had nothing, save one little ewe lamb, which he had bought and nourished up’; I am not sure whether or not in the punctuation of today the comma should stand after lamb.

⁷⁶ Sturtevant punctuated this clause as non-restrictive in TAPA 58.11, but changed in Chr. 111.

⁷⁷ As in KBo 5.8.3.24 (29) the numeral ‘9 thousand’, as well as the demonstrative uni, helps to identify the troops under discussion, but so too in all probability does the relative clause, which must therefore almost certainly be regarded as restrictive.

⁷⁸ At least according to Sturtevant’s very plausible suppletion, 1.44 [2 NINDA KUR₄.RAm]a kur-kán-zi na-as har-zi ‘[bu]t they keep [2 loaves], and she has them’; at all events, we know that the sacrificers started off with 16 loaves (1.5) and disposed of 14 (1.42-3). Incidentally, the use of two coordinate clauses here gives a rather choppy and clumsy effect, as is suggested by Sturtevant’s translation (Chr. 109), ‘(They keep two ordinary loaves—she has these)’. It seems as if a non-restrictive clause, like the one in the Pap. passage discussed just below (40), ‘they keep two loaves, which she has’, or possibly even ‘she has two loaves, which they keep’, would have produced a smoother result; but there are no non-restrictive relative clauses in Ann. Does this suggest an earlier date for Ann. than for Pap. (cf. fn. 79)? See also fn. 92 for another example of awkward coordination instead of subordination by a non-restrictive relative.

⁷⁹ It is amazing that out of the four relative clauses in Pap., three are of the rare nonrestrictive type (cf. the figures below, in the Appendix); this may suggest a late date for the document. The fourth example, 1.8, already treated as a typical example of the relative clause (fn. 4), provides an excellent contrast for 41 and 42.

⁸⁰ My translation is an attempt to render the force of the -sk- form of the verb, which is rather difficult; see Sommer, Pap. 21-5, and Bechtel, Hittite Verbs in -sk- 51. Since everything used and everything done in this part of the ritual seems to involve pairs or fours— note especially 2 DA.A.AN (on which see Sommer ad loc., Pap. 21) or 2 TA.A.AN in 30, 34, 37, and 46—I think the force of the verb here is ‘distributive’ rather than durative as Bechtel believes (his refusal to admit the possibility of a merely iterative use seems borne out by Pap. 3.42). At all events, the question of the force of the verb does not affect that of the type of the relative clause.

⁸¹ Probably the same problem is presented by the mutilated passage Pest 1.1.19 nam-ma- as-si ku-e-es ŠEŠ.MEŠ-SU … ku-en-nir ‘and furthermore his brothers(,) who …(,) they killed’.

⁸² Already discussed above (32). See especially fn. 69.

⁸³ On this see further below, the concluding paragraph of the paper.

⁸⁴ This reminds me of 35 Al. A.3.73, with the phrase ‘to the Mala River’ corresponding to the demonstrative ki.

⁸⁵ Cf. Friedrich ad loc., Vert. 2.161.

⁸⁶ My only reason for not saying ‘certain’ rather than ‘highly probable’ is my fear of being misled by the English translation. The writer may feel that he has sufficiently identified the community in question by saying simply URU Kà-as-kà-as, which to us seems inadequate since we must call it ‘a Kaskan state’ not ‘the Kaskan state’ (Hittite, like Latin, does not mark this particular distinction); and the statement about occupying Mt. Ashar- paya and cutting off the Pala roads may be added to provide additional information rather than identification. In that case, however, the clause acquires a strong causal flavor, since if it does not serve to identify the state, it does serve to give a reason for the king’s attack upon it; and thus it still, like 3, 4, 5, 6, and perhaps 78 (on which see fn. 103), approaches the restrictive rather than the non-restrictive. Exactly the same problem is presented by three closely parallel passages just below, ib. 43, 58, and 61.

⁸⁷ We may contrast this with such a passage as Hatt. KBo 6.29 B.1.17-8 (NBr 46) ERUM- an-ni-ya-mu ku-e KER.KUR.MEŠ pí-ya-an e-sir nu-mu-kán a-pi-e-ya ar-ha da-at-ta-at ‘and the countries which had been given to me in servitude these too were now taken away from me’, where, since the names of the countries involved are not specified, the relative clause is clearly restrictive.

⁸⁸ For the shift in number from plural to singular, see fn. 51 on 22.

