Word Order, Focus, and Clause Linking in Greek Tragic Poetry Bruce Lovat Fraser Darwin College A dissertation submitted for the degree of Doctor of Philosophy Cambridge University January 1999 ii Monsieur Jourdain: Il n'y a que la prose ou les vers? Maître de philosophie: Non, monsieur: tout ce qui n'est point prose est vers, et tout ce qui n'est point vers est prose. Monsieur Jourdain: Et comme l'on parle, qu'est-ce que c'est donc que cela? Molière, Le Bourgeois gentilhomme, Act 2 Scene 4 iii For Katherine and Cassie, with all love iv Declaration No part of this dissertation is being or has been submitted for any other degree or qualification at any university, and it is the author's sole work, including nothing which is the outcome of work done in collaboration. The word total is under 79,900, exclusive of the bibliography, and of the three appendices, of approximately 18,000 words in total, for which an extension was granted by the Degree Committee of the Faculty of Classics. The appendices contain formatted textual passages and some statistical tabulation of data. All results and summaries appear in the body of the thesis. Texts Oxford Classical Text editions are used for Greek citations, and Loeb Classical Library editions for Latin. English translations are the author's own, and are phrased to reflect the Greek syntax as closely as possible. Homeric translations are based on those of Lattimore (1951, 1965). All statistical collation of textual material is original, unless otherwise stated. Word searches were carried out with the help of Thesaurus Linguae Graecae, CD-Rom (D) edition (1992), published by the University of California at Irvine. Acknowledgments A great debt is owed to the superlative resources of the Cambridge University Library, and to the British Academy and the Cambridge Faculty of Classics, both of whom gave financial support. I was fortunate to be supervised by specialists in two fields: Geoffrey Horrocks, who devoted an enormous amount of time to reading the drafts and provided detailed linguistic criticism, and Richard Hunter, who gave invaluable stylistic and literary advice. Pat Easterling and Peter Matthews kindly read parts of the work in progress. Anthony Bowen, James Diggle, and Ian McAuslan have my gratitude for helping me understand something of the beauty and form of the Greek language. Sincere thanks are offered to Joseph Butler, Mark Hogarth, Joyce Reynolds, and Christine Salazar for their friendship and practical help. v Sigla Abbreviations and symbols are defined at first use, and are also noted here: AI Accusative and infinitive (in a non-finite dependent clause) CG Classical Greek Comp Head position in a CP CP Complementizer phrase (the extended clause structure) FP Focus Phrase (an alternative visualisation of the CP, for languages without complementation) IE Indo-European IP Inflection phrase (the main-clause structure) OV Object>verb ordering OVO Object>verb>object hyperbaton PIE Proto-Indo-European NP Noun phrase SOV Subject>object>verb ordering SV Subject>verb ordering SVO Subject>verb>object ordering SVS Subject>verb>subject hyperbaton TP Topic phrase (in some models, distinguished from the FP above) VO Verb>object ordering VP Verb phrase VS Verb>subject ordering X' [X-bar] Intermediate phrasal structure (and also the description of the linguistic framework which uses the category) * (with italic letters) marks reconstructed PIE stems * (with roman letters) marks a hypothetical sentence which is not grammatically well-formed [ ] enclosing a phrase mark constituent boundaries in citations > marks a regular sequence of words or phrases subscripti .... i mark co-referent elements in citations Standard abbreviations of titles are used when citing ancient texts. vi Contents Introduction Page 1 Part I: Word order 26 1 The order of subject, verb, and object 27 2 The presentational cadence: word order and phonological weight 60 3 Intra-clausal poetic syntax: phrasal tmesis in the Oresteia and other texts 88 Part II: Focus 125 4 Focus, particles, and the clause start 126 Part III: Clause linking 168 5 Subordination: clause order and focalization 169 6 Complementation: verb transitivity and focalization 201 7 Inter-clausal poetic syntax: focus and the discourse functions of complementation 230 Conclusion 258 Appendices 1: Subject and verb order 261 2: Hyperbaton 287 3: Complementation 305 Bibliography 336 1 Introduction Scope The dissertation comprises an investigation of three aspects of sentence structure in Classical Greek (henceforth CG) dramatic poetry: order of the main sentence elements (subject, verb, and object) within the clause, the emphatic position at the start of the clause, and the structure of inter-clausal linking. It is argued that these three features, usually considered separately, are interdependent, and that intra-clausal word order is directly related to the structure of compound and complex sentences.1 The discussion undertakes a systematic survey of subject, verb, and object order in a corpus of texts,2 proposes an explanation for the observed order, and develops a model which explains how prominence within the clause is exploited in clause linking to produce the complement structures observed in Homeric and tragic complementation. The problems 1) The primary problem is to explain the high degree of consistency in the order of the main sentence elements in what is traditionally considered a ‘free word order’ language.3 Ancient discussions usually described word order as an aspect of suvnqesi" (‘composition’);4 and concentrated on unusual orders rather than the norm. Modern studies, though paying more attention to ‘basic word order’,5 have not identified structural motivation for the regularities, and generally attribute variations to pragmatic determinants.6 1These terms refer to sentences of more than one clause, in a relation of conjunction or embedding respectively. See Lyons (1968: 266). 2In Chapters 1 and 2. Homeric order was discussed by Ammann (1922: 1924), Friedrich (1975), and Conrad (1990), and observations on tragic word order were made by Thomson (1938, 1939b). However, no systematic survey of tragic word order has previously been made. 3As by Kühner (1904: 595) and Dover (1960: 31). 4As Aristotle (Rhetoric), Cicero (Orator), Dionysius of Halicarnassus (De Comp.), and Quintilian (Institutio). See also Denniston (1952), Scaglione (1972) and Dover (1997). 5As Kieckers (1911), Fischer (1924), Frisk (1932), Thomson (1939a), and Chantraine (1952). 6As Goodell (1890), Loepfe (1940), Dover (1960), Dunn (1988), and Dik (1995, 2007). 2 2) The prosody of the clause start is standardly considered separately from its structure: either as an emphatic first position, or in terms of enclitic elements in second position.7 However, the structural relation between these two features has not been investigated. 3) Complementation has been extensively analysed in terms of its formal structure,8 and its historical development has been surveyed in terms of the introductory conjunctions,9 but not in terms of the relationships between these conjunctions (henceforth complementizers)10 and the semantic categories of main verb types. The process by which a whole clause, rather than a referring expression within it, came to function as an argument of the main verb,11 remains unexplained. The proposal The three problems have a unified explanation, because word order, emphasis, and clause-linking are inter-dependent. Intra-clausal word order has a morphological and a prosodic trajectory, with larger words tending to be placed later. However, there is also a position for prominent elements at the clause start (henceforth P1), which are emphatic, not only as a consequence of their initial placing, but also because they reinforced by light words in second position (henceforth P2), which mark emphasis in one of two ways. Enclitic particles are cohesive focalizers, combining phonologically with the initial constituent, so creating a larger unit. Interrogative and relative pronouns are separated from the P1 unit by an intonation break, and also create a contrast with it, by reason of their small size. In both cases the whole focal unit is separated from the basic clause by an intonation break. 7An initial emphatic position was noted by Thomson (1938: 367) and Denniston (1952: 44). Particles are analysed functionally by Denniston (1954), Ruijgh (1971, 1990), and Rijksbaron (1997a); structurally by Hale (1987, 1996), Schäufele (1991), and Wills (1993); and prosodically by Halpern (1992), Hock (1982, 1996) and Taylor (1996). Hale, Schäufele, Halpern, and Hock concentrate on Vedic Sanskrit. Other references are given in Chapter 4. 8Notably in the X-bar approach described below in Section 1. 9Most thoroughly by Monteil (1963). Other studies are cited in the main text. 10The term ‘complementizer’ to denote a complement-introducing conjunction derives from Rosenbaum (1967). 11The term ‘argument’ is used to identify the subject or object of a verb. For its sense in predicate calculus to denote the function of names in propositions, see Lyons (1977: 148–9). 3 In complex sentences, the trajectory of ‘weight to the right’ within the main clause combines with ‘prominence to the left’ in the following subordinate, so the focal element has a functions in two clauses simultaneously. Its prominence is linked with its exophoric (non-linguistic) reference: the grammaticalization of o{ti from a referring expression to a textually-deictic object involves a loss of specific reference, marked by an indefinite affix which is also a cohesive focalizer (‘say whatever you like’); while the change of function of wJ" from an adverbial to a complementizer is accompanied by the change of the preceding main verb object from a referring expression (‘I know you how you are’) to a textually-deictic pronoun (‘say this, how...’). In both constructions, the transition to complementation involves a circumstantial construction, where a verbal object is combined with a modifying clause.12 A distinction is made between intransitive (‘say of x that it is y’) and transitive circumstantials (‘know x as being y’), because they are regularly associated with different main verbs (of speech and cognition respectively).13 The circumstantial form of tragic complementation is transitional, though it is regular in tragedy and Plato: full complementation involves a loss of prosodic prominence, as the complementizer becomes a conventional marker. Consequences A unified framework has a number of advantages: it accords with the word order observed in the corpus of texts, it explains the presence of prominent words at the start of the clause and morphologically heavy ones at the end, and it shows a direct relation between the prosodic and pragmatic features of the sentence. It also provides a possible aetiology of complementation, by suggesting a semantic motivation for the change of complementizer function from a referring expression to a grammatical word,14 and for changes in word order (SV to VS, and OV to VO) and the change from a pitch to a stress accent between Ancient and Mediaeval Greek. 12Circumstantial constructions may be defined as verbal modifiers giving information on the circumstantial roles associated with a situation (see Halliday 1970b, Lyons 1977: 497). For their application to CG participial subordination, see Smyth (1956: 456–471). 13The distinction here depends on the form of the main, rather than the subordinate, clause. 14Meillet (1912) describes ‘grammaticalisation’ as one of the two processes by which grammatical forms evolve (he identifies the other as ‘analogie’). For a recent discussion on grammaticalization in Ancient Greek, see Cristofaro (1998). 4 Summary of the argument The argument is organized in seven chapters, grouped in three sections: I intra-clausal word order, II initial focus, and III inter-clausal linking. I: Intra-clausal word order Chapter 1: An examination is made of the order of subject and verb in poetic texts, including a comparison with previous studies of prose order, to discover whether genre has a significant effect on word order. Poetic and prose order are comparable, though the level of variation between texts is high in both. No syntactic constraints are observed. Chapter 2: Morphological determinants of order are examined. It is found that the order of subject, verb and object varies according as the verbal argument is pronominal or nominal, and also depends on the size of nominal arguments. Increasing word length is associated with SV and OV, and prosodic prominence with VS and VO. The latter is inferred from metrical structure: constructions with decreasing order are associated with a disyllabic word at the line end. Word size and the disyllabic ending are combined in a tendency of increasing ‘phonological weight’. Chapter 3: A study is made of a regular hyperbatic pattern, here termed ‘phrasal tmesis’, of a demonstrative-noun combination separated by a verb. In its two distinctive features, a phrasal constituent extending over the second colon of the trimeter line and a disyllabic line ending, it may be seen as a poetic exploitation of rightwards phonological weight, and also demonstrates formal similarities between intra-clausal word movement and subordination. It shows how the tendency towards rightwards weight has been exploited by ancient authors. 5 II: Focus Chapter 4: The prominence of words at the clause start (in P1 position) is attributable to their position and prosodic isolation from the body of the clause, and to the function of light words in P2.15 Enclitic particles emphasize the initial element, by cohesive focalization: forming part of the initial word grouping, so adding their morphological weight to it. The cohesive focalizers include emphatics, connectives, indefinite pronouns, and adverbials. Interrogative and relative pronouns, however, are separated from the initial element by an intonation break, but also mark it as being focal. P2 is, then, composed of two parts, divided by the intonation break. The structure of the whole sentence may be schematized as in Fig. 1 (=Chapter 4, Fig. 4), with the P2 groupings placed on either side of the main intonation break:16 [ { P1 } cohesive focalizers] [ interrogatives and relatives [personal pronouns {clause} ] ] What is new about this analysis is that the functions of P1 and P2 elements are integrated with their prosodic relationship: the mechanism by which co- ordinating particles, for example, link clauses is by focalizing the initial element in their clause, so connective and emphatic particles are interpreted as belonging to the same semantic group. The ‘meaning’ of individual particles cannot, of course, be reduced to their focal function, but it is argued that particles with a wide variety of meanings all function in a prosodically analogous way. Cohesive focalization is a structural as well as a prosodic relationship, and marks the inter-clausal link in adverbials and conditionals. III: Clause linking Chapter 5: The pattern of focalization in relative and completive clauses is shown to form a developmental sequence. The development from free relatives to o{ti-complements in Homer involves two distinctive features: focalization at the clause join, with o{" te, o{sti", and o{ti, and a regular association with verbs of speech and cognition. The relative pronoun functions as main verb object, so an indirect question becomes complement- like when it is neuter (‘say what you mean’). The change to a complement may 15Discussion of studies of P2 enclitics, from Bergaigne (1877) to Rijksbaron (1997a), may be found in Chapter 4. 16The P2 placing of personal pronouns and enclitics is discussed more fully in Chapter 4. 6 be seen in the use of a cognitive main, but not subordinate, verb, with o{ti functioning as an argument of the main verb only (‘I know that x is y’). Chapter 6: Explicit objects in wJ"-complementation serve the same function: the Homeric adverbial clause becomes a completive by the addition of main verb objects, which are regular throughout complements in the corpus. In the corpus, complementation remains prosodically focal, so is a transitional, circumstantial, form. This aetiology suggests that participials rather than accusative and infinitive complements are the precursors of finite complementation, which developed through a convergence of indirect questions with o{ti following speech verbs and adverbial wJ"-clauses following cognitive verbs (‘I know you, how you are’), into transitive circumstantials with either complementizer. Chapter 7: The poetic form and discourse functions of complements are linked with the focalization of the inter-clausal link. Metrical regularities in the positions of the complementizers accord with it, as does the use of complementation evident throughout the corpus: as a rhetorical tool rather than simply as indirect speech. Comments and theory The major points of the argument are summarized, and technical terms defined, in the sections below. Some details of linguistic structure are illustrated by diagrams, which are also reproduced in the body of the text: 1: Theoretical approach 1a: Subordination 2: Key concepts 2a: Weight 2b: Focus and focalization 2c: Subject, topic, theme, topicalization 2d: Topic position and the CG clause 2e: The intonation break and the clause break 2f: Free relatives and cohesive focalization 2g: ”W"-complementation and explicit objects 2h: Prolepsis 7 3: Aetiology of complementation 4: Justification of the hypothesis of focal linking 5: Metrical constraints 6: Textual scope 1: Theoretical approach The approach is structural, but largely informal, owing a constant debt to Jespersen (1924, 1937), Dover (1960) and Lyons (1968, 1977). Works on Greek morphology and syntax by Jannaris (1897), Riemann and Goelzer (1897), Kühner (1904),17 Meillet and Vendryes (1927), Smyth (1956), Chantraine (1958, 1963), and Liddell and Scott (1968), are cited throughout. In the diagrams, structural details are described using the terminology of X' [X-bar] syntax, in which grammatical relations are visualised as a projection of underlying logical form, with hypothetical movement to explain the observed word order.18 While the existence of other structural and functional approaches is noted,19 the X' model enables the relationships between word order, clause linking, and logical form to be described rigorously. Only the general principles of the X' model are adopted. Its central assumptions are that syntactic relations may be described structurally,20 and that phrases may be defined as endocentric.21 Figure 2 (adapted from Jackendoff 1977: 17) shows the configuration of the phrasal units of specifier, head, and complement, which define the structural relation between the head 17Usually cited as Kühner-Gerth (1904). 18As by Jackendoff (1977), Chomsky (1981, 1986, 1992), Webelhuth (1995) and others. 19The phrase structure grammars of Gazdar, Klein, Pullum and Sag (1985) and Pollard and Sag (1988) are similar to the X-bar model in being based on constituency, while the word- based grammar of Hudson (1984) concentrates on grammatical dependency relations. The drawbacks of functional models are discussed in Chapter 4, section 1a. 20Summarized as the principle of transparency, which assumes that logical and phonological form have a high degree of correspondence (see Lightfoot 1979: 121–141). 21The term, to describe phrases with a head word governing the other elements, originates with Bloomfield (1933). See discussions by Lyons (1977: 389–394), Jackendoff (1977: 7–27), and Radford (1988: 259ff.). Exocentricity (classical bahuvrīhi) is interpreted as a transformation: that is, movement to a position adjoined to the clause-structure. See Radford (1988: 545–8). 8 word and the other elements in a phrase:22 Specifier Head Complement Maximal Projection Intermediate Projection (XP) (X') Figure 2 In this framework, the head is defined structurally, as the central element of the phrase. While the possibility of other definitions is recognized,23 the basis of the schema is that words and phrases may be structural sisters.24 In Fig. 2, the pattern is shown as right-branching, but is reversible: there is no a priori reason why specifiers should not follow a complement-head pair, producing a left-branching structure, and the variable placing of some modifiers in CG (as the variable demonstrative position in NPs cited in Chapter 3, Section 6a) implies that this does occur.25 The basic clause is analysed as an Inflection Phrase (IP), with the verb inflection as its head.26 Preposed elements and complement-introducing conjunctions are placed outside the IP, in the Complementizer Phrase (CP) as in Fig. 3.27 If phrases are shown without an intermediate level, they are schematized as triangles, so specifier 22The semantic relationship of a complement to its head may be as a functional argument, or a (restrictive or non-restrictive) modifier (Jackendoff 1977: 57). 23These may be semantic (the head is the word which assigns semantic roles: as Chomsky 1982: 6); or in terms of government, that it is the controlling element in a phrase (Hudson 1984: 76); or distributional, that it is substitutable for the complete phrase (Lyons 1968: 233, 1977: 391). On their compatibility, see Lyons (1977: 392), Williams (1981), Zwicky (1985), Hudson (1987), Horrocks and Stavrou (1987), and Dwyer (1992). 24The term canonically used for elements at an equal level in the clause structure. 25This view is challenged by Kayne (1994), who argues that all syntax is right-branching, and that left branches are always produced by adjunction. 26The advantage of this analysis is that it enables the binary phrasal structure of Fig. 2 to be applied throughout clause structure. See Chomsky (1965: 106–110, 1986). 27In languages without complementation, this could be analysed as a Focus Phrase (FP): see Horvath (1986) and Kiss (1995b). The label ‘CP’ is used here, as FPs are associated with enclitic verbs in P2. An FP is more appropriate to early IE, where verbs were often enclitic. 9 and head are not distinguished. Word ‘movement’ within the structure is indicated by orthogonal lines. The identification of phrasal constituents can be ambiguous in CG. Because there is no indefinite article, and the definite article may be omitted in tragic language, a noun may ‘stand for’ either an intermediate or a maximal projection. This is particularly relevant to the discussion of hyperbaton in Chapter 3. 1a: Subordination The term ‘subordination’ is used in this discussion in two senses: a semantic and a structural one. In the examination of word order in Chapters 1 and 2, any clause in a semantically dependent relation to another is categorized as subordinated. In Chapters 4–7, subordination is considered more narrowly, as a structural category, with a linking word which defines the relation in which one clause modifies a constituent in another (as relatives and adverbials), or complementation, where one clause is a constituent in the other. Since Chomsky (1957) this has standardly been visualized as a recursive relation. Structural subordination may be contrasted with four other relations:28 i) Co-ordination: a relationship between two clauses of equal status, with a link marking the relation, which may be a single word (‘and, but, or’) or a pair (‘both ... and, either ... or’). ii) Apposition: the relationship between two clauses or phrases containing co-referent elements (‘Odysseus, son of Laertes’). There may be an asymmetrical dependency relation. iii) Adjunction: the juxtaposition of an element to a clause, as may be involved in the preposing of interrogative elements (‘whati will he doi?’). iv) Correlation: having an element in each clause which serves a syntactic function within it, as well as marking its co-ordinating relation to the other. The functions may be adverbial, as in temporal correlatives (‘when ... then’), or pronominal (‘which .... that’). These categories are not mutually exclusive (see Matthews 1981: 144 and 222), and Lehmann (1989) describes them as forming a continuum of independence. Correlation, subordination, and adjunction may all express co- 28This typology is derived from discussions by Matthews (1981), Chomsky (1981), Quirk et al. (1985), and Shopen (1985). 10 referent and adverbial relationships, co-ordination and correlation are similar, and both apposition and adjunction may be seen as a type of subordination. Recursive subordination is intimately connected with the mechanism of focal linking, because both involve an element functioning in two clauses simultaneously. 2: Key concepts 2a: Weight The examination of intra-clausal word order identifies a correlation between order and two lexical factors: i) Word size, judged by number of syllables: longer words are placed later. This accords with the principle of end-weight, proposed by Behaghel (1909: 138–139), and attributed to cognitive efficiency —an explanation followed by Hawkins (1983) and Dik (1978).29 The feature accords with a link often noted between the end of the sentence and pragmatic emphasis (Delbrück 1900, Kühner 1904, Denniston 1952). ii) A disyllabic word regularly ending the poetic line. The regularity of this feature in both trimeters and hexameters (Raalte 1986: 21, 29) suggests that it is not only a metrical effect, sensu stricto, but exemplifies the prosodic prominence which has been observed to accompany the end of a period (Quintilian Institutio IX.iv.29, Demetrius De Eloc.39, Raalte 1986). While SV and OV constructions are overwhelmingly associated with factor (i), VS and VO are associated with factor (ii). This is partly explicable because CG nouns are generally shorter than verbs, but the lower connection between VS and factor (ii) in prose suggests that it is also metrically constrained. The two features are combined in a principle of phonological weight. Other things being equal, a longer word is judged to have more weight than a shorter one, and a word at a prominent position in the stichic line is judged to have more weight than one placed elsewhere. Heavier elements tend to be placed to the right. The link between morphological and phonological weight may be explained in terms of a stress component in CG.30 29Other suggested explanations are the postponing of new information (Behaghel 1929, Mallinson and Blake 1981), or the tendency for inflections to be to the right (Gil 1982). 30‘When pitch prominence is reinforced by duration, it becomes stress’ (Devine and Stephens 1994: 216). On the possibility of a stress component in CG, see also Allen (1987: 131ff.). 11 2b: Focus and focalization Focus is regularly used in functional models to describe an initial position dedicated to marking certain kinds of information as new (Halliday 1967: 204, Jackendoff 1972, Lambrecht 1994: 208) or most salient or relevant (Dik 1980: 16, Sperber and Wilson 1986: 202–217). Dik (2007: 10) describes it as 'the most salient piece of information in the clause'. It is argued in Chapter 4, section 1a, that these are subjective categories, which can lead to circularity and imprecision. The term is here used to describe the initial position P1 as occupied by presentationally-prominent elements, irrespective of the kind of information they convey: they may indeed be grammatical words. Prominence is defined as prosodic emphasis, whether by quantity or stress. The term focalization is used to describe the reinforcement of prosodic prominence by enclitics which add their weight to the initial element, or pronouns which create a contrast with it. The prominent element may be preposed from within the basic clause, but that is not always so: in subordination, the focalized word may even function syntactically in the preceding clause. It has regular and systematic links with clause structure: it is here identified with the X-bar position of Specifier of the CP (complementizer phrase: see figs.6–11). Focalization of the initial element affects clause structure in five principal ways: 1) Enclitics (cohesive focalizers) are prosodically part of the P1 word group, on the left of the main intonation break at P2. 2) Focal prominence usually involves phonological weight, with the focalized element having more syllables than the following element. As grammatical words in P1 are often monosyllables, they are often followed by more than one cohesive focalizer (eij me;n gavr, etc.: see Chapter 4, Section 3c). 3) Enclitic pronouns in P2 are prosodically part of the main-clause intonation-group, and contrast with the P1 element both prosodically (because they are separated by the intonation-break) and morphologically (because they are lighter). 4) The initial element is regularly prominent in contrast with the preceding text, as well as the following clause. 5) This use of 'focus' is in accordance with the ordinary-language meaning of the term, as 'focus' of attention. The features which make it prosodically prominent create communicational emphasis too: it may be said that ‘loud’ implies ‘important’. 12 When there is both a focalizer and an interrogative, the start of the clause may be schematized as in Fig. 3 (=Ch. 4, Fig. 2):31 3 Focus [+WH] tiv dei' s j ejmoi; levgein… kai; nu'n ta; mavssw me;n P1 IP CP C' C Obj. 2c: Subject, topic, theme, topicalization In order to justify the prosodic definition of P1, the alternatives must be identified. Some categories by which the function of initial elements is analysed are inadequate because they are defined circularly (as topic, theme, and subject —see Li and Thompson 1976: 464; Lyons 1977: 507), and assume rather than demonstrate identity between structural and pragmatic categories (as in ‘topicalization’: Emonds 1976). Structural, semantic and textual categories are distinguished here as follows: 1) ‘Topic’ is the semantic category of ‘the person or thing about which something is said’ (Lyons 1968: 335), which may be expressed in a sentence by the grammatical subject. 2) ‘Subject’ is defined by case and relation to a verb, as the nominative nominal or pronominal specifier of a finite verb inflection. 3) ‘Theme’ is interpreted as a textual category, referring to an element in a sentence which specifies textual relevance: a pragmatic and not a structural description.32 It is hard to see how a referring expression can be described as thematic, except insofar as topical referents, as defined in (1), are likely to be expressed at several points in a text. As its etymology suggests, a ‘theme’ is generally expressed by a noun. 31The citation is from Ag.598 (‘And now, for the full story, what need have you to tell me it?’). 32The semantic relation in which ‘theme’ corresponds to ‘patient’, and so is associated with the syntactic object in accusative languages, is described in Chapter 6, Section 1b. See Halliday (1967, 1968), Fillmore (1968), and Lyons (1968: 350–359). 13 4) ‘Topicalization’ is the placing of absolute-like constructions in any position adjoined to the clause structure, with no necessary connection with topic or theme.33 The central feature of this interpretation is that there is no dedicated structural position for the topic or theme, but there is a focus position for prominent elements: P1. Of course, the word expressing the topic may be focalized,34 and this may even be normal in declarative main clauses. However, no support is found in these texts for a position determined by thematic factors, or for the function of any Greek particles as topic or theme markers, as the Japanese wa.35 2d: Topic position and the CG clause The view of the Greek clause as having an initial focus position is in contrast with the common view, as expressed by Kiparsky (1995), that there was a dedicated position in early Indo-European (henceforth IE) languages for the topic, and that interrogatives are in a focus position which follows this, as in Fig. 4 (=Ch. 4, Fig. 1; adapted from Kiparsky 1995: 153, Fig. 33, with the same quotation as in Fig. 3 above):36 4 S Topic [+WH] Focus tiv dei' s j ejmoi; levgein… S" S' kai; nu'n ta; mavssw me;n Obj. This interpretation is similar to that of most functional grammars, but has drawbacks as a model of CG structure: 1) The prosody does not accord with it, because the second-position element is always less prominent than the initial one. Though interrogatives bear an accent, tiv 33This view is justified in Chapter 4. 34Also noted by Devine and Stephens (1994: 459). Dik (2007: 32-3) calls this 'contrastive topic.' 35See Bach (1971), and Bakker (1993). 36Kiparsky’s use of the phrasal categories S, S', and S'' reflects his assumption that early IE did not have a complementizer. His schema shows the topic as the specifier of the focus (a simplified version of the X-bar model). See also Hale (1996). 14 in Fig. 3 above is not prominent in comparison with the initial element, and in Chapter 4 it is argued that interrogatives are not prominent unless they are preposed to focus position. 2) In tragic complementation, the regular presence of focal elements preceding wJ" may be better modelled if wJ" is not itself focalized. 3) A regular topic> focus sequence is semantically impossible if topic is not a structural category. The phrase kai; nu'n ta; mavssw could be both topical and focal (though not, of course, ‘topicalized’ and focal). 2e: Intonation break and clause break It is argued in Chapter 4 that P2 is not only the normal site for co-ordinating particles and other words normally regarded as P2 clitics, but also for interrogatives, which are regularly preceded by focal elements (as in Fig. 2 above), and by an intonation break. The prosodic relation between a relative pronoun and its antecedent is the same as that between an interrogative pronoun and its host, though in relative constructions the intonation break coincides with the clause boundary. The structure is shown in Fig. 5 (=Ch. 4, Fig. 3):37 PP IP pro;" tou;" kratou'nta" ou}" Fig. 5 NP P2 CP P1 i[doim j ejgwv pote Obj. 37The citation is from Choe.267: ‘[Someone might tell this] to the rulers, whom may I see die one day’. 15 The prosodic relation between the head noun and relative is similar to that between a specifier and head in a CP, but the focal pattern spans the clause break,38 so the structure differs from a correlative in having a prosodic link. 2f: Free relatives and cohesive focalization Indefinite pronouns regularly introduce free (headless) relatives. It is not only the lack of a head noun which creates loss of specificity, but also the association with verbs of speech and cognition:39 a few Homeric free relatives are associated with verbs of giving or taking (Il.15.109, Od.1.316) or showing (Il.22.73), but almost all involve verbs of speech or cognition, so may be termed indirect questions. In free relatives, te, ti", and -ti all function as indefinite markers, cohesive focalizers, and adverbial links. Epic te is modelled as the head of the CP in Fig. 6 (=Ch. 5, Fig. 1):40 VP Figure 6 gnwvmenai o{" te pefeuvgoi CP IP P1 Focus NP Object C' 38In the corpus, even restrictive relative clauses generally have a discernible intonation break, unlike restrictives in English. A phonetic motivation may be involved (‘the man that broke the bank at Monte Carlo’, rather than ‘who’). On the aspirate, see Chapter 4, Section 2cii. 39A relation between indefinite reference and interrogation is observed by Dover (1960: 12), Monteil (1963: 150, 154), and Lyons (1977: 761–2), and is discussed in Chapter 5, Section 2d. 40The citation is from Il.21.609: ‘To find out who had got away’. 16 ”O" is here metrically prominent at the C2 caesura.41 It is not only subject of pefeuvgoi, but also has an object-like relation with gnwvmenai, in terms of its (missing) antecedent (‘[the ones] who got away’). This is illustrated by the inclusion of the object NP in the diagram. The structure is therefore close to a correlative, with nominative case retained (see Chapter 5, Sections 2b and 2c). A functional parallel is created by -ti" affixes. Although ti" is etymologically a pronominal and not a conjunction, it may still be modelled as head of the CP (because indefinite and linking functions are semantically similar, and ti" is not co-referent with an element in the subordinate clause).42 The construction is schematized in Fig. 7 (=Ch. 5, Fig. 2):43 VP Figure 7 o{" CP IP P1 Focus e[k t j ejrevonto ti" tw'nd j ei[h basileuv" C' Here, the function of the P1 element as main verb object is less clear (which is why the main clause NP position is omitted in the diagram), though the CP is structurally the main verb complement. The structure is closer to a direct question, though o{" still functions as subordinate verb subject, in a focalized relation with ti". It is metrically prominent at the start of the line. 41After the fourth foot: see Fraenkel (1955), and Chapter 3, Section 4. 42Similarly, the English complementizer ‘that’ is also etymologically a pronominal (and may still retain the logical form of one: see Davidson 1968). 43The citation is from Od.10.109–10: ‘And they asked who was king of these people...’. 17 The next stage in the development of complementation is the use of neuter pronouns in indirect questions, because in these cases the pronoun can be interpreted syntactically as functioning in either clause, and there is also a semantic ambiguity: ‘say [the thought] which you are thinking/ speak [the subordinated proposition]’. The structure is shown in Fig. 8 (=Ch. 5, Fig. 3):44 VP Figure 8 o{ ti au[da fronevei" Obj. CP Focus P1 IP NP C' Obj. Because in complementation o{ti is not an argument of the subordinate verb, the form is encouraged by the use of a cognitive main, but not subordinate, verb (‘I know that x is y’). The difference with respect to the main verb is that the complementizer does not only denote the textual object (the following subordinate proposition), but also deictically ‘refers’ to it.45 The prominence created by focalization is a prosodic marker which announces the following 44Il.14.195=Od.5.89: ‘Say what you are thinking.’ 45This may be termed ‘impure textual deixis’ (Lyons 1977: 668). Such a ‘sententialist’ interpretation of complementation was first suggested, for English, by Davidson (1968). See also Quine (1968) and Davidson (1979). 18 clause, as in Fig. 9 (=Ch. 5, Fig. 4), where the construction may be translated as emphatic (‘for I know well this, that you are all sick’):46 VP Figure 9 o{- CP IP P1 Focus ti eu\ ga;r oi\d j nosei'te pavnte" NP Object C' It may be remarked that the complementizer is here also prosodically highlighted by its position in the last foot of the trimeter line (a position noted above, in Section 2a, as emphatic). 2g: w{"{{{ -complementation and explicit objects The interpretation of the linking element as focalized is justified, not only on semantic grounds (the connection between interrogation and indefinite reference), but also by analogy with wJ"-complements. In the tragic texts of the corpus, a majority of wJ"-complements are preceded by explicit main verb objects. In many, the textually-deictic function is regularly performed by demonstrative objects, which themselves have a cohesively-focalizing suffix. The intonation break and clause break are coincident, and wJ" functions as a 46 OT.59–60. 19 focus marker. The structure is shown in Fig. 10 (=Ch. 6, Fig. 3):47 IP VP Fig. 10 tovde wJ" pa'" ti" auJto;n tou' pevla" ma'llon filei' Adv. V a[rti gignwvskei" CP C' Focus P1 Here, the focal element functions syntactically in the main clause, but deictically points to the following subordinate proposition. The ‘catadeictic’ demonstrative o{de is always involved in such constructions (rather than the ‘anadeictic’ ou|to").48 Monteil (1963, 251) interprets a demonstrative object as emphasizing the ‘substantival’ force of completives, but the construction is seen here as representing a development from the transitive circumstantial (‘know x as being y’), though involving a deictic element as main verb object. Pragmatic as well as prosodic emphasis may be inferred (‘Have you only just learned this, that...’). 2h: Prolepsis A close analogue of the transitive circumstantial is evident in proleptic constructions, where the main verb object is expressed by a referring expression which is co-referent with the subject of the subordinate verb. The 47Medea 85–6, ‘Have you only just learned this, that each loves himself more than his neighbour?’ 48It is also involved in the hyperbatic pattern discussed in Chapter 3. 20 structure is schematized in Fig. 11 (=Ch. 6, Fig. 2):49 IP VP tav krivna tou' ajgrou' pw'" aujxavnousin Fig. 11 CP C' P1 NP katamavqete This is, grammatically, irregular (because only one semantic role is standardly assigned to each verbal argument),50 but the anacoluthic structure seems semantically and prosodically accurate: its irregularity demonstrates the inchoate stage of complementation. Co-reference persists even in modern languages (‘consider the liliesi of the field, how theyi grow’). 51 The phonological contrast evident in prolepsis (that it almost always involves wJ" or other light conjunctions)52 is central to the interpretation. Prolepsis is a regular feature of CG, as in the oi\dav se o}"/wJ" ei\ construction,53 and it is argued in Chapter 6 that it exemplifies the developing structure of complementation, rather than being simply a stylistic curiosity. 49The citation is from Matthew 6.24: ‘Consider the lilies of the field, how they grow’. 50The q-criterion: see Chomsky (1981: 36). Q-roles are defined in terms of causality or agency (Tesnière 1959, Fillmore 1968), or of spatial or temporal goals (Gruber 1976, Jackendoff 1983). 51The case is ambiguous here, but in other proleptic constructions it is clearly accusative, as at Il.2.409: h[/dee ga;r kata; qumo;n ajdelfeo;n j ;j ;j ; wJ" ejponei'to (for he knew in his mind his brother, how he was troubled). 52As at Eum.587 and Med.39 (eij); Frogs 41 (mhv). 53As Il.9.527–8, Eum.454. 21 3: Aetiology of complementation As evident from the examples in the previous section, there is an ambiguity in the position of the linking element, as being in main verb object position with a following conjunction, or in subordinate focus position (P1) with a particle in P2. It is proposed that ambiguity reflects the origins, and that there is a prosodic and functional parallel between focus and object, and between P2 and the complementizer position. All complementation requires a reporting verb with the linking word as object. There must therefore have been an expansion in the transitivity of these verbs, from taking only exophorically-referring objects, to textually- referring ones. It is generally assumed that the development of complementation involves a historical sequence of main verbs: of emotion> cognition/ perception> speech (Chantraine 1963, Monteil 1963). The aetiology developed in Chapters 5 and 6 suggests some modifications of this sequence: 1) The development of o{ti-complements from free relatives required an indefinite object constituting a focal link, and a main verb of speech, since the meaning of an indirect question is semantically intermediate between a relative and a complement (‘they asked what troubled him’), as the interrogative can be interpreted as functioning in both clauses. Only with a cognitive verb, however, is the structure unambiguously completive. 2) Complementation with wJ" developed from Homeric interrogative clauses following cognitive and speech verbs (‘know how’ and ‘tell how’), to a transitional stage circumstantials following speech verbs, either with indirect objects (‘singing about Ares and Aphrodite, how they lay together’), or following cognitive verbs, with direct objects (‘knowing x as being y’).54 The transition is exemplified by proleptic constructions. 3) Main verbs of emotion are less important to the structure, because they do not appear to be earlier than the others, and in fact increase in frequency into the fourth century. The causal meaning of o{ti, ‘because’, may in fact derive 54The expression ‘cognitive verb’ is used to denote verbs of perception and judgment. 22 from completives with o{ti, so ‘rejoice because’ post-dates ‘rejoice that’, rather than the converse.55 4: Justification of the hypothesis of focal linking The proposed relation between the prosody and the syntactic structure of the inter-clausal link is justified in two ways: in terms of its explanatory power, and because it makes a testable prediction. It provides a unified explanation for a number of features of CG word order and clause linking: i) A structural difference accompanying the order of conditionals. ii) The regular position of relative and complement clauses following their main clause. iii) The similarity of free relative constructions with o{sti" and complements with o{ti. iv) The high frequency of verbs of speech in Homeric free relatives. v) A functional parallel between the affix of indefinite/interrogative pronouns and epic te. vi) The use of two distinct complementizers, o{ti and wJ", and their variants. vii) The high frequency of main verb objects with wJ"-complements, and the association of proleptic constructions with wJ". The proposed mechanism of focal linking also combines with the intra-clausal principle of phonological weight as a unified explanation of word order, as the ‘heavy’ element at the end of one clause is in emphatic position in the other. The analysis also provides a structural motivation for VO order (since an element which is focalized in the subordinate clause is likely to follow the verb in the main clause), which accords with a transition between an OV and a VO order, and so might be added to the proposed motivations for a ‘rightwards drift’ in IE languages.56 55Monteil (1963) holds the second view, and Cristofaro (1998) the first. 56See Ross (1970, 1973), Lehmann (1973, 1974, 1978, 1986), Vennemann (1974, 1984), Watkins (1976), and Bauer (1995). 23 Testability. The hypothesis makes a testable prediction: that, in CG, there is no regular initial emphatic position within the syntactic structure of subordinate clauses which follow their main, and that preposed elements function syntactically within the main clause. No evidence to the contrary is found in the texts studied here:57 if it were found in contemporary texts, the hypothesis would require serious revision. It is, of course, likely that as the syntactic structure became regular, so it also became less prosodically prominent. As wJ"-complementation became common, so it omitted main clause objects except in emphatic constructions, and the complementizer o{ti also came to bear progressively less emphasis. The circumstantial is the inchoate form of complementation. 5: Metrical constraints The discussion concentrates on poetic texts, but the following presumptions may be made about the relationship between poetry and prose: 1) Poetic language is based on the same prosodic principles as prose.58 2) The same syntactic rules apply in poetry and prose, even if the exploitation of these rules is subject to specific constraints in different genres. 3) If prosody reflects syntax,59 then metrical form, which constitutes particularly visible prosodic constraints on language, may help identify underlying syntactic structure. This does not imply that poetic and prose order will necessarily be the same: constraints on rhythm also constrain word placing, yet all language has a rhythmic component, and Greek metrical patterns are likely to represent ‘a stylization or normalization of the natural rhythm of language’.60 In order to determine the extent of constraints, prose texts are used as controls, and 57The possible counter-examples are discussed in Chapter 6, Section 4b. 58This assumption is made by Allen (1987: 132), Liberman and Prince (1977), Ruijgh (1990), and Devine and Stephens (1994: 100–1), and is implicit in the word groups discussed by Dover (1960: 17). See also Quintilian, Institutio IX.iv.79 on the existence of feet in prose rhythm. 59The converse is also possible: Liberman and Prince (1977) propose that constituent organization is itself motivated by prosodic stress. 60Allen (1987: 132), quoting Meillet. 24 comparisons are made with previous studies of prose order (Frisk 1932, Dover 1960 and Dunn 1988) and of Homeric order (by Friedrich 1975). In the corpus studied here, no consistent variations between epic, tragic, and prose texts are observed.61 Similarly, no consistent variations are found in comparison with previous studies of prose, so the findings on word order may be applicable to CG more generally. Focal patterns of clause linking are observable in Plato, though they are naturally less regular than in the poetic texts. A more extensive examination of fifth century prose texts would be needed to determine their extent. 6: Textual scope The discussion of subordination draws extensively on the Homeric and tragic corpora: all subordinating instances of o{ti in Homer and in tragedy are examined, and the frequency of other complementizing conjunctions is analysed. A systematic analysis is undertaken of all finite complementation in a corpus of texts, constituting a text base of 11,343 poetic lines, plus two prose texts as controls. The texts chosen are: Iliad 9, Odyssey 9, Septem, Agamemnon, Choephoroi, Eumenides, Oedipus Tyrannus, Medea, Cyclops, Frogs, Thucydides History 5, 85–113 (the Melian Dialogue), and Crito.62 This data-set is also used for the systematic study of word order in Chapters 1–3. These texts are chosen so as to concentrate on tragedy, yet also to identify genre differences with epic and prose. The corpus is not intended to be a representative sample of tragic texts,63 but to provide the most varied sample. It includes the Oresteia, as the only extant trilogy, one tragedy by Sophocles and one by Euripides, and the only complete extant satyr play. Septem is included in order to determine whether the unusual features of word order observed in the Oresteia reveal a general trait of Aeschylean style.64 Frogs is 61Differences between order in lyric and spoken passages in tragedy are discussed in Chapter 1, Section B4. 62Traditional titles of ancient texts are given. While this may result in inconsistencies, the justification is advanced that even transliteration of Greek forms would not be ‘authentic’, because many titles are derived from secondary sources. 63This would be an impossible aim, since it could represent only surviving texts. 64The high proportion of Aeschylean language in the corpus reflects the centrality of his work in the tragic genre. 25 chosen as the comedy because it combines informal style with explicit parody of tragic language, which may demonstrate how the high style was seen at the time. Two prose dialogues are included as controls, but not historical narrative, as it has been systematically studied, in works cited earlier (in footnotes 5 and 6). Complete texts are chosen (as by Dover 1960), rather than selections (as by Frisk 1932), in order to observe larger-scale discourse patterns as well as syntactic detail. Comparisons are occasionally made between selected texts within the corpus, when there is no reason to believe the results would differ significantly with a more extensive data-set. All figures are given with accompanying percentages, in order to facilitate inter-textual comparison. In the interest of clarity, percentages, other than those cited from other works, are rounded to the nearest integer. Part I Word Order 27 Chapter 1 The order of subject, verb, and object Summary This chapter contains an investigation of the relative order of the main sentence elements (subject, verb, and object) in the selected corpus. The principal conclusions are that poetry and prose show comparable word order patterns, but that no syntactic rule captures both the regularities and the variations from them. Chapter Sections A: Previous approaches to the study of word order are discussed, and their results summarized, in the following sections: 1: Stylistic interpretations of word order 2: Pragmatic interpretations 3: Structural interpretations 4: Word order in poetry and prose B: The order of subject and verb in the textual corpus is collated. The results accord with those of Frisk (1932), Dover (1960), and Dunn (1988) for prose, in both the prevalence of SV, and also the frequency of variation from it, demonstrating that subject-verb order is not more variable in poetry than in prose. Part B is subdivided as follows: 1: Textual data: frequencies of SV and VS 2: Collation of data by clause type 3: Textual differences 4: Genre differences: lyric 5: Summary of word order patterns C: Possible syntactic explanations are considered. Neither clause type nor clause order is shown to determine word order. The position of the object in relation to subject and verb is also examined: a principle of verb centrality is considered, but no support is found for it. No structural constraints on order are observed. 28 Section C is subdivided as follows: 1: Possible structural motivation for VS 1a: Clause type 1b: Verb preposing 1c: Questions 1d: Passives 1e: Clause order 2: Verb centrality and word order 2a: Verb centrality: subject, verb, and object 2b: Verb centrality: conjunction, verb, and subject 2c: Verb centrality: conjunction, verb, and object 3: Syntactic determinants: summary A: Previous approaches to the study of word order This section is included to present the context to the discussion, by sketching different ways in which word order has been studied. It gives a brief survey of, and bibliographic references to, previous work. A 1: Stylistic interpretations of word order Ancient writers and grammarians paid most attention to compositional techniques by which word order may be manipulated, and less to the nature of the regular order. This is presumably, as Matthews (1994: 101) notes, due to a separation between grammar and rhetoric: suvnqesi" (composition) was primarily a rhetorical concern. Some grammarians did discuss a natural order: Dover (1960: 9) cites Dionysius of Halicarnassus (De Compositione Verborum 5) as thinking there was a natural criterion, ta; ojnovmata tavttein pro; rJhmavtwn (to put nouns before verbs), because substance should precede accident. However, the passage (De Comp. 5.17–18) also decides against such an order (in a sentence which exemplifies its argument): 1) piqano;" oJ lovgo", ajll j oujk ajlhqh;" e[doxen ei\naiv moi The argument [is] persuasive, but did not seem correct to me. 29 Demetrius (De Elocutione 199) considered there to be a natural order, fusikh; tavxi", with what might be defined as the element expressing the topic, to; peri; ou|, preceding.65 And yet, at 200 he writes (again iconically): 2) givgnoito me;n ou\n a]n kai; to; e[mpalin ... ouj ga;r pavnth tauvthn dokimavzomen th;n tavxin Of course the order might be reversed ... we do not absolutely approve the one order. Rhythm was also frequently mentioned: Dionysius believed composition to be based on stylistic principles of rhythm and period. Similarly, Cicero (Orator, 54) cites ‘numerus’ as the criterion. However, these principles are open to a variety of interpretation, as analyses of the opening of the Republic by a number of writers, both ancient and modern, illustrate.66 At one point in De Elocutione (21), Demetrius describes the opening sentence as a dialogic period in which the elements show little regularity, ejpevrriptai ga;r ajllhvloi" ta; kw'la ejf j eJtevrw/ e{teron (for the members are flung each upon the other). Later (De Eloc.205) he describes the opening as composed of trimevtra kovmmata, and makes a general link between the structure of ijscnov" carakthvr (the plain style) and the iambic line. Quintilian (Institutio 8.6.62–65) attributes the choice of order of the first four words (from order ‘ad necessitatem’) to rhythm, as ‘Nec aliud potest sermonem facere numerosum quam opportunata ordinis per mutatio’ (it is impossible to make our prose rhythmical except by artistic alterations in the order of words). Weil (1869: 57) defines the passage as a ‘descending construction’, in which governing words precede the governed (a principle which is considered further in Chapter 2, Section B4a). Denniston (1952: 41) analyses the first eight words as two equal commata: katevbhn cqe;" eij" Peiraia' | meta; Glauvkwno" tou' jArivstwno", of which the first has a symmetrical pattern of two monosyllables flanked by trisyllabic words.67 Cadence and rhythm were frequently discussed as important to the meaning of the sentence, yet no generally agreed structural principles were proposed. 65A connection with the syntactic subject may be implied, but cannot be proved. See the Introduction, Section 2c, and Chapter 4, Section 1. 66The passage is of Plato, Republic (327A): Katevbhn cqe;" eij" Peiraia' meta; Glauvkwno" tou' jArivstwno" proseuxovmenov" te th'/ qew'/ ...... ‘I went down yesterday to the Piraeus with Glaucon the son of Ariston to make my prayers to the goddess ...’ 67One might also note a rhythmic contrast, in the syllabic inequality of the commata, which creates a sense of acceleration. 30 The only generalization which was universally implied (though not explicitly stated) was a link between government and proximity.68 This was usually discussed in terms of the exceptions. ÔUperbatovn (hyperbaton) is mentioned or described by Plato (Protagoras 339B–343E), Aristotle (Rhetoric III.v.2), Longinus on the Sublime 22, Dionysius of Halicarnassus (De Thucydides 31.27, 52.22), Quintilian (Institutio 8.6.62–65), and Philodemus (Rhetorica 1.160S). In Protagoras (339B – 343E), Plato has Socrates develop an argument based on the possibility that the adverb ajlaqevw" in the sentence (from an ode by Simonides) a[ndr j ajgaqo;n me;n ajlaqevw" genevsqai calepovn (it is [...] hard for a man [...] to become [truly] good) is an example of hyperbaton, being ojrqw'" ejp j ejscavtw/ keivmenon (properly placed at the end). He does not define the meaning of ojrqw'", but it presumably implies a relation between sense and the proximity (or adjacency) of the words. Other discussions of composition make the same assumption of proximity. Philodemus (Rhetorica 1.160S) considered that the interval between phrasal elements must not be too great, but again order is not mentioned. At Rhetoric 1407a26–30, Aristotle describes a construction in which ejgwv is separated from its verb by too great an interval as ajsafev" (unclear), both on the grounds of correct ordering of protasis and apodosis, and of proximity between subject and verb. A specifically poetic word order was not identified. Dover (1997: 96–112) describes how Greek rhetoricians and grammarians generally distinguished poetry and prose not by order but by lexical choices, such as poetic words, absence of the article and of prepositions with locative datives, and the use of attributive and compound adjectives (though Dover observes that such features also appear in prose).69 A 2: Pragmatic interpretations 68Structural linguistic theory makes the same assumption, formalized in the ‘transparency principle’, as noted in the Introduction, Section 1, footnote 20. 