Download - Prehistoric Contacts between Hittite and Luvian: The … · Prehistoric Contacts between Hittite and Luvian: The Case of Reflexive Pronouns* Ilya Yakubovich University of Chicago

Prehistoric Contacts between Hittite and Luvian:The Case of Reflexive Pronouns*

Ilya YakubovichUniversity of Chicago

The origin of the Hittite reflexive clitic =za represents an unsolvedproblem in Anatolian and Indo-European Studies. The aim of thepresent paper is to demonstrate that this problem receives astraightforward solution once we admit the possibility ofprehistoric structural contacts between Hittite and Luvian. Hittite=za owes its existence to the paradigmatic generalization of theborrowed Luvian 2/3sg. reflexive clitic *=ti/*=di while the Luvianform must be derived from the Indo-Hittite dative clitic *=toi ‘tothee’.

1.0 IntroductionIt is a well-known fact that both the Hittite and the Luvian

languages belong to the Anatolian group of the Indo-Hittite (EarlyIndo-European) phylum.1 It is also commonly accepted that theHittite language of the Empire period underwent partial restructuringunder the influence of the Luvian vernacular, and borrowed a large

*I am grateful to Th. van den Hout (Chicago) and A. Kassian (Moscow), who readthrough the first draft of this article and made valuable remarks, to C. Melchert(Chapel Hill), who supplied me with useful references, and to Th. Wier (Chicago),who helped me to improve the style of my contribution. None of the above is to beblamed for my shortcomings.1 Ivanov (2001) attempts to demonstrate that the Anatolian group of languagesshould be regarded as a linguistic area rather than a genetic unit. It is true that someof the linguistic features traditionally taken as common Anatolian innovations can bereinterpreted as secondary contact-induced changes, and the present article is, in fact,dealing with one such feature. There are, however, several non-trivial commonAnatolian innovations whose contact-induced nature is unlikely to be ever proven,such as the generalization of u \ in *amu \ ‘me’ by analogy with *tu \ ‘thou’, the fusionof the prefix pe- with the verbal root ai ‘to give’, or the enclitization of thedemonstrative pronoun *o-. This does not mean that such innovations could notspread by way of diffusion in a continuum of already differentiated dialects, but thisclaim cannot be tested. The comparative method requires that a given commoninnovation must be treated as genetically inherited unless one can demonstrate itssecondary character.

Prehistoric Contacts between Hittite and Luvian78

number of Luvian lexemes.2 The question about the extent oflinguistic contacts between Hittite and Luvian in the interim periodappears to be more controversial, and is tied up with the discussion ofthe Hittite and Luvian core areas. Bryce (2003) put forward ahypothesis according to which the local homeland of the Luvians waslocated in western Anatolia, while their presence on the CentralAnatolian Plateau can be accounted for by assuming eastwardmigrations in the 2nd millennium BC. If we follow the communisopinio according to which the local homeland of the Hittitesstretched along the upper and middle course of the Halys river(Melchert 2003a:15), Bryce’s suggestion logically implies a gap inLuvo-Hittite contacts spanning the period between the collapse ofAnatolian genetic unity and the migrations that brought Luvians backto Central Anatolia.

The existence of such a gap does not appear to be borne out bythe linguistic evidence. Melchert (2005) has been able to detectdozens of Luvian lexical borrowings attested already in Old Hittitetexts. Luvian personal names appear on the Old Assyrian businesstablets written in Kanes, which was clearly a part of the Hittite corearea, in the 20th-18th centuries BC. Luvian incantations wereoccasionally embedded into Old Hittite magic rituals, and the countryof Luviya is mentioned in the Old Hittite Laws as a land under Hittitejurisdiction, although we do not have any evidence that Hittite kingsever extended their dominion to Western Anatolia before the Empireperiod. These arguments, taken together, provide weighty cumulativeevidence for geographic adjacency between Hittites and Luvians thatextended back to the prehistoric period. One must, however,emphasize the statistical nature of such argumentation, whoseindividual elements remain vulnerable to attacks by devil’sadvocates. Lexical borrowings may occur between geographicallydistant languages, itinerant Luvian merchants could occasionallycross the Anatolian peninsula in search of lucrative trade, while theHittite conquest of Western Anatolia in the Early Old Kingdomperiod may theoretically represent a short-lived episode thathappened not to be reflected in the preserved part of Hittite annalisticliterature.

The borrowing of grammatical morphemes provides evidence ofa completely different kind. The grammatical interference ofdifferent linguistic codes must be conditioned by a bilingualenvironment, which alone may skew the language acquisition processin a large population group. In societies that have no access to mass 2 For the latest appraisal of Luvian lexical influence on New Hittite see van den Hout(2006).

Ilya Yakubovich 79

media and do not enjoy benefits of universal public education, asituation of pervasive language contact could arise only between thegeographically adjacent dialects. Therefore, it is enough to prove asingle episode of grammatical borrowing in order to drawconclusions about the geographic proximity between the two ancientlanguages at the time it took place.

The purpose of this paper is to build up a case for a prehistoricborrowing from Luvian into Hittite that transformed the system ofHittite reflexive pronouns. In Section 2, I will discuss thedevelopment of reflexive pronouns in Hieroglyphic Luvian, whichcan be accounted for without assuming external influence. Section 3will be dedicated to the analysis of the situation in Hittite. I willargue that the Hittite reflexive particle =z(a) resists internalderivation from any of the attested Proto-Anatolian clitics, but can beexplained as an early adaptation of the Luvian pronominal clitic*=ti/*=di . In Section 4, I intend to dwell on a sociolinguisticsituation that could render such an adaptation possible, and toconstruct a relative chronology combining the postulated contactepisode with the relevant internal developments within the history ofHittite.

2.0 Development of Reflexives in LuvianThe Luvian language is known through Bronze Age ritual

incantations recorded on the Hittite cuneiform tablets, as well asthrough later texts written in an original “hieroglyphic” syllabary,mostly in the 1st millennium BC. The Hieroglyphic Luvian textualmaterial is more abundant and heterogeneous than its cuneiformcounterpart and, therefore, provides a convenient departure point forthe discussion of Anatolian reflexives. The system of HieroglyphicLuvian pronominal clitics can be summarized in Table 1.3

The accusative and dative pronominal clitics are bound by non-local antecedents, and occupy syntactic positions that are otherwisereserved for the nouns in the accusative and dative cases respectively.The function of the (so-called) reflexive clitics is much less trivialand will be described in some detail below.4

3 All the forms in the cells below are actually attested in Hieroglyphic Luvian texts.See Morpurgo-Davies (1980:89, n.3) for the attestations of the rare second personpronominal, reflexive and non-reflexive, clitics. The formal peculiarities of Luvianplural clitics will not be addressed in this section. Here and below, I use z instead ofthe IPA [ts] for the transcription of Hittite and Luvian forms.4 The functional distinction between the first and second person reflexives =mi and=di and their (mostly) non-reflexive counterparts =mu and =du has been cogentlyargued in Melchert (1988:41-42), while a similar distinction between 3sg. =di


Contrary to what one might expect, Luvian reflexive clitics donot appear to denote locally bound direct objects. A periphrasticphrase headed by a noun atari- ‘person, self’ is used in this function.5

This construction was not fully grammaticalized, since the nounatari- could appear in the same syntactic position without a binder.

nom. acc. dat. refl.1sg. (none) =mu =mu =mi /=mu2sg. (none) =du > =ru =du > =ru6 =di > =ri3sg. =as (c.)/= ada (n.) =an (c.)/= ada (n.) =du > =ru =di > =ri1pl. (none) =anz(a) =anz(a) =anz(a)2pl. (none) =manz(a) =manz(a) =manz(a)3pl. =ada ==ada =manz(a) =manz(a)Table 1: Hieroglyphic Luvian pronominal clitics.

(1) MARAS ç 4, §3-15 (Hawkins 2000:257)7

|wa/i-tá VIR-ti-i-zi-i (“PES”)pa-ti-zi |ARHA(“MANUS+CULTER”)REL+ra/i-ha-´ |(INFANS)ni-wa/i+ra/i-ni-zi-pa-wa/i-za |(“*474”)u-si-na-si-zi-i |i-zi-i-ha-´ |wa/i-mi-i|á-mi-na (“COR”)á-tara/i-i-na |á-pa-ara/i |BONUS-li-ia-nu-wa/i-haI cut off the men’s feet and made children eunuchs to us, andthereby I exalted myself

(2) KARKAMIÉ A2, §5, (Hawkins 2000:109)*a-wa/i-sa mu-´ ka-tu-wa/i-ia kar<-ka>-mi-si-za(URBS)REGIO DOMINUS-ia “COR”-tara/i-na POST-ni a-táBONUS-li-ia-taBut for me, Katuwa, the Karkemisian country-lord, he exaltedmy person.

