<<

Grammatical formatives in a generative lexical theory The case of Modern Greek και*

†Robert Ingria Το και ενώνει λέξεις και προτάσεις με πολλαπλή σχέση μεταξύ τους “και joins words and with varied types of relations between them”. — Τριανταφυλλίδης, Νεοελληνική Γραμματική Πολλές φορές η σύνδεσις δυο προτάσεων με το και είναι εντελώς ιδιόρρυθμη “Many times the of two clauses with και is quite idiosyncratic”. — Τζάρτζανος, Νεοελληνική Σύνταξις

Work within the framework of the Generative Lexicon (GL) has, to this point, concentrated on the lexical categories: , verbs, and adjectives. Serious research on the non-lexical (minor) categories has not yet been undertaken. In this paper, I examine the polysemous behavior of Modern Greek και (ke “and”) and examine the ways in which GL theory can be used to account for its behavior, as well as the limitations of GL theory in treat- ing και. Since this paper raises as many questions as it answers with regard to some of the more striking uses of και, it should be considered a report on work in progress.

Keywords: connectives, underspecification, generative lexical theory, complementation

. Rationale for this research

Until recently, work in lexical semantics had concentrated almost exclusively on verbs. Beginning with the initial efforts within Generative Lexical (hence- forth GL) theory (e.g. Pustejovsky and Anick 1988, Pustejovsky 1989), research has extended to the full range of lexical categories (in the sense of Chomsky 1965:74): nouns, verbs, and adjectives. However, the remaining minor cat- egories (or grammatical formatives) have remained virtually unexamined. Aside from Ginzburg and Kolliakou (1995), who utilize mechanisms from GL

Journal of Greek Linguistics 6 (2005), 6–0. Downloaded from Brill.com10/01/2021 01:32:53AM via free access issn 1566–5844 / e-issn 1569–9856 © John Benjamins Publishing Company 62 Robert Ingria

to analyze selection in Modern Greek, there has been little, if any, work on minor categories, within the framework of GL or any other. In this paper, I present an initial examination of the behavior of the Modern Greek conjunction και (ke “and”). This study represents another effort to ex- tend the framework of GL to the analysis of minor categories. Aside from the programmatic goal of extending the reach of GL to minor categories, there are several other reasons for undertaking this research. και is polysemous in varied and interesting ways, making it an excellent candidate for such analysis. Moreover, there are pre-existing paradigms of the behavior of και, in particular, Τζάρτζανος (Tzartzanos 1946), Mackridge (1987), and Canakis (1996), which are available for re-analysis within the framework of GL. Moreover, previous studies of και were essentially taxonomic and did not ad- dress the issue of specifying the mapping from linguistic form to meaning rep- resentation, nor of separating out the contributions of sentential /seman- tics, discourse context, and pragmatics. Finally, the polysemous behavior of και is open-ended and ubiquitous, as we can see in the following example, which is drawn from a naturally occurring EMail text discussing the performance of a well-known actor in a new role: [JMS is the writer of the television series].1 — Tromeros XARAKTIRAS, yes — Tromeros ITHOPOIOS, no! terrific character, yes terrific actor, no 30 xronia sto epaggelma, veltiosi miden! 30 years in-the profession, improvement zero Eftixos o,ti kai na kanei ton sozei to senario. (Megale JMS!) Lucky whatever and SUBJ he-does him it-saves the script Great JMS Kai toy veltiosan kai to peroykini!!!! And to-him it-betters and the wig — A terrific CHARACTER, yes — A terrific ACTOR, no! 30 years in his profession, no improvement! He’s lucky that no matter what he does the script saves him. (JMS rules!) And they’ve even improved his wig! NB: this example originally appeared in ASCII EMail text, exactly as it appears here, so there is no corresponding character transcription. Of the three occurrences of και in this passage, only one, the second, is a use as a simple conjunction. The first occurrence is used to reinforce the open- ended character of the free relative while the third is an intensifier use. The frequency of the phenomenon, coupled with the fact that different uses have slightly different senses, adapted to the current context, cries out for a genera- tive treatment.

Downloaded from Brill.com10/01/2021 01:32:53AM via free access Grammatical formatives in a generative lexical theory 63

2. Some Caveats

One objection to this line of research might be that grammatical formatives do not strictly speaking have lexical semantics, but are more properly treated as part of syntax. While this is undoubtedly true, it is equally true that the mecha- nisms of lexical semantics will almost certainly not be disjoint from those of compositional semantics. Moreover, given the existence of GL mechanisms such as co-composition and coercion, which blur the strict line between lexi- cal and compositional semantics, there is much less of a distinction between the two. As we will see below, coercion plays a major role in some of the con- structions involving και, demonstrating the utility of analyzing it from a GL perspective. There are some phenomena related to conjunction that will not be treated here. Among these are the collective–distributive distinction (as discussed by e.g. Scha and Stallard 1988) and the issue of generalized quantifiers (Barwise and Cooper 1981). These are important aspects of conjunction, but this study is focused on the wider range of uses of και, in particular, those in which it has senses quite different from that of simple conjunction.

3. Some Preliminaries

Note that και also appears as κι (ki) before vowels and can elide to κ’ (k). All these forms will appear in the examples that follow; they represent standard phonetic and orthographic variants of και and do not entail any syntactic or semantic differences. One important feature of Greek morpho-syntax that interacts with the ef- fects of και is the tense-aspect and mood system. For the purposes of this paper, we may treat the relevant facts as follows. (For fuller discussions, see Warbur- ton 1970, Setatos 1983, Veloudis and Philipaki-Warburton 1983, Philippaki- Warburton and Veloudis 1984, Veloudis 1984, Veloudis 1985, Mackridge 1987, Joseph and Philippaki-Warburton 1987, among others.) – Modern Greek does not typically indicate mood by morphological inflec- tion, but through the use of “modal particles”. The two most important modal particles, which will appear in example sentences here, are: θα (θα) which is a future marker; and να (na) which indicates the subjunctive.

Downloaded from Brill.com10/01/2021 01:32:53AM via free access 64 Robert Ingria

– Modern Greek verbs may be viewed as having two stems, a present stem, used to form the present and imperfect tenses, and a past stem, used to form the past tense and past perfect participle. e.g. γράφω (γrafo) “I write” [present] έγραφα (eγrafa) “I was writing” [imperfect] έγραψα (eγrapsa) “I wrote” [aorist/past] έχω γράψει (exo γrapsi) “I have written” [present perfect] είχα γράψει (ixa γrapsi) “I had written” [past perfect] – Whereas, in the indicative (unmarked) mood, the present and past stems can contain a temporal interpretation, when they follow a modal particle, or appear in an imperative form, they are interpreted aspectually, with the present stem receiving a habitual or iterative interpretation and the past stem receiving a punctual interpretation. e.g. θα γράφω καλά βιβλία θa γrafo kala vivlia “I will write good books” (habitually) θα γράψω ένα καλό βιβλίο θa γrapso ena kalo vivlio “I will write a good book” (once) θέλω να γράφω καλά βιβλία θelo na γrafo kala vivlia “I want to write good books” (habitually) θέλω να γράψω ένα καλό βιβλίο θelo na γrapso ena kalo vivlio “I want to write a good book” (once) Modern Greek lacks an infinitive or other non-finite form that can appear in complement position. Because of this, the Modern Greek equivalents of verbs that take infinitival complements in English and other European languages, take subjunctive complement clauses, marked by να and with a null comple- mentìzer. Example sentences that are drawn from pre-existing reference grammars and papers will have their sources indicated by the following key: Ca = Catone (1967) CC = Canakis (1996) HKK = Householder, Kazazis, and Koutsoudas (1964)

Downloaded from Brill.com10/01/2021 01:32:53AM via free access Grammatical formatives in a generative lexical theory 65

JPW = Joseph and Philippaki-Warburton (1987) M = Mackrídge (1987) T = Thumb (1964) Τρ = Τριανταφυλλίδης (1941) Τζ = Τζάρτζανος (1946)2 The remainder of this paper will be devoted to an analysis of the range of uses of και in Modern Greek. I will begin by analyzing its simple conjunction and intensifying uses, including its multi-functional behavior as a clausal conjunc- tion. Next, I will turn to a more extended use of και: its conditional use. Finally, I will turn to the most interesting and most difficult data, complement uses of και.

4. Basic and Extended Conjunction Uses

Like “and” in English, και can be used as a simple conjunction, conjoining two or more constituents of the same type. When there are more than two conjuncts, it may appear only before the last conjunct (1), or it may appear between each pair of conjuncts (2). Moreover, και can function as its own cor- relative conjunction; i.e. και… και functions much like “both … and …” in English, except that the correlative use of και is not limited to the conjunction of two elements (3). (1) ο Γιάννης, ο Νίκος κι ο Πέτρος γέλασαν o γianis o nikos ki o petros γelasan the John, the Nick, and the Peter they-laughed “John, Nick, and Peter laughed” (= JPW, §1.3.1.2, p. 60, 8b) (2) ο Γιάννης κι ο Νίκος κι ο Πέτρος γέλασαν o γianis ki o nikos ki o petros γelasan the John and the Nick and the Peter they-laughed “John and Nick and Peter laughed” (= JPW, §1.3.1.2, p. 60, 9a) (3) κι ο Γιάννης κι ο Νίκος κι ο Πέτρος γέλασαν ki o γianis ki o nikos ki o petros γelasan and the John and the Nick and the Peter they-laughed “John and Nick and Peter (all) laughed” (= JPW, §1.3.1.2, p. 60, 9b) Unlike English “and”, but much like English “even” or “also”, και can also appear before a single constituent, indicating emphasis.