69For the use of lovgo" and its variants to identify prose, see Dover (1997: 185–6). The crucial distinction may be between sung and spoken language: see Dover (1987: 1–15) and Section B4 below. 31 In the twentieth century, the search for a structural clause model has led to the relative order of subject, verb, and object in prose receiving most attention. Frisk (1932: 14), Denniston (1952: 43) and Dover (1960: 25) all describe the normal declarative order of Classical Greek prose as subject before verb. However, just as the motivation for SV is not fully established, so the high degree of variation from it has remained somewhat puzzling. Using a text base of selected passages of prose texts, Frisk (1932: 16) finds percentages of between 64–87% of SV as a total of [SV+VS]: an average (by texts) of 76%, and a variation of 23% (though if only classical texts are included, the variation is 71–87=16%). Denniston (1952: 44) concentrates on logical and rhetorical factors rather than syntactic ones, since ‘The grammatical order of precedence is modified at every turn by the claims of logical coherence and of rhetorical emphasis: and these factors, again, at every turn conflict with one another’. Denniston analyses that conflict in terms of hyperbaton, period structure and proportion, different types of antithesis (anaphora, chiasmus), and the repetition of words between clauses. Variation is described psychologically, as ‘a love of pattern-weaving for its own sake’, which Denniston (1952: 59) attributes to Plato. Dover (1960: 67) notes a similar ‘desire to achieve variety’, in Herodotus. However, Dover (1960) also suggests a number of general principles, in his systematic analysis of the interaction between logical, syntactic, and stylistic determinants of word order in three prose texts (Herodotus 3.61-87, Lysias 12, and Plato, Laches), in which words are categorized as prepositives, postpositives, or as mobile. The proportions of SV out of total subject and verb clauses in Dover’s texts are: Herodotus 3, 59%; Lysias 12, 83%; Laches, 74%.70 Dover identifies four general principles (1960: 65): i) Indispensability to sense: essential elements (‘nuclei’) tend to precede optional ones (‘concomitants’). ii) Demonstratives are preferentially prepositive (so precede the verb, whether they are subject or object). iii) Verbs used as copulatives are rarely initial. 70These percentages are derived from the totals given by Dover (1960: 29). 32 iv) Dispensable subjects may be expressed through the verb inflection, while nominal subjects are usually essential to the sense (and so, by principle i, precede). Dover’s categorization of a word as nucleus or concomitant is based on predictability, judged from the context, so is similar to the comment and topic model, but makes a converse claim, since topic is normally considered to precede comment. However, Dover’s principle (iv) suggests that even thematic nominal subjects would precede the verb. All four principles motivate SV, while only (ii) motivates OV (and even then permits hyperbaton about the verb), so there is a much stronger motivation for SV than OV, as would be expected from textual observations. The factors combine structural and discourse criteria: subject-first is in Dover’s system a discourse phenomenon, while verb and object order is a feature of clause structure. Variations are attributed to authorial choice. A pragmatic interpretation of structure necessarily requires the pragmatic principles to be identified, yet they have been formulated in radically different ways. It is possible that new information is postponed because it is cognitively most difficult to process, as suggested by Behaghel (1929). This order accords with the Prague model of functional sentence perspective, in which theme is understood as preceding rheme.71 However, it has also been suggested that thematic words might be delayed, either because more urgent information precedes (Givón 1983: 20), or because postponement creates emphasis: Denniston (1952: 46) observes that ‘often an emphatic word placed at the end of a work, or of an important section of a work, strikes the keynote of the whole thought.’ He suggests that such a keynote might correspond to Aristophanes’ kefalai'on rJh'ma (Ra.854).72 Similarly, Fraenkel (1950: 39) notes that the first 19 lines of the Agamemnon parodos consist of one period ending (at Ag.59) in jErinuvn, ‘a word heavy with meaning placed at the very end.’ This is linked by Fraenkel to a ‘tendency of archaic narrative not to display at the outset the most important uJpokeivmena in their entirety, but to introduce the elements at the moment when they give rise to a new element in the story’. 71See Firbas (1964), and the functional grammar of Dik (1978, 1980, 1989). 72A possible connection with the gri'fo" (riddle) is noted in Section C2a. 33 The possibility that an element may be placed in two quite different parts of the clause weakens the argument for pragmatic motivation, unless one position can be shown as regular and the other as emphatic.73 However, the regularity of an emphatic position at the start of the CG clause, recognized by all commentators, weakens the view that there is a parallel between ‘regular’ and ‘unemphatic’. Further, clauses do not always package information into known and unknown categories: as Dover (1960: 38) notes, the opening sequence of the Republic has no clearly thematic element.74 One would, then, expect it to have an unusual clausal trajectory, but it has been more often quoted as typical of the CG clause (as by the authors cited in Section A1). Even more importantly, textual relevance is not necessarily expressed explicitly. The work of Strawson (1952), Karttunen (1973), Sperber and Wilson (1986), Grice (1989), and others has shown that relevance is standardly implied in underlying presuppositions or implicatures, rather than being always marked by specific referring expressions. The identification of thematic elements appears to be inescapably subjective. While underlying pragmatic motivation is inevitable (because all languages are, presumably, structured to maximise communicational effectiveness), it gives a rather general interpretation (an absolute constraint on the positioning of information within the clause has never been suggested). The possibility that prepositives may be placed initially for presentational rather than informational reasons is considered in Chapter 4. A 3: Structural interpretations of word order Dunn (1988) adopts a more general approach (in a smaller textual sample, of Herodotus 1), analysing the order not only of words but also of phrases and finite clauses, through a structural model of syntactic dependence, in which elements are categorized as either head or modifier.75 Dunn observes a number of regularities: 73Canonically categorized as ‘unmarked’ and ‘marked’ (see Lyons 1977: 503). 74Horrocks (1983: 103) considers that the first sentence of a text can never be a ‘topic.’ 75These terms are defined in the Introduction, Section 1. 34 i) Finite subordinate clauses normally follow the main, except for temporal and conditional clauses, which generally precede. ii) Participial constructions normally follow the main, except for genitive absolute, datives, and nominative aorists. iii) Noun phrases (NPs) generally precede the verb, though those functioning as complements of copulas follow. Objects, instrumental datives, and accusatives functioning as subjects of infinitives show random ordering. Dunn’s findings for word order include the observation (1988: 75) that subjects of finite verbs precede their verb with approximately the same frequency (71.31% of total [SV+VS]) as do indirect objects and manner adverbs, but direct objects precede or follow with nearly equal frequency. The SV figures differ somewhat from those in the prose texts studied by Dover (1960), being significantly higher than Dover’s figure for Herodotus 3 and lower than the figure for Lysias 12, but are quite close to those in the poetic texts studied here, as summarized in Section B below. Dunn’s results show that regularities in clause order are statistically much more significant than those in word order: percentages are typically in the 80s or 90s, and even the most variable, relative clauses and dependent infinitives, have an average regularity in the 70s. However, the unification between clause order and word order which Dunn attempts to achieve by collating his data in terms of modifier and head ordering seems unproved. Dunn’s conclusion (1988: 78) is that, since 33.33% of the modifiers tested normally precede the verb, while 44.44% follow, this demonstrates that ‘from the point of view of modifier/head placement the Greek sentence emerges as verbicentric, i.e. having the verb at the centre with modifiers on either side’. This is equivalent to the Mittelstellung observed by Kieckers (1911), and adopted by a number of analysts in the functional tradition (as Dik 1995: 12). By itself, such a conclusion is incomplete, since it shows only that some verbal modifiers precede the verb and some follow, but pays no attention to the co- occurrence of both, or to the possible semantic or pragmatic motivations of verb centrality if it does exist. There are also more general reasons to doubt the generalization. As Frisk (1932: 24) notes, verbal Mittelstellung is a feature of Hellenistic Greek, and to 35 categorise CG in the same way fails to explain the high level of OV order.76 Further, a statistical approach necessarily gives a general result, and leaves the less common order unexplained. Dunn explains the variations in terms of stylistic markedness, which creates a motivational problem: it is unconvincing to use results which do not fit a statistical pattern to argue that the pattern must be a datum from which deviations gain their force (unless a particular ratio has some inherent stability, or unless positive reasons for variation are given). Yet the goal of a more systematic explanation remains attractive, and the analysis of the textual corpus, in Section C below, includes a search for structural determinants, including that of verb centrality. A 4: Word order in poetry and prose It was noted above, in Section A1, that poetry and prose were distinguished by the ancient grammarians in large part by lexical choices. Poetry was also differentiated as having metre: Gorgias (AS B vii. 39.9), Isocrates (ix.10), Aeschines (i.141), Plato (Gorgias 502.c), and Aristotle (Rhetoric 1408b21–26, 30– 31, 1409a22–23) all distinguish poetry as being e[mmetro" (metrical), and prose as a[neu mevtrou (without metre).77 Yet prose shares the feature of rhythm: in his description of prose, Aristotle (Rhetoric 1408b21–2) considers that to; de; sch'ma th'" levxew" dei' mhvte e[mmetron ei\nai mhvte a[rruqmon (the form of diction should be neither metrical nor arrhythmic), and there is no evidence that poetic rhythm was considered fundamentally different in kind from prose rhythm. The comparison of Demetrius (De Eloc.204–205) between the length of an iambic trimeter line and an ideal prose clause suggests that poetry and prose were perceived as similar in kind. The descriptions of iambic rhythm by Aristotle as mavlista lektikovn (the best for speech —Poetics 1449a24–5), and hJ levxi" hJ tw'n pollw'n (the language of the many —Rhetoric 1408b19–20), are well known, yet he also considered the paeon as the best rhetorical rhythm (Rhetoric 1409a8–9):78 3) ajpo; movnou ga;r oujk e[sti mevtron tw'n rJhqevntwn rJuqmw'n, w{ste mavlista lanqavnein 76Kieckers categorized the position of the verb with respect to any other sentence element, not simply its arguments, so his rule is not absolutely incompatible with OV. 77Noted by Dover (1997: 183). 78The paeon may be defined as a cretic (- v -) with either long syllable resolved. Aristotle (Rhetoric 1409a10–21) discusses the contrasting rhetorical effects of initial and final resolution. 36 for alone of the rhythms mentioned, it is the only one without metre, so most easily undetected. This implies that the iambic rhythm is, by contrast, visible. Aristotle’s objection to visible metre is that it creates predictability,79 but he seems to allow this in ordinary speech (see Poetics 1449a24–5). The attribution of different rhythms to different kinds of speech also assumes an overlap between prose rhythm and metre. Devine and Stephens (1994: 100–1) agree that ‘The rhythms of Greek verse are simply more highly constrained versions of rhythms already existing in Greek speech: the rJuqmizovmena [rhythmic systems] of verse are a selection of the most amenable rJuqmizovmena of prose. The basic principles of the two rhythmic systems are the same.’ If that is so, it might be expected that poetry and prose would have the same basic word order, yet the order of the main sentence elements in Greek poetry has not previously received much attention. Only Goodell (1890: 47) seems to have contemplated the possibility of comparing poetic and prose order. Homeric word order is examined by Ammann (1922, 1924), who describes the prevalence of OV order, and by Friedrich (1975), who uses a relatively small (and not precisely specified) database: passages from Books 1, 5 and 9 of the Iliad, and data from grammars by Schmidt (1885), Monro (1891), Cunliffe (1924), and Chantraine (1958, 1963). Friedrich’s interests are typological: to ascertain whether the model of change from OV in Proto-Indo-European (henceforth PIE) to VO in Greek, as proposed by Lehmann (1973), is a tenable hypothesis. A study of hyperbaton in epic and lyric poetry by Conrad (1990) is considered in Chapter 3, but no other studies of Greek poetic word order have been undertaken. In the next chapter, poetic word order is examined in terms of word morphology and sentence rhythm. In the remainder of this chapter, data from the textual corpus is collated,80 and possible syntactic motivation for the regularities and variations is discussed. 79Rhetoric 1408b21–26. 80This work follows the practice, attested from 1807 by the Oxford English Dictionary (Supplement 4, 1972: 737), of treating ‘data’ as a collective noun. This seems justified, especially since ‘datum’ has a quite different meaning (used above in Section A3). 37 B: Collation of subject, verb and object order in the corpus B 1: Textual data. Frequencies of SV and VS Subject and verb order in main and subordinate clauses is first examined, to determine overall frequencies and patterns of variation. The data is collected in Appendix 1A, and summarized below. B 1 (a): Caveat 1. Categorization of subordination Main and subordinate clauses are here differentiated semantically, and no assumption of structural subordination is made: all finite clauses having a semantic dependency relation to another, including adverbials and correlatives, are categorized as subordinated. This includes even clauses with gavr (which are, however, collated separately in Appendix 1B, to facilitate an alternative analysis). B 1 (b): Caveat 2. Restriction to finite clauses Non-finite clauses are not included. The inter-textual variations in their number may be inferred, very approximately, from the ratio of finite clauses to the number of lines. On average, there is one finite clause every 1.2 lines in the poetic texts. The frequency is lowest in Aeschylus. The number of lines per finite verb in the poetic texts are: Oresteia 1.7, Septem 1.6, Il.9 1.2, Od.9 1.2, OT.1.2, Medea 1.2, Cyclops 1.2, Ra.1.1. B 1 (c): Caveat 3. Explicit subjects In most texts, the majority of clauses do not have explicit subjects: the average proportion of finite verbs with explicit subjects is 33%. The Homeric texts and the Melian Dialogue have the highest proportion, over 50%, while Frogs has 38 the lowest, at 18%. The figures are shown in Table 1: Iliad 9: 374 main verbs, 174=47% with subjects Odyssey 9 330 “ 174=53% “ Septem 566 “ 205=36% “ Oresteia 1,712 “ 545=32% “ OT. 731 “ 180=25% “ Medea 806 “ 182=23% “ Cyclops 452 “ 96=21% “ Frogs 976 “ 174=18% “ Melian Dialogue 59 “ 32=54% “ Crito 304 “ 70=23% “ The percentage of subordinate clauses with explicit subjects is comparable, at 31%. OT. has the lowest proportion, at 21%, and Septem the highest, at 50%, as in Table 2:81 Iliad 9: 239 subordinates, 73=31% with subjects Odyssey 9 150 “ 59=39% “ Septem 108 “ 54=50% “ Oresteia 519 “ 189=36% “ OT. 519 “ 111=21% “ Medea 361 “ 86=24% “ Cyclops 137 “ 31=23% “ Frogs 380 “ 86=23% “ Melian Dialogue 73 “ 19=26% “ Crito 207 “ 71=34% “ The figures show that there is no greater probability for either main or subordinate clauses to have subjects.82 81Subordinate clauses in the corpus are collated in Appendix 1A. 82There is however, variation between nominal and pronominal subjects. This is discussed in Chapter 2. 39 B 1 (d): Caveat 4. Exclusion of SVS The collation excludes the small number of clauses with SVS hyperbaton, which are included here for reference.83 In SVS, the head nouns usually follow the verb, so the type is more similar to VS than to SV, as in Table 3: Iliad 9 Main 19 Subordinate 5 Odyssey 9 “ 13 “ 9 Septem “ 11 “ 3 Oresteia “ 32 “ 11 OT. “ 9 “ 18 Medea “ 10 “ 4 Cyclops “ 1 “ 0 Frogs “ 5 “ 1 Melian Dialogue “ 0 “ 0 Crito “ 0 “ 0 The high level of SVS in Homer and tragedy may be noted. Phrasal hyperbaton of this type is considered in detail in Chapter 3. B 1 (e): Main/subordinate ratios Of clauses with explicit subjects, the average percentage of SV of [SV+VS] is 70%, with quite modest variation (11%, between 65–76%). Listed from high to low: Melian Dialogue 88%, Odyssey 76%, Frogs 76%, Cyclops 75%, OT.72%, Septem 71%, Ag.70%, Iliad 69%, Crito 66%, Medea 65%, Eum.65%, Choe.65%. The proportion in the Oresteia is 67% overall, with Ag. distinctively higher than the others. The proportion of SV which is in main clauses in each text is usually comparable to the proportion of main clause VS within that text, as may be seen in Table 4 (next page). 83SVS constructions are exemplified by Eum.676: hJmi'n me;n h[dh pa'n''' tetovxeutai bevlo"vvv (for our part, every arrow has now been shot). 40 Table 4: Percentage of VS of [SV+VS] Main clause Subordinate clause Average 28 27 Iliad 9 33 26 Odyssey 9 22 20 Septem 29 27 Oresteia 27 48 OT. 29 26 Medea 37 29 Cyclops 24 29 Frogs 23 27 Melian Dialogue 9 16 Crito 48 24 The results may be summarized as follows: 1) Main clauses constitute 70% of total [SV+VS] clauses. This figure varies among the poetic texts by 14% (from 65 to 79%). 2) The prose texts have a significantly lower ratio of main/ subordinate clauses (main clauses constitute 40% of total [SV+VS] in the Melian Dialogue, and 59% in Crito). 3) SV clauses constitute 71% of [SV+VS] clauses, with quite modest variation (11%, between 65–76%, excluding the Melian Dialogue). 4) The poetic level of SV is comparable to Dunn’s figure for Herodotus I (71.31%), to Dover’s figure for Laches (74%), and to the average for Frisk’s prose texts (76%). 5) Order is highly regular across every type of clause, with variations at a generally comparable level. VS is always less common than SV, except in subordinate clauses in Eumenides (where it constitutes 55% of [SV+VS]). 6) The prose texts differ between themselves: Crito has a percentage of SV comparable to the poetic texts, at 66% of [SV+VS], but the Melian Dialogue has much higher SV (88%). 7) VS is slightly more common in subordinate clauses (31% of [SV+VS]) than in main (28% of main [SV+VS]). 8) VS is strikingly frequent throughout subordinate clauses in the Oresteia, constituting 48% of total [SV+VS] subordinates. As noted in (5), the proportion is highest in Eumenides. 41 9) VS is equally frequent in main clauses in Crito, constituting 48% of total [SV+VS] main clauses. The overall concord between Crito and the poetic texts (and Dunn’s figure of 71.31% for Herodotus 1), the different frequencies of the Melian Dialogue, and the variations in Frisk’s and Dover’s prose texts (23% and 24%), all show that genre cannot be an absolute determinant of order: there is a constant tendency for the subject to precede its verb, yet also a high level of variation. The motivation must be either structural or pragmatic. B 2: Proportions of clause types The proportion of subordinate clauses varies by text in the following order: Septem 21% of total [SV+VS], Odyssey IX 23, Cyc.25, Oresteia 26, Iliad IX 30, Medea and Frogs 33, OT.35, Melian Dialogue 45, Crito 56%. An approximate correlation between style and proportion of subordination may be observed: the archaic style has less subordination (though Iliad 9 has a high level). The types of subordinate clauses which predominate in different texts do not correlate directly with variations in subject and verb order. The data is collected in Appendix 1B. The results may be summarized as follows: Conditionals: the proportion of conditionals remains approximately the same throughout the poetic texts (12–16% of subordinates), though Odyssey 9 has a lower proportion than the others (9%). The prose texts have a higher frequency, with 22% in the Melian Dialogue and 25% in Crito. This may reflect the different rhetorical concerns of the dialogues, where argument rather than description predominates. Adverbials: in Homer, tragedy, and Aristophanes, adverbials constitute about half of the total of subordinate clauses. Of these, about half are adverbials with gavr, which are often indistinguishable from main clauses with co- ordinated links. In the Melian Dialogue and Crito, the frequency of adverbials is reduced (and fewer conjunctions are used). The Oresteia has an unusually high proportion of adverbials, constituting 60% of subordinates. This is almost entirely at the expense of relative clauses: conditional and complement frequencies are, though low, close to the average of tragedy. Aeschylean clauses with gavr, however, are less frequent than in other authors. 42 The number of conjunctions used with adverbial clauses differs between texts. There is not simply a reduction over time: about 17 conjunctions are in regular use in Homer, the Oresteia, and Cyclops, while about 10 are used in Septem, the Melian Dialogue, and Crito. Medea and OT. show a greater variety (26 and 34 respectively, if prepositional phrases with relative pronouns like ejx ou| and ejf j oi|" are included, or 24 and 24 if not). Complements and indirect questions show a temporal increase, from 8% of the total of subordinates in Homer to 20% in Crito. They average about 10% in the poetic texts. The increase, and a corresponding reduction in adverbials, represents a movement from semantically causal to purely formal clause linking, as subordinate clauses come to function as the objects of reporting verbs. Relative clauses: as may be seen from Appendix 1B, the percentage of relatives is fairly constant throughout the texts, being usually in the mid 20s (it is highest in Iliad 9, at 32% of total subordinates). The Oresteia has an unusually low proportion of relatives: 16% of total subordinates. This is in contrast with all other texts, even Septem, which has a high frequency (29%). Although there is no constant correlation between clause type and subject/verb ordering, it is shown in Chapters 5–7 that clause linking in complements and indirect questions does affect verb and object order in main clauses. B 3: Subject and verb order: textual differences As shown in Section B1e, a 70/30 ratio of SV/VS is the norm in the corpus. While, as noted in Section B 2, VS is marginally more common in main than in subordinate clauses (31% as against 28% of [SV+VS]), the difference is small and there is no constant differential. VS is particularly common in main clauses in Iliad 9, Medea, and Crito, and in subordinate clauses in the Oresteia and the Melian Dialogue (where the sample is very small). There are three very marked features: i) There is a near absence of VS order in the Melian Dialogue, in both main and subordinate clauses. 43 ii) There is a high proportion of VS in main clauses (48% of [SV+VS]) in Crito. iii) There is an equally high proportion of VS in subordinate clauses in the Oresteia (48% of [SV+VS]). B 3 (a): Details of the variations from a 70/30 ratio (i) Melian Dialogue The almost total absence of VS order in the Melian Dialogue, in both main and subordinate clauses, is a consequence of the typically early placing of subjects, and postponement of verbs to the end, not only of the clause, but of the sentence. Subordinate clauses are typically placed in centrally embedded structures, which are sometimes highly complex.84 The extreme separation of subject and verb may be described as a form of hyperbaton (Aristotle’s strictures against it in Rhetoric III 5.2 are noted above, in Section A1a). (ii) Crito The very high proportion of VS order in main clauses in Crito (48%, together with the high proportion of SV in subordinate clauses: 76%), shows that VS is not simply a poetic trait, and is not only a feature of subordinate clauses. (iii) The Oresteia The other extreme variation from the norm of 30% is the very high proportion of VS in subordinate clauses in the Oresteia. Of the 513 [SV+VS] main clauses, 80% have the subject first, while in the 178 [SV+VS] subordinate clauses (if the 34 with relative pronoun subjects are discounted) SV/VS numbers are therefore nearly equal: 85 are VS, compared with 93 SV.85 84See Chapter 3, Section 5. 85If relative clauses with relative pronoun subjects are included, the figures for subordinate SV/VS are SV 126 (59%) and VS 86 (41%). Relative clauses are mainly SV if relative pronouns are included in the subject total, as would be expected, though there is one instance (Eum.7) where the relative pronoun follows its verb. 44 Such a high frequency of subordinate VS cannot be simply a consequence of subordination, as may be seen from its appearance in Crito main clauses. Possible structural motivations, such as the presence of interrogatives, verb voice, noun class of subject, or specificity of reference, are explored in Section C below (where it is concluded that the proportions of SV and VS are not caused by any syntactic factor). B 4: Genre differences: lyric The data collation shows that there are no consistent differences between epic, prose, and tragedy. The possibility remains that word order might be different in lyric and stichic poetry. Lyric passages in the Oresteia show higher levels of VS, as may be seen from Table 5: Ag.: SV 176 (inc. 57 lyric: 32%), VS 52 (inc. 30 lyric: 58%) Choe.: SV 103 (inc. 36 lyric: 35%), VS 50 (inc. 23 lyric: 46%) Eum.: SV 100 (inc. 31 lyric: 31%), VS 37 (inc. 15 lyric: 41%) The reason why lyric main clauses might show a higher level of VS is that the verb may, for some pragmatic reason, be regularly preposed. In fact, VS often appears in lyric constructions which express a general statement or maxim, as Choe.402: 4) ai|ma: boa'/ ga;r loigo;" jErinu;n ... for murder calls on the Erinys and Choe.637: 5) sevbei ga;r ou[ti" to; dusfile;" qeoi'". for none reveres the thing detested by the gods. These verbs might be interpreted as preposed in a gnomic version of existential ordering: ‘there calls murder...’. Similarly at Ag.392–4, the subject seems to be pai'", as Fraenkel (1950: 206) believes, though Blomfield (1818) suggests wJ" should be understood to precede it, and Lloyd-Jones (1979: 39) translates the construction as a simile, ‘for he is like a boy that pursues a flying bird’: 6) melampagh;" pevlei dikaiwqeiv", ejpei; diwvkei pai'" potano;n o[rnin, he is , black-clotted, condemned, since [he is] a boy [who] chases a flying bird 45 Such constructions suggest one reason why lyric might have a higher level of VS. However, in the Oresteia, the effect appears mostly in main clauses: there is no higher proportion of subordinate VS in lyric sections, and most subordinate VS is not in lyric, as shown in Table 6: Subordinate clauses Ag.: SV 49 (15 lyric), VS 41 (15 lyric) Choe.: SV 21 (4 lyric), VS 20 (8 lyric) Eum.: SV 28 (14 lyric), VS 30 (7 lyric) In the Oresteia, then, VS is associated with subordinate clauses in stichic metres, and with main clauses in lyric. On the other hand, in Medea, the high level of VS in main clauses (64 of 88 total VS=73%) is only slightly higher in lyric (12 in 241 = 1 every 20 lines) than in stichic metres (52 in 1178 = 1 every 23 lines). The high proportion of main clause VS in Crito also shows that VS is not only a feature of lyric. Again, genre does not appear to be a statistically significant determinant of subject and verb order: these differences must be due to authorial choice. B 5: Summary of word order patterns i) Subject and verb order in CG poetry is comparable to that in prose. There is no correlation between variation in the SV/VS ratio and type of text (as prose, epic, or tragic). The high level of variation is also comparable: the most extreme of the variations are the high levels of VS in Oresteia subordinate and Crito main clauses. These results accord with those in the studies by Frisk (1932), Dover (1960), and Dunn (1988), which do not reveal a distinctively different pattern of subject and verb order: Frisk (1932: 16) gives percentages of between 71–87% for SV in his classical texts, while Dover (1960: 29) gives proportions of SV order as: Herodotus 3, 59%; Lysias 12, 83%; Laches, 74%. The high level of variation even within authors is evident from their figures: for Herodotus, Frisk gives a SV percentage of 74.1,86 compared with Dover’s 59, and Dunn’s 71.3.87 86Frisk uses a large number of short passages as his data-base. For example, his data on verb and object order in Herodotus is taken from: 1. 6–36, 2.151–176, 3.118–141, 4.118–142, 5.82–102, 7.1–9 & 121–137, 8.113–144. As Frisk (1932: 18) points out, the results might vary with a different sample. 87Dunn (1988: 75). The text base is Herodotus I. 46 ii) The similarity of subject and verb order in poetic and prose texts, and between hexameters and trimeters, suggests that metre is not a major determiner of subject and verb order. Although lyric main clauses in the Oresteia have a somewhat higher VS level, a converse relation holds in Medea, so there is no consistent metrical effect. This supports the comment of Denniston (1952: 57) that ‘The Greeks stylized everything; and it is the most difficult thing in the world to point to any Greek which may be regarded as “natural” ... Probably parts of Aristophanes are the best examples of spoken Greek. Certainly the metre must have had some influence on the word-order: but, as far as one can see, not much.’88 While it is possible, as Dover (1960) does, to attribute regularities primarily to logical determinants, a structural explanation would also be advantageous, since it would give more information about the language. In Section C, possible structural motivations for subject, object, and verb order are explored. C: Syntactic explanations for variations C1: Structural motivation for VS C1 (a): Clause type A constant difference in order between main and subordinate clauses would be significant, since in some languages, such as German, the subordinate order is held to be the basic one.89 Frisk (1932: 38–39) believed that Greek relative clauses are ‘frei von Affekten’, and so demonstrate ‘natürliche Wortfolge’, and Kiparsky (1995: 162 n.2) considers this to be a cross-linguistic rule. However, the opposite view has also been proposed: Denniston (1952: 43) considers that ‘order in subordinate clauses is particularly subject to influence from the context.’ In fact, the texts studied here show no difference between word order in main and subordinate clauses. 88Bers (1984: 12), also citing Aristophanes as attesting colloquial Attic usage of the late fifth century, considers versification ‘must have caused at least some divergence from everyday language’, but points out that this itself varies. 89Proposed by Bach (1962), and the basis for much subsequent work on V2 in German main clauses, summarized by Zwart (1997). See Weerman (1989) for a theoretical overview. 47 A detailed analysis of clause types in the corpus was therefore undertaken, to explore the possibility of structural motivation. The data, collected in Appendix 1B, shows that in Homer and Sophocles, most subordinate VS is in adverbial clauses. In Euripides and Crito, other subordinates (especially conditionals and relatives) have more VS. The proportions in Frogs are about equal. The figures for Aeschylus vary: in Septem, other subordinates (particularly relatives and indirect questions) have more VS, while in the Oresteia, adverbials (and relatives, though they are few in number) have most VS, but the proportion of VS is high in all types of subordination. It may be concluded: a) VS is not caused only by some structural feature of adverbials. Although there is a high level of adverbials in the Oresteia, and a particularly high proportion of VS in those clauses, VS is high with all subordinates. Gavr clauses often have preposed words, yet contain a much lower proportion of VS than other adverbials (70% compared with 148% of SV totals). The individual conjunctions associated with the highest VS frequencies, wJ", ejpeiv, and ejpeidhv, are not numerous enough to explain the percentages. b) In relative clauses in the Oresteia in which the relative pronoun is the object of its clause, the subject tends to go on the opposite side of the verb to the pronoun. However, Sophoclean practice is different: OT. relatives are predominantly SV. The low number of relatives in the Oresteia shows that word order in relative clauses does not contribute significantly to the overall subordinate VS figures. c) Conditionals: though Aeschylean conditionals have quite high VS, the proportion is matched by Medea, which does not have a high level of VS in subordinates generally. The figures therefore show no statistical correlation between VS and type of subordinate clause. VS may be highest either in adverbials with gavr, in other adverbials, in other types of subordinates, or (in Crito) in main clauses. 48 C 1 (b): Verb preposing The possibility that preposing some element other than the object might motivate VS suggests that subordinates might be more likely to have VS, because an adverbial particle might attract the verb next to it. This type of structural focalization is discussed by Kiss (1995a) and Horvath (1995). It appears to have occurred in Vedic, Mycenaean, and early Greek (see Horrocks 1990: 36), and to have become a standard feature of post-classical Greek (Horrocks 1997: 209). The very high frequency of VS in adverbials in the Oresteia would accord with it, but the comparative rarity of adverbial VS in other authors shows that it cannot be a general structural feature of CG. C1 (c): Questions VS is not associated with questions. In interrogative constructions in the Oresteia, SV is more common than VS. Out of 196 questions, 151 are finite, with an explicit question word. The vast majority have no subject, but of the 48 which do, most are SV: 34 (in 16 of which the subject is the interrogative), compared to 10 VS (in another 4, VS constructions are headed by a question word modifying the subject, creating SVS order). Finite questions with an explicit subject (other than wh-subjects) in the Oresteia stichomythia are listed in Appendix 1C. 24 have SV order and 6 have VS, of which one is existential and one indefinite. C1 (d): Passives The possibility that passivization might be linked to VS was tested by examining passive constructions in the Oresteia. 109 out of 2,231 finite verbs are passives (=5%).90 42 have no explicit subject, and of those which do, 33 are SV, 27 VS, and 7 SVS. The VS constructions therefore constitute 45% of the [SV+VS] totals, which is very high, compared to the VS/SV ratio of all finite verb and subject constructions in the Oresteia, which is 231 out of 709 (=33%). There is clearly an association between passivization and VS. The motivation may be connected with animacy: a tendency for animate nominals and pronominals to precede inanimates was noted by Silverstein (1976: 113) in Australian languages, generalized by Mallinson and Blake 90The highest proportion of passives is in Choephoroi (41 out of 689 = 5.9%). 49 (1981: 80), and proposed as a universal linguistic principle by Tomlin (1986: 102): ‘In a transitive clause, other things being equal, there is a tendency for the most “animated” NP to precede other NPs.’ The motivation may derive from a semantic link between agent and animate, proposed by Fillmore (1968: 24), and perhaps from ergative features in early IE (see Lyons 1968: 351–378). The link with VS could, then, be that there is a tendency for the subjects of passive verbs to be late in the clause because they are inanimate. This may be so in tragedy: the SVS constructions in tragedy discussed in Chapter 3 have late subjects. It is, however, of minor statistical importance for a model of CG word order, since the proportion of passive verbs is so low. Conclusion on clause type The lack of correlation between word order and clause type is inconsistent with a motivation based on clause structure, along the lines of German V2, as noted in Section C1a. The explanation must either be stylistic, or some constant structural factor(s), realized in each text according to authorial choice. C1 (e): Clause order As noted above, in Section A 2, Dunn (1988) posited that the ordering of head and modifier might be generalized to include both word and clause ordering. In order to see whether this is applicable to tragedy, the position of subordinate clauses in the Oresteia, OT. and Medea was collated. It was found that VS subordinates are ordered as in Table 7: Preceding Interpolated Following Ag.: 9 2 30 Choe.: 1 2 16 Eum.: 4 2 24 OT.: 7 3 18 Medea: 9 2 15 50 However, SV subordinates also preferentially follow their main clauses, as in Table 8: Preceding Interpolated Following Ag.: 11 2 35 (17 excluding gavr clauses) Choe.: 4 4 13 (or 8 “ ) Eum.: 1 5 21 (or 10) OT.: 12 16 56 (or 40) Medea: 13 3 49 (or 30) Clause order therefore has no general relation with subject and verb order. Nor is there a correlation in terms of subordinate clause type. While the proportion of conditionals and other adverbials which precede or follow their main clause varies between texts, complements and relatives overwhelmingly follow. If a modifier and head polarity were universally applicable, as in the model of Dunn (1988) noted in Section A2, then VS would be more common when the subordinate clause follows the main (since Dunn 1988: 64 analyses subjects as verbal modifiers), and so would be more common in complement and relative clauses than in conditionals and adverbials. The absence of such a correlation shows that Dunn’s generalization is too extensive: a modifier and head contrast does not operate outside the domain of the clause. It must be concluded that word order cannot simply be mapped onto clause order. C2: Verb centrality C2 a: Verb centrality: subject, verb, and object C2 a (i) Collation of data The possibility that VS may be motivated by the preposing of some other element before the verb, so that the verb is central, would accord with the verbicentric model of the Greek clause, the ‘Mittelstellung’ of Kieckers (1911) noted in Section A2. Though Kieckers did not attempt to explain why the verb might be central, it could be explained structurally in terms of ‘competition’ between subject and another element for the same position in the clause. 51 In order to identify possible preposed elements, clauses containing S, V, and O are considered (it will be seen that these are quite rare). A categorization of clauses in the Oresteia, Medea, and Crito is shown in Table 9: Ag. Main Subordinate SOV 23 SOV 12 SVO 13 SVO 3 OSV 11 OSV 1 VSO 0 VSO 4 VOS 2 VOS 1 OVS 9 OVS 6 Choe. Main Subordinate SOV 13 SOV 3 SVO 6 SVO 1 OSV 5 OSV 1 VSO 3 VSO 2 VOS 3 VOS 0 OVS 5 OVS 1 Eum. Main Subordinate SOV 12 SOV 3 SVO 11 SVO 1 OSV 8 OSV 1 VSO 3 VSO 1 VOS 1 VOS 1 OVS 9 OVS 3 Med Main Subordinate SOV 11 SOV 16 SVO 8 SVO 1 OSV 7 OSV 4 VSO 3 VSO 1 VOS 2 VOS 1 OVS 7 OVS 7 52 Crito Main Subordinate SOV 4 SOV 3 SVO 1 SVO 1 OSV 0 OSV 1 VSO 0 VSO 0 VOS 1 VOS 1 OVS 3 OVS 0 These figures show: 1) SV orders are more common than VS orders, except in constructions with a preposed object. With a preposed object, both OVS and OSV orders occur, with comparable frequency. 2) The SV/VS ratio is very similar to the overall ratio for all clauses with explicit subjects (noted above in Section B 2, and repeated here, in the right- hand column, for reference). The ratios are shown in Table 10: SV VS SV% Total SV% Ag. 47+16 =63 11+11=22 74 70 Choe. 24+5=29 11+3=14 67 65 Eum. 31+5=36 13+5=18 67 65 Med. 26+21=47 12+9=21 69 65 Crito 5+5=10 4+1=5 67 66 3) SOV is substantially more common than SVO. 4) SOV is particularly common in subordinate clauses in Medea. This partially supports the generalization of Friedrich (1975: 23) noted below (Section C 2c). 5) Of the VS orders, OVS is most common, and VSO and VOS are rare, in both main and subordinate clauses. The rarity of VSO and VOS suggests that VS might motivate the preposing of a preposed object. The converse, however, is not likely (since OVS and OSV are equally common). Further, the Greek clause is unlikely to be verbicentric, if SOV is the regular order. 6) The VO/OV ratio is less similar to the overall figure for all clauses with objects (collated below in Section C2c, and repeated here, in the right-hand column, for reference). In all texts other than Choe., OV is higher when there is an explicit subject, as shown in Table 11: 53 OV VO OV% Total OV% Ag. 43+19=62 15+8=23 73 62 Choe. 23+5=28 12+3=15 65 65 Eum. 29+7=36 15+2=17 68 61 Med. 25+27=52 13+3=16 76 52 Crito 7+4=11 2+2=4 73 68 The significance of this difference is that SOV, OSV, and OVS are, together, more common than SVO, VSO, and VOS. CG therefore appears to be more strongly SOV than simply OV. The rarity of SVO and VSO is especially striking, since these orders became so common in Hellenistic Greek (see Horrocks 1997: 59). C2a (ii): Three-element VS orders in the Oresteia Details may be seen in the constructions in the Oresteia. Of VS subordinates, 18 clauses have all three main sentence elements (discounting relative pronoun objects and subjects). OVS constructions constitute over half (10) the total of VS with objects. They occur at Ag.106ff., 222, 1035–6, 1186, 1424, 1432, Choe.755, Eum.309, 597, 647. The motivation for the position of the subject appears overwhelmingly metrical, because it is always disyllabic, and occupies a prominent position in the line: i) In five constructions, it occupies the last foot of the line, as at Eum.647–8:91 7) ajndro;" d j ejpeida;n ai|m j ajnaspavsh/ kovni"vvv a{pax qanovnto", ou[ti" e[st j ajnavstasi". but when the dust has drunk up the blood of a man once dead, there is no resurrection. 91The traditional category of the foot is used as a convenient means of identifying positions in the poetic line: cf. West (1987: 5). However, feet may represent real prosodic features, even in prose (cf. Quintilian, Institutio IX.iv.79). 54 and Ag.1424–5: 8) a[rcein: eja;n de; tou[mpalin kraivnh/ qeov"vvv , gnwvsh/ didacqei;" ojye; gou'n to; swfronei'n. [you will rule,] but if a god ordains the opposite you will learn, taught late, wisdom anyway. Disyllabic subjects are also positioned in the last iambic foot at Ag.1186, 1432, and 1433. ii) In the other 5 constructions, the subject is placed immediately after the penthemimeral caesura, as at Ag.1036–7:92 9) ejpeiv s j e[qhke Zeu;";;; ajmhnivtw" dovmoi" koinwno;n ei\nai cernivbwn, pollw'n meta; since Zeus has kindly made you a sharer of lustral water in our house .. and Eum.597: 10) ajll j ei[ se mavryei yh'fo"''' , a[ll j ejrei'" tavca. if the verdict catches you, you will soon say otherwise At Ag.1035, Choe.755, and Eum.597, the subject also immediately follows the penthemimeral caesura. VSO subordinate clauses total 7. They occur at Ag.267, 392–4, 563, 970, Choe.402, 637, and Eum.420. The construction often follows another preposed element, as at Ag.563: 11) ceimw'na''' d j eij levgoi ti" oijwnoktovnon, and were one to tell of bird-killing winter and Ag.267: 12) Priavmouvvv ga;r hJ/rhvkasin jArgei'oi povlinvvv . the Argives have taken the city of Priam. 92Fraenkel (1950: 468–9) discusses the problem of which verb ajmhnivtw" modifies. 55 At Ag.970, Zeuv" is in a metrically conspicuous position following the penthemimeral caesura, echoed by yu'co" in the same position in the next line: 13) o{tan de; teuvch/ Zeu;";;; ajp j o[mfako" pikra'" oi\non, tovt j h[dh yu'co"''' ejn dovmoi" pevlei, and when Zeus makes wine from green grapes, it is cool in the house However, where VSO clauses have verbs preceding a linking particle, and so prosodically prominent, they introduce a general statement or maxim, as in the lyric constructions at Choe.402 and 637 (cited above in Section B4). The rarity of VSO is surprising, because, as already noted, it became regular in Hellenistic Greek. The evidence of the corpus suggests that it is not a default order in CG, but an unusual one, always motivated pragmatically. VOS also appears to be motivated pragmatically, by the postponing of the subject. There are only two in subordinate clauses in the Oresteia: one, at Ag. 1435–6 (... e{w" a]n ai[qh/ pu'r ejf j eJstiva" ejmh'" / Ai[gisqo" ‘as long as Aegisthus lights the fire on my hearth’), has a postponed subject, which appears prominent at the start of the next line (it has been noted, by Fraenkel 1950, that the postponement of names is common in tragedy, perhaps to create suspense).93 The other subordinate VOS construction (at Eum.12–13) has a following participial clause in apposition to the subject (also a name: pai'de" ÔHfaivstou). As with the other VS constructions, metrical or pragmatic considerations appear to motivate subject and object position. C2a (iii): OSV The possibility that the position of the subject is determined distributively, through ‘competition’ with the object, is examined by considering all OSV constructions in the Oresteia, Medea, and Crito. There are 39 such constructions, 23 of which have pronominal subjects or objects94 (including 93As at Ag.681–7, 877–9, 1436, S.El.957, and E.El.764. Fraenkel (1950: 328, 394, 677) interprets the feature as a kind of gri'fo" (riddle). However, observations of these texts suggest that the names of gods are more likely to appear pre-verbally (perhaps as their names are not NPs). 94Ag.330, 594, 1291, 1397, 1643, Choe.189, 224, 594, 953, 1063, Eum.116, 299, 459, 643, 735, Med.74, 310, 362, 546, 759, 1339, 1389, Crito 50b6. 56 two constructions, at Choe.224 and Eum.643, with both), and 16 with nominal subjects and objects.95 The presence of preposed objects, even in front of subjects, demonstrates that subjects do not compete directly for an initial position. The comparable frequency of OVS and OSV orders shows there is no automatic tendency for nominal subjects to follow when the object is preposed. Again, there is no evidence for verb centrality. Summary of S, V, and O ordering In all constructions with three elements, the motivation appears to combine pragmatic and metrical features: either the preposing of an element into a prominent position (in the line as well as the clause), or, less often, the postponing of an element (usually the subject). There is no evidence for a tripartite typology based on structural determinants. C2 (b): Verb centrality: conjunction, verb, and subject Since subordinate clauses frequently have a conjunction in first position, then subordinate VS could be motivated by the position of the verb in the clause, if it has a natural Mittelstellung. The percentages of SV and VS in subordinate clauses, discussed in Section B 2, show that VS cannot be motivated by the simple presence of a conjunction before the verb. C2 (c): Verb centrality: conjunction, verb, and object A principle of verb centrality would imply that not only VS but also VO would be more common in subordinate clauses, where there is a conjunction preceding the verb. To consider whether this is the case, verb and object order in the three texts with highest VS (the Oresteia, Medea, and Crito) is collated below. 95Ag.127, 198, 284, 320, 700, 951, 1022, Choe.334, Eum.334, 506, 529, 850, Med.1003, 1073, 1192, 1321. 57 The figures show that the proportion of VO in subordinates is similar to that in main clauses, and, further, that the proportion of VO in subordinates compared to main has no direct correlation with the proportions of subordinate and main VS. Neither figure supports a criterion of verb centrality, as may be seen from Table 12: Agamemnon OV total 140 main 112 subordinate 28 VO 85 72 13 VO/(VO+OV) 38% 39% 32% (compared with VS/(VS+SV) 30% 24% 45% Choephoroi OV total 74 main 63 subordinate 11 VO 77 62 15 VO/(VO+OV) 51% 50% 58% (compared with VS/(VS+SV) 35% 33% 45% Eumenides OV total 106 main 89 subordinate 17 VO 67 63 4 VO/(VO+OV) 39% 41% 19% (compared with VS/(VS+SV) 35% 27% 55% Medea OV total 145 main 96 subordinate 49 VO 134 112 22 VO/(VO+OV) 48% 54% 31% (compared with VS/(VS+SV) 35% 37% 29% Crito OV total 44 main 34 subordinate 10 VO 21 18 3 VO/(VO+OV) 32% 35% 23% (compared with VS/(VS+SV) 34% 48% 24% 58 It may be observed: 1) OV is the normal order in both main and subordinate clauses, except in Choe.and Medea main clauses, and is most predominant in Crito. 2) OV is usually more common in subordinate than main clauses, especially in Medea (by 18%) and Crito (by 13%). The predominance of OV in subordinates weakens the case that order might be based on a principle of verb centrality. This result is in accord with the opinion of Friedrich (1975: 23) that OV order is more frequent in subordinate than in main clauses, and with the data collected by Frisk (1932: 28–31), which shows a higher level of OV in relative, temporal and conditional clauses in a number of CG prose texts. It does not, however, support a view that OV is the basic order. Nor do the triple orders or subordinate constructions give any support to an ordering principle based on verb centrality. The observations of Kieckers (1911) must therefore either correlate with the position of the verb with respect to sentence elements other than the subject and object, or must be a stylistic feature. The latter conclusion is supported by his figures, which show that Mittelstellung varies by up to 38% in a corpus of historical and gospel texts.96 C3: Summary of syntactic determinants None of the possible syntactic determinants of subject and verb order discussed above predicts the observed distribution of variations in word order. The connection between passivization and VS noted above in Section C1 (d) may be attributable to animate-first order, but the number of passive verbs is too small to explain the figures. The figures for verb and object order also suggest no syntactic motivation. It may, then, appear that CG is a true ‘free word order’ language, with all regularities in ordering motivated by logical/cognitive or pragmatic/stylistic determinants. However, structural motivation does not involve only syntactic 96The percentages of central verbs are, derived from the totals of Kieckers (1911: 5): Herodotus 58%, Thucydides 52%, Xenophon 56%, Polybius 71%, Matt. 47%, Mark 51%, Luke 37%, John 33%, Theophanes 57%. 59 patterning, especially in a highly inflected language, where morphological and prosodic factors may be even more important in determining order. They are investigated in the next chapter. 60 Chapter 2 The presentational cadence: word order and phonological weight Summary In Chapter 1, it was demonstrated that subject, verb, and object order in tragedy is comparable to that in prose, so implying that it is not constrained by genre-specific factors. Nor was evidence found for any syntactic motivation. However, structure is not identical to syntax, and, rather than concluding that word order must be purely pragmatically motivated, the possibility that features of the words themselves determine order is examined in this chapter. The conclusions are that SV and OV are motivated by the comparative length of nouns and verbs, that VS and VO are associated with a stress component, and that a syntactic component is additionally involved in object placing. These regularities are combined as a principle of ‘phonological weight’. Possible underlying cognitive reasons for this feature are considered, and a prosodic description adopted. Chapter Sections The discussion is organized in two sections, the first concentrating on pronominal subjects and objects, and the second on the morphology of nominals: A: Pronominal subjects and objects. A direct correlation between word order and the use of nominal or pronominal subjects is observed. It appears that a typology of word order which does not take account of pronominals is incomplete, though the feature does not fully explain variations in order. The section is subdivided as follows: 1: Pronominal subjects and word order 2: Lack of correlation 3: Pronominal objects 4: Summary: subjects and objects 61 B: Phonological weight. A morphological and a rhythmic feature are found to be associated with constructions having nominal subjects and objects: i) Number of syllables. In the vast majority of SV constructions, the verb is longer than the subject, so there is an ‘ascending’ morphological trajectory. In VS constructions, however, descending sequences are equally common, though usually less than [ascending + constant] sequences. OV and VO show a similar correlation. ii) Disyllabic endings. In VS and VO, the subject or object is frequently a disyllabic word in the last foot of the poetic line. This is not wholly a metrical constraint, since it applies in both trimeters and hexameters. A third, minor, feature is the presence of appositional phrases. A subject which is modified by a following phrase tends to be last in its clause, so appositional NPs are often associated with VS. This factor does not, however, have a major effect on the frequencies, as the converse does not hold (most VS is not associated with appositional NPs). Part B is subdivided as follows: 1: Nominal subjects: syllable number 1a: The results 1b: Apposed phrases 2: Nominal objects 3: Phonological weight 4: Cognitive motivation 4a: Right-branching syntax 4b: Rightwards phonological weight 5: Prosodic motivation for rightwards weight 5a: Prosodic motivation for poetic VS 5b: Prosodic motivation for prose VS 5c: VS, VO, and a stress accent 6: Phonological weight and authorial choice Conclusion: phonological weight, prosody, and metre 62 A: Pronominal subjects and objects A correlation between pronominal subjects and SV order is observed in all texts. A clause may end with a verb or a noun, but rarely a non-emphatic pronoun other than ejgwv (often in the phrase savf j oi\d j ejgwv).1 This encourages SV order with pronominal subjects. The data for pronominal subjects, collated by main and subordinate clauses, is collected in Appendix 1D. The results, summarized below, demonstrate that the proportion of pronominals varies considerably between texts. The most striking variation is that Aeschylus uses fewer pronouns than other authors. This is clearly a stylistic choice, but it does not, overall, affect word order, since Aeschylean order is comparable with that of the other authors. However, the proportion of pronominals may differ between main and subordinate clauses, and this does have a correlation with word order, as in Table 1: Pronominal subjects Main Subordinate Iliad 9 49 [SV 47, VS 2] 18 [SV 17, VS 1] Odyssey 9 80 [all SV] 15 [SV 14, VS 1] Septem 53 [SV 48, VS 5] 8 [SV 6, VS 2] Oresteia 153 [SV 141, VS 12] 29 [SV 21, VS 8] OT. 75 [SV 61, VS 14] 49 [SV 41, VS 8] Medea 58 [SV 47, VS 11] 30 [SV 25, VS 5] Cyclops 51 [SV 44, VS 7] 15 [SV 10, VS 5] Frogs 103 [SV 86, VS 17] 35 [SV 26, VS 9] Melian Dial. 11 [SV 11] 7 [SV 7] Crito 18 [SV 11, VS 7] 34 [SV 29, VS 5] A 1: Word order and pronominals: correlation A correlation between SV and VS variation and the presence of pronominal or nominal subjects may be seen in all texts. Details may be seen from a consideration of the Oresteia, OT., Medea, and Crito. 