Genitive Constructions in Sumerian Gabor Zolyomi Journal of Cuneiform Studies, Vol. 48 (1996), 31-47. Stable URL: htp//links jstor-org/sici?sic!=0022-0256%281996%29489%3C31%3AGCIS%3E2.0,CO%3B2-B Journal of Cuneiform Studies is currently published by The American Schools of Oriental Research, ‘Your use of the ISTOR archive indicates your acceptance of JSTOR’s Terms and Conditions of Use, available at hhup://uk,jstor.org/abouvterms.himl. JSTOR’s Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you ‘may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at hup://uk,jstor.org/jourmals/asor. hm Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the sereen or printed page of such transmission, STOR is an independent not-for-profit organization dedicated to creating and preserving a digital archive of scholarly journals, For more information regarding JSTOR, please contact support @jstor.org. hups//uk jstororg/ ‘Thu Ape 14 12:57:43 2005GENITIVE CONSTRUCTIONS IN SUMERIAN Gabor Zélyomi The Oriental Institute, University of Oxford | ELTE, Budapest Scholars describing the Sumerian noun phrase usually use the expression “chain” (“Wortkette”). As far as T can judge, this expression has its origin in Poebel’s statement: “seinem aggluti- nnierenden Character gemiiss driickt das Sum- erische die grammatische Zusammengehdrigkeit der einzelnen Worter oder Satzelemente im all- gemeinen nur dadurch aus, dass sie dieselben zu Wortketten aneinander reiht” (Poebel 1928, 85 §98). While Hungarian is also characterized as an agglutinative language, I am not familiar with any Hungarian grammar that would use this expression, And indeed the Hungarian phrase (1) a nagy kirdly-ok fényiiz6 palotéjs-0-ban! the big king! opulent palacepossSrd persloe in the opulent palace of the big kings would not remind anybody of @ chain, although its Sumerian counterpart perhaps could: (2) egal mab lugal galeneaka An earlier version ofthis paper was read atthe 3rd meet {ng ofthe Sumerian Grammar Discussion Group in Oxford, I~ 8 July 1992. woud lke to thank the participants J. A Black, 0. Eézard, Th Jacobsen, B, Jagersma, J Krecher, P. Miche Towski, CL Wilcke, M. Yoshikawa) fo thir useful comments ‘on the pape Iam especially indebted to Bram Jagersma who ead several versions ofthis article and his detailed eriteism ‘nd suggestions had impact on ts fnal form. Lam also grateful to KE Kiss for her kindness to comment onthe linguistic side ofthis paper. Nedless to say, all errors are mine I gratefully acknowledge the financial support covering my trivel ex penses between Budapest and Oxford given by the “Alapit: ‘ny a Magyar Felsioktatasért és Kutatiser” Foundation 1. For Hungarian examples I follow the description and segmentation of Szaboes\ (1981 and 1983). Her segmentation eventually goes back to MeTeuk (1973) Pss stands fr a pos sessive marker. 8 isa sufix ageing herein person sith the possessor. 81 T think that phrases similar to (2), in which all affixes stand at the end of the phrase, could have led Poebel to use the term “Wortkette” My point is that this feature has nothing to do with aggluti- nation, rather itis the consequence of the order of constituents of the noun phrase and of the nature of various affixes in Sumerian, 1, The Noun Phrase in Sumerian LLL. In the generative grammar of the 80s (called the "Government and Binding Theory” GBI) various phrases are assigned a structure based on the X-bar theory? So, a phrase like (Ga) the great king’ small house will have the following structure: (8b) Pm NPS N, 4 \ /\ aw ON | I the great king’ small house or, using labeled bracketing instead of tree- diagrams? 2. For information on the existence of different phrase levels in X-bar theory one ean consult Radford (1988, “Chap: ter 4 Noun Phrases") or Haegeman (1981, "Chapter 2; Phrase Structure") A short, popular, and yet very informative acount of GB, including the Xcbar theory can be found in Bickerton (1980, 57-74). ‘3. Actually house should be an Nin English because of| its possible complement (eg, house with red ro). In Sumer fan this position does not exist and therefore for the sake of| simplicity I take it as N. It does not have any beating on our tople. 105 45 10832 GABOR ZOLYOMI (Be) tye, bythe great kingly larsmallliy housel] Nis the head of the phrase, AP is in the position of the modifier and the NP, dominated immedi- ately by the NP}, is in the specifier position, so the possessor occupies the specifier position 1.1.2. Let us try to assign a similar structure to Sumerian nominal phrases (4a) e tur lugal gala the small house of the great king (4b) xm, a ™ x NP, SN, oe e tur lugal galak (46) typ, lol ella trl lugal gaa] ‘As we can see from these examples, the two phrases differ only in the order of their constitu- ents. In English the specifier and the modifying adjective are in front of the noun, while in Sume- rian, itis the other way round—the two structures are mirror images of one another. As far as the genitive marker is concerned, in English it is attached to the NP and not to the N, which can be deduced from phrases like: (8) the king of England’s house (lis the king that possesses the house and not England) If we assume that in Sumerian the potential af- fixes (ie. possessive suffix, plural and case mark- ers) are attached in the same way to phrases’ then it can be claimed that the main difference between the two languages lies in the order of 4, One possible way to describe the intuition underlying the three levels of phrases (head, modifiers, specifier) isthe following: “the three layers of X-bar structure represent, re= speetvel, (a generic class, X;(b) the properties peculiar to particular members ofthat cass (large with a dark red cover, ‘of Mary’) and (the specification ofthe complete individual in terms of abstract relations such as quantity, proximity, familiarity, and so on (a, this. three)” (Bickerton 1980, 195) 5 This sequal to saying that all these alixes are in effet clities. I think tha this s the ease T return to this topie in de- tall below (See 2 their constituents. In order to make my point more clear let us compare the following two phrases: (6a) John’s little cats green house (ob) “lw a \ /\ Nwy Y aww ZN ey Johns tle cats greenhouse (60) Iyebselve, John’ litle cat) green house] (7a) ninda.za mu-ka-key Sulgi P, Section b 1.9) the offering of the New Year m) um wry b\w, 4 (7) Ip, minda fy, 24 yp, mua] ak] el What (7) shows is that the succession of genitive markers is rather ostensible, Each genitive marker belongs to a different phrase and each phrase is embedded into another one, “because phrases are not, as they might appear to be, strung, together serially, like beads on a string, Phrases are like Chinese boxes, stacked on inside another” (Bick- erton 1990, 60). The reason why in Sumerian, unlike in English, all the genitive markers could fall at the end of the whole phrase is that the order of constituents is reversed compared to the order in English. 1.13. Poebel’s description and illustration of the same phenomenon (“é-dumu-(lugal-ak)-ak}- a [Poebel 1923, 184 §367-68; see also note 9 herel) shows some similarity to mine, but as he used a different framework and terminology he ‘was not able to point out the cause of this seem- ingly unique way of forming noun phrases and its relation to constructions in other languages.GENITIVE CONSTRUCTIONS IN SUMERIAN 33 1.2.1. Consider the following two descriptions of Sumerian noun phrases: a. Falkenstein (1959, 87 §19): “1, Substantiv, 2 attributiv Adjektiv, 8. Rektum des Genetivs, 4. pronominal suffix, 5, Plurazeichen(-ene), 6, Postposition” , Thomsen (1984, 54 $46); Regens Rectum NOUN-ADJ- NOUN-ADJ— Rectums affixes Regens' affixes POSS-PLUR-GEN— _PLUR-CASE-COP The difference between the two systems is that ‘Thomsen’s scheme cannot generate constructions like “é-nir-su-karni ‘sein Haus von Girsu’ (= 1 + 8+ 4)" or “-unwkir-ga-ko-ni ‘ihr Haus der “Heili- gen Stadt”” (= 1 + 8[= 1 + 2} + 4)” (Falkenstein 1959, 87 §19). I would like to return to this sort of construction later in the third section of the paper; here I will analyze only those phrases that fit into Thomsen’s system! ‘What is noticeable and, in my view, misleading in Thomsen’s chart, is that the system is asym- ‘metrical with respect to the affixes and cases al- lowed. A regens cannot have a possessive suffix. It cannot have a genitive marker either, only mark- ers of other cases, which seems to imply that the genitive and other cases are in a different rank (see also Thomsen (1984, 90-91 $163] and Hayes 6.1 do not treat the genitive