Downloaded from Brill.com10/01/2021 01:32:53AM via free access 66 Robert Ingria

(4) του έδωσα και την διεύθυνσή σου tu eðosa ke tin ðefthinsi su to-him I-gave and the address your “I gave him your address, too.” // “I also gave him your address.” (in addition to some other information) (= CC, 2.1) Here και functions like επίσης (episis = “also”): (5) του έδωσα επίσης την διεύθυνσή σου tu eðosa episis tin ðiefθinsi su to-him, I-gave also the address your “I gave him your address, too.” // “I also gave him your address.” (= CC, 2.1) In addition to this intensifying use of και, in which it essentially highlights one member of an at least implicit set of items, there is also a focussing use, in which και highlights an item that is not interpreted as a distinguished member of a pre-existing or discourse relevant set of items. Canakis (1996) refers to this use as “out of the blue και” or “the a propos of nothing sense”, and gives the following examples: (6) επί τη ευκαιρία, πήρα και κάρτα από το Στέλιο epi ti efkeria, pira ke karta apo to stelio on the occasion I-got and card from the Stelios “By the way, I got a card from Stelios.” (= CC, 3.1) (7) η Linda Evangelista έχει και το κορμί i Linda Evangelista exi ke to kormi the Linda Evangelista she-has and the body “Linda Evangelista has the [best] body [around].” (= CC, 3.2) These facts suggest an initial analysis of the syntax and semantics of και along the following lines. I posit that the syntactic structure of a conjoined structure is the following:

Downloaded from Brill.com10/01/2021 01:32:53AM via free access Grammatical formatives in a generative lexical theory 67

(8) XP

YP ZP

ke ZP In the typical case of simple conjunction of like categories, XP, YP and ZP will all be the same. This schema also allows for cross-categorial conjunctions, which constitute some of the more interesting uses of και. και is taken to be a dependent of the second conjunct, rather than a sister of the two conjuncts, to allow for the existence of examples such as (4), (6) and (7), in which και appears as the left-most daughter of a constituent. Such con- stituents would merely be isomorphic with the second conjunct of this general conjunction schema. This syntactic asymmetry also reflects a semantic asym- metry. I will suggest that the minimal semantic structure associated with και has the following schema: (9) [ |Y| * |Z| ] where |Y| represents the semantics of the first conjunct, |Z| represents the se- mantics of the second conjunct, and * represents the focus or “head” marking element of GL theory. That is, I posit that και introduces a semantic struc- ture that inherently contains two semantic elements (which I will refer to as “semantic conjuncts” for ease of reference and to distinguish them from the syntactic conjuncts) and that the second of these is highlighted. In the case of simple conjunction, the mapping from syntax to semantics is straightforward: the semantics of the first syntactic conjunct is instantiated as the first semantic conjunct and the semantics of the second conjunct is instantiated as the second semantic conjunct. In the case of intensifying and “out of the blue” και, the constituent of which και is a dependent provides the semantics for the second semantic conjunct, while the first must be instantiated in some other way. The examples in (4), (6) and (7) suggest that there are two modes of instantiation of the first semantic conjunct: from discourse context or by existential quan- tification. In the case of the intensifying use of και the first semantic conjunct receives its interpretation from some previously mentioned element. For example, in (4), και την διεύθυνσή σου (ke tin ðiefθinsi su) “and your address” is taken to be an addition to some previously mentioned piece of information, such as a phone number. It is this discourse element that instantiates the first semantic conjunct of the schema introduced above. In the case of “out of the blue” και,

Downloaded from Brill.com10/01/2021 01:32:53AM via free access 68 Robert Ingria

as in (6) and (7), there is no previous discourse referent to instantiate the first semantic conjunct, so it is existentially quantified, presumably with fairly vacu- ous semantics. This means that the second conjunct is highlighted relative to some radically underspecified background, and so is simply thrown into relief “on its own”, as it were. This initial sketch of an analysis has several attractive properties. First, the apparent syntactic asymmetry of και is matched by a semantic asymmetry which seems to accord with the facts. Second, the intensifying and “out of the blue” uses of και are collapsed and are treated not as two separate senses or uses, but rather as relatively simple extensions of the basic semantics of και, which differ only by the context in which they occur. Thus, there is no hard and fast distinction between these two “extended” uses of και, which seems, in fact, to be the case. For example, if (4), which was presented here as an example of intensifying και, appeared without the relevant discourse antecedent, it could be interpreted as a case of “out of the blue” και i.e. as “Oh, by the way, I gave Stelios your address.” This treatment also seems to extend to examples like (3), in which each conjunct is emphasized. One technical sticking point of this treatment is the use of * to focus the second semantic conjunct. Within GL, * is used to focus on an element at the event structure level. It would seem to be an extension of the domain of * to use it to arbitrarily highlight miscellaneous semantic elements. Its use here, however, may not, in fact, be much of an extension, since there are indications that the interpretation of και introduces temporal information. Consider the following data. The Modern Greek temporal preposition μέχρι (mexri) is in- terpreted as either “by” or “until”, depending on the event type of the verb with which it is construed: (10) θα ’χουμε φύγει μέχρι την Τρίτη θa xume fiγi mexri tin triti FUT we-have left by the Tuesday “we’ll have left by Tuesday.” (from M, p. 131) (11) θα μείνουμε ανοιχτοί μέχρι την Τρίτη θa minume anixti mexri tin triti FUT we-remain open until the Tuesday “we shall remain open until Tuesday.” (from M, p. 132)

Downloaded from Brill.com10/01/2021 01:32:53AM via free access Grammatical formatives in a generative lexical theory 69

If την Τρίτη (tin triti = “Tuesday”) in (11) is emphasized by the addition of και: (12) θα μείνουμε ανοιχτοί μέχρι και την Τρίτη θa minume anixti mexri ke tin triti FUT we-remain open until and the Tuesday the temporal nature of the constraint is emphasized: “We shall remain open up to and including Tuesday.” Note that this use of και is not precisely equiva- lent to either its simple intensifying use (throwing one member of a set of ele- ments into focus) nor to its “out of the blue” use (emphatically introducing something new). Rather, it emphasizes the temporal interpretation (inclusive endpoint) of “Tuesday”. This suggests that there may be some co-composition of the semantics of και with that of μέχρι (mexri) and την Τρίτη (tin triti). If και does, indeed, carry temporal information, the use of * in its interpretation may be no real extension of its previous uses. There are other indications that και introduces temporal constraints, when one or more of its conjuncts bear an event structure representation. As Τζάρτζανος (Tzartzanos 1946) and Canakis (1996) point out, when και conjoins two clauses with temporal interpretations, either they are both interpreted as occurring at the same time (13)–(14) or else the actions of the first are taken as preceding those of the second (15)–(16); the opposite temporal order results in anomaly (17). (13) Ο Ανδρούτσος έτρωγε κι έπινε κι έστριφτε το μουστάκι o anðrutsos etroγe ki epine ki estrifte to mustaki the Αndhrutsos he-was-eating and he-was-drinking and he-was-turning the moustache “Andhrutsos was eating and drinking and twirling his moustache” (From Τζ §219, p. 15) (14) κάθομαι και περιμένω kaθome ke perimeno I sit and I-wait “I am sitting and waiting.” (From Τρ §1032, p. 393) (15) ο Σκαλιγέρης επλήρωσε κι έφυγαν o skaliγeris eplirose ki efiγan the Skaliyeris he-paid and they-left “Skaliyeris paid and they left” (From Τζ §219, p. 15)

Downloaded from Brill.com10/01/2021 01:32:53AM via free access 70 Robert Ingria

(16) κάνε μια φωτοτυπία και δώσε μου το πρωτότυπο kane mia fototipia ke ðose mu to prototipo make! a photocopy and gíve! me the originai “Make a photocopy and (then) give me the original” (= CC 6) (17) # δώσε μου το πρωτότυπο και κάνε μια φωτοτυπία ðose mu to prototipo ke kane mia fototipia give! me the original and make! a photocopy (= CC 6) While these facts might be explained on a purely pragmatic basis, they might also be indications that και introduces temporal constraints of its own. Taken together with example (12), which seems to involve some sort of temporal co- composition, and later examples which seem to show that και is sensitive to event structure information, there is some initial evidence that the semantics of και contains constraints relating to tense, aspect, and event structure.3 In addition to its interpretation as a co-ordinating conjunction, και fre- quently appears in contexts in which it functions like various subordinating conjunctions, such as “because”, “but”, “therefore”, etc. While English “and” can sometimes appear in such extended uses, και is used much more frequently in such constructions and is the most common conjunction of any type in Mod- ern Greek.4 και = γιατί (γiati “because”) (18) φοβούμαι σ’, αδερφάκι μου, και λιβανιές μυρίζεις fovume s’ aðerfaki mu ke livanies mirizis I-fear you brother my and incense you-smell “I fear thee, my brother, for (because) thou smellest of incense” (From Τ, §261, p. 184) (19) άνοιξε, κ’ εγώ ’μ’ ο Κωσταντής anikse k’ eγo ’m’ o kostantis open! and I I-am the Constantine “Open, for I am C.” (From Τ, §261, p. 184) (20) έλα και μας περιμένει ο μπαμπάς ela ke mas perimeni o babas come! and us is-waiting the dad “Come on because dad is waiting for us.” (= CC 1)

Downloaded from Brill.com10/01/2021 01:32:53AM via free access Grammatical formatives in a generative lexical theory 7

(21) φάτο κ’ είναι καλό fato k’ ine kalo eat!-it and it-is good “eat it because it’s good” (From Ca §129, p. 165) (22) πού είναι ο Σταύρος και τον θέλω; pu ine o stavros ke ton θelo where he-is the Stavros and him I-want “Where’s Stavros? I want him!” (= M, p. 242, 21) και = αλλά/μα (ala/ma “but”) (23) Κάμνει το σοφό και δε γνωρίζει τους πιο απλούς νόμους της γλώσσας του λαού kamni to sofo ke ðe γnorizi tus pio aplus nomus tis γlossas tu lau he-makes the wiseman and NEG he-knows the most simple rules of-the language of-the people “He plays the wise man but doesn’t know the simplest rules of the language of the people” (From Τζ, §219, p. 17) (24) Είχα τόσα να πω και είχα γλωσσοδέτη ixa tosa na po ke ixa γlossoðeti I-had so-much SUBJ I-say and I-had tongue-tied “I had so much to say but I was tongue-tied” (From Τζ, §219, p. 17) (25) Όλοι αγαπούν και χαίρονται και γω αγαπώ και κλαίω oli aγapun ke xerode ke γo aγapo ke kleo all they-love and they-rejoice and I I-love and I-cry “Everybody loves and is happy but I love and weep” (From Τζ, §219, p. 17) (26) οχτ’ αδερφοί δε θέλουνε κι ο Κωσταντίνος θέλει oxt’ aðerfi ðe θelune ki o kostantinos θeli eight brothers NEG they-want and the Constantine he-wants “eight brothers do not wish it, but C. wishes it” (From Τ, §261, p. 184)

Downloaded from Brill.com10/01/2021 01:32:53AM via free access 72 Robert Ingria

(27) ήταν άρρωστος και πήγε στη δουλειά itan arostos ke piγe sti ðulia he-was sick and he-went to work “He was sick, but he went/did go to work.” (= CC 10)5 και = γιαυτό (γιafto “therefore”) (28) Είμαι φτωχό κορίτσι, ορφανό, κι έχω ανάγκη να δουλέψω, γιανά ζήσω ime ftoxo koritsi orfano ki exo ananki na ðulepso γana ziso I-am poor girl orpham. and I-have need to work in-order-to I-live “I am a poor giri, an orphan and (so) I must work in order to live” (From Τζ, §219, p. 16) (29) μπορεί να μη βρίσκει εργάτες και δεν έρχεται bori na mi vriski erγates ke ðen erxete it-may-be SUBJ NEG he-finds workers and NEG he-comes “it may be he can’t find workers and [that’s why] he’s not coming” (= M, p. 242, 18) The fact that και can be used to express this range of relations between clauses suggests that it may be underspecified vis-a-vis the various subordinating con- junctions in Greek and that, moreover, it may be underspecified compared to English “and”. Support for both these hypotheses is provided by consideration of further uses of και in the conjunction of nominals. When two conjoined nominals are taken to refer to the same individual or event, the second con- junct can appear without an article (30), and, in position, can trigger singular rather than plural verb agreement (31). (30) ο γιος της και ανηψιός τους o γιos tis ke anipsios tus the son her and nephew their “her son [who is] their nephew” (= M, p. 203, 47) (31) η έρευνα και ανακάλυψη των τάφων έχει ολοκληρωθεί i erevna ke anakalipsi ton tafon exi olokliroθi the search and discovery of-the tombs it-has been-completed “the search [for] and discovery of the tombs has been completed” (= M, p. 203, 46) A more extreme instance of the use of και to conjoin nominals appears in the fol- lowing examples, where the second conjunct in each case essentially functions