1The placing of disyllabic subjects in line-end position is considered in Section B1b. The phrase is comparable to the ‘afterthought’ constructions discussed in Chapter 7, Section 3c. 63 A 1 (a): The Oresteia The high frequency of VS in Oresteia subordinate clauses (48% of subordinate SV + VS, as noted in Chapter 1, Section B 1e) appears to be motivated by a low number of overt subject pronouns, as pronouns are less common in subordinate than main clauses. In the Oresteia, there are 29 subject pronouns in subordinate clauses (8 VS: 2 with ti", 2 with ejgwv , and one each with ou[ti", tavde, tau'ta, suv). They are listed in Appendix 1E. This compares with 153 subject pronouns in main clauses, of which 12 are in VS constructions. Subordinate subject pronouns therefore constitute 19% of main clause subject pronouns: a very low proportion. A 1 (b): OT. The much lower proportion of subordinate VS in OT. (28% of SV+VS) compared to the Oresteia is associated with a greater number of pronominal subjects. OT. has 49 subordinate subject pronouns (8 of which are VS), again listed in Appendix 1E. This contrasts with 75 subject pronouns in main clauses (of which 14 are in VS constructions). Subordinate subject pronouns are therefore much more common than in the Oresteia, totalling 65% of main clause subject pronouns. This is particularly marked in relatives. In the 23 relative clauses which have explicit subjects, almost all subjects are pronouns. There are only 3 nominal subjects, all in extensive clauses, and two with enjambement, at OT.853: 1) fanei' dikaivw" ojrqovn, o{n ge Loxiva" diei'pe crh'nai paido;" ejx ejmou' qanei'n. [as it should have been], whom Loxias declared should die at the hands of my son. and OT.1452: 2) ouJmo;" Kiqairw;n ou|to", o}n mhvthr tev moi pathvr t j ejqevsqhn zw'nte kuvrion tavfon, my mountain Cithaeron here, which my mother and father, living, set for my tomb 64 The third construction, at OT.382, has a subject immediately following the penthemimeral caesura. It might be analysed as a main clause exclamative rather than a subordinate: 3) o{so" par j uJmi'n oJ fqovno" fulavssetai, how great is the grudge nursed for you The reason why these nominal subjects are also in SV constructions is considered in Part B of this chapter. There are two pronominal VS clauses: one existential at 296 (to be expected, as indefinite subjects are standardly post-verbal), and one chiastic, at OT.1180–1: 4) kavk j eij" mevgist j e[swsen: eij ga;r ou|to" ei\ o{n fhsin ou|to"||| , i[sqi duvspotmo" gegwv". ... for if you are the man whom he says, know that you were born unfortunate. There is therefore a strong correlation between pronominal subjects and SV in Sophoclean subordinate clauses. In all instances the end of the clause or of the line is occupied by a polysyllabic word. Examples are OT.148: 5) kai; deu'r j e[bhmen w|n o{d{{{ j ejxaggevlletai. and we came here for this, which this man proclaims and OT.171: 6) w|/ ti" ajlevxetai: ou[te ga;r e[kgona with which one may defend oneself ... Subject and verb order in Sophoclean constructions cannot be explained simply as the preposing of emphatic words, or as the placing of pronouns in P2 (see the Introduction). There seems to be a prosodic explanation in the examples above, and in patterns like OT.966, where the placing of ejgwv is the same as in many Aeschylean VS clauses: 7) klavzonta" o[rnei", w|n uJfhghtw'n ejgw;j ;j ;j ; ktenei'n e[mellon patevra to;n ejmovn… [oJ de; qanw;n] ... [or] the cawing birds, by whose teachings I was to kill my father, ... Similarly, in one of the rare Sophoclean relatives with a nominal subject, at OT.853 (cited above), Loxiva" ends the line, though not the clause. 65 A 1 (c): Medea The data from Medea shows a similar correlation in subordinate clauses. Subordinate VS constitutes 29% of total [SV+VS]: rather similar to the proportion in OT. There are 30 subordinate subject pronouns, of which 5 are in VS constructions, in contrast with 58 subject pronouns in main clauses, of which 11 are VS. Subordinate subject pronouns total 52% of main clause subject pronouns: rather fewer than in OT., but much more than the Oresteia. The only VS relative (at 228) is existential, 4 of the 7 SV relatives have pronominal subjects, and the other three have emphatic nominal subjects. A 1 (d): Crito In Crito, a strong correlation between pronominal subjects and SV in subordinates contrasts with a lower proportion of pronominals in main clauses. There are 34 subordinate subject pronouns out of 70 subordinate subjects (49%), compared to 18 main clause subject pronouns of 50 main clause subjects (36%). Only 5 of the subordinate pronoun subjects are in VS constructions. The presence of pronominals therefore correlates with the low level of subordinate clause VS (24% of total [SV+VS]), and a high level of main clause VS (48% of total [SV+VS]). The highest level of pronominal VS occurs in the rather small total in Crito main clauses, where 7 of 18=39% are VS. However, a number of constructions with pronominal subjects alternate SV and VS order. This is examined below in Section B 1b, in the discussion of appositional phrases. Frogs and the Melian Dialogue also show a correlation between pronominal subjects and SV. In Frogs main clauses, nominal subjects constitute 66 of 169=39% of the [SV+VS] total, and, of these, VS constitute 22 of 66=33%. In subordinate clauses, nominal subjects constitute 50 of 85=59% of the [SV+VS] total, and of these, VS constitute 14 of 50=28%. The proportions of VS in main and subordinate clauses are comparable. 66 A 2: Pronominals and word order: non-correlation Word order is not, however, determined simply by the choice of pronominal or nominal subjects. The proportion of pronominal VS is variable, being as high as 11/58=19% in Medea main clauses, and 7/18=39% in Crito main clauses. In both texts, this reflects a generally high level of main clause VS (as shown in Chapter 1, Section B1e). Further, the proportion of VS with nominal subjects is also variable, as may be seen from Odyssey 9 and Septem: A 2 (a): Odyssey 9 In main clauses, nominal subjects constitute 88 of 168=52% of the [SV+VS] total, and, of these, VS constitute 37 of 88=42%. In subordinate clauses, nominal subjects constitute 35 of 50=70% of the [SV+VS] total, and of these, VS constitute 10 of 35=29%. Though there are more nominals in subordinate clauses than in main, VS is much higher among main clause nominals than among subordinates. A 2 (b): Septem In main clauses, nominal subjects constitute 141 of 194=73% of the [SV+VS] total, and, of these, VS constitute 51 of 141=36%. In subordinate clauses, nominal subjects constitute 43 of 51=84% of the [SV+VS] total, and of these, VS constitute 12 of 43=28%. To a somewhat lesser extent than in Odyssey 9, VS is higher in main clause nominals, though, again, there are more nominals in subordinate clauses. Summary It may be concluded that pronominals are overwhelmingly associated with SV, but two provisos must be made: i) the pronominal SV/VS ratio is not constant between texts. ii) the presence of pronominals is not enough to explain the SV/VS ratios. The proportion of VS in nominals varies between authors, and between clause type. Before possible reasons are discussed, the relation between pronominal objects and word order is considered. 67 A 3: Pronominal objects A 3 (a): OV and pronominal objects If word order is affected by the presence of pronominals, then it might be expected that there would not only be a correlation between SV and pronominal subjects, but also between OV and pronominal objects. The proportions in the Oresteia, Medea, and Crito show this to be the case. However, there is considerable variation, most notably in the high level of pronominal VO in Medea main clauses, as shown in Table 2: Ag. Pronominal objects OV VO Main 44 35 9=20% Subordinate 9 8 1=11% Choe. Main 37 28 9=24% Subordinate 5 5 0=0% Eum. Main 31 21 10=32% Subordinate 6 6 0=0% Medea Main 74 38 36=49% Subordinate 25 20 5=20% Crito Main 34 24 10=29% Subordinate 8 7 1=13% A 3 (b): Medea and pronominal OV A consideration of Medea shows that, although there are fewer pronominal objects than subjects, there is indeed a higher proportion of pronouns with OV than with VO, but the proportion (58 of 99=59%) is far less than for pronominal subjects, given in Section A 1. When pronominal VO does occur, the pronoun always follows immediately on the verb, suggesting a structural motivation. The reason does not appear to be metrical, since line position is quite variable. It can be summarized in three categories: 68 1) Second word in the line:2 12 instances (344, 394, 476, 489, 505, 636, 692, 709), including repetitions of dravsw tavd j, 4 times, at 268, 927, 1019, and 184: 8) dravsw tavdvvv j: ajta;r fovbo" eij peivsw I shall do it, but there is doubt whether I shall persuade ... 2) Last position in the line (6 instances). Pronouns here do not seem to carry pragmatic emphasis, though they do have prosodic prominence, as may be seen at Med.311: 9) misw': su; d j, oi\mai, swfronw'n e[dra" tavdevvv . I hate him. But you, I think, did this wisely and 961: 10) [dokei'" de; crusou'…] sw'/ze, mh; divdou tavdevvv . ... keep them; do not give them. Similar patterns occur at 1057 and 1131. However, pragmatically emphatic pronouns appear in final position, as in two constructions with se which both contrast with pronouns earlier in the line, at Med.1058:3 11) ejkei' meq j hJmw'nJ 'J 'J ' zw'nte" eujfranou'siv se. living there with me, they will gladden you. and 515: 12) ptwcou;" ajla'sqai pai'da" h{{ {{ t j e[sw/sav se. ... the children wandering as beggars, and I who saved you. This construction echoes the enjambement four lines earlier, at 510–511: 13) e[qhka" ajnti; tw'nde: qaumasto;n dev se e[cw povsin kai; pisto;n hJ tavlain j ejgwjjj v, ... but in you I have a wonderful husband, and a faithful one, poor me. 2At Med.344, 636 and 709, there are second-position particles or initial ajll j, so the object is, strictly speaking, the third word. 3Lines 1056–80 are bracketed by Diggle (1984). The textual problems are discussed by Kovacs (1986). 69 3) Position elsewhere in the line (9 instances); at Med.326, 332, 351, 585, 613, 908, and 1040, and 306–7: 14) su; d j ou\n fobh'/ me: mh; tiv plhmmele;" pavqh/"… oujc w|d j e[cei moi, mh; trevsh/" hJma'"J 'J 'J ' , Krevon, But you fear me: what harm are you afraid of? It is not so with me; do not fear me, Creon In three constructions, the pronoun functions as antecedent to a constituent within a complement clause, at Med.85–86, 168, and 39–40: 15) pavscous j: ejgw\/da thvnde, deimaivnw tev nin mh; qhkto;n w[sh/ favsganon di j h{pato", ... I know her and fear her, lest she drive a whetted sword through her innards It may be concluded that there is a correlation, but a much looser one, between pronominals and OV than between pronominals and SV. The position of pronominal objects suggests that the same phonological motivation applies to object as to subject pronominals, but that there is a contrary, syntactic, tendency for objects to follow their verbs (see also Dover 1960: 18, and Luraghi 1998: 192). Clearly, head-government is much more prevalent in constructions with pronominal objects. This would accord with the diachronic movement of enclitics from P2 to adjacency to their head words, between CG and Hellenistic Greek (see Chapter 4, Section 3b). A 3 (c): Pronouns and word order: summary In all texts, subject pronouns are rarely associated with VS order, and this explains some variations of order in the corpus. The Oresteia has a very low proportion of subject pronouns in subordinate clauses compared to the other texts, and a high frequency of subordinate VS. The greater number of subordinate pronominal subjects in the OT. is matched by a very high SV frequency. Medea has a slightly lower frequency of pronouns, but an equivalent subordinate SV frequency. The low frequency of main clause subject pronouns in Crito correlates well with the high main clause VS levels. Most pronominal VS constructions involve either ejgwv, most commonly at the line end; indefinite ti"; or suv. Pragmatic emphasis is not evident: the position of ejgwv in the phrase oi\d j ejgwv seems motivated prosodically. 70 Correlation between pronominal objects and OV is less marked than between subjects and SV. The reason appears to be that object pronouns are likely to follow immediately on their verb. The position of objects, then, is determined by both prosodic and syntactic determinants, while the placing of subjects appears to be determined prosodically. B: Nominal subjects and objects B 1: Nominal subjects: syllable number It was noted above, in Section A3, that the presence of pronominals is not sufficient to explain the SV/VS ratios, and that the proportion of VS in nominals must therefore vary between authors. In this section, all clauses in the corpus with nominal subjects are examined, and subject and verb order is collated in terms of word size, judged in terms of number of syllables, and consequently categorized as ascending (longer words to the right), descending (the converse), or constant order. The data is collected in Appendix 1F, and summarized below in Table 3. The following points should be noted: 1) In order to create a consistent test, only the subject noun, rather than the phrase (NP), is considered. 2) The syllable number of noun and verb rarely differs by more than one. 3) A consideration of complete phrases may change the figures in two ways: i) a descending VS order (in terms of the subject noun) may become an ascending one, if the whole phrase is considered. ii) an ascending SV order (in terms of the subject noun) may become a descending order, in terms of the whole phrase (these are collated in Appendix 1F). Table 3: Nominal subjects Iliad 9 Total Ascending Descending Constant Main SV 57 34 15 8 VS 49 24 17 8 Sub. SV 33 14 10 9 VS 17 4 8 5 71 Odyssey 9 Main SV 51 28 12 11 VS 37 5 13 19 Sub. SV 25 14 1 10 VS 10 0 6 4 Septem Main SV 90 52 19 19 VS 51 12 19 20 Sub. SV 31 19 4 8 VS 12 1 3 8 Oresteia Main SV 231 139 44 48 VS 129 55 53 21 Sub. SV 72 35 21 16 VS 77 33 32 12 OT. Main SV 61 38 15 8 VS 35 16 13 6 Sub. SV 28 16 5 7 VS 16 9 4 3 Medea Main SV 61 36 11 14 VS 53 23 23 7 Sub. SV 34 22 2 10 VS 19 5 8 6 Cyclops Main SV 27 15 4 8 VS 17 4 10 3 Sub. SV 12 7 2 3 VS 4 2 1 1 Frogs Main SV 44 24 8 12 VS 22 8 7 7 Sub. SV 36 16 8 12 VS 14 10 2 2 72 Melian Dialogue Main SV 10 4 2 4 VS 2 1 0 1 Sub. SV 9 4 3 2 VS 3 2 0 1 Crito Main SV 15 9 4 2 VS 17 7 3 7 Sub. SV 24 10 7 6 VS 12 3 5 4 B 1 (a): Summary of results 1) Descending constructions are in the minority in all texts, compared with [ascending + constant] orders. 2) An ascending order is more common with SV in all texts, in both main and subordinate clauses, with the exception of subordinate clauses in Eumenides. 3) A few ascending SV orders may be categorized as descending if the whole phrase is considered. However, in no text does this create a majority of descending SV. 4) The ascending / descending ratio is much more even in VS constructions, descending order often being in the majority. Descending order is more common than ascending in Iliad 9 subordinates, all clauses in Odyssey 9 and Septem, Ag. and Choe. subordinates, Eum. main, Medea subordinates, Cyclops main, and Crito subordinates. However, descending order rarely outweighs [ascending + constant] together: it does so only in Odyssey and Choe. subordinates and Cyc. main clauses. 5) Many descending VS constructions include a disyllabic subject in a prominent prosodic position, most commonly, the end of the line. 6) Some descending VS becomes ascending, if the complete NP is included. 7) A few descending VS constructions are followed by phrases in apposition. This is considered further in Section B1b. 73 The most significant prosodic element accompanying VS is factor (5). If disyllabic subjects in line-end position are discounted, there is a majority of ascending VS constructions in almost all texts. The more even figures actually observed suggests that a rhythmic element may compensate for a morphologically ascending trajectory in these constructions. It may be concluded that SV is closely associated with ascending order. VS is, however, equally strongly associated with an ascending order only if a prosodic feature, of emphasis, is allowed. The same correlation is observed with verb and object ordering (see Section B 2). B 1 (b): Apposed clauses There is a somewhat higher probability of VS when there is a following phrase in apposition. In the Oresteia, a high frequency of NPs in apposition to subordinate clauses may contribute to the high VS level in the trilogy. Out of 92 subordinate VS constructions, 22 have appositional phrases following (=24%). This compares with 6 out of the 99 subordinate SV constructions (=6%). It may be connected with the low number of finite relative clauses in the Oresteia.4 Appositional phrases might preferentially follow subordinate clauses for stylistic or cohesive reasons: there can be a certain clumsiness (or at least lack of cohesion) when SV is followed by such a phrase, as at Septem 24 (the translation mimics the effect): 16) nu'n d j wJ" oJ mavnti"J vJ vJ v fhsivn, oijwnw'n bothvr, Now as the prophet says, the shepherd of birds, ... However, there is no correlation between VS and following phrases in Crito. Nine VS constructions (out of 41 = 22%) have following infinitives or participle: 4 in main clauses, and 5 in subordinates. They are listed in Appendix 1H. The VS total compares with 16 SV constructions out of 79 (=20%). A structural motivation would be possible only for the participles, which are in agreement with the subjects, in contrast with the infinitives (separated from their controlling verbs by the subject NPs). 4Noted in Chapter 1, Section B 2. 74 An appositional phrase may be included in the prosodic trajectory, and so included in a principle of increasing word size. However, the presence of appositional phrases can be only a contributory factor: the figures from the Oresteia and Crito show that most VS constructions are not followed by non- finite phrases. For this reason, a statistical study of the phenomenon is not undertaken here. Object and verb order is, however, affected by the presence of clauses following: in section B6c it is observed that complementation is associated with main clause VO. B 2: Nominal objects As with pronominals, the OV/VO ratio is more variable with nominal objects than subjects. In the five texts, VO varies between 27% and 71% of [OV+VO], and may be higher either in main clauses (in Agam, Eum.and Med) or in subordinates (Choe.and Crito), as shown in Table 4: Ag. Nominal objects OV VO Main 140 77 63=45% Subordinate 32 20 12=38% Choe. Main 88 35 53=60% Subordinate 21 6 15=71% Eum. Main 121 68 53=44% Subordinate 15 11 4=27% Medea Main 134 58 76=57% Subordinate 46 29 17=37% Crito Main 22 14 8=36% Subordinate 5 3 2=40% Verb and object order, like subject and verb order, shows an ascending order (longer words to the right). As with SV, so OV is overwhelmingly ascending. The figures for VO, however, are even more weighted towards descending order than is VS. The correlation may be seen in the Oresteia, Medea, and Crito (the data is collated in Appendix 1G). 75 It has been shown that SV and OV orders are predominantly ascending, VS is quite evenly balanced, and VO is predominantly descending. The features associated with descending VO include the same metrical determinants as with VS (disyllabic nouns in the last foot, or an extended phrase), but there appears to be an extra, presumably syntactic, motivation for the object to follow its verb. This accords with the positions of object pronouns in Medea, described above in Section A3. The proposed explanation, that SV and OV are motivated by the morphology of the words, while VS and VO have a stress component (associated with metrical prominence), is discussed below. The first factor is examined in Sections B3 and B4, and the second in Section B6. B 3: Phonological weight As discussed above in Sections B1 and B2, SV and OV constructions show a trajectory of increasing word size. This accords with the observation of Behaghel (1909, 1929), that a ‘Gesetz der wachsenden Glieder’ is evident in Indo-European languages. It has also been observed, by Frisk (1932: 44, 87, and 94) and Schwyzer (1950: 691), that, in Greek prose subordinate constructions, longer words tend to follow shorter ones. Chantraine (1952: 72) suggests that it may be a general rule of Greek for the longer term to follow the shorter. A tendency for word size to increase in English sentences is noted by Jespersen (1949: Chapter 2) and Quirk, Greenbaum, Leech, and Svartvik (1985: Chapter 14), and termed ‘end-weight’. Behaghel (1909: 138–139) initially suggested a cognitive motivation: that longer words follow shorter ones, in the absence of iconic factors like temporal or causal sequence, because a more complex task tends to be delayed. In a subsequent article, Behaghel (1929) suggested that the principle may be associated with the postponing of new information. Subsequent commentators have usually followed either the cognitive or the pragmatic motivation, and sometimes both: Mallinson and Blake (1981: 151–157) suggest that heavy elements place fewer demands on short-term memory if they are later, and that light elements, ‘typically a pronoun or a simple noun phrase’, occur at the start of the clause for reasons of textual cohesion (which they describe as topic to the left). Reasons for doubting the value of a pragmatic approach were outlined in the Introduction, Section 2c, and Chapter 1, 76 Section A2. The possibility of a cognitive explanation is discussed in the next section. B 4: Cognitive motivation B 4 (a): Right-branching syntax The notion that word order may reflect principles of logical ordering has been suggested by both ancient and modern critics. The stylistic divisions posited by Demetrius (De Eloc.12–14, 36), the free (dialelumevnh) and the periodic or compacted (katestrammevnh), were reinterpreted by Weil (1869: 51–67) as ‘constructions descendantes’ or ‘ascendantes’ according as the governing word precedes or follows the governed. Because Weil’s terminology is unrelated to the morphologically descending and ascending sequences described in Section B1, the symbols G> and , and VS and OV in Weil’s terms, while, in the X' schema, SV is emphasizes the ideas which the individual words represent (he notes its frequency in Aristotle’s definitions, and, as noted in Chapter 1, Section A1, he identifies the opening of the Republic as an example), while is easier to remember: Yngve (1960) showed that when constituents are organized in a pattern of dependency, a speaker has to remember each until the utterance is complete, and the memory load required will vary according to structure; a feature he characterized as ‘sentence depth’. Yngve’s model predicts that right-branching structures, which correspond to Weil’s G> (SV and VO), are easier to remember than ‘flat’ or left-branching ones. Though influential, the explanation has several problems: i) Psycholinguistic: it is not proven that binary-branching structures are easier to learn or to process than ‘flat’ structures (see Frazier and Fodor 1978, Matthei 1982, Slobin 1986).7 ii) Typological: the fact that left-branching languages are not markedly less common than right-branching ones implies that many languages are ‘inefficient’, which is counter-intuitive.8 iii) Diachronic: the model cannot in itself explain language change. iv) Variation: non-standard orders are not always emphatic, and the level of suspense they create does not necessarily constitute a great load on the memory. Goodell (1890: 10) noted that Weil’s model does not explain variations in order within CG, or between CG and Latin, and Yngve’s model is open to the same objection. v) Structural: The model is purely linear, and does not model optional constituents or prosodic groups (Frazier 1985: 155). A cognitive basis for word order appears attractive, as it links a structural generalisation with stylistic choice, but word order clearly involves more than the existence of right-branching syntactic structures. What does seem likely is that a consistently-branching structure is cognitively simpler than a mixed one: Kuno (1974) notes that centre-embedding structures are particularly difficult to process. The significance of this is that changes in order might be cumulative. B 4 (b): Rightwards phonological weight The connection between a cognitive explanation for right-branching and for rightwards phonological weight is in the interaction between syntax and 7Binary-branching systems may, however, be inherently more efficient: see Simon (1962). 8See Kayne (1994) for discussion of the structural implications. 78 morphology (especially evident in an inflecting language).9 Memory load is clearly applicable to large (phrasal) constituents, as ‘Heavy NP Shift’ (the rightwards adjunction of complex NPs —see Ross 1967), and extraposition (rightwards placing of subject clauses —see Jespersen 1924, Koster 1978). The most rapid recognition of even small-scale phrasal constituents may also be achieved by postponing heavier elements (see Hawkins 1983, 1990, 1994; Dik 1978: Ch. 9; id. 1989).10 Parsing efficiency suggests an explanation of why weight might be placed to the right, though analyses of textual corpora have not yet shown strong correlation between word order and constituent size (see Siewierska 1993).11 Wasow (1997) argues that a better explanation is provided by a production- based explanation: that it is in a speaker’s interest to ‘keep options open’ as long as possible, so an order of words which is less predictable to a listener might be preferred by a speaker (as postponing heavy elements generally delays a speaker’s commitment to the final structure). The prospect of a direct cognitive explanation for linguistic structure is attractive, but cognitive models have so far had little predictive power, perhaps because they concentrate on constituent ordering12 rather than word size or prosodic prominence (only Dryer 1992 has addressed the relationship between constituent and word ordering).13 In a highly inflected language, weight could be analysed as a morphological feature, as words are often full phrasal constituents. The data here accords with either view: the figures in Section B 1(a) show that the predominance of 9The relation between the two is described by Baker (1985) as the ‘mirror principle’ . 10Hawkins proposes that order is motivated by Early Immediate Constituent (EIC) recognition, while Dik (1989: 351, 369) suggests a weight-based ‘language-independent order of constituents’ (LIPOC), of [clitic> pronoun> NP > adpositional phrase> subordinate clause]. 11Siewierska examines only subject and object order, in terms of number of words in an XP. 12This may change with progress in machine-based parsing, on which see Marcus (1980), Fodor (1983), Berwick and Weinberg (1984) Hausser (1989), Berwick, Abney, and Tenny (1991), and Bunt and Tomita (1996). 13Dryer (1992) describes word order in terms of phrases and single words, associating VO with word> phrase order. This creates the problem of defining a phrase, which Dryer (1992: 112–4) restricts to full (XP) projections. 79 ascending SV is similar in both frameworks, though categorization by constituents somewhat reduces the predominance of descending VS. Single words are used here as the unit of analysis, in order to concentrate on the smaller-scale features. In either framework, inflections add rightwards weight: the nominative is often phonologically zero, and generally precedes the non-zero accusative (see Gil 1982: 133–4). The inflectional system may then motivate subject> object order, and also SV and OV, due to the relative size of noun and verb inflections. This feature is particularly relevant to the figures for pronominal order discussed above: the proportional difference between a monosyllabic and a disyllabic constituent is greater than that between two constituents with, say, 5 and 6 syllables, so pronominals are especially sensitive to case- marking rules (see Gil 1982: 134). However, the lack of correlation between VS and ascending order (described above in Section B1a) shows that more is involved in rightwards phonological weight than simply number of syllables. A prosodic component appears to be involved too. B 5: Prosodic motivation for rightwards weight B 5 (a): Prosodic motivation for poetic VS There does not appear to be a simple rightwards trajectory of increasing emphasis: the traditional view (as Thomson 1938: 18, Denniston 1952: 44) is that emphasis declines over the course of the CG clause. However, the clause end is also generally agreed to have some prominence. The conclusion must be that a purely linear model is inadequate, and prosodic and intonational groupings should also be considered. Three metrical groupings are of particular relevance: the prominence of the start of the line, a more regular metrical organisation in the second colon than in the first, and a rhythmic effect in the last foot: i) The start of the clause is universally agreed to be emphatic, and this is reflected in the typical prosodic structure of the stichic line, with a single intonation break, discussed by Fraenkel (1932, 1933), De Groot (1935), and 80 Allen (1973).14 Allen (1973: 115) describes the colon (or its delimiting caesurae) as ‘a metrical feature, based on grammar, and manifested in composition.’ Ruijgh (1990: 229–230) notes the frequency with which the line break and the penthemimeral caesura define the phonetic frontier between ‘l’expression thématoïde’ (by which he means topical expressions, including adverbials) and ‘la phrase proprement dite.’ The start of the clause is examined further in Chapter 4. ii) It is observed by Allen (1973: 106) that the second colon of the stichic line is normally defined more rigidly than the first. This appears to be a very ancient feature: West (1987: 6) notes that, in early IE poetry, quantities were free, except towards the end of the verse. Its appearance in tragic iambics is exploited at Ra.1198ff., where Euripides is criticized for the Lekythion pattern commencing at the penthemimeral caesura. A different criticism of Aeschylus, that his lyrics include dactylic refrains, is made at Ra.1264–5 (=Myrmidons 1): 17) Fqiw't j jAcilleu', |tiv pot j ajndrodavi>kton ajkouvwn ijhv, kovpon, ouj pelavqei" ejp j ajrwgavn… Phthian Achilles, | why hearing the man-slaying -ah- blows, do you not join to help? This implies that, in Aeschylean lyric, the second period of a clause is prosodically predictable, even in a following line.15 The connection with phonological weight lies in the highlighting effect which a predictable rhythm has on linguistic form (discussed by Aristotle, Rhetoric 1408b21–26, and implicit in the ‘poetic function’ of Jakobson 1987: 69). The impression of solidity resulting from a highly visible form may be seen in hyperbatic constructions, as at Choe.773: 18) ejn ajggevlw/ ga;r kupto;" ojrqou'tai lovgo". For in the messenger the crooked word is made straight. 16 14Bibliographies may be found in De Groot (1919: 200–217), Allen (1973: 361–389), and Devine and Stephens (1984: 142–147, 1994: 498–562). 15Dover (1993: 345) notes that the satire is directed not only to Aeschylus’s fondness for dactylic rhythm, ‘but also his use of refrains, which sometimes consist of only a few words ... but may also constitute short stanzas.’ 16Garvie (1986: 253) discusses the textual problems of the line. 81 Here, the hyperbaton may be considered to give more weight to the whole constituent kupto;" lovgo". The metrical regularity of this type of hyperbaton is considered in detail in Chapter 3. iii) The last word: while weight to the right does not involve only a continuum of increasing emphasis, the final position itself appears to be prominent (cf. Quintilian, Institutio IX.iv.29). Pragmatic reasons why this might be so (mostly involving the notion of a key or thematic word) have been suggested by Delbrück (1900: 110), Kühner (1904: 597), Thomson (1938: 19), and Denniston (1952: 45).17 However, the prominence of the last word of the stichic line may serve a prosodic function: the marking of a period. Quintilian (Institutio IX.iv.91–3) notes that long syllables, which carry more auctoritas (dignity), create a sentence ending which is firmissima (strongest), Demetrius (De Elocutione 39) describes a long final syllable in prose as megalei'on (grand), and a general tendency for the last syllable of a word group to be lengthened is noted by Allen (1973: 204–207) and Devine and Stephens (1984: 25–28). Final emphasis is particularly clear in poetry: though Thomson (1938: 368) argues that ‘the end of the line, as such, is never emphatic’, the iambic line and the clause regularly coincide (as is implied by the fact that enjambement is worthy of note, and by the correlation proposed by Demetrius, De Eloc.204– 205, between the trimeter line and the ideal length of a prose clause, noted in Chapter 1, Section A4). Final emphasis is encouraged by the rhythmic principle, canonically described in terms of ‘bridges’,18 that repeated coincidence of word boundaries with metrical units should be avoided. This presumably also motivates the normal division of trimeter and hexameter lines by caesura within the metra (such as the penthemimeral and hephthemimeral caesurae) rather than diaeresis (when word and metron end coincide). However, metrical and word boundaries regularly coincide in the final foot, with a high frequency of final disyllabic nouns throughout trimeters in tragedy: Raalte (1986: 207, 214 Table XX) shows that more than 75% of tragic trimeters have word breaks before the penultimate or final foot. The late 17See Chapter 1, Section A 2. 18An extensive discussion is given by Devine and Stephens (1984). 82 position is the more common, with similar proportions in Aeschylus, Sophocles, and Euripides (with percentages in the mid-50s, rather than the 30s after the second metron). It is interpreted by Raalte (1986: 21) as a return to the initial rising movement of the iambic. It cannot be simply a metrical feature, because it is also a feature of the hexameter line, so it may be not only the coincidence of word and final foot, but the disyllabic beat itself, which is ‘a rhythmic index of verse-end’ (Raalte 1986: 29).19 These three features (initial prominence, increasing rhythmic regularity, and final prominence) are further discussed in the remainder of this work, and it will be shown that, in connected texts and in complex sentences, initial and final prominence are combined. B 5 (b): Prosodic motivation for prose VS Since, as shown in Chapter 1, prose and poetic order are comparable, similar prosodic motivation might be expected. This is partly borne out in the prose texts of the corpus: in the Melian Dialogue and Crito, SV is always associated with ascending order (see Appendix 1F). However, VS is also associated with ascending or constant order (in both texts, though only in Crito main clauses is there a significant number of examples: 17, of which 7 are ascending, 3 descending, and 7 constant). The same predominance of ascending constructions may be seen in verb and object ordering in Crito (see Appendix 1G). It may be inferred that there is the same association of noun and verb length and morphological weight in both genres, but that the last position of the stichic line is more prominent than the final word of a prose period, as might be expected. Prosodic groupings can, of course, be observed in prose too.20 In Crito, there seems to be a final rhythmic component in VS clauses with the subject oiJ povlloi, where the repeated long syllables create prominence. The 19It seems to be a contrastive effect, because final trisyllabic words are permitted when preceded by a monosyllable, as described by Porson’s Bridge, which disfavours [- - |- v -||], again ensuring final emphasis. See Porson (1802, reproduced in Allen 1973: 308–9). 20References to discussions of the ‘clausula’ (the last 4 or 5 syllables preceding a pause) are given by Dover (1996). 83 constructions include what would, in verse, be hiatus after the noun, as at Crito 44c3–4: 19) ouj ga;r peivsontai oiJ polloi; [wJ" su; aujto;" oujk hjqevlhsa" ajpievnai ejnqevnde hJmw'n proqumoumevnwn. For the many will not be persuaded that you yourself did not wish to leave here, while we were willing 44d2–3: 20) o{ti{{{ oi|oiv t j eijsi;n oiJ polloi; [ouj ta; smikrovtata tw'n kakw'n ejxergavzesqai ...that the many are able to achieve not the smallest of harm..... and 48a11: 21) oi|oiv tev eijsin hJma'" oiJ polloi; [ajpokteinuvnai. And the many are able to destroy us. In these constructions, VS does not appear motivated structurally or pragmatically, by the preposing of emphatic elements or the postponing of thematic ones, but by the creation of a rhythmic cadence, which would not exist if the subjects preceded the verbs. It may also be noted that in the three VS constructions with hJ povli", the short final syllable is lengthened by following consonants, at 50c1, 51b9, and 53a3–4: 22) ou{tw soi diaferovntw" tw'n a[llwn jAqhnaivwn h[resken hJ povli" te kai; hJmei'" oiJ novmoi dh'lon o{ti: So much more than the other Athenians did the city and we the laws please you, it is clear. Further evidence of prosodic groupings might be adduced from the phonetic gap created by the aspirated article21 and the greater phonological prominence of nouns compared to verbs (see Devine and Stephens 1994: 352). How far tonal intonation may be identified remains uncertain, though it was proposed as a motivation for prose word order by Loepfe (1940), who associated the G> and plovkamo"j > vj > vj > v tevles j jHwv"vvv , but when rose-fingered dawn brought the third day Od.9.318=424=Il.9.94: 105) h{de{{{ dev moi kata; qumo;n ajrivsthj vj vj v faivneto boulhv: and this seemed the best plan to me in my mind This accords with a common Homeric pattern in which formulaic epithets start at the B2 position, noted by Parry (1928: 12–13). 3) Alternatively, the tmetic pattern may start at the B1 caesura, as Od.9.214=515: 107) a[ndr j ejpeleuvsesqai megavlhnvvv ejpieimevnon ajlkhvnj vj vj v that I would encounter a man endowed with great strength 4) There is no pattern with demonstrative and noun: the first element is usually an adjective. 5) When at either B caesura, the adjective is emphatic, rather than being simply a traditional epithet, as in the constructions observed by Conrad (1990: 115 50), who identifies a pattern with the noun at the B caesura and the adjective at the line end as common. Conrad interprets the prosodically emphatic element as being also pragmatically emphatic, but this is not really sure: even though meilicivoisi is a traditional epithet, it does not have less emphasis than e[pessi at Od.9.63: 108) kai; tovte dhv min e[pessi [[[ proshuvdwn meilicivoisivvv : and then I addressed him with honeyed words Conversely, the adjective at Od.9.381 is pragmatically, though not metrically, prominent: 109) i{stant j: aujta;r qavrso"vvv ejnevpneusen mevgavvv daivmwn. .... but a daimon breathed great courage upon [us]. 6) A striking difference from tragic practice appears in the much higher proportion of subject hyperbaton about the verb. Iliad 9 has 21 instances of SVS, 22 of OVO, and 9 of indirect object hyperbaton, and the proportion of subject hyperbaton is even higher in Odyssey 9 (15 SVS, 7 OVO). In both, hyperbaton is frequently associated with formulaic passages. This does not necessarily conflict with a principle of animate>inanimate ordering, because it is due to a lower frequency of transitive constructions. As noted above (Section 1b), it could reflect a change from topic-prominence in Homer to subject-first in tragic practice. 7) Verbal tmesis constitutes a distinctively Homeric type of hyperbaton, which is associated with nominal hyperbaton in several cases, sometimes in adjacent lines, as at Il.9.211–2: 110) pu'r de; Menoitiavdh"vvv dai'en mevga ijsovqeo"j vj vj v fwv". .... aujta;r ejpei; kata; pu'r ejkavhj vj vj v kai; flo;x ejmaravnqh, and the god-like son of Menoitios made the fire blaze greatly. And when the fire had burnt out and the flame subsided Od.9.296–7: 111) aujta;r ejpei; Kuvklwy megavlhnvvv ejmplhvsato nhdu;n;;; ajndrovmea krev j e[dwn kai; ejp jj jj jj j a[krhton gavla pivnwnvvv , but when the Cyclops had filled his great stomach feeding on human flesh and drinking down unmixed milk 116 and, in one case with verbal and phrasal tmesis interwoven, at Od.9.375: 112) kai; tovt j ejgw; to;n moclo;n uJpo ; spodou' h [lasa pollh'"''' , and then I drove the beam under the deep cinder-pile Three other instances of verbal tmesis occur at the line-end, Il.9.92=222: 113) aujta;r ejpei; povsio" kai; ejdhtuvo" ejxjjj e[ron e{nto{{{ , but when they had put aside their desire for drinking and eating and Od.9.558: 114) h\mo" d j hjevlio" katevdu kai; ejpi;j ;j ;j ; knevfa" h\\ \\lqe, but when the sun set and darkness came over The verbs involved in these constructions all have a close grammatical connection with the internal noun. As in phrasal tmesis, it involves the interaction of the three phrasal elements of head, modifier, and complement.21 Three constructions represent a milder form of hyperbaton, with no lexical word intervening, with an interpolated clitic at Il.9.653: 115) kteivnont j jArgeivou", katavv vv te smu'xai''' puri; nh'a". ... killing the Argives, and will darken the ships with fire The second and third involve anastrophic tmesis,22 at Il.9.539: 116) w\\ \\rsen e[pi[[[ clouvnhn su'n a[grion ajrgiovdonta, ... sent on them the wild boar with shining teeth and Od.9.534: 117) ojye; kakw'" e[lqoi, ojlevsa" a[poj v [j v [j v [ pavnta" eJtaivrou", may he come late, in misfortune, having lost all his companions Homeric phrasal tmesis shares some of the prosodic regularity of tragic practice, and also favours the line-end position for the second element, which may be a disyllabic noun. It may be noted that the syntactic and prosodic patterns of verbal and nominal tmesis are similar: in both constructions it is 21See the Introduction, Fig. 2. 22Quintilian (Inst.8.6.65) defines hyperbaton of two words as ajnastrofhv: ‘verum id cum in duobus verbis fit, ajnastofhv dicitur, reversio quaedam’ (when hyperbaton involves two words, it is called anastrophe: a kind of reversal). His examples are mecum, secum, and quibus de rebus. 117 the specifier (demonstrative, adjective, or verb prefix) which is out of expected sequence, and these elements are usually phonetically light. 5: Phrasal tmesis in prose Denniston (1952: 51) notes the frequency of hyperbaton about the verb in CG prose. He calls the insertion of a verb between an adjective and noun in agreement one of ‘the milder forms’ of hyperbaton, ‘common in all authors’, and cites 19 examples from Herodotus, Isocrates, Plato, Demosthenes and Lysias. These show a variety of grammatical dependencies, and a variation in the number of intervening words. Ten involve the separation of noun and adjective in agreement, and in only one example, Plato Critias 116c, are they separated by a verb alone: 118) marmaruga;";;; e[conti purwvdei"vvv having fiery flashings Other instances of separation of words in agreement involve only pronoun insertion, or are of wider scope. All other examples of verb interpolation involve a noun and a dependent genitive, which create a less striking effect.23 The function of ‘genitival tmesis’ could often be seen as thematic (since both the dependent and governing words are nouns). When the genitive precedes, it signals its dependence on a noun, so providing the context for it, as at Plato Leg.812c: 119) ... eij" ajreth'"j 'j 'j ' e{pesqai kth'sin''' to accompany them in the acquisition of virtue and Isoc. 5.1–2: 120) ... tou' lovgou' v' v' v poihvsomai th;n ajrchvn; j v; j v; j v ... ... I shall give the start of the speech ...24 In contrast, when the governing noun precedes, it is frequently an evaluative word, which requires a genitive to follow it, as at Plato Phaedr.240d: 121) ... ejp j e[scaton[[[ ejlqei'n ajhdiva"j vj vj v to go to the utmost of unpleasantness 23These may be compared with the Aeschylean constructions cited by Fraenkel (1950: 827–8). 24 One might contrast the identical sequence of words in Isoc.10.16, which have a different grammatical structure, with the genitive governed by a prior instance of th;n ajrchvn. 118 and Lys. 32.11: 122) to; mevg; v; v; v eqo" aujth;n ajnagkavsei tw'n sumforw'n' '' '' ' the size of the misfortunes would force her... 5a: Hyperbaton in Thucydides ‘Longinus on the Sublime’ (22.3) describes Thucydides as: 123) kai; ta; fuvsei pavntw'" hJnwmevna kai; ajdianevmhta o{mw" tai'" uJperbavsesin ajp j a[llhvlwn a[gein deinovtato" most clever in separating by hyperbaton both ideas unified by nature and indivisible alike. However, most hyperbaton in the Melian Dialogue is wide-scope, and only occasionally involves a nominal phrase, as Thuc. 5.100: 124) «H pou a[ra, eij tosauvthnvvv ge uJmei'" te mh; pauqh'nai ajrch'" kai; oiJ douleuvonte" h[dh ajpallagh'nai th;n parakinduvneusin; v; v; v poiou'ntai, Surely then, if you and your subjects brave so great a risk, you so as not to lose your empire and they, already your slaves, so as to be rid of it ... The patterns of subordination might be analysed as a sort of hyperbaton based on the clause rather than the word. The primary separation is that of subject and verb. This may be seen in a construction at Thuc. 5.86: 125) hJ me;n ejpieivkeia [tou' didavskein kaq j hJsucivan ajllhvlou"] ouj yevgetai, the fairness [of the proposal of mutual instruction at leisure] is not objectionable ... In a non-finite construction at Thuc. 5.89, the subject is separated from its verb by a finite complement clause and by another, governing verb, (ajxiou'men): 126) [ou[q? uJma'"J 'J 'J ' ] ajxiou'men [ [ h] o{ti Lakedaimonivwn a[poikoi o[nte" ouj xunestrateuvsate h] wJ" hJma'" oujde;n hjdikhvkate ] levgonta" oi[esqai[[[ peivsein] nor do we think that you [will think it possible to persuade us by saying [either that being colonists of the Spartans you did not become allies, or that you have done us no wrong] ] The subject pronoun in the non-finite clause is early, and the verb is late. This is the type of pattern criticized by Aristotle at Rhetoric 1407a26–30 as ajsafev". Denniston (1952: 50) suggests two motivations for hyperbaton in prose: preposing an emphatic word, and the creation of a rhythmic pattern. 119 Thucydidean usage may be considered as a third type: the establishment of subject reference as early as possible in dialogic exchanges. 5b: Contrast between poetic and prose hyperbaton The similarity of hyperbaton and subordination is shared by prose and poetic constructions, as discussed further below, in Section 6d. However, a condensed and prosodically regular pattern is distinctive to epic and tragic poetry. The expansion of phrases to fill the second colon of the poetic line has a similar effect to the postponement of phonological weight, as is considered below in Section 6c. 6: Phrasal tmesis and prosody It is clear from the patterns of emphasis that in wide-scope hyperbaton the demonstrative is preposed, and in demonstrative-last constructions postponed, to a prominent position in the clause. This description accords with a pragmatic explanation for hyperbaton, in terms of marked elements (see Dik 1995: 7). However, in phrasal tmesis, there is no single emphatic element, and an explanation in terms of movement of complete phrases, as by ‘scrambling’25 or topicalization,26 is also inappropriate, because the pattern involves the separation of two elements of a phrase. In fact, no purely structural model provides a satisfactory explanation, because ‘movement’ cannot be judged except in relation to an ‘original’ position. The regular placing of the demonstrative and the noun suggests a prosodic explanation. 6a: Demonstrative position The demonstrative o{de has variable placing in CG, being categorized by Dover (1960: 23) as ‘preferential’ (Ma): that is, tending towards the front of the clause, though quite frequently appearing in Herodotus and Plato as the last word of a clause. However, the position of the demonstrative in the trimeter line is remarkably regular. In the trimeters of the Oresteia, there are 67 non- hyperbatic phrases with a noun and demonstrative (collected in Appendix 25Suggested by Ross (1967: 75ff.) as the formalisation of an intuitively-simple notion, and developed by Williams (1984), Webelhuth (1984), and Grewendorf and Sternefeld (1990). 26Defined in the Introduction, Section 2c. 120 2I): in 35 the demonstrative follows the noun, and in 32 it precedes. There seem no clear pragmatic differences: gh'" th'sde and th'sde gh'" both occur. There is, however, great prosodic regularity. In noun-first constructions, the demonstrative is framed between the penthemimeral and hephthemimeral caesura in over half the instances (22 out of 35), as at Ag.18: 127) klaivw tovt j oi[kou tou[[[ 'de''' sumfora;n stevnwn - - v - |- : - v : - | v - v - then I weep, lamenting this house’s misfortune, Ag.906: 128) e[kbain j ajphvnh" th'sdej v 'j v 'j v ' , mh; camai; tiqeiv" come down from this wagon, not setting [your foot] on the ground Ag.1039: 129) e[kbain j ajphvnh" th'sdej v 'j v 'j v ' , mhd j uJperfrovnei: get down from this wagon, and do not be proud Ag.1071: 130) ei[kous j ajnavgkh/ th'/dej v / '/j v / '/j v / '/ kaivnison zugovn. yielding to this constraint bear the new yoke. The other constructions are collected in Appendix 2I. The regularity of the demonstrative position does not necessarily imply a marked ordering with respect to the noun, because demonstrative-first instances occur later in the line precisely because the demonstrative is in the same position, in 22 out of the 32 instances of demonstrative-noun order, as Ag.24: 131) pollw'n ejn “Argei th'sde sumfora'"' '' '' ' cavrin. in Argos for the sake of this event Ag.33: 132) tri;" e}x balouvsh" th'sde''' v moi fruktwriva"vvv : this beacon-watching having thrown a triple Ag.619: 133) h{kei su;n uJmi'n, th'sde gh'"' '' '' ' fivlon kravto". he will come with you, dear ruler of this land 121 Choe.85: 134) ejpei; pavreste th'sde prostroph'"' '' '' ' ejmoi; since you are here in this supplication to attend me The other 18 constructions are collected in Appendix 2I. Since this is also the position of the demonstrative in phrasal tmesis, it appears that the position of the demonstrative between the penthemimeral and hephthemimeral caesurae, is basic. Such a constant position, even with noun-demonstrative order, implies that metrical motivation is primary. The regular position of the demonstrative is not, therefore, a feature unique to hyperbaton, or even particularly associated with it, as the 90 Aeschylean instances of this central position are divided about equally between D >N, N>D, and hyperbatic patterns. Nor is the central demonstrative position distinctive to Aeschylus: it is also evident in Medea (th'sde 272, 353, 702, 709, 729, 916, with 2 instances of alternative placing) and OT. (th'sde 54, 418, 601, 764, 811, 1043, with 5 alternative placings). It may be motivated metrically: a word which is disyllabic, but readily elided, and with a compound accent, is highly suited to the central position of the trimeter line. 6b: Noun position and narrative function It was noted above, in Section 1a, that the pattern is overwhelmingly line- final, with a disyllabic noun in the last foot. The importance of the clause-final noun may be inferred from its frequency even in the Homeric examples, where it is most often a subject. The stress component of a double beat at the line end was discussed in Chapter 2, Section B6 (a). These final words could, perhaps, sometimes be interpreted as thematic. In Aeschylus, sequences with demonstrative and lovgon or movron appear 12 times, and pathvr 10 times (mostly in Choephoroi). Possibly thematic final words appear in Sophocles and Euripides too, as the prominence of povli" and kakav in OT., and lovgo", tevknon, and cqovna in Medea, demonstrates. If these words are analysed as thematic, tmesis might be considered a technique by which the narrative line is highlighted. However, as has been argued in the Introduction, Section 2c, and Chapter 2, Section B4, the category of theme is not very informative: any noun has a high probability of being thematic. 122 A narrative function may, however, be observed in the position of the formulaic tmetic patterns within the texts: the repetitions of complete hyperbatic lines in Medea appear to constitute a sort of ring composition (as discussed in Section 2b above), and several of the texts start with a tmetic line: Od.9.2 (the first line of Odysseus’s speech), Ag.1 (with ring composition at Ag.20), Choe.1, Cyc.1 and 2. The former seems to be a cohesive device, but the latter feature may reflect a tendency for the start of texts to be especially formal, as suggested above in Section 2a. 6c: Phrasal tmesis and phonological weight Phrasal tmesis can be seen as a syntactic feature rather than a thematic pattern. It was noted above in Section 1a that the effect of the tmetic pattern is to define the second part of the line more rigidly than the first, by creating a constituent which is co-extensive with the second colon of the line. The visibility of the artifice, evident from the Aristophanic quotations, is essential to this effect. The author of Longinus on the Sublime (22.1) considers that the effects of hyperbaton in prose should be imperceptible: 135) tovte ga;r hJ tevcnh tevleio", hJnivk j a]n fuvsi" ei\nai dokh'/, hJ d j au\ fuvsi" ejpituchv", o{tan lanqavnousan perievch/ th;n tevcnhn. for art is perfect when it looks like nature, and equally nature successful when containing art hidden within her. However, the extreme regularity of tragic hyperbaton is, in contrast, very conspicuous (as Aristophanes’ citations of it demonstrate). It contributes to Aeschylean o[gko" (weight), since it gives prominence to the whole phrase. Cognitive effort is involved: parsing a sentence involves recognizing the phrasal constituents, and the extra interpretative effort created by the separation of the elements emphasizes the phrasal structure, because the constituent is recognized as early as possible, but the reference is delayed. This creates suspense, which could be reinforced by rarity of the article, common in wide-scope hyperbaton27 (demonstratives are less predictable than articles because they may be interpreted as pronouns or adjectives), the typical neuter case of adjectives (pavnta is especially common),28 and the rarity 27There is of course a metrical constraint: tov would be possible at the penthemimeral caesura only with resolution. 28 Ag.582, 599, 1210; Eum.501; Pers.246; OT.291, Aj.480, Tr.484, Phil.1240; Med.487; IA.97, 1249. 123 of explicit subjects (only 17 of the 81 OVO constructions in the Oresteia have an explicit subject).29 Yet any ambiguity is only a minor effect: phrasal tmesis appears primarily to be a poetic exploitation of interaction between prosodic patterning and constituent structure, using metre to emphasise a syntactic unit, and to give weight to the line end, by drawing attention to the linguistic form. 6d: Tmesis and subordination Denniston (1952: 52) notes the frequency in Greek prose of hyperbaton with separation of article and noun, and compares it with German nominalization, in a jocular construction: ‘The every year all England with excitement filling Oxford and Cambridge boat race.’ This demonstrates a similarity with relativization: Denniston’s example could be rephrased as ‘The Oxford and Cambridge boat race, which fills...’. The tmetic structure is similar to a relative clause with agreement attraction, where the relative takes the case that a correlative in the main clause would have, creating a structure similar to a ‘contact’ relative (without a relative pronoun, as ejcrh'to ai\\ \\" ei\ce bivbloi" ‘he used those books he had’). The prosodic pattern is the same as if the verb functioned adjectivally, qualifying the NP like an attributive relative clause: ‘for the general this [you planned] death’ (Ag.1627), ‘over one not yet lying down this [you boast] speech’ (Eum.590). Though not a true relative construction, since it is the main verb which is interpolated, this type of hyperbaton creates a subordinate-like construction, which nominalizes the whole clause. The comparison with subordination is very close: in his discussion of agreement attraction in relative clauses, Gonda (1954a: 29) categorizes a construction at S.El.762–3 as Attic attraction,30 although it is actually tmetic, as it does not depend on a main verb, but is in agreement with an adjective: 136) ajlgeinav, toi'" d j ijdou'sin, oi{per ei[domen, mevgista pavntwn [ w|n||| [ o[pwp j ejgw; ] kakw'n.''' ] ... piteous, but for those seeing it, as I saw it, the greatest of all evils I have seen. 297 relative pronouns (Ag.934, Choe.615, 991, Eum.3, 58, 639, 760), and 10 which appear emphatic (Ag.281, 1212, 1248, 1275, 1588, Choe.254, 401, 760, 765, 927). 30Attic attraction is discussed in Chapter 5, Section 2b. 124 The the parallel with subordination extends to the use of the demonstrative. As noted in the Introduction, Section 2i, the demonstrative o{de is involved in the development of wJ"-complementation, because it has an anticipatory textual function, and the demonstrative in the tmetic pattern is almost always o{de too: the rarity of ou|to" in this position cannot be only metrically motivated, as it appears in four Aeschylean and two Sophoclean constructions,31 and is also at the caesura in 14 non-tmetic constructions in the Oresteia. The function of anticipatory o{de is considered further in Section 4b of Chapter 4. That chapter, which constitutes Part II of this work, is devoted an examination of prosodic and syntactic details of the start of the CG clause. In Chapter 5, the analysis will be extended to the function of anticipatory o{de as a clause-initial feature. 31Elided tou't j appears at Ag.547; Choe.991, Eum.58, 743, and OT. (138, 1033). Part II Focus 126 Chapter 4 Focus, particles, and the clause start Introduction The importance of a rightwards trajectory of phonological weight in determining word order was discussed in Chapters 2 and 3. In this chapter, the prosody and structure of the start of the clause are investigated, and a unified explanation proposed for the regular initial placing of emphatic elements and the functions of enclitics following. It is argued that particles with a wide variety of functions all fulfil their function by emphasizing the initial element, and that coordinating particles are also focalizers. The problem The beginning of the CG clause has two distinctive features: an emphatic element in first position (P1), followed by an enclitic word or words with a variety of different syntactic functions in second position (P2). As noted in the Introduction, the prosody and structure of these elements are usually considered separately. P1 prominence is canonically explained in terms of logic, textual relevance, or emphasis. P2 has traditionally been studied functionally, though its prosody has more recently been investigated (in work cited below, in Section 2c). The interrelation between P1 and P2 elements has, however, always been considered in purely prosodic terms. A functional description of the relation would be desirable. The proposal The P1>P2 sequence can be described by a unified model which integrates prosody and structure. The prominence of initial words is attributed to focalization by enclitics in P2. These are divisible into cohesive focalizers and operators (interrogative and relative pronouns). Emphatics, coordinating, and adverbial particles all function as cohesive focalizers, being part of the same intonation group as the P1 word. Operators are separated from the word in P1 by an intonation break, and mark it as focal, also usually being in morphological contrast with it. 127 Consequences The proposed [P1>P2] structure suggests a mechanism of inter-clausal linking: 1) The first constituent of a main clause is regularly focalized. 