construction deseribed in Yoshikawa (1992). [think it would be misleading not to take ino account that Sumerian changed during its history. None of the modern grammars of English, for example, would describe the verb inflection merging the Old> and Moder English paradigms into one system. As far as ean se, the rl evant examples ofthe article considered all come from later periods ofthe language, so the construction is very likely to be Secondary development in Sumerian, One cannot rule out the impact of another language, namely Akkadian. I ageee with the author in identifying the element -¢ as a demonstrative enelitic. From a formal point of view, the construction differs from the common noun phrase of Sumerian in thatthe de- rmonstrative clement occupies the slot af modifier (eg the place of adjctives) instead of that of specter. Its easy to notice that in Akjadian the demonstrative anniim behaves similarly to adjectives: bitwn anni, bitum sirum, This explanation ‘would consider only the formal side ofthe problem As far as the function ofthe construction is considered, that is stilt be investigated. [1990, 54-55]). This is suggested by Jacobsen as well: “That -ak- is... seen to precede casemarks, to be compatible with them rather than being re~ placive for them, sets it apart as representing, or belonging to, a separate and different rank class It is not one of them” (Jacobsen 1973, 162). The same can be said against Jacobsen's argument already mentioned in 1.1.2: A genitive case and a casematker following it belong to different phrases. The occurrence of double or triple geni- tive markers (Jacobsen 1978, 163) also does not entail that the genitive must be set apart from other markers. Cases, in my opinion, express syn- tactical relations between constituents of a sen- tence. Cases that are usually classified in Sumerian as adverbal cases mark a relation between a NP and a verb, Since every clause contains only one verb, itis natural that the markers of these rela- tions occur only once in a sentence.’ The genitive marker, however, marks the relation of two NPs, and since somebody (or something) that belongs to somebody (or something) can have something belonging to it, the marker of this relation can ‘occur more than once in a clause. So, in the case of the genitive element, only the related constit- uuents are different compared to the adverbal ‘easemarkers; the character of the relation is not different. 1.2.2. Consider the following phrase: (8a) dumu lugalakenera for the sons of the king (8b) fyey lly dumullfyp,lugalak] ener) A different version of (8) is the so-called “antici- patory genitive” [AG]; (©) lpg luealakllyp, byy duro ani] enea) In (9) the rectum precedes the regens and a pos- sessive suffix agreeing in class, person, and num ber with the possessor is attached to the regens. Which position is occupied now by -ani? Act cording to Thomsen’s chart, a regens cannot have a possessive suffix. Is -ani in the position that would 7. Adverbial cases behave differently inthis respect but [NPs marked by these cases, in a stricter sense, are not om strued by the verb34 GABOR ZOLYOMI “otherwise be occupied by the rectum’s possessive suffix? Examples like (10) €g4 nt-gal-bi (Gudea cyl. 9:17) the great awesomeness of my house show that the answer must be in the negative. I think that -ani occupies the place of the whole rectum: it is in the specifier position of NP, sox in (9) is NP, ie, it is in complementary distri- bution with the rectum, It is reasonable then to assume that Sumerian possessive suffixes always ‘occupy the specifier position of NP. Taking this for granted, we have a simpler grammar because the asymmetry that the regens cannot have a poss, sufi disappears, and in this grammar, the genitive isin the same rank as the other cases. 1.23, The Sumerian noun phrase then has the structure (11) NP + N“XPNP (head modifier specifier) (12) NP + PLURAL + CASE, The head is the only obligatory constituent of a NP. The modifier and the specifier positions are ‘optional, and there can be more than one modi- fier. XP can be frst of all an adjective phrase, but the various non-fnite and nominalized finite verb forms can also fill this position. The specifier po- sition can be occupied either by a full NP or by 1 possessive pronoun. NPs occupying the specifier position are in the genitive case, so the difference between genitive and other eases is not in their rank but in the syntactic position of the NP they are attached to, eg, ergative case is characteristic of NPs that are subjects of transitive verbs. NPs are followed by two “slots” Slot 1 is that of a plural marker, slot 2 is that of a casemarker. The chain-like character of Sumerian NP is due to the order of head, modifier, and specifier, to the ma- ture of various affixes, and to the fact that the NP occupying the specifier position of another NP, that is the rectum, is formed after the same rules. Therefore it can have a modifier and a specifier, and it could be in the plural and have a case (namely the genitive). Both Falkenstein’s and Thomsen’s descriptions of the Sumerian NP, as must be apparent by now, fail to capture an important characteristic of the language: the rectum and the possessive pronoun occupy the same position. Moreover, Thomsen, with Jacobsen and Hayes, thinks that the genitive and the other cases are not in the same rank. In both cases the explanation for this lies in the implicit assumption® that the “language is some Kind of serial stringing process” (Bickerton 1990, 61) as opposed to the “Chinese boxes” principle, Another reason why Falkenstein’s and Thomsen's descriptions of the distribution of the rectum and the possessive pronoun are incorrect lies in not recognizing the real status of an element in the language. If the possessive pronoun is classified asa kind of sufi,’ it really could seem odd that it shares its slot with a NP, but if one assumes that this element is a clitic then everything finds its place. The next part of the paper attempts to lucidate the role of clitics in Sumerian, 2. Clities in Sumerian, 2.1.1, As itis clear from 1.2.22, the possessive suffixes are in complementary distribution with NPs Syntactically they are words. This property makes it very likely that our usage of the term “suffix” is not entirely appropriate. We should con- sider whether it would be better to classify them as clities. 8. This assumption is not even tacit since Falkenstein uses the term “Kettenbildung” and Thomsen names the NPas "nominal chain.” 8. These examples can also serve as an argument against some scholars who think that moder linguistic approach provides only a new terminology without adding anything to ‘ur knowledge and understanding of Sumerian. Interestingly ‘enough, Poebel's description of Sumerian difers from the two schemes concerned with respect tothe distribution ofthe rex- tum and the passessive pronoun. In his scheme, the “besitean- 2eigendes Fuirwort” and the “besitzanzeigender Genetiv” oe cupy the same slot (Poebel 1923, 35 $08). In connection with the string quoted above in 11.3, Poebel states that “in diesem Wortkomplex am Anfang die drei Substantiva und am Schluss die drei Fostpositionen zusammenstehen, u2-in einer dear igen Anordnung, dass das erst Substantiv und de lette Post position, das mittlere Substantiv und die mitterePostposiion, das letzte Substantiv und die erste Postposition zusammen fehiren, die zusammengehsrigen Paare also konzentrsch ‘neinander geschachtelt sind” (1023, 184 §968) In stating this, Poebel was very near to freeing himself from the constraints ofthe view tha teats constituents as beads ona string,GENITIVE CONSTRUCTIONS IN SUMERIAN, 35 21.2, First [ will try to review the different definitions of cities since their various treat- ‘ments or classifications could give entirely differ- ent results with respect to whether an element in a given language can be perceived as a clitc. What one can ascertain from these treatments is the following: Clities are elements that share the features both of fully-fledged words and of affixes. They “give the syntactic properties of words but the phono- logical properties of affixes” (Spencer 1991, 391), Clitieness is a graded continuum; there are sev- eral parameters to take into account, and in most cases a given element will behave in one case as a ‘word but in another case as an affix. Cliticness is an intermediate status between being a word and developing into an affix. Usually the following considerations are taken into account when deciding the status of an clement: “Prototypical elities are syntactically like words, in that they are relatively independent of the words they attach to (that is, they are not specifically selected by their bases), they have a straightforward meaning, they tend not to show idiosyncratic allomorphy themselves, and they do not condition idiosyneratic allomorphy on their hosts. Inflectional affixes, on the other hand, are highly dependent on the bases they attach to, ex- hibit multiple exponence, and show much allomor- phic variation including suppletion. In addition, they frequently condition idiosyncratic allomor- phy on the stems to which they attach” (Spencer, 1991, 881), An element attached after a clitic is a dltic itself; clities are not followed by affixes. Clities can gather into clusters (eg, Serbo-Croat pronominal clitics) and in these cases they have a particular order. ‘There are several classifications of cities. Zwvicky (1977) classifies the clities into three groups. Sim- ple clitics or reduced forms of words, come into being because of speech rate, level of formality, and the like. They can be substituted by the full forms, unlike the special clitics, which are the allomorphs of full form words, but are not de- rived from their equivalents as a result of phrase phonological reduction processes. Bound words do not have a full form equivalent but need a host and can be restricted to a particular sentence position, Another classification is that of Klavans (1985) She is not interested in the origin of clities be- cause “the source of clitis is probably a language particular fact... However, what is strictly con- strained is the position where a cltie ean occur in a tree structure; ie, the locus of cliticization, and how the clitic attaches to a host; ie. the morphol- cay of cliticization” (1985, 96).She defines clities as phrasal affixes (ie. affixes attaching to phrases and not to lexical heads) that have a domain. This domain is their syntactic host. The notion of a phonological host is not dropped, but this could be different from the syntactic host. Three binary parameters concerning the place of the attach- ment of clities are set up. Pl defines whether a clitic attaches to the initial or final constituent of its syntactic host (Initial Final). P2 determines whether a clitic attaches before or after this con- stituent (Before/After). PS determines where a clitic attaches phonologically, ie. to the left or to the right (Enclitic/Proclitie)." 22.1, The Sumerian possessive element evi- dently attaches toa phrasal category. This has the consequence that it can attach to a range of word classes depending on whether there is something in the modifier position of the NP and what it looks like. It is an allomorph of a full form word, namely the personal pronoun, and is in comple- mentary distribution with NPs. It does not condi- tion an idiosyncratic allomorphy on its host (at least the writing does not reflect it) It does not exhibit multiple exponence, and its meaning can be said to be straightforward. One of the syntac- tic tests proposed by Zwvieky concerns movement According to him “in an X + Y combination if either X or ¥ can be moved without the other, then X and Y are words, neither of them is clitie (1985, 289). (10) above shows that this test defines the possessive element as not an independent word. After these considerations I think itis rea- sonable to conclude that the Sumerian possessive 10. One of the corollaries of these parameters is that a lite, the syntactic domain of which i 4 NP, can, in theory, attach phonologically to a word immediately before the (Initia Before/Eneliti,36 GABOR ZOLYOMI element is a clitic. Zwicky’s classification puts it among the special clitics. Following Klavans' ty: pology, one can claim that it is a Final/After/ Enelitieclitic, which (Klavans’ 7th) is one of the ‘most frequent types as itis very iconic. Its domain is’ 22.2. If the Sumerian possessive element isan enclitic as we concluded, we cannot avoid looking {nto the plural element and the casemarkers. I have already mentioned, following Zwicky and Pullum (1983, 502), that elities can be followed only by clities, but I think other criteria also make it likely that these two elements must also be regarded as clities. The distribution of both is similar to the possessive encliti (ie. they are also phrasal affixes), but some of the casemarkers can also turn up in the verbal prefix-chain just like the plural element, which, in addition, could be a verbal suffix as well. They differ from the pos- sessive enelitic in that they are not in com- plementary distribution with full form words Concerning the full form equivalents, as far as T can see, only the comitative clement is likely to have an equivalent in the Sumerian with which ‘we are familiar. Both the plural marker and the ‘casemarkers can show idiosyncratic allomorphy, and the former and some of the casemarkers (eg, {genitive marker) condition idiosyneratic allomor- phy on their phonological host. The plural marker tnd some of the casemarkers (eg, dative) show se~ ‘mantic idiosynerasy in that they can be attached only to a particular class of words, namely to animate nouns, After Zwicky’s classification, the plural marker is most likely a special clitic, but I ‘would hesitate to put all the casemarkers into this class instead of putting some of them among the bound words. In Klavans’ system their domain is the NP They are different in this respect from the possessive enclitic. The parameters classify them similarly as Final/After/Bnclitic cities. In concl sion, if we suppose a scale between “wordhood’ and “affishood:” I would propose the following ‘order: possessive—plural__marker—casemarkers. That is, the possessive enclitic is the most word like while the easemarkers are the most affix-like, but I would certainly classify all three elements asclitic, ‘Treating these three elements as elite is again not only a matter of terminology. It has bearing on the definition of category of word in Sumerian. In some languages, this definition is an easy task because various (semantic, syntactic, phonologi- cal) criteria all define the same unit, but in lan- guages where clitics play a role, the definition is not that simple (see eg, French). Is lugalanira one word, but lugal galanira two and lugal gal uruakenera three words? Counting the number of the lexemes the answer is yes. But -ani, as we have argued, at least syntactically, has an equal status with ugal, so from this point of view the three expressions contain 2, 3, and again 3 words, respectively: The element -ene in the third ex- pression obviously expresses a category (namely the number) of lugal but not of uru, the same can be stated of -ra. If -ene and -ra were affixes one should conclude lugal gal uruakene-ra is one word. Weitings like lugala-né-ér, at least, imply that from the point of view of phonological eri- terium, a word ends in Sumerian after the last ceasemark.All these problems can be realized and dealt with, although not solved unanimously if ‘one recognizes the clitie status of these three ‘elements 8. Indefinite Genitive 8.1.1. In an unpublished paper, Jagersma pre- sented evidence for the existence of what he called ‘a “compounding” genitive construction (Jagersma 1992). According to him, this construction differs from other genitive constructions in word order and the place of the genitive casemark. In a com- pounding genitive construction, “the genitive post- ‘modifier precedes any attributive adjective” and “the genitive postposition of the genitive post- modifier is attached to the last word belonging to the ‘compounding’ genitive construction, regard less of whether itis the last word of the phrase 1, So,eg, Lean not see any bass for transliterations like lugalgal-urwkavar. Would anybody write the greatking of England's palace? In my view, each lexeme ofa noun phrase, ‘except of lexiealized compounds, should be written separately and only the lis must be hyphened tothe last lexeme ofthe phraseGENITIVE CONSTRUCTIONS IN SUMERIAN 37 or not” (Jagersma 1992, 8). Jagersma claims that “compounding” genitives “express the same kind of meanings as compounds in English, German, or Dutch” Two of his examples follow in (13) and (14): (18) ki stim-ma aSog gibil turka mu-sur he planted them in the onion plot of the Small New Field” (STH 1:52 iti 1-2) (14) sag apin-na dumu-dumurkegne for the chief ploughmen of the children (WAT 4861 vi 1) (15) lypall sag apinak]ype dumudumuakleners) In contrast to Jagersma,!? I would attribute simi- lar properties to constructions (a) Noun nam + Nounak (eg, igi nam-tirka-ni Gudea St C 2:12), (b) Noun GNak (eg. é uru-kirgarka-na Gudea St D 8:17). I shall call Jagersma’s compounding genitive and constructions like (a), (b) non-possessive geni- tive, but later (31.8) anew term will be proposed.” As can be seen in (15), the problem is that in the position of the head of the NP we find a phrase, which according to our grammar, is an NP itself (since being a genitive phrase). But I ‘wonder whether the fact that it looks like an NP ‘means that it behaves like an NP. Let me explain my point using a Hungarian example. In Hungar- ian a genitive construction looks like: (16) a biz ablak-a0 the house window.possrd pers. the window of the house 12. In personal communication (Letter 29/07/82), Jage cersma pointed out that his “compounding genitive” does not (b) His argument is based on ex: “amples where the frst noun of the construction is followed by ‘an adjective but atthe same time there is pronominal encltic after the genitive marker. My opinion about these examples is detailed inn. 16. 