Downloaded from Brill.com10/01/2021 01:32:53AM via free access Grammatical formatives in a generative lexical theory 73 as an appositive, either providing a different description of the first conjunct or further specifying a subtype of the first conjunct. This “pleonastic” use of και, as Tzartzanos (1946) calls is, is not limited to two conjuncts, as (35) shows. (32) Δάσκαλε, πού ’ναι ο Κωσταντής και πού ’ναι το παιδί μου; ðaskale pu ne o kostantis ke pu ne to peði mu teacher where he-is the Konstantine and where he-is the child of-me “Teacher, where is Constantine? Where is my child?” (From Τζ, §219, p. 18) (33) να πάω κι’ εγώ με τα πουλιά και με τα χελιδόνια na pao ki eγo me ta pulia ke me ta xeliðonia SUBJ I-go and I with the birds and with the swallows “May I too go with the birds, with the swallows” (From Τζ, §219, p. 18) (34) Πατήσανε τα Γιάννινα και τη μεγάλη χώρα patisane ta yanina ke ti meγali xora they-trod the Yannina and the great city “They conquered Yannina, the great city” (From Τζ, §222 p. 24) (35) Καραβοκύρη κι’ αδερφέ και πολυαγαπημένε karavokiri ki aðerfe ke poliaγapimene skipper and brother and much-beloved “Skipper, beloved brother” (From Τζ, §222, p. 25) Note that, in each case, the English equivalents omit “and” and that the pres- ence of “and” would not allow for identity or subtyping among the conjuncts.6 Taken together, the data in (18)–(35) suggest the following:

– και, as well as various subordinating conjunctions, are complex elements, at least semantically. – και has an underspecified semantics, compared to that of overt subordinat- ing conjunctions such as μα (ma “but”), τότε (tote “then”), γιαυτό (γiafto “therefore”), etc. This allows it to be used in situations where these con- junctions can appear, since its semantics is non-distinct from theirs in the appropriate respects. One way of characterizing the situation is to view each of the various subordinating conjunctions in Modern Greek as indi- cating a particular rhetorical relation (in the sense of Mann etc.),7 while και can mark any one of a range of rhetorical relations.

Downloaded from Brill.com10/01/2021 01:32:53AM via free access 74 Robert Ingria

– και is also underspecified in with English “and” in that it al- lows for identity and subtype relations among its conjuncts where English “and” permits no such interpretations. – και cannot be completely underspecified, since it is distinct from negative conjunctions such as ούτε (ute “nor”) and disjunctions such as ή (i “or”). While the hypothesis that και is semantically underspecified both within Mod- ern Greek (compared to the subordinating conjunctions) and cross-linguisti- cally (compared to English “and”) seems promising, it is merely a programmat- ic statement at present. Giving it substantive content would require specifying what the elements are that make up the complex semantic structures associated with και and other co-ordinating and subordinating conjunctions, in Modern Greek and in English. This is one of the areas suggested by this study in which more work must be done. These examples of clausal conjunction with και and the hypothesis that και can fulfill the function of various subordinating conjunctions because it is underspecified raises one final question. If και is, in effect, substituting for μα, γιαυτό, γιατί, and other subordinating conjunctions, how can it do so if it ap- pears in the adjunction structure sketched in (8)? Presumably, subordinating conjunctions appear as daughters of S′ and sisters of S, as in (36), rather than adjoined to S′. (36) S′

S(′) S′

ma S How can και receive a complementizer interpretation if it appears in a non- complementizer position? One solution would be to posit a non-parallelism in phrasal and clausal conjunction. That is, phrasal conjunction would have the schematic structure given in (8), while clausal conjunction would have a struc- ture similar to, if not identical with, that of adjunct clauses, as in (37). (37) S′

S′ S′

ke S

Downloaded from Brill.com10/01/2021 01:32:53AM via free access Grammatical formatives in a generative lexical theory 75

Given this structure, the ability of και to receive interpretations identical to those of subordinating conjunctions would be less problematic. As we shall see, there are other structures involving και in which it would be useful to treat it as appearing in complementizer position. Moreover, even in English, there are differences between the behavior of clausal and phrasal conjunction. “And” can occur with co-relative conjunctions in phrasal conjuncts that cannot appear in clausal conjuncts. (38) Both Bill and Hillary will be there. (39) * Both Bill was at Oxford and Hillary was at Wellesley. Moreover, even in cases where the same co-relative conjunctions can appear in clausal and phrasal conjuncts, there are differences in behaviour between the two cases. (40) Not only Bill but also Hillary will be there. (41) * Not only Bill was at Oxford but also Hillary was at Wellesley. (42) Not only was Bill at Oxford but Hillary was also at Wellesley. (43) * Not only Bill went to Oxford but also Hillary went to Wellesley. (44) Not only did Bill go to Oxford but Hillary also went to Wellesley. Note in particular the subject-aux inversion in the first conjunct in (42) and (44), which is a phenomenon that is often induced by elements in COMP. The fact that “but also” envelops the subject in the second clause is also notewor- thy, and would be consistent with “but also” being part of the specifier system of S′. I will tentatively, then, adopt the structure in (37) for clausal conjunction.

5. Conditional και

One of the more interesting uses of και is illustrated by the following example: (45) το παραμικρó και φύγαμε/φεύγουμε to paramikro ke fiγame/fevγυme the slightest and we’re-gone-(Pst./Prf)/ we’re-leaving-(Prs./Imprf.) “If you do the slightest thing, we’re leaving.” (= CC 5)

Downloaded from Brill.com10/01/2021 01:32:53AM via free access 76 Robert Ingria

This construction raises several questions: l. What is the nature of the constituent that spans the phrase and the clause? 2. Where does the event interpretation of the initial noun phrase come from and why is the noun phrase required to have an event interpretation in the first place? 3. What forces the conditional interpretation of the entire structure? In answer to question (1), I would propose that we have a structure like that in (46), a variation on the structure presented in (37): (46) S′

NP S′

ke S That is, the entire conjoined structure is an S′, and the initial NP appears in “sat- ellite” position, similar to a topicalization structure, at least at the syntactic lev- el. (If, in fact, the structure is identical to a topicalization structure, this might suggest that certain constructions share the same syntactic structure, with their differing interpretations being the result of the contribution of different lexical semantics. In the example under discussion, this would be the difference be- tween a COMP occupied by και, with its semantics, and an empty COMP.) As for the answer to question (2), concerning the event reading of the ini- tial nominal, I believe this follows from the asymmetrical nature of the syn- tactic and semantic structures associated with και. In examples stich as (45), an NP, which is semantically not an event, is conjoined with a clause, which semantically is an event, producing a disagreement in the semantic types of the conjoined elements. By virtue of the asymmetrical nature of the syntactic and semantic structures, it is the type of the second conjunct that prevails. That is, in order for the conjoined structure to be well-formed, both conjuncts must be events. Since the initial NP conjunct does not denote an event, it must be coerced into an event, in a process well familiar from the GL literature. In a case when a nominal must be coerced into an event reading, the reconstructed event is derived from the qualia structure of the nominal. In this construction, the event appears to be derived from the agentive role. This, however, raises the question of why the event should be reconstructed from the agentive quale, rather than some other.

Downloaded from Brill.com10/01/2021 01:32:53AM via free access Grammatical formatives in a generative lexical theory 77

Finally, I would propose the following answer to question (3), regarding the conditional interpretation of the construction. Note that the event recon- structed from the nominal’s quale lacks a temporal anchor, given that the nom- inal has no event structure. This suggests that the event so derived is given an irrealis interpretation, just like the protasis of an overt conditional in Modern Greek, in which the main verb receives an aspectual interpretation, without a temporal anchor. Support for this position comes from a related construction, in which the first conjunct is in the imperative; in Modern Greek, imperative forms are also marked for aspect but not tense.8 (47) μη με σκοτώνης κι εγώ θα σε κάμω να πάρης τη βασίλισσα mi me skotonis ki eγo θa se kamo na paris ti vasilissa NEG me you-kill and I FUT you I-make SUBJ you-get the queen “Don’t kill me and I will have you marry the queen” (From Τζ, §219, p. 18) (48) Σώσε μας, καπετάνε, και να γενούμε σκλάβοι σου sose mas kapetane ke na γenume sklavi su save! us captain and SUBJ we-be slaves your “Save us, captain, and we’re your slaves” (From Τζ, §219, p. 18) (49) λύσε με και σ’ το δείχνω lise me ke s’ to ðixno release! me and to-you it I-show “untie me and I[’ll] show you (it)” (= M, p. 242, 23) While the analysis sketched out here seems to provide a good start to an ex- planation of the basic properties of this construction, there are still open ques- tions. In particular, it seems that while the properties mentioned here — the coercion of the semantics of the nominal to an event, and the irrealis nature of the first conjunct inducing a conditional interpretation — may be necessary to produce the interpretation observed, they may not be sufficient. Let us observe the analogous construction in English. Note that the initial nominal conjunct seems to require a particular type of quantification, and that even structures where a pragmatically plausible inter- pretation could be constructed are ill-formed if they do not meet this condition — contrast (52) and (53). (50) Another outburst and I’ll clear the court!