2) The inter-clausal link is also focalized: in co-ordinated links, the focal element is in the second clause, focalized by the P2 element. In subordination, the final element of the main clause, standardly the object, may be focal, marked by the subordinating conjunction. 3) Subordinate clauses which precede their main have a focal element in P1, while subordinates which follow their main do not. Clause order therefore directly affects clause structure and prosody. Inter-clausal linking in relatives and adverbials is also discussed in this chapter. A survey of conditional clause order with respect to main clauses is also undertaken, and it is observed that, throughout the corpus, there is a correlation between clause order and patterns of focalization, which creates formal similarities between following conditionals and indirect questions. Chapter Sections 1: P1 1a: Prominence 1b: Logic 1c: Topic 1d: Textual relevance 1e: Clause structure 1f: Definition of focus 2: P2 2a: Summary of P2 enclitic functions 2b: P2 as an operator position 2b (i): Interrogatives 2b (ii): Questions in Choephoroi 2b (iii): Interrogatives and the intonation break 2b (iv): Relatives 2c: P2 and the intonation break 2c (i): Order in collocations 128 2c (ii): The cohesive focalizer/operator division 3: Cohesive focalization 3a: Overlaps in particle function 3b: Diachronic changes in position and function 3c: Gavr and dual function 3d: Other cohesive focalizers 4: Focalization and linking 4a: Mechanism 4b: Pragmatic functions of focalization 5: Focus and clause order 5a: Focus and conditional order 5b: Conditional order and discourse function 5c: Conditional order and indirect questions Summary: focus and prosody 1: P1 The start of the clause is always taken to be communicatively special, but there are reasons to doubt the canonical descriptions (of topic, focus, and theme), as noted in the Introduction. They can be summarized as an over- rigid association of logical and textual categories with sentence structure. In order to present the background to the proposal of an initial focus position, alternative descriptions of P1 are presented in terms of three criteria: prominence, logic, and textual cohesion. The relations between these categories and clause structure are discussed briefly, and the categorization of the initial position as focal is justified. 1a: Prominence The emphasis of the beginning of the CG clause has been interpreted in two ways: as the start of a continuum of gradually declining emphasis, or as a unique emphatic position at the beginning. Denniston (1952: 44) adopts the first approach, considering that ‘the weight of a Greek sentence or clause is usually at its opening, and the emphasis tends to decline as the sentence 129 proceeds.’ Thomson (1938: 367) holds the second view, believing that the basic order of CG is predicate-first, as in copulative sentences (aJplou'" oJ mu'qo" simple [is] the story, Choe.554), and so SV is always an emphatic order. This view accords better with the principle of increasing phonological weight discussed in Chapter 2 than does Denniston’s interpretation, and is assumed by most subsequent commentators (though not necessarily for the same reason as Thomson). In itself, emphasis is a purely formal category, involving phonological prominence, which is either morphological, positional, or marked by stress or intonation. The emphatic element may be the intonational centre of the sentence (Chomsky 1971: 202), and may carry stress, perhaps even in CG (see Chapter 2, Section B 5a). Emphasis is usually also taken to have a communicative function, so ‘loud’ implies ‘important’.1 Emphasis has been linked to a variety of communicative functions, such as emotional stimulation or expositive power (Dover 1960: 32), or the marking of information as new (Halliday 1967: 204, Jackendoff 1972, Lambrecht 1994: 208) or most salient or relevant (Dik 1980: 16, Sperber and Wilson 1986: 202–217). These functions can be identified only subjectively, with the possible exception of new information, though they underlie the observable phonological feature. There is, of course, no reason why emphasis should not serve a variety of functions. In Section 1e, a prosodic basis for the prominence of P1 will be proposed, and in Section 4a, a functional description will be given. 1b: Logic Grammatical words are likely to be placed early if they are operators: that is, quantifiers having scope over the basic clause (more precisely, binding variables within it).2 These include interrogatives and relatives (‘wh-words’). Operator position is often identified with the category of sentence focus (Chomsky 1976, Kiss 1995a: 15). In this interpretation, focus is co-referential with a ‘trace’ (empty position) in the basic clause,3 so is regarded as having 1The Prague model of communicative dynamism (CD) exemplifies the connection: Firbas (1964: 270) defines CD (somewhat circularly) as ‘the extent to which the sentence element contributes to the development of communication’. 2See Lyons (1977: 454), Kiss (1995a: 15). 3The trace theory of movement, a central feature of X' theory, derives from Chomsky (1973). It is exemplified by interrogatives (Whomi did you see -i?). 130 moved from within the clause. It will be argued here, in Section 2b, that the operator and focus positions are not the same in CG: though operators can be emphasized by movement into focus position, their regular position is not prosodically emphatic. 1c: Topic Alternatively, a lexical word might be placed early because it is the logical subject, which in propositional logic standardly precedes the predicate. This schema is associated by Hockett (1958: 201), Li and Thompson (1976), and Lyons (1977: 503) with the division between topic and comment. Lyons (1977: 501) describes the difference as categorial: ‘The subject, then, is the expression which refers to and identifies the topic and the predicate is the expression which expresses the comment.’ However, the parallel between topic, logical subject, and grammatical subject is not exact, and the topic or subject does not necessarily precede comment or predicate, even in declarative sentences (as may be seen from the citation from Choe.554 in Section 1a above). 1d: Textual relevance A third reason why elements might be initial is that they are thematic, and express known information (Mathesius 1939: 234). Theme is standardly associated with topic, and so taken to provide an association between the intra-sentential and textual organization of information: Kuppevelt (1995: 140) analyses ‘bound discourses’ (such as literary texts) as sequences of ‘discourse topics’ in which the sentence topics are embedded, and, in discourse analysis, texts are standardly visualized as organized by ‘clause chaining’ and ‘thematic paragraphs’.4 A congruence of textual theme and sentential topic is, however, very restrictive: a topic may be a theme, but there is no reason to consider that it must be thematic. Textual cohesion is typically expressed by linking words in P2, and themes are not necessarily expressed explicitly, but are often communicated through presuppositions and implicatures (see Chapter 1, Section A 2). 4See Greimas (1966), Halliday and Hasan (1976), Fries (1981), Halliday (1982), Brown and Yule (1983), Givón (1983), and Coulthard (1994). In the functional grammar of Dik (1978, 1980, 1989), theme is also considered to be structurally more peripheral than topic. 131 1e: Clause structure The pragmatic categories of topic, theme and focus are, clearly, of great value in describing function: their formulation by Dik (1978, 1980, 1989) has stimulated much recent work on the functional organization of Ancient Greek (by Ruijgh 1971, 1990; Rijksbaron 1989, 1997a; Wakker 1994, 1997; and others cited below, in this and subsequent chapters). However, they have less power to explain word order or details of clause structure. In fact, they are regularly associated with different structural positions: emphatic information follows unemphatic, in the theme-rheme model (Firbas 1964: 170), but precedes it in terms of task urgency (Givón 1983: 20). The only reliable definition may be that of Halliday (1967: 212), that ‘Basically, the theme is what comes first in the clause’. This is, however, simply a (linguistically uninformative) way of describing a common sequence in declarative sentences. In structural terms, an initial element can be modelled within or outside the basic clause: a position outside the clause is usually described as ‘topicalized’, defined here, following Emonds (1976), as movement to a position adjoined to the sentence, with a purely co-referential link. In Modern Greek, topicalized words typically have a resumptive pronoun in the main clause.5 The use of the word ‘topicalization’ to describe this sort of adjunction is unfortunate, because it implies an association with the pragmatic category of topic, yet topicalization cannot be always topical (since it is possible to have multiple topicalized elements adjoined to one clause).6 Neither theme nor topic appears to be a structural category of CG: there is no evidence from the corpus that any syntactic position is determined by thematic factors, or that any Greek particles are topic markers. There may be an association between topic and definite NPs (Li and Thompson 1976, Gundel 1988), but not all definite NPs are topical, and (in tragedy) not all topical NPs are explicitly definite. The conclusion that P1 is not a structural topic position in CG contrasts with the view, expressed by Kiparsky (1995: 153), that there was an initial 5See Horrocks (1983) and Philippaki-Warburton (1985). 6See Horrocks (1983: 104) and Kiss (1995a: 11). 132 structural topic position external to the sentence in early IE, preceding a focus position, as in Fig. 1 (=Introduction, Fig. 4):7 1 S Topic [+WH] Focus tiv dei' s j ejmoi; levgein… S" S' kai; nu'n ta; mavssw me;n Obj. It is not clear that kai; nu'n ta; mavssw mevn is here semantically a topic (though part of it, t;a mavssw, might well be), or that only a topic can be placed initially. Nor does the prosody support Kiparsky’s interpretation: the interrogative appears less, not more, prominent than the initial phrase, which is emphasized by the intonation break following.8 A similar objection may be raised to the description of thematic being applied to other word groups which are separated from the rest of the sentence by an intonation break, as at Hdt. 1, 10.2 (discussed also by Dover 1960: 17 and Ruijgh 1990: 229): 1) kai; hJ gunh; J; J; J ; | ejpora'/ min ejxiovnta et (quant à) la femme, elle le voit sortir 9 Ruijgh (1990: 229) defines hJ gunhv as the theme of its sentence, an interpretation which could perhaps be plausible for an NP, but scarcely for a 7Adapted from Kiparsky (1995: 153, Fig. 33), with the quotation, from Ag.598 (‘And now, for the full story, what need have you to tell me it?’), added. 8This interpretation is supported by the metre: Raalte (1986: 186) considers that a sequence of two monosyllables following the penthemimeral caesura emphasizes a word-boundary after the third foot (here, between mevn and tiv). 9The translations of this and the following citation are taken from Ruijgh (1990: 229). 133 temporal phrase at Od.8.55–56 which he defines similarly: 2) uJyou' d j ejn notivw/ thvn g j w{rmisan: aujta;r e[peitaj ; [j ; [j ; [ bavn rJ j i[men ... ....... Mais quant à ce qui se passa après cela, ils se mirent en marche pour aller .... Ruijgh’s translation puts quite a strain on a two-word phrase, but its length implies emphasis, and the view taken here is that these constructions are emphatic, either as part of the basic clause (as the first could be) or perhaps topicalized, but certainly not thematic: they rather create a contrast with the preceding text. Nor can anaphoric elements, as the demonstrative at Eum.649–650, be interpreted as occupying a topic or thematic position, though they may be initial in the clause: 3) touvtwnvvv ejpw/da;" oujk ejpoivhsen path;r ouJmov", . . . for these matters, my father has not set charms. Such elements could be interpreted as adjoined, but there is no regular position for them: they may appear anywhere in the sentence, as at Eum.199: 4) aujto;" su; touvtwnvvv ouj metaivtio" pevlh/, you yourself, for this you are not only jointly responsible ... or Eum.932–3: 5) oJ ge mh;n kuvrsa" barew'n touvtwnvvv oujk oi\den o{qen plhgai; biovtou: and yet the one meeting the hostility of these does not know from where [are] the blows assailing his life These pronouns may be structurally topicalized, but they cannot be topics, which have a unique function in the clause. The rarity of such adjoined elements also precludes a model of the sentence as having regular focus and topic position: almost all CG sentences have an emphatic initial element with a particle following, and topicalized elements appear much less frequently. 134 1f: Definition of focus Initial position is regularly occupied by constituents which are prosodically prominent, either by phonetic assimilation of a P2 enclitic, or by morphological contrast with a word in P2, and such prosodic prominence is here identified with focalization. Enclitic particles are part of the same intonation group as the P1 word, and have a syntactic as well as phonetic relation with it, so are here classed as ‘cohesive focalizers’. Operators, classed as ‘focus markers’, create a contrastive prominence with the P1 element by virtue of relative morphological size, accentuation pattern, and the intonation break separating the elements. P1 is therefore defined as prosodically focal. The functional consequence is that it is always presentationally prominent too.10 The proposed structure is shown in Fig. 2 (=Introduction Fig. 3), with a cohesive focalizer (mevn) and an operator (tiv) in P2: 2 Focus [+WH] tiv dei' s j ejmoi; levgein… kai; nu'n ta; mavssw me;n P1 IP CP C' C Obj. The P1 element here is preposed, but this is not a necessary condition of focalization: the presence of an initial prominent position in practically every CG main clause implies that at least some of the variations must represent ‘basic’ word order, and indeed, in the corpus, initial focalization makes no noticeable difference to word order: in gavr-clauses, the grammatical subject can be, and frequently is, placed in P1, but this does not affect order 10The view that P1 is a focal, rather than a topic, position is in accord with the view of Luraghi (1998: 195) rather than that of Steele (1977). 135 statistically.11 Any motivation for preposing must therefore affect all elements equally, and not only the subject. Nor is the P1 element normally an operator: that function is regularly fulfilled by a P2 word. Evidence for distinguishing the focus position from the interrogative position is examined below, in Section 2b. As discussed by Hock (1996: 202ff.), a P2 enclitic may follow a prosodic host in a preceding clause, and this might seem to justify a purely prosodic model of the P1>P2 ‘string’. It is, however, an assumption of this work that there is a close connection between prosody and syntactic function, which is evident in inter-clausal linking: the prosodic relation between the elements in P1 and P2 determines their function, not only within the clause, but inter-clausally. Prominence creates contrast with the preceding text as well as the word following it. P1 is therefore seen, not simply as the initial position of the clause, but as a valency position which may function in two clauses simultaneously. When the distinctive P1>P2 prosody spans the clause boundary, it creates a structural link too. CG may have ‘discourse-configurational’ features, but only with respect to main clauses: in complex sentences, the function of the P1 element in the main clause is established through focalization by P2 elements in the subordinate. The functions of these particles are discussed next. 2: P2 In this section, the traditional groupings of P2 enclitics are described, and a division is made between particles and pronominals, based on their intonational and syntactic relationship with the P1 element. 2a: P2 categories The regular presence of grammatical words in P2 is associated with languages having largely free word order, including early IE languages and 11The relation between gavr clauses and SV order in the corpus is collated in Appendix 1A. A majority of 140 SV constructions (81 = 58%) have a subject in P1. The figures are: Il.9, 3 out of 7; Od.9, 5/10; Septem, 9/15; Agam., 12/20; Choe., 5/9; Eum., 14/17; OT., 14 /23; Medea, 13/23; Cyclops, 2 /4; Frogs, 14/19; Melian Dialogue, 2 /3; Crito, 4/6. 136 the Australian language Warlpiri,12 which suggests that the position has a regular syntactic function. P2 words may be grouped in four functional categories: i) Linking. A connective function is implied in the term suvndesmo" (conjunction), as used by Aristotle (Rhetoric 1407a20), who gives mevn and dev as examples. Aristotle (Poetics 1456b36–1457a4) identifies two features, a lack of denotative meaning and a postpositive placing, in his definition of suvndesmo" as fwnh; a[shmo" h} ou[te kwluvei ou[te poiei' fwnh;n mivan shmantikh;n ejk pleiovnwn fwnw'n pefukui'an suntivqesqai (a sound without meaning which neither hinders nor causes the formation of a single sound or phrase from several sounds), which is not put at the beginning of a phrase which is by itself (h}n mh; aJrmovttei ejn ajrch/' lovgou tiqevnai kaq j auJtovn).13 The function is subcategorized by Denniston (1954) as additional (kaiv), adversative (me;n ou\n, ajllav, dev, mhvn), confirmatory (gavr), or inferential (tw', toigavr, ou\n, a[ra). ii) Adverbial. Some words, which Demetrius (De Eloc.II, 55.1) terms paraplhrwmatikoi; suvndesmoi (expletive conjunctions), seem to have a primarily stylistic function within the clause: ajrktiko;" ga;r teqei;" oJ suvndesmo" kai; ajpospavsa" tw'n protevrwn ta; ejcovmena megalei'ovn ti eijrgavsato (‘for, as the conjunction is set initially and separates what follows from what precedes, it creates a certain dignity’: Demetrius De Eloc.II. 56.4–5).14 Words like ‘subtlety’, ‘nuance’, ‘elusive’, ‘colour’, and ‘bouquet’ are often used to describe this group,15 as they have the most subtle shades of meaning. Adverbials may be subcategorized, as by Denniston (1954: xxxvi–xl), into affirmative, intensive, determinative, and limitative particles. Alternatively, a binary division may be made into adverbials with scope over one word, usually termed emphatics, and those with sentence scope, which may be 12The data for Warlpiri is described by Hale (1976, 1983, 1992), Kashket (1991), and Simpson (1991). 13Poetics 1456b38–1457a10 is marked by Kassel (1965) as ‘corrupta et confusa’, but the general sense may be discerned. Simpson (1991: 69) notes a similar feature in Warlpiri, that the auxiliaries normally appearing in P2 occur sentence-initially in connected speech. 14These include orthotonic words as well as P2 particles: Demetrius’s examples include dhv, nu, provteron, feu', and even poi'ovn tiv ejstin, but this discussion concentrates on the enclitics. 15These epithets are used by Denniston (1954) and Smyth (1956). 137 called sentence adverbials. It will be shown below, in Section 2c (ii), that not only do these function similarly, but that they belong to the same semantic group: the focalizers. It may be noted that Aristotle (Poetics 1456b36–1457a4) appears to view the category of suvndesmo" as including connectives and adverbials, as he gives as examples dhv and toi, as well as mevn and dev. iii) Anaphoric textual reference. Bergaigne (1877) and Delbrück (1878) described P2 as the natural place or Haus of pronominals as well as connective particles. Bergaigne (1877: 177–178) suggested that anaphoric pronouns come as early as possible in the sentence because, like conjunctions, they have a linking function, and that first and second personal pronouns are placed there by analogy. However, the argument by analogy seems weak: pronouns might be placed in P2 for purely phonetic reasons, or for reasons of scope (as is suggested by the high frequency of subject pronouns in P1 or P2, as described in Chapter 2, Section A2). Dik (1980: 23) defines P2 as the preferred placing for pronominals on the basis of a phonetic weight criterion, as described in Chapter 2, Section B5b, and Mallinson and Blake (1981: 151) combine this explanation with a pragmatic one, based on theme-first. iv) Scope. Elements may be placed early because they have scope over the whole clause. Wackernagel (1892: 34–5) pointed out that enclitic verbs are regularly placed in P2 in early Greek and Vedic (see also Hock 1982). The reason may be that the verb inflection has scope over the clause, of which it is the head (in the X' model).16 A prosodic explanation appears necessary to explain verb placing in P2 rather than P1 (see Anderson 1993, Adams 1994b), and this could be modelled as a Focus Phrase, with the verb as head and the focalized element as specifier (Horvath 1986, Kiss 1995b). This accords with a principle of morphological weight: as inflections became more complex, so verbs moved rightwards. v) Prosody. Wackernagel (1892) implicitly attributed the placing of light elements in P2 (to which he famously gave the status of ein Gesetz) to prosody, by defining P2 clitics in terms of their lack of accent, and so concentrating on the P2 position itself as attracting different categories of elements, including pronominals, linking words, and verbs.17 16It may be noted that in Warlpiri, P2 is regularly occupied by the tense marker: see Hale (1992: 65). 17For a recent discussion of his model, see Anderson (1993). 138 Wackernagel’s criterion (that the element in P2 is accent-less) implies a contrast with the element in P1, which is therefore emphasized. It follows that a complete description of P2 clitics requires an analysis of the function of the P1 element too. Explanations of why these varied elements are placed in P2 may be divided into syntactic and prosodic interpretations, which are discussed further in Section 2c (i) below. Clitics are canonically categorized as reduced forms of orthotonic words — simple and special clitics in the schema of Zwicky (1977)18 —or independent enclitics with no orthotonic form: ‘bound words’, including all words traditionally termed ‘enclitic particles’. It is proposed that the most important division should be drawn between elements which can appear in either P2 or P1, and those which are always enclitic. The principal contrast is between operators (‘special clitics’ when they are in P2 and do not occupy the same positions as full forms), and cohesively-focalizing particles (always P2).19 Operators are considered first. 2b: P2 as an operator position 2b (i): Interrogatives Although focalized constructions are standardly analysed as semantically similar to interrogatives and relatives (as Schachter 1973, and Chomsky 1981), a distinction may be drawn in CG between focus position and P2, where interrogatives and indefinite pronouns appear. Evidence will be adduced that interrogatives appear in initial position only by being focalized: that is, foregrounded in P1 by a focalizer in P2. This conclusion is stimulated by the observations of Thomson (1939b) on the very common placing of emphatic constituents preceding interrogative pronouns in tragedy and Aristophanes, in over 400 constructions. Thomson described this pattern as the postponement of the interrogative, but his 18The basic distinction between simple and special clitics is that the former have no distinctive placing or syntactic restrictions: clitics other than bound words which appear regularly in P2 are therefore special clitics. 19The terms particula and movrion were applied by the ancient grammarians to phonetically light words, including lexical ones, so correspond more closely to enclitics than particles. For references, see Schenkeveld (1988), and, for a historical overview, Sluiter (1997). 139 examples show that the elements in P1 are always prominent. He identified a number of regularities: 1) If one of the clauses in a mevn–dev construction is interrogative, the pronoun is postponed, as at OT.1232–3: 6) Leivpei me;n oujd j a} provsqen h[/demen to; mh; ouj baruvston j ei\nai: pro;" d j ejkeivnoisin; j j v; j j v; j j v tiv fhv/"… There lacked nothing in what we knew before of lamentable matters: and in addition to that, what may you say? and Ag.598 (also cited in Figs. 1 and 2 above): 7) kai; nu'n ta; mavssw me;n' ; v ;' ; v ;' ; v ; tiv dei' s j ejmoi; levgein… And now the full story, what need to tell it me? 2) A word repeated, either by one speaker or in a stichomythic exchange, is emphasized, as at Eum.94: 8) eu{doit j a[n, wjhv: kai; kaqeudousw'n''' tiv dei'… Do sleep on, hey — and of sleeping, what use is there? Sometimes there is no actual repetition, and a synonym is used, as Ra.628–30: 9) .... jAgoreuvwvvv tini; ejme; mh; basanivzein ajqavnaton o[nt j: eij de; mhv, aujto;" seauto;n aijtiw'. øAI.Ø Levgei"vvv de; tiv… I say that I should not be beaten, being a god. Otherwise, you will blame yourself. (Ae.) What are you saying? 3) The second of two questions usually has a preposed element, as Eum.678: 10) tiv gavr… pro;" uJmw'n; J '; J '; J ' pw'" tiqei's j a[momfo" w\… What then? In regard to you, how may I arrange matters so I may be blameless? 4) Shifts of focus between speakers in dialogue are regularly emphasized, as Ra.1430: 11) Eu\ g j, w\ Povseidon. Su; ;;; de; tivna gnwvmhn e[cei"… Brilliant, by Poseidon. And as for you, what is your opinion? 140 Shifts of focus between ideas are similarly emphasized, as Choe.924–5: 12) o{ra, fuvlaxai mhtro;" ejgkovtou" kuvna". ta;" tou' patro;" ; ' ;; ' ;; ' ; de; pw'" fuvgw parei;" tavde… (Clyt.) Look, watch out for your mother’s spiteful hounds (Or.) Those of my father, how may I escape them, neglecting this? Expression of a contrast is also associated with the same pattern, as Cyc.525: 13) qeo;";;; d j ejn ajskwjjj '/ pw'" gevghq j oi[kou" e[cwn… But a god in a wineskin, how can he be happy to have his home there? Thomson identifies other functions: to mark stages in an interrogation, to make an objection, or to recall something in the previous text. His figures demonstrate that the pattern is least common in Aeschylus and most frequent in Aristophanes and the later works of Euripides (and Cyclops), which suggests that it is a developing feature. Metrical convenience cannot be the explanation, as the ‘postponed’ pattern is common in Plato too, as at Crito 44a9:20 14) «Hn de; dh; tiv to; ejnuvpnion… What was it, the dream? Thomson’s ‘postponement’ always involves the preposing of the element in P1.21 It seems possible to go further, and infer that P2 is the regular position for interrogatives, and that they appear first in the clause only by being focalized, in which case they are always followed by particles, as at Ag.1286: 15) tiv dh't''' j ejgw; kavtoikto" w|d j ajnastevnw… Why then do I make this pitiful lament? and Ag.1643–4: 16) tiv dh;; ;; to;n a[ndra tovnd j ajpo; yuch'" kakh'" oujk aujto;" hjnavrize", ajlla; su;n gunhv, Why then with your cowardly heart did you not yourself kill this man, .... Here, focalization appears to express an adversative force: all Thomson’s examples cited above involve a contrast with the preceding text. It is argued in Section 4b that focalization is always, in discourse terms, contrastive. 20Thomson (1939b: 151) cites other Platonic examples at Rep.349b, Crat.388a, Ap.20c. 21This accords with the X' schema, in which rightwards movement of the interrogative is impossible, because a ‘trace’ cannot precede its antecedent. See Chomsky (1981). 141 2b (ii): Questions in Choephoroi In order to test the hypothesis that P2 is the regular position for interrogatives in CG, the 80 finite interrogatives in Choephoroi were examined. If the 25 polar questions22 are discounted (because they are not marked as questions, or are introduced by the prepositives h\ or povtera),23 the 55 remaining interrogatives may be categorized in two groups: a) Line-initial, with respect to a clause start which coincides with the line: i) With particles following: 21.24 ii) With other enclitics following: 3 (87, 778, 844). iii) As quantifier in an NP: 2 (10, 530). iv) Other: 3 (88, 858, 871, all with interrogative pw'"). v) In P2: 12.25 b) Clause-initial, not coinciding with the line start: vi) With particles following: 6.26 vii) With other enclitic following: 1 (883). viii) As quantifier in an NP: 2 (12 and 885). ix) Other: 2 (88 and 871, both with pw'"). x) In P2: 3 (256, 408, 778). These figures show that most interrogatives are clause-initial, and followed by P2 enclitics. The unspecified ‘enclitics’ noted at (ii) and (vii) are all verbs, except the element at Choe.844 (the demonstrative tau't j). Pw'" is clearly a stronger prepositive than tiv, since it is the only wh-word to appear in first position without following enclitics, though Dover (1960: 12) classes it with the other interrogative/indefinite pronouns and adverbs. It is phonetically different from the others in having a long closed syllable. No other interrogative appears in first position without being followed by an enclitic, 22The expression derives from Jespersen (1924). Polar questions may also be termed ‘yes/no’ questions: see Lyons (1977: 754). 23These are: Choe.14, 90, 92, 112, 120, 122, 177, 220, 222, 224, 297, 339, 418, 495, 496, 526, 774, 775, 845, 894, 899, 909, 912, 1010, 1074. 24Choe.48, 110, 114, 123, 169, 171, 338, 418, 569, 638, 720, 732, 766, 847, 885, 900, 916, 994, 997, 1051, 1075. 25Choe.179, 214, 216, 218, 315, 394, 528, 532, 594, 855, 899, 925. 26 Choe.10, 187, 338, 388, 703, 880. 142 unless it functions as quantifier in an NP, in one construction at Choe.10–11: 17) tiv crh'mav 'v 'v ' leuvssw… tiv" poq j h{d j oJmhvguri" steivcei gunaikw'n favresin melagcivmoi" What thing do I see? what company of women conspicuous in black robes comes? 27 The line and clause start usually coincide, but this does not affect the pattern. Because clause-initial position does not occur without following particles (unless the interrogative is part of an NP), it is likely that the emphasis is created by the P2 particles themselves, and that interrogatives are in P1 only when focalized. Although that is their most common position (approximately double the number of interrogatives which are in P2), it may still be always emphatic.28 2b (iii): Interrogatives and the intonation break A semantic as well as morphological similarity may be noted with indefinites, which occur regularly in P2: interrogative ‘who?’ presupposes indefinite ‘someone’, and both are presupposed by relative ‘who’ (for references see Chapter 5, Section 2c). The difference in placing between the two may be identified through the existence of an intonation break: interrogatives are preceded by one, while indefinites are not. Dover (1960: 12–13) questions whether Ag.1344 (si'ga: tiv" plhgh;n aju>tei' kairivw" oujtasmevno"… Silence: who/someone tells of a blow, mortally wounded) is really a question or an indefinite statement. Yet, although tiv" is in second position in the line, it is first in the clause, following an intonation break, suggesting an interrogative, though the pattern of emphasis creates ambiguity (Ag.553–4 tiv" de; plh;n qew'n / a{pant j ajphvmwn to;n di j aijw'no" crovnon… ‘who except the gods is without pain all his life’ is more clearly a question). Conversely, at Ag.449, tiv" is truly enclitic to si'ga, 27Dindorf (1851: Vol. 1) adds before line 10, which could provide a prosodic host. Dindorf (1876: 97) justifies this by the construction at Prom.298: e[a: tiv crh'ma… kai; su; dh; povnwn ejmw'n .... (Ah, what [is] the matter? Have you too come [to gape] at my torture?) 28Left-dislocated constituents regularly precede wh-clauses in Vedic too (see Luraghi 1998: 192). 143 with no intonation break, so an indefinite meaning is more likely:29 18) tavde si'gav ti" bauv>zei, these things someone /who mutters in a whisper Indefinite ti" is part of the same word group as a preceding word, so may be considered a cohesive focalizer, while it is itself focalized at Ag.553 by dev. 2b (iv): Relatives In relative constructions, the prosodic relation between an antecedent and the relative pronoun is analogous to that between a P1 word and an interrogative, as discussed above. This is particularly clear in constructions where there is also a metrical break before the head noun, when it is line- initial, as at Od.9.197: 19) hJdevo"J vJ vJ v , o{n moi dw'ke Mavrwn, Eujavnqeo" uiJov", sweet [wine], which was given me by Maron, son of Euantheos... Ag.1433: 20) “Athn jErinuvn qj vj vj v j, ai|si tovnd j e[sfax j ejgwv, by Ruin and the Erinys, to whom I sacrificed him and Eum.484: 21) qesmovnvvv , to;n eij" a{pant j ejgw; qhvsw crovnon. an oath, which I shall set for all time. Here, the prominence of the head noun creates a particularly strong contrast with the relative pronoun. However, the prosodic relation between head noun and relative is contrastive, even when there is no intonation break before the noun, as at Eum.661: 22) e[swsen e[rno"[[[ , oi|si mh; blavyh/ qeov". She preserves the offspring for such as a god does not harm Septem 426: 23) puvrgoi" d j ajpeilei' deivnvvv j, a} mh; kraivnoi tuvch: He threatens our towers with terrors, which may fortune not fulfil 29The position of ti" as third word implies that tavde is topicalized, like the demonstratives in Section 1d. 144 and Ag.525–6: 24) Troivan kataskavyanta tou' dikhfovrou Dio;" makevllhvvv /, th'/ kateivrgastai pevdon. he who uprooted Troy with the pick of avenging Zeus, with which the earth has been worked over. An intonation break between head noun and relative pronoun may be observed in all these constructions, including Eum.661 (a restrictive relative, which in English would not have ‘comma intonation’). The close relationship between prosody and meaning is demonstrated by the rare constructions where the head noun is not adjacent to the relative. Such a purely prosodic relation is stylistically uncomfortable, as may be seen from Ag.1221–2: 25) su;n ejntevroi" te splavgcn j, ejpoivktiston gevmo", prevpous j e[conte"[[[ , w|n path;r ejgeuvsato. ... and with the entrails, the viscera, a pitiful load, they seem to be holding, which their father tasted. The clause boundary is schematized in Fig. 3 (=Introduction, Fig. 5):30 PP IP pro;" tou;" kratou'nta" ou}" Fig. 3 NP P2 CP P1 i[doim j ejgwv pote Obj. 30The citation is from Choe.267, cited also above: ‘[someone might tell this] to the rulers, whom may I see die one day’. 145 This demonstrates that the P1>P2 sequence is a relational one, which holds across syntactic boundaries (as noted by Hock 1996), and creates a prosodic link, additional to the co-referential one. 2c: P2 and the intonation break The mechanism of cohesive focalization is described below in Section 4. Focalizing particles may also be distinguished prosodically from pronominals by their ordering in P2 collocations. 2c (i): Function and order in P2 collocations It was observed by Delbrück (1900: 51) that, in early Greek, connectives precede pronouns, and Watkins (1964: 130ff.) generalizes that feature as common to all early IE languages. Denniston (1954: lx) notes that sentence adverbials usually follow connectives. Recent models have focussed on the relative order of enclitics in collocations, especially the ‘initial string’ in Vedic (Hock 1982, 1996; Hale 1987, 1996; Schäufele 1996), Hittite (Luraghi 1998), and Homeric and Koine Greek (Ruijgh 1990, Wills 1993, Taylor 1996). As noted above in Section 2a, interpretative approaches tend to be polarized between the syntactic and the prosodic. The former is exemplified by the schema of Ruijgh (1990: 223) for Homeric particle clusters: 1) Adverbials with single-word scope (per, ge, mav, mavn, mevn). 2) Preparatory co-ordinating connectives (mevn, te). 3) Connectives (dev, gavr, te, mevn). x) [see below] 4) Sentence adverbials (a[ra> nu, epic te > ke, a[n >qhn, ou\n> dhv> au\). 5) Indefinite pronouns and adverbs. 6) Personal pronouns. Ordering like this seems both over-rigid and incomplete, and Ruijgh (1990: 225) admits the exception that a[ra and nu may follow dhv, and the categorization of gavr only as a connective and ou\n as an adverbial is partial. A gap (x) has been left in the listing above, where interrogatives and relatives are placed (see Hale 1987: 42, Wills 1993: 72). 146 Ruijgh (1990: 217) explains the sequence in terms of syntactic domain, because ‘un postpositif suit immédiatement le mot initial de la séquence qui constitue son domaine’, so in a complex sentence at Il.3.396–8, the postpositives are ordered by increasing domain, with the connective t j at 398 preceding the adverbial a[r j because its domain is the whole line (qavmbhsevn ... ojnovmaze), while t j has scope only over the first clause of it (qavmbhsevn ... e[peita): 26) kaiv rJJ JJ j [ [ wJ" ou\\ \\n ejnovhse qea'" perikalleva deirh;n sthvqeav q j iJmeroventa kai; o[mmata marmaivronta, ] qavmbhsevn t j a[r j[ j[ j[ j e[peita, e[po" t j e[fat j, e[k t j ojnovmaze:] And then, [ [ as she recognized the round sweet throat of the goddess and her desirable breasts and her eyes that were full of shining,] she both wondered, and spoke a word, and called her by name] Organization by increasing domain clearly has some validity, as it suggests why particles like ge, per, and adverbial w{" (which have scope over single words) precede others in collocations. The position of preparatory particles and adverbials may also be explained by their domain: preparatory mevn and te (in te kaiv) precede connectives because their function is limited to the first clause of a compound sentence, while a connective like gavr has scope over both: Ruijgh (1990: 218) illustrates this by the use of me;n ga;r at Il.11.825.31 However, a principle of increasing domain does not explain the standard (decreasing) sequence of [indefinite > personal pronoun], as Il.4.245:32 27) ... eJsta's j, oujd j a[ra tiv" sfivvv meta; fresi; givgnetai ajlkhv. (connective> adverbial > indefinite > personal pronoun) .... stand still, and there is no heart of courage within them? Nor does it explain the decreasing sequence of d j and a[ra at Il.5.47: 28) h[ripe d j ejx ojcevwn, stugero;" d j a[ra minj [j [j [ skovto" ei|le (connective > adverbial > personal pronoun) he dropped from the chariot, and the hateful darkness took hold of him Position by domain does not, therefore, give a full description. An alternative analysis of P2 in terms of word movement has been suggested for Vedic by 31Wills (1993) adopts a similar analysis, observing that particles which emphasise single words are normally adjacent to them, so precede connectives and sentence adverbs. His examples are ge and per, and he notes that mevn, dhv and nu may be placed in the same position. 32Although tiv" is here an adjective, it occupies the same position as would a pronominal. 147 Hale (1987), who proposes that sentence adverbials follow connectives because they are enclitic to P1 as defined after wh-movement (the preposing of interrogatives and relatives) but before the focalization of lexical words,33 while connectives are, as it were, ‘inserted’ last, as discourse rather than sentential elements, so precede question words. Emphatics are ‘cliticized to the constituent they will emphasize, and ... following this the emphasized element can be topicalized’ (Hale 1987: 46). This model is formalized by Hale (1996) in terms of two maximal projections: a topic phrase and a focus phrase (the latter being the position for pronominals and adverbials). The sequence accords with the observed order [emphatic > connective > wh-word> adverbial], so agrees with the collocation sequence more accurately than Ruijgh’s principle. However, it is inadequate in two respects. It does not match the prosody, and Hale (1996: 178) admits to being ‘not entirely comfortable with “focus” as a general name for that function’. Secondly, it is uninformative: no explanation is given of why focalization should occur ‘after’ wh-movement, or why indefinites should precede demonstratives. Neither model explains the overlap between interrogatives and indefinite pronouns, which may be placed in a different sequence, either preceding or following sentence adverbs. Enclitic position has also been described in terms of prosody, as in the proposal of Halpern (1992), that P2 elements are, structurally, the leftmost elements in the clause, adjoined to the IP, but are placed in P2 by a prosodic ‘flip’, which allows an enclitic to move one place rightwards if it lacks a host, due to an intonation break to its left. In Halpern’s view, the sequence [P1 phrase>P2 clitic] results from phrasal preposing (to the specifier position in CP), but the sequence [P1 word>P2 clitic] results from an automatic ‘prosodic inversion’ of the two elements. The difference is that only the first, phrasal, construction involves pragmatic motivation. However, it has also been suggested that phrases as well as single words can undergo prosodic inversion (see Hock 1996, Taylor 1996). Observations of the variable positions of gavr in the corpus texts, discussed in Section 3b below, suggest that phrases and single words are focalized analogously. 33The focalized word may of course itself be an interrogative, as at Il.5.703: e[nqa tivnavvv prw'ton, tivnavvv d j u{staton ejxenavrixan… (then whom first and whom last did they slaughter?). 148 Prosodic analyses have usually been presented as alternatives to syntactic interpretations: one may detect a rather confrontational tone in the way Hock (1996) and Hale (1996) comment on each other’s models. It seems more useful to investigate the relation between prosodic and syntactic approaches, and a combined approach seems particularly relevant to the structure of the clause break in subordination, as is discussed below. 2c (ii): The cohesive focalizer/operator division Prosody and syntax have generally been considered separately, since P2 enclitics are not always syntactically linked to their prosodic host. In particular, it is usually held that, in Homeric Greek and CG, only emphatics follow their host in terms of both phonological liaison and syntactic precedence (as Hock 1996). However, it is argued here that connectives, adverbials and indefinite pronouns do too, and the only syntactic division is between the cohesive focalizers (emphatics/ connectives/ indefinites / adverbials) and operators or ‘wh-words’ (interrogatives/ relatives); a division which is also marked prosodically by an intonation break. The whole sentence, including the P1>P2 string, may then be schematized as in Fig. 4 (=Introduction, Fig. 1): [ { P1 } cohesive focalizers] [ interrogatives and relatives [personal pronouns {clause} ] ] The cohesive focalizers include connectives, indefinite pronouns, emphatics and other adverbials. Enclitic verbs may be grouped with the operators, since their scope is analogous (as noted in Section 2a above). P2 then consists of two word groups divided by domain and by the intonation break. This gives a different ordering from the canonical one in two principal respects: i) Particles normally defined as sentence adverbs (a[ra, nu, epic te, ke, a[n, qhn, ou\n, dhv, au\) form part of the initial intonational group. ii) Indefinite pronouns precede operators (unless the latter are focal, in P1). The morphology of the P2 element accords with this interpretation: the distinctively harsh articulatory onset of a voiceless central fricative (like ou{" 149 and other relatives) may reinforce the intonation break.34 Of course, [/h/] is not necessary to create an intonational pause, since a boundary also occurs with interrogative tiv, but it is suggestive that not only do relative pronouns start with an aspirate, but complementizers do also. The accent borne by an operator is, presumably, also indicative of increased vocal friction compared to an indefinite or other accent-less enclitic. It might be objected that a prosodic division after the cohesive focalizers does not correspond to syntactic domain: after all, emphatics have a single-word domain, connectives link two clauses, and adverbs modify a verb or a whole clause. This is what prosodic interpretations like those of Hock (1996) and Taylor (1996) assume. However, it can be shown that emphatics and connectives in fact have the same domains, and the same can be attributed to sentence adverbs too: their position in collocations may vary according as their prosodic ‘target’ is the word in P1, or a non-preposed wh-word in P2. In either case, their function is primarily local. It will be shown in Section 4 that this is true for all cohesive focalization, and, in Chapter 5, that indefinites also function cohesively. An alternative analysis is possible for the position of operators: they could be in P1, preceded by topicalized elements. As noted in the Introduction, Section 2d, this analysis is not adopted, because it fails to model the prosodic prominence of the initial elements in contrast with the operator, and it does not explain the interrogative patterns discussed above in Section 2b, or the regular presence of focal elements preceding wJ" in tragic complementation (discussed in Chapter 6). The focalizer-operator division gives a more unified analysis. In the next section, evidence for a single function of ‘cohesive focalizer’ is discussed. 3: Cohesive focalization Cervin (1990: 59–65) notes that CG particles have a regular emphatic effect on constituents to their left (or complex constituents in which they appear late). It is argued here that the emphatic and linking functions are always shared by the same particles, and particles with one function are always associated with 34Cavity friction would certainly disrupt the intonational pattern. See Allen (1987: 18–19) on the articulation of CG aspirates. 150 both. Further, overlaps may be observed in the linking and adverbial functions too. 3a: Overlaps in particle function A coincidence of emphatic and connective function has often been noted. Denniston (1954) suggests an aetiology for three groups of particles. Though tentative (cf. Rijksbaron 1997b), it makes the case for a causal connection between the two functions: 1) Some (ge, dhv, mhvn, toi) may have originally expressed thoughts in isolation, but also have connective functions (Denniston 1954: xxxvii). 2) Conversely, primarily connective particles may have emphatic functions. Denniston (1954: 359) believes that ‘the primary function of mevn, as of mhvn, is emphatic’. Other connectives with emphatic force include gavr (considered below, Section 4b) and ou\n (Denniston 1954: 416–425). Responsive particles like dev and te also emphasise the preceding word. 3) Apodotic function in conditionals (i.e. in the consequent clause) may involve adverbials (ge, h\, mevntoi, dhv) and paratactic connectives (kaiv, ou\n, toivnun). Denniston (1954: xl) suggests an adverbial origin for this. 3b: Diachronic changes in position and function The interpretation of variable particle position as historically transitional (as in Denniston’s schema described above)35 suggests the possibility of a functional connection between emphatics and connectives, in three respects: i) Origin. The remark by Denniston (1954: xl–xli) that ‘it is by no means certain that the connective sense of any Greek particle is the original sense’ implies that at least some particles changed function from adverbial to connective. A general development from adverbial to connective functions is a plausible inference, since intra-clausal meaning might be expected to precede a linking function chronologically (since it precedes logically). ii) Rightwards movement. The converse development is better established: in post-Homeric Greek there was a diachronic movement of enclitics from P2 to 35See also Ruijgh (1990: 221ff.) on pevr, a[ra dhv and dhv rJa, and Wills (1993: 63n7) on a[ra. 151 adjacency to their head words (see Dover 1960: 15–19, Horrocks 1990).36 Rightwards movement is also influenced by phonetic factors: Horrocks (1990: 39) notes that in CG the enclitic pronoun moi may be placed next to the most emphatic sentence constituent, rather than to its governing verb (as at Xen. Resp. Ath. 3.10 and Dem. 37.23). Movement may also be motivated in part by an intonational change: postponement is linked by Dunn (1989) with the movement of a postulated sentence accent (as the Satzmelodie of Schwyzer 1950) from Wackernagel’s position in CG (Comrie 1980: 86) to fall on the main verb in Hellenistic Greek (Dunn 1989: 11). Adams (1994a) observes a similar placing for weak pronouns in Classical Latin. There appears also to have been a rhythmic component (see Marshall 1987). iii) Rightwards movement and subordination. A second motivation for postponement may be found in the loss of connective function, which was increasingly fulfilled by subordinating conjunctions like ejpeiv, eij, wJ", and o{ti. The discussions of eij below in Section 5a and of wJ" in Chapter 6 suggest that this was a gradual process, with conjunctions showing enclitic features before occupying an independent position in a developed system of subordination. There does, then, appear to be a historical progression, which strengthens the case for a functional explanation. In Section 4, it is argued that the causal connection may be found in the mechanism of focalization. In the next section, a constant emphatic/connective effect is noted in constructions with gavr, and, in Section 3d, the same link is noted for other particles. 3c: Gavrvvv and dual function Denniston (1954: 56–57) considers that the core function of gavr is causal, all instances having a connective function. However, noting that ‘few Greek connecting particles started their careers as conjunctions’, he suggests that ‘an earlier, asseverative, force lay behind the causal sense’, and links this to an etymology of gavr from ge + a[ra. Smyth (1956: 638) has the same view, suggesting that ge originally gave prominence ‘either to the word it followed 36Head-government appears as early as Homer too, explaining a feature which Wills (1993: 66) describes as ‘third position’, adducing constructions like Il.1.81: ei[[ [[ per gavr te covlonvvv ge kai; aujth'mar katapevyh/ (and suppose even for the day itself he swallow down his anger ...). These, however, are better explained as head-governed: ‘[covlon ge] [kai; aujth'mar]’. Cf. Ruijgh (1990: 219–20) on ge at Il.2.703. 152 or to the whole clause, while a[ra marked this prominence as due to something expressed or latent in the context.’ Such an interpretation raises the possibility that retrospective and preparatory function may reflect the same force, and indeed anticipatory (preparatory) use is considered by Denniston (1954: 68) to be a stylistic variant, rather than an independent asseveration (which was the view of Kühner 1904: 332–3). Retrospective and anticipatory gavr simply reverse the dependency ordering: ‘since a, therefore b’ is logically equivalent to ‘b, because a’, and parenthetical gavr following the main subject could be translated as either (or ‘as’), as Ag.1069: 29) ejgw; d j, ejpoiktivrw gavrvvv , ouj qumwvsomai but I, because/since I pity her, will not be angry and Eum.230: 30) ejgw; d j, a[gei ga;r;;; ai|ma mhtrw'/on, divka" mevteimi tovnde fw'ta kajkkunhgevth". but I, since a mother’s blood urges, shall pursue this man to punish him, and be his tracker 37 Throughout the corpus, causal force is always combined with pragmatic emphasis, as at OT.409: 31) i[s j ajntilevxai: tou'de''' ga;r kajgw; kratw': ... arguing equally, for in that I rule too Med.1376: 32) pw'" ou\n… tiv dravsw… kavrtavvv ga;r kajgw; qevlw. How? What shall I do? For I very much wish that too. 37Denniston (1954: 58) believes that retrospective use of gavr is ‘commoner in writers whose mode of thought is simple than in those whose logical faculties are more developed. The former tend to state a fact before investigating its reason, while the latter more frequently follow the logical order, cause and effect’. In the corpus, there are in fact fewer instances of gavr in Crito than the other texts: one per 320 words, compared with 100 in OT. (c. per 12 lines), Med.112 (15 lines), Ra.150 (16 lines). Frequencies in the Oresteia are less: 170 (one every 22 lines), and in Septem even less: 251 (40 lines). However, this seems based on presentational rather than logical criteria: see Sicking and van Ophuijsen (1993: 21ff.). 153 and Crito 50b6: 33) pro;" tau'ta kai; a[lla toiau'ta… polla; ga;r a[n ti" e[coi, towards this and other similar things, for one might have many things [to say] While the identification of pragmatic emphasis must be a subjective judgment, it is sufficient for this argument that all constructions are consistent with an emphatic interpretation of the initial word. The relation between pragmatic and prosodic emphasis may be interpreted in terms of factive rather than causal force, as in the suggestion by Sicking and van Ophuijsen (1993: 24) that gavr is responsive in the sense of ‘making explicit a presupposition of the previous sentence.’38 Such a unified explanation in fact accords with the range of functions which Denniston (1954: 63, 65) identifies for causal gavr, referring loosely to the preceding discourse, or more specifically to a single clause or individual word. The categorization of gavr by Ruijgh (1990) as purely co-ordinating is therefore incomplete. The effect of focalization appears to be inherently local: the view of Halpern (1992), cited above, that only phrasal focalization is pragmatically motivated, while focalization of individual words is an automatic ‘flip’, is not supported by the constructions in the corpus, where gavr is always adjacent to the word or phrase it emphasizes pragmatically, though when gavr follows NPs, the whole phrase may be regarded as prominent, as at Ag.32: 34) ta; despotw'n; '; '; ' ga;r eu\ pesovnta qhvsomai for after my master’s lucky throw, I shall make the next move and Ag.461: 35) tw'n poluktovnwn' v' v' v ga;r oujk a[skopoi qeoiv, ... for of the killers of many, the gods are not unwatchful In contrast, when gavr appears within an NP, as at Eum.334, only the demonstrative seems emphatic, so a specific pragmatic effect may be understood: 36) tou'to''' ga;r lavco" diantaiva Moi'r j ejpevklwsen ejmpevdw" e[cein, For this lot, piercing Fate spun for us to be permanent 38Factivity may be defined as the transference of presuppositions between clauses. See Kiparsky and Kiparsky (1970), Karttunen (1971), Lyons (1977: 599–606, 794–809) and the discussion in Chapter 6, Section 1c. 154 Following a VP at Ag.222, the adjacent element, rather than the whole phrase, appears pragmatically emphatic (as Dover 1987: 61 suggests): 37) brotou;" qrasuvneivvv ga;r aijscrovmhti" tavlaina parakopa; prwtophvmwn: Woeful madness, the first cause, suggesting evil, emboldens men; The variable position of connective particles with respect to an initial phrase is described by Dover (1960: 16) as ‘the result of a compromise between pattern and principle’ in the interaction between prosody and syntax. If connectives are always also emphatic, as proposed here, variable placing may be understood as reflecting variable emphasis, which is presumably open to authorial choice. Though gavr may emphasise cohesive material (as at Ag.222 and Eum.334 above), Dover (1987: 61–3) argues that it tends to emphasise new information. The reason is prosodic: an emphatic function certainly has a contrastive effect, and when the emphatic and linking functions are in tension, the former is dominant, as at Eum.797, where it is adjacent to ejk Diov", though its linking function is related to that of ajll j: 38) ajll j ejk Dio;"j ;j ;j ; ga;r lampra; martuvria parh'n, and yet, from Zeus, there was clear witness, The primacy of the emphatic function may also be seen when gavr follows grammatical words which themselves are primarily connective, as in eij gavr,39 and h\ gavr, which appears to have a sense of urgency in Plato (Erp Taalman Kip 1997). Its dominance follows from the mechanism of cohesive focalization (described below in Section 4a). 3d: Other cohesive focalizers Other connectives and adverbial particles also have a regular emphatic force. The high frequency of collocations beginning with mevn (as mevn rJa, mevn ge, me;n dhv, me;n ou\n, mevntoi), not only in Homer, but also tragedy and prose, accords with a combined emphatic and linking function. This may be seen 39See Misener (1908). 155 even when me;n gavr follows a monosyllabic grammatical word, as at Il.9.515: 39) eij me;n ga;r mh; dw'ra fevroi ta; d j o[pisq j ojnomavzoi and yet, if he were not to bring gifts and to name still more hereafter... and Od.9.132: 40) ejnjjj me;n ga;r leimw'ne" aJlo;" polioi'o par j o[cqa" for in [it] there are meadows near the shore of the grey sea ... The functional overlap is at least partly due to position: if the most emphatic constituent of the sentence is standardly initial, a particle following it may have either function: it cannot, for example, be known by a listener at the moment of utterance that there will be a responsive clause, so every instance of mevn must carry the possibility of emphasis.40 The contrastive quality of the emphatic function may be seen in the force of dev, which, Bakker (1993) argues, regularly marks a variety of discourse boundaries, both continuative and contrastive.41 The importance of ‘adversative’ function is also observed by Basset (1997), Slings (1997), and Jacquinod (1997), in studies of orthotonic conjunctions and P1>P2 sequences.42 However, dev does not simply mean ‘but’: Davies (1997) shows that, in Arcadian, dev changes from a ‘seriously’ adversative particle into a continuative one, functioning as a textually cohesive marker.43 As noted in the previous section, the probable origin of gavr from ge and a[ra depends on their functions as making some sentence element prominent, and other adverbial particles also seem to be cohesive focalizers, placed in the same word group as their prosodic host.44 Denniston (1954) identified emphatic uses of qhn, ou\n, and dhv, and the pragmatic functions which Sicking (1997) identifies in Platonic interrogatives always involve the relative 40The frequency of particle collocations following grammatical words may also have a morphological significance, in providing the weight to focalize a monosyllable. 