1S. Pocbel (1928, $168) describes a genitive construction that is similar to non-possessive genitive. He calls it "be- schreibender Genetiv” and puts it into the same slot with Ad- fectives (se his scheme on p. 85 [§98))..The problem with his ‘description is that in his scheme one cannot put an Adjective after Substantive + beschr Genet, nother words, is sytem fs not reeursive—he could not capture that side of the phe= ‘nomenon that a word level category is brought about (again ‘only bis linguistic framework suily”)So a Subst. + beschr Genetiv is construction that should behave lke a Substan- tive in is scheme. There can be words between the rectum and the regens, but in this case the rectum receives a dative case: (17) a hiia-nak nagy volt az ablak-a® the house dat. ig was the window,poss rd pers, the window of the house was big ‘There are some street and square names in Hun- garian that are genitive constructions: (18) Hésk tered heropl squarepossrd pers. ‘square of heroes It is not possible to say (if Tam speaking about that particular square): (192) “Hés-dkenek szép volta tered hheropldat nice was the square possSrd pers. ‘but (19b) a Hésok tere szép volt ‘The square of heroes was nice So I think the phrase sag apin.na behaves in the same way as Hésok tere in the Hungarian exam ple, Syntactically it is an N, the x in (15) is N, but regarding its origin, it isa genitive phrase. 3.1.2. There exists another way of looking at the problem. One of the questions that can be posed is whether, for example, sa apin.na is a lexicalized phrase listed in the lexicon or if it is built according to the rules of Sumerian genitive constructions. In other words, is sag apin.na or ki ‘stim-ma a compound (ie. a word level category) ‘ora phrase? It is possible not to answer this ques- tion if we suppose that syntactic rules ean pro- duce expressions that are word level categories. ‘An example of this can be found in Turkish. In this language there are two constructions express- ing possession, The construction called indefinite ‘izafet (20a) has the form Noun + Noun-poss (the possessive affix agrees with the possessor in per- son and number), The definite izafet (20b) looks like Noun-gen + Noun-poss. (20a) yatak odasi bed room-its bedroom 14. follow the description by Spencer (1991, 919-19. All my examples come from this book His analysis is based on Lewis (1967).38 GABOR ZOLYOMI (200) wman-n rapor-u expertof reporthis the expert’ report ‘The relation between the elements of an indefi- nite izafet “can't sensibly be called that of ‘pos: session. Rather, the possessive affix simply marks some sort of attributive relation between the head and the modifying noun, a relation which is signaled by simple concatenation in English” (Spencer 1991, 315). The indefinite izafet behaves differently in some respects compared to its defi- nite counterpart. First, itis the indefinite izafet that tends to become lexicalized, Second, it is not possible to modify its head. If one wants to modify the head by using an adjective, this can be done only by forming a definite izafet construction. (@la) Istanbul camiler-t T. mosquesits Istanbul mosques, (21b) Istanbul-un tarihi camiler-i Lof historie mosquesits Istanbul's historic mosques Inan indefinite izafet the “non-head loses its ref= erentiality and becomes simply a modifier of the head, losing at the same time many of its syntac- tic properties. Moreover, the nonchead may only be a word or another indefinite izafet, suggesting that the indefinite izafet itself is a word level cat- egory” (ie. a noun) (Spencer 1991, 318), (22) Tirk Diki Dergist Turk language-isjournal-its [[Turkish Language] Journal] 3.18, As I tried to show in 81.1. above, the Sumerian non-possessive genitives are also word level categories. In Sumerian also, the non-head loses its referentality. In ki stim-ma the word suim does not refer to any onion of the world, it rather describes a kind of field. As far as Lean see, there is no example for a non-possessive genitive where the head is modified by an adjective, that is an expression like *ki gibil simak aSag turak}® ‘meaning something like “the new plot of onion of 15, See also uresagean-na libirraié (Letter Bet 1.7 (Al 1964, 71), the small field” is excluded on principle. If one is to modify only ki, the phrase must look like ki gibil sim a’ag turak.ak, Similarly, phrases like *ki sim gibilak aSag turak “the plot of new ‘onion of the small field” cannot oceur either” Itis possible to find a construction in Sumerian ‘equivalent to the Turkish example (22), (23) 26 tig angeka sumun (tig gt ane} ak] sumun 26 old donkey neck clothes (TSA 31 i 4, quoted by Jagersma [1992 3) The differences again are only due to the reversed order of constituents in Sumerian NP. Non-possessive genitives also can become lexi- calized, One possible way to test it isto count the number of genitive casemarks in a phrase. In (27) below, there is a form @nin-gér-su-kacka, ie. ningirsukakak Sumerian does not write out a third genitive casemark (see Poebel 1923, 136 {§870), so not surprisingly the conclusion is that ningirsuk has already become lexicalized. So, apart from the differences in the morpho- logical marking of the constructions, Turkish and Sumerian show an interesting similarity from the point of view of function, Accordingly, I would suggest a similar terminology in Sumerian, that is, I would call non-possessive genitive indefinite genitive, while the other construction could be named definite genitive construetion.* 16 One could mention two phrases frm the Gude texts that seem tobe counterexamples: a ¢ urarkrga-harna (eg SUD 317, St 115.7), 0) Pddrgornay naman (StE 43-4 StF 38) Inthe case of), Tepid uruk s ot an Noun * Adjective construction Syntacialy it must be a noun, similar to New York Regarding () is posible to argue the ther way around Since dr garmah aceupes slot which, in other compounding geniives, ean be oxeupet only by 2 ‘oun (a word level category) think, itis reasonable to con clude that syntactically dr-gormal should also be eae asanoun Le asa blackboard or égatype compound) Inthe Case of mals ths would ot be exceptional a 1 is demon Stated by Akkadian word lke Kimahu kirmah 17, One can wonder whether the translation of “nin gr sas "the lord of Gia can bejustifed Should this phrase be Interpreted as an indeite genitive then, inmy view could rather mean “Gis Lord” dike “Oxford Universy Pres) the easeaf lida) Se (4) below) the transition ike "the ‘nan of bousbuldig” Caer Mann des Tempe Balen) se Steible 1961, Tl 42, note 15) also seems to be not entire comet. This expression should simply be translated at “theGENITIVE CONSTRUCTIONS IN SUMERIAN 39 4. Anticipatory Genitives 4.1.1. We have already quoted one example of AG. In the case of (9), which is a simple genitive construction, there is only one rectum that can get in front of the whole phrase. If we take into account the double genitive constructions the fol- lowing variations can be assumed:!