⁸⁹ Note 3-4 just above, ku-it-ma-an-za-kán A-NA ^GIŠGU-ZA A-BI-YA na-wi e-es-ha-at nu-mu a-ra-ah-zé-na-as KUR.KUR.MEŠ ^LÚKÚR hu-u-ma-an-te-es ku-u-ru-ri-ya-ah-hi-ir ‘before I took my seat on my father’s throne, all the surrounding enemy countries had made war on me’. But the epithet is probably thrown in casually and is not meant to make any appreciable difference in the picture: there is no reason to assume that the ‘surrounding enemy countries’ of 3-4, 9-10 nu a-ra-ah-zé-na-as KUR.KUR ^LÚKÚR ki-is-sa-an me-mi-ir ‘the surrounding enemy countries spoke thus’, 19 (60), and 23-6 (61), are a group more restricted than ‘the enemy countries’ of which we hear in 6-7 ma-ah-ha-an-ma KUR.KUR.MEŠ ^LÚKÚR ^IAr -nu-an-da-an ŠEŠ-YA ir-ma-an is-ta-ma-as-sir nu KUR.KUR.MES ^LÚKÚR ku-u-ru-ri-ya-ah-hi-is-ki-u-an da-a-ir ‘when the enemy countries heard (that) my brother A. (was) ill, the enemy countries began to make war’ and 8-9 (59). That this is the case seems conclusively proved by the final lines of this introductory portion of the Ten- Year Annals: 28-9 nu-za ki-e a-ra-ah-zé-na-as KUR.KUR.MEŠ ^LÚKÚR I-NA MU 10.KAM tar-ah-hu-un na-at-kán ku-e-nu-un ‘and these surrounding enemy countries I conquered and defeated in the course of 10 years’.

⁹⁰ However, the causal force and emotional flavor may render it restrictive. See the earlier discussion of the entire passage, and cf. fn. 37.

⁹¹ Note that the similar term in KBo 5.8.2.17 and KUB 14.16.2.12 ^LÚKÚR ^URUKà-as-kà-as ‘the Kaskan enemy’ certainly does not exhaust the entire body of Kaskans. Cf. sup. on 86.

⁹² That the non-restrictive clause is not yet a really natural form of expression is suggested by the fact that just above, where we might have expected one, we find a coordinate clause instead: 3.53-4 pa-ra-a-ma I-NA ^URU La-ak-ku pa-a-un nu ^URULa-ak-ku-us URU-as BÀD-an-za e-es-ta ‘but I went to Lakku, and Lakku was a fortified town’. Cf. fn. 78, on the similar clause in Ann. (1.44).

⁹³ Cf. sup. on 16, 17, and 18; also fn. 40. On 63 itself, see further below, fn. 97.

⁹⁴ I am correcting Tenner’s ‘gegrundet (?) werde’, rejected by Götze, AM 257.

⁹⁵ If Götze’s suppletion of ku-is after ^INIG.BA-^DU-as is correct, we have still a second non-restrictive clause, ‘N., who was his first(-born) son’.

⁹⁶ Or perhaps Ή., who at this time was governing’ (so Götze, AM 19). If the adverb preceding the antecedent really belongs in the relative clause, this furnishes fresh proof that the antecedent itself belongs there.

⁹⁷ Cf. fn. 96, also 40. It should be noted, however, that in all these examples, the antecedent precedes the relative, so that at least they are a step beyond 63, where, though the clause is clearly restrictive, the antecedent is imbedded within the relative clause.

⁹⁸ Cf. instances noted above of adjective modifiers of the antecedent that followed the relative (23-28).

⁹⁹ However, I think we must class with it 73 (unless we exclude this for the reason noted in fn. 96) and 74, for to rule them out on the grounds of the extra subject alone might be to overdo logic in an effort at precise analysis. Probably the troublesome noun or pronoun that serves to repeat the antecedent at a later point is to be explained here—and also in 41, 42, 78, 79, and 84—as a mere bit of clumsy tautology somewhat like the repetition of a slightly different type noted below in regard to 76, 77, 78, and 82; in other words, its presence in the main clause need not force us to assign the antecedent to the subordinate clause along with the relative if there is no other compelling reason for doing so. But this ambiguity doubtless helped in the evolution of the older type into the newer one.