(reflexive) and 3sg. =du non-reflexive has been commonly accepted since the dawnof Luvian studies. The semantic analysis given below is, however, my own.5 On the meaning and etymology of atari- see Yakubovich (2002:94-97) and van denHout (2002).6 The rhotacism *d>r was probably a sound change in progress in Iron Age Luvian,whose implementation could depend on a sociolinguistic register. The exact picturehere is difficult to assess since the conservative spelling could have skewed thephonetic distribution of original vs. rhotacized forms in Hieroglyphic writing. Forthe general phenomenon of rhotacism in Hieroglyphic Luvian see Morpurgo-Davies(1982).7 All the Hieroglyphic Luvian texts quoted here and below are cited according to thenames given to them in Hawkins (2000), or a later editio princeps. I have modifiedthe transliteration and translation of several passages.

Ilya Yakubovich 81

In a few cases, reflexive pronouns are used for locally boundindirect objects. The rarity of such forms is probably conditionedsolely by the nature of our corpus, since I know of no other way toconvey the same meaning in Hieroglyphic Luvian.

(3) MARAS ç 4, §9-10 (Hawkins 2000:256)|wa/i-ti-ia-ta |NEG2-a-ha |tá-ti-i-sa |NEG2-a-ha AVUS-ha-sá

|sá-ta |wa/i-mi-ia-tá |EGO ITONITRUS.HALPA-pa-CERVUS2-ti-ia-sa |sá-a-ha

Neither (my) father, nor my grandfather “allowed it tohimself”, but I Halparuntiyas “allowed it to myself.”

In a significantly larger number of cases, a reflexive pronounstands for the locally bound possessor of a verbal complement,usually a direct object. Thus, in example (1) above the reflexive clitic=mi duplicates the possessive pronoun ami- ‘my’ (and not the directobject atari- ‘self’!), while in example (4) below the reflexivereplaces the missing possessive pronoun *abassa/i- ‘his’. This usagemust be considered together with the other cases of possessor raisingin Hieroglyphic Luvian. In the case of nouns and free pronouns, theraised constituent normally appears in the dative, as in (5).8 Thisdistribution allows one to treat reflexive pronouns in (1) and (4) as asubset of dative reflexives.9

(4) KARKAMIÉ A1b, §2-3 (Hawkins 2000:92)*a-wa/i-ti *a-mi-i-sa VIR-ti-i-sa REL-i-ta REL-i-ta || |á-ta5-

ma-za i-zi-i-sa-ta-i |mu-pa-wa/i-ta-´ || |BONUS-sa5+ra/i-ti

CUM-ní i-zi-i-sa-ta-IWherever my husband honors his own name, he shall alsohonor me with respect.

8 This dative clitic, in its turn, may undergo case attraction and acquire the casemarking of the possessum in Hieroglyphic Luvian. The pertinent examples attestedin our corpus appear to be limited to the cases of inalienable possession (Yakubovich2002:193-194). The phenomenon of case attraction remains beyond the scope of thepresent paper and cannot be discussed here in any more detail.9 The reservations of Boley (1993:209-219) about the rarity of dative relexives in theearliest Hittite texts lose their cogency if we consider raised reflexive possessivesmarked by Wackernagel clitics together with this group. Therefore, going somewhatahead, I must state that no descriptive evidence contradicts the semantic derivationof “ethical dative” reflexives from regular dative reflexives in Luvian.


(5) KULULU 5, §10 (Hawkins 2000:486)wa/i-tu-u [á-ta4/5-ma-za ARHA] DELERE-nú-tu-uLet them destroy his name

Another large group of examples comprises the cases whereLuvian reflexive clitics appear to have no correspondence in thesentence logical form. So far as one can judge, the sole purpose oftheir usage is stylistic, perhaps stressing the agency of the subject.Thus, in example (6), the king intones that setting up the gods’statues represents his own pious accomplishment, and therefore usesthe reflexive pronoun =m(i), whereas the same pronoun is omitted ina very similar context in (7), presumably since no special emphasis isintended. This syntactic pattern can be taken as a subset of a moregeneral phenomenon called “ethical dative.” It is used with referenceto dative pronouns that have no place in the clause argumentstructure, but indicate a party which is presumed to be emotionallyinvolved in a situation.10 The usage of reflexive “ethical datives”brings about the identification of this party with the clause subject.

The development of dative pronouns with a reflexive functioninto reflexive “ethical datives” is closely paralleled by certainAmerican English dialects (mostly spoken in southern Appalachia),where oblique personal pronouns are used with ditransitive verbsdenoting a self-directed transfer to an agent (e.g. I bought me a gun)but also with transitive verbs that do not require a beneficiary (e.g. Iate me a possum). It is remarkable that the reflexive/emphatic usageof personal pronouns never extends to direct objects in these dialects.

(6) KARKAMIÉ A11b, §17 (Hawkins 2000:101)*a-wa/i-ma-tá |za||-ti-i |(“PODIUM”)hu-ma-ti |(SOLIUM)i-sà-nú-wa/i-haI myself established them (i.e. the gods) on this podium

10 Compare the following Latin example: Hic mihi quisquam misericordiam nominat“Someone here is mentioning compassion.” The ethical dative mihi, lit. ‘to me’,reflects here the speaker’s indignation towards mentioning compassion. In AtticGreek, 2sg. ethical dative =toi was grammaticalized as a particle used in directspeeches with the general meaning “let me tell you, mark you, look you” used inthose cases where special attention of the person addressed is desired.

Ilya Yakubovich 83

(7) KARKAMIÉ A23, §10 (Hawkins 2000:119)*a-wa/i-ta (DEUS)ku+AVIS-pa-na |kar-ka||-mi-si-za-na(URBS) MAGNUS.DOMINA-sa5+ra/i-na |POST-ni

|SOLIUM-nu-wa/i-haI re-established Kubaba, queen of Karkamié

The emphatic usage of Hieroglyphic Luvian reflexivesunderwent a secondary grammaticalization in two distinct groups ofcases. The first group comprises nominal sentences with first/secondperson subjects. Since the present tense verbal copula normally doesnot find an overt expression in Luvian, and since the use of freesubject pronouns is optional there in all contexts, the “ethical dative”reflexives emerge as the only obligatory way of expressing thesubject of a nominal predicate.11 Thus we would not know the subjectof the first clause in (9) if not for the reflexive pronoun =ri. Thisfunctional device would be, however, redundant in the case of thirdperson subjects in nominal clauses, since these are always expressedeither by nouns, or by subject enclitic pronouns, as in (10).12

(8) KARKAMIÉ A1b, §1 (Hawkins 2000:92)

EGO-mi-i IBONUS-ti-sa Isu-hi-si-i REGIO-ní DOMINUS-ia-i-sa |BONUS-mi-sa || FEMINA-ti-i-saI am BONUS-tis the Country-lord Suhis’ dear wife.

(9) ASSUR f+g, §11-12 (Hawkins 2000:536)

wa/i-ri+i-i |ku-ma-na |ha-tu-ra+a wa/i-za |ni-i-´ |ma-nu-ha

|ARHA-´ |(“*69”)sa-si-i

When (thou) are to write, do not abandon us by any means.

(10) KARATEPE 1, §60 (Hawkins 2000:56)

ní-pa-wa/i-sa ICAPUT-ti-sáOr (if) he (is) a prince.

11 For the account of Luvian connective clitics in nominal sentences as etymologicalethical datives, see Eichner (1974:68). The reservations of Melchert (1988:42) areunfounded in view of (6) and similar examples.12 This is a corollary to a more general principle, known as the Watkins-Garrett rule,according to which the third person subject enclitic pronouns cannot be dropped inclauses with unaccusative predicates in all the Anatolian languages. For a detaileddiscussion of this rule, in its application to Hittite, see Garrett (1996).


The second group is limited to certain telic verbs, which arehomophonous with verbs denoting states. The only verb belongingto this group which can be discussed without resorting tocomparative evidence is *asa- (normally written with a logogramSOLIUM). Its meaning with and without the reflexive particle is ‘tosit down’ and ‘to sit, live’ respectively.13 One can hypothesize thatthe telic marker in (11) and similar cases represents a furtherdevelopment of the “ethical dative” function of the reflexivepronouns, stressing the agency of the subjects, because telic verbstend to be closer to prototypical transitives than their stativecounterparts (Hopper and Thomson 1980:252). The English clause“He thought to himself,” which can be used with reference toindividual acts of thinking, but not to one’s permanent convictions,provides an approximate parallel to the usage of Luvian reflexives inthis function.

(11) IZGIN 1, §2 (Hawkins 2000:315)wa/i-mi-ta-´ mi-i-´ |tá-ti THRONUS-tara/i-ti REL SOLIUM-haWhen I sat down on my father’s throne...

(12) KARATEPE 1, §24 (Hawkins 2000:51)|á-TANA-wa/i-sa-wa/i(URBS) || |REL-ti |(BONUS)wa/i+ra/i-ia-ma-la |SOLIUM-MI-i... so that Adanawa might dwell (lit. sit) peacefully.