® i NP; yp, N’lyp, POSS) i NP,NP, lyr, Np, POSSI] fi, NP, bypyN’byry Ny POSSI ik N° NP! iy N’ Ips POSS iii yp, N’ NPal NP yy lyr POSSI iv. NB bye, N’hyn BOSSlive,N'vry POSSI Attestations: i @4) Sningirsurka namnir-galni (Gudea GLB 24:12) += ningirsukak namnirgalant += namningal ningirsukak authority of Ningirsa éaten-kike, gis-burt (Gudea Cyl. A 17:17), += eak NP, ishurbi = gisbur eak fan of the house eavlilk li S-ga-ne-kam (Gudea Cyl AIT) += nlilak Iu Saganiakam = ln kagenlilakakam he isa man of Enlil’s heart [8] ente-merna [4] gives [5] “nin girsirkerka [6 digirra {7] Sul-utulyyramg (Ent 35, 8:3-7) += Entemena giskesradua Ningirsukakak digirani 4 digirE,gikesradua Ningirsukakak si mune-si i (26) hhousebuilder” or “the architect” Not recognizing the differ- ence between a definite and an indefinite genitive can have a ‘mate serious effect on translation: the phrase ia-érin-nakarnt (Gilgamesh and Agga 1.81; 99) s translated by Jacobsen ss “in the midst of his troops" Jacobsen 1987, 359-54). Apparently Jacobsen tok it asa definite genitive although the possessive fenelitie would imply an indefinite one, as itis understood bby Kramer who translated it as “his soldierly hear” (Kramer 1949, 12), 18. Owing to the fact that Sumerian is an extnet lan _gunge ie there does not exist any native speaker of it, we have to suppose tht attested forms are correct and only they are correct Atthe same time Tam fully aware ofthe imitations of this approach but think this sa useful working hypothesis if the texts come from before the OB times ‘The (personal) god of E, the dambuilder of N. (89) gi-dé-a [40] énsi [41] laga’lka [42] fi inim-ni tka (St B 8:39-42) = Gudea ensi lagagakak NP, inimani inimG ensi lagasakak Gudea, ensi of Lagash’s word {1] €a diva mul-kirba {2] gir marraardé (Gudea Cy. A 61-2) = eakdubiak mul kubia = dueakak mul kubia = mul kudueakaka holy star of the building of the house (This construction could be coined as a double an- ticipatory double genitive)! 4.1.2. There is one more construction we should consider. Poebel (1923, 112-13 §307-8), Sollberger (1966, 183) and Hayes (1990, 111; ef. 1991) de- scribe a construction which, according to them, would be an AG, with the structure: % le, NU NP fey N’ burs POSSI (30) gal aruba immrba Sul 52,6) ‘Tog anub.ak immubiak = log immu anuibdakak King ofthe four quarters ‘They think that -bi would agree with an-ub-da. A similar phrase turns up in the NG teats: (81) distita diskus lugal imin-ba (NG 117, 20-21), dickus lugal is a genitive construction as we can see in (92) dickus Ingal-keg Ii-ts-sé-ra bala in-na-an-sum, (NG 113, 37-38) Should (81) have the structure 0, we ought to have di-kus lugal-ka (ie. di-ku lugalakak®”), My 19, An interesting late development of AG is mee erne- ‘am kien (SBH Nr. 58, | 68). Poebel analyses this phrase as [AG Pocbel 1926, 264). Oberhuber apparently has a diferent ‘view saying “zum merkwirdigen Possessivausdruck (Personal pronomen + Besitz + Possessivepronomen) vl typologsch lungar. az én szavam ‘mein Wort” (Oberhuber 1990, 324, v.61 meemu).If he means that Hungarian has the con” struction described by him, then he is wrong, In the Hungar- fan phrase az én szava'm the morpheme -misan agreement ‘marker and not a posessive pronoun, So gig in the Sumer fan phrase isthe equivalent of inthe Hungarian. 20, The same applies to examples like alan &ga 8:ba- kam (instead ofthe expected alan €ga-ka) (DP 58 ix 14 [Nik 23 x17; TSA 1 ix 11). Further examples are éamar-ubary Stim gal abi (TCS 1, Ne 821, 3) and (?) udu lam-lam-ma 101-b1 (FCS 1, Ne.170,4,40 GABOR ZOLYOMI assumption is that the function of -bi in these phrases is similar to its role in the word imin-bi; sibittu “a group of seven’! ie. it is a possessive cenclitie attached to numerals that are not stand- ing attributively to nouns (Thomsen 1984, 88 $142). The construction numerical + pronominal enclitic makes the phrase to which it stands in apposition definite? (81) would mean literally “decision of [judge of the king, seven of them)” that is, “decision of the seven judges of the king” In (80) lémmu is in apposition to an-ub-da. Liter- ally it means “king of [quarter, four of them’ that is “king of the four quarters” Accordingly (30) is neither double nor anticipatory genitive. 4.1.8, Thave tried to avoid making any definite statement about the status of the possessive en- clitie in AGs. I do not think that we need any definite approach to the matter, this is only a technical problem in this context and whatever the outcome, it would not have any bearing on ‘our topic. However, some qualifications are nec- essary before we proceed further. (83) AGs derive from “simple” definite genitive constructions. If we do not accept this assumption, we would have to suppose that NPs can receive the genitive case in a position different from NP specifier. (4) AnAGisa marked construction compared to its simple equivalent! By this I mean that AG has a function that is dif- ferent from its counterpart’ Our main concern in the following will be to ascertain this function of AG. 421. The question to consider is on which level this phenomenon has relevance. One fea- ture shared by all AGs is that the rectum never moves backwards. In all cases the rectum gets in 21. See CAD, p. 230 and also erbettu in CAD E,p. 256, 22 Although Poebel analyses this construction diferent, he fas the same view about its funetion: “dle Verbindung Int... sets determinierende Kraft” (Pocbel 1925, 112 §307) 23. Keenan's definition seems to apply to our ease “a sy tactic structure + is semantically more basic than s syntactic structure yifand only if, the meaning of y depends on that ‘That is, t0 understand the meaning of y i is necessary to understand the meaning of x (Keenan 1976, 807), front of the regens. The key examples are those that have either the structure ia or iia. In these ceases, since the regens is the object of the verb and there is an overt subject in the sentence, itis apparent that the rectum fills a position that is different from both subject and object. It is usual to differentiate syntactically between two sorts of processes by which a constituent can occur in front of a sentence: topicalization and left-dislocation ‘These labels define these processes only in syn- tactic terms while they do not tell anything about their function in a given language. (85) Peter, Lean’ stand. (86) Simon, don’t like him, In the case of the latter, there is a pronoun co- indexed with the NP in the sentence. I suggest that we should treat AG as a left-dislocation [= LD] phenomenon. My assumption is that its function in Sumerian is to move the rectum into the topic position of a sentence® Before we go any fur- ther, a short excursus is needed in order to ck: what is meant by topic. jonally, the topic is the constituent de- noting what the sentence is about, ie, in the logi- cal sense, itis the subject of the sentence. This logical interpretation of the term subject, going back to Aristotle, is also called “notional subject’ Although the topic, or notional subject, is very often identical with the grammatical subject, this is not necessarily so. Sentences that have a gram- matical subject can be subjectless in the logical sense Based on the difference between the no- 24 Should the function of AGS be to move a con {nto the opi position ofa sentence, then the lack af tru Tike e has an easy explanation. In constructions like rectum cannot reach the sentence intial postion, which i the landing ste aimed at by AG, since it does not even move out from the NP It must be noted that if we were nat o treat the constituents inthe genitive ease standing in front ofthe sen- tence as the result of some kind of derivation, it should lead to claiming that in Sumerian a topic marker exists, its form being ak 25. A sentence being subjeetless in the logial sense is eg It is raining Li and Thompson give a list of those features In which subject and topie ean be sid tobe different (1976, 461-86), Some of their observations: a. Topics are always de finite, subjects do not need to be definite Topi “need not hhave a selectional relation with any verb in a sentence; that i ventGENITIVE CONSTRUCTIONS IN SUMERIAN, 4 tional and grammatical subject, it is possible to differentiate between semantic and syntactic pred- ication. There are languages in which it is the semantie predication that determines the syntac- tie structure of a sentence (ie. the word order) In these “discourse-configurational” languages “the (iscourse-) semantic topic} serving to foreground a specific individual that something will be pred- ‘cated about, is expressed through a particular structural relation (in other words, itis associated with a particular structural position)” (E. Kiss 1995, 6) In the following I shall try to demonstrate through examples from Sumerian texts that AG really functions as a device to move a constituent {nto topie position, and that the rectum occupying the topic position can be interpreted as a “no- tional subject” In the case of AG we have a mor- phologically readily discernible syntactic structure that