41Bakker’s analysis is based on discourse framing: for which see Goffman (1974) and Tannen (1993). 42Basset examines ajllav; Slings ajllav, ajlla; gavr, mevntoi, and kaiv toi; and Jacquinod kaivtoi. 43For discussions of the pragmatic functions of dev, see also Dover (1987: 59–61) and Rijksbaron (1997c). 44Assimilation to a larger phonological unit may be seen in a capacity to function as affixes (dev: oijkovnde, o{de, hjdev; dhv: ejpeidhv; and te: o{ste); or with affixes added (dh'ta). 156 emphasis of the P1 and P2 elements (ou[koun as expressing disbelief while oujkou'n elicits assent, tiv dev as expressing a new discourse topic while tiv ou\n marks logical coherence). It is argued in Chapter 5 that epic te also focalizes its relative host, and a contrastive textual effect is implied in its function of marking a digression from the main narrative.45 A unified view of focalization may seem at odds with the great variety of particle functions: as Rijksbaron (1997b, 12) argues, ‘instead of treating “the” particles as one monolithic block in “the” Greek literature, there is a need for studies dealing with (groups of) particles in specific dialects, genres, authors and discourse types.’ However, the intention here is not to deny the value of such specificity, but to suggest a single mechanism which may underlie the multiplicity of functions. This is described in the next section. 4: Focalization and linking In this section, emphatic, connective, and adverbial particles are shown to share the same prosody and function, though the syntactic domain of each appears quite different, because they all function in the same way: by creating a unified prosodic group centred on a focalized word in P1. 4a: Mechanism Although the discourse functions of grammatical words have been intensively studied, their meanings have usually been considered as purely conventional (as Grice 1989), and their phonology has not been investigated. Conjunctions are usually described simply by their logical function: as ‘signals’ (Hockett 1958: 153–4), ‘markers’ (Matthews 1981: 60–69), or ‘co- ordinators’ (Quirk et al. 1985: 918–1007). It is proposed here that the connective function is a kind of emphasis, because, just as a flag attracts attention by being physically prominent, so connectives also function, by making a contiguous element phonologically prominent. The proposed mechanism has three principal aspects: 1) Phonological. The cohesive focalizers function by assimilation, adding their morphological weight to the prosodic host. This difference correlates with 45Gonda (1954b: 274) compares it to an actor’s ‘aside’. The classic study of ‘te digressif- permanent’ is by Ruijgh (1971). The earliest is by Bäumlein (1861: 227–235). 157 order in P2 collocations: primary emphatics like ge and per precede connectives for reasons of scope or domain, as in Ruijgh’s model, because emphasis is inherently short-range (no particle emphasizes a preceding sentence, though responsive gavr may come close). It may be noted that, though pronominals are usually separated from the P1 group by an intonation break, the pronominal toi can function cohesively as a particle, though properly an ethical dative.46 2) Cognitive. The word in P1 is consequently prominent, not only through focalization by the particle following, but in contrast with the preceding text. It therefore attracts attention to its own meaning. If it is a referring expression, it introduces a new referent into the discourse. If it is a pronominal, focalization creates indefinite reference, so raises the question ‘what?’. 3) Textual. The prominence of the P1 element has two pragmatic effects: of marking its own meaning as important, new, or in contrast with the preceding text, and of introducing the new proposition of which it is a part. Focalization is therefore, in pragmatic terms, epiphenomenal: it is not the presentation of any particular category of information, but the intensification of that presentation: it could be described as phonological amplification. Its pragmatic effects vary according to the P1 element, as well as the discourse context. The observation that particles have a great range of pragmatic functions does not preclude the possibility that they may function in a similar way. The core meanings of individual particles are compatible with a unified model of focalization, in terms of what element is in P1. 4b: Pragmatic functions of focalization The discourse functions of cohesive focalization are realized in four ways: 1) When connectives are P2 enclitics, the clause needs an initial, prominent element functioning as an attention-getting device. If the P1 element is a lexical word, the link is automatically contrastive, because new information is being foregrounded. This explains the typically contrastive character of CG coordination, and also the association of focus with new information, noted above in Section 1a: the word chosen as highlighted in contrast with the preceding text is likely to be emphatic in a communicative sense, even though 46Liddell and Scott (1968: 1801) note that its function corresponds to focal stress in English. 158 this is not the primary motivation for its prominence. A focal word is functionally necessary, but the pragmatic effect varies according to the word selected. 2) Subordinate clauses, however, lack a P1 element. In relative clauses, the pronoun is syntactically initial, though prosodically in P2 (as schematized in Fig. 3 above). This creates a tension between prosody and syntax, which may lead to ambiguity. In relatives with a nominal head, the relative pronoun is clearly a P2 element, so is not followed by enclitics, but in free relatives, where the relative pronoun can be interpreted as in P1, there is nearly always an enclitic or affix following. This creates a combined [P1>P2] focal grouping. 3) While the reference of a pronoun may be anaphoric or anticipatory, focalized demonstratives have indefinite reference when they function as interrogatives, which are anticipatory (o{sti"). The same focal and indefinite function is performed by indefinite ti" in free relative clauses introduced by o{sti", and epic te (which marks a clause as denoting a class rather than an individual)47 when it follows an interrogative pronoun (see Chapter 5). 4) The demonstratives therefore function by ‘impure textual deixis’ (Lyons 1977: 668), as they point to a proposition, rather than to a referring expression in the text.48 A few constructions in tragedy (with ou|to") are retrospective, as Ag.1303, 1320, and 1617–8: 41) su; tau'ta''' fwnei'", nertevra/ proshvmeno" kwvph/, .... you speak this, seated at the lower oar However, the majority of demonstratives are anticipatory, with o{de. The typical tragic form is as Pers.356–7: 42) ejlqw;n e[lexe paidi; sw/' Xevrxh/ tavdevvv : wJ" eij melaivnh" nukto;" i{xetai knevfa" ”Ellhne" ouj menoi'en .... 47Though Ruijgh (1971: 9) argues that its meaning is not simply generic. 48A similar view of complementizer function as textually-deictic is the basis of the influential model of Modern English complementation advanced by Davidson (1968). On deixis, see Bühler (1934, 1982), Lyons (1977, 1982), Jarvella and Klein (1982), Ehlich (1982), Rauh (1983), Morel and Danon-Boileau (1992), Létoublon (1992a), and Green (1995). On the distinction between textual deixis and anaphora, see Lyons (1977: 442–3) and Ehlich (1982). 159 [A Greek,] coming to your son Xerxes, told him this, that when the darkness of black night should come, the Hellenes would not remain... The demonstrative has a double effect: a local one of de upon tav, and the effect of tavde on the clause wJ" . . . menoi'en. It is shown in Chapters 5 and 6 that functions (3) and (4) are central to the development of complementation. Functions (1) and (2) not only affect co- ordinated linking, but also link prosodic structure to clause order. This is examined in the next section. 5: Focus and clause order Throughout the textual corpus, there is a correlation of clause order with the placing of the focal element: 1) When a dependent clause precedes, the focal element is syntactically within it. The P1 element may be the conjunction itself, as at OT.1266–7: 43) cala'/ kremasth;n ajrtavnhn: ejpeijjj ; de; gh'/ e[keito tlhvmwn, deinav d j h\n tajnqevnd j oJra'n. [... he loosened the halter]; and when the wretched woman lay on the ground, it was terrible to see what happened next. Alternatively, a focal element may precede the conjunction, as at Ag.866–8. The focal element is clearly not thematic (‘wounds’ is a new topic): 44) kai; traumavtwn me;n; v ;; v ;; v ; eij tovswn ejtuvgcanen ajnh;r o{d j wJ" pro;" oi\kon wjceteuveto favti", tevtrhtai diktuvou plevw levgein: and of wounds if this man was enduring as many as rumour was channelled to the house, he has now more holes to count in him than a net. 2) When a dependent clause follows, it is the last element of the main clause which is focalized. As discussed in Section 2b (iv), this effect may be seen in relativization, when the head noun in the main clause is emphatic, but the same pattern is also evident in conditionals, even though there is no main clause head noun. 160 The structural difference between (1) and (2) may be motivated by the requirements of clause linking, but the effects are to focalize the initial element of the first clause, whether the main or the dependent, and to make a (dependent) second clause more prosodically integrated with the main. It will be shown that it can make it more syntactically integrated too. 5a: Focus and conditional order Prosodic integration can be seen especially clearly in conditional constructions, where clause order is highly variable. Friedrich (1975: 20 n.10) states that conditionals normally precede their main clause in Homer, but it has also been observed (by Houben 1977, Dunn 1988), that conditionals more often follow than precede their main clause in Homer and Herodotus 1. In the texts of the corpus, conditionals precede the main clause more often than they follow it: Frogs is the only text to have a higher proportion of conditionals following. The overall ratio is about 60/40, but there is considerable variation, with a particularly high proportion of conditionals preceding in Odyssey 9, Septem, OT. and Cyclops. Conditional order with respect to the main clause is schematized in Table 1:49 Preceding Interpolated Following Iliad 9 17 12 (=41%) Odyssey 9 7 3 (=30%) Septem 8 3 (=27%) Agamemnon 23 2 13 (=34%) Choephoroi 10 9 (=47%) Eumenides 7 6 (=46%) OT. 61 25 (=29%) Medea 23 4 17 (=39%) Cyclops 9 4 4 (=24%) Frogs 23 5 27 (=49%) Melian Dialogue 6 4 5 (=33%) Crito 30 1 20 (=39%) Totals 224 20 144 (=37%) 49The percentages are given with respect to the total number of conditionals in each text. 161 A very strong correlation between position and structure is evident. As noted in the previous section, when a conditional precedes the main, the conjunction may be cohesively focalized by P2 particles, as at Ag.1025–9: 45) eij de; ;;; mh; tetagmevna moi'ra moi'ran ejk qew'n ei\rge mh; plevon fevrein, profqavsasa kardiva glw'ssan a]n tavd j ejxevcei: and if one fate, appointed by the gods, did not restrain another from going too far my heart outstripping my tongue would pour this out. However, preceding conditionals more often have a substantial focalized constituent before the conjunction, as at Ag.345–7: 46) qeoi'" d j ajnamplavkhto"' j j v' j j v' j j v eij movloi stratov", ejgrhgoro;" to; ph'ma tw'n ojlwlovtwn gevnoit j a[n, .... and even if, without wandering from the gods, the army return, the awakened pain of the dead might arise: ..... and Ag.563–4: 47) ceimw'na d''' j eij levgoi ti" oijwnoktovnon, oi|on parei'c j a[ferton jIdaiva ciwvn, and winter, if one were to speak of it, bird-killing, how intolerable the snow of Ida made it ... 50 These constructions are analogous to the interrogative patterns discussed in Section 2b (i), and demonstrate that eij also regularly appears in P2. The semantic effect may be, as in Ag.563 and 671–2, to increase inter-clausal cohesion by foregrounding a referring expression which is co-referent with (Ag.671), or even syntactically part of (Ag.563), a constituent in the main clause. 50The sentence continues with further subordinate clauses: a possible main clause occurs at Ag.567 (tiv tau'ta penqei'n dei'… ‘Why should one lament for this?’). 162 Interpolated conditionals, however, always follow emphatic elements in the main clause, as Ag.37–8: 48) bevbhken: oi\\ \\ko" d j aujtov"j j vj j vj j v , eij fqoggh;n lavboi, safevstat j a]n levxeien: wJ" eJkw;n ejgw; The house itself, if it had a voice, could best say... and Medea 66: 49) sigh;n;;; gavr, eij crhv, tw'nde qhvsomai pevri. for silence, if necessary, I shall keep silence about this. Similarly, when the conditional follows the whole main clause, initial emphasis occurs in the main clause, as at OT.1166: 50) “Olwla", ei[ se tau't j ejrhvsomai pavlin. You die, if I ask you this again. Emphasis may be purely formal, with no pragmatic force, as at Medea 1134–5, where the contrast seems rather to be between tevryeia" and pagkavkw": 51) levxon dev: pw'" w[lonto… di;" tovson ga;r a]n tevryeia" hJma'"J 'J 'J ' , eij teqna'si pagkavkw". Say then, how did they die? For you will give twice as much pleasure to me, if they died horribly. Only one following construction in the corpus was found to have focalized elements syntactically within the conditional, at OT.120–1. Here, the demands of line integrity apparently outweigh textual cohesion, demonstrating that a conditional may be structurally very peripheral:51 52) To; poi'on… ’En ga;r povll j a]n ejxeuvroi maqei'n, ajrch;n bracei'anj ; 'j ; 'j ; ' eij lavboimen ejlpivdo". ..... for one thing might be the way to learn many, if we could gain narrow ground for hope. 5b: Conditional order and discourse function As discussed above in Section 4b, focalization appears to function as an attention-getting device, and may be said to introduce a new ‘participant’ in the discourse. In preceding conditionals, this is usually (before following 51Conjunctive wJ" is also sometimes followed by enclitics when it is line-initial, as wJ" is at Il.9.311 (wJ" mhv moi), Od.9.394 (w}" tou'), and Cyc.628 (wJ" mhv). 163 conditionals, always) a referring expression rather than the conditional itself. Clause order therefore functions as a way of modulating emphasis. This view has not been suggested in other discussions of conditional clause order, since conditionals are not considered to be structurally subordinated,52 and are consequently studied either in semantic terms (Strawson 1952, Traugott et al. 1986, Sanford 1989),53 or pragmatically (Wakker 1994), in respect to their informational relationship with the main clause. It has been observed (by Haiman 1978, and Sweetser 1990) that there is a pragmatic similarity between conditionals and topics, as the conditional gives the context for the main clause proposition. A topic function is likely to motivate a conditional to precede, since topics precede and comments follow in the Prague FSP model, and because a protasis may precede its apodosis for iconic reasons (Greenberg 1963: 84–5, Lewis 1973, Comrie 1986). A topic function could underlie the semantic similarity between conditionals and indefinite relatives which has been observed by Clapp (1891),54 Kühner (1904: 423), and Chantraine (1963: 245), and extended to temporals by Wakker (1992: 49), who notes the equivalence of the constructions: o}" a]n aJmavrth/, qeou;" quvesi paratrevpousi / ejavn ti" aJmavrth/, / o{tan ti" aJmavrth/ (anyone who errs, sacrifices to the gods). Such generic conditionals tend to precede the main clause. However, analogous relatives may follow, as at Od.1.352–3: 53) th;n ga;r ajoidh;n ma'llon ejpikleivous j a[nqrwpoi h{ti"{{{ ajkouvontessi newtavth ajmfipevlhtai For surely people give more applause to that song which [/if it] is latest to circulate among the listeners. 52Matthews (1981) notes that the ‘if...then’ construction is correlative. 53Relations between verb mood and semantic type are discussed by Gildersleeve (1876), Goodwin (1889), Koppers (1959), Gonda (1980: 149–196), Greenberg (1986), Bakker (1988b), and Horrocks (1995). 54Clapp collates 3,226 conditionals in tragedy, of which there are 398 in Aeschylus (=49.8 per 1,000 lines), 1,039 in Sophocles (=103.1 per 1,000 lines), and 1,743 in Euripides (=69.7). As well as using fewer conditionals, Aeschylus also uses comparatively fewer hypothetical ones. 164 It appears that, even though the generic type might be considered topical in the sense of giving contextual information, order with respect to the main clause may still be determined by emphasis. The predominance of preceding conditionals in the corpus (noted above) could, perhaps, be attributed to a topic function, but the large proportion of following conditionals would then lack explanation. A pragmatically-based taxonomy of conditional placing has been developed by Wakker (1994: 49), who divides conditionals into three types: predicational (where the protasis gives the condition for the realization of the state of affairs described by the main clause: ‘if it rains, I’ll take an umbrella’), propositional, (where the protasis gives the condition for its truth: ‘if the story is true...’), and illocutionary (where the protasis gives the condition for its speech act felicity: ‘if I may say...’).55 Her figures (Wakker 1994: 58, 60, 89) indicate that predicational conditionals generally precede the main clause in Greek, except in Homer (where they precede and follow in equal proportions), while propositional and illocutionary types more often follow (except in Thucydides and Xenophon).56 Wakker (1994: 84–88) associates preceding conditionals with an ‘orientation’ function (similar to theme) and following conditionals with an ‘elaboration’ function.57 The categorization into propositional, predicational and illocutionary types appears somewhat over-rigid. This may be seen by considering a preceding conditional at Il.9.604–5: 54) eij dev k j a[ter dwvrwn povlemon fqishvnora duvh/" oujkevq j oJmw'" timh'" e[seai povlemovn per ajlalkwvn, But if without gifts you go into the fighting where men perish, your honour will no longer be as great, though you drive back the battle. Wakker would, presumably, categorise this as predicational (since it gives the condition for the realization of the state of affairs designated by the main 55Speech act felicity conditions are defined by Austin (1962) and Searle (1969). Wakker (1994: 49n10) notes that her categories are roughly equivalent to the content, epistemic and speech act conditionals of Sweetser (1990). 56Wakker gives percentages only for the predicational type, so it is not possible to compare it with the [propositional + illocutionary] total, and so determine the overall order of conditionals in her data-set. 57Functions similar to, but wider than, the Functional Grammar categories ‘theme’ and ‘tail’. 165 clause). However, it also (obviously) gives the condition for the truth of the main proposition, so it could also be categorized as propositional. Conversely, a preceding conditional at Crito 45c2–3 is explicitly categorized by Wakker (1994: 84–5) as illocutionary: 55) eja;n de; bouvlh/ eij" Qettalivan ijevnai, eijsi;n ejmoi; ejkei' xevnoi oi{ se peri; pollou' poihvsontai ... and if you wish to go to Thessaly, I have friends there who will make much of you 58 However, this conditional also gives the condition for the realization of the state of affairs described by the main clause, and it therefore appears somewhat partial to categorize it as illocutionary only. The evidence from the corpus suggests that there is a causal component involved, because there is a strong tendency for past unreal conditionals (‘he would have ... if he had not ...’)59 to follow their main clause. The following conditionals in Homer cited by Houben (1977) are also overwhelmingly past unreal conditionals,60 so a pragmatic explanation appears more likely than a logical one (since there is no real cause>effect sequence in a counterfactual). This does not support Wakker’s argument, since these are propositional (and so should precede). The only pragmatic function which is always apparent is an emphatic one: it is clearly possible to place any conditional clause before or after its main clause, and the reasons for placing must therefore be logically independent of the semantic structure of the clause itself. Conditionals, then, share this feature with true subordinates: following clauses are prosodically more integrated (since less prominent). Because conditionals have a loose syntactic relation to their main clause, their clause ordering shows greater variety than that of other subordinates, but the basis on which order is based appears to be the same. 58To avoid any misunderstanding, the translation is taken from Wakker (1994: 84). 59See Smyth (1956: 518–520), and Sanford (1989: 182ff.), who terms them ‘backtracking conditionals.’ 60Houben calls them ‘contrary-to-fact conditionals’, but his examples are only of the past unreal type. 166 5c: Conditional order and indirect questions Prosodic integration leads to syntactic integration, because the conjunction can be interpreted as the complement of a transitive main verb. Conditionals following speech verbs may then be formally indistinguishable from indirect questions, as at Choe.105: 56) levgoi" a[n, ei[[ [[ ti tw'ndæ e[cei" uJpevrteron. Say whether you have anything better than that [or: speak, if ....] Choe.668: 57) xevnoi, levgoitæ a]n ei[[ [[ ti dei':...... Strangers, say whether anything is lacking [or: speak, if ...] and Choe.755–7: 58) ouj gavr ti fwnei' pai'" e[tæ w]n ejn spargavnoi", eij limov", h] divyh ti", h] liyouriva e[cei: For a child still in swaddling clothes does [not say anything/speak at all], whether it has hunger or thirst or the desire to make water The humour of Choe.755–7 would be greater with a complement interpretation (‘a child in swaddling clothes does not say whether...’), but a real ambiguity exists in all these constructions. At Eum.587, it is obviated only by the textual context: 59) th;n mhtevr j eijpe; prw'ton eijj jj katevktona". Say first whether you killed your mother (or: speak, if ...) Ambiguity exists in these constructions because of the ambiguous transitivity of speech verbs.61 Following cognitive verbs, there is no indeterminacy, as at Eum.142: 60) ijdwvmeq j ei[ ti tou'de froimivou mata'/. Let us see whether any of this prelude is faulty 61The interpretation of these constructions depends on the transitivity of speech verbs and the status of the conditional conjunction. 167 and Choe.890: 61) eijdw'men eijj jj nikw'men, h] nikwvmeqa. Let us know whether we are to be victors or vanquished The position of indirect questions following their main clause gives some support to the view of Houben (1977: 5) that ‘postposition of if-clauses occurs under the influence of completive clauses’. The persistence of conditionals which precede their main clause shows that there is no purely structural rightwards ‘drift’, but the high frequency of following conditionals in Frogs (noted above) suggests that there may be a diachronic effect. This is likely to be cognitive: the tendency for a following conditional to be an afterthought is encouraged by speech patterns in which main clauses tend to precede. It may also be noted that the ‘illocutionary’ conditionals of Wakker (1994) are likely to be afterthoughts, and so to be textually non-prominent. Summary: focus and prosody The start of the clause was mapped in terms of prosodic focalization of the P1 element by P2 enclitics. Cohesive focalization is associated with connectives and emphatics, while interrogative and relative pronouns mark a focal element across an intonation break. The link functions as a form of deixis by which the textual prominence of the P1 element signals the entry of a new ‘participant’ in the discourse: either a referring expression, foregrounding some information, or a grammatical word, emphasizing the relation of the new clause to the preceding text. Evidence that early subordination is not merely a clausal sequence, but an overlap, may be seen in the focalization patterns of free relatives and completive clauses. These are examined in Chapter 5. Part III Clause linking 169 Chapter 5 Subordination: clause order and focalization Summary In Chapter 4, the regular focalization of the clause start was discussed, and extended to the inter-clausal link in relatives and conditionals. In this chapter, the development of complementation is considered by examining the grammaticalization of o{ti from an indefinite relative pronoun to a complementizer. All constructions in extant Homeric and tragic texts are considered, and comparisons are drawn with wJ"-complements. The problem It is generally accepted, following Meillet and Vendryes (1927), Chantraine (1963), and Monteil (1963), that the development of complementation involved a convergence of causal and transitive features in subordinate clauses dependent on verbs of feeling, thought, and speech. Complements are canonically defined formally either as completives (functioning as main verb object), or substantivals (having nominal function).1 They can be described pragmatically as constructions which report propositions, so their ‘objects’ are not referring expressions, as in relatives, but clauses. Yet they are generally analysed as a subdivision of relativization.2 There is certainly a semantic similarity between complements and indefinite relatives (common in the Rigveda texts and in Homeric Greek), and also a morphological one with o{ti-complements, since both are introduced by the PIE pronoun *yo- (Sanskrit ya and Greek o{").3 However, an explanation of the transition from one type to the other must demonstrate how the relative 1Discussed more fully in Chapter 6, Section 1a. 2As by Delbrück (1900: 295–345), Gonda (1954a), Benveniste (1954), Chantraine (1963: 236– 249), Monteil (1963), Touratier (1980), and Lehmann (1984). 3See Delbrück (1900: 311–314, 406–7), Porzig (1923), Schwyzer (1950: 639), Sihler (1995: 400). The Ionic and Doric use of *so/to- as a relative is discussed by Monteil (1963: 5 and 15). For discussions of the Italic use of the indefinite *kwo in correlatives, see Kühner (1914), Haudry (1973), and Lehmann (1989). 170 pronoun developed into a complementizer. Further, wJ"-complements must have a different origin, since they do not derive directly from relatives. A convergence of two syntactic structures must be involved. The proposal Complement structure developed from adverbial and relative forms which depend on main verbs of speech and cognition, through the focalization of a nominal or pronominal element placed at the clausal interface, having a double function (as object of the main verb and marker of the subordinate clause to follow). Complementation therefore developed as a circumstantial construction, with a focalized object and a modifying clause. This sequence explains how completive structure emerged from two distinct sources: a regular focal element is observed in complements with both o{ti and with wJ". ”Oti-complementation The development of o{ti-complementation is mapped in terms of phonological effects at the clausal interface, involving the placing of o{sti" and o{ti as main verb objects, combining an indefinite and a linking function. Four factors are central to the grammaticalization of o{ti from indefinite relative pronoun to complementizer: 1) The position of relatives following the main clause, which creates phonological interference at the clausal interface, causing case indeterminacy. 2) The regularity of free relative constructions following main verbs of knowing, which include an indefinite pronoun functioning as a (focalized) interrogative conjunction (‘to find out who had got away’). 3) Loss of gender marking, which renders case ambiguous, so the complementizer may be interpreted as functioning in both clauses. The construction involves indirect questions depending on main verbs of speech (‘say what you are thinking’), which are common in Homeric constructions. 4) A purely textually-deictic function for the indefinite pronoun in full complements. This requires a cognitive verb in the main, but not the subordinate, clause (‘for I know well that you are all sick’). 171 ÔW"-complementation Complementation with wJ" developed from Homeric adverbial clauses following cognitive and speech verbs (‘know how’ and ‘tell how’), to a transitional stage of intransitive circumstantials, with indirect objects following speech verbs (‘singing about Ares and Aphrodite, how they lay together’). These are the precursors of circumstantial completives with cognitive verbs and direct objects (‘know x, how it is y’), which are discussed in Chapter 6. Chapter Sections 1: Correlatives, relatives, and clause order 2: ”Osti" and o{ti: from relatives to complements 2a: The position of relative clauses (i): Correlatives, indefinite reference, and focalization (ii): Subordinated relatives preceding the main verb 2b: Subordination and case attraction 2c: Free relatives and focalization 2d: Interrogation, indefinite reference, and focalization 2e: ”Oti with free relatives in Homer 2f: From relatives to complements in Homer 2g: Subordinating o{ in Homer 3: Homeric o{ti and main verb type 4: Complementation with wJ" 4a: ÔW" in Homer 4b: ÔW" and circumstantial constructions Summary: subordination in Homer with o{ti and wJ" 1: Correlatives, relatives, and clause order As schematized in the Introduction, Figure 2, a structurally subordinated clause functions as complement to a phrasal head, and (in IE languages) is standardly to its right. A clause to the left of the phrasal projection is usually described as apposed to the structure, with a co-referential link, which may 172 be unmarked morphologically (in parataxis) or marked with grammatical words, in a relation of co-ordination (mevn ... dev) or of correlation (when the linking words have an intra-clausal as well as linking function). Much early correlation is temporal (o[fra ... tovfra, h\mo" ... th'mo", h\mo" ...tovte, e{w" ... tevw", o{te ... tovte), and the ‘if ... then’ construction may also be construed as forming a correlative group.4 The early correlative type is exemplified by the IE generic relative, as RV 1.36.16 (cited by Kiparsky 1995: 156): 1) yó mártyaḥ sísite áty aktúbhir, mánaḥ sá ripúr isvata which mortal makes himself too sharp by night, [may] not us that trickster dominate. The correlative elements may be identified with the PIE pronouns *yo- and *so-/to-, relative and demonstrative respectively, though they function here as determiners rather than pronouns.5 The constituent structure of the early IE sentence appears not to have included structural subordination: Kiparsky (1995: 153) expresses the common view that ‘The key difference [between PIE and Germanic] is that there were no complementizers, and therefore no CP, and no embedding.’ There must have been a position to the left of the basic clause to which emphatic and interrogative elements could be preposed, but these, presumably, simply had scope over the basic clause, rather than serving a subordinating function.6 It follows that correlatives have a semantic but not a structural link. While the semantic function of all relatives is the establishment of co-reference (Touratier 1980: 34), a subordinated relative has the additional property of 4As noted by Matthews (1981: 239). 5Determiners are here described, following Quirk et al. (1985) and Lyons (1977) as the class of words, including articles and demonstratives, which restrict the reference of nouns. For discussions of their functions in NPs, see Lyons (1977: 452–455) and Horrocks and Stavrou (1987: 100–101). 6Perhaps as a Focus Phrase: see Introduction, footnote 27. Further discussion of IE relativization, and bibliography, may be found in the monograph by Kurzová (1981). The situation in Vedic is examined most fully by Hettrich (1989: 467–790), and a cross-linguistic comparison of correlative constructions is undertaken by Downing (1973). 173 supplying a variable which identifies the head.7 Though, as noted in the Introduction, a complement can theoretically be to the left of its head, the presence of correlative elements in almost all main clauses which have a preceding relative clause suggests that Homeric and Classical Greek were right-branching. The development of subordinated constructions therefore requires a change in clausal position, from preceding to following the main clause.8 While correlation does not involve only subordinates which precede the main, even in the IE type (Gonda 1954a: 40–41), the majority do; and, conversely, subordinated relatives always follow a head in the main clause. This may affect intra-clausal word order: while a subordinated relative may follow any main clause nominal, there is a cognitive advantage if the relative follows the whole main clause, so avoiding centre-embedding (the man who mistook his wife for a hat suffered from prosopagnosia).9 The relative clause will be to the right of whole clause, if it modifies either the main verb object in a VO clause, or the subject of a VS construction.10 Head-relative adjacency has phonological effects which may be described as ‘agreement attraction’, discussed below in Section 2b. In Section 2c, this feature is linked to focalization, through the free relative construction, which is interpreted as the precursor of o{ti-complementation. 2: ”Osti" and o{ti{{{ : from relatives to complements 2a: The position of relative clauses The greater ease of parsing a relative clause which follows the main verb is reflected in its typical position in Ancient Greek: Dunn (1988: 69) notes that, in Herodotus 1, relative clauses follow their main verb with a frequency of 7Both explicative (appositive, or non-restrictive) relatives and determinative (restrictive) ones have comparable syntactic structure: see Keenan (1985: 169). 8See Porzig (1923), Benveniste (1954), Chantraine (1963: 236), Haudry (1973), and Lehmann (1974). 9As noted in Chapter 2, Section B5a, these constructions are difficult to process. 10The consequent motivation for VO rather than OV has been noted in Chapter 4, Section 6, and is discussed in Chapter 6, Section 4b, and Chapter 7, Section 2c. Motivation for VS rather than SV is equally plausible. 174 79.3%, and Friedrich (1975: 19) observes that 92% of relatives of a sample from the Iliad and 97% from the Odyssey follow their main verb.11 The textual corpus studied here also shows a preponderance of relative clauses following the main verb, as schematized in Table 1: Preceding Following Following an NP Iliad 9 8 69 (=90%) Odyssey 9 1 34 (=97%) Septem 4 23 (=77%) 3 Agamemnon 14 17 (=55%) Choephoroi 6 14 (=70%) Eumenides 4 25 (=86%) OT. 28 94 (=72%) 8 Medea 16 63 (=75%) 5 Cyclops 4 21 (=66%) 7 Frogs 2 77 (=94%) 3 Melian Dialogue 4 13 (=72%) 1 Crito 11 46 (=81%) Totals 102 496 (=83%) 27 Relatives preferentially follow the main verb in all the texts, with a higher proportion in Homer and Aristophanes than in tragedy or prose. The highest proportion of preceding relatives is in tragedy, most of all in Agamemnon (which in this respect appears more like early IE than does Homer). Relative clauses appear in five positions with respect to the main clause: a) Preceding the main clause, in a correlative relation. b) Following a main clause NP but preceding the main verb. c) Following a main clause NP and the main verb. Some relatives in stichomythia follow a verb in a previous utterance, as at OT.1120: 2) h\ tovnde fravzei"… Tou'ton, o{nper{{{ eijsora'/". (Oed.) Do you speak of this man? (Mess.) This one, whom you see. d) A small number of relative clauses depend on NPs which are either in 11Friedrich’s sampling and definition are somewhat imprecise: he defines clauses as preposed or postposed without defining the terms, and his sample is of 254 unidentified relatives from the Iliad and 167 from the Odyssey. The correct percentages are given here, rather than the errors of the published version. 175 apposition to other phrases, or apostrophes, as Med.496: 3) feu' dexia; ceivr, h|"||| su; povll j ejlambavnou, Oh right hand, which you often grasped ... e) Free relatives following the main verb, without depending on an NP. The semantic and phonological effects associated with these different positions are discussed in Sections 2a (i, ii) and 2b. 2a (i): Correlatives, indirect reference, and focalization A small number of preceding correlatives persist, in prose as well as tragedy, and are presumably motivated pragmatically (as are correlatives in modern languages). Aeschylus uses the correlative construction with indefinite reference, in general statements. This is the same function as in the IE-type correlative (though in tragic constructions the correlative markers are full pronouns rather than adjectives), as at Ag.501–2: 4) o{sti"{{{ tavd j a[llw" th'/d j ejpeuvcetai povlei, aujto;"j ;j ;j ; frenw'n karpoi'to th;n aJmartivan. Whoever prays otherwise for this city may he himself reap the error of his mind. Eum.316–320: 5) o{sti"{{{ d j ajlitw;n w{sper o{d j aJnh;r cei'ra" foniva" ejpikruvptei, mavrture" ojrqai; toi'si qanou'sin paragignovmenai pravktore" ai{mato" aujtwjjj '/ televw" ejfavnhmen. but whoever, wandering like this man, hides his bloody hands standing by as true witnesses for the dead as blood avengers to that one we appear in power. 176 and, with the Homeric relative tov, at Eum.336–9:12 6) qnatw'n toi'sin''' aujtourgivai xumpevswsin mavtaioi, toi'"''' oJmartei'n, o[fr j a]n ga'n uJpevlqh/: qanw;n d j ... of mankind, to whom there happen wanton kin-murders to pursue those, until he goes under the earth ... The other authors, however, use correlation in constructions with specific reference, as at OT.68–9: 7) h}n}}} d j eu\ skopw'n eu{riskon i[asin movnhn, tauvthnvvv e[praxa:..... and which one remedy that I could find by careful searching, that I effected. and, in a non-finite construction, at Thuc.5.98.5–6: 8) o{soi{{{ ga;r nu'n mhdetevroi" xummacou'si, pw'" ouj polemwvsesqe aujtouv"j vj vj v , For who are now allies of neither, how shall you not make enemies of them? At Med.14–15, the main clause is a temporal correlative: 9) h{per{{{ megivsth givgnetai swthriva, o{tan{{{ gunh; pro;" a[ndra mh; dicostath'/. This is the greatest security, when a woman does not disagree with her husband. It may be noted that the P1 element in the relative clause is always focalized, and that this is associated with indefinite reference in the relative clause: specific reference is established only with the main clause. The regular inter- clausal movement from the general to the specific may be marked by modal subordinate verbs, as at Crito 45d6: 10) a{per{{{ a]n ajnh;r ajgaqo;" kai; ajndrei'o" e{loito, tau'ta''' aiJrei'sqai, what a good and brave man would choose, take that ... 12Monteil (1963: 23ff., 82ff.) describes the Homeric and Herodotean uses of relative tov. 177 and Crito 50e5–7: 11) kai; a{tt{{{ j a]n hJmei'" se ejpiceirw'men poiei'n, kai; soi; tau'ta''' ajntipoiei'n oi[ei divkaion ei\nai… and whatever we undertake to do to you, you think it right to retaliate to that? The correlative remains significant as a rhetorical form into the fourth century. Its prosodic effect, of focalizing the relative pronoun, is analogous to the conditional pattern discussed in Chapter 4, Section 5a, and to the structure of preposed complements discussed in Chapter 7, Section 4a. 2a (ii): Subordinated relatives preceding the main verb Most relatives which precede a main verb follow a main clause NP, functioning as its complement (category ‘b’ above, p.174). As discussed in Chapter 4, Section 2b (iv), the head noun is focalized by the relative pronoun. There may sometimes be a main clause correlative element, as OT.449–451: 12) Levgw dev soi: to;n a[ndra tou'ton; [ '; [ '; [ ' [ o}n pavlai zhtei'" ajpeilw'n kajnakhruvsswn fovnon to;n Laiv>eion, ] ou|tov"| v| v| v ejstin ejnqavde, But I say to you: this man [ whom you have been seeking threatening him and loudly proclaiming him as murderer of Laius,] this one is here. The resumptive element here creates emphasis as well as clarity, after the extended relative clause. However, most relatives have no resumptive element (if the verb inflection is discounted). The head noun is usually the main verb subject, as Il.9.60–1: 13) ajll j a[g j ejgwvnj vj vj v , o}" sei'o geraivtero" eu[comai ei\nai, ejxeivpw kai; pavnta diivxomai: But come, I, who declare myself to be older than you, may I speak and go through the whole matter ... Occasionally, the object is the head, as at Il.9.59–61: 14) povll j ajpemuqeovmhn: su; de; sw'/ megalhvtori qumw'/ ei[xa" a[ndra fevriston[ v[ v[ v , o}n ajqavnatoiv per e[tisan, hjtivmhsa", eJlw;n ga;r e[cei" gevra": .... but you, giving way to your proud heart’s anger, dishonoured a great man, whom even the immortals honoured ... 178 The other type of relative which may (though rarely) precede the main verb is the ‘free relative’ construction, where there is no head noun. Kiparsky (1995: 157) cites two early IE constructions (RV 10.27.11 and KUB XI IV 19ff.) where the relative precedes. Both have a loose syntactic relation with the main clause. The construction is regularly associated with aphorisms, as at Ba. 881=901, where the copula is (standardly) omitted:13 15) o{ ti kalo;n fivlon ajeiv. What [is] honourable [is] always precious The same function may be seen in modern languages: ‘qui dort, dîne’.14 In English, these constructions share a morphological feature with correlation, because they are always third person singular, so the verb inflection could be regarded as a correlative marker (‘who dare-si, win-si’): the construction is therefore similar to the IE-type generic correlative (as RV 1.36.16 in Section 1c, citation 1, above). There are only four instances of preceding free relatives in the corpus, all with an initial focal element.15 In two constructions, focalization does not involve the relative pronoun, and the reference is consequently more specific. At Od.9.94–5, the relative is in P2, preceded by a focalized anaphoric pronoun: 16) tw'n''' d j o{" ti" lwtoi'o favgoi melihdeva karpovn, oujkevt j ajpaggei'lai pavlin h[qelen oujde; nevesqai, But of them, whoever ate the honey-sweet fruit of lotus, was no longer willing to take any message back or go away... At Septem 1046, the adversative element rather than the relative pronoun is prosodically focalized: 17) ajll jj jj jj j o}n povli" stugei', su; timhvsei" tavfw/… But whom the city hates, you will honour with a tomb? The third construction, at Med.453–4, seems to be purely appositional: 18) a} d j ej" turavnnou" ejstiv soi lelegmevna, pa'n kevrdo" hJgou' zhmioumevnh fugh'/. But (as for) what you said against the royal family, consider yourself lucky you have been punished with exile. 13For the omission of the copula in tragedy, see Havelock (1978). 14Cyrano de Bergerac, IV.1. 15Constructions elsewhere in tragedy include E.Hipp.191 and 193, Hel.822, Ba.515, IA.1014. 179 At OT.486, however, indefinite reference is marked by o{ ti in a construction involving speech and cognitive verbs: 19) .... o{ ti levxw d j ajporw' but what I can say I do not know This last construction shows the close connection between indefinite reference and interrogation (discussed below, in Section 2d), which is clearer when the relative follows the main clause (‘I am at a loss what to say’). The fact that almost all free relatives follow the main verb is particularly important to the development of complementation, because of the phonological and structural features of the clausal interface. These may be described in terms of case attraction. 2b: Subordination and case attraction The different orderings of relative and main clauses are aligned with different meanings: the IE-type correlative is indefinite because it precedes the head noun supplying the specific reference, and even those preceding correlatives with specific reference cited above in Section 2a (i) do not have their reference established until the appearance of the head in the main clause. Subordination of relatives requires the correlative elements to be adjacent, so creating a tension between referential and relational marking (because case- marking is a syntactic marker, and so varies according to intra-clausal function, while markers for number and gender remain inter-clausally constant between co-referent elements).16 Proximity, however, encourages phonological attraction of case marking, which occurs regularly in Homeric and Classical Greek. There are four possible permutations: either the head noun or the relative pronoun may be attracted to the case of the other, or the head may be omitted and the relative stand either in the case proper to the head, or to its own case (see Gonda 1954a, Chantraine 1963: 237–9). Though Gonda (1954a: 29) cites Greek constructions involving case attraction as demonstrating that the dependency of the o{"-clause on the main is symptomatic of pragmatic prominence (rather 16Failure of number agreement occurs in relatives at Il.11.367, 16.368–9, and Od.23.121, and of gender agreement in a coordinated construction at Il.11.237–9. 180 than being syntactically fixed), it will be shown that the patterns have a direct relationship with clause order, and prosodic prominence too. The types may be seen as forming a historical sequence, because the first requires the relative to precede the main verb, while the fourth, the free (headless) relative, almost always follows, and is semantically similar to complementation with o{ti. The other three are first discussed briefly, to demonstrate the effect of clause order on focalization: 1) When the subordinate precedes, the antecedent typically takes the case of the relative, in ‘inverse attraction’. This is the most common type in the Rigveda,17 in Homer (Gonda 1954a: 30, Chantraine 1963: 237), and in Old Latin (Vonlaufen 1974: 29), which suggests it may be an early form. It gives prominence to the relative clause, since antecedent and relative are both case- marked with respect to their function in the subordinate clause, as at Il.14.75– 6: 20) nh'e"''' [ o{sai{{{ prw'tai eijruvatai a[gci qalavssh" ] e{lkwmen... Let us drag those ships which are beached in the first line near the sea... and Il.14. 371–2: 21) ajspivde"j vj vj v [ o{ssai{{{ a[ristai ejni; stratw'/ hjde; mevgistai ] eJssavmenoi ... Putting on the shields which are best in all the army and biggest Most instances, like those above, involve the attraction of the case of the main verb object to that of the relative. Adjacent subjects may also be attracted, as at Il.10.416–7: 22) fulaka;";;; d j [a}"}}} ei[reai, h{rw", ] ou[ ti" kekrimevnh rJuvetai strato;n oujde; fulavssei Those guards which you ask of, hero, there is no detail which protects the army and guards it This last instance could be apposed, since there is a resumptive element (ti") in the main clause. Havers (1926) analyses the construction as substituting for an emphatic nominative, while Gonda (1954a: 32n115) interprets it as thematic. 17RV 6.74.2; 3.37.11; 5.30.15; and others cited by Gonda (1954a: 33). 181 In this type, the antecedent is isolated from the rest of the main clause, so its function as an argument of the main verb is less noticeable. An antecedent may even be positioned inside the relative clause, as at Il.18.429–30, where qeavwn might be expected, so o{sai appears to be a determiner in the NP o{sai qeaiv, and the construction is similar to the IE-type correlative:18 23) ... h\ a[ra dhv ti", o{sai{{{ qeaivv vv eijs j ejn jOluvmpw/, tossavd j ejni; fresi;n h|/sin ajnevceto khvdea lugra; Is there any of the goddesses who are on Olympus, who in her heart has endured such grim sorrows? Inverse attraction creates phonological emphasis, because the case agreement constitutes homoioptoton,19 and its echoic pattern may be considered a form of focalization, which may also draw attention to the meaning of the words.20 2) In the ‘Attic’ type, the relative pronoun is attracted to the case of the head NP, as OC 334: 24) xu;n [ w|/per|/ |/|/ ei\con ] oijketw'n pistw'/ movnw/'/ v /'/ v /'/ v / (I came) with the one true servant that I had. and X. An.1.7,3: 25) o{pw" e[sesqe a[ndre" a[xioi th'" ejleuqeriva"' j v' j v' j v [ h|"||| kevkthsqe ] so you will be men worthy of the freedom which you possess 21 Here, there is a similar phonological effect to type 1, as Havers (1931: 72) observes. Although antecedents do not always precede immediately (Gonda 1954a: 29), the relative clause always follows the main verb. The syntactic effect is to integrate the relative clause in the structure of the main, with a clearer adjectival function. Such constructions do not exist in Homer, but non- 18Chantraine (1963: 238) and Gonda (1954a: 21) differ on whether this should be termed inverse attraction, because Gonda thinks it represents a single prosodic constituent unified by the pronoun. Sihler (1995: 396) notes that the placing of the antecedent in the relative clause is also typical of Hittite and Tocharian. 19See Rhetorica ad Herennium IV.xx.28 and Quintilian, Institutio IX.iii.77–80. 20Cf. Pope, An Essay on Criticism 365: ‘The Sound must seem an Eccho to the Sense.’ See also Jakobson (1987: 86ff.) on paronomasia and ‘the internal nexus between sound and meaning’. 21These constructions are cited by Kühner (1904: 407). 182 finite analogues do, and Chantraine (1963: 237) and Gonda (1954a: 25) categorise them as attraction, as at Il.1.262–3:22 26) ouj gavr pw toivou"vvv i[don ajnevra"j vj vj v oujde; i[dwmai oi|on||| Peirivqoon te, Druvantav te, poimevna law'n For I never yet have seen nor shall see again such men as Peirithoos, and Dryas, shepherd of the people A different effect is created if the ‘antecedent’ noun is positioned within the relative clause, as the noun at Xen.An.1.9.14: 27) touvtou" kai; a[rconta" ejpoivei [ h|"||| [ katestrevfeto ] cwvravvv " ] He made them rulers of the land he subdued The great inter-clausal integration of this construction is described by Gonda (1954a: 29) as ‘Verschmelzung’ (melting). It should not really be considered as attraction, because it does not involve adjacency, and its effect depends rather upon constituent recognition than on phonetic similarity. It is very similar to the intra-clausal tmesis discussed in Chapter 3, and Gonda (1954a: 29) categorizes a tmetic construction at S.El.762–3 as Attic attraction, though it is not subordinated to a verb, but is in apposition to an adjective (and here homoioptoton is reinforced by very strong assonance throughout): 28) ajlgeinav, toi'" d j ijdou'sin, oi{per ei[domen, mevgista pavntwnvvv [ w|n||| [ o[pwp j ejgw; ] kakw'n.''' ] ... piteous, but for those seeing it, as I saw it, the greatest of all evils I have seen. Attraction with the antecedent positioned in the subordinate clause may reflect the earlier function of the relative as a determiner rather than a full pronoun (Section 1b above), though the dependent clause is here in modifier position within the NP. 3) The relative clause normally follows the main also in constructions where the relative has its case assigned with respect to the main clause, but the head noun is omitted, as at S.El.1048. This is similar to type 2 in functioning in the main clause, but differs in that the relative functions as a full pronoun, so is properly ellipsis rather than attraction: 29) fronei'n e[oika" oujde;n w|n||| ejgw; levgw You seem to hear nothing of what I say. 23 22Similar constructions occur at Od.9.322, 9.325, 10.113, 10.167, 11.25, and 19.233. 183 The construction highlights the governing relations in the main clause, and (in this construction) the assonance in the subordinate. The similarity between types 2 and 3 is that the main clause case is realized because it precedes, so cohesion in the main clause varies according to clause order with respect to the main verb. Although a true relative modifies a nominal head, so having no direct connection with the main verb, phonological factors encourage a movement from correlative to modifying elements when the subordinate clause follows the main verb. A different effect is created by a fourth type of attraction, where the relative pronoun retains its proper case in the subordinate clause, despite the absence of a head noun, so the subordinate clause is more prominent. As with type 3, this is a form of ellipsis, so draws attention to what is not there, and therefore to the meaning of the matching clause, rather than its phonology. This is discussed in the next section. 2c: Free relatives and focalization The ‘free relative’, where the relative stands in its proper case (with respect to the subordinate clause), without a head word in the main clause, appears to be central to the development of subordination, for a number of reasons: i) The structure occurs regularly, in contrast with the other types of attraction. ii) The subordinate clause almost always follows the main clause. iii) The syntax retains the syntactic integrity of the subordinate, rather than the main, clause, yet the subordinate follows main clauses without correlative elements, so the relative has a semantic function in the main clause. iv) The construction has a regular association with the same types of verbs on which complement clauses depend. v) The pronouns are indefinite. This is interpreted as a transitional stage between exophorically-referring relatives and textually-deictic complement conjunctions. vi) Indefinite affixes function as cohesive focalizers, making the pronoun more prominent. 23See also S.Phil.1227, Demosthenes 37.2, and other constructions cited by Kühner (1904: 408). 184 In tragedy, the pronoun almost always stands for the main clause object, though an instance of subject function occurs at Septem 452 (and it may be noted that this encourages VS order in the main clause): 30) o[loiq j o}"}}} povlei megavl j ejpeuvcetai May he perish, he who boasts greatly against the city. The characteristic feature of a free relative functioning as object is dependence on a cognitive verb (of perception or judgment). The construction may therefore be analysed as an ‘indirect question’.24 The pronouns always have generic reference, marked by epic te, -ti", or ke, as at Il.2.365–6: 31) gnwvsh/ e[peiq j o{" q j{ j{ j{ j hJgemovnwn kakov", o{" tev{ v{ v{ v nu law'n hjd j o{" k{{{ j ejsqlo;" e[h/si ... Then you will see which of your leaders is bad, and which of your people, and which also are brave... Il.21.609–10: 32) mei'nai e[t j ajllhvlou", kai; gnwvmenai, o{" te{{{ pefeuvgoi o{" t{{{ j e[qan j ejn polevmw/ ... ... to wait for each other and find out which one had got away and who had died in the battle... and Od.3.184–5: 33) oujdev ti oi\da keivnwn, oi{ t j{ j{ j{ j ejsavwqen jAcaiw'n oi{ t{{{ j ajpovlonto. .... and I knew nothing of those Achaians, which had survived, which had perished. There are also 12 similar indirect questions with o{sti" following cognitive verbs,25 and only one following a speech verb, at Od.10.109–10: 34) oiJ de; paristavmenoi prosefwvneon, e[k t j ejrevonto o{" ti"{{{ tw'nd j ei[h basileu;" kai; oi|sin ajnavssoi. My men stood by her and talked with her, and asked her who was king of these people and who was lord over them. 24This may be defined as indirect speech or thought with an interrogative or relative pronoun, but no head noun. 25At Il.3.167, 3.192, 11.219, 14.509, 16.424, 20.363, Od.4.380=423=469, 4.552, 8.28, 9.331–2. There are therefore 16 construction in Homer, as against 51 constructions with other verbs (=31%). 185 The reason for the rarity of dependence on a speech verb in Homer is not only that indirect speech is uncommon in epic, because, as shown in Chapter 7, it is uncommon in tragedy too: as is argued in Chapter 6, speech verbs could not take animate direct objects, but cognitive verbs could (‘know him’).26 Though the pronouns are masculine, their function with respect to the main verb is like that of an object. However, it will be shown below, in Section 2f, that constructions with speech verbs and neuter pronouns are quite common in Homer, and their use represents a further stage in the development of the relative pronoun from a referring expression to a textual deictic (‘say what...’ to ‘say that...’). 2d: Interrogation, indefinite reference, and focalization A semantic connection between indefinite and interrogative pronouns is noted by Dover (1960: 12), Monteil (1963: 150, 154), and by Lyons (1977: 761– 2), who points out that an interrogative with tiv" presupposes the truth of an indefinite statement with ti".27 The semantic link is reflected lexically: Sihler (1995: 396–7) observes that the PIE stem *kw- has indefinite and interrogative function in every IE branch, while its relative function was usually lost, and speculates that the link between relative and interrogative might be that *kw- was a focus marker.28 The functional parallel drawn here between ti" and ‘epic’ te is supported by the analysis of Ruijgh (1971: 9), who notes that epic te is regularly associated with digressive, non-restrictive, relatives (after o{", oJ, o{qi, o{qen, o{te, oi|o", wJ") and co-ordinated clauses (after dev, ajtavr, gavr, mevn, kaiv, ajllav). It is central to his definition of the function of epic te as ‘digressif-permanent’ that it has a linking function and that it marks non-specific reference.29 26Jussives are analysed differently, with a semantically indirect object. 27See also Monteil (1963: 65), Ruijgh (1971: 310), and Biraud (1985: 162). Indefinite reference appears to involve the loss of a causal connection between clauses: as Kiparsky and Kiparsky (1970: 167) note, there is a connection between truth (and factivity) and specific reference. 