Downloaded from Brill.com10/01/2021 01:32:53AM via free access 78 Robert Ingria

(51) * The outburst and I’ll clear the court! (52) * An outburst and I’ll clear the court! (53) If there is an outburst, I’ll clear the court! A complete analysis of the conditional use of “and” would explain the reason for this constraint. Similarly, it needs to be ascertained whether similar constraints hold of the Modern Greek construction. Note that the nominal in (45) has a definite determiner, but that the lexical semantics of παραμικρό (paramikro “the least little thing”) seems to provide a similar kind of cumulative interpre- tation that the determiners and quantifiers provide in the English examples.9

6. Complementizier uses of και

In this section, we turn to one of the most striking uses of και in Modern Greek: και as complementizer, that is, as a formative that introduces comple- ment clauses. Let us begin with cases that are potentially ambiguous between conjunction and complement analyses and then move on to cases that are more clearly complements. (54) βρήκα το Βασίλη και διάβαζε vrika to νasili ke ðiaνaze I-found the Vassilis and he-was-reading “I found Vassilis reading” (= CC 4.1) This is given an interpretation parallel to that of (55), in which που (pu) is one of three complementìzers introducing finite indicative complement clauses in Modern Greek. (55) βρήκα το Βασίλη που διάβαζε vrika to vasili pu ðyaνaze (= CC 4. 1)10 Sentences such as (55) are a common form of complement structure for per- ception verbs in Modern Greek. As was mentioned above, Modern Greek does not have infinitives and many perception verbs in Modern Greek can appear in the context: (56) ___ NP [S′ pu … ]

Downloaded from Brill.com10/01/2021 01:32:53AM via free access Grammatical formatives in a generative lexical theory 79 where the direct NP is construed as a nominal within the complement clause. Typically, this is the subject, but construal is possible with other posi- tions as well; see Ingria (1981) for discussion. I will not consider the question of how the construal is effected in this and subsequent constructions, since this is a general problem of Modern Greek syntax, and is not peculiar to clauses in- troduced by και. Examples of και appearing in this construction are numerous, and it occurs with assorted perception verbs. (57) ακούν πουλιά και λένε akun pulia ke lene they-hear birds and they-say “they hear birds saying (and they say)” (From Τ, §261, p. 185) (58) βλέπει το φτωχό κ’ έρχεται vlepi to ftoxo k’ erxete he-sees the poor and he-is-coming “he sees the poor man coming” (From Τ, §261, p. 185) (59) τον είδα και πήγαινε ton iða ke piγene him I-saw and he was-going “I saw him going” (= JPW, §1.3.1.1, p. 59, 5) (60) βρίσκει την κ’ εχτενίζουνταν vriski tin k’ extenizudan he-finds her and she-was-combing-herself “he finds her as she was combing herself” (From Τ, §261, p. 185) (61) βρήκε τα παιδιά της και κοιμώτανε vrike ta peðia tis ke kimotane he-found the children her and they-were-sleeping “he found her children sleeping” (From Ca §129, p. 165) (62) θωρώ το πρόσωπό σου κ’ έγινε σαν τη φωτιά θoro to prosopo su k’ eγine san ti fotia I-see the face your and it-became like the fire “I see how thy face became like fire” (From Τ, §261, p. 185)

Downloaded from Brill.com10/01/2021 01:32:53AM via free access 80 Robert Ingria

και can also appear in this use with να (na), a formative analogous to voilà in French and ecco in Italian. (This να is distinct from the να that marks the subjunctive.) (63) να τον και κατέβαινε na ton ke katevene ecco him and he-descended “behold how he descended” (From Τ, §261, p. 185)11 Such examples, as Householder, Kazazis and Koutsoudas (1964) point out, are ambiguous as to whether they are actually complementizer uses of και or simple conjoined structures. In these examples, the interpretation of the clause following και might simply be inferential. For example, in (54), if I found Vas- silis and he was sleeping at the same time that I found him, as the imperfective tense or the second clause indicates, then, by inference, I found him sleep- ing. This overlap of events in the first and second conjuncts, as indicated by an imperfect or present tense in the clause following και, is true of examples (57) through (61). In (62), on the other hand, the tense in the second clause is a simple past, which might make the inference a bit more difficult. Such examples need to have their structures disambiguated by further tests, such as extraction by WH-movement from the putative complement clauses. The pos- sibility of extraction would support the complement analysis; its impossibility would suggest a simple conjunction analysis.12 In addition to such ambiguous cases, there are other examples of και in- troducing apparent complement clauses that are less susceptible to a simple conjunction analysis. Just as in English perception verbs can appear either fol- lowed by a noun phrase and a non-finite clause, as in “I saw John leaving” or by a finite clausal complement, as in “1 saw that John was leaving”, so, too, in Modern Greek do perception verbs appear followed either by an NP in the ac- cusative case and a complement clause, or simply by a complement clause. και also appears in the latter context, apparently introducing a simple complement clause. (67) άκουσα κ’ έλεγε akusa k’ eleγe I-heard and he-was-speaking “I heard that he was speaking” (From Ca §129, p. 165)

Downloaded from Brill.com10/01/2021 01:32:53AM via free access Grammatical formatives in a generative lexical theory 8

(68) ήκουσα και σε μάλωνε η κερά σου ikusa ke se malone i kera su I-heard and you she-scolded the mother your “I heard how thy mother scolded thee” (From Τ, §261, p. 185) (69) βλέπω και χαμογελάς vlepo ke xamoγelas Ι-see and you-laugh “I see that you are smiling” (From Τζ, §220, p. 19) (70) Ξάφνου θαρρεί κι’ επέρνα μια λάμψη βιαστική ksafnu θari ki eperna mia lampsi viastiki suddenly he-perceives and he-was-passing a flash sudden “Suddenly he perceives that he was passing (like) a sudden flash” (From Τζ §220, p. 19) Note also the following example, in which the constituency of απ’ το σταυρό (ap to stavro “from the cross”) is unclear. (71) Και βλέπω απ’ το σταυρό και βγαίνει αίμα ke vlepo ap to stavro ke vγeni ema and I-see from the cross and it-comes-out blood “And 1 see blood dripping from the cross” (From Τζ §220, p. 19) In all these cases, since the initial verb requires a complement of some sort, de- riving the complement interpretation of the clause following και by inference is less likely, since the initial verb would be left with its subcategorization unsatis- fied if the second clause were not its complement. Applying the inference anal- ysis in these cases, then, would also require positing an elliptical analysis of the clause containing the initial verb, with the second conjunct filling in the elided complement. Again, extraction tests should decide this matter, but, indepen- dent of such tests, the inference analysis is much less likely in these cases. We now come to examples where a complement interpretation of the clause following και seems impossible. These are sentences involving cases of apparent subject control (i.e. construal of the subject of the matrix clause with the subject of the complement clause). One of the interesting features of the in- terpretation of these constructions involves the question of whether the event denoted by the complement clause is interpreted as merely potential or hav- ing actually occurred. The verbs that participate in this construction typically

Downloaded from Brill.com10/01/2021 01:32:53AM via free access 82 Robert Ingria

make a complement clause whose main verb is subjunctive. Since Modern Greek verbs in the subjunctive receive an aspectual interpretation (iterative vs. punctual) but no temporal interpretation, the event denoted by the comple- ment clause in such a case is always interpreted as potential (irrealis). Comple- ment clauses introduced by και, on the other hand, allow their main verb to be in the indicative, and, hence, to receive a temporal interpretation. This allows complement clauses introduced by και in these apparent control cases to con- vey the information that the event denoted by the complement clause actually took place. We can see examples of this phenomenon in (72)–(80). (72) Δύνεσαι και διώχνεις την παγωνιά ðinese ke ðioxnis tin paγonia you-can and you-send-away the frost “You can send away the frost (and you are sending it away)” (From Τζ §220, p. 20) (73) μού ζήτησε κ’ ήπιε νερό mu zitise k’ ipie nero to-me he-asked and he-drank water “He asked me for a drink of water (and got it).” (From HKK §7.132, p. 171) (74) μπορούσε και το ’κανε boruse ke to kane s/he-was-able and it s/he-did “(s)he was able to do it and (s)he did it” (= M, p. 242, 14) (75) γιαυτό εμπόρεσε κ’ έφκειασε τόσα κάστρα γiafto eborese k’ efkiase tosa kastra therefore he-was-able and he-built so-many castles “therefore he was able to build (and he built) so many castles.” (From Τ §261, p. 185) (76) Δεν ντράπηκες και του το είπες; ðen drapikes ke tu to ipes NEG you-were-ashamed and to-him it you-told “Weren’t you ashamed to tell him that?” (From HKK §7.132, p. 171)

Downloaded from Brill.com10/01/2021 01:32:53AM via free access Grammatical formatives in a generative lexical theory 83

(77) πρόφτασε και του ’πιασε το χέρι proftase ke tu piase to xeri he-was-in-time and him-he-caught-hold-of the hand “he was in time to catch him by the hand” (From Ca §129, p. 165) (78) Προσπαθώ και τα μαθαίνω μόνος μου prospaθo ke ta maθeno monos mu I-try and them I-learn alone of-me “I’m trying to learn it myself” (From HKK §7.132, p. 171) (79) Τόλμησε και στο είπε αυτό; tolmise ke sto ipe afto he-dared and to-you-it he-told that “He dared (to) tell you that?” (From HKK §7.132, p. 171) (80) Φροντίζω και της στέλνω εγώ τα χρήματα frontizo ke tis stelno eγo ta xrimata I-take-care and to-her I-send I the money “I take care to send her the money myself” (From HKK §7.132, p. 171) και also appears introducing complement clauses to αρχίζω (arxizo), the Mod- ern Greek equivalent of English “begin”, which has been analyzed as ambigu- ous between raising and control uses in English. (81) Σιγά το κύμα αρχίζει και μερώνει siγa to kima arxizi ke meroni gradually the wave it-begins and it calms-down “The wave is gradually beginning to become calm” (From Τζ §220, p. 19) (82) άρχισε κ’ έβρεχε arxise k’ evrexe it-began and it-was-raining “it began to rain” (From Ca §129, p. 165) Householder, Kazazis, and Koutsoudas (1964, §7.132, p. 171) provide the fol- lowing list of verbs that permit complement clauses introduced by και with construal of the matrix subject. The lists of tenses following some of these verbs

Downloaded from Brill.com10/01/2021 01:32:53AM via free access 84 Robert Ingria

seems to indicate the tenses of that verb that permit the use of και but they do not explicitly state what these annotations mean. (83) αγωνίζομαι (aγonizome) “struggle, contend” αξίζω (aksizo) “deserve” αξιώθηκα (aksioθika) “manage” αποφασίζω (apofasizo) “decide” αρχίζω (arxizo) “begin” (aorist and imperfect) ευκαιρώ (efkero) “be at leisure to, have a chance to” έχω (exo) “have” ζητώ (zito) “demand, beg” καρτερώ (kartero) “wait” καταντώ (katando) “end up” καταφέρνω (kataferno) “succeed, manage” κατορθώνω (katorθono) “accomplish, succeed” μαθαίνω (maθeno) “learn” (aorist and present) μπορώ (boro) “can, be able” νοιώθω (nioθo) “notice feel” ντρέπομαι (drepome) “be ashamed” ξέρω (ksero) “know” πλησιάζω (plisiazo) “approach” προλαβαίνω (prolaveno) “be in time to, have enough time to” προσπαθώ (prospaθo) “try” προσφέρομαι (prosferome) “be suitable” προφταίνω (profteno) “be or act in time” συνηθίζω (suniθizo) “be accustomed to, be used to” τολμάω (tolmao) “dare” φροντίζω (frontizo) “take care to, see to” There seems to be some dialect difference or other factor that produces un- certainty as to whether a verb can appear in this construction. For example, Householder, Kazazis, and Koutsoudas claim that θέλω (θelo “want”) does not participate in this construction (84), while Tzartzanos presents an example of just such a use (85).13 (84) Ήθελε και το έκανε iθele ke to ekane he-wanted and it he-did “He did it because he wanted to.” Not: “He wanted to do it” (After HKK §7.132, p. 171)