helps us to examine it, but should the rectum of AG function as a topic then it is likely that other, not easily perceptible structures, also ful- fill the same function. As is clear from the defini- tion of topic, the “notional subject” is basically a foregrounded constituent. Foregrounding is also an important concept in the case of the passive. The passive basically foregrounds the undergoer of an action and/or moves the previous actor to the background?" The passive in Sumerian is one of the perplexing problems of grammar that has eluded a coherent description so far. If Sumerian it need not be an angument of a predictive constituent” “Topicpredicate agreements vey rare and we know of no language in which it is widespread or obligatory.” Al. though the surface cading of the topic may involve sentence position as wel as morphological markers, tis worth noting that the surface coding f the topic in all languages we have ‘examined always involves the sentence inital position” £ “The subject but not the topic plays a prominent role in such pro- cesses are reflexivizaton,pasivzation, Eque-NP deletion, verb serialization, and imperativization” 26, It seems typical to me that earlier, when attention was paid solely to syntactic predcation, Hungarian, a discourse ‘configurational language, was characterized as a free word order language. Cf. Thompsen (1984, 51 §48): “The order of| the various nominal chains fergative, dative, terminative ete) Inthe sentence i, however, rather fre, but the verb is always atthe end ofthe sentence 27, See Foley and van Valin, J (1985) for describing the passive in these terms. proves to be a language in which topic plays a role, one should describe it in a completely differ- ent way compared to passives in Indo-European languages 4.2.2 Consider first the following passage from Gudea St B: (87) 18:39] gi-dé-a [40] énsi [41] lagad!ka [49] 1 inim-ni tb-kir-a [43 di ku. Vib-bal-ea [4d] ance... [9:4] digituio nin i8-2i-da-keg [5] nam-tarraeni bédabrkir-ne [6] gueginz (7) upnena bé-gae 18] am-ginz [9] & bus-na bédabs Gudea, ensi of Lagash (topic), the man who changes his decision, and alters his judgment (topic), let An,..., my personal god Ningish- zida change his fate; like an ox let him be Killed immediately and like a bull let him be seized by his terrible arms* (Gudea St B 89-99) In this thematic unit there are two AGs. Lines 39-41 are the rectum of inim-ni. Lines 89-48 are the rectum of nam-tar-ra-ni. In both eases the rectums are in the topic position of the clause and the sentence respectively. In both cases the rectums stand in front of the subject. In the rela- tive clauses Gudea is the topic, and although in the main sentence the grammatical subjects are the various gods, itis the offending person that is the real theme of this unit. Of course up to this point this has been a matter of interpretation, but 28, The following examples from Brown and Yule (1988, 127) demonstrate clearly that foregrounding of constituents and the proper interpretation of the various foregrounding devices in a language do conteibute alot to the meaning, AI though examples a-g express the same content, the meaning ‘ofthe sentences is rather different, 1 John kissed Mary bh Mary was hissed by Jon « Ttwas John who hissed Mary 4. Teas Mary who was kissed by John ‘© What John did was kiss Mary. Who John kissed was Mary. , Mary, Job kissed her. 20. am avware that this and the following translations sound clumsy or are even not entitely correct in English but English being a Sp language makes it very difficult to give back topic-comment structure. My intention was to make ap parent shih constituent I think tobe the top Li and Thomp: son (1976) however emphasize that in languages where topic does play a role these sentences are not marked or special in any respect.42 GABOR ZOLYOMI there are other points in favor of this. Li and Thompson list among the characteristics of ‘Tp languages that “the topic and not the subject, typ~ ically controls co-referential constituent deletion” (Li and Thompson 1976, 469-70), Their example is from Mandarin: (88) Nei kai tin dai zg hen da suSyi__hén zhigidn that piece land rice grow very big so__very valuable that piece of land (topic), rice grows very big, 0 it (the land) is very valuable In this example, the deleted constituent is the NP in the topic position and not the subject. Turning to our example, in my opinion, we find the same phenomenon in the case of II, 9:6-9. In the previ- ous sentence the offending person is neither the subject nor the object, but a topic. In both sen- tences of Il. 9:6-9, this topic will be the subject The two verbs have a passive meaning marked by both the agreement (a zero suffix in this ease) and the ham(u root (see Edzard 1971, 214). The same phenomenon can be described with a slightly dif- ferent terminology. Ina paper written eight years after Li and Thompson's, Givén introduces the term “thematic paragraph” The thematic para- graph is“a chain of equi-topie clauses, ie.a string of clauses whose main/primary topic remains the same...” (Givén 1983, 9). He speaks about the “de- gree of topic accessibility” by which he actually ‘means the “identifiability” of a topic in the differ- cent parts of thematic paragraph. The topic is ‘more identifiable or continuous in the middle or at the end of a thematic paragraph and accordingly a scale can be set up regarding the phonological size of a constituent functioning as topic. The scale looks like: (89) zero anaphora - unstressed/bound pronouns (Cagreement”)- - stressed/independent pronouns = full NPs Zero anaphora is the most continuous topic; full NPs usually stand at the beginning of a thematic paragraph, In the case of our example in Il, 6-9 a verbal agreement identifies the topic, which, in this case coincides with the subject. The chosen coding device tells us that the topic is easily iden- tifiable. I wonder whether this would be the case if the topic of the previous sentence were different 4.2.8, Consider the following example: (40) [22] &ba-bag [23] ki-bé gigarda (24) De-Balbi [1] part AK-da [2] Bddr-darlaga’lka [3] suhus-bi gina-da (4) gairdéa 5] énst {ol lagas-ka [7] gid inim gina [8] urna galada [9] namet-la-na NO} wg i staeda 11] digieraeni [12] ‘nin gis-z-da [13] ba-bay [14] éuruki-ga-ckana (15] mu-nardarkuy kg (LE 7:22-8:15) ‘The house of B.having been restored: its abun- dance having been made apparent; the throne cof Lagash (topic) its base being frm;G.ensi of Lagash (topic), the scepter of firm word being held in his hands is topic) life, its days being rade lasting; his (opie) personal god N. was ‘going to B into her Eurukuga, What we have in this excerpt of St E is a chain of equi-topic clauses. This chain starts from 1. 8:4 and lasts until the end of the passage. First, the topic is a full NP (I 8:4-6). In the subsequent clauses Gudea remains the topic but is referred to by a bound pronoun. So, in L 8:9 nam-tila-na ‘moves in front of the sentence because the pos- sessive enclitie refers to Gudea; strictly speaking it is the possessive enclitie referring to Gudea that moves, but being a clitic, it cannot move without its host (the same would apply to Il. 13 14 in [41]). The problem of digir-ra-ni*nin-gid-zi- da (which contains two constituents standing in apposition) in I. 8:11-12 is more complicated, In the majority of cases this phrase occurs in a re- versed order (St I 8:7-8; St P 8:8-9; St Q 1:1-2; Gudea 64, 1-2; 65, 1-2; 66 Betschrift 1:1-2; 67 1-2; 68, 1-2; Cyl A 18:15; St B 8-4-5). With the exception of the last two occurrences, in these examples, ¢nin-git-zéda digir-ra-ni is construed with a -na infix in the verbal prefix-chain and occupies the beginning of either the whole text or 80, Consider the following examples from English a Julia has helped Marys mother, 8 Shey ge; must be grateful fia As for Mary, Julia has helped her, mother, fill She, must be gratefulGENITIVE CONSTRUCTIONS IN SUMERIAN, 43 a clause. It occurs in the same order three more times, in St B 9:4 as digir2uyo ‘nin Bis-zi-da-key, in Cyl A 5:20 as digir-zunin-git-zi-darkey, and in St G 2:8-9 as digirra-ni Snin-gis-zi-da, Below in (62) I will treat the so-called “hierarchy of sali- ence? which has something to do with the inher- ent information status of NPs Speaker/addressee, as the most salient constituent, can be found on the top of this hierarchy. It therefore seems prob- able that in St B 9:4 and Cyl A 5:20 this hierarchy is responsible for the order. In the case of our text and St G 2:8-9 this solution does not work. A more likely