28A view anticipated by Delbrück (1897: 511ff.), who considered that, following a pronoun, *kwe always had an emphatic function, linked to indefiniteness, which was prior to a connective function (see also Bernert 1940: 78). 29The view of Sihler (1995: 401) that -te is added to relative pronouns ‘without any apparent change in meaning’ is not, therefore, followed here. On the meaning of epic te, see also Bäumlein (1861: 227–235), and Gonda (1954b), who analyses it as digressive. 186 There may be an etymological connection too: though Gonda (1954b: 181) considered that the similarity of *kwe and the indefinite interrogative adjective *kwo-/*kwe- is mere ‘phonetic coincidence’, ti" and te may be, as Sihler (1995: 160) suggests, ‘perhaps ultimately related’ etymologically.30 The difference between definite and indefinite reference parallels that between anaphora and textual deixis: as noted by Ehlich (1982), when pronouns are used phorically, their function is to sustain the listener’s focus of attention, whereas deictic use alters the focus. The phonological effect of both indefinite reference markers therefore accords with the description of the P1>P2 sequence described in Chapter 4: the particle is prosodically in P2, focalizing the relative pronoun.31 The structure accompanying epic te is shown in Fig. 1 (=Introduction, Fig. 6):32 VP Figure 1 gnwvmenai o{" te pefeuvgoi CP IP P1 Focus NP Object C' The object NP is included in the diagram to illustrate the function of o{" as object of the main verb. The construction may be considered a circumstantial version of type 4 case attraction. 30Te appears to be derived from the stem *kwo-/*kwe-: see Meillet (1898), Kühner (1904: 236, 241), Schwyzer (1950: 573ff.), Gonda (1954b), Monteil (1963: 109–111), and Ruijgh (1971). 31For a discussion of the focal effect of ke and a[n in indefinite clauses, see Howorth (1955). 32The citation is from Il.21.609: ‘To find out who had got away’. 187 The loss of gender marking creates the possibility of a completive interpretation. Focal o{ te occurs in similar constructions following cognitive verbs at Il.5.331: 35) gignwvskwn o{ t j a[nalki" e[hn qeov"... knowing her, that she was a god without warcraft and Il.8.251: 36) oi} d j wJ" ou\n ei[donq j o{ t j a[p j ejk Dio;" h[luqen o[rni" but when they saw the bird, that it had come from Zeus Similar constructions occur at Il.17. 623, 17.626–7 and Il.1.411–2 (=16.273–4). All are indistinguishable from complements, though Monteil (1963: 263) notes they could be also interpreted either as relatives or temporals, and that only Il.5.331 (cited above) is clearly completive, because relative o{ is excluded by the gender (qeov" refers to Aphrodite). The use of neuter o{ti in relative constructions is discussed in the next section. A similar pattern occurs with ti", as in Fig. 2 (=Introduction, Fig. 7):33 VP Figure 2 o{" CP IP P1 Focus e[k t j ejrevonto ti" tw'nd j ei[h basileuv" C ' The structural difference between this and the construction with epic te is that ti", being pronominal, is less likely to be in head position in the CP. 33The citation is from Od.10.109–10: ‘And they asked who was king of these people...’ 188 However, the etymological and functional link between the two words, noted above, suggests that a semantic parallel may be drawn. The connection between these functions and the forms -ti" and te signifies, then, not only a semantic link with interrogation, but also the structural effect of cohesive focalization, which co-occurs with a contrastive feature: the intonation break between the relative pronoun and the main verb. 2e: ”Oti with free relatives in Homer The grammaticalization process is encouraged by the loss of gender marking (which involves a further loss of referential specificity) with non-animate objects.34 Of the 128 Homeric constructions in which o{ti appears, 39 are relatives or free relatives which have generic reference (see below in Section 2i).35 A few of these constructions depend on verbs of giving or taking, as at Il.15.109 (where the relative is adjectival): 37) tw; e[ceqæ o{tti{{{ v ken u[mmi kako;n;;; pevmph/sin eJkavstw/. you must each take whatever evil he sends you. and Od.18.112–3: 38) Zeuv" toi doivh, xei'ne, kai; ajqavnatoi qeoi; a[lloi, o{tti{{{ mavlistæ ejqevlei" kaiv toi fivlon e[pleto qumw'/, may Zeus and the other gods, stranger, give you whatever you want most and is dearest to your spirit. However, most of the constructions depend on verbs of speech, as at Il.1.85: 39) qarshvsa" mavla eijpe; qeoprovpion o{ ti oi\sqa Speak very boldly the prophecy that you know Without a correlative, constructions with speech verbs constitute indirect questions, as Il.14.195=Od.5.89: 40) au[da o{ ti fronevei" .... Say what you are thinking 34”Oti (and its variants o{tti and o{ ti) is understood as the accusative singular neuter of o{sti", corresponding to the PIE *yot kwid. See Monteil (1963: 247) and Sihler (1995: 400). 35The function of mood in marking indefinite reference may also be noted. For references, see Chapter 7, Section 3a, n. 49. 189 The absence of gender marking, and consequently of case marking, facilitates the interpretation of the relative as functioning in either the main or the subordinate clause. In Fig. 3 (=Introduction, Fig. 8) the pronoun can be interpreted as accusative in both: VP Figure 3 o{ ti au[da fronevei" Obj. CP Focus P1 IP NP C ' Obj . The high frequency of speech verbs with relative o{tti extends even to the verb in subordinate clauses which depend on main verbs of other classes. There is an evident metrical component (|- v v|- -||) in the frequency of o{ttiv ken ei[pw and its variants at the line end, as at Il.1.294=14.190=Od.1.158=1.389: 41) eij dh; soi; pa'n e[rgon uJpeivxomai o{ttiv ken ei[ph/": if I must carry out every order which you might say It is significant that either the main or subordinate clause may contain a speech verb, and sometimes both do, as at Od.8.548–550: 42) tw' nu'n mhde; su; keu'qe nohvmasi kerdalevoisin, o{tti kev sæ ei[rwmai: favsqai dev se kavlliovn ejstin. ei[p[[[ j o[nom j o{tti se kei'qi kavleonvvv mhvthr te pathvr te... do not keep hiding now with crafty purposes what I ask you. It is better to speak . Tell me the name which your mother and father called you there... 190 This pattern shows that the linking element is properly the object of both verbs. The subordinate at Od.8.550 above is the only definite Homeric relative with o{tti (it may be influenced by the use of the same pronoun with indefinite reference in the previous line, at 549). Explicit main verb objects are common with relative o{tti, occurring in 14 out of the 32 constructions. However, reference appears no more specific: there seems little difference between its reference at Od.19.378: 43) ............ ajll j a[ge nu'n xunivei e[po"[[[ , o{tti ken ei[pw: But come, listen to whatever word I speak. and at Od.2.25=2.161=2.229=24.454: 44) kevklute dh; nu'n meu, jIqakhvsioi, o{tti ken ei[pw. Hear now, Ithacans, whatever I tell you. Failure of number agreement occurs occasionally, always accompanied by a subjunctive verb, and involving neuter relatives. It is not a sign of a complement, because it occurs in relatives at Od.18.142 and at Il.10.207–208: 45) h[ tinav pou kai fh'min''' ejni; Trwvessi puvqoito a{ssa{{{ te mhtiovwsi meta; sfivsin, h] memavasin or he might learn some report of the Trojans, what they deliberate among themselves, ... 2f: From relatives to complements in Homer Free relatives are associated with verbs of speech and thought, which frequently both appear in the same construction. The semantic significance of speech verbs is that their reference can combine exophoric and textual features: in ‘hear what I say’, the referent is also the linguistic object itself, even when there is a correlative pronoun, as at Il.1.294: 46) eij dh; soi; pa'n e[rgon' [' [' [ uJpeivxomai o{ttiv ken ei[ph/": if I must accept every point which you might say If the pronoun expresses the subordinate subject, the subordinate is semantically closer to a complement, because the pronoun then has scope over the whole clause, as at Od.9.402: 47) iJstavmenoi dæ ei[ronto peri; spevo", o{tti eJ khvdoi: and standing about the cave they asked what troubled him. 191 The similarity with complementation stems from a structural ambiguity: a transitive complement clause does not have a referential gap as a relative does (the place thati I know -i), but a relative subject would, if it existed, have a position indistinguishable from that of the relative pronoun, so it is not clear that any element has moved (the man thati -i broke the bank at Monte Carlo), 36 and the subordinate clause appears formally more independent. The key step to a completive is in the use of the neuter pronoun with a cognitive main, but not subordinate, verb (because then o{ti cannot be interpreted as an argument). The structure of o{ti-complementation is shown in Fig. 4 (=Introduction, Fig. 9). The CP is not now a site for elements preposed from the subordinate clause, so they now have only one q-role, as main verb object, and the structure is syntactically regular:37 VP Figure 4 o{- CP IP P1 Focus ti eu\ ga;r oi\d j nosei'te pavnte" NP Object C' 36The point is that the co-referent positions are indistinguishable: Chomsky (1986: 48–54) defines this as ‘vacuous movement’. 37The citation is from OT.59–60: ‘ For I know well that you are all sick...’ 192 2g: Subordinating o{{ {{ in Homer Occasional Homeric relatives involving the definite pronominal o{ also demonstrate inter-clausal focalization.38 Some have a specifying function, and these have an intonation break before the pronominal, so the prosodic pattern is similar to the following conditionals discussed in Chapter 4. The absence of an indefinite marker in these constructions results in a weak inter-clausal bond: the semantic relation of the subordinate to the main clause is appositional, as at Od.1.382 (=18.411=20.269): 48) Thlevmacon qauvmazon, o}} }} qarsalevw" ajgovreue. they wondered at Telemachos, at the daring way he had spoken and Od.21.289–90: 49) oujk ajgapa'/", o}} }} e{khlo" uJperfiavloisi meq j hJmi'n / daivnusai Is it not enough that you dine in peace among us arrogant people However, after cognitive verbs, the conjunctive meaning is stronger, because the pronoun is focalized by P2 pronouns. It may be interpreted as having an adverbial force (‘how’ rather than ‘that’), as at Od.17.545: 50) oujc oJrava/", o{{ {{ moi uiJo;" ejpevptare pa'sin e[pessi… do you not see how my son sneezed for everything I have spoken? and Il.15.248: 51) oujk aji?ei" o{ {{{ me nhusi;n e[pi prumnh'/sin jAcaiw'n ou}" eJtavrou" ojlevkonta boh;n ajgaqo;" bavlen Ai[a" did you not know how, by the Achaians’ grounded ships, Aias of the great war cry struck me as I killed his companions... A number have pronominal main verb direct or indirect objects (never adjacent to the relative pronoun), as at Il.8. 362: 52) oujdev ti tw'n''' mevmnhtai, o{{ {{ oiJ mavla pollavki" uiJo;n teirovmenon swveskon uJpæ Eujrusqh'o" ajevqlwn. he does not remember that I often protected his son when the tasks of Eurystheus were too much for his strength 38Monteil (1963: 77–78) cites 24 instances in Homer. Delbrück (1900: 311–5) derives o{ from the correlative pronoun *so-/to-, while Chantraine (1963: 166) identifies it with both the article and the demonstrative pronoun. 193 Il.9.493: 53) ta; fronevwn o{ {{{ moi ou[ ti qeoi; govnon ejxetevleion thinking that/how the gods would not bring to birth any children ... and Il.20.466=Od.3.146: 54) nhvpio", oujde; to;; ;; h[/dh o}} }} ouj peivsesqai e[mellen: ... and did not see that there would be no way to persuade him These are very similar to constructions with o{ti. The presence of the ethic datives is particularly revealing: whatever their pragmatic function, they have the prosodic effect of identifying o{ as in P1, and so of focalizing it. 3: Homeric o{ti{{{ and main verb type As noted for relative constructions with case attraction, the local phonological effects at the inter-clausal link are also dependent on clause order (main first) and syntactic relations: principally the transitivity of the main verb. It is argued here that the ‘indirect question’ with verbs of speech is central to the development of complement structure. The types of verbs involved in Homeric subordinate clauses introduced by o{ti are collated below. There are 128 constructions with o{ti,39 including 41 so spelt in the Oxford texts, together with 72 instances of the variant o{tti,40 and 15 of o{ ti.41 The constructions are schematized according to the type of main 39The constructions are cited in Appendix 3A. 40Sihler (1995: 400) derives Homeric o{tti from the same PIE stem (*yot kwid) as o{ti. 41Monteil (1963: 254) notes that the graphological difference between o{ ti and o{ti ‘ne repose en fait sur aucune tradition ancienne.’ 194 verb they follow, in Table 2:42 Main Verb o{ti{{{ o{tti{{{ o{ ti{{{ TOTALS Emotion 13 11 0 24 Cognitive (perception or judgment) 14 11 (6 relative, including 5 with sub. speech vbs.) 1 (relative: Il.10.503) 26 Speech 2 (Od.16.131, & Il.17.641–2, dep. on speech noun) 12 (6 relative, including 4 with sub. speech vbs.) 6 (4 relative, all with sub. cog. vbs.) 20 Other verbs, (relative o{ti) 20 (including 8 with sub. speech vbs.) 8 (inc. 2 with sub. cog. vbs.) 28 Other verbs, (o{ti causal ‘because’) 12 5 17 Other verbs, (adverbial o{ti) 10 (o{tti tavcista) 10 Dependent on an adjective 3 3 TOTALS 41 72 15 128 Three principal features may be noted: i) 53 out of 128 constructions (=41%) are completive, 24 following verbs of feeling, 19 following cognitive verbs, and 10 following speech verbs (or the noun ajggelivh"). ii) 39 out of 128 constructions (=30%) are relatives or free relatives. iii) If relatives are included, 20 constructions depend on speech verbs, and in 17 of the relative constructions, the relative clause itself contains a speech verb. There is therefore a stronger connection between relatives and speech verbs than of any other type. It has been proposed that subordination with verbs of emotion is the precursor of true complements, either because these verbs take causal completives (Chantraine 1963), or because the subordinate clause is a specifying substantival, in an appositive relation (Monteil 1963). Chantraine 42As the constructions include completives and relatives, and are listed by verb type and variant pronouns, the data-set and analysis differ from Monteil (1963: 399ff.), where 36 constructions are categorized as substantival, and between 19 and 31 as causal conjunctions. 195 (1963: 288–299) believes that declaratives following verbs of thought, speech, or perception are derived from causal completives following verbs of feeling, as Od.19.247–8: 55) .............tiven dev min e[xocon a[llwn w|n eJtavrwn jOduseuv", o{ti{{{ oiJ fresi;n a[rtia h[/dh. Odysseus prized him above his other companions, in that their thoughts were in harmony. Monteil (1963: 248) agrees, though categorizing the structure as substantival. He interprets a construction at Il.14.406–7=22.291–2 in the same way: 56) .............. cwvsato d j ”Ektwr o{tti{{{ v rJav oiJ bevlo" wjku; ejtwvsion e[kfuge ceirov" and Hektor was angered that his swift weapon had been loosed from his hand in vain. This would also be a causal completive in Chantraine’s terms, and Monteil considers that ‘il est hors de doute que la subordinée tout entière sert de regime au verb principal: “Hector s’irrita du fait que son trait avait vainement quitté sa main.”‘43 The key factor here is the main verb, because substantivals following other types of verb are not complements, as the causal construction at Il.9.75–7: 57) ...... mavla de; crew; pavnta" jAcaiou;" ejsqlh'" kai; pukinh'", o{ti{{{ dhvi>oi ejgguvqi nhw'n kaivousin pura; pollav: ...... for there is great need for all the Achaians of good close [counsel], in that close to the ships the enemy burn many fires. However, the substantival function appears less central to the development of complementation than does the completive: as the figures in Table 1 show, all but 20 of the non-relative subordinating constructions in Homer are completives following emotional, cognitive, or speech verbs. The view that the earliest completives are those dependent on verbs of emotion is not borne out by the Homeric o{ti-constructions: although there are rather more completives which follow verbs of emotion than of any other single type, there is a strong association of o{ti with speech verbs, even in 43Monteil (1963: 249) derives his interpretation from the ‘Substantivsätzen’ of Kühner (1904: 354–377) and Schwyzer (1950: 645). Monteil (1963: 257) identifies the substantival function of o{ti with that of an adjectival article, following Benveniste (1954: 188–192). 196 Homer, as shown at (iii) above. The common presence of speech verbs in the subordinate clause, as well as the main, shows that the transitivity of both main and subordinate verbs is crucial: relative o{ti usually functions as object in both clauses. The observation of Monteil (1963, 249–250) that ‘La prééminence des verbes “de pensée” sur ceux “de déclaration” continue à se manifester après Homère et jusqu’à la fin du ve siècle’ may be valid for complements, but not for all indirect questions. There is, on the other hand, no evidence that causal completives following verbs of emotion are earlier than the others: extant constructions are contemporary with, not earlier than, completives following cognitive and speech verbs. The consequence is that the pronominal meaning of o{ti may well precede its causal meaning (‘because’),44 and the causal sense could derive from causal constructions with reporting verbs. The meaning is ambiguous in constructions such as Il.10.503: 58) Aujta;r o} mermhvrize mevnwn o{ ti{{{ kuvnteron e{rdoi But he waited, divided as to what more daring he might do and Il.14.220–1: 59) .............. oujdev sev fhmi a[prhktovn[ v[ v[ v ge nevesqai, o{ ti{{{ fresi;n h\/si menoina/'. Nor do I say that you are going unsuccessfully in whatever you desire. In the analysis of Monteil (1963) these constructions would be substantivals, but it is significant that both involve a verb of thought or speech, even if that is not the governing verb. Conversely, indirect questions dependent on speech verbs quite often contain verbs of emotion in the subordinate clause, as at Od.8.577: 60) eijpe; d j o{ ti klaivei" ..... Tell me why you weep/what you bemoan and Il.1.64: 61) o{" k j ei[ph/ o{ ti tovsson ejcwvsato Foi'bo" jApovllwn Who can tell why Phoibos Apollo is so angry/what Phoibos Apollo is angered about 44This view is also held by Cristofaro (1998: 72). The high level of causal o{ti in Euripides supports it: see Chap. 7, Section 2b (iii). It may be noted that English ‘because’ is also substantival in origin, originally focalized: ‘by cause that’ (OED Vol. 1: 746). 197 It appears, then, futile to attempt to establish primacy among the verb types: verbs of thought, speech, and emotion are approximately equally involved. However, their regular appearance in both the subordinate and the main clauses shows that it is the transitivity of the verbs rather than a substantival clause function or a causal interpretation of the pronoun which is central. The importance of free relatives in the development of o{ti-complementation depends on the pronominal rather than the causal sense of o{ti, though, as Monteil (1963: 251) notes, both are involved with verbs of emotion, which are ‘verbes exprimant un procès dont l’objet est de toute façon la cause’ (and as noted in fn. 44, ‘because’ may derive from a focalized substantial). 4: Complementation with wJ"JJJ 4a: ÔW" in Homer The principal syntactic difference of wJ"-complements from those with o{ti is a closer connection with verbs of thought, partly because wJ" does not introduce relative clauses. However, wJ" appears after speech verbs too. There are over 2,000 instances of wJ" in Homer, most either with adverbial meaning, or as a textually deictic link, especially after a speech. Completive, causal, and interrogative uses of wJ" analogous to those of o{ti may be identified.45 Completive constructions are more common than are substantival: Monteil (1963: 355) considers there to be only one sure substantival construction, at Il.17.450=Od.2.312: 62) h\ oujc a{li" wJ" kai; teuvce j e[cei kai; ejpeuvcetai au[tw"… [Is] it not enough that he has the armour and so exults in it? As this sort of construction is so rare in Homer, it may, in fact, not be early. A few constructions appear with o{ti and the same predicator (a{li"), as Il.23.670: 63) h\ oujc a{li" o{tti mavch" ejpideuvomai… [Is] it not enough that I am lacking in battle skills? Similar appositive constructions, where the subordinate clause expresses the subject (formally, the complement of an unexpressed copula), become more 45As Riemann and Goelzer (1897: 497–499) and Monteil (1963: 351–364). 198 common in Aristophanes, Plato and Aristotle, as in the use of dh'lon noted in Chapter 7, Section 3a. A minimal main clause frame is more characteristic of late than emerging complementation. Completives, though more frequent, are much less common in Homer than those with o{ti. Monteil (1963: 354) cites 16 instances of wJ"-completives following cognitive or speech verbs (7 and 9 respectively), and 5 of completives following verbs of emotion (in view of their rarity, his argument that these last are forerunners of the substantival category is therefore even weaker than for constructions with o{ti). The low frequency of Homeric completive wJ"-constructions is somewhat puzzling, as they became so common in tragedy (as discussed in Chapter 7). 4b: ÔW" and circumstantial constructions A clue to the origin of wJ"-complements may be seen in the difference of position according as the conjunction follows a cognitive or speech verb. ÔW" normally follows directly after cognitive verbs, as Il.4.360: 64) oi\da ga;r w{"{{{ toi qumo;" ejni; sthvqessi fivloisin h[pia dhvnea oi\de: ta; ga;r fronevei" a{ t j ejgwv per. for I know how the spirit in your secret heart knows ideas of kindness only; for what you think is what I think. Il.10.160: 65) oujk aji?ei" wJ"JJJ Trw'e" ejpi; qrwsmw'/ pedivoio ei{atai a[gci new'n, ojlivgo" dæ e[ti cw'ro" ejruvkei… do you not know how the Trojans at the break of the land are sitting close to our ships, and narrow ground holds them from us? and Il.15.204: 66) oi\sqæ wJ"JJJ presbutevroisin jErinuve" aije;n e{pontai. you know how the Furies forever side with the elder However, after speech verbs, wJ" is typically line-initial, and generally does not follow the verb immediately, as Od.4.376: 67) ejk mevn toi ejrevw, h{ ti" suv pevr ejssi qeavwn, wJ"JJJ ejgw; ou[ ti eJkw;n kateruvkomai, .... so I will tell, whoever you may be of the goddesses, how I am not detained of my free will 199 Od.8.75: 68) nei'ko" jOdussh'o" kai; Phlei?dew jAcilh'o", w{"{{{ pote dhrivsanto qew'n ejn daiti; qaleivh/ ejkpavglois j ejpevessin, ... the quarrel between Odysseus and Achilles son of Peleus, how these once contended, at the gods’ generous festival, with words of violence, ... and Od.8.266–9: 69) aujta;r oJ formivzwn ajnebavlleto kalo;n ajeivdein ajmf j “Areo" filovthto" eju>stefavnou t j jAfrodivth", wJ"JJJ ta; prw't j ejmivghsan ejn ÔHfaivstoio dovmoisi lavqrh/: ... Then he struck the lyre and began singing well about Ares and sweet-garlanded Aphrodite, how they first lay together in the house of Hephaistos secretly ... There seems no metrical reason for this regular difference in position, and a structural motivation seems likely: that there are different origins for constructions with cognitive verbs and those with verbs of speech. The former involve indirect questions (‘know how’), and the latter take intransitive circumstantials, with main verb prepositional phrases (‘singing about Ares and Aphrodite, how they lay together’), where the objects delay the subordinate clause. This second type creates a focal link, comparable to that of the constructions with o{sti" and o{ti described earlier in this chapter. This is discussed further in Chapter 7, Section 2a. This aetiology is supported by rare Homeric constructions which Monteil (1963: 399) considers as the instrumental use of wJ",46 as at Od.4.389–390. This may be interpreted as a transitive circumstantial, with an early use of a speech verb with a direct object: 70) o{" kevn toi ei[ph/sin oJdo;n kai; mevtra keleuvqou, novston q j wJ"JJJ ejpi; povnton ejleuvseai ijcquoventa ... he could tell you the way and the length of the journey, and the homecoming, how you could venture on the fish-swarming sea. 46Biraud (1985: 163) follows Monteil in considering that all non-substantival Homeric instances of wJ" (and o{pw") may be interpreted as instrumental pronouns. On whether wJ" was originally instrumental or ablative, see Cristofaro (1998: 66, 85n4). 200 It is likely that adverbials following cognitive verbs and (with indirect objects) following speech verbs were both early forms, and that the development of complementation involved the transitive circumstantial construction following cognitive verbs, which could take a direct object (‘know x as being y’). In Chapter 6, it is shown that direct objects are extremely common in tragic wJ"-complements in tragedy, and that their placing accords with the function of wJ" as a focal marker. Summary: subordination in Homer with o{ti {{{ and wJ"JJJ The discussion of o{ti-clauses concentrated on two aspects of relative and completive subordination: 1) Clause order, leading to phonological effects at the clausal interface, involving a linking element functioning as a pronominal in one clause, as well as a conjunction in the other. The presence of a focalized verbal object creates a transitive circumstantial. 2) A change in the semantic class of object, from a referring expression to a textually-deictic word. The transition involved the use of indefinite pronouns, and reporting verbs, in both the main and subordinate clauses. Complementation with wJ" has a different origin, from adverbial clauses: either directly following cognitive verbs, or in circumstantials with indirect objects following speech verbs. In circumstantial constructions, the subordinate clause has an adverbial rather than substantival function. The central factor is the transitivity of the verbs, which is central to the development, not only of finite complements with o{ti and wJ", but also of non-finite complements (discussed in Chapter 6, Section 1d). The changes in transitivity of reporting verbs during the epoch covered by the Homeric and tragic texts are considered in the next chapter. 201 Chapter 6 Complementation: verb transitivity and focalization Summary In Chapter 5, o{ti-complementation was discussed in terms of clause order and the resultant phonetic features. The prominence of the inter-clausal link was interpreted as a focalized object in Homeric free relatives with o{sti" following cognitive verbs and o{ti following speech verbs. ÔW"-completives were described as adverbials following cognitive verbs, or circumstantial-like constructions following speech verbs. In this chapter, focalization of the object is considered from the point of view of main verb transitivity. The development of wJ" as complementizer is not, as with o{ti, the grammaticalization of a referring expression, but a change of grammatical function from an adverbial to a completive. The process involves the regular presence of explicit main verb objects, again creating a circumstantial structure. Two aspects of the change of function of these objects, from referring expressions to textually-deictic markers, are discussed: 1) The influence of non-finite on finite complementation. The key constructions appear to be circumstantial participials: accusative and infinitive (henceforth AI) constructions appear peripheral to the development of finite complements. 2) The presence of a focal main verb object in the majority of wJ"-completives in the tragic texts of the corpus. These constitute a structure comparable to that of o{ti-complements. i) In the early type, the objects are proleptic referring expressions. Their rarity following speech verbs is explained by restrictions on transitivity:1 such verbs do not take referring expressions as (semantically) direct objects (*’say them’). ii) In tragedy, textually-deictic elements regularly precede wJ": these are occasionally nouns explicitly naming ‘these words’, but more commonly the anticipatory demonstratives tovde or tavde. 1Standardly termed ‘selection restrictions’. See Katz and Fodor (1963), Chomsky (1965: 113ff., 1981, 1986), and Jackendoff (1983). 202 This analysis shows how finite complementation may have developed by an expansion in verb transitivity, and provides a new interpretation of prolepsis. In Section 5, the hypothesis of focal linking is put under test by examining possible counter-examples. As noted in the Introduction, Section 3, it is predicted that there is no initial emphatic position in subordinate clauses which follow their main, and that preposed elements function syntactically within the main clause. An examination is made of constructions in which the focal element in a subordinate clause does not function as main verb object. It is concluded that these structures do not invalidate the hypothesis. Chapter Sections 1: Complements and main verb transitivity 1a: Definition of complementation 1b: Transitivity of the introductory verbs 1c: Transitivity and factivity 1d: Non-finite complements 1d (i): Participial complements 1d (ii): AI complements 1e: Finite complements 2: Animate objects in finite complementation 2a: Prolepsis and Homeric specifying constructions 2b: Prolepsis 2c: Prolepsis and subordinating structure 3: Transitional constructions 3a: Double constructions 3b: Verbs of witness 4: Textual objects in wJ"-complementation 4a: Pleonastic objects 4b: Pronominal objects 5: Testing the hypothesis of focal linking Summary: focus, transitivity and speech verbs 203 1: Complements and main verb transitivity 1a: Formal definitions of complementation2 As discussed in Chapter 5, complements may be defined as completives or substantivals. The first category concentrates on the relationship of the subordinate clause to the main verb, as ‘the syntactic situation that arises when a notional sentence or predication is an argument of a predicate ... [and in particular]... if it functions as the subject or object of that predicate’ (Noonan 1985: 42). This is equivalent to the traditional definition of completives as propositions which are logically the subject or complement of a main verb (Meillet and Vendryes 1927: 661, Riemann and Goelzer 1897: 449). Chantraine (1963: 288) gives a more cautious definition, describing completives as propositions which ‘énoncent une notion indispensible à l’expression du verbe principale’. The substantival criterion is used by Kühner (1904: 348–377), who categorizes Substantivsätzen in terms of their nominal function. This definition concentrates on the meaning of the complementizer as ‘the fact’, rather than its subordinating role, and is not tied to particular classes of main verb, so is closer to a specifying meaning (‘in that’). This approach is adopted by Schwyzer (1950), Monteil (1963), and Lehmann (1984: 153–156). The substantival category is independent of the meaning of the main verb, but models the pattern of focalization: emphasis is regularly associated with specificity, as in demonstratives (see Chapter 4, Section 4b). It was argued in Chapter 5 that the completive categorization gives the best description of the development of the form. However, neither category explains the semantic types of complement-introducing main verbs, or their modality and factivity.3 Functionally, they do not distinguish indirect speech and rhetorical use. Nor do they explain why so many pronominals and adverbials are used as complementizers. In this chapter, the dependence of complements on particular types of verbs, and the high frequency of explicit main verb objects, are examined in terms of their grammatical relations. 2There have been no formal syntactic studies of finite CG complementation: generative studies of classical complementation (Lakoff 1968, Lightfoot 1971, Quicoli 1982) have all concentrated on the AI construction. 3Factive verbs presuppose the truth of their clausal complements. See Section 1c below. 204 1b: Transitivity of the introductory verbs The principal complement-introducing verbs are those of judgment, perception, and speech.4 The function of taking a clausal complement is a purely formal type of transitivity, or valency (Tesnière 1959), as a complement does not fulfil a non-linguistic thematic function (a ‘q-role’ in the X' schema).5 Since grammatical relations appear to derive from thematic functions,6 the development of complementation involves an expansion in the transitivity of the introductory verbs, to taking textually-deictic as well as thematic objects. The semantic reason is that clausal complements are not normally analysed as having thematic roles.7 Complement-introducing verbs do, of course, regularly take thematic object NPs (‘see/know someone’), but agentive and spatial terms seem inapplicable to clausal complements. Fillmore (1968: 85–6) considers them to be semantically vacuous (‘dummy’) factitives, but they might also be analysed as participants which limit the scope of the verb.8 In both o{ti- and wJ"-subordination in the corpus, a referring expression regularly serves as main verb object, with a subordinate clause as adverbial modifier. 1c: Transitivity and factivity As transitive constructions became progressively more common in IE languages (Coleman 1989, Bauer 1993b), increasing formalization could be seen as a loss of factive force: that is, the semantic property of presupposing 4A fuller categorization is given by Noonan (1985: 10–133), who lists complement introducing predicates (CTPs) as utterance predicates (‘say’), propositional attitude predicates (‘believe’), pretence (‘imagine’), commentative or factive (‘regret, be significant’), knowledge (‘know, see’), manipulative (‘persuade, let’), and others. 5Defined in terms of causality or agency (Tesnière 1959, Fillmore 1968), or of spatial or temporal goals (Gruber 1976, Jackendoff 1983). 6See Gruber (1976) and Jackendoff (1983). 7Gruber (1976: 128) and Jackendoff (1983: 203) identify even the complements of speech verbs as thematic: an utterance is interpreted as a thematic entity moving from the speaker [agent] to the hearer [goal]. It is, however, difficult to see how such an analogy can be sustained. See Munro (1982) and Amberber (1996) for further discussion. 8This constitutes a specifying function, which is pragmatically and prosodically focal. 205 the truth of a subordinate proposition.9 Since verbs of emotion presuppose the truth of their clausal complements (‘happy that [x is y]’ presupposes [x is y]), while cognitive verbs may be factive (‘see that x is y’ presupposes [x is y]) or non- factive (‘know x is y’ asserts [x is y]),10 and speech verbs are non-factive (‘say that [x is y]’ asserts [‘x is y’]),11 there is progressively less causal force in the sequence of [emotional> cognitive> speech] verbs. This appears to be taken by Chantraine (1963) as indicating the historical sequence,12 though the Homeric evidence suggests that completives depending on emotional verbs are no earlier than the others (see Chapter 5, Section 3). 1d: Non-finite complements If, as proposed in Chapter 5, early complements have a circumstantial form, then non-finite complementation would be expected to show the same precedence of circumstantials over other forms, such as jussives, and participial constructions are more likely to be the precursors of finite complements than are AI constructions. The evidence is surveyed in the next sections. The semantic difference between participial and AI constructions involves the factivity of the main verb: factive verbs are typically followed by participials and non-factives by infinitives.13 With verbs that can take either infinitives or participles, the infinitive construction normally has an imperative, inchoative, or final sense,14 and verbs of thought which have a connotation of judgment, as dokw', nomivzw, eijkavzw, pisteuvw, and uJpopteuvw, usually take infinitives, while 9See Kiparsky and Kiparsky (1970), Karttunen (1971), and Lyons (1977: 599–606 and 794–809). 10This view is justified below. Lyons (1977: 794), however, analyses ‘know’ as always factive, on the basis of epistemic necessity (if the proposition is negated, the presupposition remains). 11Verb person determines the difference here. 12Chantraine (1963: 289–290): ‘Les propositions déclaratives sont issues de propositions complétives de cause. Les propositions causales complétives se développent à la suite de verbes exprimant un sentiment. C’est de propositions causales de ce type que sont issues les propositions exprimant un jugement...’ See also Schwyzer (1950: 645). 13Kiparsky and Kiparsky (1970) note this feature in English too. 14Examples discussed by Smyth (1956: 474–476) include aijscuvnomai (ashamed at doing/to do), tlavw, a[rcomai (start/continue), gignwvskw (recognize/decide, determine), deivknumi (show/show how to), ejpilanqavnomai (forget /forget how to). 206 participles are normal with verbs of perception like oJrw', ajkouvw, aijsqavnomai, punqavnomai.15 The classification of ‘know’ as assertive rather than factive (suggested in the previous section) accords with this structural difference. As Kiparsky and Kiparsky (1970: 147) argue, the sentence ‘it is true that [John is ill]’ does not presuppose, but asserts, the truth of [John is ill], so is non-factive. The sentence ‘he knows that [John is ill]’ could also be analysed as assertive, while ‘he sees that [John is ill]’ presupposes, rather than asserts, the truth of the subordinate proposition, so is factive. Since many CG cognitive verbs can have either sense, as noted above, they may be interpreted as factive when taking participial complements. Their assertive use is a common feature in the corpus (see Chapter 7, Section 3). The implication for complementation is that participials can always be associated with factive verbs, which may take referring expressions as direct objects in circumstantial participial constructions, while non-factive verbs usually take infinitive complements.16 Cognitive verbs like oi\mai take direct thematic objects when functioning as verbs of perception (‘see John as being ill’), but only (semantically) indirect objects when functioning as verbs of judgment (‘believe [about] John to be ill’). Similarly, speech verbs take only (semantically) indirect human objects (‘it is necessary for them to go’ ... ‘order [to] them to go’). In both constructions, there is an accusative element with a double function, but the presence or absence of thematic objects mirrors the semantic difference between participial and infinitive constructions. It is argued below that the difference is reflected in the origin of non-finite complementation, and that participial complements precede AI complements. This may be seen from an examination of the semantic function of the accusative elements. 1d (i) Participial complements Participial constructions have two characteristic structural features: the subordinate clause is typically phonologically heavy, due to the inflection, 15See Riemann and Goelzer (1897: 687–8). 16The converse does not hold absolutely: occasional Homeric AI complements depending on cognitive verbs are used factively. See below in Section Id (ii). 207 and is also more syntactically integrated in the structure of the main clause than are infinitive constructions, because the inflection, in addition to any explicit subordinate subject, ordinarily agrees with whichever of the arguments of the main verb it is co-referent.17 Participial complements appear to derive, as Smyth (1956: 471) suggests, from circumstantial use following cognitive verbs, so ouj ga;r h[/desan aujto;n teqnhkovta (‘they did not know him as being dead’) may be interpreted as completive (‘... know that he was dead’). The accusative subject of such a construction is functionally the object of the main verb, so its case may be explained by assignment from the main verb in the canonical way.18 The function of the main verb in assigning case is less clear when the verb does not normally take an accusative. Participle and subordinate ‘subject’ may agree with the indirect object of a main verb, or take an accusative in a complement-like construction, as in the alternatives suvnoida soi eu\ poihvsanti or suvnoida se eu\ poihvsanta. In fact, Kühner (1904: 49) notes that an accusative construction with suvnoida is rare: the prefix suvn implies an indirect object, though an accusative construction may be more likely when there is also an indirect object, as at Dem. 61.23: suneidw;" tw'n ajqlhmavtwn' j v' j v' j v douvlou" metevconta" (knowing about the contests, that slaves participate in them). In that construction, differentiation of case may have a pragmatic motivation: of creating clarity. The opposite effect seems to be achieved in adjacent lines at Choe.216–7, where the change to accusative creates ambiguity, hiding the object ( jOrevsthn) among the other accusatives: 1) kai; tivna suvnoisqav moi kaloumevnh/v /v /v / brotw'n… suvnoid j jOrevsthn pollav s j ejkpagloumevnhnj j vj j vj j v . (El.) And whom among men do you know of me that I call upon? (Or.) I know that it is Orestes whom you very much admire. The assignment of case by the main verb, rather than as an accusative default, is the more likely, since the accusative construction seems to be 17In reflexive constructions, case is assigned with respect to the main clause, as lanqavnw ejmauto;n poiw'n''' ti (I am doing something unawares), since *lanqavnw ejmauto;n poiou'nta''' ti (I do not know that I am doing something) would be self-contradictory: see Kühner (1904: 50). 18The current syntactic model involves morphological ‘feature checking’ between a specifier and head: see Chomsky (1992). 208 chronologically later: none of the accusative constructions with suvnoida cited by Kühner (1904) predates the fifth century, while suvnoida with dative occurs in Herodotus. Case can always be explained as assigned by other main verbs too. Clearly, the accusative form developed by analogy, either with the majority of participial constructions depending on transitive verbs, or with the other non-finite form, the AI. This is considered in the next section.19 1d (ii): AI complements Case attraction requires adjacency, if it is a phonological process, and this is normally taken to be applicable to AI constructions:20 Horn (1985) considers that AI exists only in languages where the subjects of finite complements occupy the same apparent position as the object of the main verb. This is not strictly applicable to CG, which may have OV ordering in main clauses with a dependent AI clause, as in a co-referent construction at E. Alc. 641: 2) kaiv m j jjj ouj nomivzw [ ] pai'da so;n pefukevnai and I do not count myself as any true child of yours However, non-contiguous elements may be interpreted as preposed within the main clause: and the pronominal often follows the main verb, as at Pl. Hipp. Maior 282E: 3) ... kai; scedovn ti oi\\ \\mai ejmejjj ; pleivw crhvmata eijrgavsqai I know well enough I earn more money... Case assignment in AI is usually described on the analogy of control (jussive) sentences, where a semantically indirect but accusative object has a thematic relation with the main verb.21 Variations of case following a main verb taking an indirect object in jussive constructions (devomaiv sou ejlqei'n or devomaiv se ejlqei'n, I beg you to go) are usually explained with the accusative as the default, and other cases as the result of case attraction to that of the main verb object (Kühner 1904: 24, Smyth 1956: 438–440). This description is incomplete, as it does not explain the origin of the supposed accusative default, other than by analogy with the transitive sense of jussives (keleuvw meaning ‘urge on’, proei'pon 19For further discussion of participials, see Kühner (1904: 49ff.) and Quicoli (1982). 20By Kühner (1904), Rosenbaum (1967), Lakoff (1968), Lightfoot (1971), Andrews (1971), Miller (1974), and Quicoli (1982). 21The contrast between ‘control’ and ‘exceptional’ constructions, corresponding to jussives and non-finite complements, derives from Chomsky (1981). 209 ‘proclaim’, nouqetevw ‘advise’). Yet the very high proportion of jussives which take datives (kwluvw, boavw, ejpistevllw, etc.) weakens the argument for an accusative default: there must have been another, transitive, construction which provided the analogy, and this is likely to be the circumstantial following cognitive verbs (since the accusative is the thematic object). The view that AI complements are the precursors of finite constructions (Meillet and Vendryes 1927: 589, Aitchison 1979: 53) therefore lacks plausibility. There is, in fact, little textual evidence for it. Participial complements are common in Homer (Kühner 1904: 49), but AI complements are even less frequent than finite complements. This may be demonstrated by grouping Homeric AI constructions into four categories:22 i) Ditransitive constructions, where the accusative functions as the main verb object after jussives, as Il.17.30: 4) ajllav s j e[gwg j ajnacwrhvsanta keleuvw / ej" plhqu;n ijevnai but I myself tell you to get back into the multitude and Il.2.11: 5) qwrh'xaiv eJJ JJ kevleue kavrh komovwnta" jAcaiouv" Bid him arm the flowing-haired Achaians Meillet and Vendryes (1927: 561) consider this to be the earliest type.23 However, as noted above, the accusative element is semantically an indirect object (‘goal’ rather than ‘patient’), and its case must be motivated by some other construction. ii) The accusative functions as the infinitive subject in impersonal constructions, as at Il.1.126 (ejpevoike), 2.24 (crhv), Od.14.193 (ei[h), and after the prepositions privn and pavro", or in final clauses. In these constructions, the accusatives could have a thematic function as a ‘goal’ or as ‘patient’. Some are similar to ditransitive factitives (‘appoint him general’), as at Od.4.209–210: 6) wJ" nu'n Nevstori dw'ke diampere;" h[mata tavnta, aujto;nj ;j ;j ; me;n liparw'" ghraskevmen ejn megavroisin, As now he has given to Nestor, forever, all his days, 22Following Monro (1891: 202–203), Kühner (1904: 26–33), Meillet and Vendryes (1927: 561ff.), and Chantraine (1963: 312–318). 23Similar constructions occur at Il.14.62, Od.10.531–3, and Od.23.258. 210 for himself to grow old prosperously in his own palace, ... In the final and prepositional types, the accusatives may have a thematic function as goals, as they appear to do after verbs of movement in Homer.24 In the impersonal constructions (‘it seems that/it is necessary that [x]’), the accusative element is semantically an indirect object. This is explicit at Thuc. 1.120: 7) ajndrw'n ajgaqw'nj ' j 'j ' j 'j ' j ' ejstin ajdikoumevnou"j vj vj v ejx eijrhvnh" polemei'n It is right for good men, being wronged, to fight instead of peace. The accusative may be used to avoid ambiguity, as with the participial at Dem. 61.23 cited above in 1d(i). Otherwise, it must occur by analogy with some other construction. Impersonals, which may be termed modal subordinating predicators (Lyons 1977, 793–809), could be considered as similar to jussives, and another construction must have provided an analogy for their form. iii) Accusative constructions following desiderative, perceptual, and judgmental verbs. These constitute a type of complement, because the infinitive describes a fact or action, and the accusative element is logically the object of the main verb, as at Il.4.247: 8) h\ mevnete [ Trw'a"''' scedo;n ejlqevmen ]… Are you waiting for the Trojans to come close? The construction appears occasionally with perceptual verbs, when it has the circumstantial meaning normally associated with participials, as at Il.6.386–7: 9) ... ou{nek j a[kouse / [ teivresqai Trw'a"''' ], ... Because she heard that the Trojans were being pressed hard It is most plausible that this construction is an analogue of the more common participial circumstantial. iv) Chantraine (1963: 312) cites one reporting construction depending on a speech verb, at Il.1.521: 10) neikei', kaiv tev mev fhsi mavch/ Trwvessin ajrhvgein she accuses [me], and speaks of how I help the Trojans in battle 24The thematic roles of accusatives following Homeric verbs of movement (bavllw, i{kw, iJkavnw) are discussed by Kühner (1898: 303), Haudry (1977), and Boel (1988). 211 Here the accusative element is placed in the P2 collocation, preceding the speech verb and following connective particles, so focalizing neikei', and functions as its object, as well as subject of ajrhvgein. This construction is therefore a transitive circumstantial. Il.1.521 is the only citation by Monro (1891), Kühner (1904), or Chantraine (1963) of a AI complement depending on a speech verb which predates the fifth century. Although Moorhouse (1955: 181) finds 20 instances of AI (all dependent on fhmiv)25 in Iliad Books 1–10, he notes that in another 6 infinitive constructions the subject is omitted, even though it differs from the main verb subject, suggesting that the AI construction is a late development.26 An AI complement at Od.2.171 is cited by Coleman (1985: 327), who describes AI as common in Homeric Greek. Here, the order highlights the accusative element at the line end, which is followed by wJ" (a pattern discussed in Chapter 7): 11) kai; ga;r keivnw/ fhmi; teleuthqh'nai a{panta{{{ w{" oiJ ejmuqeovmhn... and about him I say that everything will be accomplished as I said... The circumstantial form of this construction may be seen from the datives, and from the prominence of the accusative element, and its word order may be influenced by the finite adverbial clause following. Of these types, the jussives (i) cannot, as argued above, motivate the accusative use in complementation. The ‘goal’ accusatives (ii) and the transitive circumstantials (iii) constitute the most plausible candidates. Spatial relations may provide the earliest analogue: it has been proposed (in the ‘thematic relations hypothesis’ of Gruber 1976 and Jackendoff 1983) that all thematic relations derive ultimately from spatial ones. However, the circumstantials are semantically closer, because they involve reporting verbs, and so are likely to have been involved in the development: probably by analogy with the more common circumstantial participials. This aetiology 25Moorhouse (1955: 180) considers that the predominance of fhmiv is explicable by its meaning as a cognitive rather than speech verb. 26Monteil (1963: 405) and Miller (1974: 241–2) describe AI constructions as more common than finite complements in Classical Latin too, so even if there was a linear development, it must have taken place independently in different languages. 212 would also explain the circumstantial form of early finite complementation observed here. 1e: Finite complements As noted above, in participial constructions, the subordinate element is phonologically heavy. In finite complementation, prominence is motivated syntactically, by the function of the linking element as focal main verb object. In Chapter 5, it was argued that o{ti functions as the object. The following sections describe how wJ"-complements regularly include an explicit object, either as referring expressions, or as textually-deictic demonstratives. 2: Animate objects in wJ"JJJ -complementation 2a: Prolepsis and Homeric specifying constructions A number of Homeric constructions with indirect objects and o{ti are similar to intransitive circumstantials, such as Il.1.56: 12) khvdeto ga;r Danaw'n''' , o{ti rJa qnhvskonta" oJra'to. for she pitied the Danaans, because/in that she saw them dying. and Il.23.555–6: 13) ’W" favto, meivdhsen de; podavrkh" di'o" jAcilleu;" caivrwn jAntilovcwvvv /, o{ti oiJ fivlo" h\en eJtai'ro": So he spoke, and brilliant swift-footed Achilles smiled, favouring Antilochos because/in that he was his dear companion These are also similar to the specifying completives following verbs of emotion occurring in Homer, which also follow main verbs with an explicit object, as at Od.11.102–3=13.342–3: 14) ... o{ toi kovton e[nqeto qumw'/ cwovmeno" o{ti{{{ oiJ uiJo;n fivlon ejxalavwsa". ... who holds anger against you in his heart, angry because/that you blinded his own son. These are the causals which Chantraine (1963) and Monteil (1963) consider to mark the origin of complementation. However, a completive interpretation (‘she grieved that the Danaans.../rejoicing that Antilochos...’) requires an object in the 213 main clause. It also involves greater inter-clausal integration, so an object is almost always followed by the phonologically light wJ", rather than o{ti.27 2b: Prolepsis Referring expressions may function as accusative objects in finite completives, especially with cognitive verbs, which regularly take a human object. The construction occurs in Homer, as at Il.2.409: 15) h[/dee ga;r kata; qumo;n ajdelfeo;n j ;j ;j ; wJ" ejponei'to he knew in his mind his brother, how he was troubled Prolepsis, as this structure is standardly termed,28 has usually been discussed in purely pragmatic terms, whose value is somewhat diminished by the contradictory interpretations which have been proposed:29 either that a preposed ‘subject’ is pragmatically prominent (Kühner 1904: 577–8, Gonda 1958), or that it has reduced emphasis because it is a theme (Panhuis 1984, Slings 1992: 105).30 The first view is more accurate in terms of the structural relationship between the two clauses: just as case attraction of the relative pronoun to that of the main clause gives more prominence to the latter (Gonda 1954a: 29), so prolepsis demonstrates the integration of the subordinate in the main. In any case, pragmatic indeterminacy (between emphatic or thematic function) can exist only in relation to the subordinate clause: the object is always prominent in the main clause. The view that ‘there is nothing emphatic or vivid etc. in a sentence containing a prolepsis’ (Panhuis 1984: 38) is especially inappropriate to minimal clauses of the oi\dav se o}"/wJ" ei\ pattern, as Il.9.527–8: 16) mevmnhmai tovde e[rgonv [v [v [ ejgw; pavlai ou[ ti nevon ge [ wJ"JJJ h\n: ...] I remember this behaviour of old, it is not a new thing, how it was. 27Two exceptions are cited in Section 2b. 28For its early rhetorical sense, of ‘anticipation’, see Hermogenes, Meth. 10, and other references in Liddell and Scott (1968: 1488). 29A bibliography is given by Slings (1992: 105n46). 30Slings (1992: 106) defines theme as syntactically disjunct from its clause, yet ‘articulating the focal information’, so defining theme, topicalization, and focus circularly. 214 and Eum.454: 17) gevno"vvv de; toujmo;nj ;j ;j ; [ wJ"JJJ e[cei ] peuvsh/ tavca. you will soon learn my race, how it is Such constructions always reduce the subordinate clause to a parenthesis, and emphasise the accusative. As with other, more extensive, types of prolepsis, the construction is most common in Euripides: wJ" e[cei occurs 15 times (Alc. 280, Her.956, Ion 1416, Troi.394, 923, 931, 1144, El.427, IA.106, 446, and in the fragments), as against two in Aeschylus (Eum.454 above and Fr. 726g), three in Sophocles (Trach.622, OT.1172, El.791), and one in Aristophanes (Eq.153). ÔW" h\n occurs once in Homer (Il.9.527–8 above), and in tragedy only in Euripides (And.381, HF.27, El.690, IT.532, Phoen.1280). Constructions with o{pw" e[cei occasionally occur, as at Ra.75, and the type occurs also with polar indirect questions, as at S.Phil.444: 18) tou'ton''' oi\sq j eij zw'n kurei'… Do you know of him, if he is alive? The prosody supports the interpretation of prolepsis as focalization, because it almost always involves phonologically light complementizers: proleptic o{ti-complements are rare, and in such constructions the subordinate clauses appear more peripheral, as at Od.8.461–2: 19) cai're, xei'n j, i{na kaiv pot j ejw;n ejn patrivdi gaivh/ mnhvsh/ ejmeijjj ' j, o{ti{{{ moi prwvth/ zwavgri j ojfevllei". Good-bye stranger and think of me sometimes when you are back at home, how I was the first you owed your life to. and Eum.970–1 (Monteil’s substantival type —see Chapter 5, Section 3): 20) .... stevrgw d j o[mmata Peiqou'"[ '[ '[ ' o{ti moi glw'ssan kai; stovm j ejpwvpa ... ... and I rejoice in the eye of Persuasion, that it was guiding my tongue and lips The function of the accusative element may be seen in its early dependence on cognitive rather than speech verbs: Homeric prolepsis appears always to involve cognitive verbs. Aeschylus occasionally uses prolepsis after other verbs, such as the verb of emotion at Eum.970–1 cited immediately above, after a speech verb at Septem 375–6, and at Choe.851–3, where the proleptic 215 element functions as object of a cognitive and a speech verb: 21) ijdei'n ejlevgxaiv t j eu\ qevlw to;n a[ggelon; [; [; [ , ei[t[[[ j aujto;" h\n qnhv/skonto" ejgguvqen parwvn ei[t[[[ j ejx ajmaura'" klhdovno" levgei maqwvn: I wish to see and question the messenger, whether he was near, being there, or whether he speaks, learning from a faint rumour. This is a partial analogue of the jussive construction, as to;n a[ggelon is semantically an indirect object of the second verb in the main clause (ejlevgxaiv, although though not the first). The only construction in the Oresteia where a speech verb governs an element preposed from the subordinate clause is an ‘object-to-object’ construction at Eum.308–311: 22) mou'san stugera;n ajpofaivnesqai dedovkhken, levxai te lavch ta; kat j ajnqrwvpou"v ; j j vv ; j j vv ; j j v wJ" ejpinwma'/ stavsi" aJmhv. it seems appropriate to show our grim song, and to speak of the lots among men, how our band apportions them. Sophoclean prolepsis also usually involves cognitive verbs, although an object-raising construction at OT.604 appears transitional (depending on the categorization of punqanovmai as a cognitive or speech verb): 23) peuvqou ta; crhsqevnt; v; v; v j eijj jj safw'" h[ggeilav soi: Enquire [about] the oracles, whether I declared them truly to you. In contrast, prolepsis may follow speech verbs in Euripides, even with human objects, as at Med.248–9: 24) levgousi d j hJma'" J 'J 'J ' wJ" ajkivndunon bivon zw'men kat j oi[kou", oi} de; mavrnantai doriv: They speak [of] us, that we live a safe life at home, while they fight with the spear Med.452: 25) levgous jj jIavson jv jv jv j wJ" kavkistov" ejst j ajnhvr: saying [of] Jason, how he is the worst of men 216 and Med.669 (the preposed genitive emphasizes that the object is human): 26) paivdwn ejreunw'n spevrmvvv j o{pw" gevnoitov moi. to ask [of] the seed of children, how I might have them The relatively late use of prolepsis following speech verbs reflects their transitivity: while both cognitive and speech constructions are used circumstantially, so in translation requiring the addition of a subject pronoun in the subordinate clause which is co-referent with the preposed element (‘consider the liliesi of the field, how theyi grow’), speech verbs have the additional difference of not functioning transitively, so the proleptic element is semantically an indirect object (‘they speak of usi, that wei live a safe life’). Prolepsis following speech verbs is, therefore, semantically similar to constructions with syntactically indirect objects, as at Med.1246–7: 27) kai; mh; kakisqh'/" mhd j ajnamnhsqh'/" tevknwnvvv , wJ" fivltaq j, wJ" e[tikte": ... and do not weaken, or remember about these children that you love them, that you bore them ... The early use of proleptic elements depending on cognitive verbs exploits their double valency, in being able to take a transitive circumstantial complement. This may be seen in constructions with verbs of fearing, as at Med.39–40: 28) pavscous j: ejgw\/da thvnde, deimaivnw tev nin mh; qhkto;n w[sh/ favsganon di j h{pato" ... I know her, and fear her, lest she drive a sharpened sword through her liver It may be demonstrated that the double transitivity of such constructions was perceived by speakers of CG, from the evidence of an Aristophanic joke at Ra.41, where a transitive sentence is turned into a complement main clause by the addition of an unexpected subordinate clause. The humour depends on the double transitivity of devdoika, so ‘afraid of’ becomes ‘afraid that’: 29) ÔW" sfovdra m j e[deise. Nh; Diva, mh; maivnoiov ge. (Dion.) How terribly afraid of me he was. (Xan.) Yes, [afraid] that you were mad. Prolepsis, then, appears to be a regular effect of ambiguous transitivity, which reveals the developing structure of complementation. The same ambiguity 217 appears in following eij-clauses, which may be interpreted as either conditional or completive (as discussed in Chapter 4, Section 5a). 