Based on the discussion above, the reconstructed development ofreflexive pronouns in Luvian can be summarized in the followingtree:

Raised Reflexive Possessives

Indirect Object “Ethical dative” 1/2 Person Markers

Reflexives Reflexives in Nominal Sentences

Markers of Verbal Telicity

Thus, although the pronouns functioning as indirect objectreflexives represent but a small proportion of all the clitics =mi/=di

13 Hawkins (1992:262) convincingly argued that Luv. muwa- accompanied byreflexive pronouns represents a precise functional equivalent of Hitt. =za ... tarh #- ‘toovercome, conquer’. Unfortunately, the meaning of Luv. muwa- without reflexivepronouns cannot be clearly determined out of context, even though the suggestionthat it is close to that of Hitt. tarh #- ‘to be powerful, dominate’ remains likely.

Ilya Yakubovich 85

attested in the Luvian corpus, they can be taken as a logical startingpoint for further semantic developments within this group, whichvindicates their traditional designation as reflexive clitics. The nextstep is to compare the results of our semantic reconstruction with thesystem of pronominal clitics traditionally reconstructed for Proto-Indo-European (see Table 2 below).14

The reflexive stem sw-, which definitely existed in Late Indo-European, does not appear to have any clear counterparts inAnatolian. Whether this represents a common Anatolian innovation,or special reflexive clitics emerged at a stage of Indo-European thatpostdates the separation of Anatolian, one must assume that Proto-Anatolian had different means to express the co-reference betweenthe subject and the other arguments within a clause. Proto-Luvianindirect object reflexives *=mê (1sg.) and *=tê/=dê (2/3sg.) representa straightforward phonetic outcome of the Indo-European clitics*=moi ‘to me’ and *=toi ‘to thee’ respectively, while the subsequentcontraction of long vowels in these forms probably reflects acommon tendency toward eliminating long vowels in Indo-Europeanclitics.15 Therefore, one can advance a hypothesis that the ancestordialect of Luvian extended the usage of dative pronominal clitics tothose instances when they were bound by the local antecedent. Thisdevelopment can be summarized in Table 3:

acc. dat. acc. refl. dat. refl.1sg. *=me *=moi *=me (or *=swe?) *=moi (or *=swoi?)2sg. *=te *=toi *=te (or *=swe?) *=toi (or *=swoi?)3sg. *=se *=soi *=swe *=swoiTable 2: Proto-Indo-European clitics.

14 Cf. Meier-Brügger (2000:209) and Fortson (2004:129-130). I cannot agree withattempts to take *=se and the reflexive *=swe (resp. *=soi and *=swoi) as twovariants of one and the same reflexive pronoun. The comparison between Hittite =seand Indo-Iranian =se ‘to him’ assures the reconstruction of *=soi as a dative non-reflexive (anaphoric) clitic. The fact that the reflexes of *soi and *se acquired areflexive meaning in several branches of Anatolian and Indo-European does not needto reflect anything more than the recurrent typological tendency to form newreflexive pronouns based on the original personal pronouns.15 Thus the Indo-European accusative clitics *=me ‘me’ and *=te ‘thee’ are to beprobably taken as “irregular” contracted variants of the stressed pronouns me \:, te \:.Compare also Greek particles dÆ vs. = de and = µan/ = µhn vs. = µen. Note thatCuneiform Luvian plene spellings militate against the assumption that all thesecondary long vowels were always contracted in this language in unstressedpositions (Melchert 1994:278)


nom. acc. dat. (dat.) refl.1sg. (none) =mu =mi → =mi2sg. (none) =tu/=du =ti/=di → =ti/=di3sg. =as (c.); =ada (n.) =an (c.); =ada (n.) =si → =siTable 3: Development of Luvian clitics, Stage 1.

Melchert (2003b:204, n.31) suggested the next step when hewrote: “The introduction of the enclitic forms -mu, -tu, and -tu asdatives [in Luvian– I.Y.] would have severely restricted the use ofthose with i-vocalism, leading to their synchronic analysis asreflexives”.16 The Luvian forms =mu (1sg.) and =du (2sg.) must betaken as the reflexes of the Indo-Hittite accusative clitics *=me and*=te, whose vocalism was modified by analogy with independentpronouns amu\ ‘I’ and tu \ ‘thou’. The merger of dative and accusativefirst/second person pronominal clitics in Luvian finds a functionalparallel in the system of the plural nominal declension, where theaccusative ending *-ans > -anz(a) spread to the dative in prehistorictimes. This clitic merger failed, however, to be implemented in thethird person because of an earlier enclitization of the demonstrativepronoun *o- in the nominative and accusative cases, resulting in thecreation of a new anaphoric clitic (nom.sg.m. =as, acc.sg.m. =anetc).17 Since this pronoun was marked for case in the singular in asimilar way to singular nouns and adjectives, it was immune tofurther case mergers within the clitic system.

nom. acc. dat. (dat.) refl.1sg. (none) =mu → =mu =mi2sg. (none) =tu/=du → =tu/ =du =ti/=di3sg. =as (c.); =ada (n.) =an (c.); =ada (n.) =si =siTable 4: Development of Luvian clitics, Stage 2.

16 Melchert cites Luvian clitics in transliteration, rather than phonetic transcription,within this quote. We know that this form underwent lenition and was pronounced as/du/ (vel sim.) in Common Luvian, both due to the fact that the consonant is neverdoubled in Cuneiform Luvian orthography and because it underwent rhotacism inHieroglyphic Luvian. The phonetically distributed allomorphs =tu and =du (resp. =tiand =di) have to be assumed for the earlier stages of the language, on which seeSection 4 below. On the phenomenon of Luvian lenition, see Morpurgo-Davies(1982).17 This new enclitic stem was restricted to the nominative and accusative in all theAnatolian languages except for Lycian and Milyan (we lack information aboutCarian). In the Lycian and Milyan languages, the innovative dative form =i ‘tohim/her/it’ and locative adverb =ije ‘therein, thereon’ arose secondarily (cf. Melchert2004:26-28, 116).

Ilya Yakubovich 87

The final change that one needs to assume in order to account forthe common Luvian stage is the merger between the second and thirdperson singular dative clitics, both reflexive and non-reflexive. To besure, this change is difficult to justify in functional terms, and yet onehas no choice but to accept it in view of its systematic character.Besides the parallelism between the functional extension of =di and=du, one has to mention the existence of the Hieroglyphic Luvianplural clitic =manz(a) ‘to you, to them; to yourselves, to themselves’.Apparently, all the Luvian dative clitics came to be characterized bythe opposition first/non-first person. The accusative clitics wereexempt from this development because the anaphoric forms=an/=ada ‘him/her/it’ and =ada ‘them’ had a paradigmatic supportfrom their nominative counterparts. I will have more to say about thisprocess when we discuss the development of Hittite clitics.

nom. acc. dat. (dat.) refl.1sg. (none) =mu =mu =mi2sg. (none) =du3sg. =as (c.); =ada (n.) =an (c.); =ada (n.)

=tu/=du ↓=tu/=du

=ti/=di ↓=ti/=di

Table 5: Development of Luvian clitics, Stage 3.

The last change to be mentioned occurred already in thehistorical period. The late Hieroglyphic Luvian inscriptions bearwitness to the spread of =mu , at the expense of =mi, to the firstperson reflexives, as the following example shows:

(13) KARATEPE 1, §63-4 (Hawkins 2000:57)ARHA-wa/i-ta “*69”(-)i-ti-wa/i || (LITUUS)á-za-ti-wa/i-tà-sáá-ta5-ma-za PORTA-la-na-ri+i zi-na wa/i-mu-ta || á-ma-za á-

ta4-ma-za a-ta tu-pi-wa/i

I will delete Azatiwatas’s name from this gate, and will incisemy own name.

3.0 Development of Reflexives in HittiteThe set of Old Hittite pronominal clitics can be summarized in

Table 6.


nom. acc. dat. refl.1sg. (none) =mu =mu =z(a)18

2sg. (none) =tu/=ta19 =tu/=ta =z(a)3sg. =as/=ad =an/=ad =se =z(a)1pl. (none) =nas =nas =z(a)2pl. (none) =smas =smas =z(a)3pl. =e/=e =us/=e =smas =z(a)Table 6: Old Hittite pronominal clitics.

