Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://imk.msu.ru/Structure/Linguistics/yakubovich/download/lydian.pdf
Äàòà èçìåíåíèÿ: Fri Apr 30 15:11:28 2010
Äàòà èíäåêñèðîâàíèÿ: Mon Oct 1 20:04:29 2012
Êîäèðîâêà:

Lydian Etymological Notes
Lydian philology is not so prominent in Anatolian studies now as it was in early sixties in the last century, and for a good reason.1 Even though our grasp of Lydian texts still does not go beyond understanding the most stereotypical inscriptions and formulae, further achievements of combinatory analysis in this field do not seem to be possible without new sources. Although keenly aware of the pitfalls of etymological discussion that is not based on solid synchronic data, I still feel that the comparative method can nowadays offer more rewarding results to a scholar of Lydian, than any other approach. Recent progress in Anatolian studies, especially in our understanding of Luwic languages,2 leaves hope that some of its by-products can also be gleaned in the field of Lydian. The morphemes and lexemes I have selected for the present discussion all belong, likely or assuredly, to the vague category of grammatical elements. The reasons for this decision were purely practical. Grammatical elements are, on average, more stable than their lexical counterparts and, in addition, are normally subdivided into small classes of segments characterized by uniform (morpho)-syntactic behavior. This makes their comparative analysis less dependent on understanding the full meaning of all contexts in which they occur. In this article, I have indeed frequently avoided translating the contexts cited, even in those cases where such attempts have been made before. Doing otherwise would have forced me to take sides in unsettled disputes that do not have much to do with the essence of my account, and eventually could have rendered this paper unreadable. § 1. Prefixes da-, EQ-, and EW- can be shown to have existed in Lydian by distributional analysis. The following minimal pairs testify to the existence of d(a)- in verbal forms:3

Usual disclaimers apply. I am deeply obliged to Th. van den Hout (Chicago) and C. Melchert (Chapel Hill), who commented on the substance of this paper, and to Th. Wier (Chicago), who took pains to improve its style. 2 The Luwic language group embraces Luwian proper, Lycian, and probably Carian, as per Melchert 2003: 177, fn.7. 3 Here and below, the Lydian inscriptions are numbered as in Gusmani 1964.
1

76 3sg. 3sg. 3sg. 3sg. / / / / p p p p l l l l . . . . pr pr pr pr t. da-ca-l (10.4) t. fa4-d-o-l (3.2) es. fa-d-in-t (5.5) t. fa-d-i-l (15.1)

Ilya Yakubovich

3sg. 1sg. 3sg. 3sg.

/pl. pres. ca-t (11.6) pres. fa-o-w (24.2, 24.18) /pl. prt. in-t etc. (23.20, 24.11) /pl. prt. i-l (12.6, 13.1)

In addition, the existence of the same prefix can be surmised in the following forms, both verbal and nominal da-troL- (nom. 11.8, 13.3, acc. 22.7),5 3sg./pl. prt. da-wic-il (13,7),6 part.(?) nom. n. da-cuwe-U-t (23.1), 3sg./pl. pres. »a-krod (44.4). More questionable is the status of verbal forms 1sg. prt. fa-dofi-dv (10.11), 3sg./pl. prt. fa-difi-l (24.4), and 3sg./pl. prt. f-EGDYR-t (54.3), which admit of different morphological interpretations. It is clear from the examples above that the prefix dacan lose its vowel in sandhi before the root vowel. With all due reserve caused by our poor understanding of Lydian texts, one can claim that the verbal roots combined with da- belong to the "positive" semantic field. The earlier meaning assigned to ca- was `to set up, dedicate', while C. Melchert believes that verbs formed from this root can mean both `to promise, pledge' and `to agree, assent'.7 The verbal root o- also seems to refer to some sort of verbal activity.8 The root of the cognate stems i- and in- is probably the same as that of Hitt. ie- and Luw. - `to do, make', and the meaning of the Hittite and Luwian verbs suits the appropriate Lydian contexts.9 Lyd. cu(we)- is routinely compared with Luw. tuwa- `to put'.10 The hypothesis that da-

The prefix fa-, which appears always at the very beginning of a verbal form, can also introduce the clause-initial enclitic chain, in this case frequently having the meaning `then' (e.g. in 1.6, 3.5, 4a.3, 4b.4 etc.). The two functions of this element can be explained together if it was cognate with Hitt. SSD that both occupies the first place in a chain of preverbs, and can appear clause-initially with the meaning `then'. This solution, offered as one of possible alternatives in Carruba 1959: 33, 35, is preferable to the splitting approach of Heubeck (1959: 45-6), who connected the Lydian prefix fa- with the Hittite orientation prefix pe-, and linked the Lydian particle fa= with the Luwian contrastive particle =pa. Note that Hitt. pe- cannot be followed by any preverbs. 5 Cf. da-WURLV NVDU\ (13.3) vs. saris-WUR (11.1, 11.7). 6 For the analysis of ¥wic- see Melchert 1997: 47 ff. 7 Melchert 1997: 39-41. 8 Gusmani 1964 plausibly translates faow as `I announce (vel sim.)'. Melchert 1991: 135 (fn. 7) derives Lyd. fa-do-l `to put in (vel sim.)' from virtual *tuwV- `to put, place', but a more credible reflex of the same root seems to be Lyd. cu(we)- (Melchert 1994: 334). The sentence ardEF DODUP IDGRO ZWD is rather obscure, but the translation `I declared it my ardEW during my lifetime' would fit it as well as `I made it my ardEW during my lifetime'. 9 Melchert 1992: 42. 10 Cf. Melchert 1994: 334 with ref.
4