2c: Prolepsis and subordinating structure Gonda (1958: 119) considers proleptic structure to be ‘a more or less mechanical reproduction of an originally paratactic supplementation to a short sentence’, presumably because it appears to be an early form, but the accusative element in fact provides evidence of the developing form of subordinating structures: ‘they speak about us that we live a safe life’ is not a mechanical reproduction of the forms [‘they speak of us’] + [‘how we live a safe life’], but is rather a development of transitive circumstantials (‘I know you how/who you are’), initially following cognitive verbs. If prolepsis represents the preposing of the subordinate subject to the main clause object position (Panhuis 1984: 26, Christol 1989), the configuration might, provisionally, be described as in Fig. 1: CP IP VP Verb NP (of the main clause) (of the subordinate clause) Object Figure 1 katamavqete ta; krivna tou' ajgrou' aujxavnousin pw'" In this configuration, the preposed phrase is not focalized, since pw'" is interpreted as in initial position in the subordinate clause, and emphasis would, as Panhuis (1984) and Slings (1992) assume, be somehow debarred from crossing the clausal boundary. However, the placing also creates rightwards weight within the main clause. The weakness of the Fig. 1 structure derives from its failure to model the prosody of the inter-clausal 218 link, and in particular the start of the subordinate clause and the absence of emphatic elements following the proleptic element. The subordinate subject may better be modelled as preposed to the focus position within the subordinate clause, which is adjacent to the main verb object position, and followed by a conjunction which is the head of the CP. Semantically, the focal element functions as the object of the main verb, so the structure is as in Fig. 2 (=Introduction, Fig. 11): IP VP tav krivna tou' ajgrou' pw'" aujxavnousin Fig. 2 CP C' P1 NP katamavqete The double function of the focal element is, as noted in the Introduction, formally anacoluthic, because (in X' terms) it can have only one q-role. However, the valency of verbs of knowing allows the pronoun to be interpreted intra-clausally as a direct object, so the double structure, though formally broken-backed, is perfectly intelligible, and, as noted earlier, is mirrored in English translation (‘consider the liliesi of the field, how theyi grow’). The difference between the structures in Figs. 1 and 2 affects the position of pw'", which is in complementizer position in Fig. 2, but may be in complementizer or focus position in Fig. 1. 219 Prolepsis therefore demonstrates a stage in the development of a separate complementizer and focus position. The contrasting pragmatic views described at the start of this section are reconciled by the model of focal linking: the proleptic element is indeed in the main verb object position, but it is also in subordinate focus position. 3: Transitional predicators 3a: Double constructions In Homer, o{ti may be interpreted as possessing causal force when it is governed by verbs of perception. After intransitive verbs of emotion, both o{ti and wJ" must be interpreted adverbially, mostly in a causal sense (‘because’), while after cognitive verbs, the conjunctions may be interpreted as manner adverbs (‘how’). Speech verbs, by contrast, were not followed by transitive circumstantial constructions, because they took as ditransitive complements only infinitive jussives (they did, of course, also take relative subordinates, both with and without head nouns). The change from Homeric practice to the high frequency of speech verb objects in tragic complements implies a semantic shift in the relation of the object to the subordinate clause (from a circumstantial to a textually-deictic one). The double transitivity of speech verbs may be seen at Od.19.463–4, where (after a verb of emotion taking an indirect object) a speech verb takes two explicit objects, e{kasta and oujlhvn: 30) cai'ron nosthvsanti kai; ejxereveinon e{kasta{{{ , oujlh;nj ;j ;j ; o{tti pavqoi:... they rejoiced in his homecoming, and asked about everything, and his wound, how he suffered it... Here, both e{kasta and oujlhvn are referring expressions, though e{kasta is indefinite. Oujlhvn could, perhaps, be parsed as the head noun of a following relative (‘the wound that he suffered’), though the modality of the subordinate clause would be inappropriate to such an interpretation. In constructions with wJ", however, the circumstantial meaning is unambiguous. Homeric complements following verbs of speech may be preceded by an 220 adverbial phrase, as at Od.8.266ff. (also cited in Chapter 5, Section 4b): 31) Aujta;r oJ formivzwn ajnebavlleto kalo;n ajeivdein ajmfjjj j “Areo" filovthto" eju>stefavnou t j jAfrodivth", wJ"JJJ ta; prw't j ejmivghsan ejn ÔHfaivstoio dovmoisi lavqrh/: .... Then he struck the lyre and began singing well about Ares and sweet-garlanded Aphrodite, how they first lay together in the house of Hephaistos secretly ...... In tragedy, there is a regular use of explicitly transitive constructions, with accusative objects. The characteristic association of speech verbs with explicit objects may be due to the influence of relative-type structures, and their restricted transitivity: the only object they may take is a textual marker (‘I say this...’). Because the referent of a speech verb’s object is itself a linguistic entity, the object functions somewhat like an introducer of direct speech. The semantic similarity between complementation and direct speech is noted by Kiparsky and Kiparsky (1970: 157n.7), and in many non-IE languages the complementizer itself is etymologically related to a speech verb.31 In Homer, of course, words are standardly reported in direct speech, followed by wJ" e[faq? or its cognates. 3b: Verbs of witness A form intermediate between cognitive and speech predicators occurs in a few constructions involving the notion of witness, as a verb or nominal predicator. At Med.619–20, the accusative is really factitive (‘invoke them as witnesses’): 32) ajll j ou\n ejgw; me;n daivmona"vvv martuvromai, wJ" pavnq j uJpourgei'n soiv te kai; tevknoi" qevlw: Well, I call as witnesses the gods how I am willing to help you and the children in every way 31This feature is noted by Lord (1976) in a number of African and Asian languages, including Yoruba, Tamil and Burmese. 221 The subordinate clause functions adjectivally as a modifier of the nominative mavrtu", at Choe.988–9: 33) wJ" a]n parh'/ moi mavrtu"vvv ejn divkh/ pote; wJ"JJJ tovnd j ejgw; meth'lqon ejndivkw" movron so he may be present as my witness in the trial at some point, how I justly pursued this fate (of my mother) The clause may also modify a gerund, as at Ag.1505–6. Here the preposing of the clause highlights its adverbial force: 34) wJ"JJJ me;n ajnaivtio" ei\ tou'de fovnou tiv" oJ marturhvswnJ vJ vJ v … that you are innocent of this murder, who [will be] the witness? Aeschylus uses the verb transitively, with a deictic object, at Ag.494–6: 35) ... marturei' dev moi kavsi" phlou' xuvnouro" diyiva kovni" tavdevvv , wJ"JJJ ou[t j a[naudo" ou[tev soi daivwn flovga u{lh" ojreiva" shmanei' kapnw'/ purov": and the neighbouring brother of mud, thirsty dust, witnesses to me this, how he is not voiceless, nor for you kindling the flame of mountain wood will he signal with smoke of fire These constructions are somewhat similar to completives depending on verbs of showing (deivknumi, dhlovw, kathgorevw, mevmfomai, staqmavomai, tekmhvrion), which are a feature of Herodotean complementation (Neuberger-Donath 1982: 260–263). The Aeschylean construction, however, demonstrates an additional feature: the use of a textually-deictic object. 4: Textual objects in wJ"JJJ -complementation In the epic and tragic texts considered here, the subordinating conjunction wJ" is regularly preceded by an accusative element functioning as main verb object, and in the tragic and prose texts the majority of wJ"-complements have antecedents of some kind: every instance of completive wJ" in the Oresteia, almost all in OT. and Crito, and most in Medea, follow a main clause accusative element. Citations are given in Appendix 3B. The elements may be 222 divided into three categories: the proleptic elements described above, and nominals or pronominals with purely textual reference. Although o{ti is occasionally preceded by tov or tou'to (Ant.61, 98, 188; Prom.377), accusatives are associated almost exclusively with wJ". The reasons may be etymological as well as phonological: Monteil (1963: 329) describes wJ" as originally anaphoric: ‘Ancienne forme casuelle d’instrumental, wJ" a dû primitivement se référer à un substantif antécédent, à l’intérieur d’un énoncé anaphorique puis relatif.’ Accusatives are more frequent with verbs of speech than verbs of knowing. The explanation is not that wJ" is more likely to be used after verbs of speech and o{ti after verbs of knowing: there is an increase in complements dependent on verbs of speech with both conjunctions. The preference for a specifying object seems to be connected with verb valency: cognitive verbs may take a human object, while speech verbs do not, except as indirect objects in jussive constructions. This motivates their association with pronominals with purely textual reference, while cognitive verbs may take objects which have exophoric reference, in proleptic constructions. 4a: Pleonastic objects The textual reference may be explicit, and expressed by nominals following speech predicators, as Med.776–7: 36) molovnti d j aujtw'/ malqakou;";;; levxw lovgou"vvv , wJ" kai; dokei' moi taujtav, kai; kalw'" e[cein when he comes I shall speak soothing words, that the matter seems the same to me.... Or.892–3: 37) lovgou"vvv eJlivsswn, o{ti kaqistaivh novmou" ej" tou;" tekovnta" ouj kalouv": ... ... twisting words, that he set precedents dangerous for parents 223 and in lyric at IT.1092–3 (one of two instances in these texts where complementizing o{ti stands at the head of the poetic line):32 38) eujxuvneton xunetoi'" boavnvvv , o{ti{{{ povsin keladei'" ajei; molpai'" a cry intelligible to those who can understand that you mourn your husband with songs In these constructions, the object position is filled by a nominal which explicitly categorizes the subordinate (as boav or lovgoi). NPs like malqakou;" lovgou" do not only specify the syntactic function of the subordinate clause, but also comment on its form as speech, along the lines of the Homeric e[pea pteroventa.33 Sophocles uses the construction at OT.790–1: 39) kai; deina; kai; duvsthna proujfavnh; ; v j v; ; v j v; ; v j v levgwn, wJ" mhtri; me;n creivh me micqh'nai, .... and saying terrible and lamentable revelations that I was to wed my mother ... and at S. El.44: 40) lovgwvvv / de; crw' toiw'/d j'/ j'/ j'/ j, o{ti xevno" me;n ei\ ... Use this story, that you are a foreigner... In a participial, circumstantial-like, construction, at OT.1287–90, the NP does not refer catadeictically, but retrospectively (to the unspoken head noun in the phrase ‘to;n mhtro;" ...’): 41) boa'/ dioivgein klh'/qra kai; dhlou'n tina toi'" pa'si Kadmeivoisi to;n patroktovnon, to;n mhtro;": aujdw'n ajnovj vj vj vsi j oujde; rJhtav moij j ; J vj j ; J vj j ; J v , wJ" ejk cqono;" rJivywn eJautovn, ... he calls for someone to undo the bolts, and show him to all the Cadmeans, his father’s killer, his mother’s — saying unholy things, unutterable by me, that he will cast himself out of the land 32The other is Ra.599, cited in Chapter 7, Section 3c. There are also appositive constructions with line-initial o{ti at Eum.971 and Ra.20, 742. 33‘Winged words’, as Il.1.201, 2.7, 3.155, 4.92, and many other instances. 224 These nominals are always in tragedy associated with speech verbs, because they can take only textual objects. One Homeric construction involves a verb of perception, at Il.17.641–2: 42) .....ejpei; ou[ min oji?omai oujde; pepuvsqai lugrh'" ajgg' j' j' j elivh"vvv , o{ti oiJ fivlo" w[leq j eJtai'ro". ... since I think he has not yet heard the terrible news, that his dear companion has perished. 4b: Pronominal objects The most common textually-deictic object, however, is the demonstrative, as at Ag.494–7 (cited above in Section 3b) and OT.729–30: 43) “Edox j ajkou'sai sou' tovdvvv j, wJ"JJJ oJ Lavi>o" katasfageivh pro;" triplai'" aJmaxitoi'". I thought I heard you say this, that Laius was killed where three roads meet. 34 The complementizer may be seen as enclitic upon the pronominal, in the contrastive pattern described in Chapter 4, Section 2b. The deictic force of the demonstrative (itself prominent through cohesive focalization), creates an emphatic effect, drawing attention to the subordinate proposition, as at Medea 85–6: 44) ...... a[rti gignwvskei" tovdevvv , wJ"JJJ pa'" ti" auJto;n tou' pevla" ma'llon filei', Have you only just now learned this, that/how each loves himself more than his neighbour? The positions of both the pronominal and of wJ" with respect to the line end are quite regular, as shown in Chapter 7. Constructions in which the clause break does not coincide with the line end occur mostly with cognitive/perceptual verbs, as at OT.729–30 (cited above). In these instances, the object may be preposed within the main clause, as at Med.1405: 45) Zeu', tavdvvv j ajkouvei" wJ"JJJ ajpelaunovmeq j, Zeus, do you hear this, that/how I am driven away... 34On the significance of the definite article, see Dawe (1982: 165). 225 Constructions with pronominals do not generally mark factivity, as they assert, rather than presuppose, the truth of the subordinate: the pronouns mean ‘these words’ rather than ‘this fact’, so their function is that of impure textual deixis (referring to a proposition: see Chapter 4, Section 4b). There is an extremely high frequency of accusatives in tragedy: all complements with wJ" in the Oresteia (11), and most in OT. (10 out of 15) and Medea (10 out of 14), have them. Some are cited in Chapter 7, Section 2c, and other constructions with pronominals are cited in Appendix 3B. The structure is modelled in Fig. 3 (=Introduction, Fig. 10):35 IP VP Fig. 3 tovde wJ" pa'" ti" auJto;n tou' pevla" ma'llon filei' Adv. V a[rti gignwvskei" CP C' Focus P1 The construction is analogous to the o{ti-complements discussed in Chapter 5, and involves a similar focal pattern: the demonstrative typically occupies the same line-final position as o{ti. The functional difference is that o{ti itself performs both the interrogative and object functions, while wJ" and a main verb object share the interrogative function. With both, however, the significance of an indefinite suffix is the reduction in causal connection between the clauses (see Chapter 5, Section 2d, n.27). 35The citation is from Medea 85–6, ‘Have you only just learned this, that each loves himself more than his neighbour?’ 226 5: Testing the hypothesis of focal linking The use of proleptic and demonstrative accusatives accords with the mechanism of focal linking described in Chapter 5. As noted in the Introduction, this requires there to be no focus position in structurally subordinated clauses in CG. A study of the corpus texts and all epic and tragic o{ti-constructions revealed five constructions which appear to be counter-examples. They are examined here. In four constructions, an element is preposed before the conjunction, but remains nominative. The most perplexing construction occurs at OT.779–780: 46) ajnh;r ga;r ejn deivpnoi" m j uJperplhsqei;" mevqh/ kalei' par j oi[nw/ plasto;";;; wJ"JJJ ei[hn patriv. For at dinner a man overfilled with drink called me, drunk, that I was counterfeit to my father. There are three possible explanations for the nominative: i) It could be an effect of ambiguity between the adverbial and conjunctive functions of wJ" (as also with the difficult tranw'" jAtreivdhn eijdevnai kurou'nta o{pw" at Ag.1371).36 ii) It might function as a pragmatic marker, identifying the speaker from the subjects in the surrounding text, as a ‘hanging’ nominative.37 However, this does not accord with the presence of m j in the previous line. iii) It may mimic direct speech, perhaps because the presence of an accusative object in the previous line debars plastov" from occupying the object position. The last seems the most plausible, though it does not preclude the first, especially as speech verbs do not, before Euripides, generally take accusative proleptic constructions (as noted above, Section 2b). This interpretation is supported by three similar constructions with o{ti which also follow speech verbs. These constructions are formal announcements, so a parallel with direct speech is likely. 36 ‘... to know definitely how it is faring with the son of Atreus’ (?): see Denniston and Page (1957: 195). 37Nominative pronominals occur in extended infinitive constructions, as at Thuc. 4.114 and Dem. 21.204. See Smyth (1956: 439). The third episode of OT. has a very high frequency of subjects: 45 in 165 lines (=1 every 4 lines). 227 The nominative constructions occur at Hel.1491–4: 47) karuvxat j ajggelivan Eujrwvtan ejfezovmenai, Menevlew"vvv o{ti Dardavnou povlin eJlw;n dovmon h{xei. Proclaim the message as you perch on Eurotas that Menelaus has taken the town of Dardanus and will come home Ba.173–4: 48) i[tw ti", eijsavggelle Teiresiva"vvv o{ti zhtei' nin: ... Someone go and say that Teiresias is looking for him. and Ra.519–20: 49) “Iqi nun, fravson prwvtista tai'" ojrchstrivsin tai'" e[ndon ou[sai" aujto;"j ;j ;j ; o{ti eijsevrcomai. Go now, and first tell those dancing-girls inside that I myself am coming in A more serious counter-example to the hypothesis of focal linking is constituted by a preposed dative phrase in the complement at OT.525–526: 50) Tou[po" d j ejfavnqh tai'" ejmai'" gnwvmai"' j ' v' j ' v' j ' v o{ti peisqei;" oJ mavnti" tou;" lovgou" yeudei'" levgoi… But was the word clear [ ], persuaded by my advice, that the prophet gave false answers? It is clear that tai'" ejmai'" gnwvmai" is emphatic (see Dawe 1982: 147), and may therefore be interpreted as focalized, so appears to be a counter-example to the claim of identity between main verb object and subordinate focus. It may, however, be observed that there is no main verb object here, and the pragmatic motivation for preposing the dative (together with a metrical one, in the Sophoclean regularity of putting o{ti in the last foot of the iambic line) may be realizable structurally precisely because of the absence of a main verb object. 228 Though all are syntactically irregular, only the first construction has a preposed focus co-occurring with an explicit main verb object, and this puts a clear stress on its syntax (since all commentators regard it as problematic). It is therefore concluded that none of these constructions disproves the hypothesis that an element in subordinate focus position is indistinguishable from a main verb object.38 It may be noted that accusative elements which are clearly preposed from the subordinate clause may be also preposed within the main clause: the possibility of OV ordering with pronominal objects has been noted above, and the same order may occur with proleptic nouns, as at OT.842–3: 51) Lh/sta;"/ ;/ ;/ ; e[faske" aujto;n a[ndra"[[[ ejnnevpein w{" nin katakteivneian..... you were saying of robbers, that he said that they killed him..... and OT.955–6: 52) jEk th'" Korivnqou, patevra to;n so;nv ; ;v ; ;v ; ; ajggelw'n wJ" oujkevt j o[nta Povlubon, ajll j ojlwlovta. ...from Corinth, announcing your father Polybus as no longer living, but having perished. Preposing the object within the main clause requires it to move out of subordinate clause focus position. However, the extra interpretative difficulty of such constructions create a tension between meaning and structure, which does not exist with VO ordering. Summary: transitivity and speech verbs The importance of verb transitivity has been demonstrated in completive constructions with explicit main verb objects, including textually-deictic pronominals and proleptic nominals. Though transitivity presumably originally reflected non-linguistic agentive relations, its expansion to 38A contrast may be drawn with the focalized element within a following conditional at OT.120–1 (cited in Chapter 4, Section 5a), demonstrating the difference between a subordinated and a peripheral clause. A somewhat similar construction at Crito 51c6, cited in Chapter 7, Section 3a, appears motivated by interference between features of direct and indirect speech. 229 encompass textual objects as well as referring expressions is evident in both non-finite and finite Homeric and tragic complementation. The importance of main verb objects is that they provide an explanation of the change of function of wJ" from an adverbial to a completive, through the intermediate stage of transitive circumstantial constructions. The proposed historical sequence is: participial intransitive circumstantials, transitive circumstantials following verbs of knowing; then a convergence of indirect questions with o{ti following speech verbs and adverbial wJ"-clauses following cognitive verbs; a transitive circumstantial structure with a focalized object, with either complementizer, and eventually with either verb type. The AI construction appears peripheral to the aetiology. The explanation advanced here is based on the interaction of structural and prosodic features. However, the development of complementation presumably had pragmatic motivation, and, in poetry, its prosodic patterns are expressed in metrical form. The relationships between prosodic features of the poetic line and the presentational functions of complementation are discussed in Chapter 7. 230 Chapter 7 Inter-clausal poetic syntax: focus and the discourse functions of subordination Summary In Chapter 6, the overlap between main verb object and subordinate focus was examined in terms of verb transitivity. In this chapter, the clausal overlap is considered in terms of prosodic patterning and discourse function. The focalization of the inter-clausal link, discussed in Chapters 5 and 6 in terms of its prosodic pattern, is discussed in terms of the positions of o{ti and wJ" in the hexameter and trimeter line. Though complements are often described in terms of indirect speech,1 that is not a common function of complementation in the corpus, which is in epic principally the management of point of view, and, in tragedy and Plato, asseveration and rhetorical persuasion. Relationships between clause order and function are also examined, in a discussion of preposed complements. Discourse function is seen to be reflected in the patterns of focalization. Chapter Sections 1: Complementizer use and meaning 1a: Distribution within the corpus 1b: Meaning: o{ti and wJ" 2: Complementizers and focus 2a: Prosodic features of Homeric complements 2b: Prosodic features of tragic o{ti-complements 2c: Prosodic features of tragic wJ"-complements 2d: Prosody and syntax 3: Discourse functions of complementation 3a: Structural implications of function 3b: Discourse function and clause order: preposing 3c: Complementizer meaning in preposed complements Conclusion 1As by Jannaris (1897: 453), and Smyth (1956: 580–1). 231 1: Complementizer use and meaning 1a: Distribution within the corpus The most unexpected finding is that there is a diachronic increase in the number of complementizers between Homeric and 4th century texts. This is in contrast with the view of Monteil (1963: 400) that there are in effect only two completive conjunctions by the end of the fifth century: an assumption based on conjunction frequency rather than variety. The prose works considered here, the Melian Dialogue and Crito, demonstrate a narrowing of conjunction frequency (o{ti is the principal introducer), but not a reduction in variety (o{pw" also occurs in Crito and diovti elsewhere in Plato). It may be that by the fourth century the meanings of subordinating conjunctions had become more established than in Homer. However, complementation structure remains highly fluid, and retains its focal emphasis even in Plato. The distribution of the principal complementizers (o{ti, wJ", o{pw" and eij) in the corpus (plus Prometheus) is schematized in Table 1 (which excludes free relative and final constructions): o{ti{{{ wJ"JJJ o{pw"{{{ eijj jj Totals Iliad 9 1 5 2 0 8 Odyssey 9 1 2 1 2 6 Oresteia 1 11 4 11 27 Other Aesch. works 0 12 1 0 13 Prometheus 7 9 2 1 19 OT. 5 11 5 2 23 Medea 1 14 1 7 23 Cyclops 2 2 0 2 6 Frogs 7 4 4 (1 Euripidean quote) 15 Melian Dialogue 7 4 0 0 11 Crito 33 14 2 4 53 65 88 22 29 204 232 Four principal features are evident: i) There is a low level of finite complementation in Homer and tragedy. ii) An increase in prose is accompanied by a movement from wJ" to o{ti (also evident in Frogs). This may be a more general feature of Attic prose: Lysias also demonstrates a great preference for o{ti.2 iii) Finite complementation is particularly rare in Aeschylus, and the use of o{ti is very low in both Aeschylus and Euripides (the rarity of complementation in Aeschylus may be an archaism, since it is also very rare in Cyclops). iv) There is a higher level of indirect questions with eij in the Oresteia and Medea than in the other texts.3 v) The frequency of o{ti in Prometheus is atypical of Aeschylus, and is closer to Sophoclean frequency (on which see also Section 2a.i below). Other complementizers include mhv (in most texts), oi|on, o{pa/, o{ph (Oresteia), ejavn (Prometheus, Crito); o{pou (OT.), oi|a (Medea), oi|o" (Cyclops), o{tihv (Frogs). Even causal complementizers like oJqouvneka and ou{neka (which Monteil 1963, 400 regards as insignificant, because they are restricted largely to tragedy) do not decline in frequency: they are uncommon even in Homer (not occurring in the Iliad). A few complements are introduced by ou{neka in Homer, Sophocles, and Euripides,4 and by oJqouvneka in Sophocles (Trach.813, OT.572, 1271, El.47, 617, 1308, OC.852, 944, 1005) and Euripides (Alc.796). Similarly, the frequency of o{pw" as a complementizer remains at a constant low level, rather than declining from the 7th to the 4th C. In Homer, only 1 out of 30 subordinating constructions is a complement (Il.10.491–2). Aeschylus uses it in at least three complements (Supp.289ff., Ag.105ff., Eum.591),5 and Sophocles 3 times in OT. (OT.548, 1058, 1366). There are 5 instances in Medea (171, 322–3, 669, 1060, 1099–1102). Subordinating constructions with o{pw" in Frogs are mostly final: 4 out of 12 introduce complements. The use of adverbials may reflect the persistence of circumstantial constructions.6 2Monteil (1963: 399) finds 293 instances as against 135 for wJ". 3Constructions in the Oresteia are cited in Chapter 6, Section 4c. In Medea, polar indirect questions occur at 184–5, 346, 931, 941, 1319, 492–4, and 1103–4. 4 Od.5.215–6, 7.300, 15.42, 16.300, 16.379, S.Phil.232, Ant.63, OT.708, S. El.1478, E.IA 102. 5Ag.1371 is another possible example: see Denniston and Page (1957: 195). 6A survey of the use of o{pw" may be found in Amigues (1977). 233 In addition to its variant o{tti, two compounds of o{ti occur: oJtihv and diovti. The former, defined by Liddell and Scott (1968: 1265) as a colloquial form of o{ti in its causal meaning, appears occasionally in the fifth century, but not in tragedy. It is used to introduce complements in Aristophanes, at Nu.331, Ra.1146, and Plutus 48, and appears once in Plato, at Philebus 58a1: 1) Dh'lon oJtih; ... pa'" a]n thvn ge nu'n legomevnhn gnoivh It is clear that ... everyone would know what has just been said. Diovti is much more common, though only in prose (apart from Aeschylean fragment 19 321.b1). It occurs in Herodotus, Isocrates, Isaeus, and Demosthenes, usually with causal meaning, though an appositive, specifying, meaning is also evident. It introduces complements in perhaps 2 of the 14 Herodotean constructions (2.50.2, 6.86.24), 1 of 5 in Isaeus (3.50.6), and 8 of 26 in Demosthenes.7 Only Isocrates uses it primarily to introduce complements after cognitive verbs, in 10 out of 12 constructions.8 It becomes much more common in the fourth century, appearing in Plato 62 times, and over 460 times in Aristotle. It appears, then, that there is an increase, rather than a decline, in the number of complementizers by the fifth century, even in prose. The only post- Homeric reduction is the almost total abandonment of o{ te (explicable by the restriction of te to a co-ordinating link). Surprisingly, relative o{ is as frequent in Aeschylus as in Homer, and occasionally occurs elsewhere in the fifth century, usually followed by focalizers.9 1b: Complementizer meaning: o{ti{{{ and wJ"JJJ ”Oti is generally thought to have a more objective meaning than wJ": Humbert (1960: 185) proposes that ‘en général, on emploie wJ" quand le jugement énoncé comporte des réserves —comme quand le verbs principal est négatif ou quand on ne prend pas à son compte ce que dit quelqu’un’, while Smyth (1956: 582) notes that the subordinate verb may be the negative one. Monteil (1963: 356) considers that, ‘tandis que o{ti insiste sur la réalité du fait, wJ" 7Cor. 155.11, 167.5, Or. 46.16.1, 47.42.3, 58.36.10, 58.42.1, 59.111.4, Erot. 38.1. 8 In Call. 1.6, 31.1, In Loch. 8.2, De Big. 43.5, Paneg. 48.3, Plat. 23.2, Arch. 24.5, De Pac. 14.1, Antid. 133.2, Philip 1.1. 9Monteil (1963: 399) cites 24 substantival constructions in Homer, and Sommerstein (1989: 128) gives a total of ‘nearly 30’ for relative *to- in Aeschylus. It occurs with final clauses at E.Phoen.155, Hec.13, Ar.Eccl.338; and a{, with following particles, at S.Phil.559, Trach.136. 234 exprime des nuances de doute ou de simple vraisemblance’. Chantraine (1963: 291) identifies the core value of Homeric wJ" as adverbial ‘comme’. Yet Biraud (1985: 170) notes that there is no presumption of subjectivity in the Homeric wJ" e[fato (‘so he spoke’), and suggests that there is a difference only when there is explicit contrast between the complementizers. In the characteristically emphatic structures of the corpus, a contrast in meaning is evident. Two chiastic patterns at A.Prom.259–260 and E.Cyc.321– 322 demonstrate the difference clearly. In both constructions, o{ti precedes and wJ" follows, with the former dependent on a verb of perception and the latter on a value judgment. The adverbial force of wJ" is emphasized by the clause order: as discussed below in Section 3c, it is especially clear in preposed complements, because wJ" is then emphatic in P1 (with focalizers following it). Prom.259–61: 2) ........ oujc oJra'/" o{ti{{{ h{marte"… wJ"JJJ d j h{marte", ou[t j ejmoi; levgein kaq j hJdonh;n soiv t j a[lgo" .... do you not see that you have erred? yet how you erred is not pleasant for me to speak, and pain for you A similar preposed completive occurs at Cyclops 321–3: 3) oujd j oi\d j o{ ti{{{ Zeuv" ejst j ejmou' kreivsswn qeov". ou[ moi mevlei to; loipovn: wJ"JJJ d j ou[ moi mevlei, a[kouson. o{tan a[nwqen o[mbron ejkcevh/, nor do I know that Zeus is a greater god than I. I am not concerned for the future, and how I am unconcerned, listen.... In Crito, a clear difference may be seen in a correlation between complementizer and verb type: o{ti is used after cognitive verbs (or dh'lon) in 13 complement constructions, as Crito 49d2: 4) oi\\ \\da ga;r o{ti{{{ ojlivgoi" tisi; tau'ta kai; dokei' kai; dovxei. For I know that there are few who believe or will believe this. 235 Similarly at Crito 51a7: 5) h] ou{tw" ei\ sofo;" w{ste levlhqevnv vv vv v se o{ti{{{ mhtrov" te kai; patro;" kai; tw'n a[llwn progovnwn aJpavntwn timiwvterovn ejstin patri;" kai; semnovteron kai; aJgiwvteron Or are you so wise that it has escaped your notice that your country is more to be honoured than your mother and father ... By contrast, all 8 constructions with wJ" depend on verbs of emotion or attitude, as Crito 44b9: 6) e[ti de; kai; polloi'" dovxwvvv , oi} ejme; kai; se; mh; safw'" i[sasin, wJ"JJJ oi|ov" t j w[n se swv/zein eij h[qelon ajnalivskein crhvmata, ajmelh'sai. And I shall seem to many who do not know me and you well that being able to save you if I had been willing to spend money, I neglected it. and Crito 44c3: 7) .............................. ouj ga;r peivsontavvv i oiJ polloi; wJ"JJJ su; aujto;" oujk hjqevlhsa" ajpievnai ejnqevnde hJmw'n proqumoumevnwn. For the many will not believe that you refused to get away, while we were eager to help. All 13 constructions are listed in Appendix 3C. The difference in meaning accords with the mechanism of focal linking, as o{ti is focal and draws attention to the following proposition (see Chapter 4, Section 4b), while wJ" emphasizes the preceding constituent (‘the many’), on which the prosodic emphasis is placed. The morphology of the words fits this interpretation, too: the origins of o{ti as a determiner and of wJ" as anaphoric accord with the distinction made by Benveniste (1933: 124) between demonstrative and anaphoric themes as respectively strong and weak in morphological characteristics and semantic value.10 The interpretation of the difference as focal would explain why, in the Melian Dialogue, wJ" appears to be used simply to alternate with o{ti, as in the construction at 5.89.1ff., cited below in Section 3 (where it is noted that the complementizers bear little prosodic emphasis). 10Though adverbial w{" may have a different etymology from the conjunction wJ": the former from *Ûwv" and the latter from the relative adverb yw("). See Ruijgh (1971: 856). 236 A related interpretation is suggested by Cristofaro (1998: 73), who considers that ‘o{ti typically conveys new, focalized and non-topical information ... while wJ" introduces already known, non-focalized and topical information’. Cristofaro uses these terms as pragmatic categories,11 but prosodic focalization provides an even more precise explanation, since it identifies a specific constituent as focal. 2: Complementizers and focus The regular presence of prosodic emphasis and main verb objects in the corpus shows a contrastive style, perhaps stimulated by the experimentation naturally associated with a developing syntactic form. An emphatic link is implied, not only by proleptic objects, but also by the metrical placing of the introductory conjunctions. The typical positions of both complementizers in tragedy (o{ti at the line end, wJ" at the start) give them a prosodic prominence which emphasizes their linking function, and accord with the focalization patterns described in Chapters 4–6. In both Homer and tragic stichic verse, there is often a line break between speech verbs and their complements, in contrast with cognitive verbs, which are usually in the same line as their subordinates. The position of wJ" in Homer varies according as it follows a cognitive or a speech verb, while the position of o{ti with respect to the line break is very regular in tragedy. There appears to be a structural as well as metrical motivation for these regularities. 2a: Prosodic features of Homeric complements The normal position of o{ti is immediately following the main verb (which is frequently at the start of the line). There are 4 line-initial instances of complementizing o{tti: Il.5.407, 13.675, 24.564, and Od.4.392, and 8 instances which introduce causal clauses.12 11Her terminology is based on the model of Dik (1989). 12 Il.14.407, 15.156, 16.531, 17.568, 22.292, and Od.14.52, 14.527, 21.415. 237 Both o{ti and o{tti are frequently followed by enclitic elements, which are typically dative pronouns (preceded by short-vowelled particles after o{tti, evidently for metrical reasons), as at Il.24.241: 8) h\ ojnovsasq j o{ti moi{{{ Kronivdh" Zeu;" a[lge j e[dwke Is it not enough for you that Zeus son of Kronos has given me sorrow 13 and Il.17, 567–8: 9) ’W" favto, ghvqhsen de; qea; glaukw'pi" jAqhvnh, o{tti{{{ v rJav oiJJ v JJ v JJ v J pavmprwta qew'n hjrhvsato pavntwn. So he spoke, and the grey-eyed Athene rejoiced that first among all the divinities she had been prayed to. 14 The presence of short-vowelled particles after o{tti must be metrically constrained, but the regular presence of enclitics following both forms suggests a functional motivation: that a conjunctive meaning is linked with focalization by an enclitic in P2. The line position of o{ti is quite variable, while o{tti is more often line-initial. Both features reflect metrical constraints, but more is involved, as o{tti is more frequently line-initial after verbs of emotion than after other types (as may be seen from the citations in Appendix 3A), so causal and resultative clauses are prosodically as well as syntactically more peripheral than relatives. The position of wJ" also varies according to the main verb: as noted in Chapter 6, Section 4b, it is normally placed directly after cognitive verbs, while after speech verbs it is normally line-initial, and generally does not follow the verb immediately. The Homeric constructions cited by Monteil (1963) show the same correlation, and also reveal a stylistic difference between the Iliad and Odyssey: constructions in the Iliad mostly (6 out of 7) follow line-initial verbs of perception or knowing, as Il.9.647: 10) mnhvsomai w{"{{{ m j ajsuvfhlon ejn jArgeivoisin e[rexen jAtreiv>dh" ....... I remember how the son of Atreus insulted me before the Argives 13Datives follow o{ti also at Il.1.537, 2.255, 5.326, 16.35, 23.484, 23.545, 23.556, 23.577, Od.5.340, 11.103, 13.343, 17.377, 16.130, 18.11, and 19.248. 14Particles follow o{tti also at Il.13.675, 14.407, 15.156, 16.531, 17. 411, 17.655, 22.292, 22.439, Od.14.52, 14.367, 14.527, and 21.415. 238 In contrast, the 9 constructions following verbs of speech are all from the Odyssey, and in all, wJ" is line-initial, as Od.22.373: 11) o[fra gnw'/" kata; qumovn, ajta;r ei[ph/sqa kai; a[llw/, wJ"JJJ kakoergivh" eujergesivh mevg j ajmeivnwn. so you may know in your heart, and say to another, how good dealing is better by far than evil dealing Monteil’s examples are listed in Appendix 3A. Constructions following verbs of emotion also generally (in 4 out of 5 constructions) have line-initial wJ". Similar positions may be observed for completive wJ" in Iliad 9 and Odyssey 9. The sample is rather small, and more research would be needed to discover whether the same correlation holds throughout Homer. However, the existence of a similar, though less strong, tendency in tragic trimeters, noted below, suggests that the phenomenon has a structural as well as metrical component, which accords with the aetiology suggested in Chapter 5: that completives following cognitive verbs were originally interrogative, and that the development of full complementation involved a transitional stage of circumstantial-like complements, which are also likely to be structurally peripheral (since they require an explicit main verb object). 2b: Prosodic features of tragic complements with o{ti{{{ 2b (i): ”Oti in Sophocles In the total of 35 Sophoclean complements introduced by o{ti, the most notable feature is its prosodic regularity: it is in 32 constructions at the line end. Examples include Trach.439–40: 12) oujd j h{ti" ouj kavtoide tajnqrwvpwn o{ti caivrein pevfuken oujci toi'" aujtoi'" ajeiv. ...nor [am I] one who does not know of men that we cannot ourselves enjoy constant happiness and OT.59–60: 13) proshvlqeq j iJmeivronte": eu\ ga;r oi\d j o{ti nosei'te pavnte", kai; nosou'nte" wJ" ejgw; ....... for I know well that you are all sick, ...... 239 The form of the second construction, with the clause starting at the hephthemimeral caesura, is common in Sophoclean complementation (and eu\ ga;r oi\d j o{ti occurs also at OT.1133 and Ant.1043). However, the first example shows that the line and clause may coincide.15 Almost all of the complements are introduced by a cognitive rather than speech verb.16 Two of the constructions where o{ti is not line-final are in choral odes, so do not have the same metrical constraint. They are also structurally irregular: the complement clause is preposed at OT.499, and, at El.1070, o{ti begins a choral antistrophe, but depends on the verb bovason (or its object ojneivdh) in the previous strophe, so it functions as a cohesive link, and may have causal force: 14) katav moi bovason oijktra;n o[pa toi'" e[nerq j jAtreivdai" ajcovreuta fevrous j ojneivdh. o{ti sfi;n h[dh ta; me;n ejk dovmwn nosei'tai... (ant. a) carry below a piteous tale, to the Atreidae below carrying a reproach not to be danced. That/because now their house is sick... The iambic line-end position occurs also in all 7 complement constructions in Prometheus.17 All follow verbs of knowing, most of perception rather than judgment, and generally state a general truth or proverb, as Pr 104–5: 15) ........ gignwvskonq j o{ti{{{ tov th'" ajnavgkh" e[st j ajdhvriton sqevno". ... knowing that the strength of necessity is unconquerable The similarity of position in Prometheus and in Sophocles is noted by Griffith (1977: 192), who describes the line-end position as ‘a form of “Sophoclean” enjambement’, and attributes it to the difficulty of fitting o{ti into trimeters. 15Other full line clauses occur at Tr.904, Ant.61, 311, 779, Aj.678, El.332, 426, 1106, Phoen.325, 405, and OC.666. 16The two constructions introduced by speech verbs are at Ant.325 and El.1070. 17 Pr.104–5, 186–7, 259–60, 323–4, 328–9, 377–8, and 951–2. 240 However, it may be noted that this position is not universal for conjunctive o{ti in Sophoclean trimeters, though it is normal. At Ant.2–3, it is early in the line: 16) a\r j oi\sq j o{ ti Zeu;" tw'n ajp j Oijdivpou kakw'n oJpoi'on oujci; nw'/n e[ti zwvsain telei' Do you know which of the miseries from Oedipus Zeus does not accomplish equally on us still now alive? 18 There is evidently a metrical constraint on the position of o{ti, but there are reasons for thinking this is connected with the syntax. Griffith’s observation (1977: 192) that ‘Aeschylus chose not to use the word, whereas Prom. uses it with Sophoclean frequency and style’ is not a complete explanation, for the following reasons: 1) His assumption that this feature constitutes ‘enjambement’ begs the question of where the clause break occurs. The prosody accords with the functions of o{ti as focal and main verb object, as described in Chapter 5. 2) As is observed below in Section 2b (iii), completive o{ti is much rarer in Euripides, despite showing a greater variety of position. Euripides’ reluctance to use it is unlikely to be simply metrical (since the line-end position is available). 3) Not only o{ti, but also complementizing wJ", is rare in Aeschylus, as indeed the figures of Griffith (1977: 335) demonstrate.19 Even if the tendency for line- final placing of o{ti were entirely metrically constrained, alternative complementizers were available to the tragedians, as noted above in Section 1a. The fact that Aeschylus rarely uses any type of finite complementation is at least as significant as his reluctance to use one particular conjunction. The distinctive discourse function of Aeschylean complements is considered below in Section 3. 4) Sophoclean use of o{ti is almost always completive in trimeters,20 and always relative or causal (‘because’) in its rare appearance in other metres.21 This is in great contrast to Euripidean practice, which is described below in 18Lloyd-Jones and Wilson (1990: 183) suggest a syntactic break after 2, and a\, poi'on for oJpoi'on. 19Aeschylean complements with wJ" are discussed below, in Section 2b (ii). 20The possible exceptions (Ant.2, OT.71, Aj.513, and Tr.161) are all free relatives depending on cognitive verbs. 21Relatives occur at Tr.1278, El.155, Ph.849, OC.250; free relatives at Tr.1009, OT.486, Ph.210; and causal use at Ant.159, 1321, OT.1340, Ph.1022, 1165. 241 Section 2b (iii). A purely metrical interpretation does not explain these functional differences: the prosody also reflects the poets’ distinctive syntax. 2b (ii): ”Oti in Aeschylus As noted in the previous section, Aeschylus rarely uses finite complementation. However, the three instances of o{ti in extant Aeschylean texts suggest a link between position and meaning, as well as metre. Completive o{ti follows a speech verb at Eum.98–9, and this is line-final, in the canonical trimeter position: 17) ..... prounnevpw d j uJmi'n o{ti{{{ e[cw megivsthn aijtivan keivnwn u{po. ....... I declare to you that I am accused by them of the greatest crime 22 In a proleptic construction in anapaestic lines at Eum.970–1 (also cited in Chapter 6), the conjunction is line-initial, in the typical Homeric position following verbs of emotion: 18) .... stevrgw d j o[mmata Peiqou'" o{ti{{{ moi glw'ssan kai; stovm j ejpwvpa ... ... and I rejoice in the eye of Persuasion, that it was guiding my tongue and lips In a free relative construction, also in anapaests, at Ag.97–8, o{ ti is adjacent to the main verb, reflecting its function as object:23 19) touvtwn levxas j o{ ti{{{ kai; dunato;n kai; qevmi", ai[nei paiwvn te genou' Of these matters say what [is] possible and right, and consent and be healer There is evidently a metrical constraint on the line position of o{ti, but again the prosodic patterns accord with semantic type. 22Pragmatic and prosodic features of this construction are considered below. 23M and V have aijnei'n; F and Tr eijpei'n, either of which would make o{ti the object in both clauses. Denniston and Page (1957: 77) justify Wieseler’s conjecture dunatovn through the necessity for te to link to a preceding imperative. 242 2b (iii): ”Oti in Euripides A purely metrical explanation of line placing is usually invoked for the greater variability in the position of Euripidean o{ti, which correlates with a higher frequency of trimeter resolution.24 The position of the conjunction shows great prosodic variation, being line-final only 4 times.25 However, Euripidean use is also functionally distinctive, in three ways: 1) The most common use of o{ti (usually o{ ti) is as a relative, in 32 constructions. Of these instances, 8 are preposed (some with correlative elements),26 and 12 are free relatives, most (7) depending on speech verbs.27 2) A function intermediate between relative and interrogative, with a similar force to epic te, appears in three constructions with ejstiv, as at IA.525: 20) oujk e[st j jOdusseu;" o{ ti se; kajme; phmanei' Odysseus is not able to injure you and me. A similar, but impersonal, use occurs at IA.1453 (‘e[sq? o{ ti...Is it possible that...’), while a relative function appears at Or.418: 21) douleuvomen qeoi'", o{ ti pot j eijsi;n oiJ qeoiv. We are slaves to the gods, whoever the gods are. 3) The most common conjunctive use of o{ti (in 18 constructions) is causal (‘because’), as at Ba.944:28 22) ai[rein nin: aijnw' d j o{ti meqevsthka" frenw'n. [to lift it,] and I rejoice because/that you have changed your mind. The high frequency of causal use compared to the other tragedians accords with a late development of this meaning: as suggested in Chapter 5, Section 3, 24See West (1987: 25–26). 25At Cyc.421–2 and Med.560–1 (cognitive verbs); and Phoen.1617 and Ba.173 (speech verbs). 26Preposed relatives occur at Hipp.191 and 194, IT.20, 822 (a free relative), and 1137, Ba.430 and 881=901, IA.1014. 27Free relatives occur at Cyc.548, Hec.585, IT.760, Hel.822 and 1254, Phoen.1015, Or.150, Ba.492 and 506, IA.127, 129, and 652. 28The others are at Cyc.230, 553; Ion 831; IT.848, 1274; Hel.9, 186, 581, 1406; Or.395, 767; Ba.31, 245, 296, 616, 944; IA.506, 824. 243 the pronominal sense is likely to have been earlier.29 In some constructions, o{ti has no completive force at all, as at Cyc.229–230: 23) uJpo; tou'… tiv" ej" so;n kra't j ejpuvkteusen, gevron… uJpo; tw'nde, Kuvklwy, o{ti{{{ ta; s j oujk ei[wn fevrein. (Cyc.) By whom? Who has been boxing with your head, old man? (Sil.) By these men, Cyclops, because I would not let them take your things. The 13 non-causal complements depend on speech verbs rather than on cognitive verbs (8 as against 4).30 All constructions are cited in Appendix 3B. The rarity of completive o{ti is comparable to Aeschylean rather than to Sophoclean practice (where there are 34 constructions in a much smaller corpus), and it may be inferred that the freedom of Euripidean placing of o{ti has a semantic as well as a metrical component. Even if the explanation were (impossibly) prosodic but not structural, it could not be simply metrical, because line-final placing is not restricted to trimeters: the pattern occurs in over half the instances in Aristophanes, where 14 of the 24 instances of complementizing o{ti occur at the line end, and in Pindar, where the six instances of completive o{ti are always terminal;31 five following verbs of emotion, and one of speech, at N. 5.3–5: 24) stei'c j ajp j Aijgivna" diagevllois j, o{ti{{{ Lavmpwno" uiJo;" Puvqeva" eujrusqenhv" nivkh Nemeivoi" pagkrativou stevfanon [my song] sped from Aegina, announcing that the son of Lampon, strong Pytheas, won the crown for the Nemean pancratium The prosodic pause at the end of any stichic line is likely to be reflected in the syntax. In trimeters, the brevis in longo created by line-end o{ti32 standardly coincides with a sense-pause (see Stinton 1977a). On occasion, the break 29Aristophanes demonstrates both uses: 5 out of 15 instances in Frogs are causal. 30”Oti follows speech verbs at Ba.173, Ba.649, HF.1417, Or.8–10, Or.892–3, IT.1093–4, El.171, Hel.1493; cognitive verbs at Cyc.321 and 421–2, Phoen.1617, and Med.560–1, and once depends on plhvn (at El.1312). 31P. 2.31, P. 5.20, P. 10.69, N. 4.43, and N. 5.3; as against 16 non-completive instances, none of which is line-terminal. 32The terminology is discussed by Rossi (1963). 244 involves hiatus with the following line, as at Eum.98–9 (o{ti/e[cw), Prom.259–60 (o{ti/h{marte") and 377–8 (o{ti/ojrgh'"), and S.Ant.61–2 (o{ti/e[fumen).33 There is therefore an emphatic component to this position as a clause introducer. The regular occurrence of the expletive phrase oi\d? o{ti at the line end (OT.59, 1133, Ant.276, 1043, E.Phoen.1617, Ar.Ra.600) also shows that a sense-break can be quite normal here.34 The regular position of complementizing o{ti at the line end emphasizes its function in the main clause, as modifier or complement of the main verb. 2c: ÔW" in tragedy The position of the conjunction wJ" is more variable in the tragic texts than in Homer: initial position following speech verbs occurs in the single instance in the Oresteia (Eum.310–1), 6 out of the 7 constructions in other Aeschylean works, 4 out of 5 in Prom., and 6 out of 12 in OT, but in only 2 of the 6 constructions in Medea.35 However, the presence of main verb objects is very striking: every instance of complementing wJ" in the Oresteia, almost all in OT. and Crito, most instances in the Medea,36 and half of the Aeschylean instances outside the Oresteia are preceded by accusatives (see Appendix 3B). There is a somewhat higher frequency following main verbs of speech than those of knowing (the structural details are discussed in Chapter 6, Section 4). Though wJ" is not preceded by hiatus as often as o{ti is followed by it, hiatus does occur, sometimes also with brevis in longo, as at Ag.494–6 (tovde/wJ"), Choe.987–8 (pote;/wJ"), Pers.287–8 (pavra/wJ") and 356–7 (tavde/wJ"), OT.712–3 (a[po/wJ"), and Med.85–6 (tovde/wJ").37 Of the 23 completive clauses with wJ" in Aeschylus (approximately equal to the number of resultative and final constructions with wJ"), almost all have a proleptic or pronominal object, or depend on a nominal, and most have non- finite subordinate clauses (they are cited in Appendix 3B). Most Aeschylean 33Hiatus with o{ti occurs three times in Pindar too: at P. 5.20–1, P. 10.69–70, and N. 5.46–7. 34See Section 3c for the function of these phrases as ‘afterthoughts’. 35The single instance of wJ" following a speech verb in Cyclops, at 452, is also line-initial. 36As Med.248, 447, 452, 530, and 1405. 37Hiatus occurs also with eij, as at Choe.181–2 (tavde/eij) and Eum.466–7 (kardiva/ eijj). 245 constructions depend on cognitive verbs, and conjunction and verb are usually (as in Homer) in the same line, as Choe.1034–6: 25) kai; nu'n oJra'tev m jj jj, wJ"JJJ pareskeuasmevno" xu;n tw'/de qallw'/ kai; stevfei prosivxomai mesovmfalovn q j i{druma, Loxivou pevdon, And now look at me, how equipped with this branch and wreath I shall approach the shrine at the earth’s navel, the land of Loxias Eum.454: 26) gevno"vvv de; toujmo;nj ;j ;j ; wJ"JJJ e[cei peuvsh/ tavca. and my race, how it is, you will soon learn. and Choe.492: 27) mevmnhso d j ajmfivblhstronj vj vj v wJ"JJJ ejkaivnisa". remember how you first used/they devised a net 38 The only wJ"-completive in the Oresteia which follows a verb of speech is in (anapaestic) line-initial position, at Eum.310–1 (also cited in Chapter 6, Section 2b): 28) levxai te lavchvvv ta; kat j ajnqrwvpou" wJ"JJJ ejpinwma'/ stavsi" aJmhv. and to speak of the lots of men, how my band apportions them. There are 12 complements with wJ" in other Aeschylean works. Four depend on cognitive verbs (Pers.287–9, 525, 599–600, Sept.617), 7 on speech verbs (Supp. 390–1, Pers.356–7, 754, 819–20, Sept.375–6, 468–9, 922–3), and one on a verb of showing (Sept.176). Again, the position of the conjunction varies with the main verb: 6 of the 7 constructions dependent on speech verbs are line- initial, as against 1 of the 4 following cognitive verbs. Similarly in Prometheus, none of the four constructions following cognitive verbs (Prom.359, 442–4, 1002–3, 1093) is line-initial, but 4 of the 5 following speech predicators are (Prom.211–3, 296–7, 842–3, 889–90, 1073–5). 38Garvie (1986: 179), following Blomfield (1824), argues for this construction (rather than the dative w|/ in the Laurentian manuscript M, which would reflect oi|" in 491). 246 The positions of the conjunctions are more regular than authorial practice in the placing of other prepositives in the iambic line would suggest: West (1982: 83–4) notes great differences between the tragedians in the use of prepositives at the line end, and postpositives at the beginning:39 Aeschylus occasionally places prepositives at the line end,40 Euripides almost never does, while Sophocles uses enjambment very freely,41 even putting a postpositive at the line start, at Aj.986 (dh'ta, after several lines of ajntilabhv) and OT.1085 (pot j, though the previous line ends with e[ti, creating a metrical pause). 2d: Prosody and syntax These examples show that there is regularly a prosodic pause after o{ti and before wJ", and that there is also normally a sense-pause, even with enjambement. If prosody reflects syntactic organization, then o{ti would be placed in the main clause group, and wJ" in the subordinate. The frequent presence of main verb objects before wJ" accords with this interpretation. It was noted above in Section 1 that the meaning of the two complementizers is consistent with their etymologies as demonstrative and anaphoric (and consequent prospective or retrospective force respectively), and this accords with their positions with respect to the line break. The opposition of [o{ti / - ] and [ - / wJ" ] may therefore be interpreted as a metrical reflection of the focal patterns of the circumstantial construction. The pragmatic function of emphasis (‘I tell you this’) accords with the typical use of complementation in the corpus as a foregrounding technique,42 discussed in the next section. 3: Rhetorical functions of complementation Complementation is not only, or even primarily, indirect speech. In Homer, it is normally a perspectival technique (‘he did not know that...’; ‘Hector was angry that...’). In the tragic texts of the corpus, it is overwhelmingly used for rhetorical effect and persuasion. The rarity of reported speech is very striking: the change of the proportion of main verbs, from cognitive verbs in Homer to speech verbs in tragedy, discussed in Chapter 6, does not represent an 39These terms are used as distributional categories, as by West (1987: 9n2) and Dover (1960: 12–13). 