explanation would rely upon the con- textual salience of the person to whom the pro- noun refers St G 2:8-9 does not exclude such an interpretation,” and in the case of our text this would seem natural since Gudea occupies the topic position of the two clauses standing before Wsil-14 In the next example, the 3rd person possessive cenclities refer to the topic of the first sentence. ‘The topic remains the same throughout the the- matic unit. (41) 5] fi Eanrnarta [6] sb-twabeea [7] fb-ze-re-a [8| musara-ba Su bi [9] inanna {10} nin kur-kurrarkey [11] sag-@é-ni ukkin-na (12) nam hé-ma-kusre [3] gu-za gub-ba-na [14 suhus-bi (15) na-an-gine (16] numun-ani het [17] bal-ani bé-kus, (stC45-17) ‘The man who removes it [= the statue] from Eanna, who tears it out, who erases its inserip- tion (topic), let Inanna, lady of all the lands ‘curse his head in the assembly, his (topic) erected throne, she should not make firm its base, Let his (topic) seed come to an end, Let his (opie) reign be cut off 4.24, The word referring to the Eninnu in the Gudea texts (6 &ninnu, 2 éninnu) tends to move 431, See Falkenstin (1968, 102 [about St G 28-8) “Bei der Ubergabe der ‘Hochzeitsgnben’ an Bab, die Ningirsu selbst dberbrachte, eat sein (d Gudeas) Got dahinter of Fensichtich um dabei als Vertreter des Stadtfirsten daruber ‘zu wachen, das sich der Akt in der rechten Weise abspele™ in front of the sentence when iti the rectum of a genitive phrase (42) [4] €a gis-burbi imgees (Cyl A 5:4) ‘The house (topic), the) was setting down its plan. (43) [5] Snin-DUBskam &agié-bur-ba immi-sese- ge (Ov A 65) Tt was N, The house (topic), (he) was making its plan (44) {10} € ga us ki garra-bida (11) bé-gAl besdargen (Gudea Cyl 4 11:10-11) My temple (topic), may with the laying ofits foundation abundance come. (45) én Semikeg gi-burbi si murnesé (Cyl A 177) ‘The house (topic), Enki prepared its plan for him, (46) [14] a ni-galbi (25) kalam-ma mi (16) katar-ra-bi (17) kurre ba-ti [18] é-ninnu ni-bi kurkur-ra tog: ging im-duly (CyLA 2914-18) ‘The house (topic), its great awesomeness set- tles upon the country: Is glory reaches the highland, Eninnw (topo) its awesomeness like garment covers all the lands. (47) L) ursag ningie-su éa mint-kuckuy [2] €a lugal-bi immagen (Cyl B5:1-2) Warrior N. entered the house. The house (topic) its king has arrived. (48) [18] 28 éninmurna dirba za-ra me-ra-an-duy (CyLA 5:18) His shrine, Eninnu (topic he ordered me to build it up. (49) (17) €gé ni gak-bi kur-kur-ra muti (Cyl A 9:17) My house (topic), its great awe was placed ‘upon the country: (50) [9] € lugal-na-ka da-bi [10] éninnu an kita Dad-bi [11] igia mu-na-a gal (Cyl A 209-11) His king’ house (topic), its building; Eninnu (topic) its separation of heaven from earth appeared for him in the vision 82, Ttake it as a form deriving from sd—sb-sé» summuru “to lot plan” This would explain the loative af github See Sjoberg (1969, 103-4),44 GABOR ZOLYOMI See also (29) above. One apparent exception of this tendency is (81) [4] enna 8 an-giny si-ni [5] Snin-gir-su dumu Seni ka zarra marae yur Gana maracphd-pad-de (Cyl A Lord (top), his, N. son of E's heart which is as inconceivable as the heavens, will be appeased for you. He will reveal the plan of his house for you. ‘The topic of this short unit isthe lord Ningirsu as it can be seen in I, 4. Ningirsu remains the topic in 1.6 as well, and as a result ¢ in 1. 6 is not left- dislocated, One can wonder why the assumption, that a possessive enclitie would be able to make its host move before a sentence, does not seem to apply here. The important difference between the examples of 4.2.3, and (51) is that there, in 1.9-10 of (40) and Il 18-14 of (41), the topic and the subject are different. However, in case of (51), the topie and the subject are different in the first sentence (Il 4-5) but coincide in the second cone (1.6) 4.25. Another notion which seems to be appli- cable to our subject is the information status of NPs or constituents (Foley and van Valin, Jr. 1985, 283-91), The more prominent a constituent is, the ‘more likely that it will be the topic of a sentence. The information status of NPs depends on con- textual factors and on the inherent properties of their referents. An NP can be referential or non- referential, definite or indefinite, it can refer to “a participant being introduced into the discourse (new information) or to a participant already established in the discourse (given information) (Foley and van Valin, Jr. 1985, 286). As far as the inherent properties are concerned, a hierarchy of salience can be established: (52) speaker/addressee > Srd person pronouns > human proper nouns > human common nouns > other animate nouns > inanimate Givén approaches the same phenomenon in a slightly different way. He speaks about “the uni- versal hierarchy of topicality, i, the likelihood of various NP arguments being the topic of sentences...” (Givin 1975, 152). His hierarchy actually takes on the form of binary hierarchic relations: (68) a human > non-human, b. definite > indefinite more involved participant > less involved participant 4d. Ist person > 2nd person > Srd person From these considerations, only the definite vs. indefinite feature could cause some problem, since there is no article in Sumerian, Taking into ac- count the hierarchies above, the behavior of ¢ in ‘examples (42)-(50) seems to be interpretable. In all the examples we can assume that is the “more involved participant” or “more given” com- pared to its regens since the text itself describes the building of Eninnu, In case of (46) and (49) the regens is an abstract noun and itis likely that in (48) and (50) the building of the house (dit) takes a position that is lower down in the hier~ archy than the house. As can be seen in the case of (51), other prineiples can interfere with these tendencies. Similar considerations seem to apply to the following examples: (64) [18] dea [19] 1 é ddvarka [1] nam-tita-ni hes [2] mus® murnasay (StC3:18-4:2) G, the housebuilder (topic), let his life be ong-he named it (the statue) for her® (55) (7 atm ciginrenea [28] en Sine (29) nam-mabya-ni [90] kalam-e bé-zu-zu (StB9:27-30) The steadfast of gods, lord N. (topic), let his greatness be known by the lan. (66) [12] ¢nin-ir-sucka nam-nirgéleni {13} kurkur-e zura (Cyl B 24:12-18) Ns authority (topic), being known by all the countries, See also (87) (Gudea vs. inim). 4.2.6. In examining the AGs according to the role their regens play, one finds that there does not seem to be any restriction’ Most often it is 38, Also StP 53-7, See further St15:8-7;St ASs4-43 34. According to Li and Thompson (1976, 470): “In ce tain Sp languages, the topic-comment type of sentence isGENITIVE CONSTRUCTIONS IN SUMERIAN, 45 the object of the verb, Nor are there any differ- ences with respect to subjects of transitive or in- transitive verbs. There are some cases when the regens is in a case other than ergative or absolu- tive, In (44) itis construed with -da; in St E 8:6-8 (41) and in (48) with locative. Another example of a regens in locative is the following: (57) Il] urwna G-KA*UDNI 2-bia muda-e-nd- am (Cyl B 18:1) His city (topie), the unclean?) slept in its outskrt, 42.7. If the rectum contains several NPs, being in apposition, there are some variations regarding the AG. In (51) (repeated here for the sake of con- venience as [58)), parts of the original constituent are separated by the regens. (55) [4] en-na shamgin; sini 15] Sninegirsu dum fenviL-46-ka2a-ra marae 74-6) 69) dumu “eni-4é en ‘nin-gir-su (Cyl A821;93) ‘Taking, (59) as the initial form the following deri- vation can be imagined: (60) a regens [a + [b+ cl}-gen. Bangin, si-ra [dumu em-illé + fen + ain Bir-sul}-gen, (be regens [fb + e+ all-gen. Degen. regens e+ al gen, Using this notation, (55) would look like: (61) a regens a + [b + cllgen. nnam-mab (gabargal digir-rene + [en + Snin- Sir-sull-gen. bo agen. [b + clgen.regens To ascertain the rules, if there are any, that gov ‘em these constructions, one needs more exam- ples. I would venture the assumption that the highly constrained in terms of what can serve as the topic con- stituent” They mention Indonesian, in which only the subject tnd its rectum can be the topic. quote their example because ofthe striking similarity with Sumerian ‘Anak ity, iburnja- membelt sepata child that, mather-poss buy shoe “That child, his mother bought shoes reason in (61b) each of the two phrases in front of the regens receives a case is that they are treated as distinct constituents, ie. that they are ifferent slots in front of