Downloaded from Brill.com10/01/2021 01:32:53AM via free access Grammatical formatives in a generative lexical theory 85

(85) Τι ήθελα και σ’ έστελνα στη ξενιτειά ti iθela ke s estelna sti ksenitia why I-wanted and you I-send to-the abroad “Why did I want to send you abroad (and I did send you)” (From Τζ §220, p. 20) και also appears introducing complement clauses to verbs in which the direct object of the matrix verb is interpreted as the subject or other nominal con- stituent of the complement clause. (In (87) the matrix object is construed with the object of the complement clause.) (86) αναγκάζεται και στέλνει anagazete ke stelni he-is-forced and he-sends “he is forced to send” (From Ca §129, p. 165) (87) τους αφήνει και τους πιάνουν τα λαγωνικά tus afini ke tus pianun ta laγonika then he-lets and them they-catch the hounds “he allows them to be caught by the hounds” (From Τ §261, p. 185) (88) τους βάνει κάθε νύχτα και οργώνουν τα χωράφια του tus vani kaθe nixta ke orγonun ta xorafia tu them he-makes every night and they-till the fields his “he makes them every night till (and they till) his acres” (From Τ §261, p. 185) (89) με γείπε και το bσείρισα me γipe ke to bsirisa me he-said and him I-took-off the lice “he said to me to take the lice off him [and I did so]” (From Τ §261, p. 185) (90) με κάνεις κι ανατριχιάζω me kanis ki anatrixiazo me you-make and I-shudder “you make me shudder.” (From Ca §129, p. 165)

Downloaded from Brill.com10/01/2021 01:32:53AM via free access 86 Robert Ingria

(91) μ’ έκανες και σε πίστεψα m ekanes ke se pistepsa me you-made and you I-believed “You made me believe you” (= M, p. 242, 15) (92) δεν ξέρω τι με κρατάει και δε σας πετάω έξω ðen ksero ti me kratai ke ðe sas petaο ekso NEG Ι-know what me it-prevents and NEG you I-throw out “I don’t know what’s keeping me from getting rid of you” (From Ca §129, p. 165) As in cases of subject construal, the indicative nature of the main verb of the complement clause allows its associated event to be interpreted as having actu- ally occurred (86). It is also possible for the complement clause introduced by και to contain a subjunctive verb, in which case there is no difference in inter- pretation from the identical complement clause introduced by να: (93) μη με κάνεις κι’ αρχίσω πάλι mi me kanis ki arxiso pali NEG me you-make and I-start again “don’t make me start again” (M, p. 278, 17)14 Householder, Kazazis, and Koutsoudas (1964, §7.132, p. 171) provide the fol- lowing list of verbs that permit complement clauses introduced by και with construal of the matrix object. Note that their list also includes some of the perception verbs we saw earlier.15 (94) ακούω (akuo) “hear” (aorist and imperfect) αναγκάζω (anagazo) “compel, oblige” αφήνω (afino) “let” βάζω (vazo) “get (to do)” βιάζω (viazo) “force” βλέπω (vlepo) “see” (aorist and imperfect) βρίσκω (vrisko) “find” (aorist and imperfect) διατάζω (ðiatazo) “order” διδάσκω (ðiðasko) “teach” (aorist and present) κάνω (kano) “make” μαθαίνω (maθeno) “teach” (aorist and present) μηνώ (mino) “send (for)” παραγγέλνω (paragelno) “order”

Downloaded from Brill.com10/01/2021 01:32:53AM via free access Grammatical formatives in a generative lexical theory 87

στέλνω (stelno) “send” συνηθίζω (siniθizo) “accustom” υποχρεώνω (ipoxreono) “oblige” και also introduces complement clauses to impersonal verbs and other imper- sonal constructions. (95) Αν λάχη και ξενιτευτώ an laxi ke ksenitefto if it-happens and I-live abroad “If it happens that I live abroad” (From Τζ §220, p. 19) (96) πώς συνέβηκε και το ξέχασες; pos sinevike ke to ksexases how it-happened and it you-forgot “how did it happen that you forgot?” (From Ca §129, p. 165) (97) μα τυχερό μου και τη στιγμή εκείνη ακούστηκεν η φωνή του βλησιδιάρη μας ma tixero mu ke ti stiγmi ekini akustiken i foni tu vlisiðiari mas but fortunate for-me and the moment that it-was-heard the voice of-the treasurer our “But it was lucky for me that at that moment the voice of our treasurer was heard” (From Τζ §220, p. 19) (98) Ώρα είναι και σχόλασε το βελόνι ora ine ke sxolase to veloni time it-is and it-stopped-work the needle “It is time for the needle to come to rest (and it has come to rest)” (From Τζ §220, p. 20) These examples raise many interesting questions. The proposal that και- ap pears in complementizer position within a clause that was made at the end of Section 4 at least explains why it is that και can introduce complement clauses. Still to be explained, however, is how και can appear in a complement position without violating the selectional restrictions of the main verb. There is also the question of delimiting the class of verbs that allows και to introduce their complements, since there seem to be lexical restrictions on the occurrence of και in this use.

Downloaded from Brill.com10/01/2021 01:32:53AM via free access 88 Robert Ingria

We can make a start at explaining why και satisfies the selectional restric- tions of the verbs with which it occurs by considering several other cases in which και appears as a complementizer. The first such case involves the use of και introducing a . (99) μια φορά ήταν ένας βασιλιάς κι’ είχε δυο κόρες mia fora itan enas νasilias ki ixe ðio kores one time there-was a king and he-had two daughters i. “Once there was a king and (he) had two daughters.” ii. “Once there was a king who had two daughters.” (= CC 9) (100) Στα στήθη μου έχω φωτιά κι’ ανάφτει την καρδιά μου sta stithi mu echo fotya ki anafti tin karðya mu in-the breast my I-have flame and it-inflames the heart my “In my breast I have a flame that inflames my heart” (From Τζ §220, p. 22) (101) μπας και είναι τουρκόπουλο και δε νοιώθει τη γλώσσα μας; bas ke ine turkopulo ke de nioθi ti γlossa mas perhaps and he-is Turk-(Dimin) and NEG he-understands the language our “Perhaps he’s a Turkish youth who doesn’t understand our language?” (From Τζ §219, p. 15) Note that Modern Greek has two relative clause formation strategies. One in- volves movement of a (slightly archaic) ; the other, by far more frequent, involves the use of a resumptive pronoun in a subordinate clause in- troduced by the complementizer που (pu). και can also introduce result clauses. (102) είναι τόσο κουτός και δεν το καταλαβαίνει ine toso kutos ke den to katalaveni he-is so stupid and NEG it he-understands “he is so stupid that (and) he does not even grasp it” (From Τ §261, p. 185) (103) Τι μπορεί να τους έφταιξε η Ιφιγένεια, και θέλουν τώρα το θάνατό της; ti bori na tus eftekse i Ιfιγenia, ke θelun tora to θanato tis what she-can SUBJ them she-did the Iphigenia and they-want now the death her “what harm can Iphigenia have done them for them now to desire her death?” (= M, p. 242, 19)

Downloaded from Brill.com10/01/2021 01:32:53AM via free access Grammatical formatives in a generative lexical theory 89

(104) τι έχει το μωρό και κλαίει ti exi to moro, ke klei what it-has the baby and it-cries “what’s wrong with the baby that it’s crying?” (From Τ §261, p. 185) (105) τι έπαθες και δε μιλάς ti epaθes ke ðe milas what you-suffered and NEG you-speak “What’s the matter with you? Why don’t you speak?” (= M, p. 242, 17) Again, result clauses in Modern Greek are more typically introduced by the complementizer που (pu). Finally, και can also introduce purpose (“in order to”) clauses, which are normally introduced by για να (γia na “for to”) or simply by να (na), the by now familiar subjunctive marker and closest Modern Greek equivalent to English infinitival “to”. As in the complement cases we have seen, και allows its clause to denote an event that has actually occurred, resulting in semantic anomaly in examples such as (108). (106) Αν έρθη κανείς και σε ρωτήση, μην του πης τίποτα an erθi kanis ke se rotisi min tu pis tipota if he-comes someone and you he-asks NEG to-him you-say nothing “If someone comes to ask you a question, don’t tell him anything” (From Τζ §220, p. 21) (107) ερχόταν και σας έβλεπε κάθε μέρα erxotan ke sas evlepe kaθe mera he-was-coming and you he-was-seeing every day “Ηe used to come to see you every day.” (= CC 4.2) (108) # ερχόταν και σας έβλεπε κάθε μέρα αλλά λείπατε… erxotan ke sas evlepe kaθe mera ala lipate… he-was-coming and you he-was-seeing every day but you-were-away (= CC 4.2a) (109) πήγε και κάθησε piγe ke kaθise go! and sit! “go (to) sit down” (from Ca §129, p. 165)

Downloaded from Brill.com10/01/2021 01:32:53AM via free access 90 Robert Ingria

(110) ήρθε και πήρε irθe ke pire he-came and he-took! “he came and took it.” (From Ca §129, p. 165) In all these instances of the use of και as a complementizer, i.e. in complement, relative, result, and purpose clauses, it appears in contexts in which either the complementizer που (pu) or the modal particle να (na) might occur instead. Given the hypothesis at the end of Section 4 that και is underspecified in com- parison to other Modem Greek , this free variation suggests that και is also non-distinct from που and να. In this case however, I believe we can pin down the ways in which και resembles που and να in somewhat more detail by building on work by Ginzburg and Kolliakou. As was mentioned above in passing, που is one of the three complementiz- ers in Modern Greek that introduce complement clauses. The other two are ότι (oti) and πως (pos). Ascertaining the criteria by which different verbs select one or another of these complementizers has been one of the long-standing questions in Modern Greek syntax. Ginzburg and Kolliakou (1995) propose an analysis of Modern Greek complementizer selection, at least of ότι and που, based on event types. According to their analysis: – “A pu clause unambiguously describes an event.” – “An oti clause denotes a proposition.” Their hypothesis that clauses introduced by που denote events provides a start to explaining why και can appear in contexts where που does. I have already suggested that και is underspecified. If we assume, as seems likely, that claus- es by default denote events unless coerced to another semantic type, then an underspecified complementizer like και, which presumably does not coerce the clauses which it introduces, will also introduce clauses that denote events. Therefore, if Ginzburg and Kolliakou are correct that complementizer selec- tion is done in terms of semantic type, it seems natural that και can appear in contexts in which που does. As for και appearing in clausal contexts in which να does, again, it ap- pears possible to build on Ginzburg and Kolliakou’s analysis. Like clauses in- troduced by που, clauses containing να also seem to denote events, rather than propositions or questions (another semantic type borne by clauses in Ginzburg and Kolliakou’s analysis). However, unlike clauses introduced by που, clauses containing να have an irrealis interpretation. Perhaps we should take them as