A number of formal features of pronominal clitics are morearchaic in Hittite than they are in Luvian. Hittite =ta ‘(to) thee’ maybe directly derived from the Proto-Anatolian accusative clitic *=te‘thee’.20 Hitt. =ad ‘it’ directly reflects Anat. *=od, while Luv. =ada‘it’ owes its final vowel to the analogical influence of other neuterpronominal forms ending in -a. Hitt. =smas ‘(to) you, them’ directlyreflect Anat. *=smos, while HLuv. =manz(a) ‘(to) you, them’ betraysthe influence of the Common Luvian nominal dat./acc.pl. ending*=anz(a). The discrepancy between Hitt. =nas and Luv. =anz(a) ‘us’admits of different explanations, but my personal preference is toposit a Proto-Luvian form *=nanz(a), which later underwentreanalysis in a position after word-forms ending in -n. It is likely thatthe tonic stem anz- ‘we /us’ played a role in this reanalysis.21

Once we account for these formal distinctions, the initialevolution of Hittite and Luvian clitics is amenable to a uniforminterpretation. My reconstruction of structural changes in Hittiteleading to the partial merger of dative and accusative clitic pronounsis identical to the Luvian “stages” 1 and 2. Assuming the traditional

18 There is no doubt that that the phonetic realization of the Hittite reflexive particlewas [ts], as per Kühne (1988) and Yoshida (2001). The transliteration =z(a) isretained here solely in order to make this paper more accessible to a broad range ofHittitologists.19 The synchronic distribution between the allomorphs =ta and =tu in Hittite ismorphologically conditioned. The variant =tu occurs before the clitics =z(a) and=san, while =ta occurs before =kan and in chain-final position. On the likelydiachronic interpretation of this distribution, see section 4 below.20 For the change *e>a in Hittite posttonic open syllables see Melchert (1994:137-138).21 Alternatively, one can argue that =anz(a) directly reflects *=n …s, the zero-gradeform corresponding to *=nos. This reconstruction would obviously result in a morestraightforward derivation of HLuv. =anz(a), but would leave us with a questionabout the relationship between the two ablaut grades at the Proto-Anatolian level.Since Proto-Anatolian plural clitic pronouns did not have a case paradigm, wheredifferent grades could alternate, one would have no choice but to assume that theLuvian 1pl. clitic =anz(a) represents a shortened form of the respective tonicpronoun anz- that has ousted the proto-Anatolian clitic *=nos.

Ilya Yakubovich 89

Indo-European proto-form *=moi and *=soi, this reconstruction canbe seen in Tables 7 and 8.22

The isomorphism between Hittite and Luvian “stages” 1 and 2allows us to view both “stages” as common Anatolian innovations.Nothing precludes a relative chronology according to which all thesound changes and analogies that are responsible for the formaldistinctions between Luvian and Hittite postdate the implementationof these functional changes. Therefore, one can assume that the twotables contain the provisional reconstruction of the actual Proto-Anatolian forms. The internal reconstruction of Luvian given insection 2 should be modified accordingly.

Nom. acc. dat. (dat.) refl.1sg. (none) =mu =me\ → = me\2sg. (none) =te/=de =te\ /=de\ → =te\ /= de\3sg. =as (c.); =ad (n.) =an (c.); =ad (n.) =se\ → =se\Table 7: Development of Anatolian clitics, Stage 1.

nom. acc. Dat. (dat.) refl.1sg. (none) =mu → =mu =me\2sg. (none) =te/=de → =tu/ =du =te\ /= de \3sg. =as (c.); =ad (n.) =an (c.); =ad (n.) =se\ =se\Table 8: Development of Anatolian clitics, Stage 2.

Once we move to Luvian stage 3, we can observe that ananalogous change was not consistently implemented in Hittite.Hittite, alone of all the Anatolian languages, preserves the dativeanaphoric clitic =se < *=soi ‘to him’, which indeed serves as a basisfor reconstructing this element in Proto-Anatolian. We have nochoice but to assume that a merger between the second and thirdperson singular clitics in individual Anatolian languages postdatesthe common Anatolian stage. Luvian and Hittite both participated inthis process, but the scope of their involvement in it was different ineach of the three cases. Hittite generalized the reflexive =z(a), butfailed to expand the dative clitic =tu ‘to thee’ to the third person.Luvian extended the usage of both the reflexive pronoun *=ti/*=diand the personal pronoun *=tu/*=du.

22 Eichner (1974:31-32) proposed the reconstruction *=mei ‘to me’ and *=tei ‘tothee’ for Proto-Anatolian, trying to account for the vocalism of Hittite possessiveclitics =mis ‘my’ and =tis ‘thy’. Eichner’s claim that Hittite possessive clitics aroseout of the historical dative clitics has been meanwhile undermined by the fact thatthe extension of the old accusative clitics to the dative represents a commonAnatolian phenomenon, whereas the cliticization of possessive pronouns is aspecifically Hittite development.


The extension of the second person pronominal forms to the thirdperson is not a common typological phenomenon, and it becomeseven more puzzling once we assume that it happened several times indifferent branches of Anatolian. One has to look for peculiarities ofthe common Anatolian pronominal system that could trigger thisprocess on such a large scale. The only element that qualifies as asuitable starting point for the analogical spread of this merger is thecommon Anatolian plural clitic *=smos ‘(to) you, to them’. To besure, the common Anatolian status of this clitic was challenged in thepast, but the close functional parallelism between Hitt. =smas ‘(to)you, to them’ and HLuv. =manz(a) ‘(to) you, to them’ pleads in favorof this solution.23 One may in fact try to project the existence of sucha form back into Indo-Hittite, since the oblique plural clitics TA -mand TB -me represent a perfect phonological match to Anat. *=smosand can be used with reference to all the three persons (‘(to) us/(to)you /(to) them’).24 This argument, however, cannot be pursued herein any more detail because the linguistic reconstruction beyond theAnatolian level does not constitute the topic of this paper.

Because Proto-Anatolian plural clitics could function both aspersonal pronouns and reflexives, the analogy triggered by theseforms could likewise target both reflexive and non-reflexive clitic

23 For a detailed discussion of homonymous second and third person pronominalforms in Indo-European see Katz (1998:234-253). The author comes to theconclusion that the synchronic identity between such forms represents a historicalaccident in all the instances treated in his dissertation, and specifically suggests twodifferent etymologies for Hitt. =smas ‘(to) you’ and Hitt. =smas ‘to them’, the latterbeing cognate with HLuv. =manza ‘to them’. He is not able, however, to provide aphonetic account for HLuv. =manz(a) ‘to you’ and is therefore forced to assume thatwe are dealing with an extension of a third person form to the second person withinLuvian (Katz 1998:241). He suggests further as an alternative that the second personforms =smas and =manz(a) may be in fact cognate, in which case they “bothrepresent a true, and not purely Luvian, extension from the third to the secondperson” (Katz 1998:241-242). I believe that this last alternative is to be preferred,because it alone provides us with a foundation for explaining further changes withinthe system of Anatolian singular clitics.24 Katz (1998:156, 247) suggests that TA =m/B =me referring to the third person dogo back to the clitic *=smos, whereas its first and second person counterparts goback to the aphaeretized forms of stressed pronouns *n …sme ‘us’ and *usme ‘you’respectively. This reconstruction entails formal problems concerned with thevocalism of Tocharian B, and in addition there is no obvious advantage in derivingTocharian oblique clitics from Late Indo-European stressed object pronouns, anddisconnecting them from oblique clitics in Anatolian. In my view, one canreconstruct the system of Indo-Hittite dative/accusative plural clitics as =nos (1pl.)/=smos (2pl.)/=smos (3pl.), and assume that Tocharian generalized the second/thirdperson form to all persons. The Late Indo-European 2pl. clitic *=wos ‘you’ can beexplained as an innovation based on stressed forms and obtained through aproportional analogy of the type no\s/=nos ~ wo\s/X.

Ilya Yakubovich 91

pronouns. In practice, it targeted both, but in different languages andat different times. The analogy that may have occurred early in thehistory of “non-Hittite” Anatolian can be described by the proportion*=(s)mas/*=(s)mas ~ *=tu/ X. In prose, the formal identity betweenthe clitics meaning ‘to you’ and ‘to them’ triggered the functionalextension of *=tu ‘to thee’ to the third person, at the expense of theoriginal clitic *=se\ ‘to him’. This innovation also occurred in Palaicand can probably be reconstructed for the history of Lydian. Anotherapplication of the same type of analogy resulted in the expansion of2sg. reflexive =di to 3sg. in Luvian. Neither Palaic nor Lydianparticipated in this process since the reflexive clitic *=si < *=soi ispreserved in both languages (Melchert 1991:142).

The only clitic the spread of which at “stage 3” remains formallyunaccounted for is the Hittite reflexive =z(a), phonetically /ts/, whichis routinely reconstructed as *=ti (Kimball 1999:191 with ref.). Thereis simply no form in the Proto-Anatolian reflexive paradigm thatcould develop into Hitt. *=ti >=z(a) according to the standard soundlaws. And yet, the connection between the Hitt. =z(a) and the Luvianpronominal clitic =di belongs to the common stock of knowledge ofAnatolianists at least since Laroche (1959). In view of the compellingevidence for the common origin of Luvian =di and the Indo-European second person dative clitic, presented above and supportedby parallels from the other Anatolian languages, only two solutionsremain possible. One must either reject the connection between Hitt.=z(a) and Luv. =di altogether, or assume that Hitt. =z(a) was derivedfrom Indo-Hittite *=toi or Anatolian *=te\ with the help of changesother than sound laws.