the sentence. In (55) the whole object phrase is before the subject. As is clear from numerous votive inscriptions where the beneficiary stands at the beginning the sen- tence, AG must be only one of those devices that works on discourse level in Sumerian. Therefore, the existence of other topicalization rules cannot be ruled out. Accepting this, one possible expla- nation of the structure of (55) could be that the object is moved before the subject as a result of topicalization, and in a second step, AG comes about because the rectum is more salient (in terms of those brought up in 4.2.5). Again one can only guess about the nature of the various intonation patterns accompanying these rules, But if it were really the case that gaba-gdl digir-re-ne-ka and en ‘nin-gir-surka are viewed as two distinet con- stituents, that would entail the assumption that a sentence can have more than one topie®* Then, in (62) the subject could be perceived as moved out from its original position in front of the object. (62) [5] Snin-tu [6] ama digirrene-key [7] girdéa [I] la € divarka [2] nam-tterni [8] mu-ke murna-say (SLAS4-43) N,,mother of all gods (?),G, the man of house building (topic), she made his life lasting— hhe named it the statue) for her. This explanation does not sound so forced if we Took into (63), where part of the subject is placed in front of the rectum of a genitive construction. (63) [26] lugal mu-ni-s kur KUKU-e [27] givdé-a en nin-gir-suckey [27] gu-za-ni murgi (Cyl A 28:26-27) ‘The king, at the name of whom the moun- tains tremble (topic), Gudea (topic), lord N. made his throne firm. 85, This is not exceptional Kiss (1987 assumes a sen tence initia slot Tor topic in Hungarian. Ths slot can Be oe pied by more than one constitaent.E. Kis (1995, 11) names Catalan, Grek, Quetcha, and Somat in adition to Hangar fan as languages in which there can be multiple topes.46 GABOR ZOLYOMI 5. Conclusions First, we tried to apply the X-bar theory of recent generative grammar to Sumerian noun phrases and generative constructions. This ap- plication had the following results: (i) Sumerian pronominal suffixes are in complementary dis- tribution with noun phrases, ie. syntactically they should be regarded as words; (ii) in contrast to earlier views, the genitive marker shares its rank with other cases. As a consequence of (i) it be- came clear that pronominal suffixes are in fact enelities. The same was proposed about the plural marker and the various casemarkers, It was also claimed that problems concerning the definition ‘of word in Sumerian can be realized and dealt with, although not solved unanimously, if one ree- ‘ognizes the clitic status of these three elements. Taking Jagersma’s “compounding genitive” as a starting point, we ascertained the existence of two basic types of genitive constructions in Sumerian. ‘We proposed to call these two types definite and indefinite genetive respectively. The terms come from Turkish grammar, and our proposal was based on formal and functional similarities. From the point of view of formal characterization, the similarity lies in that indefinite genitives result in a word-level category (N) in both languages as op- posed to definite genitives, which are phrases (NP). Regarding the function, indefinite genitives are used to express a sort of attribution, while defi- nite genitives mark possession in both languages. In the last part of the paper we treated antici- patory genitives that are variations of definite genitives. We claimed that, formally, AG is a left dislocation. The function of AG was identified as topicalization. That is, the rectum of AG becomes the topic or one of the topies of the sentence, We tried to prove that Sumerian seems to be sensi- tive to the information status of constituents and shows properties that characterize those languages that are recently called discourse-configurational languages. References. AWEA 1964 Sumerian Letters: Tico Collections from the (Old Babylonian Schools. PRD. diss, Univer sity of Pennsylvania, Bickerton, D. 1990 Language and Species. Chicago: University of Chicago. Brown, G,and G Yule 1983. Discourse Analysis. Cambridge Textbooks in Linguistics. Cambridge: Cambridge University Edzard, D.O, 1971. hamtu, mardi und freie Reduplikation beim sumerischen Verbum, ZA 61: 208-32 B Kiss, 1987 Configurationality in Hungarian. Studies to Natural Language and Linguistic Theory, 3 Budapest: Akadémiai Kind 1995. Discourse-Configurational Languages. Intro- duction, In. Discourse Configurational Lan- guages, ed. KE Kiss, Oxford Series in Comparative Syntax. Pp. 3-27. Oxford: Oxford University in, A ‘Das Sumerische. Handbuch der Orientalistik, Abt, 2.Bd, 1/2 Abschn, Lief. 1 Leiden: Brill 1966 Die Inschriften Gudeas von Lagas, L Ana- lecta Orientalia, 80. Rome: Pontifcium Inst- tutum Biblicum Foley, W.A and R.D.van Valin, J 1985. Information Packaging in the Clause. In Lan- guage Typology and Syntactic Description, Ill ed, T Shopen. Pp, 282-864, Cambridge: Cambridge University. Giv6n,T. 1975 Topic, Pronoun, and Grammatical Agreement In Subject and Topic, ed. C. N. Li Pp. 251-85. New York: Academic 1983 Topic Continuity in Discourse: An Introdue- tion. In Topic Continuity in Discourse: A Quan- titative Cross Language Study, ed. T. Givén ‘Typological Studies in Language. Pp. 1-41 Amsterdam: Benjamins Haegeman, I. 1991 Introduction to the Government and Binding Theory Oxford: Blackwell. Hayes, LL. 1990 A Manual of Sumerian Grammar and Texts. ‘ARTANES, 5, Malibu: Undena, 1991 Some Thoughts on the Sumerian Ge ‘AS 13: 185-94GENITIVE CONSTRUCTIONS IN SUMERIAN 7 Jacobsen, T. 1973 Notes on Sumerian Genitive. JNES32: 161-66. 1987 The Harps that Once. ..: Sumerian Poetry in Translation, New Haven: Yale University Jagersma, B. ‘unpub, Some Observation on Word Order in Noun Phrases. Paper presented atthe third meeting of the Sumerian Grammar Discussion Group, Oxford, 1-3 uly 1992 Keenan, EL. 1976 Towards a Universal Definition of “Subject” In Subject and Topic, ed. C.N. Li. Pp, 308-89, ‘New York: Academic. Klavans, JL. 1985 ‘The Independence of Syntax and Phonology in Clitieization, Language 61: 95-120. Kramer, S.N. 1949" Gilgamesh and Aggn. AJA 54: 1-18. Lewis. 1967 Turkish Grammar. Oxford: Oxford University LCN, ed. 1976" Subject and Topic. New York: Academic. Li,C.N,and S.A Thompson 1976 Subject and Topic: A New Typology of Lan- uage. In Subject and Topic, ed. C.N. Li. Pp. 457-89, New York: Academic. Mel?cuk, L 1973. 'On the Possessive Forms of Hungarian (oun. In Generative Grammar in Europe, ed. F Kiefer and N. Ruwet. Reidel: Dordrecht Oberhuber, K 1990 Innsbrucker Sumerisches Lexikon, I. Sumerisches Lexikon =u “George Reisner, ‘Sumerisch—babylonische Hymnen nach Ton- tafeln griechischer Zeit (Berlin 1896)" und 2u verwandten Texten. Innsbrucker Beitrige zur Kulturwissenschaft, Sondetheft 70. Inns- Druck: Institut fiir Sprachwissenschaft der Universitit Poebel, A. 1923 Grundstige der sumerischen Grammatik Rostocker orientalistische Studien, 1. Rostock: Selbstverlag 1926 Sumerische Untersuchungen IL ZA 27: 161- 76, 245-72. Radford, A 1988 Transformational Grammar. A First Course. Cambridge Textbooks in. Linguistics. Cam- bridge: Cambridge University. Shopen, Ted 1985) Language Typology and Syntactic: Descrip tion, III1 Cambridge: Cambridge University. Sjoberg, AW. and E. Bergmann 1969" The Collection of Sumerian Temple Hymns Texts from Cuneiform Sources, 8. Locust Valley, NY: Augustin, Sollberger, E 1966 The Business and Administrative Correspon- dence under the Kings of Ur. Texts from Cuneiform Sources, 1. Locust Valley, NY: Augustin, Spencer, A. 1991 Morphological Theory. An Introduction to Word Structure in Generative Grammar. Oxford: Blackwell Steible,H 1991 Die neusumerischen Bau- und Weihin- schriften, IIL Freiburger Altorientalische Studien, 9, HI Stutigat: Steiner Seabolesi, A 1981 The Possessive Construction in Hungarian: A Configurational Category in a Non-Configura- tional Language. Acta Linguistica Academiae Scientiarum Hungaricae 31: 261-88. 1983. The Possessor that Ran Away from Home. The Linguistic Review 8: 89-102. ‘Thomsen, M-L, 1984 The Sumerian Language. An Introduction to its History and Grammatical Structure. Mesopotamia, 0, Kobenhagen: Akademisk. Yoshikawa, M 1992 Sumerian Genitival Construction in Terms of "Status Constructus” ASJ 14: 408-6, ‘Dwicky, AM. 1977 On Clitcs. Bloomington: Indiana University Linguistics Club 1985. Clitics and Particles, Language 61: 283-305, ‘Zowvieky, A.M. and G.K. Pullum 1983" Clitcization vs. Inflection: English n°t Lan- ‘guage 59: 502-18,