Lydian Etymological Notes
WURL

77

- contains the stem tro- `to offer, grant (?)', on which see immediately below, is not assured, but likely. The existence of the EQ- prefix can be demonstrated with the help of the following verbal minimal pairs. 3sg./pl. 3sg./pl. 3sg./pl. 3sg./pl. f-EQ-t pres EÓ-u-d (23.21) prt. f-EQ-ani-l (15.2) prt. f-EQ-ca-l (50.4) prt. EQ-tro-l (14.16), ro-l (45.2) 3sg. 3sg. 3sg. 3sg. / / / / p p p p l l l l . . . . pr pr pr pr t. k t. k es. es. at-u-l (10.7) at-avi-l (2.3) kan-ca-t (13.10) kan-tro-d (11.12)

Of equal probative value is the adjectival pair EQ-tarfo-d `belonging' (12.1) vs. b-tarwU[d] `id.' (1.5). The connection between the adjectival (?) forms cawo-(12.2) and EQ-2awo- (14.18) is more questionable.11 The forms where EQ- is to be surmised for structural reasons include 3sg./pl. pres. EQ-sibi-d (3.3, 4a.5), f-EQ-sibi-d (passim), 3sg. med. (?) EÓ-sarb-tat (10.7), acc.sg. (?) f-EQ-tasEQD-v (12.10, 22.4). Most of the verbal stems of this group again have "positive" semantics. Stems u- and o- treated above are thought to have similar meanings12 and may even be cognate. ani-/ avi- has been compared with Hitt. an(n)iya-, Luw. DQQ¯- `to work, make', and the meaning `to make' nicely matches the context of (2.3).13 The stem tro(d)- `offer, grant' cannot be separated from Luw. (pari) WDUZL\D- `to deliver'.14 At the same time, the verb (f)-EQ-sibi-, occurring more than fifteen times in Lydian inscriptions, clearly has a negative meaning `to cause harm, damage' (vel sim.). It is also noteworthy that the prefix EQ- interchanges

With Melchert 1997: 34 2 and c arose in some cases from the palatalisation of *-tand *-d- respectively. I am skeptical, however, with regard to attempts of explaining the synchronic contrast between 2 and c as voiceless vs. voiced. A mere look at the vocabulary part of Gusmani 1964 suggests that c- occurred word-initially more than five times more frequently than 2-, which would indeed be peculiar if c- were a voiced fricative. Perhaps, c- represents a laminal affricate [t,], or even a palatalized stop [tj]. 12 Gusmani 1964: 31. Gusmani (op.cit.: 217-9) translated Lyd. u- as `to write' using both combinatory and etymological evidence. While the meaning `to write' is compatible with the contexts available, the suggested comparative evidence is no longer tenable. HLuw. **KZD¯ `to write' is now read as REL-za- `incise' (Hawkins 2000: 70). This stem can be read /kwanza-/ and represents a transparent derivative of Anat. *gwen`strike'. 13 Gusmani 1964: 124 with ref. 14 Melchert 1992: 53.
11

78

Ilya Yakubovich

in several cases with the prefixes kat- and kan-, cognate with Hitt. katta `down' and kattan `below' respectively.15 The morpheme EW- was apparently less widespread than both da- and EQ-. The stems attested with EW- do not occur with other prefixes, and yet its existence can be safely posited on structural grounds. It is attested in part. nom.sg.n. EW-R-U (2.3), 3sg./pl. pres. EW-ol-t (14.5), 3sg. pres. med. (?) EW-qra-tad (23.17), 3sg./pl. acc. sg. EW-wers-n (24.14, with enclitics 23.11), and 3sg. med.(?) f-EW-win-tat (12.4).16 It is likely but not certain that the noun or adjective EWDPY- (10.18, 11.5, 23.19, 23.21) and the derived verbal form 1sg. prt. f-EWDPYL-dv (23.18, 24.6) contain the same prefix.17 The prefix EW- is synchronically distinct from EQ-, which does not undergo any morphophonological changes in a position before -t (cf. EQ-tro-l (14.16) and other examples above). While the existence and distribution of these prefixes is generally agreed upon, their etymology remains problematic. The most credible is Carruba's claim that Lyd. EW- directly corresponds to the Hittite preverb anda.18 In the same article, O. Carruba also maintained that EQ- is cognate with the Indo-European prefix *en- reflected in Gk. ±2- and Lat. in-. Although this suggestion is phonetically possible,19 it makes one wonder why Indo-Hittite *en- left abundant traces in one Anatolian language of the first millennium BC, but is not directly attested anywhere else within Anatolian. The prefix da- was compared with Slavic preverb and preposition *do-, or, alternatively, with the element *to attested e.g. in the Old Hittite clause-initial particle ta.20 Since, as

The comparison between Lyd. kat- and Hitt. katta is a commonplace. For the connection between Lyd. kan(tro)- and Hitt. kattan see Gusmani 1982: 63 with ref. The regular correspondence of Hitt. kattan would be Lyd. *katn, but it is only natural to assume that the consonant cluster -tn- was simplified to -n- before the root-initial consonant. 16 The last of these stems was compared with the putative 3 sg. med. win-tad (23.2) in Gusmani 1964: 126, but the last letter of this word admits of several readings (see the remarks in Gusmani 1982: 108). 17 My personal preference is to analyze EWDPY- `decree, designation (vel sim).' as *EW-ta-m(n)-, derived from ÿta `to put, place'. A possible alternative is to assume that it was formed from a verb cognate with Hitt. handai- `to entrust, fix, etc.'. The choice between these two possibilities is contingent upon our better understanding of the meaning of the words involved. I reject the persistent attempts to connect EWDPY- with the Anatolian word for `name' (Hitt. laman (n.), Luw. */D/DPDQ-/ (n.)), which are unfounded both phonologically and semantically. 18 Carruba 1959: 32. 19 Note, however, that it implies an old univerbation of the prefix and the root (otherwise one expects *en > EY). 20 Gusmani 1980: 48. The last comparison still lingers in Melchert 1994: 338, accompanied by two question marks.
15