The success of the first approach would be contingent uponfinding alternative plausible cognates of Hitt. =z(a) in Anatolian orIndo-European. At the present time, however, I am not aware of anyalternative etymological suggestions backed by the adequatesemantic analysis of =z(a). The claim of Carruba (1964:429-430) that=z(a) should be connected with the Indo-European demonstrativestem *to- stumbles upon the lack of any Anatolian inflected formsthat can be related to this Indo-European stem, and in any event is nothelpful for tracing the inflectional prehistory of =z(a).25 Josephson’sassertion that Hittite =z(a) should be identical with 3rd sg. dat. =sa of

25 Carruba (1969:45-46) attempted to corroborate his earlier suggestion by adducingthe Luvian clitic =du ‘to him’, allegedly derived from Indo-European *=to(i) (velsim.) built on the demonstrative stem *to-. This suggestion is phonetically unlikelygiven the change of non-final *oi to ê in Luvian, as per Melchert (1994:279). Note inaddition that Carruba has not presented any arguments that would motivate thereplacement of the inherited *=soi with *=toi in the prehistory of Luvian.


the possessive stem si/a- ‘his’ (Houwink ten Cate and Josephson1967:137) does not take into consideration the actual pronunciationof =z(a) as /ts/. The scholars who studied =z(a) in the subsequentyears tended to abstain from suggesting extra-Anatolian cognates ofthis Hittite particle. One can mention Neu’s confession: “DieGemeinanatolische Partikel *-ti, auf die heth. -z und ‘mediopassives’-ti zurückgehen, vermag ich bisher noch nicht an Indogermanischesanzuschliessen” (Neu 1968:145), and Oettinger’s verdict: “ZurEtymologie lässt sich nur sagen dass -z auf uranatolisch *-ti vonoffenbar sehr ähnlicher Bedeutung zurückgeht. Dessen Herkunft istunklar” (Oettinger 1997:412). The absence of plausible alternativesis not due to the lack of interest for the origin of =z(a) on the part ofAnatolian scholars, but rather due to objective difficulties withfinding suitable Indo-European reflexes of Indo-Hittite **=ti, aputative direct ancestor of Hitt. =z(a).

The second approach is exemplified by the recent work ofElisabeth Rieken. Rieken (2004:183) was the first scholar toexplicitly connect both Hittite =z(a) and Luvian =di with the LateIndo-European =toi ‘to thee’.26 Rieken was well aware of the factthat the direct reflex of *=toi would be Old Hittite **=te, just as*=soi ‘to him’ developed into =se in Old Hittite (Kimball 1999:214).Therefore she proposed that *=toi had changed to *=ti already inProto-Anatolian, and attributed this change to the analogicalinfluence of the nominal dative-locative ending *-i. This suggestioncan be viewed as an extension of an earlier proposal by Melchert(1994:218), who explained the vocalism of Palaic =ti ‘to thee’ as aresult of analogical leveling triggered by the dative-locative nominalending.

Rieken’s approach has much to recommend itself from thefunctional point of view. Now one can hypothesize that *=ti > =z(a)originally functioned in Hittite as the second singular reflexive, whileits spread to the third singular represented the first stage in itsproliferation. This could be yet another proportional analogymotivated by the homonymy of the second and third plural reflexiveclitics (=smas / =smas ~ =z(a) / X), and parallel to the spread of =diin the ancestor dialect of Luvian. Later, however, the reflexive =z(a)was generalized from the third singular to all the other persons andnumbers, and ousted even those plural reflexive clitics that wereresponsible for its original spread. The tide reversed in the Middle

26 Rieken gave credit for this suggestion to Melchert (1994:218), but the claims ofthe two scholars were in fact substantially different. Melchert limited his analysis tothe etymology of Palaic =ti ‘to thee’ (cf. below), which he does not connect withHitt. =z(a) or Luv. =di anywhere in the Anatolian Historical Phonology.

Ilya Yakubovich 93

Hittite period, when the plural reflexive clitics =nas and =smas werereintroduced from the non-reflexive paradigm under the direct impactof Luvian.

As for the formal side of Rieken’s proposal, it appears to besignificantly more problematic. Unlike Melchert’s suggestionregarding the origin of Palaic =ti, which can be corroborated by theidentical vocalism of the reflexive =si in Palaic, a would-be analogyleading to the development of =ti in Proto-Anatolian loses itscredibility in the face of the preservation of =se ‘to him’ in OldHittite. The paradigmatic leveling could not have been selectivelyapplied only to some members of the rhyming group of dative clitics(*=me \ /*=te\/*=se\). It is even less credible that it was applied to *=te \after this pronoun was specialized as a reflexive in Proto-Anatolian(stage 2), but failed to be applied to the true dative anaphor *=se \. Asa matter of fact, we can observe the effects of an analogy thatchanges =se into =si in the history of Hittite, but this processoccurred by the end of the Old Hittite period, and obviously hasnothing to do with the putative Proto-Anatolian change discussed inthis paragraph (cf. Melchert 1994:218).

If the development *=te \ > *=ti cannot be explained eitherthrough sound law, or by analogy, one may try the third optionoffered to us by the comparative method. The form *=ti could beborrowed from an Anatolian dialect where the change *toi > *ti wasphonetically regular or analogically motivated. All the Anatolianlanguages other than Hittite probably qualify as possible sources ofborrowing according to this formal criterion, but Luvian also does iton sociolinguistic grounds. On the one hand, we have abundantevidence for the structural influence of Luvian on Middle and NewHittite, which includes the specific case of reflexives mentionedearlier in this section. On the other hand, we have enough data toposit intensive lexical contacts between the two languages in theprehistoric period. Neither of the two kinds of evidence is availablein the case of Palaic or Lydian, so they can be eliminated as sourcesof borrowing by Occam’s razor.

The most likely scenario for the development of reflexives withinHittite is represented below:

nom. Acc. Dat. (dat.) refl.1sg. (none) =mu =me\2 sg. (none) =te/ =de3 sg. =as (c.); =ad (n.) =an (c.); =ad (n.) =se\

=te\/=de\ ↓=te\/=de\

Table 9: Functional convergence with Luvian, Stage 3.


nom. Acc. Dat. (dat.) refl.1 sg. (none) =mu =mi2 sg. (none) =te/ =de3 sg. =as (c.); =ad

(n.)=an (c.); =ad (n.) =se\

=ti/=di=ti/=di

Table 10: Formal convergence with Luvian, Stage 4.

nom. Acc. Dat. (dat.) refl.1 sg. (none) =mu2 sg. (none) =te/ =de3 sg. =as (c.); =ad (n.) =an (c.); =ad (n.) =se\

=ti/=di=ti/=di ↑=ti/=di

Table 11: Generalization of the reflexive particle, Stage 5.

The rational for ordering the formal convergence with Luvianafter the implementation of “stage 3” is the necessity to eliminate thereflexive pronoun *=se \ in Hittite prior to the borrowing of the -ivocalism. If we assume the opposite order, we could expect that thecontact-induced change = se \ > =si spread to the anaphoric = se \ ‘tohim’, which is flatly contradicted by the Old Hittite evidence. Bycontrast, the relative chronology given above implies a stage when allthe Hittite singular reflexive clitics were formally distinct from theirnon-reflexive counterparts. Under such conditions, the contact-drivenformal changes in the reflexive paradigm would not need to have anyrepercussions for the rest of the system.

At the same time, it is very likely that the implementation of“stage 3” in Hittite was motivated not only by the proportionalanalogy (=smas / =smas ~ *=te \ / X), but also by the similar changein the territorially adjacent Luvian dialects. As was alreadymentioned above, the spread of second person clitics to the thirdperson, whatever its motivation, is typologically unusual, and sowhen one observes a cluster of such changes in a group of closelyrelated dialects, an areal explanation should be seriously considered.The Luvian language, where the innovations of “stage 3” wereimplemented most consistently, could generate a “linguistic wave”that would spread across the continuum of already differentiatedAnatolian dialects triggering similar functional changes, but on asmaller scale. This scenario would, however, remain a merepossibility, if we did not have independent evidence for the formalconvergence between Hittite and Luvian reflexive clitics. Now thatsuch evidence is available, the interpretation of “stage 3” in arealterms must be preferred to assuming a set of typologically similar,but unrelated changes.

Ilya Yakubovich 95

Grammatical borrowing: when and why. I am keenly aware ofthe fact that the claim about an isolated contact-induced formalchange in the pronominal system of any language is strong enough tolook suspicious. Critics may wonder if the argumentation byexclusion provides a sufficient justification for assuming foreigninfluence of what is frequently regarded as the most stable part of thecore vocabulary.27 Fortunately for my account, there is anindependent confirmation that this part of the vocabulary was not sostable in the prehistory of Hittite as purists might think. It comesfrom the analysis of the clitic complex =(t)tu=za attested from theMiddle Hittite period onwards.28

(14) HKM 71 Obv 9-10, MH/MS (Alp 1991:254)namma=tu=za UGULA NIMGIR.ERÍN.MEÉ halzissanziThey call thee “overseer of military heralds” (a high-rankingmilitary title).

(15) HKM 56 Obv 11-12, MH/MS (Alp 1991:224)nu=ttu=za=kan ka\sa sanzaI am now angry with thee.