Downloaded from Brill.com10/01/2021 01:32:53AM via free access Grammatical formatives in a generative lexical theory 9 denoting a subclass of events, potential events.16 Note that the semantic types of που and να must be compatible, since να can appear in result and relative clauses introduced by που. At any rate, και must be compatible with the irrealis interpretation of clauses containing να, since it permits subjunctive verbs, as we saw in (93). Under an analysis based on Ginzburg and Kolliakou’s, we might view com- plementizers like ότι and πως as transducers that coerce the semantic type of the clauses in which they occur from an event to a proposition or other type. Since και is regarded as underspecified under the hypothesis being propounded here, it would lack the semantic structure to effect such coercions and, there- fore, would be incompatible with contexts that select for ότι and πως.17 While this approach is attractive, and seems to account for the bulk of the data, there are apparent counterexamples.18 (111) Μη λέτε κι’ είμαι νιόνυφη mi lete ki ime nionifi NEG say! and I-am newly-married “Don’t believe that I am a newlywed” [I follow Tzartzanos in glossing lete as “believe” here.] (From Τζ §220, p. 19) (112) έμαθα και παντρεύτηκες emaθa ke pantreftikes I-learned and you-got-married “I learned that you got married” (From Τζ §220, p. 19) (113) δε ξέρουν οι γιατροί κι απ’ε το βασιλόπουλο γιατρεύεται ðe kserun i γiatri ki ape to vasilopulo γiatrevete NEG they-know the doctors and from-what the king-let he-is-cured “the physicians do not know by what means the royal child is cured” (From Τ §261, p. 185)19 It is not clear that the clauses introduced by και in each of these cases denote an event rather than a proposition. In any case, they certainly require a closer examination. Finally one outstanding issue in the analysis of complement clauses in- troduced by και is the nature of the verbs that permit this construction. The analysis sketched so far suggests that these should be just the set of verbs that select for που or να. While this seems to be true from the cases observed so far, a much more intense, and, preferably, corpus-based, study of this question needs to be done.20

Downloaded from Brill.com10/01/2021 01:32:53AM via free access 92 Robert Ingria

7. Summary and Future Work

In this paper, I have presented paradigms of data exemplifying some of the range of uses of the conjunction και in Modern Greek. I have suggested that a good deal of its behavior can be explained by the hypothesis that, in its clausal uses, it is underspecified, both with respect to other complementizers in Mod- ern Greek and vis-à-vis “and” in English. I have also presented syntactic and semantic frames in which και and its semantics appear, and have suggested how these structures interact with its underspecified character to produce the interpretations observed. Many questions still remain, however. It is not clear whether all the apparent complement introducing uses of και presented in Sec- tion 6 should be analyzed in the same way or whether the different classes of complement have different analyses. Moreover, there is the question of how the semantic schema (9) interacts with these apparent complement structures. There are also questions regarding the differences in the behavior of Modern Greek και and English “and” which does not appear in the range of comple- ment structures that και does. Under the hypotheses sketched here, the differ- ence would presumably be that “and” is more specified than και, so that, just as it does not have the same wide range of interpretations presented at the end of Section 4, it also does not appear in complement structures like those in Section 6. While this seems promising, further analysis of such cross-linguistic differences is necessary. All that said, I believe, nevertheless, that this paper has demonstrated con- clusively that grammatical formatives can be a rich field of inquiry for lexical semantics and, further, that GL provides tools that are necessary for the analy- sis of grammatical formatives. I hope that this paper will mark the expansion of the areas studied by lexical semantics, even as GL marked an expansion of the field of lexical semantics from earlier work.

Notes

* This paper was presented in a Workshop on Lexical Semantics in Courmayeur, Italy, 7–10 September 1996, where Bob Ingria was an invited speaker. The conference organizers had expressed their intention to publish the papers from the workshop together with audience comments, but there is no indication that the paper was ever submitted for publication. The paper was posted on Ingria’s website immediately after the conference as an initial draft (http://world.std.com/~ingria/cintro.htm). The paper remained there unaltered until his death in 2003 and is no longer accessible.

Downloaded from Brill.com10/01/2021 01:32:53AM via free access Grammatical formatives in a generative lexical theory 93

Nick Nicholas sent Ingria comments in October 1996 on the paper; though he expressed intent to repeatedly, he never responded to them. In preparing this paper for publication, Nicholas has left the body of the paper intact, apart from silent corrections of spelling and translation (and switching from the original inconsistent Romanization to a phonemic IPA). Nicholas is responsible for the comments to the paper appearing as footnotes; there were no footnotes in the original paper.

. In conventional orthography: Τρομερός ΧΑΡΑΚΤΗΡΑΣ, yes — τρομερός ΗΘΟΠΟΙΟΣ, no! 30 χρόνια στο επάγγελμα, βελτίωση μηδέν! Ευτυχώς ό,τι και να κάνει τον σώζει το σενάριο. (Μεγάλε JMS!) Και του βελτίωσαν και το περουκίνι!!!!

2. Thumb’s, Triantafyllidis’ and Tzartzanos’ examples, on which so much of this paper hing- es, predate the standardization of Modern Greek, and many are unacceptable in the current standard. This does not mean that such usage should not be accounted for; but this would be most illuminatingly done within a dialectological framework, which is beyond the scope of the present paper.

3. It has long been established, as Ingria alludes, that the temporally sequential sense of and in English (and by extension in Greek) results from the Gricean maxim of manner (“be orderly”), and is defeasible. This does not preclude a co-composition of the type proposed for (12); but the applicability of μέχρι και in a locative as well as a temporal sense (έφτασε μέχρι και την Ασία, eftase mexri ke tin asia ‘it reached even Asia (inclusive)’) may mean that the “up to and including” sense does follow directly from the intensifying sense after all: the implicit |Y| in [ |Y| * |Z| ] would be simply “the time/distance up to and not including Z”. So by itself (12) does not resolve the difficulty in applying * to και NP.

4. Such usage is not widely attested for Classical Greek, where clause chaining was done with participles instead (but see Trenkner 1960); however it is plentifully attested for Hel- lenistic (Aejmelaeus 1982) and Late Middle Greek (Ljungvik 1932:55–62).

5. The more difficult it is to infer the incompatibility of the two conjuncts, the more marked the contrastive reading becomes, so that it relies increasingly on the appropriate intona- tion (specifically an intonation break between the conjuncts). So whereas (26) is obviously contrastive (“NEG want” vs. “want”), (27) relies for its contrastive interpretation on the im- plicature “sick people don’t go to work”, and would characteristically be uttered with an intonation break before και. This does not contradict Ingria’s contention that και is under- specified, but it confirms that recovering a specific sense in such instances is contingent on pragmatics.

6. Identity at least of conjuncts outside of vocatives (32, 34) is unacceptable in Standard Modern Greek; this may be a feature specific to the folk song genre, which features strong use of parallelism.

7. The standard reference to Mann & Thompson’s Rhetorical Structure Theory is Mann and Thompson (1988); see also the RST web site at http://www.sfu.ca/rst

8. As is common cross-linguistically, the imperative is used here conditionally. The “or-else” use of ‘and’ after prohibitives is an intriguing extension of the construction: the prohibitive

Downloaded from Brill.com10/01/2021 01:32:53AM via free access 94 Robert Ingria

imperative “Don’t X” can mean “if you don’t X, then…”, as in (47), but it can also mean the opposite: “unless you don’t X = if you do X, then…”:

(47) a. Μην πας και θα σε δείρω min pas ke θa se ðiro NEG go and FUT you I-beat ‘Don’t go or I’ll beat you.’ (‘Unless you don’t go, I’ll beat you’)

The motivation for this construction is the use in the same context of γιατί ‘because’: μην πας γιατί θα σε δείρω ‘don’t go because I will beat you [if you do]’. This use of ‘because’ for ‘or else’ is a Balkanism, also attested in Albanian and Arumanian (Fiedler 1987:61), and we have already seen that και can substitute for γιατί in (18)–(22). The construction is even extended to prohibitives that are merely implicit in interjections — so that the event interpretation is coerced onto an element which is not overtly either verbal or nominal:

(47) b. Αλτ και σ’ έφαγα alt ke s efa>a halt! and you I-ate ‘Halt! or you’re dead.’ (‘Unless you halt, I will have eaten you = I will kill you instantly’: “Halt!” = “Don’t move!”)

Both (47) and (47a) involve a tenseless imperative reinterpreted as a conditional, and both have the same original prohibitive protasis. Crucially, the conditional involves not the illo- cutionary act of the directive (which after all would be infelicitous), but the perlocutionary act resulting from the directive (“if indeed you do not kill us…”) But the polarity of the con- ditional appears to be determined purely by conversational implicature: a rational speaker is assumed to reward compliance and to punish disobedience. So (47), where the apodosis is a reward, is the apodosis of compliance to the directive — the speaker’s desired state of affairs. But (47a), where the apodosis is a punishment, is the apodosis of disobedience to the directive: “unless my directive is carried out”. Incidentally, while this “or else” sense is encompassed by English or (halt or I’ll shoot!), it is not encompassed by Greek i; so Ingria is still right that ke is not so underspecified as to include negative conjunctions or disjunctions, particularly given that Greek treats this as a causal connection rather than a disjunction. Although the implicit, perlocutionary protasis is negated in “or else”-ke, the on-the-record protasis, the illocutionary directive, is indeed asserted; and the construction is disjunctive only in the trivial truth-conditional equivalence of ¬A ⇒ B ≡ Α ∨ Β — but certainly not in its illocutionary force.

9. As in English, the constraint appears to be that the NP is quantified as a minimal amount, either explicitly or implicitly; the implication exploited by the use of the construction as a warning is that punishment will follow from even a minimal deviation, so the addressee should not attempt any deviation at all. The Greek equivalent to (50) is μία/(?μια) ακόμα κουβέντα mia akoma kuveda ‘one more word’ οr έστω και μία/μια κουβέντα esto ke mia kuveda ‘even hypothetically one word’: the appearance of accented μία [Ámi.a] rather than unaccented μια [mja] confirms that this is the quantifier ‘one’, and not the indefinite article. The literal equivalents of (51) and (52), η κουβέντα (i kuveda) and κουβέντα (Ø kuveda) with a null determiner respectively, are just as ungrammatical as in English. The minimal quantification of paramikro is of course lexical.

Downloaded from Brill.com10/01/2021 01:32:53AM via free access Grammatical formatives in a generative lexical theory 95

0. (55) is itself ambiguous between complementizer and relativizer readings of pu.