Lydian Etymological Notes

79

Melchert 1997 has shown, all the initial coronal stops merged into *t- in Proto-Lydian, neither of these two otherwise questionable etymologies can be regarded as phonetically satisfactory. Since Lydian is usually regarded as a rather isolated language within Anatolian,21 one can understand why scholars persistently try to explain Lydian facts through the Indo-European data. It is nevertheless surprising that nobody, to my knowledge, has tried so far to compare the system of prefixes in Lydian and Lycian. In our case, this comparison brings an immediate result: Lyd. da- is matched by Lyc. ßte `in(-)', ultimately cognate with Hitt. anda, CLuw. QGD (IH. *endo) ; Lyd. EQcorresponds to Lyc. EQE `under(-)', cognate with CLuw. annan ; Lyd. EWcan be compared with the heretofore isolated Lyc. EWL `down'.22 It is easy to observe that the suggested scheme reveals two isoglosses between Lydian and Lycian that are not shared by Luwian. The aphaeresis in Lyc. ßte and Lyd. da-, probably caused by some sort of sandhi, contrasts with the preservation of the word-initial vowel in CLuw. anda,23 while Lyc. EWL and Lyd. EW- seem to lack direct cognates in Luwian. Future research will show whether these similarities reflect secondary contacts between Lydian and Lycian, betray their original dialectal isoglosses, or are pure coincidences. None of the three correspondences is laden with irresolvable phonological difficulties once we assume that Lydian preverbs once functioned as independent phonetic units.24 The Lydian apocope takes care of Lyc. EWL ~ Lyd. EW-. Lyc. EQE can correspond to Lyd. *EQY with an expected syncope of an unaccented vowel in a final syllable;25 it is natural to assume that *EQY was simplified to *EQ before consonants, once the univerbation took place, and then, analogically, also before vowels. The comparison between Lyd. da- and Lyc. ßte allows one to advance a hypothesis about the new origin of the mysterious initial d- in Lydian it can go back to word-initial pre-nasalized coronal stops.26 In fact, there

21 For the survey of modern views on the place of Lydian within Anatolian see Melchert 2003a. 22 Melchert 1993a: 24, 49. 23 See Melchert 1993b: 18 for the graphic variants and attestations. 24 The recent character of the agglutination of preverbs EQ- and EW- with the following verbal forms can be independently deduced from the aberrant phonological behavior of prefixed forms. They may contain two vowels that were historically accented according to Eichner's rules (Melchert 1994: 350 with ref.). 25 Melchert 1992: 38-39. 26 According to Melchert 1997: 44, the initial d- in Lydian must be discussed together with the rare cases of the initial dd- in Lycian. Unfortunately, none of the Lycian words belonging to this group has a convincing meaning assignment or etymology, and many

80

Ilya Yakubovich

are reasons to think that pre-nasalized coronal stops can loose their nasal component also in Lycian. It is likely that some or all Lycian words beginning with dda- and dde- contain in fact the historical preverb ßta- / ßte- that merged with the root. It is also noteworthy that another Late Anatolian language, Carian, had a special grapheme for /nd/,27 which indicates that this segment could be perceived as a monophonemic combination in this language. In Lydian, however, the development went further. Since neither pre-nasalized, nor voiced stops were a part of the Lydian phonological inventory, the secondary word-initial */nda-/ yielded */da-/ and then //a-/ (graphically da-). The semantic side of our correspondences cannot be fully assessed until we know the exact meaning of the Lydian verbs involved. Yet, what one can glean from the available data, speaks rather in their favor. The alternation between EQ- and kan-/kat- receives an explanation if we assume that these prefixes belonged to the same semantic domain (the direction or location downwards).28 The fact that the "negative" verb (f)EQ-slibi- `to damage (vel sim.)' is endowed with a prefix of this group is also compatible with my hypothesis. Finally, it is certainly comforting to think that the cognate of Hitt. anda is not a relatively little-attested Lyd. EW-, but a common prefix da-. While the suggested scheme of correspondences has its obvious intrinsic merits, it gains additional support from considering two other Lydian words beginning with d-. The first one is the clause-initial particle dum-, traditionally translated as `but' or `while'. Vetter (1959: 26 ff.) convincingly argued that this word, usually appearing at the beginning of an enclitic chain, has purely syntactic function, but his exact translation of this word has less to recommend itself. In four out of the seven occurrences of this word (10.3, 12.2, 14.1, 22.2), it introduces the second sentence of an inscription, immediately following basic information

of them resist morphological analysis. Yet cf. dderije (44b.17) vs. erije (29.7) and ddezedu (107b.2) vs. zedi (26.19, N324.29), as well as PN. Ddaqasa (88.1,2) vs. qas- `to strike repeatedly' and Ddepßnewe (98.1 etc.), probably to be analysed as dde-epß-newe. 27 Usually transliterated as /, e.g. in Car. trqu/H, WUT/- < Tarhund-. Adiego (2000: 145) puts forward a hypothesis that the grapheme / was used for the etymological *nd that was synchronically realized as a voiced stop /d/ (distinct from the correspondent IULFDWLYH / d). This hypothesis definitely cannot be extended to all Carian dialects ; cf. a Carian PN. NL/EVLDWWHVWHGDV /.% "LQ*UHHNWUDQVPLVsion. 28 For the co-occurrence of both groups in the same language cf. Hieroglyphic Luwian preverbs SUB-(n)an `down' and INFRA-ta `id.' usually compared with CLuw. annan and Hittite katta respectively (see e.g. Hawkins 2000 II: 456). The ideograms SUB and INFRA were rendered by the same sign in the Luwian writing system.