One can see from the example above that =tu represents anallomorph of the Hittite clitic =ta ‘(to) thee’. It could be used both fordirect and indirect objects, as in (14) and (15) respectively. Thefactor conditioning the synchronic choice between the allomorphs=tu and =ta is the presence vs. absence of the particles =z(a) or =sanimmediately to the right (the combination =(t)tu=za occursapproximately five times more frequently than =(t)tu=(s)san does).It is clear that the origin of such a distribution cannot be phonetic,nor can one find a plausible source for the analogical extension of=tu , since a clitic of such a shape is not otherwise attested inHittite.29 The assumption of a Luvian borrowing into Middle Hittite

27 The general claim that pronouns are never or extremely rarely borrowed wasshown to be false long ago (see Thomason 2001:83-84 for a set of examples). Itremains nevertheless a descriptive fact that Indo-European personal pronouns areresistant to borrowing from unrelated or distantly related languages.28 I am grateful to Y. Gorbachev (Harvard), who was the first to indicate theimportance of the allomorph =tu for my argument. Note that the sequence =tu=za isattested in a late copy of the Old Hittite Telibinu proclamation (KUB 11.5 Rev. 2).29 F. Josephson suggested that Hitt. =tu(=za) represented a relic accusative clitic,whose function was otherwise taken over by the original dative clitic =ta (Houwinkten Cate and Josephson 1967:136-7). This hypothesis is untenable in view of thedirect correspondence between Hitt. =ta and the Indo-European accusative clitic*=te.


would be equally gratuitous, unless one can explain why theborrowed form is used exclusively in a position before two pariclesinherited from Old Hittite.

Nothing, however, contradicts the assumption that the variant=tu= had existed already in Old Hittite, exactly with the samedistribution that is attested in Middle and New Hittite. We simplylack passages featuring the second person clitic in front of=z(a)/=san in the Old Script texts. This is probably to be explainedby the absence of letters and the paucity of treaties written on theearliest Hittite tablets available to us, since these two textual genresdisplay the largest concentration of second person forms in a laterperiod, and dominate among the sources of attestation of =(t)tu=za.In the absence of positive evidence, it is more logical to assume thatMiddle Hittite has inherited an old distribution, than to posit aninnovation for which we cannot account.

Once we relegate the problem to the prehistoric period, itreceives a likely explanation in terms of language contact. Hittiteshad apparently adopted not only individual Luvian reflexive cliticsbut also the whole sequence *=tu=ti in lieu of the inherited *=te=te \.A factor that no doubt facilitated this contact-induced change was theclose genetic similarity between Hittite and Luvian and, morespecifically, the phonetic resemblance between the relevantpronominal forms. The same factor had to be at play in the borrowingof the Old Norse third-person plural pronouns into the Anglo-Saxondialects (cf. Old Norse πeir / πeim / πeira ‘they / them / their’ withtheir Old English equivalents hêe / him / hiera). Later the clitic =tucame to be perceived as an allomorph of =ta, and its usage wasextended to the position before =san, perhaps for the reason of partialphonetic similarity between =z(a) and =san.

The parallel between the prehistoric contacts between Hittite andLuvian on the one hand, and the historical contacts between OldEnglish and Old Norse on the other hand, can be carried beyond thisparticular episode. Dawson (2003) has convincingly argued that thedefining factor in the lasting impact of Norse on the structure andbasic lexicon of the English language was not so much the number ofDanes that settled in England, or the long period of contacts, as thegenetic and typological similarity between the two languages thatwould enable the speaker of one language to understand the otherlanguage without learning to speak it. She appropriately refers to theconvergence triggered by this type of contact as koineization, andstresses the fact that in such conditions it is frequently difficult todraw a line between borrowings sensu stricto and the contact-drivenmodification of linguistic items that had been already available in the

Ilya Yakubovich 97

target language. This approach helps to explain why the Norsepresence in medieval France, or the conquest of England by theassimilated French-speaking Normans failed to result in acomparable structural convergence: the mutually incomprehensiblecharacter of Romance and Germanic languages precluded theformation of a new koine.30

If we assume that Proto-Hittite and Proto-Luvian populationgroups lived side by side in sufficient numbers and intensivelycommunicated each in their respective language at some point shortlyafter their genetic separation, it is easy to envisage the situation whena new generations of speakers would perceive these two closelyrelated dialects as stylistic varieties of the same code, and could drawon either of them in the process of language acquisition. If thespeakers of Hittite were in the majority in a given area, it isreasonable to expect that Hittite elements would predominate in thekoine of this area, but it does not impose any restrictions on theclasses of words that could be adopted from Luvian. A very generalprediction that can be made is that formally simpler and functionallymore transparent forms would have better chances to enter the newkoine. Thus one can hypothesize that the Luvian clitic =ti would begiven preference over its Hittite counterpart =te \ in view of thegeneral tendency toward eliminating long vowels in clitics. Likewiseone could conjecture that the sequence *=tu=ti would be givenpreference over *=te=ti in order to avoid potential confusion betweentwo phonetically similar clitic pronouns. The explanatory force ofsuch accounts should not overestimated: we cannot make exactforecasts about elements that would be borrowed into a koine in aparticular contact situation. The point of comparison with medievalEngland is rather to stress that basic lexical items can be easilyadapted from various sources in a context of koineization.

We do not have historical records that would shed light on theearly convergence between Hittite and Luvian. One may, however,try to place it within a relative chronology of phonological andmorphosyntactic changes that occurred on the way from Proto-Anatolian to Hittite. While the relative chronology does not directlytranslate into the absolute chronology, a significant number ofchanges that occurred between the postulated contact episode and theoldest attested stage of the Hittite language would preclude thehypothesis that the reconstructed koineization of Hittite was

30 The only member of Swadesh’s one hundred word list that was borrowed intoEnglish from French is mountain, an item of short supply in southern England. Thewords belonging to this list that were either borrowed from, or influenced by, OldNorse include at least die, egg, give, and sky.


connected with the expansion that resulted in the creation of theHittite Old Kingdom. We have already seen that the generalization ofHitt. =z(a) as a reflexive particle co-occurring with all types ofsubjects (stage 6) represents one of such changes. The relativechronology of other developments is contingent upon theinterpretation of the Hittite medio-passive extension -di (with asynchronic variant -d).

The inventory of Old Hittite medio-passive verbal endings, withtheir traditional grammatical interpretation, extracted from Kassian(2002), is given in Table 12 below.

One can see that the present and preterit medio-passive endingshave an identical first part for each person/number, but differ withrespect to their extensions. According to received wisdom, thesecond part of Old Hittite present endings is zero or -ri (rarely -ni),while the preterit endings are extended by the elements -di or -d. Thisdistribution is somewhat modified in the later history of Hittite,where we encounter present endings -tadi (2sg.), -wastadi (1pl.) and-tumad (2pl.) alternating with the variants in -ri. At the same time,the extension -ri remains restricted to the present tense throughoutthe history of Hittite (Neu 1968:24-30).

Present Preterit Imperative Sg. Pl. Sg. Pl. Sg.

1 -ha; -hari; -wasta (1x) -had N/A N/A2 -ta (1x) -tumari (1x) -tad N/A –hud3 (I) -ta; -tari

(II) -a; -ari; -ani

-anta; -antari

(I) -tad(II) –adi

-antadi (I) –taru(II) –aru

Table 12: Old Hittite medio-passive verbal endings.

The lack of strict association between the dental extension andthe preterit tense was confirmed through my examination of the OldScript corpus. Contrary to the established opinion, the formsuppiyahhadi, occurring in the script of an Old Hittite ritual, shouldbe interpreted as a present tense verb:

(16) KBo 25.112 ii 14'-15', OS, CTH 733.II.4 (Neu 1980:191)LUGAL-ué=za éuppiyahhadi ANA [hal]pudi ma\nha\nda ma\ldike\=a QATAMMAThe king purifies himself. As he chants before the halpudi-altar, so also (he chants) these things in the same way.

The first components of Hittite medio-passive endings, markingperson and number, have, in their majority, a good pedigree going

Ilya Yakubovich 99

back to a set of reconstructed Indo-Hittite inflectional markerslabeled as “perfect,” “stative,” or “proto-middle” by differentscholars and reflected in the perfect and medio-passive conjugationof various Indo-European languages.31 The equation between theHittite medio-passive extension -ri and the Indo-European element -rwith an identical function, augmented by the hic et nunc particle -i, islikewise well established (Yoshida 1990). By contrast, the indicativeextensions –di/-d lack direct counterparts in Indo-European, and mustbe considered a Hittite or Anatolian innovation.32 According to theplausible suggestion of Kury`owicz (1964:69), Hittite 3sg. pret. -ta-drepresents a combination of the old medio-passive and the new activepreterit ending. This attractive hypothesis also allows one to explain1sg. pret. -had and 2sg. pret. -tad as *-hadi and *-tadi (attested inlater Hittite texts), which had undergone paradigmatic leveling. Thisanalogical development, in its turn, could bring about the generalconfusion between the extensions -di and -d, which came to beperceived as two synchronic variants of the same morpheme, andwere eventually associated with present and preterit respectively inlater Hittite.33