¹⁰⁰ This relative (and also those in 82 and 88, treated below) is of the sort of stuff from which the Latin ‘connecting relative’ is made, but I do not think Hittite reached this development, which even in Latin is rare before the classical period (cf. Hofmann, StolzSchmalz 709). A test of the difference is that the connecting relative clause cannot naturally be interpolated into the clause containing its antecedent, but must follow it. Cf. fn. 106.

¹⁰¹ Cf. Lg. 22.82 fn. 80; and to the list of 12 examples there referred to, and listed ib. 76 fn. 43, add the present one, 87.

¹⁰² Cf. sup., fn. 58. For a restrictive clause we may contrast the parallel passage KBo 3.4.4.40-1 nu-za ^DUTU-ŠI ku-in NAM.RA I-NA É LUGAL ú-wa-te-nu-un na-as 3 LI-IM NAM.RA e-es-ta ‘and the captives that my Majesty brought to the palace, they were 3 thousand captives’. It might be urged concerning our passage that there were other captives from D. beside the particular 3000 that the king took to the palace, in view of the statement in the Ten-Year Annals (3.4.4.39-43, just quoted in part) that after the assaults on Aripsa and Dukkama the king personally took 3000 captives to the palace, whereas his officers and his army took further numbers unspecified of captives and also cattle and sheep; but probably this general collection of booty was confined to Aripsa, and the only captives taken at Dukkama were the king’s 3000, since we learn from KBo 4.4 that Aripsa had been handed over to the whole of Hattusas to be plundered (4.13), and the troops had seized captives, cattle, and sheep (4.14-5), whereas Dukkama by surrendering specifically to the king saved itself from the same unhappy fate (18-24).

¹⁰³ There may be a causal element in the relative clause (cf. sup., on 3, 4, 5, 6, and also fn. 86 on 57), the idea being that the Azzians were frightened when they saw the king assailing fortified towns inasmuch as they themselves possessed a number of these.

¹⁰⁴ The normal order would be : ma-ah-ha-an-ma LÙ.MEŠ ^URUAz-zi ku-i-e-es URU.AŠ.AŠ.ḪI.A BÀD… har-kir nu LÚ.MEŠ ^URUAz-zi a-ú-e-ir … na-at na-ah-sa-ri-ya-an-da-ti ‘but when the Azzians, who held fortified towns …, saw …, they were frightened’.

¹⁰⁵ This passage has already been compared with KBo 2.5.4.7-9 (63), but characterized as less clear. Götze (AM 99) interprets ^VRUU-ra-as ku-is URU-as simply as ‘welche Stadt Ura’—which may well be right, in view of ^VRUU-ra-an URU-cm, right below (3.24). But in that case the interpolation of kwis between the name of the city and the appositive URU seems a little odd: the latter usually follows directly after the name (or the name plus initial particles if these are present), as in KBo 2.5.4.23, 3.4.2.19, 2.29, 4.4.2.15, 18, 23, 61, 3.41, 4.24, 5.8.1.28, 2.15-6; I have met just one exception, KUB 14.15.2.5 ^URUA-pa-as-sa-an ŠA ^IUh-ha-LÚ URU-an ‘Apassa, U.’s city’ (contrast the parallel passage, KBo 3.4.2.19 ^IU-uh-ha-LÚ-ya ^URUA-pa-a-sa-an URU-aw ‘U.’s Apasa the city’).

¹⁰⁶ This is the Latin idiom, as exemplified in Caesar, BG 1.10.1 non longe a Tolosatium finibus absunt, quae civitas est in provincia. Cicero, Att. 5.20.3 ad Amanum contendi, qui Syriam a Cilicia in aquarum divertio dividit; qui mons erat hostium plenus, which—in misleadingly abbreviated form—is cited as the typical example by most of the grammars (e.g. Stolz-Schmalz 711, Gildersleeve-Lodge §616.2, Lane §1809), seems to me quite different: mons is not used with the first qui, and the second qui is a connecting, not a subordinating, relative; cf. fn. 100.

¹⁰⁷ Cf. also Vert. Targ. 1.14-7, now mutilated but doubtless originally parallel with the others.

¹⁰⁸ The meaning of this word is doubtful. There has been considerable debate concerning it (see especially Friedrich, Vert. 1.76-7; Götze, Madd. 140-1; Friedrich, K1F 1.106-7, especially 106 fnn. 2 and 3; Sommer, AU 342-5, 395); but this does not affect our problem.