Lydian Etymological Notes

81

about an inscribed stele (Lyd. mru-). It may be appropriate to quote these contexts in full: 10 12 14 22 t] .:ud karo katowalis bil (2) wora[ ka]rol katowa (3) wora b [bilk] iskubto. dum =m=is sellis (a large space) (4) ZHIDNZRUDGDFDO. m ru siwÖml saro EQWDUIod. (2) dum ==is nid qaasl ZHIDDUDUP cawR. (1) m ru=m=it baal. dum =is q[...] factot (1) est m rud IDUGEQ2 facviris mLPQD ZLFY29 (2) GXP P ebad fawnEULVDPEGIDUGEWDWUNR (1) es[ karo empty (1) est

In none of the four passages cited above does the second sentence have a secure interpretation.30 It is difficult, however, to imagine that in all these four examples, immediately upon learning the purpose of each stele, the readers are presented with new information that contrasts with, or contradicts to one previously stated.31 It is also suspicious that the particle dum= does not appear in those cases where we understand the Lydian text well enough in order to surmise a contrastive discourse (e.g. blessing vs. curse formulae in 3.3-5, or the prescription to carry out the orders followed by a curse against those who fail to comply (?) with them in 23.17-22). Yet another troublesome fact is the absence of any ideas about the origin of this putative contrastive particle,32 while most of the other Lydian clause-initial elements have been already provided with etymologies, in many cases quite obvious. Anatolian particles form a closed set, and introducing a new element, with no external parallels, to this set must represent the last resort of a comparative linguist. These problems can be successfully resolved if we assume that dum = is a particle introducing a new topic, cognate with Hitt. anda=ma `in

Melchert (1997: 43-44) suggests that line 1 of inscription 22 contains two clauses, the second one beginning with facviris. 30 In spite of numerous attempts to interpret well-preserved inscriptions 10, 12, and 22, I am not aware of a coherent translation of these texts as a whole. The intuitive interpretation of individual fragments, such as the tentative translation of 12.1-2 in Vetter 1959: 37, does not have much probative value. 31 The other three contexts featuring the element dum= (12.9, 14.14, 44.8) are too poorly understood to allow for any kind of generalizations. 32 One must prophylactically reject the comparison between Lyd. dum= and Lat. dum `while' as phonologically impossible. All the original word-initial coronal stops normally fall together into t- in Lydian, as per Melchert 1992.
29

82

Ilya Yakubovich

addition, furthermore'.33 Given the suggested sound change */nd-/ > //-/ in Lydian, the consonantal correspondences are regular. The vocalic development suggests that Lyd. *d=um originated in a context where no enclitics were attached to it; =um is a regular allophone of Lyd. =m= ~ Hitt. =ma= in word-final position. After the univerbation of dum, its shape was extended to the cases where it formed one phonetic unit with subsequent clitics. The semantic side of our comparison is unproblematic: after giving us the basic information about the stele, the authors of the inscriptions provide additional details. For example, they could mention those who assisted in its construction, or prescribe who can or cannot use it as a burial monument in addition to the sponsor. If some contrast were present in one or two of the examples quoted, it would not be very surprising: after all, Lyd. =m probably had a contrastive function. Another Lydian etymology to be clarified is that of the postposition (Dat.) + dÖv `from', the meaning of which was firmly established by Melchert (1997: 35-8). It is a phonetically perfect match to the Hittite postposition andan `in, on', while the Carian pre- or postposition /en /nden/ (1x, MY 2) shows the intermediate stage of its phonetic development. The semantic change in Lydian is certainly striking, but such functional shifts are not without parallels in world languages. Thus, the English preposition from goes back to Old Engl. fram, originally an adverb `forward', while Russ. ot `from' is cognate with the Lithuanian prefix at-, the original meaning of which was the direction towards the speaker.34 In our case, the construction *Abl. + dÖv could originally mean `from within smth.' or `from smth. inward' , while dÖv acquired the ablative meaning after the Lydian case system was remodeled. If my above conclusions are correct, the aphaeresis in the Lydian reflexes of *endo was extended to its derivatives. This is in perfect agreement with the situation in Lycian, where the aphaeresis occurs in ßte-pi `in', ßtewE < *ßte-tewE `toward' etc.35 The new sound law Lyd. */nd-/ > */d-/ > //-/, consistent with the phonetic evolution of other Anatolian languages, makes it possible to provide a unified explanation for three morphemes da-, dum=, and (-)dÖv, which previously lacked Anatolian etymologies. It does not infringe on the other sources of the voiced coronal phoneme(s) in this

33 anda could also be used as a discourse particle without =ma, both in Old Hittite and Luwian. See Yakubovich 2002: 209 for the examples. 34 One can further compare the Hittite ablative ending in -az < -(o)ti going back to the adverbial particle *-ti with a likely locative meaning (cf. Melchert 1977: 454). 35 Differently Melchert 1993: 50.

Lydian Etymological Notes

83

language. Thus, Lyd. /// could develop from /*t/ by lenition, as in Lyc. taada- `father' ~ Lyc. tedi-, Car. ted `id.' < Pre-Anat. (baby word) *tÀta- `daddy', or from /y/ as in 1sg. verbal ending -idv < IH. *-iyo-m.36 Since Luw. annan is prototypically a postposition, and only secondarily acquired the function of a preverb, the existence of the Lydian preverb EQ- would be hard to believe, if the homonymous postposition were not attested in the same language. Fortunately for my hypothesis, the independent lexeme EQ occurs in 12.7 (nÖ=v sirma EQ Ön ismEQOLV IDWR), where the context does not leave any doubt that we are dealing with a postposition.37 It is also possible that EQ DW IDIEQZY in 13.8 shows the fronting of the preverb EQ-. The context of EQY (or EQ Y?) in 44.15 is completely unclear, and so one can only speculate whether this word represents an older form of the postposition/ preverb EQ(-).38 I conclude this section by listing the Lydian prefixes and related words mentioned in it in a tabular format. The last three comparisons, which are not my own, have been included into this table for the sake of completeness. da(-)dÖv dum=
EQ-, E E Wfa kankatQ