. There is a grammaticality distinction at work betweenke and pu after this formative: both can appear in na ton pu/ke katevene, with a direct object of na (whose behavior is verblike: Joseph 1981). But without the direct object, only pu is acceptable: na pu/*ke katevene. On the other hand, the medieval equivalent of na, eðe (Kriaras 1968–1997 s.v. έδε; etymologically the imperative ‘see!’) can occurs as both eðe (ο)pu and eðe ke, without a direct object (though this appears restricted for ke to exclamatory clauses with ti ‘what! how!’); and Hellenistic Greek uses the even earlier idou kai increasingly with a clausal rather than nominal referent (although usually kai is a focus particle in this context, meaning ‘also’ — as in (4)):

(64) a. έδε και τι χολομανώ και τι βαρεά στριγγίζω eðe ke ti xolomano ke ti varea stri:izo see! and what I-rage and what heavily I-sigh ‘See how I rage and how I sigh heavily.’ (Michael Glykas, Verses Written While Confined to Prison 293; ca. 1158) b. έδε και τι μ’ εμάρανεν της ξενιτείας η στράτα eðe ke ti m emaranen tis ksenitias i strata see! and what me it-withered of-the exile the path ‘See how the path of exile has withered me.’ (On Exile 46; early 15th century) c. έδε οπού μας έδωκαν οι θεοί στρατιώτην βασιλέα eðe opu mas eðokan i θei stratiotin vasilea see! that us they-gave the gods soldier king ‘See how the gods have given us a soldier–king.’ (Achilleid, Naples ms. 273; 15th century) d. καὶ ἰδοὺ καί γε Σαδωκ καὶ πάντες οἱ Λευῖται μετ’ αὐτοῦ kai idou kai ge sado˜k kai pantes hoi leuitai met autou and see! and at-least Zadok and all the Levites with him αἴροντες τὴν κιβωτὸν διαθήκης κυρίου airontes t7˜n kibf˜ton diath7˜k7˜s kyriou lifting the ark of-covenant of-Lord ‘And lo Zadok also, and all the Levites were with him, bearing the ark of the covenant of God.’ (Septuagint, 2 Sam 15:24: nominal argument — “see Zadok who was lifting…”) e. Ἰδοὺ καὶ ξύλον ἡτοίμασεν Αμαν Μαρδοχαίῳ idou kai ksylon he˜toimasen aman mardokhaiff˜i see! and wood he-prepared Haman to-Mordecai ‘Behold also the gallows which Haman had made for Mordecai.’ (Septuagint, Esther 7:9: clausal argument — “see that Haman has prepared a gallows”)

The explanation is straightforward in terms of coercion. As a relativizer, pu routinely at- taches clauses to non-clausal heads, so there is little syntactic difficulty in it attaching to an interjection-like formant such as na. But ke is constrained to conjoin clauses to clauses. The requisite verbal interpretation is readily available for eðe and idou (which are still verbs), and is also accessible when na is acting like a verb, taking a direct object. When this is not the case, na has to be coerced back into verb status to attach to a ke-clause; this has ended up dispreferred, possibly because the alternative na pu is not as problematic.

Downloaded from Brill.com10/01/2021 01:32:53AM via free access 96 Robert Ingria

2. Delveroudi (1994:281–282) provides other tests establishing the complementizer sta- tus of ke: (1) ordering restrictions, under which the reverse ordering is not merely anoma- lous like (17) but nonsensical: αρχίζει και καταλαβαίνει/*καταλαβαίνει και αρχίζει arxizi ke katalaveni/*katalaveni ke arxizi ‘she starts to realize’. (2) Restrictions on person in the ke-clause, indicating subcategorization by a matrix predicate: *μπορώ και διαβάζει boro ke ðiavazi ‘I-can and she-reads’. (3) Restrictions on negation: *δεν ξέρει και ράβει ðen kseri ke ravi ‘she doesn’t know how to sew (and she does sew)’, contradicting the factivity of ke as a complementizer, cf. (70)–(80) — which again would not arise unless the ke-clause was a complement and not just a conjunct. For the particular test of WH-extraction, cf:

(65) a. Τι είδες ότι/πως διάβαζε; ti iðes oti/pos ðiavaze what you-saw that he-was-reading Ø ‘What did you see that he read?’ b. ?? Τι τον είδες που διάβαζε; ti ton iðes pu ðiavaze what him you-saw that he-was-reading Ø ‘What did you see him reading?’ c. Τι τον είδες να διαβάζει; ti ton iðes na ðiavazi what him you-saw SUBJ he-reads Ø ‘What did you see him reading?’ d. Τι τον/Ø είδες και διάβαζε; ti ton/Ø iðes ke ðiavaze what him/Ø you-saw and he-was-reading Ø ‘what did you see him reading?’

To my judgement, WH-movement for pu is marginal, but WH-movement for ke is accept- able. It is difficult to find instances of such WH-movement, but the following instance of raising out of a ke-complement proves the same point:

(66) e. To ma8ntiko kivnma, molis twra arxise kai Ø dnmiourgeite… the student movement only now started and it-is-being-created Το μαθητικό κίνημα, μόλις τώρα άρχισε και δημιουργείται… ‘the student movement is only now starting to be created’ (‘Rocean’, “Ma8ntikes Kouvitopoinseis (Was: RE: Klassika **** sobaro”, Hellas mailing list, 29 Sep- tember 1999; USENET: bit.listserv.hellas)

3. θelo ke is widely used in standard Greek; it is prominent in the fixed expression θέλεις και τα παθαίνεις θelis ke ta paθenis (‘you want to suffer these things (and you do)’ = ‘you willingly suffer these things’ = ‘serves you right’). But it is certainly true that the accept- ability of complementiser-ke is highly variable: the construction itself is now marked as colloquial, and under retreat before the more explicit complementizers of the standard lan- guage. Of the verbs in (83), my impression is that the verbs which already presuppose their complements remain most acceptable with ke-complements — so that the ke-complement does not end up with a different distribution from the unmarked na-complement. Of the 26

Downloaded from Brill.com10/01/2021 01:32:53AM via free access Grammatical formatives in a generative lexical theory 97 listed, 12 presuppose their complements outright (aksioθika, arxizo, efkero, katado, katafer- no, katorθono, drepome, prolaveno, profteno, siniθizo, tolmao, frodizo; e.g. (80)), and another four are semi-factive (Karttunen 1971: maθeno, boro, ksero, prosferome). Delveroudi disputes the bona fides of the other verbs — verbs of obligation (e.g. ofilo ‘I am supposed to’), intent (e.g. apofasizo ‘decide’), or attempt (e.g. prospaθo ‘try’). She con- tends that these verbs are not really asserting their complements, but rather that the ke- clause is an elliptical reading: “I decided to go and I did go”; Tzartzanos himself in his ex- amples speaks of this as “brachylogy”, and Thumb glosses them similarly. Such an “elliptical” reading would be impossible for factive verbs like kataferno: ??“I managed to go and I did go”. The elliptic reading is activated according to Delveroudi when context minimizes the distance between the two events, e.g. deciding to go and going, when these form part of a narrative sequence (with the matrix in either the aorist or the historical or habitual present), but not if the matrix event is in the imperfect: the assertion of successful outcome lies out- side the process of deciding, so locating the clause during the deciding phase does not allow the outcome assertion to be made — whereas if this were a true factive complement, this should not present an obstacle (*αποφάσιζα και πήγα apofasiza ke pi>a ‘I was deciding to go and I went’; αποφάσιζα και πήγαινα apofasiza ke pi>ena ‘I kept deciding to go and going’ is only permissible because it has an iterative reading. By contrast, cf. factive κατάφερνα κι αποκοιμώμουνα την ώρα που τηλεφώνησες kataferna ki apokimomuna tin ora pu tilefonises ‘I was managing to get to sleep when you rang’.) It is difficult, as Ingria mentions, to differentiate complementizer-ke from conjunct uses (which semantically are more akin to resultatives: (102)–(105)). For instance, the idiomatic translation of (85) is “what was I thinking, to have sent you abroad”. If ti ‘what’ is indeed the direct object of θelo, and not a variant of >iati ‘why’, then the ke-clause cannot be a comple- ment of θelo: ke here looks more like a conjunct than a complementizer, with a resultative in- terpretation (“what was I thinking, as a result of which I sent you abroad”). I would interpret (92) in the same way, particularly as (92) cannot be expressed with an overt complementizer (*τι με κρατάει να μην σας πετάξω έξω ti me kratai na min sas petakso ekso ‘what is holding me back from throwing you out’ — cf. τι με εμποδίζει να σας πετάξω έξω ti me eboðizi na sas petakso ekso ‘what prevents me from throwing you out’).

4. The class of connectives that can introduce an aorist subjunctive clause is restricted, though subject to analogical pressure — e.g. although the relativiser που (pu) cannot normal- ly introduce a subjunctive on its own, κάθε που (kaθe pu) ‘each (time) that’ can do so (Mack- ridge 1985:258), by analogy with όποτε (opote) ‘whenever’. (93) shows such analogy with na underway. Moreover, (93) is incompatible with a conjunct reading for ke, and is only consis- tent with a complement reading, since the conjunction of different moods is not normally allowed: *πηγαίνω και να έρθω pi>eno ke na erθo ‘I go and that I should come’. (When it does occur, as with (47)–(49), an implicature needs to be formed which is modally consistent.)

5. Again, of the sixteen verbs in (94), only the four verbs of command (ðiatazo, mino, paragelno, stelno) and the verb of obligation (ipoxreono, see foregoing discussion) do not presuppose their complements. The verbs of compulsion (anagazo, vazo, viazo, kano) do presuppose them: “I order him to read” does not require that the target ends up reading, but “I force him to read” does. Obviously the two classes are semantically similar, which is why ke is used with both.

Downloaded from Brill.com10/01/2021 01:32:53AM via free access 98 Robert Ingria

6. The literature consistently agrees that na and pu introduce the same kind of entity (events), and Christidis (1982; 1983) distinguishes them through presupposition, and through the pu-event delimited and bound to a specific time and place, while na is not. In- gria analyses the ke-clauses (72)–(80) as temporally delimited, in the same way as Christidis proposes for pu-clauses. Subsequently the trend has been to shy away from describing the semantics of Greek complementizers, and has taken on the more tractable task of describing the semantics of complement-taking predicates. Ginzburg and Kolliakou (1997:468) in particular set up an ontology for predicates according to the complements they take as ±e(vent), ±f(act), ±p(roposition), under which verbs taking pu-complements select for +e and either +f or +p, while verbs taking na-complements select +e but neither +f nor +p. This implies that a na-complement is an event that cannot be coerced into a fact or a proposition — which is consistent with the na-complement not being temporally delimited.