As for the medio-passive extension -di, its origin is not to besought in the verbal morphological inventory. Neu (1968:145)connected this morpheme with the pre-Hittite reflexive clitic*=ti/=di. The stranding of the original Wackernagel clitic to the post-verbal position has several parallels within Indo-European, but thesuffix -s’a/-s’ marking the Russian reflexive voice and going back tothe common Slavic Wackernagel reflexive clitic *=s e ç provides aparticularly close analogy. Example (16) cited above can serve as an 31 See Jasanoff (2003:1-29) for a survey of previous opinions about the reconstructedmeaning of this conjugation, and the rest of the book for Jasanoff’s own views.32 Pedersen (1938:110), followed by Oettinger (1997:417), connected Hitt. -di/-dwith the Indo-European imperative suffix *-dhi. This etymology remains very likelyfor the Hittite 2sg. middle imperative ending -hu-d, even if it requires the assumptionof an irregular apocope. Carruba (1992:15-31) linked -di with medio-passiveinfinitive suffixes Ved. -dhyai and Gk. -syai, and boldly postulates the existence ofan Indo-European medio-passive suffix *-dh- with various language-specific formalextensions and functional restrictions. The common handicap of both theories is thenecessity to invent more or less artificial analogies that would account for thetypologically unlikely spread of dental suffixes from peripheral forms to theindicative.33 Alternatively, one can hypothesize that the variant -d was always due to theapocope of the earlier -di, or to its analogical reanalysis in preterit forms. Thishypothesis, however, appears less adequate for the purpose of explaining thecomplementary distribution of -di and -d in specific personal endings in Old Hittite.In addition, if Pal. éar-ku-ta-at (KUB 35.165 Obv. 6) is indeed a medio-passiveform, as per Kammenhuber (1959), then the -d extension represents a Luvo-Palaicinnovation, whereas the -di extension must be exclusively Hittite.


instance of a syntactic construction in which such a stranding mighthave originated.34 Neu’s approach allows one to explain why theextension -di does not appear to have been restricted to a particulartense in Hittite.

Neu’s suggestion was not immediately accepted by all scholars.In recent years, however, it received a striking confirmation from theanalysis of Hieroglyphic Luvian sentences with an element -siappended to verbal endings. The medio-passive character of suchforms is not immediately clear in each and every context, but can beunambiguously inferred from the comparison between (17) and (18),where -si alternates with -ru (3sg. impv.) within the paradigm of thepassive verb iziya- ‘to become’.

(17) ÇINEKÖY , §7 (Tekog'lu and Lemaire 2000)|hi-ia-wa/i-sa-ha-wa/i(URBS) |su-ra/i-ia-sa-ha(URBS)“UNUS”-za |DOMUS-na-za | i-zi-ia-siQue/Hiyawa and Assyria became one house.

(18) KARATEPE, §53 (Hawkins 2000:55)REL-pa-wa/i za (“CASTRUM”)há+ra/i-ní-sà-||za i-zi-ia-ru(DEUS)BONUS-sa (DEUS)VITIS-sá-háThen let this fortress become (one) of the Grain-god and theWine-god.

Oshiro (1993:54) was the first one to link the Luvian medio-

passive marker -si with the Palaic Wackernagel clitic =si, whichprobably had a reflexive meaning. This idea found furtherdevelopment in the work of Rieken (2004:183-184), who drew adirect parallel between the spread of Hitt. *=ti/=di and Luvian =si tothe post-verbal position, thus vindicating Neu’s theory.35 There is,however an important difference between the relative chronologies ofthe two syntactic processes. The spread of Luv. =si must haveoccurred before “stage 3,” which effectively eliminated thisWackernagel clitic. The rarity of verbal forms in -si in HieroglyphicLuvian texts, together with their complete absence in Cuneiform

34 For the reconstructed duplication of the Wackernagel accusative anaphoric clitic inother positions in the history of Lycian compare the account of Garrett (1991).Alternatively, one can envisage the scenario according to which Wackernagel cliticswere first attached to medio-passive forms in verb-initial sentences of the typedescribed in Sidel’tsev (2002). What speaks against such a solution is the rarity ofthis syntactic type in the original Hittite compositions.35 A more remote parallel is the creation of the Lydian emphasizing particle =s ;,which can be analyzed as the original reflexive marker =s ;/=is stranded from theexpected second position, as per Melchert (1991).

Ilya Yakubovich 101

Luvian, bears witness to the vestigial character of this formation. Onthe contrary, the stranding of Hitt. *=ti/=di must have postdated“stage 3,” and probably “stages 4/5” as well. Otherwise, it would bevery difficult to explain why the second person clitic *=te\/*=de \ wasgeneralized as a medio-passive marker, and then underwent anirregular contraction to *=di under the influence of a Luvian cliticthat never had this function.

The formal relationship between the Wackernagel clitic *=ti andthe verbal extension -di in Hittite should be discussed with referenceto Proto-Anatolian lenition. According to the standard formulation ofthis sound law, Proto-Anatolian voiceless stops were voiced afteraccented long vowels, or between two unaccented vowels (cf.Melchert 1994:60 with ref.). Its effects are best preserved in theverbal system of Luvian, while in Hittite it was obfuscated by anumber of subsequent analogical changes. Nevertheless, occasionalsynchronic alternations within Hittite verbal paradigms (e.g. a \:gi‘(s)he dies’ vs. akkánzi ‘they die’) vindicate the existence of aphonologically conditioned lenition at the earlier stages of the Hittitelanguage. The distribution of the clitic allophones *=te\/*=de\ etc.,reconstructed for Pre-Luvian and Pre-Hittite in the tables above, waspresumably governed by the laws of lenition. The voiceless variantwould be favored in a position after another consonant while thelenited variant would be preferred in a position between theunstressed vowels of the clitic chain.

This distribution must have been alive in Hittite till the momentwhen *=ti/*=di became licensed in postverbal position and wasspecifically associated with the medio-passive conjugation. Since allthe unextended medio-passive markers ended in unstressed vowels,the variant =di was automatically chosen in this environment. Laterin the history of Hittite, the laws of lenition lost their synchronicstatus, and all the second-position clitics generalized the unlenitedallomorphs. This is clear from the orthographic variants withconsonant doubling, such as =tta and =ttu, which occur in positionswhere lenition is otherwise predicted.36 This analogicalgeneralization was not applied, however, to the medio-passiveextension -di, which presumably had already lost the synchronicconnection with its original source. By contrast, Luvian hasgeneralized the lenited variant of the Wackernagel pronominal clitics(=di > =ri). This fairly late development was not shared by the

36 Numerous variants such as namma=tu=za (for the expected **namma=ttu=za) in(14) must be regarded as instances of simplified spelling. This phenomenon, brieflydiscussed in Yoshida (1998:607), deserves much fuller investigation aimed atdetermining its licensing conditions.


closely related Lycian language, which patterns with Hittite inpreserving the voiceless variant =ti (Melchert 2004:65).

The final accord that brings us to the stage attested in the earliestHittite texts is the affricativization /t/>/ts/_/i/, accompanied by theapocope of -i in word-final position.37 The posited sound changemust have occurred after the lenited variant *=di was wiped out ofthe Hittite clitic system, because otherwise *=ti would have beenretained under analogical pressure. The fact that this sound changefailed to occur in the case of the medio-passive extension -di refutesthe contention of Yoshida (1998:614) that both *t and *d, as opposedto *dh, were palatalized before the high front vowel. Note that, evenif we make an unlikely assumption that Hitt. =z(a) and Hitt. -di arenot related, we still have to admit that affricativization was a fairlylate change because it applied to the clitic =ti/=di borrowed fromLuvian. It would require a vivid imagination to believe that IH *dh,which did not otherwise yield reflexes distinct from *d in any of theattested Anatolian languages, was preserved in Pre-Hittite at such alate stage.38

The observations made above can be summarized in thefollowing relative chronology:

1. Hittite borrows Luvian *=ti/=*di.2. Hitt. *=ti/=*di is generalized to all persons and numbers.3. Hitt. =di spreads to the postverbal position, where it looses its

clitic status.4. The variant =ti is generalized in the second position.5. /t/ > /ts/ /_/i/; /tsi/ > /ts/ /_#

37 Cf. Kimball (1999:191-192). I find it difficult to believe that the apocope of -i wasalso regular in a position after d , since the preservation of the medio-passiveextension -di militates against this assumption. In my opinion, the apocope in theimperative suffix -dhi must be viewed as one of those irregular shortenings thatfrequently occur in imperative forms.38 The philological considerations adduced by Yoshida (1998) in favor of the soundchange */d/>/dz/_/i/ do not hold water, in my opinion. It is true that some verbalforms in *-e \zzi are sometimes spelled with -e \zi/-ezi in Old Hittite, but, in those caseswhere we have enough data (e.g. ie \zzi ‘he makes’ or pehudezzi ‘he brings’) thisspelling appears to be less frequent than -e \zzi/-ezzi (Kassian 2002:99, 116). In such asituation, I prefer to think that the variants -e \zi/-ezi do not testify to the existence of aseparate phoneme /dz/ but rather represent yet another instance of simplifiedspellings. This is precisely the explanation adduced by Yoshida (1998) foroccasional Old Hittite 3sg. verbal forms in -nuzi occurring in free variation with-nuzzi The argument that the sign IZ is too simple to be omitted by less diligentscribes is not compelling because, after all, it takes more effort to write even a three-stroke sign than to write nothing!