Hitt. anda, CLuw. QGD, HLuw. a-ta, Luc. ßte `in(to)' Hitt. andan, CLuw. andan, HLuw. a-ta-an `in' Hitt. anda=ma `furthermore'. Cf. HLuw. a-tÀ-ha /anda=ha/ `id.' CLuw. annan, HLuw. SUB-na-an, Lyc. EQE `under' Lyc. EWL `down' Hitt. SSD `behind, back; then', CLuw. SSD `back, again' Hitt. kattan `below, along', HLuw. CUM-an `with' (?) Hitt. katta, HLuw. INFRA-ta `down'

Cf. Melchert 1997: 44-5. The postposition EQ occurring in this context was compared in Gusmani 1964: 105 with an unclear Hittite adverb annan that occurs with a gloss-wedge in a letter of Puduheba (Friedrich-Kammenhuber 1975-, 76). Should my etymology of this postposition be correct, the meaning of VLUPDEQ is likely to be not `in the temple', as per Gusmani (loc. cit.), but `near the temple', or perhaps `under [ the authority of] the temple'. 38 At the same time, I must make it clear that a definite answer with regard to the etymology of Lyd. EQ can be arrived at only through the refinement of our understanding of Lydian inscriptions. Should it become clear that the preverb/postposition EQ signifies an inward location, Carruba's connection with IE. *en `in' must be reinstated in its rights. Unlike in the case of da- (and, by extension, EW-) my etymology of EQ is not tied in with the interpretation of other Lydian data.
37 36

84

Ilya Yakubovich

§ 2. Lyd. arlylli-, arlili- `(one's) own' is attested three times in relatively well-understood quasi-parallel inscriptions 23 and 24. A pair of parallel curse formulae is cited below: 23.10-11. fak=m=it=in qdÖn=k artimu=k katsarlokid bu bil=k arlyll qyra `... then QdÖn and Artemis will curse (?) him, him and his arlilproperty'. 24.12-14. IDN P LDUWLPXLELPVLVNDWVDUORNLGEELO NDUOLOTLUD `... then Ephesian Artemis will curse(?) him, him and his arlil- property'. Lyd. qira- (1.8) is matched in the Bilingual 1 by Aram. qnyn`property'. Right in the next clause of each inscription, the same concept is expressed without mentioning the word qyra-itself: 23.11-14 nik bili `He will he, nor
EWZHU=k=in saroka=k esvÖv ciwav nikumEN VDZEQW QLN ELV s arlyllis never enjoy(?) ... and protection(?)39 from these gods, neither his arlil- `.

24.14-16 artimu=k=in EWZHUVQ VDURND N QLNXPEN VDZEQW QLN ELV QLN ELOIEQLV" `He will never enjoy (?) ... and protection (?)from Artemis, neither he, nor his goods' (?).40 Thus, the same word arlili- (with a scribal variant arlylli-) can either modify the word for `property', or be used as its substitute. It makes good sense to follow the analysis of Gusmani (1964: 61) and translate this word as `own, proper' in its adjectival meaning, and `property', when it is substantivized. An alternative school of thought assigns the meaning `relative' to sfEQ9-; this interpretation destroys the unity of curse formulae in inscriptions 23 and 24, but may better account for the dat. pl. sfEQDY in the fragmentary context 42.5. In such a case, one has to assume that arlili- could be used in an elliptical construction `one's own

I tentatively take saroka- as a cognate of sarEWD- `protector (?)'. I tentatively take sfEQL- as a derivative of asfa- `property, goods (vel sim).' and regard both words as ultimately related to Hitt. assu- `good'. This comparison is, of course, not assured because the conditions of aphaeresis in Lydian remain unknown to us. In any case, cf. Lyd. sfEQGD- that probably also refers to material possessions, as per Gusmani 1964: 204.
40 39

Lydian Etymological Notes

85

[relatives]'.41 Whatever of the two interpretation one chooses, basic meaning of arlili- must have been `own, proper'. Now, there is another Lydian word that is taken to be cognate with arlili-. This is the declined form alarm- that occurs exclusively in funerary inscriptions with the variants ararm- and alarma-, and has an adjectival derivative alarml-. The full list of its occurrences is given below: 3.2 10. 12. 12. 13. 14. 14. 16. 80. ardEFDODUPIDGROZWD 16-17 fa=k=um Ön islodaalarmnNDWDFY 2 GXP LVQLGTDDVOZHIDDODUPFDZR 6 LDLODODUP2HD 1-2 kud=mÖk cEQVLGY DODUPD amu=k nÖsi=k E
ZWDD

PLV LVWDPLQOLV

WROW

5 11 4 1

ak=um adEQY)>a] larmlv E ...] DUDUP facato[.] ...] ODUPVLG> a] lDUP bis fatis[?] .[

None of the contexts cited above has been fully elucidated. Vetter (1959: 34-39), having discarded the earlier translation of alarm(a)-, `founder', opted for taking this word as a pronominal element `-self' (`selbst'). Although this hypothesis was not based on the strict application of combinatory analysis, it brought important insights to our understanding of Lydian funerary inscriptions. Presumably, the word alarm- appears in those cases where the sponsor of the tomb designated it for himself, not only for some of his relatives. Hajnal's tentative translation of 13.1-2 `where I myself during my lifetime consecrated myself and my family members' reflects this basic intuition.42 The apparent similarity between the roots arlili- `own, proper' and alarm(a)- increases the chances that Vetter's translation of the latter is essentially correct. Unfortunately, his morphological analysis of this word can not any longer be maintained. Vetter took ararm- as an original stem and hypothesized that it contains the reduplication of the basic element ar-, which can be seen also in ardEF `with his own means (??)' (3.2) and arwol `to appropriate (??)' (2.9). Since that time it has become clear that **DUDUP in 12.2 is to be read DODUP,43 and DUDUP

41 42 43

Gusmani 1980: 32. Hajnal 2001: 9. Gusmani 1980: 32.