7. Delveroudi (1994:289) independently dissociates ke from the propositions introduced by oti and pos, and further dissociates it from pu-complements, which are obligatory for emotive predicates (χαίρομαι που/*οτι/*και ήρθες xerome pu/*oti/*ke irθes ‘I am glad that you came’). In her scheme oti-complements are an assertion subordinate to the matrix asser- tion, while pu-complements are an assertion of a state of affairs, and the emotive matrix is an assertion of a reaction to it (so the matrix is pragmatically subordinate to the complement). Being non-propositional, na-clauses do not constitute assertions about states of affairs in the world distinct from the matrix — the matrix–complement complex is asserted as a unit. The ke-complement on the other hand involves an assertion distinct from the matrix, but Delveroudi concludes that its assertion is not subordinate or superordinate to the matrix: both clauses are situated in the same state of affairs, so both are true without one being contingent on the other.

8. Whereas (94), with ke standing in for na, is acceptable in (colloquial) Standard Mod- ern Greek, (111)–(112) are not. Delveroudi (1994:290) explicitly rejects (111) (and 70) as genre-restricted. (On (113) see below.) I recall a Cretan relative saying θυμούμαι κι ήτονε ένας φακίρης θimume ki itone enas fakiris, ‘I remember there was a conjurer’; this seems to me more acceptable, but θimame can take pu-complements in the standard language, whereas leo ‘say’ does not. On the other hand, λες και les ke, ‘say that’ = ‘you’d think that; as if’ is a standard idiom (with no presupposition of the complement!); this suggests that ke- complements were formerly more widespread, and have become restricted to this particular collocation. (Delveroudi 1994:288 accounts for this non-presuppositional usage by having the ke-clause asserted as true in the world — which she posits as the distinguishing feature of ke-complements — but as being so asserted by a fictional everyman speaker (the second person), placing the claim in the realm of the hypothetical. I admit to not finding this con- vincing: why can’t the everyman claims be common and undisputed knowledge?) I would not wish to say that such examples as (111)–(112) do not need to be accounted for. Rather, the extent of underspecification of ke appears to be variable in space and time (as one might expect of competing complementizers with overlapping semantics): cf. the varying proportions of pu vs. oti/pos as complementizers in different dialects outlined in Nicholas (1999 §8).

Downloaded from Brill.com10/01/2021 01:32:53AM via free access Grammatical formatives in a generative lexical theory 99

9. (113) would be syntactically strange indeed if Thumb’s translation were accurate, but it is not: απέ is a dialectal variant of από ‘from’ (the text is from Epirus) which can be used adverbially (‘from that point’ = ‘afterwards’, possibly as a metanalysis of από + εκεί ‘from there’ > απέκει ‘thence, since’: Kriaras 1968–1997 s.v. από II, with bibliography). So the first bird in Thumb’s (1964:235) tale asks: “Aren’t you upset that the prince is ill and the doctors have decided so?” The second bird responds with (113): “The doctors don’t know [any such thing], and soon the prince will be cured”. The first bird then asks: “With what medicine will he be cured?” — leading to Thumb’s misunderstading.

20. As already noted, see Delveroudi (1994).

2. Asterisked items appear only in Nicholas’ commentary.

References21

*Aejmelaeus, Anneli. 1982. Parataxis in the Septuagint: A Study of the Renderings of the Hebrew Coordinate Clauses in the Greek Pentateuch. (Annales Academiae Scientiarum Fennica, Dissertationes Humanarum Litterarum 31) Helsinki: Suomalainen Tiedeaka- temia. Canakis, Costas. 1996. “Kai: The Polysemous Conjunction”, Talk presented at Harvard Uni- versity, February 22, 1996. Catone, Nicola. 1967. Grammatica neoellenica. Roma: Edizioni dell’ Ateneo. Chomsky, Noam. 1965. Aspects of the Theory of Syntax. Cambridge, Mass.: MIT Press. *Χριστίδης, Αναστάσιος–Φοίβος [Christidis, Anastasios-Phivos]. 1982. “Πρόσθετες παρα- τηρήσεις στις συμπληρωματικές προτάσεις της ΝΕ [Further notes on complement clauses in Modern Greek]”. Studies in Greek Linguistics 3. 47–72. *Χριστίδης, Αναστάσιος–Φοίβος [Christidis, Anastasios-Phivos]. 1983. “Παρατηρήσεις στη σύνταξη των «αισθήσεως σημαντικών» στα νεοελληνικά [Observations on the syntax of emotive verbs in Modern Greek]”. Studies in Greek Linguistics 4. 115– 125. *Delveroudi, Rhea. 1994. “Ο συμπληρωματικός δείκτης και [The complementizerke ]”. Stud- ies in Greek Linguistics 15. 281–291. *Fiedler, Wilfried. 1987. “Zur Typologie der Grammatischen Interferenz zwischen den Balkansprachen im Bereich der Konnektive”. Studien zur Morphologie und Syntax der Balkansprachen, 1: 45–68. (Linguistiche Studien Reihe A Arbeitsberichte 157) Berlin: Akademie der Wissenschaften der DDR, Zentralinstitut für Sprachwissenschaft. Ginzburg, Jonathan & Dimitra Kolliakou. 1995. “Events and Facts: the semantics of pu and oti clauses”. Paper presented at the 2nd International Conference on Greek Linguistics, Salzburg, September 22–24, 1995. Published as: Ginzburg, Jonathan & Dimitra Kolliak- ou. 1997. “Events and Facts: the semantics of pu and oti clauses”. Greek Linguistics ’95: Proceedings of the 2nd International Conference on Greek Linguistics, ed. by G. Drach- man, A. Malikouti-Drachman, J. Fykias & C. Klidi, II:459–470. Graz: W. Neugebauer. Householder, F. W., Kostas Kazazis & Andreas Koutsoudas. 1964. Reference Grammar of Literary Dhimotiki. Bloomington, Ind.: Indiana University Publications. (= IJAL 30.2, Part II.)

Downloaded from Brill.com10/01/2021 01:32:53AM via free access 00 Robert Ingria

*Joseph, Brian D. 1981. “On the Synchrony and Diachrony of Modern Greek NA”. Byzantine and Modern Greek Studies 7. 139–154. Joseph, Brian D. & Irene Philippaki-Warburton. 1987. Modern Greek. London and New York: Routledge. Kalokerinos, Alexis. 1995. “The semantics of enhancement: a comparative study”. Paper pre- sented at the 2nd International Conference on Greek Linguistics, Salzburg, September 22–24, 1995. Published as: Kalokerinos, Alexis. 1995. “The semantics of enhancement: a comparative study”. Greek Linguistics ’95: Proceedings of the 2nd International Confer- ence on Greek Linguistics, ed. by G. Drachman, A. Malikouti-Drachman, J. Fykias & C. Klidi, II:505–514. Graz: W. Neugebauer. *Karttunen, Lauri. 1971. “Some Observations on Factivity”. Papers in Linguistics 4. 55–69. *Κριαράς, Εμμανουήλ [Kriaras, Emmanouil]. 1968–1997. Λεξικό της μεσαιωνικής ελληνικής δημώδους γραμματείας [Dictionary of Medieval Vernacular Greek Writing]. Thessalonica. *Ljungvik, Hermann. 1932. Beiträge zur Syntax der Spätgriechischen Volkssprache. Uppsala: Almqvist & Wiksell. Mackridge, Peter. 1987. The Modern Greek Language. Oxford/New York: Oxford University Press. *Mann, William C. & Sandra A. Thompson. 1988. “Rhetorical Structure Theory: Toward a functional theory of text organization”. Text 8(3). 243–281. *Nicholas, Nick. 1999. The story of pu: The Grammaticalisation in Space and Time of a Modern Greek Complementizer. Unpublished Ph.D. Thesis, University of Melbourne. http://www.opoudjis.net/work/thesis.html. Philippaki-Warburton, I. & G. Veloudis. 1984. “The subjunctive in complement clauses”. Studies in Greek Linguistics: Proceedings of the 5th Annual Meeting of the Depart- ment of Linguistics, Faculty of Philosophy, Aristotelian University of Thessaloniki, 149–167. Θεσσαλονίκη: Εκδοτικός Οίκος Αφών Κυριακίδη. Pustejovsky, James. 1989. “Current Issues in Computational Lexical Semantics”. Proceed- ings of the 4th Conference of the European Chapter of the ACL, Manchester, England, xvii–xxv. Pustejovsky, James & Peter Anick. 1988. “On the Semantic Interpretation of Nominals.” Pro- ceedings of COLING–1988. Budapest. Scha, Remko & David Stallard. 1988. “Multi-Level Plurals and Distributivity.” 26th Annual Meeting of the Association for Computational Linguistics: Proceedings of the Confer- ence, Association for Computational Linguistics, Morristown, NJ, 17–24. Setatos, M. 1983. “The perfect tense in Common Modern Greek.” Studies in Greek Lin- guistics: Proceedings of the 4th Annual Meeting of the Department of Linguistics, Faculty of Philosophy, Aristotelian University of Thessaloniki, 101–113. Θεσσαλονίκη: Εκδοτικός Οίκος Αφών Κυριακίδη. Thumb, Albert. 1964. A Handbook of the Modern Greek Language. Chicago: Argonaut, Inc., Publishers. *Trenkner, Sophie. 1960. Le style kai dans le récit attique oral. Assen (Netherlands): Van Gorcum. Τριανταφυλλίδης, Μανόλης. 1941. Νεοελληνική Γραμματική (της Δημοτικής) [ (of Demotic)]. Αθήναι: Οργανισμός Εκδόσεως Σχολικών Βιβλίων.

Downloaded from Brill.com10/01/2021 01:32:53AM via free access Grammatical formatives in a generative lexical theory 0

Τζάρτζανος, Αχιλλέας. 1946. Νεοελληνική Σύνταξις (της κοινής Δημοτικής) [Modern Greek Syntax (of Common Demotic)], v. 2. Αθήναι: Οργανισμός Εκδόσεως Σχολικών Βιβλίων. Βελούδης, Γιάννης. 1984. “Η υποτακτική στις αναφορικές προτάσεις” [“The subjunctive in relative clauses”]. GLoSSoLoGia 2–3 (1983–1984):111–135. Veloudis, G. 1985. “Tensed na complements in Modern Greek.” Studies in Greek Linguistics: Proceedings of the 6th Annual Meeting of the Department of Linguistics, Faculty of Philosophy, Aristotelian University of Thessaloniki, 183–198. Θεσσαλονίκη: Εκδοτικός Οίκος Αφών Κυριακίδη. Veloudis, I. & I. Philippaki-Warbuiton. 1983. “The subjunctive mood in Modern Greek.” Studies in Greek Linguistics: Proceedings of the 4th Annual Meeting of the Depart- ment of Linguistics, Faculty of Philosophy, Aristotelian University of Thessaloniki, 151–168. Θεσσαλονίκη: Εκδοτικός Οίκος Αφών Κυριακίδη. Warburton, Irene P. 1970. On the Verb in Modern Greek. Bloomington, Ind.: Indiana Uni- versity Publications.

Downloaded from Brill.com10/01/2021 01:32:53AM via free access Downloaded from Brill.com10/01/2021 01:32:53AM via free access