Ilya Yakubovich 103

One cannot exclude that some of the postulated changes (e.g #2and #3) occurred simultaneously. Yet the sheer complexity ofphonological and morphosyntactic evolution that postdated thediscussed contact episode prompts one to posit a period of severalhundred years between the adaptation of Luv. *=ti/=*di and the stageof Hittite attested in the Old Script texts. Prehistoric contacts betweenHittite and Luvian are probably to be dated back to the late 3rd

millennium BC.References

Alp, Sedat1991 Hethitische Briefe aus Masçat-Höyuk (Türk Tarih Kurumu Yayınları

VI-35). Ankara: Türk Tarih Kurumu.Boley, Jacqueline

1993 The Hittite Particle –z/-za. Innsbruck: Institut der Sprachwissenschaftder Universität Innsbruck.

Bryce, Trevor2003 History. In: The Luwians. H. Craig Melchert (ed.). Leiden/Boston:

Brill, 27-127.Carruba, Onofrio

1964 Hethitisch –asta-, -(a)pa- und die anderen Ortspartikeln. Orientalia(New Series) 33:405-436.

1969 Die Satzeinleitenden Particeln in den indogermanischen SprachenAnatoliens. Roma: Ateneo.

1992 Die Endungen auf –ti des hethitischen Mediums. Studii Micenei eEgeo-Anatolici 29:15-31.

Dawson, Hope2003 Defining the Outcome of Language Contact: Old English and Old

Norse. Ohio State University Working Papers in Linguistics 57: 40-57. Available for download at www.ling.ohio-state.edu/ publications/osu_wpl/osuwpl_57/ hope_wpl.pdf.

Eichner, Heiner1974 Untersuchungen zur hethitischen Deklination (Dissertation

Teildrück). Erlangen.Fortson, Benjamin W. IV

2004 Indo-European Language and Culture: An Introduction. Oxford:Blackwell.

Garrett, Andrew1991 The Lycian Nasalized Preterite. Münchener Studien zur

Sprachwissenschaft 52:15-26.1996 Wackernagel’s Law and Unaccusativity in Hittite. In: Approaching

Second: Second Position Clitics and Related Phenomena, A.L.Halpern and A.W. Zwicky (eds.). Stanford: CSLI Publications, 85-133.

Hawkins, J. David1992 The Inscriptions of the Kızıldag' and the Karadag' in the light of the

Yalburt Inscription. In: Hittite and other Anatolian and Near EasternStudies in Honour of Sedat Alp, Ekrem Akurgal et al. (eds.). Ankara:Türk Tarih Kurumu, 260-275.

2000 Corpus of Hieroglyphic Luvian Inscriptions. Volume I. Part I, II:Texts; Part III: Plates. Berlin/New York: W. de Gruyter.


Hopper, Paul J., and Sandra A. Thomson1980 Transitivity in Grammar and Discourse. Language 56(2):251-299.

van den Hout, Theo2002 Self, Soul and Portrait in Hieroglyphic Luvian. In: Silva Anatolica.

Anatolian Studies Presented to Maciej Popko on the Occasion of his65th Birthday, Piotr Taracha (ed.). Warsaw: Agade, 171-186.

2006 Institutions, Vernaculars, Publics: The Case of Second MillenniumAnatolia. In: Margins of Writing, Origins of Culture. NewApproaches to Writing and Reading in the Ancient Near East, SethSanders (ed.). Chicago: Oriental Institute Press, 217-256.

Houwink ten Cate, H.J. Philo, and Folke Josephson1967 Muwatallis’ Prayer to the Storm-God of Kummanni. Revue Hittite et

Asianique 25:101-140Ivanov, Vyacheslav V.

2001 Southern Anatolian and Northern Anatolian. In: Greater Anatoliaand the Indo-Hittite Language Family, Robert Drews (ed.).Washington: Institute for the Study of Man, 131-183.

Jasanoff, Jay H2003 Hittite and the Indo-European Verb. Oxford: Oxford University

Press.Kammenhuber, Annelies

1959 Das Palaische: Texte und Wortschatz. Revue Hittite et Asianique17:1-92.

Kassian, Alexei2002 Glossary of Verbal Forms and Derivatives from Published Old Hittite

Texts. In: Anatolian Languages, Vitaly Shevoroshkin and Paul J.Sidwell (eds.). Canberra: Association for the History of Language,72-136.

Katz, Joshua1998 Topics in Indo-European Personal Pronouns. Harvard University

PhD dissertation.Kimball, Sara

1999 Hittite Historical Phonology. Innsbruck: Institut derSprachwissenschaft der Universität Innsbruck.

Kühne, Cord1988 Über die Darstellung der hethitischen Reflexivpartikel –z. In:

Documentum Asiae Minoris Antiquae. Festschrift für Heinrich Ottenzum 75. Geburtstag, Erich Neu and Christel Rüster (eds.).Wiesbaden: Harrassowitz, 203-233.

Kury`owicz, Jerzy1964 The Inflectional Categories of Indo-European. Heidelberg: Carl

Winter.Laroche, Emmanuel

1959 Dictionnare de la langue louvite. Paris: Adrien-Maisonneuve.Meier-Brügger, Michael

2000 Indogermanische Sprachwissenschaft. Seventh, fully revised edition.Berlin/New York: Walter de Gruyter.

Melchert, H. Craig1988 “Thorn” and “Minus” in Hieroglyphic Luvian. Anatolian Studies

38:29-42.1991 The Lydian emphasizing and reflexive particle -s ; / is. Kadmos

30:131-142.1994 Anatolian Historical Phonology. Amsterdam/Atlanta: Rodopi.

Ilya Yakubovich 105

2003a Prehistory. In: The Luwians, H. Craig Melchert (ed.). Leiden-Boston:Brill, 8-26.

2003b Language. In: The Luwians, H. Craig Melchert (ed.). Leiden-Boston:Brill, 170-210

2004 A Dictionary of the Lycian Language. Ann Arbor/New York: BeechStave Press.

2005 The Problem of Luwian Influence on Hittite. In: Sprachkontakt undSprachwandel. Akten des XI Factagung der IndogermanischenGesellschaft, 17-23 September 2000, Halle an der Saale, G. Meiserand O. Hackstein (eds.). Wiesbaden: Reichert, 445-459.

Morpurgo-Davies, Anna1980 The personal endings of the Hieroglyphic Luvian verb. Zeitschrift für

Vergleichende Sprachforschung 94:86-108.1982 Dentals, Rhotacism and Verbal Endings in the Luvian Languages.

Zeitschrift für Vergleichende Sprachforschung 96:245-270.Neu, Erich

1968 Das hethitische Mediopassiv und seine indogermanischenGrundlagen (Studien zu den Boghazköy-Texten 6). Wiesbaden:Harrassowitz.

1980 Althethitische Ritualtexte in Umschrift (Studien zu den Boghazköy-Texten 25). Wiesbaden: Harrassowitz

Oettinger, Norbert1997 Der Partikel –z des Hethitischen’ (mit einem Exkurs zu den

Medialformen auf –t, -ti). In: Berthold Delbrück y la sintaxisindoeuropea hoy, E. Crespe and J.L. García-Ramón (eds.). Madrid:Ediciones de la UAM / Wiesbaden: Reichert, 407-420.

Oshiro, Terumasa1993 Notes on Hieroglyphic Luvian. Orient 29:45-56.

Pedersen, Holger1938 Hittitisch und die anderen indoeuropäischen Sprachen. Copenhagen:

Munksgaard.Rieken, Elisabeth

2004 Das Präteritum des Medio-Passivs im Hieroglyphen-Luvischen.Historische Sprachforschung 118:179-88.

Sidel’tsev, Andrej2002 Inverted Word Order in Middle Hittite. In: Anatolian Languages,

Vitaly Shevoroshkin and Paul J. Sidwell (eds.). Canberra: Associationfor the History of Language, 137-188.

Tekog'ln, R., and A. Lemaire2000 La Bilingue Royale Lonvito-Phénician de Çineköy. Comptes

Rendues de l’Académie des Inscriptions 2000:961-1006.Thomason, Sarah J.

2001 Language Contact: An Introduction. Washington: GeorgetownUniversity Press.

Yakubovich, Ilya2002 Nugae Luvicae. In: Anatolian Languages, Vitaly Shevoroshkin and

Paul J. Sidwell (eds.). Canberra: Association for the History ofLanguage, 189-209.

Yoshida, Kazuhiko1990 The Hittite Medio-passive endings in –ri. Berlin/New York: Walter de

Gruyter.


1998 Hittite verbs in –Vzi. Acts of the IIIrd International Congress ofHittitology, Sedat Alp (ed.). Ankara: Grafik, Teknik Hazırlık UyumAjans, 605-614.

2001 Hittite nu-za and Related Spellings. In: Akten des IV. InternationalenKongresses für Hethitologie, Würzburg, 4-8 Oktober 1999, GernotWilhelm (ed.) (Studien zu den Boghazköy-Texten 45). Wiesbaden:Harrassowitz.

Ilya [email protected]

Download - Prehistoric Contacts between Hittite and Luvian: The … · Prehistoric Contacts between Hittite and Luvian: The Case of Reflexive Pronouns* Ilya Yakubovich University of Chicago

Top Related