86

Ilya Yakubovich

in 14.11 remains the only instance of a "non-dissimilated" form. Since the adjective a]larmlv is attested in the same inscription, DUDUP is best taken as a nonce creation caused by a scribal anticipatory error. I believe that Lyd. arlili- `own, proper' and alarm(a)- `-self' must be considered together the family of Luwic words that includes HLuw. (COR) at(a)ri- `person, self; soul', Lyc. atra-/atla- `self', and Car. ot2r`person, self'.44 The Proto-Lydian forms could be *atarÌli `proper', endowed with a common adjectival suffix, and *atÀrama- `personal(ly)' probably containing genetically the same suffix as Luw. -ama/i- seen in /massanama/i-/ `one who belongs to the god' or /tiwadama/i-/ `of the Sun-god'.45 Later, the root *atar- underwent an irregular sound change *-t- > -l-, synchronically attested in the variation antola / anlola `statue', and 3 sg./pl. pres. silawad vs. fa-sitawad,46 and this yielded the intermediate forms *al(a)rili and *alar(a)ma-. Finally, after the syncope took place, the unusual consonant cluster -lr- was simplified through metathesis in the case of arlili-.47 The part of this analysis that may seem least founded is the assumption of an irregular sound change *t > l. There is an obvious temptation to regard the variation t ~ l in Lydian as a late and dialectal phenomenon. And yet, there is one more case where its extrapolation to the history of Lydian seems to be necessary. I believe that it can help to resolve the problem of the Lydian 3 sg./pl. preterite ending -l, which can be put in the form of a paradox: on the one hand, common sense dictates that it must continue an Anatolian 3 sg./pl. preterite verbal ending,48 on the other hand, it cannot be derived from any of these endings by regular sound laws.49 If we start with the Proto-Anatolian 3sg. prt. endings *-t

On these Anatolian forms, see van den Hout 2002. On this suffix, see Melchert 2003b, 195. Gusmani 1980: 22. 47 The combination -lr- is not attested in Lydian, while -U- is attested only once (NDWDUHV, 5.2). The consonant cluster -rl- is attested at least 24 times in the published corpus of Lydian inscriptions. For a syncope triggering metathesis, cf. e.g. the case of Lyc. *ebehi > *ebhi > ehbi `his/her'. 48 Melchert's suggestion that 3 sg. /pl. prt. -l is akin to the Slavic preterite suffix -l- of participial origin (e.g Melchert 1994: 379) stumbles against the fact that this ending was not extended to all the preterite paradigm, as happened in Slavic, and as one would expect of a participial formation. Cf. also Oettinger 1978: 86, fn. 50. 49 Oettinger (1978: 86) argues that the /d/ of Lyd. -da < *-ta was dorsalized before back vowels and then this dorsal sound became -l in word-final position, after the apocope had taken place. I am not aware of any language where the phonemes /d/ and /a/ are present, but the phonetic segment [da] is impossible. Note that Anat. *-d is preserved in Lydian, as e.g. in the enclitic pronoun -ad `it'.
45 46 44

Lydian Etymological Notes

87

(active) or *-to > *-ta (medio-passive), and assume that they could have variants *-l or *-la in Proto-Lydian, then the generalization of the lateral reflexes can be explained by the functional necessity to differentiate Lyd. 3 sg./pl. prs. -t/-d from Lyd. 3 sg./pl. prt. -l. It is likely that the irregular change *t > l originally represented some sort of distant assimilation, as in anlola- and arlili-, but subsequent analogical changes and/or dialect mixture obfuscated its distribution.50 The suggested scenario has repercussions for the etymology of Luwic *at(a)ra/i- `soul, self'. Hajnal (1995: 244-5) mentioned Gk. ã845 `heart; seat of passion, feeling, desire' as its possible cognate, while Hawkins (2000: 460) compared it with Hitt. es(sa)ri `form, image, figure'. Finally, Kassian (2000: 80-1) established the existence of Hitt. attes (plur. tant.) `soul', which may be related to the Luwic noun at the root level. If my comparison between Luwic and Lydian forms is valid, it tips the scales in the favor of Hajnal's and Kassian's suggestions, as opposed to the one made by Hawkins. We do not have evidence that the irregular change *s > t, required in order to link Hitt. es(sa)ri with Luw. *at(a)ra/i-, occurred outside the Luwic group of the Anatolian languages.51 § 3. The Lydian particle =in is attested more than twenty times, always at the very end of an enclitic chain. In some cases, it appears to have an additive function, as the examples cited below can demonstrate: 1.3-6 ak=it n>Öqis] esPUXEXN52HVZÖnaEXNHVYDYODTLULVDYEXN LW ! NXGLVWHVZÖnaEWDUYU>G@DN W=in nÖqis qel NIEQVLILG .. . `Whoever causes harm to this stele, or to this tomb, or to this dromos or to whatever belongs to this tomb, or whoever causes harm to anything else...'

I believe that the Lydian rendering of Zeus and Demeter as lew and lamEWUX reflects a different phenomenon. It is likely that Lydian, like Lycian, did not have native words beginning with plain /d-/, but only with its "intensive" counterpart /dd-/. This variant could originally go back to pre-nasalised coronal stops (cf. the previous section), but then it could be generalized as an allophone of /d/ in word-initial position. Lydian lwas chosen as a nearest approximation of initial voiced coronal stops in foreign words. 51 On this change, see Melchert 1994: 274. In connection with the etymologies of Luw. *atra/i-, cf. also the semantic discussion in Yakubovich 2002: 194-7. 52 I derive Lyd. buk `or' from an earlier *apa=we=ku, where *apa `after, then' is an adverb normally yielding Lyd. fa=, *we continues IH. *w `or' probably forming a second part of CLuw. na-wa `not', and *ku < IH. *kwe `and'. This enclitic complex should have yielded *bok (phonologically /pok/), which later became buk in unstressed position. Cf. typologically MPers. \E `or' < Ir. *DGD Z DSL lit. `then-or-also'. For earlier etymologies, see Gusmani 1964: 87.
50