¹⁰⁹ As Sommer says (loc.cit.), the use of the plural to refer to a singular antecedent need not trouble us. Examples of this as well as of the use of the singular to refer to a plural (cf. sup., fnn. 51 and 88) are given by Hahn, Lg. 22.77 fn. 46.

¹¹⁰ Götze’s statement (Madd. 141 fn. 2), offered as a refutation of Friedrich’s interpretation, that the relative clause precedes ‘so gut wie immer’ and the antecedent is incorporated in it, of course applies to the vast majority of cases; but, as we have just seen, exceptions do occur. And Götze’s own interpretation, combining kwes kus LÚ.MEŠ kwirwanas with apedas antuhsas, seems objectionable, because of the interpolation of the hatrasi clause between the relative clause and the main clause to which the latter is supposed to apply (6, where the interpolated extra clause, KBo 3.4.1.25-6, is purely anticipatory of the main one, would not seem to justify this), and also (as in the case of 77) because of the lack of a sentence connective before the relative clause. Another possible argument—though, I admit, not a completely conclusive one—against him is that the use of apas after the combination of kas (or, for that matter, of asi or uni) with kwis does not seem normal; in cases where a pronoun follows them in the main clause (for the list, see Lg. 22.81 fn. 71), the only one that I have met is the enclitic personal pronoun (-as, -at, -si, etc.).

¹¹¹ I view this relative (and also the one in 83) as introducing a quasi-independent parenthetical clause, and thus somewhat like a Latin connecting relative (cf. fn. 100). Otherwise Sommer (AU 11), who takes the lewis clause as an ‘as to the fact’ clause (cf. Hahn, Lg. 22.77 fn. 49).

¹¹² Otherwise Sommer (AU 11), who interprets ‘(so ist) er … (nur) nicht ganz ein Schwager von mir’. But the introduction into the English translation of ‘nur’ without any basis in the Hittite seems to me unjustified and unjustifiable (cf. my refutation, Lg. 20.92 fn. 9 and 94 fn. 14, of Sommer’s similar treatment of several passages in the Bil.)

¹¹³ Already cited above as 47.

¹¹⁴ Quoted above (fn. 36) in another connection.

¹¹⁵ It refers to an already completely specified individual, just as do those in examples 69-76.

¹¹⁶ Cf. the discussion of the passage (47) above.

¹¹⁷ AM KBo 2.5.1.11, 2.4; 3.4.1.12, 1.20, 1.23, 1.32, 3.39, 3.43, 3.48, 3.58, 3.61, 3.90; 4.4.2.46; 5.8.1.1, 2.14, 2.29, 2.35, 3.24; KUB 14.15.3.28; 14.16.1.31, 3.30; 14.20.7; 19.29.4.20; 19.30.1.14; 19.37.3.36, 3.42. AU Tav. 1.32. BoSt 3.2.3.30. Hatt. 2.63, 2.70, 3.46, 3.70, 4.27, 4.62. Madd. 1.48, 1.58, 2.56, 2.85. Pest. I.1.4, 1.19, 2.43; II.3.1, 5.3, 8.4, 9.8; III.2.1.12. Tunn.1.31. Vert. Kup. §27.A.20; Man. B.3.6 and 7; Al. A.3.10, 3.31, 3.73, 4.38; Hukk. 2.56, 3.25, 3.66, 4.50, B.1.15.

¹¹⁸ AM KBo 2.5.4.7; 3.4.2.67; 4.4.2.3, 2.68; KUB 14.29.1.17; 19.29.4.11.

¹¹⁹ AM 3.4.1.10; 4.4.4.26, 4.29; 5.8.2.19; KUB 14.15.2.11; 14.17.2.31, 3.21. AU Tav. 2.73, 4.48. Pap. 1.31, 1.38, 2.15. Vert. Kup. §18.C.25; Al. A.2.77.

¹²⁰ Key-numbers go up to 86, but this is because two passages are quoted twice: 47 = 85, and 61 — part of 6.

¹²¹ As in fnn. 50, 52, 59, 61, 71, 73, 78, 80, 89, 91, and 102.

¹²² As in fnn. 46, 69, 70, 80, and 105.

¹²³ These 78 passages represent 79 examples of relatives (cf. fn. 19), since the one listed here with the key-number 9 contains two separate relatives, Vert. Man. B.3.6 and 3.7 (364 and 365 of my earlier paper).

¹²⁴ Only 20, including 6 peculiar cases (listed in fn. 118) and 14 clear-cut cases (listed in fn. 119).