88 2.2-4
DN H

Ilya Yakubovich

Z

Öna

PUX

N E DVR

N

L2 TLG NDWDYLO ODTULVD

N

i n q i d E

WRU

DGNDUROLGVDE DOLG

`This tomb, stele, b. that I made (for myself), as well as the dromos that ..., [all] that belongs to Karo, son of 6DED.' 2.4-8 ak=t=in nÖqis fEQVibid esvav mwEQG)Y LVNRQ TLG D WDPY EXN wÖnaHVEmNPUXEXNEDVRHVEXNODTULVDYEXN inDEYDZÖv ... `Whoever causes harm to all these m. that I built, or to this tomb, or to this stele, or to this b., or to the dromos, or else to any other object ...' At the same time, there are many other contexts, in which =in appears to have no function at all. Thus, the first occurrence of =in in (2.4-8), seemingly connecting this sentence with (2.2-4), can be contrasted with many other examples, where no such particle is required for the transition from the ownership information to the curse formula in a funerary inscription (e.g. 1.3, 3.3, 6.2 etc.). Even more instructive is the contrast between parallel clauses (23.10-11) and (24. 12-14), cited in the previous section, the first of which contains =in, but the second does not. Finally, in the following example, representing a complete inscription on a terracotta vase, the particle in question clearly does not have any syntactic meaning: 30 tSWL LQEPYWLVDUGYIDELODWD `Titi gave my t. to Ata Kitva-.'
NLWZD

Given the fact that =in does not occur automatically in every Lydian sentence, it cannot be taken as a structural element, the sole purpose of which is to demarcate clauses. The natural hypothesis to entertain is that it had some sort of weak pragmatic function, which could be transformed into the additive one, or vice versa. I believe that the asseverative function fully fits this description. On the one hand, asseverative particles frequently cannot be idiomatically translated into other languages; classicists know the difficulties of rendering the meanings of the Greek particle 4Â2. On the other hand, English examples like `I dislike firebreathing dragons, or indeed any kind of dragons' illustrate the additive usage of the English asseverative particle indeed in a construction "X or indeed Y" (X Y). In fact, the passage 2.4-8 can be translated as well `Whoever causes harm to all these m. that I built, or to this tomb, or to this stele, or to this b., or to the dromos, or indeed to any other object ...'.

Lydian Etymological Notes

89

The usage of Lyd. =in can be compared with that of the Hittite emphatic/asseverative particle imma. The original usage of this particle is most patent in rhetorical questions e.g. VBoT 2, 7-9 nu mQ KDQGQ DPPHO DUMU.MUNUS-YA sanhiskisi nu=tta ê/imm a pihhi pihhi=tta `if you properly seek my daughter, will I really not give her to you? I will give her to you'. At the same time, there are several contexts where the additive function (`also') cannot be excluded, e.g. KBo 5.3 iii 54-5 apas= an=kan anda auszi ABI DUTUSI=ma=kan imm a GISAB-az arha auszi `(Did not the lady's maid walk by) and he look at her? But the father of My Majesty himself/also looked out of the window'.53 In the following case, the restrictive interpretation seems to be the most likely one: KUB 13.2 iii 4-7 nu=ssan ANA Pç SISKUR2 kittari na=at=si VVDQGX DUDVkandu kuedani=ma ANA Pç SISKUR2 NU.GàL n=at=kan VDU imm a araskandu `(If) a ritual is established for a spring, let them celebrate and visit it, (but) for whatever spring there is no ritual, let them only visit (it)'.54 It is difficult to say with certainty how frequently Hitt. imma displays additive or restrictive pragmatics, but Melchert 1985 has convincingly shown that this is likely to be a secondary development. I believe that the Lydian asseverative enclitic =in and Hittite emphatic particle imma are not only roughly synonymous, but also genetically related.55 Although the word-final development -m > -n occurred in all Anatolian languages, it was still a synchronic rule at least in the Istanuwian dialect of Luwian, as Cuneiform Luwian forms in -am=san (.acc. sg. c. + 3 sg. poss.) seem to indicate.56 My hypothesis requires that it be synchronically present in Lydian at the time of the syncope (i.e. *immV >*im >in). This rule is seemingly contradicted by the existence of Lydian enclitics =(u)m and =im, cognate with the Hittite enclitic conjunction =ma and the Luwian 1 sg. enclitic pronoun =mu/=mi respectively.57 Yet, the place of these two forms was not limited to the end of the enclitic chain; this can be directly demonstrated in the case of

Puhvel 1984: 359 vs. Melchert 1985: 195. Cf. Friedrich-Kammenhuber 1975-, I: 217. 55 The only alternative suggestion known to me is a very old, but occasionally reiterated hypothesis connecting Lyd. =in with Hittite clause-initial particle nu= (Carruba 1959: 37-8 ). This claim does not seem to be supported by any positive arguments and was apparently made faute de mieux. Hitt. nu= occupies an opposite position within the enclitic chain, and does not have any discernible pragmatic functions. 56 Melchert 1994: 270. 57 The enclitic pronoun =(i)m is attested at least in the following texts: 13.1, 56, 73, 75, 76, 77, 95, 104. kim in 10.20 is completely unclear, but, in view of NL P DG in 80.9, it may contain the same clitic =(i)m.
54 53

90

Ilya Yakubovich

=(u)m58 and plausibly surmised in the case of =im based on what we know about the position of enclitic personal pronouns in Anatolian languages.59 One can hypothesize that in these two cases, a retention by analogy with chain-internal clitics bleeds the general rule *-m > -n.60 On the contrary, the particle =in, attested exclusively at the end of a phonological word, underwent the expected sound change. Some syntactic comments are in order. It is true that Hitt. imma can occur almost at any place within a clause emphasizing an immediately preceding syntactic constituent. There is, however, at least one text (KUB 31.71) where Hitt. imma consistently appears right after the clause-initial enclitic chain. The following examples are attested: nu= mu=kan imma (iii 7), nu=wa=mu imma (iii 16), ANSE.KUR.RA.MES= ma=wa imma (iii 25), nu=wa=kan imma (iv 8) and par ZD PX NDQ imma (iv 22). This text, containing a dream narrative of the queen, is likely to reflect colloquial speech, different from that of the official documents emanating from the Hittite chancellery. Melchert (1985: 196) plausibly conjectures that imma here modifies not any particular element of the sentence, but the sentence as a whole, functioning as an asseverative particle. It is easy to see that the position of imma in this text exactly corresponds to that of =in within the Lydian corpus.61 These data, allowing us to see how the enclitisation of Lyd. =in came about, represent an important confirmation of my etymological hypothesis. References
Adiego, Ignacio J. (2000): La inscripciÑn Greco-caria santuario de Sinuri. Kadmos 34, pp. 133-57. Carruba, Onofrio (1959): Studi sul nome, sui preverbi Quaderni del Istituto di glottologia dell'universitÞ di Bo in Analecta Linguistica Anatolica (Studia Mediterranea 1997. pp. 317-347. de los Hecatomnidas en el e sulle particelle in lidio. logna 4, pp 13-43. Reprinted 11). Pavia: Gianni Iuculano,

Cf. such examples as fak=m=at (14.4), ak=m=av (23.7) etc. (Gusmani 1964: 162). It is likely that mru=m=it baal in 14.1 contains not the enclitic conjunction =m, unexpected and unparalleled in the first clause of a Lydian inscription, but rather an enclitic personal pronoun =(i)m. This clause could be translated `The stele ba-ed me', or `(He) ba-ed for me the stele'. 60 In the case of =im, this was also a preventive analogy. The phonetic confusion between the enclitic pronoun =im and the asseverative particle =in would have been certainly undesirable. 61 It is not clear whether the separate writing buk in (23.19) is a scribal error (as per Gusmani 1964: 87), or rather it represents an emphatic usage of in, under which can function as an independent phonetic word.
59 58

Lydian Etymological Notes

91

Friedrich, Johannes and Kammenhuber, Annelies (1975-): Hethitisches WÆrterbuch. Zweite, vÆllig neubearbeitete Auflage. v1.-A. Heidelberg: Carl Winter. Gusmani, Roberto (1964): Lydisches WÆrterbuch mit grammatischer Skizze und Inschriftensammlung. Heidelberg: C. Winter. ___ (1980): Lydisches WÆrterbuch, ErgÄnzungsband, Lfg. 1. Heidelberg: C.Winter. ___ (1982): Lydisches WÆrterbuch, ErgÄnzungsband, Lfg. 2. Heidelberg: C.Winter. Hajnal, Ivo (1995): Der lykische Vocalismus (Methode und Erkenntnisse der vergleichenden Anatolischen Sprachwissenschaft, angewandt an das Vokalsystem einer Kleincorpussprache). Graz: Leykam. ___ (2001): Lydian: Late Hittite or Neo-Luwian? Unpublished handout (presentation made at Berkeley in May 2001). ___ (2003): ,,Jungluwisch" eine Bestandsaufnahme. Licia e Lidia prima dell' ellenizzazione, Roma: Consiglio Nazionale delle Ricerche, 2003. Hawkins, J. David (2000): Corpus of Hieroglyphic Luwian Inscriptions. Volume I. Part I, II: Texts. Berlin-New York: W. de Gruyter. Heubeck, Alfred (1959): Lydiaka. Untersuchungen zu Schrift, Sprache und GÆtternamen der Lyder. Erlangen: UniversitÄtsbund. Kassian, Aleksej S. (2000): Two Middle Hittite Rituals Mentioning fZiplantawija. Moscow: Paleograph. Melchert, H. Craig (1977): Ablative and Instrumental in Hittite, Harward University PhD dissertation. Cambridge, Mass. ___ (1985): Hittite imma and Latin LPPþ. KZ 98, pp. 184-205. ___ (1991): The Lydian Emphasizing and Reflexive Particle -LV. Kadmos 30 (1991), pp. 131-42. ___ (1992): The third Person present in Lydian. IF 97, pp. 31-54. ___ (1993a): Lycian Lexicon. Chapel Hill: Self-published. ___ (1993b): Cuneiform Luvian Lexicon. Chapel Hill: Self-published. ___ (1994): Anatolian Historical Phonology. Amsterdam-Atlanta: Rodopi. ___ (1997): PIE Coronal Stops in Lydian. Festschrift for Eric P. Hamp. D.Q. Adams (ed.), Washington: Institute for the Study of Man, 1997. ___ (2003a): The dialectal position of Lydian and Lycian within Anatolian. Licia e Lidia prima dell' ellenizzazione, Roma: Consiglio Nazionale delle Ricerche, 2003. ___ (2003b): Language. The Luwians (ed. C. Melchert). HdO 68. Leiden: Brill. Oettinger, Norbert (1978): Die Gliederung des anatolischen Sprachgebietes. KZ 92, pp. 74-92. Puhvel, Jaan (1984): Hittite Etymological Dictionary. Vol. 1: Words beginning with A. Vol 2: Words beginning with E and I. Berlin: Mouton Publishers. van den Hout, Theo (2002): Self, Soul and Portrait in Hieroglyphic Luwian. Silva Anatolica (Fs. Popko). Warsaw: Agade. pp. 171-86. Vetter, Emil (1959): Zu den Lydischen Inschriften. Vienna: Rudolf M.Rohrer. Yakubovich, Ilya (2002): Nugae Luvicae. Anatolian Languages (ed. V. Shevoroshkin and P. Sidwell). Canberra: Association for the History of Languages.

University of Chicago

I l y a Y a k ubo v i c h

Near Eastern Languages and Civilisations 1155 East 58th Street Chicago IL 60637 U.S.A. e-mail: sogdiana@uchicago.edu