40West cites Pers.486 ou|, Ag.1354 wJ", Choe.1005 mhv, and Eum.238 prov". 41As Ant.5 ouj, 171 kaiv, 324 mh;, 409 to;n, El.879 toi'", 1309 wj", OC.495 ejn, and 993 eij. 42‘Foregrounding’ translates the aktualisace of Mukarovsky: see Burbank and Steiner (1977). 247 increase of reported speech, but rather of constructions which assert or suggest the truth of a statement or the authority of the dramatic speaker (as with cognitive constructions like oi\da ga;r o{ti, eu\ i[sqi o{ti, dh'lon o{ti). Aeschylus in particular uses first person verbs of speech asseveratively, as in the performative construction at Eum.98–9 (also cited above, in Section 2b): 29) ..... prounnevpw d j uJmi'n o{ti{{{ e[cw megivsthn aijtivan keivnwn u{po. .... I declare to you that I am accused by them of the greatest crime Subordination adds greatly to the pragmatic complexity of this utterance, by giving it four arguments (‘I, you, them, the crime’), and creating a self-reflexive point of view, though adding nothing to the information expressed in the subordinate clause.43 The prominence of o{ti is increased by two prosodic features: hiatus with e[cw, and paromoiosis in the last metra (uJmi'n o{ti ... keivnwn u{po). Its emphasis suggests that o{ti may here be interpreted as ‘this fact’. Kiparsky and Kiparsky (1970) argue that such a meaning underlies factive constructions, and it is also claimed by asseverative ones.44 The low level of finite complementation in Aeschylus is offset by a use of ‘guarantee’ clauses as afterthoughts. This construction may be seen with first person verbs at Ag.269: h\ torw'" levgw… 625: ouj yeudh' levgw, and Choe.107: levxw, keleuvei" gavr, to;n ejk freno;" lovgon. Similar use is made of second person imperative verbs at Ag.680: tosau't j ajkouvsa" i[sqi tajlhqh' kluwvn, and Ag.1302: ajll j i[sqi tlhvmwn ou\s j ajp j eujtovlmou frenov". The difference with third person verbs is evident at Choe.527: ... wJ" aujth; levgei (wJ" may be used dismissively here), and, in anticipation, Eum.420: mavqoim j a[n, eij levgoi ti" ejmfanh' lovgon. The other tragedians use complementation less as asseveration than as an argumentative or emotionally expressive technique (constructions in the corpus are cited in Appendix 3B). Sophoclean complementation is often 43The performative use of language was first defined by Austin (1962), and the self-reflexive character of poetic language formulated as a linguistic principle by Jakobson (1958: 69). 44An analogous tragic use of participles depending on speech verbs to assert truth is noted by Fournier (1946). 248 argumentative, as in the pattern of anaphora mirroring the confrontation at OT.547–8: 30) Tou't j aujto' j j' j j' j j ; nu'n mou prw't j a[kouson wJ"JJJ ejrw'. Tou't j aujto' j j' j j' j j ; mhv moi fravz j o{pw"{{{ oujk ei\ kakov". (Cr.) This same thing, now first hear how I say it. (Oe.) Don’t tell me this same thing, that you are not evil. A self-referential, perhaps ironic, use of complementation appears in the double construction at OT.1366 (at the end of the second kommos, after Oedipus has blinded himself). The position of se in P2 after the complementizer, though it properly belongs in the following AI construction, is, presumably, phonetically motivated: 31) Oujk oi\d j o{pw"{{{ se fw' bebouleu'sqai kalw'". I do not know how/that I can say you have planned this well. Euripidean use is often exclamative, as at Med.1405: 32) Zeu', tavdvvv j ajkouvei" wJ"JJJ ajpelaunovmeq j, Zeus, do you hear this, how I am driven away... and Med.1060: 33) ou[toi pot j e[stai tou'q''' j o{pw"{{{ ejcqroi'" ejgw; pai'da" parhvsw tou;" ejmou;" kaqubrivsai. This will never be, that I shall leave my own children to be mocked by my enemies. Complementation in the Melian Dialogue is, not surprisingly, used as a persuasive technique: the shifts of viewpoint create the impression of common understanding, while in fact expressing a single argument about the justice of force. This is reflected in a predominance of cognitive verbs, which are generally used as assertions. The factive dhlovw, which is particularly common in Plato and Aristotle (see Section 3a below), appears twice. The extraordinary complexity of the passage at Thuc. 5.89 (part of which is cited in Chapter 3, Section 5) is increased by the embedding of the finite 249 complements inside a participle construction, itself juxtaposed to a double infinitive construction: 34) ÔHmei'" toivnun ou[te aujtoi; met j ojnomavtwn kalw'n, wJ"JJJ h] dikaivw" [to;n Mh'don kataluvsante"] a[rcomen h] ajdikouvmenoi nu'n ejpexercovmeqa, lovgwn mh'ko" a[piston parevxomen, [ou[q j uJma'" ] ajxiou'men [ [ h] o{ti {{{ Lakedaimonivwn a[poikoi o[nte" ouj xunestrateuvsate h] wJ"JJJ hJma'" oujde;n hjdikhvkate] levgonta" oi[esqai peivsein] We ourselves then will make use of no fair phrases, that either we rule justly because we overthrew the Persians, or come against you injured, offering lengthy but unbelievable arguments, nor do we think that [ you will think it possible to persuade us by saying [ either that being colonists of the Spartans you did not become allies, or that you have done us no wrong ] ] The style of this passage is the converse of the emphatic: the clauses are organized less contrastively, as the complementizers are less prominent, with no P2 particle following o{ti, and no evident focal elements before wJ". The juxtaposition of the complementizers with h[ and the use of negatives make the effect disjunctive, yet not prosodically contrastive.45 It was noted above (in Section 1b) that there is no evident difference here in the sense of o{ti and wJ", which need some emphasis to be semantically distinct. The structure is therefore more clearly completive. Negation is also common in Platonic complementation, typically involving questions, suggesting an evaluation of the subordinate clause, which may itself be negative. However, the prosody is more emphatic, with P2 focalizers, as may be seen at Crito 47a2–3: 35) oujc iJkanw'" dokei' soi levgesqai o{ti{{{ ouj pavsa" crh; ta;" dovxa" tw'n ajnqrwvpwn tima'n ajlla; ta;" mevn, Do you not think it is satisfactorily said that we ought not to honour all opinions, but only some... and Crito 50e3–4: 36) e[coi" a]n eijpei'n prw'ton me;n' ;' ;' ; wJ" oujci; hJmevtero" h\sqa kai; e[kgono" kai; dou'lo", aujtov" te kai; oiJ soi; provgonoi… Could you say to begin with that you were not our offspring and slave ... 45The sentence appears designed to be difficult to understand: a kind of cognitive bullying. 250 Other constructions from Crito are cited in Appendix 3C. The emphatic style is reinforced by the Platonic use as main verbs of peivqw (Symp.212b, Ap.27e) and ejpideivknumi (Ap.25c, Prot.320c), with both o{ti and wJ", noted by Neuberger-Donath (1982: 273). The use of preposed complementation also creates emphasis, as in the 4 constructions from Crito discussed below in Section 4. The low frequency of complementation in tragedy may be an effect of genre, because reporting is, presumably, less used in drama than other genres, and indirect speech forms might be expected to be more common in historical and forensic writing. No figures are available for the historians, but Monteil (1963: 399) observes a high level of ‘substantival’ constructions in Lysias. The frequency and interrogative form of complementation in Plato suggests it is one of the techniques by which the dialogic management of the Socratic elenchus is effected, and, in particular, a device by which one voice may appropriate the arguments of another, without departing from the dialogue form.46 3a: Structural implications of function In fact, all the tragedians use complementation to highlight a proposition by subordinating it: it might be described as a presentational frame.47 Pragmatic motivation may underlie a number of structural features, such as multiple embedding, verb person, the use of nominal predicators, and of preposed clauses. 46Relations between discourse structure and argument in Plato are discussed by Vlastos (1983), Kahn (1996), and Gill and McCabe (1996). The functions of P2 particles in Plato are discussed by Cook (1992), Sicking and van Ophuijsen (1993), Kip (1997), and Sicking (1997). 47Presentational frames are discussed by Goffman (1974), Tannen (1993), and Dik (1989: Chap. 10). 251 Double complements occur in Sophocles, at OT.1366 (cited in the previous section) and OT.1133ff., where h\mo" may be used as a subordinating conjunction:48 37) ..... eu\ ga;r oi\d j o{ti{{{ kavtoiden h\\ \\mo" to;n Kiqairw'no" tovpon oJ me;n diploi'si poimnivoi", ejgw; d j eJni; ‹ › ejplhsivazon tw/'de tajndri; trei'" o{lou" for I know well [that he knows [when about Kithairon, he with two flocks, and I one, ‹ › I was this man’s neighbour for three whole half-years ...] ] The shifting point of view created by multiple embedding is matched in a construction combining features of direct and indirect speech at Crito 51c6: 38) ÆSkovpei toivnun, w\ Swvkrate",Æ fai'en a]n i[sw" oiJ novmoi, Æ [eij hJmei'" tau'ta ajlhqh' levgomen,] o{ti ouj divkaia hJma'" ejpiceirei'" dra'n a} nu'n ejpiceirei'".Æ Observe then, Socrates, the laws might perhaps say, [if what we speak is the truth, ] that you do not justly try to do to us what you are now trying. Here, the structurally irregular position of the conditional creates a slight perspectival ambiguity (who or what are hJmei'"?). As most complement constructions have indicative verbs (in contrast with many of the free relatives discussed in Chapter 5, Section 2), the asseverative function of complementation is marked primarily by person rather than by mood or tense.49 Complements in the texts studied here show the use of first person constructions for rhetorical rather than narrative uses. Even a speech 48Though there may have been a conjunction in the lacuna following 1135: see Lloyd-Jones and Wilson (1990). 49The use of mood in expressing the attitudes and factivity of complementation is outside the scope of this discussion. See Lyons (1977: Chs. 16 and 17) for a theoretical overview, and Howorth (1955), Lightfoot (1971, 1979: 282–294), Boel (1980), Crespo (1984), and Sicking and Stork (1996) for surveys of Greek usage. The use of subjunctives and indicatives in relative clauses is considered by Vester (1989), and mood in conditionals is discussed by Bakker (1988b) and Horrocks (1995). A study of complementation modalities has been undertaken for English by Ransom (1986). 252 verb will pass on presuppositions when the main verb is in the first person: the performative use of the first person is common in Aeschylus (as cited above) and also in the Homeric eu[comai construction, as at Od.9.263: 39) laoi; d j jAtrei?dew jAgamevmnono" eujcovmeq j ei\nai We claim to be the people of Agamemnon son of Atreus. Constructions with second person verbs are, as might be expected, more confrontational, as in Sophoclean argumentation, in the constructions cited in the previous section. Both first and second person constructions are quite different from complements depending on third person verbs, which generally express no commitment to the truth of the subordinate clause. Complementation may therefore be considered as a way of mediating point of view as well as asserting truth.50 Dependence on a nominal or adjectival predicator becomes progressively more common. Monteil (1963) and Jespersen (1924) analyse these constructions as expressing the logical subject of the main verb, but recent models would consider them to be complements.51 They make a judgment, rather than simply reporting the subordinate proposition. The rare Homeric examples include a{li" o{ti at Il.5.349 and Il.23.670, and a{li" wJ" at Il.17.450 and Od.2.312 (see also Chapter 5, Section 3a). A complement with dh'lon appears once in Homer, at Od.20.333, with a pronominal object: 40) nu'n d j h[dh tovdevvv dh'lon, o{ t j oujketi novstimov" ejstin. But now this has become clear that there will never be a homecoming [for him]. In post-Homeric use, dh'lon occurs as a main clause idiom. It introduces complements three times in extant tragedy, all with wJ" and with omitted copula, at S.Phil.162–3: 41) Dh'lon''' e[moig j wJ"JJJ forbh'" creiva/ stivbon ojgmeuvei tovnde pevla" pou It [is] clear to me that he has gone to find food somewhere near here. E.Hipp.627: 42) touvtw/ de; dh'lon''' wJ"JJJ gunh; kako;n mevga. From this [is] clear how great a curse [is] woman. 50Relationships between focus, point of view and verb person in Latin are discussed by Fowler (1990) and Jones (1991). 51Following the analysis by Koster (1978) that such complements are base-generated within the VP, rather than being ‘extraposed’ from subject position as suggested by Jespersen (1924). 253 and at E.El.660: 43) ejlqou'sa mevntoi dh'lon wJ"' J' J' J ajpovllutai But if she comes home, it [is] clear that she will die. Prose use is much more extensive. Demosthenes uses dh'lon wJ" (5 instances) and dh'lon o{ti (32 instances), dh'lon w{ste (once, at Or. 21.162.6), and dh'lon diovti, in a preposed complement at Or. 58.42.1: 44) ... ejpeiv diovtivvv toujnantivon ejstivn, kai; oujc ou|to", ajll j ejgw; katestasivasmai, kai; fhsavntwn tinw'n moi sunagwniei'sqai prodevdomai dia; ta;" touvtwn eJtairevia", ejkeivnw" dh'lon''' uJmi'n e[stai For that it is the opposite case, and not he but I am libelled, and, though some have said they would aid me, I have been betrayed by their collaborators, [this] will be [made] clear to you in the following way. The predicator appears to be especially suited to philosophical argumentation (perhaps because it involves a value judgment): dh'lon occurs over 400 times in Plato, and over 1,500 times in Aristotle, introducing complements in many instances, most commonly with o{ti. As noted earlier, it is also employed as an afterthought (dh'lon o{ti) at Crito 53a3–4, and after a preposed complement with o{ti, at 53b1 (scedovn ti dh'lon). It appears in 4th century texts as the adverbial dhlonovti, occasionally in Plato and Demosthenes (8 instances), and rather more often in Aristotle (23 instances). The use of a nominal predicator may also be a consequence of the increasingly formalized and minimal main clause frame. 3b: Discourse function and clause order. Preposing As described in Chapter 5, clause order is crucial to the development of complementation, because, when the subordinate clause follows, the complementizer occupies a bivalent position. However, a small number of complements precede the main: in the corpus, there are 288 complement clauses (as shown in Appendix 1A), and of these 13 (under 5%) precede their main clause. Most are cited below.52 Preposing may be regarded as rhetorical hysteron proteron, with the order giving the subordinate clause greater prominence. 52The ones omitted are at Il.9.704 (wJ") and Eum.652 (pw'"). 254 The only Homeric instance of preposing with o{tti, at Od.23.115–6, has a correlative structure and a causal sense: 45) nu'n dæ o{tti{{{ rJupovw, kaka; de; croi÷ ei{mata ei|mai, tou[nekæ[ [[ ajtimavzei me kai; ou[ pwv fhsi to;n ei\nai. but now that I am dirty and wear foul clothing for this she dishonours me and says I am not her husband. Preposed tragic and Platonic complements are appositive rather than correlative. There is only one preposed complement with o{ti in tragedy, at OT.499–501: 46) ........ ajndrw'n d j o{ti{{{ mavnti" plevon h] jgw; fevretai, krivsi"vvv oujk e[stin ajlhqhv": but that [the word of] a seer, a man, weighs more than mine, the conclusion is not established In Crito, all four preposed complements are introduced by o{ti. Three are semantically subjects, two with dh'lon and one with mevnei. At 48b8, the subject of the subordinate is itself preposed: 47) [To; de; eu\ kai; kalw'" kai; dikaivw" ] o{ti{{{ taujtovn ejstin, mevnei h] ouj mevnei… And, the good, [living] both well and justly, that they are the same thing, does it hold or not? At 53b1 the complementizer, focalized by particles, signals the construction: 48) o{ti{{{ me;n ga;r kinduneuvsousiv gev sou oiJ ejpithvdeioi kai; aujtoi; feuvgein kai; sterhqh'nai th'" povlew" h] th;n oujsivan ajpolevsai, scedovn ti dh'lon''' : For that your friends will themselves risk banishment and loss of home or destruction of property, this [is] fairly clear. At Crito 53a3, the main clause is really an afterthought: 49) ou{tw soi diaferovntw" tw'n a[llwn jAqhnaivwn h[resken hJ povli" te kai; hJmei'" oiJ novmoi [ dh'lon''' o{ti{{{ : ] So much more than the other Athenians the city and we the laws pleased you, it [is] clear. 255 When the subordinate is the object of the main verb, it makes a greater call on the memory, as at Crito 53d7ff.: 50) o{ti{{{ de; gevrwn ajnhvr, smikrou' crovnou tw'/ bivw/ loipou' o[nto" wJ" to; eijkov", ejtovlmhsa" ou{tw glivscrw" ejpiqumei'n zh'n, novmou" tou;" megivstou" parabav", oujdei;" o}" ejrei'… But that, being an old man, having a short time to live, probably, you clung to life with such shameless greed, transgressing the greatest laws, will there be no-one who will say [this]? When wJ" is used in a preposed complement, it is also usually focalized by particles (as in the conditionals discussed in Chapter 4). There is only one correlative in the corpus, in the Melian Dialogue, 5.91.6ff.: 51) .......... wJ"JJJ de; ejp j wjfeliva/ te pavresmen th'" hJmetevra" ajrch'" kai; ejpi; swthriva/ nu'n tou;" lovgou" ejrou'men th'" uJmetevra" povlew", tau'ta''' dhlwvsomen, But how it is for the benefit of our empire that we are here, and for the safety of your city that we now propose to speak, this we shall make plain.... The main verb has a pronominal object at OT.1369–70, in a ditransitive construction (since the object is semantically indirect): 52) ÔW" me;n tavd j oujc w|d j e[st j a[rist j eijrgasmevna, mhv m j ejkdivdaske, mhde; sumbouvleu j e[ti. That/how these things have not been done in the best way, do not teach me, nor give me advice now Preposing emphasizes the causal force: Jebb (1981: 93) considers wJ" at OT.848 (and similar preposed constructions elsewhere in Sophocles) to be purely adverbial, as it ‘merely points to the mental attitude which the subject of ejpivstaso is to assume’. However, it is also conjunctive: 53) jAll j [ wJ"JJJ fanevn ge tou[po" w|d j] ejpivstaso, But how the word was thus set forth, you may understand, In a participial construction at Medea 1311, wJ" could be analysed as adverbial: 54) [wJ"JJJ oujkevt j o[ntwn ] sw'n tevknwn frovntize dhv. As being no longer alive, you must think of your children. 256 At Ag.1505–6, the construction could bear a substantival interpretation, because of the meaning of the gerund, but again a reinterpretation of the conjunction is required: 55) wJ"JJJ me;n ajnaivtio" ei\ tou'de fovnou tiv" oJ marturhvswn… how/that you are innocent of this murder who [will be] the witness? 3c: Complementizer meaning in preposed constructions As noted above, almost no post-Homeric constructions have a correlative element, in contrast with the persistence of correlative relatives in tragedy and in Plato, as cited in Chapter 5, Section 2a. In preposed complements, the complementizers bear a greater emphasis than they would when following the main clause, and are always focalized by P2 particles. The consequent emphasis of the whole clause is evident from the opening of the Apology, 17a1–2, where preposing implies a specifying interpretation, and by delaying the main clause, minimalizes the completive function: 56) ”O ti me;n uJmei'", w\ a[ndre" jAqhnai'oi, pepovnqate uJpo; tw'n ejmw'n kathgovrwn, oujk oi\da. In that (how) you have been affected by my accusers, men of Athens, I do not know. Without clues to the contrary, the conjunction may initially be interpreted in its pronominal sense (‘what you have suffered’). Such constructions may therefore create a level of irony, as ‘garden-path utterances’ (Sperber and Wilson 1986: 184), because they require re-interpretation when the main clause is heard. The semantic ambiguity of preposed constructions supports the assumption (discussed in Chapter 2, Section B 5a) that left-branching structures are more difficult to process, and, as the dependent clause is much larger than the governing clause, demonstrates the markedness of decreasing weight (at least in terms of constituents). This contrasts with ‘afterthought’ constructions, where o{ti may be interpreted as an emphatic main clause pronoun, as at S.Ant.276: 57) pavreimi d j a[kwn oujc eJkou'sin, oi\d j o{ti{{{ . and I am here unwilling to those who do not welcome me, I know that. 257 At E.Phoen.1616–7, both the main and subordinate clauses are minimal: 58) tiv" hJgemwvn moi podo;" oJmarthvsei tuflou'… h{d j h{ qanou'sa… zw'sav g j a]n savf j oi\d j o{ti{{{ . What guide will accompany the foot of the blind man? The dead girl? If she were alive, I know that well. At Ar.Ra.599–600, the conjunction is repeated, in what may be interpreted either as epanadiplosis, or with the first o{ti as a causal link, as in dialogue elsewhere (including Ra.20), with the repetition as an afterthought: 59) o{ti{{{ me;n ou\n, h]n crhsto;n h\/ ti, tau't j ajfairei'sqai pavlin peiravsetaiv m j eu\ oi\d j o{ti{{{ . that then, if anything good happened, he would want to take the clothes back from me, I know that well As with the interpretation of o{ti and wJ" in preposed complements, so the meaning of eij depends on its position. In a preposed indirect question, the conjunction will carry a conditional sense until the main clause is heard, as at E.Med.1103–4:53 60) e[ti d j ejk touvtwn ei[t[[[ j ejpi; flauvroi" ei[t[[[ j ejpi; crhstoi'" mocqou'si, tovdvvv j ejsti;n a[dhlon. And then of these, whether they toil upon worthless or whether on worthy objects, this is unclear. Conclusion This chapter discussed relationships between prosodic features and pragmatic function in complementation, showing that focalization is intimately associated with clause order and the prosody of the stichic line. Complementation is used in the corpus primarily as a presentational form rather than as indirect speech. A close interrelation between discourse function and focalization is evident, with point of view expressed through prosodically- as well as pragmatically-focal patterning. Only in the Melian Dialogue is complementation unemphatic: elsewhere in the corpus it retains focal prosody, implying that it also retains a circumstantial, adverbial, form. 53Of course, indirect questions may also be ambiguous when they follow speech verbs, as discussed in Chapter 4, Section 5b. 258 Conclusion Recapitulation Chapter 1 showed that subject, verb, and object order in epic and tragedy is statistically comparable to that observed in prose. No structural constraints on order were observed. The Greek clause cannot be regarded as verbicentric, because most finite clauses do not have both an explicit subject and object, and in those which do, SOV is more common than SVO. Chapter 2 showed that intra-clausal word order is motivated by phonetic weight rather than syntactic structure. Larger words preferentially come later in the sentence, motivating SV and OV, while VS and VO are associated with a prosodically prominent noun. The features are combined in a principle of ‘phonological weight.’ Chapter 3 discussed tragic phrasal tmesis, which demonstrates similarities between intra-clausal patterns and subordinated ones. It was treated by Aristophanes as typical of the tragic style. By its prosodic regularity, the construction highlights the form of the second colon, and so may be seen as exemplifying ‘weight to the right’. In Chapter 4, a prosodic definition of focus was proposed, and the P1>P2 sequence interpreted as a focal unit. Focalization and clause linking were analysed as the same, textually deictic, force. Focalization was also identified in relative and conditional clauses. Chapters 5–7 examined the development of complementation. The grammaticalization of o{ti involved its indefinite reference initially following speech verbs, and of wJ" its function as an adverbial in circumstantial constructions, following cognitive verbs. Because of the focal element, tragic complementation remained circumstantial, with adverbial features. The model of focalization accords with features of clause linking observed in conditional, relative, and complement constructions, and suggests a way in which a convergence of adverbial and relative structures could have developed into complements by a series of gradual steps. It also accords with the discourse functions observed in tragic and Platonic complementation. 259 In the Introduction, Section 3, it was noted that the hypothesis is testable, as it predicted that there would be no focus position in subordinate clauses which follow their main. Apparent counter-instances were discussed in Chapter 6, Section 4c, and it was concluded that they do not disprove the hypothesis. Principles of clause and word order The existence of regularities in word order has been taken to imply that underlying principles may be found by examining statistical frequencies: Dunn (1988: 78) considers that ‘Greek word order is not a matter of absolute laws, but is by its very nature a statistical phenomenon. It follows that Greek word order must be investigated by statistical methods.’ However, the findings here show that statistics do not in themselves explain variations. Nor, it appears, do purely pragmatic explanations. The similarity of word order in poetry and prose demonstrates that genre is not a significant determinant.1 Two structural factors must also be considered: rightwards phonological weight and initial focalization. The prosodic component of rightwards phonological weight also suggests a causal connection between the development of subordination and of a stress accent in Greek. Word order change Focalized linking also suggests an explanation for some historical changes in word order. While the ordering of subject, verb and object in early IE languages remains uncertain,2 it is usually thought that there was a transition in post-Classical Greek from free object and verb ordering to VO, and also from SV to VS. Suggested explanations of these changes have included ‘phonetic erosion’ (the need to preserve clarity when inflections are lost: Vennemann 1974), ‘gapping’ (when elements of the second clause in a co- ordinated structure are omitted: Ross 1970), or a general rightwards ‘leakage’, due to a global change in modifier and head ordering (Ross 1973, Lehmann 1973). 1Although, as noted in Chapter 1, Section B 4, there is a high proportion of VS in lyric passages in the Oresteia, it was also observed that this is restricted to main clauses, and is matched by the high VS level in main clauses in Crito. 2See Delbrück (1878), Watkins (1964), Lehmann (1973), Friedrich (1975), Miller (1975), and Taylor (1994). 260 This discussion has shown reason to doubt these explanations. In Chapter 2 it was shown that head-modifier ordering cannot be a complete explanation for word order, and, if preceding clauses are not structurally subordinated, it cannot be an explanation for changes in clause order either. Phonetic erosion is not applicable to CG, which retains its inflections. The structure of the subordinating link constitutes a more plausible motivation: complementation motivates VO order in the main clause, and may also motivate VS order within the subordinate clause, because P1 is syntactically in the main clause.3 The avoidance of centre-embedding in relativization also motivates VS in the main clause. It is clear that intra-clausal word order must be considered in the context of inter-clausal structure. Last words Throughout the discussion, a link between the stichic line and clause structure has been noted, with intonational and metrical boundaries reflecting the syntax. The study of ancient poetry therefore appears to be a valid approach in the study of linguistic structure. Precisely because stichic verse is prosodically very regular, underlying patterns may be the more evident. And we may expect to identify them in prose too. Poetry is not a derived form of language, but a ‘way of speaking’, a compositional technique which exploits the existing rhythms of speech by highlighting their natural regularities. This realization suggests that the answer to Monsieur Jourdain’s question cited in the epigraph of this work is not as obvious as might at first appear. The prosody of speech is at least as close to poetry as it is to prose, and poetic texts may provide our best evidence for the speech patterns of ancient times. 3 Kühner (1904: 597) notes the occurrence of VS order in complements, attributing it to the prominence of the elements, in almost the same words he uses to describe prolepsis: ‘wenn der Inhalt derselben gleichsam mehr vor das Auge gerückt werden soll’. 261 Appendix 1 Subject, verb, and object order 1A: Subject and verb order: the textual data Page 262 1B: Subject and verb order in subordinate clauses 267 1C: Questions in Oresteia stichomythia 275 1D: Subject and verb order with pronominal subjects 276 1E: Subordinates with pronominal subjects in the Oresteia and OT 278 1F: Subject and verb order with nominal subjects, collated by syllable number 279 1G: Object and verb order with nominal subjects in the Oresteia, Medea, and Crito, collated by syllable number 283 1H: VS in Crito 285 262 Appendix 1A The textual data: subject and verb order by main and subordinate clauses The whole corpus Total [SV+VS] clauses=2,437 Main [SV+VS] main clauses= 1,710 (70% of total clauses) Subordinate [SV+VS] clauses=727 SV total= 1,724 (71% of total SV+VS) SV main=1,224 (71% of total SV) (72% of main SV+VS) SV subordinate=500 (69% of subord. SV+VS) VS total=713 VS main=486 (68% of total VS) VS subordinate=227 The individual texts Iliad 9 (713 lines) has a high number of subjects (both SV and VS) in main rather than subordinate clauses. It has a higher proportion of VS than the Odyssey, in both main and subordinate clauses. Total [SV+VS] clauses=223 Main [SV+VS] clauses=155 (70% of total) Subordinate [SV+VS] clauses=68 SV total=154 (69% of total SV+VS) SV main=104 (68% of total SV) (67% of main SV+VS) SV subordinate=50 (74% of subord. SV+VS) VS total=69 VS main=51 (74% of total VS) VS subordinate=18 263 Odyssey 9 (566 lines) (a high proportion of SV and VS in main rather than subordinate clauses) Total [SV+VS] clauses=217 Main [SV+VS] main clauses=168 (77% of total) Subordinate [SV+VS] clauses=49 SV total=170 (78% of total SV+VS) SV main=131 (77% of total SV) (78% of main SV+VS) SV subordinate=39 (80% of subord. SV+VS) VS total=47 VS main=37 (79% of total VS) VS subordinate=10 Septem (1078 lines) (a high proportion of SV and VS in main rather than subordinate clauses) Total [SV+VS] clauses=245 Main [SV+VS] main clauses=194 (79%) Subordinate [SV+VS] clauses=51 SV total=175 (71% of total SV+VS) SV main=138 (79%) (71% of main SV+VS) SV subordinate=37 (73% of subord. SV+VS) VS total=70 VS main=56 (80%) VS subordinate=14 264 Oresteia (3,796 lines) All three plays of the Oresteia have a high proportion of VS subordinates, compared to SV subordinates. Main clauses are predominantly SV (68% of main clauses with subjects),1 while in subordinate clauses, SV/VS proportions are about equal. Total [SV+VS] clauses=691 Main [SV+VS] main clauses=513 (74%) Subordinate [SV+VS] clauses=178 SV total=465 (67% of total SV+VS) SV main=372 (80%) (73% of main SV+VS) SV subordinate=93 (52% of subord. SV+VS) VS total=226) VS main=141 (62%) VS subordinate=85 Agamemnon (1,673 lines) SV total=218 (70% of total SV+VS) SV main=171 (78%) (76% of main SV+VS) SV subordinate=47 (55% of subord. SV+VS) VS total=92 VS main=54 (59%) VS subordinate=38 Choephoroi (1,076 lines) SV total=125 (65% of total SV+VS) SV main=103 (82%) (67% of main SV+VS) SV subordinate=22 (55% of subord. SV+VS) VS total=68 VS main=50 (74%) VS subordinate=18 1372 of [SV 372+VS 141+SVS 32 = 545]. 265 Eumenides (1,047 lines) SV total=122 (65% of total SV+VS) SV main=98 (80%) (73% of main SV+VS) SV subordinate=24 (45% of subord. SV+VS) VS total=66 VS main=37 (56%) VS subordinate=29 OT. (1,530 lines) (a high proportion of both SV and VS in subordinates) Total [SV+VS] clauses=264 Main [SV+VS] main clauses=171 (65%) Subordinate [SV+VS] clauses=93 SV total=191 (72% of total SV+VS) SV main=122 (64%) (71% of main SV+VS) SV subordinate=69 (74% of subord. SV+VS) VS total=73 VS main=49 (67%) VS subordinate=24 Medea (1,419 lines) (a high proportion of VS in main clauses) Total [SV+VS] clauses=255 Main [SV+VS] main clauses=172 (67%) Subordinate [SV+VS] clauses=83 SV total=167 (65% of total SV+VS) SV main=108 (65%) (63% of main SV+VS) SV subordinate=59 (71% of subord. SV+VS) VS total=88 VS main=64 (73%) VS subordinate=24 266 Cyclops (709 lines) (a high proportion of SV and VS in main rather than subordinate clauses) Total [SV+VS] clauses=126 Main [SV+VS] main clauses=95 (75%) Subordinate [SV+VS] clauses=31 SV total=94 (75% of total SV+VS) SV main=72 (77%) (76% of main SV+VS) SV subordinate=22 (71% of subord. SV+VS) VS total=32 VS main=23 (72%) VS subordinate=9 Frogs (1,533 lines) Total [SV+VS] clauses=254 Main [SV+VS] main clauses=169 (67%) Subordinate [SV+VS] clauses=85 SV total=192 (76% of total SV+VS) SV main=130 (68%) (77% of main SV+VS) SV subordinate=62 (73% of subord. SV+VS) VS total=62 VS main=39 (63%) VS subordinate=23 267 Melian Dialogue (overwhelmingly SV) Total [SV+VS] clauses=42 Main [SV+VS] main clauses=23 (55%) Subordinate [SV+VS] clauses=19 SV total=37 (88% of total SV+VS) SV main=21 (57%) (91% of main SV+VS) SV subordinate=16 (84% of subord. SV+VS) VS total=5 VS main=2 (40%) VS subordinate=3 Crito (a high proportion of SV subordinates and of main clause VS, in contrast with the Oresteia) Total [SV+VS] clauses=120 Main [SV+VS] main clauses=50 (42%) Subordinate [SV+VS] clauses=70 SV total=79 (66% of total SV+VS) SV main=26 (33%) (52% of main SV+VS) SV subordinate=53 (76% of subord. SV+VS VS total=41 VS main=24 (59%) VS subordinate=17 Appendix 1B: Subject and verb order in subordinate clauses, analysed by type Iliad 9 (713 lines) Subordinates (1 per 3 lines) (239, 73=31% with subjects: SV 50, VS 18, SVS 5) [VS are 41% of SV] 268 47% Adverbials 113 (in descending order of frequency) SV 26, VS 15 (VS 58% of SV): gavr (29: SV 7, VS 4), ejpeiv (23: SV 1, VS 1), wJ" (17: SV 6, VS 3), o{te (14: SV4, VS 4), o[fra (8: SV2, VS 1), i{na (5: SV 1), eij" o{ (3: SV 1), ou{neka (3: SV 1), o{sson (3: VS 1), privn (2: SV 1, VS 1), oJppovte (2), tou[neka (1), o{ti (1: SV), o{tti (1), h/| (1), o{qi (1: SV). 32% Relatives (77: SV 15) 13% Conditionals (30: SV 9, VS 4) Complements (11: SV 1, VS 1) 8% Indirect questions (7: VS 1) Fearing (1) Omitting gavr clauses from subordinates: adverbials=84, SV 19, VS 11 (VS 58% of SV), compared to 57% VS for gavr clauses. Omitting all adverbials leaves other subordinates= 126, SV 25, VS 6 (VS 24% of SV). VS is therefore associated with adverbial clauses. Odyssey 9 (566 lines) Subordinates (1 per 3.7 lines) (150, 59=39% with subjects: SV 39, VS 11 SVS 9 [VS are 31% of SV] 57% Adverbials 85 (in descending order of frequency) SV 25, VS 10, SVS 1 (VS 40% of SV): gavr (20: SV 10, VS 4), ejpeiv (11: SV 1, VS 1), wJ" (10: SV 3, VS 2), o{te (9: SV 3), o[fra (9: SV1, VS1), h\mo" (8: SV 3, SVS 5), i{na (7: SV2, VS1), o{qi (2: SV1), e[peit∆ (1), eij" o{ (1: SV), o{ph/ (1), o{pph/ (1: SV), o{sa (1), ei|o" (1), oi|a (1: SV), ou{neka (1), o{sson (1). 25% Relatives (37: SV 5, VS 2) 9% Conditionals (14: SV 5) Indirect questions (10: SVS 2, VS 1) Final (2: SV 1) 9% Complements (1: SV) Fearing (1: SV) Omitting gavr clauses from subordinates: adverbials=65, SV 15, VS 6 (VS 40% of SV), compared to 40% VS for gavr clauses. 269 Omitting all adverbials leaves other subordinates= 65, SV 13, VS 3 (VS 23% of SV). As with Iliad 9, VS is associated with adverbial clauses. Septem (1084 lines) Subordinates (1 per 10 lines: a very low proportion) (108, 54=50% with subjects —high: SV 37, VS 14, SVS [VS are 36% of SV] 49% Adverbials 53 (in descending order of frequency) SV 25, VS 6, SVS 1 (VS 24%): gavr (27: SV 15, VS 4), wJ" (10: SV 4), ejpeiv (6: SV 2, VS 1), eu\te (2: SV), o{te (2: SVS 1), i{na (2), o{tan (1: SV ), o{mw" (1), mhv (1: VS), w{sper (1: SV). 29% Relatives (31: SV 3, VS 4: high VS) 12% Conditionals (13: SV 8) Complements (6: SV 3) 10 % Indirect questions (3: VS 2) Final (2: SV 1, VS 1) Omitting gavr clauses, adverbials=26, SV 10, VS 2 (VS 20% of SV), compared to 27% VS for gavr clauses. Omitting all adverbials leaves other subordinates= 55, SV 29, VS 12 (VS 41% of SV). In contrast with Homer, VS is associated with non-adverbial subordinates. Oresteia (3796 lines) Subordinate Totals (1 per 7.3 lines: a low proportion) Oresteia (519, 189=36% with subjects: SV 93, VS 85, SVS 11) [VS are 93% of SV]. VS is especially high for relatives and adverbials (non-gavrvvv ). 60% Adverbials (310: SV 68, VS 65, SVS 4) VS 97% 16% Relatives (84: SV 9, VS 9) VS 100% quite low number of relatives 14% Conditionals (75: SV 14, VS 11) VS 79% 10% Complements (21: SV 1, VS 2) Indirect questions (29: SV 7, VS 3) VS 63% 270 Omitting gavr clauses, adverbials=140 (27%), SV 21, VS 31 (VS 148% of SV), compared to 70% VS for gavr clauses (170=33%). Omitting all adverbials leaves other subordinates= 208, SV 31, VS 24 (VS 77% of SV). VS is associated particularly with adverbials, but not clauses introduced by gavr. Agamemnon (1673 lines) Subordinates (214, 1 per 7.8 lines, 42% with subjects: SV 49, VS 41) [VS are 84% of SV] 58% Adverbials 125 (in descending order of frequency) SV 32, VS 29, SVS 1 (VS 91%): gavr (69: SV 20, VS 17), wJ" (18: SV 4, VS 4), ejpeiv (15: SV 2, VS 3), o{tan (5: SV 2, VS 1), o{pw" (2: SV 1), eu\te (2: SV 1, VS 1), o{te (2: SVS 1, VS 1), mhv (2), eu\t∆ a[n (2: SV 1), eu\t∆ (1: VS), o{sonper (1: SV), ejpeivdh (1), o{ph/ (1), o{pa/ (1), e{w" (1: VS), o{poi (1), e[nqa (1). 18% Conditionals (39: SV 11, VS 6) 15% Relatives (33: SV 2, VS 4) 8% Complements (9: VS 1) Indirect questions (8: SV 4, VS 1) Omitting gavr clauses, adverbials=56, SV 12, VS 11 (VS 92% of SV), compared to 85% VS for gavr clauses. Omitting all adverbials leaves other subordinates= 87, SV 16, VS 11 (VS 69% of SV) Choephoroi (1076 lines) Subordinates (151, 1 per 7.1 lines, 28% with subjects: SV 22, VS 20) [VS are 91% of SV] (A lower proportion of subordinate clauses with subjects than Agamemnon or Eumenides) 62% Adverbials 93 (in descending order of frequency) SV 16, VS 16, SVS 2 (VS 100%): gavr (46: SV 10, VS 8), wJ" (15: SV 2, SVS 2, VS 1), ejpeiv (7: VS 1), o{pw" (5: SV 1, VS 2), w{sper (4: VS 1), o{tan (3: SV 1, VS 2), ejpeivdh (2: VS 1), h/| (2: SV), o{te (1), o[fra (1), e[nqa (1: VS), o{mw" (1), a{te (1), eu\t∆ (1), toigavr (1), e{w" (1), w{ste (1). 15% Conditionals (23: SV 3) 15% Relatives (22: SV 1, VS 2) 9% Indirect questions (9: SV 1) Complements (4: SV 1, VS 1) 271 Omitting gavr clauses, adverbials=47, SV 6, VS 8 (VS 133% of SV), compared to 80% VS for gavr clauses. Omitting all adverbials leaves other subordinates=59, SV 6, VS 3 (VS 50% of SV). Eumenides (1047 lines) Subordinates (154, 1 per 6.8 lines, 38% with subjects: SV 28, VS 30) [VS are 107% of SV] 60% Adverbials 92 (in descending order of frequency) SV 20, VS 20, SVS 1 (VS 100%): gavr (55: SV 17, VS 8), wJ" (10: VS 4), ejpeiv (5: SVS 1), o{pw" (4: SV 1, VS 1), ejpeivdh (2: VS), o{tan (2: SV 1, VS 1), o{pou (2), w{sper (1), e[nqen (1: VS), i{na (1), o[fra (1), ou{tw (1), o{te (2), o{mw" (1), o{qen (1: VS), e[st∆ a[n (1: SV), mhv (1: VS), ejx ou|te (1: VS). 19% Relatives (29: SV 6, VS 3 SVS 1) 8% Conditionals (13: VS 5) (High VS) 13% Indirect questions (12: SV 2, VS 2) Complements (8) Omitting gavr clauses, adverbials=37, SV 3, VS 12 (VS 400% of SV), compared to 47% VS for gavr clauses. Omitting all adverbials leaves other subordinates=62, SV 8, VS 10 (VS 125% of SV) OT. (1530 lines) Subordinates (1 per 3 lines) Great variety of conjunctions (34). (519, 111=21% with subjects: SV 69, VS 24, SVS 18) [VS are 38% of SV] 48% Adverbials 247 (in descending order of frequency) SV 42, VS 22 (Vs 52%): gavr (126: SV 23, VS 12), wJ" (27: SV 4, VS 2), e[nqa (13: SV 2, VS 1), ejpeiv (13: VS 2), i{na (7), o{pw" (7: SV 1, VS 1), w{st∆ (6: SV 1), o{tan (6: SV 1), w{sper (4: SV 1), privn (3: SV 1), o{q∆ (3: SV 1), ajf∆ w|n (3: SV 1), oi|a (3: SV 1), o{pou (2), o{mw" (2: VS 1), ejf∆ oi|" (2), uJf∆ ou| (2: VS 1), w|n ou{nek∆ (2), ejpeivdh (1), o[qounek∆ (1: VS), uJf∆ w|n (1: VS), ajnq∆ w|n (1: SV), par∆ ou| (1: SV), ejx ou| (1), w|n u{per (1), o{tou (1: SV), su;n ai|" (1), provsqen (1: SV), pot∆ (1), di∆ w|nper (1), eJw" (1), ou{tw (1), o{poiper (1), o{qen (1: SV). 272 25% Relatives (129: SV 21, VS 2) 16% Conditionals (84: SV 13, VS 1) Indirect questions (33: SV 1, VS 2) 12% Complements (20: SV 2, VS 1) Fearing (6: SV 1, VS 1). Omitting gavr clauses, adverbials=121, SV 19, VS 10 (VS 53% of SV), compared to 52% VS for gavr clauses. Omitting all adverbials leaves other subordinates=272, SV 38, VS 8 (VS 21% of SV). VS is associated with adverbials. Medea (1419 lines) Subordinates (1 per 3.9 lines) (361, 87=24% with subjects: SV 59, VS 24, SVS [VS are 38% of SV] 50% Adverbials 182 (in descending order of frequency) SV 41, VS 12 (VS 29%): gavr (95: SV 23, VS 5), wJ" (19: SV 6, VS 1), ejpeiv (17: SV 2, VS 1), o{tan (6: SV 3, VS 1), w{st∆ (4: SV 1), privn (4), o{mw" (4), o{pw" (4: SV 1, VS 1), toigavr (3), o{te (3: SV 1), h|/ (3: SV 1), oi|a (3), h{nik∆ (2), ejpeivdh (2: VS 1), i{n∆ (2: VS), ejnqa (1: SV), ou{tw (1), h]n mhv (1: SV), oi|on (1), ejx w|n (1: SV), w{sper (1), w{nper ou{nek∆ (1), o{ph/ (1), mhv (1), ou{nek∆ (1), h{ (1). 24% Relatives (86: SV 7, VS 1) 12% Conditionals (45: SV 11, VS 7) (conditionals are the main reason for high VS level) Indirect questions (27: SVS 1, VS 1) 14% Complements (16: SV 4, VS 2) Fearing (7: VS 1). Omitting gavr clauses, adverbials=87, SV 18, VS 7 (VS 39% of SV), compared to 22% VS for gavr clauses. Omitting all adverbials leaves other subordinates=181, SV 22, VS 12 (VS 55% of SV). VS is associated with non-adverbial subordinates. 273 Cyclops (709 lines) Subordinates (1 per 5 lines) (137, 31=23% with subjects: SV 22, VS 9, SVS 0) [VS are 43% of SV] 51% Adverbials 70 (in descending order of frequency) SV 13, VS 2, SVS 1 (VS 15%): wJ" (24: SV 5, VS 3), gavr (17: SV 4), ejpeiv (8: SV 1), i{na (5: SVS 1), o{tan (4: SV 1), ejpeivdh (1), o{pw" (1), povt∆ (1), w{ste (1), mhv (1), e[ste (1: SV), eij gavr (1), privn (1), wJseiv (1: SV), o{ti (1), o{tih (1), o{pou (1: VS). 26% Relatives (35: SV 2, VS 4) 12% Conditionals (16: SV 3) 12% Indirect questions (12: SV 1) Complements (4: SV 2, VS 1) Omitting gavr clauses, adverbials=53, SV 9, VS 2 (VS 22% of SV), compared to 60% VS for gavr clauses. Omitting all adverbials leaves other subordinates=67, SV 8, VS 6 (VS 75% of SV). As in Medea, VS is associated with non-adverbial subordinates. Frogs (1533 lines) (Very even percentages) Subordinates (1 per 4 lines), with 2nd greatest variety of conjunctions (28). (380, 86=23% with subjects: SV 62, VS 23, SVS [VS are 36% of SV] 52% Adverbials 199 (in descending order of frequency) SV 42, VS 14, SVS 2 (VS 33%): gavr (95: SV 19, VS 6), i{na (19: SV 4, SVS 1, VS 1), wJ" (16: SV 6, SVS 1, VS 2), o{tan (9: SV 2), hJnivk∆ (7: SV 1), ejpeivdh (7: SV 2), o{pw" (6: SV 1), o{te (5: SV 1, VS 1), oJtihv (4), w{ste (3), privn (3: SV 1), ejpeiv (3), e{w" (2: VS 1), ejpeidavn (2: VS 1), o{mw" (2: SV 1), ou| (2: VS 1), o{ti (2: SV 1), w|nper e{neka (2), o{per (1), pw" (1), w{sper (1: SV), o{ph/ (1), oi| (1), ou|per (1: VS), h/|per (1), o{qen (1), oJpovtan (1), oJpovson (1: SV). 22% Relatives (83: SV 12, VS 8, SVS 1) 15% Conditionals (58: SV 7, VS 1) Indirect questions (24: SV 3) 11% Complements (14: SV 2, VS 1) Fearing (2: SV 1) 274 Omitting gavr clauses, adverbials=104, SV 22, VS 8 (VS 36% of SV), compared to 30% VS for gavr clauses. Omitting all adverbials leaves other subordinates=181, SV 25, VS 10 (VS 40% of SV). VS is not associated with any one type of subordinate clause. Melian Dialogue (about 200 lines) Subordinates (1 per 2.7 lines) (73, 19=26% with subjects: SV 16, VS 3, SVS 0) [VS are 24% of SV] 37% Adverbials 27 (in descending order of frequency) SV 9, VS 2 (VS 22%): gavr (12: SV 3, VS 1), wJ" (4: SV 1), ejpeivdh (2: SV), o{ti (2: SV 1), o{sw/ (2), ejpeivdan (1: VS), o{pw" (1: SV), w|/ (1), w{sper (1: SV), o{tan (1). 25% Relatives (18: SV 5) 22% Conditionals (16: SV 2, VS 1) 16% Complements (11: SV 1, VS 1) Indirect questions (1) Omitting gavr clauses, adverbials=15, SV 6, VS 1 (VS 17% of SV), compared to 33% VS for gavr clauses. Omitting all adverbials leaves other subordinates=46, SV 8, VS 2 (VS 25% of SV). Crito (about 560 lines) Subordinates (1 per 2.7 lines) (207, 70=34% with subjects: SV 53, VS 17, SVS 0) [VS are 32% of SV] 27% Adverbials 56 (in descending order of frequency) SV 15, VS 4 (VS 27%): gavr (17: SV 6, VS 4), wJ" (16: SV 5), i{na (6), ejpeivdh (5: SV 3), w{sper (3: SV 1), o{poi (3), w{ste (2), h|/ (2), o{mw" (1), ejpeivdan (1). Much lower than tragedy, and fewer conjunctions. 28% Relatives (58: SV 18, VS 6) 25% Conditionals (52: SV 12, VS 4) Complements (30: SV 7, VS 3) 20% Indirect questions (8) (Much higher than tragedy) 275 Fearing (3: SV 1) Omitting gavr clauses, adverbials=39, SV 9, VS 0 (VS 0% of SV), compared to 67% VS for gavr clauses. Omitting all adverbials leaves other subordinates=151, SV 38, VS 13 (VS 34% of SV). Appendix 1C: Questions in Oresteia stichomythia SV (not including wh-subjects) Ag.626: povteron ajnacqei;" Ê ejmfanw'" Ê ejx ∆Ilivou, h] cei'ma, koino;n a[cqo", h{rpase stratou'… Ag.630: povtera ga;r aujtou' zw'nto" h] teqnhkovto" favti" pro;" a[llwn nautivlwn ejklhv/zeto… Ag.942: h\ kai; su; nivkhn thvnde dhvrio" tivei"… Ag.1198: kai; pw'" a]n o{rkou ph'gma gennaivw" page;n paiwvnion gevnoito… ............. Ag.1204: mw'n kai; qeov" per iJmevrw/ peplhgmevno"… Ag.1251: tivno" pro;" ajndro;" tou't∆ a[co" porsuvnetai… Ag.1286: tiv dh't∆ ejgw; kavtoikto" w|d∆ ajnastevnw… Ag.1542: h\ su; tovd∆ e[rxai tlhvsh/, ....... Ag.1643–4: tiv dh; to;n a[ndra tovnd∆ ajpo; yuch'" kakh'" oujk aujto;" hjnavrize", ajlla; su;n gunhv, Choe.179: kai;; pw'" ejkei'no" deu'r∆ ejtovlmhsen molei'n… Choe.394–5: kai; povt∆ a]n ajmfiqalhv" Zeu;" ejpi; cei'ra bavloi, Choe.532: kai; pw'" a[trwton ou\qar h\n uJpo; stuvgou"… Choe.775: ajll∆ eij tropaivan Zeu;" kakw'n qhvsei potev… Eum.717–718: h\ kai; pathvr ti sfavlletai bouleumavtwn prwtoktovnoisi prostropai'" ∆Ixivono"… Eum.744: w\ Foi'b∆ “Apollon, pw'" ajgw;n kriqhvsetai… OSV Choe.224: wJ" o[nt∆ ∆Orevsthn gavr s∆ ejgw; prossennevpw… 775: ajll∆ eij tropaivan Zeu;" kakw'n qhvsei potev… 276 VS Ag.272: e[sti tw'ndev soi tevkmar… 276: ajll∆ h\ s∆ ejpivanevn ti" a[ptero" favti"… 278: poivou crovnou de; kai; pepovrqhtai povli"… 935: tiv d∆ a]n dokei' soi Privamo", eij tavd∆ h[nusen… Choe.177: mw'n ou\n ∆Orevstou kruvbda dw'ron h\n tovde… 528: kai; poi' teleuta'/ kai; karanou'tai lovgo"… Non-finite VS 545: poqei'n poqou'nta thvnde gh'n strato;n levgei"… SVS Ag.1306: tiv d∆ ejsti; crh'ma… tiv" sj ajpostrevfei fovbo"… 280: kai; tiv" tovd∆ ejxivkoit∆ a]n ajggevlwn tavco"… 547: povqen to; duvsfron tou't∆ ejph'n stuvgo" stratw'/… Eum.125: tiv soi tevtaktai pra'gma plh;n teuvcein kakav… Appendix 1D: The data for pronominal subjects, collated by main and subordinate clauses Iliad 9 (67 pronominal subjects out of 223 SV+VS=30%) Main pron 49 [SV 47, VS 2] (out of a total of 155 SV+VS main subjects = 32%) Subordinate pron 18 [SV 17, VS 1] (out of a total of 68 SV+VS subordinate subjects = 26%) Odyssey 9 (95 pronominal subjects out of 218 SV+VS=44%) Main pron 80, all SV (out of a total of 168 SV+VS main subjects = 48%) Subordinate pron 15 [SV 14, VS 1] (out of a total of 50 SV+VS subordinate subjects = 30%) Septem (61 pronominal subjects out of 245 SV+VS=25%) Main pron 53 [SV 48, VS 5] (out of a total of 194 SV+VS main subjects = 27%) Subordinate pron 8 [SV 6, VS 2] (out of a total of 51 SV+VS subordinate subjects = 16%) 277 Oresteia (182 pronominal subjects out of 691 SV+VS=26%) Main pron 153 [SV 141, VS 12] (out of 513 SV+VS main subjects = 30%) Subordinate pron 29 [SV 21, VS 8] (of a total of 178 SV+VS subordinate subjects = 16%) OT. (124 pronominal subjects out of 264 SV+VS=47%) Main pron 75 [SV 61, VS 14] (out of 171 SV+VS main subjects = 47%) Subordinate pron 49 [SV 41, VS 8] (out of 93 SV+VS subordinate subjects = 53%) Medea (88 pronominal subjects out of 255 SV+VS=35%): Main pron 58 [SV 47, VS 11] (out of 172 SV+VS main subjects =34%) Subordinate pron 30 [SV 25, VS 5] (out of 83 SV+VS subordinate subjects = 36%) Cyclops (66 pronominal subjects out of 126 SV+VS=52%) Main pron 51 [SV 44, VS 7] (out of a total of 95 SV+VS main subjects = 54%) Subordinate pron 15 [SV 10, VS 5] (out of a total of 31 SV+VS subordinate subjects = 48%) Frogs (138 pronominal subjects out of 254 SV+VS=54%) Main pron 103 [SV 86, VS 17] (out of a total of 169 SV+VS main subjects =61%) Subordinate pron 35 [SV 26, VS 9] (out of a total of 85 SV+VS subordinate subjects = 41%) Melian Dialogue (18 pronominal subjects out of 42 SV+VS=43%) Main pron 11 [SV 11] (out of a total of 23 SV+VS main subjects = 48%) Subordinate pron 7 [SV 7] (out of a total of 19 SV+VS subordinate subjects =27%) 278 Crito (52 pronominal subjects out of 120 SV+VS=43%) Main pron 18 [SV 11, VS 7] (out of 50 SV+VS main subjects =36%) Subordinate pron 34 [SV 29, VS 5] (out of 70 SV+VS subordinate subjects =49%) Appendix 1E: Subordinate clauses with pronominal subjects in the Oresteia and OT Oresteia Ag.261 (SV suv), 423 (SV ti"), 563 (VS ti"), 671 (SV ti"), 934 (SV ti"), 944 (VS tau'ta), 953 (SV oujdeiv"), 1088 (SV suv), 1205 (SV pa'" ti"), 1403 (SV suv), 1433 (VS ejgwv), 1523 (SV ou|to"), 1630 (SV oJ), 1661 (SV ti"), Choe.192 (SV ejgwv), 224 (SV ejgwv), 265 (SV ti"), 267 (VS ejgwv), 527 (SV aujthv), 580 (VS tavde), 637 (VS ou[ti"), 853 (SV aujtov"), 897 (SV suv), 988 (SV ejgwv), Eum.420 (VS ti"), 457 (SV suv), 484 (SV ejgwv), 622 (VS suv), 804 (SV ejgwv). OT. 22 (SV aujtov"), 117 (SV ti"), 138 (SV aujtov"), 148 (SV o{de), 171 (SV ti"), 219 (SV ejgwv), 237 (SV ejgwv), 258 (SV ejgwv), 264 (SV ejgwv), 274 (SV tavde), 285 (SV ti"), 297 (SV oi{de), 339 (SV ti"), 340 (SV suv), 341 (VS aujtav, SV ejgwv), 358 (SV suv), 432 (SV suv), 445 (SV suv), 500 (SV ejgwv), 591 (SV aujtov"), 618 (SV ti"), 725 (VS suv), 731 (VS tau'ta), 786 (VS tau'ta), 799 (SV suv), 884 (SV ti"), 966 (SV ejgwv), 979 (VS ti"), 1058 (SV ejgwv), 1062 (SV suv, SV ejgwv), 1086 (SV ejgwv), 1156 (SV ou|to"), 1173 (VS h{de), 1179 (SV aujtov"), 1181 (VS ou|to"), 1209 (SV auJtov"), 1247 (VS aujtov"), 1298 (SV ejgwv), 1336 (SV ejgwv), 1361 (SV aujtov"), 1371 (SV ejgwv), 1464 (SV ejgwv), 1476 (SV ejgwv), 1485 (SV aujtov"), 1498 (SV aujtov"), 1499 (SV aujtov"), 1526 (SV ti"). 279 Appendix 1F: Subject and verb order with nominal subjects, collated by syllable number Notes 1) Only the subject noun, rather than the phrase (NP), is considered. 2) A consideration of complete phrases may change the figures in two ways: a) a descending VS order may become an ascending one: marked as [x asc.ph]. b) an ascending SV order may become a descending order: marked as [x desc.ph]. 3) Other abbreviations used: a) [x dis] = descending VS (or, more rarely, SV) in which the second element is a disyllabic positioned at the line end (or occasionally at the penthemimeral caesura). b) [x names] = descending SV order in which the subject is a name. c) [x + appos] = descending VS with a following appositional phrase The data Iliad 9 Main pron 49 [SV 47, VS 2] 106 nominal subjects (out of a total of 155 SV+VS main subjects = 68%) Total SV 57, asc. SV 34 [1 des.ph], desc. SV 15 [2 dis., 8 names], const. SV 8 Total VS 49, asc. VS 24, desc. VS 17 [7 dis.], const. VS 8 Subordinate pron 18 [SV 17, VS 1] 50 nominal subjects (out of 68 SV+VS subordinate subjects = 74%) Total SV 33, asc. SV 14, desc. SV 10 [2 dis, 4 names], const. SV 9 Total VS 17, asc. VS 4, desc. VS 8 [2 dis], const. VS 5 Odyssey 9 Main pron 80, all SV 88 nominal subjects (out of a total of 168 SV+VS main subjects = 52%) Total SV 51, asc. SV 28 [8 des.ph], desc. SV 12 [1 dis, 1 name], const. SV 11 Total VS 37, asc. VS 5, desc. VS 13 [5 dis, 5 asc.ph], const. VS 19 Subordinate pron 15 [SV 14, VS 1] 35 nominal subjects (out of 50 SV+VS subordinate subjects = 70%) Total SV 25, asc. SV 14, desc. SV 1, const. SV 10 Total VS 10, asc. VS 0, desc. VS 6 [2 dis, 4 asc. phr], const. VS 4 280 Septem Main pron 53 [SV 48, VS 5] 141 nominal subjects (out of 194 SV+VS main subjects = 73%) Total SV 90, asc. SV 52 [7 des.ph], desc. SV 19 [6 dis, 4 names, 1 asc.ph], const. SV 19 Total VS 51, asc. VS 12, desc. VS 19 [7 dis, 3 asc. ph], const. VS 20 Subordinate pron 8 [SV 6, VS 2] 43 nominal subjects (out of 51 SV+VS subordinate subjects = 84%) Total SV 31, asc. SV 19, desc. SV 4 [2 dis], const. SV 8 Total VS 12, asc. VS 1, desc. VS 3 [3 dis], const. VS 8 Oresteia Main pron 153 [SV 141, VS 12] 360 nominal subjects (out of 513 SV+VS main subjects = 70%) Total SV 231, asc. SV 139 [30 des.ph], desc. SV 44 [9 dis, 3 asc.ph, 11 names], const. SV 48 Total VS 129, asc. VS 55, desc. VS 53 [32 dis, 12 appos/metric], const. VS 21 Subordinate pron 29 [SV 21, VS 8] 149 nominal subjects (out of 178 SV+VS subordinate subjects = 84%) Total SV 72, asc. SV 35, desc. SV 21, const. SV 16 Total VS 77, asc. VS 33, desc. VS 32 [21 dis], const. VS 12 Ag. Main pron 58 [SV 56, VS 2] 179 nominal subjects (out of 237 main subjects = 76%) Total SV 115, asc. SV 69 [19 des.ph], desc. SV 25 [2 dis, 3 asc.ph, 6 names], const. SV 21 Total VS 52, asc. VS 24, desc. VS 19 [11 dis, 4 + appos], const. VS 9 Subordinate pron 14 [SV 11, VS 3] 74 nominal subjects (out of 88 explicit subordinate subjects = 84%) Total SV 36, asc. SV 19, desc. SV 7, const. SV 10 Total VS 35, asc. VS 15, desc. VS 17 [10 dis], const. VS 3 281 Choe. Main pron 43 [SV 37, VS 6] 120 nominal subjects (out of 163 main subjects = 74%) Total SV 66, asc. SV 34 [6 des.ph], desc. SV 11 [4 dis, 2 names], const. SV 21 Total VS 44, asc. VS 19, desc. VS 18 [5 dis, 7 metric], const. VS 7 Subordinate pron 10 [SV 7, VS 3] 33 nominal subjects (out of 43 explicit subordinate subjects = 77%) Total SV 15, asc. SV 9, desc. SV 3, const. SV 3 Total VS 15, asc. VS 5, desc. VS 8 [5 dis], const. VS 2 Eum. Main pron 52 [SV 48, VS 4] 92 nominal subjects (out of 144 main subjects = 64%) Total SV 50, asc. SV 36 [5 des.ph], desc. SV 8 [3 dis, 3 names], const. SV 6 Total VS 33, asc. VS 12, desc. VS 16 [8 dis, 4 + appos], const. VS 5 Subordinate pron 5 [SV 3, VS 2] 53 nominal subjects (out of 58 explicit subordinate subjects =91%) Total SV 21, asc. SV 7, desc. SV 11, const. SV 3 Total VS 27, asc. VS 13, desc. VS 7 [6 dis], const. VS 7 OT. Main pron 75 [SV 61, VS 14] 91 nominal subjects (out of 171 SV+VS main subjects = 53%) Total SV 61, asc. SV 38 [5 des.ph], desc. SV 15 [1 dis, 1 asc. ph, 1 name], const. SV 8 Total VS 35, asc. VS 16, desc. VS 13 [3 dis, 2 asc.ph], const. VS 6 Subordinate pron 49 [SV 41, VS 8] 44 nominal subjects (out of 93 SV+VS subordinate subjects = 47%) Total SV 28, asc. SV 16, desc. SV 5 [1 dis, 1 name], const. SV 7 Total VS 16, asc. VS 9, desc. VS 4 [2 dis], const. VS 3 282 Medea Main pron 58 [SV 47, VS 11] 114 nominal subjects (out of 172 SV+VS main subjects =66 %) Total SV 61, asc. SV 36 [9 des.ph], desc. SV 11 [3 names] , const. SV 14 Total VS 53, asc. VS 23, desc. VS 23 [14 dis, 2 asc. ph], const. VS 7 Subordinate pron 30 [SV 25, VS 5] 53 nominal subjects (out of 83 SV+VS subordinate subjects = 64%) Total SV 34, asc. SV 22, desc. SV 2, const. SV 10 Total VS 19, asc. VS 5, desc. VS 8 [6 dis, 1 asc.ph], const. VS 6 Cyclops Main pron 51 [SV 44, VS 7] 44 nominal subjects (out of a total of 95 SV+VS main subjects = 46%) Total SV 27, asc. SV 15 [2 des.ph], desc. SV 4 [2 dis], const. SV 8 Total VS 17, asc. VS 4, desc. VS 10 [4 dis, 3 asc.ph], const. VS 3 Subordinate pron 15 [SV 10, VS 5] 16 nominal subjects (out of 31 SV+VS subordinate subjects = 52%) Total SV 12, asc. SV 7, desc. SV 2 [2 dis, 1 name], const. SV 3 Total VS 4, asc. VS 2, desc. VS 1 [1 equ.ph], const. VS 1 Frogs Main pron 103 [SV 86, VS 17] 66 nominal subjects (out of a total of 169 SV+VS main subjects =39%) Total SV 44, asc. SV 24 [7 des.ph], desc. SV 8 [5 names], const. SV 12 Total VS 22, asc. VS 8, desc. VS 7 [2 dis, 4 asc.ph], const. VS 7 Subordinate pron 35 [SV 26, VS 9] 50 nominal subjects (out of 85 SV+VS subordinate subjects = 59%) Total SV 36, asc. SV 16 [3 des.ph], desc. SV 8 [2 dis, 2 names], const. SV 12 Total VS 14, asc. VS 10, desc. VS 2 [1 dis, 1 asc.ph], const. VS 2 Melian Dialogue Main pron 11 [SV 11] 12 nominal subjects (out of 23 SV+VS main subjects = 52%) Total SV 10, asc. SV 4, desc. SV 2 [1 name], const. SV 4 Total VS 2, asc. VS 1, desc. VS 0, const. VS 1 Subordinate pron 7 [SV 7] 12 nominal subjects (out of 19 SV+VS subordinate subjects = 63%) 283 Total SV 9, asc. SV 4, desc. SV 3 [3 names], const. SV 2 Total VS 3, asc. VS 2, desc. VS 0, const. VS 1 Crito Main pron 18 [SV 11, VS 7] 32 nominal subjects (out of 50 SV+VS main subjects = 64%) Total SV 15, asc. SV 9, desc. SV 4, const. SV 2 Total VS 17, asc. VS 7, desc. VS 3, const. VS 7 Subordinate pron 34 [SV 29, VS 5] 36 nominal subjects (out of 70 SV+VS subordinate subjects = 51%) Total SV 24, asc. SV 10, desc. SV 7, const. SV 6 Total VS 12, asc. VS 3, desc. VS 5 [3 with appos.NP], const. VS 4 Appendix 1G: Object and verb order with nominal subjects in the Oresteia, Medea, and Crito, collated by syllable number Ag. Main 140 nominal objects (out of 184 main objects =76%) Total OV 77, asc. OV 40 [11 des.ph], desc. OV 13, const. OV 24 Total VO 63, asc. VO 9, desc. VO 35 [19 dis, 17 asc.ph], const. VO 19 [pronominals 44 Total OV 35, asc. OV 31, desc. OV 3 const. OV 1 Total VO 9, asc. VO 0, desc. VO 6 [1 dis], const. VO 3] Subordinate 32 nominal objects (out of 41 subordinate objects =78%) Total OV 20, asc. OV 13 [3 des.ph], desc. OV 2, const. OV 5 Total VO 12, asc. VO 0, desc. VO 10 [7 dis, 6 asc.ph], const. VO 2 [pronominals 9 Total OV 8, asc. OV 7, desc. OV 0, const. OV 1 Total VO 1, asc. VO 0, desc. VO 0, const. VO 1] Choe. Main 88 nominal objects (out of 125 main objects =70%) Total OV 35, asc. OV 15 [4 des.ph], desc. OV 7 [1 dis], const. OV 13 284 Total VO 53, asc. VO 10, desc. VO 22 [9 dis, 8 asc.ph], const. VO 21 [pronominals 37 Total OV 28, asc. OV 25, desc. OV 0, const. OV 3 Total VO 9, asc. VO 0, desc. VO 7 [1 dis], const. VO 2] Subordinate 21 nominal objects (out of 26 subordinate objects =81%) Total OV 6, asc. OV 5, desc. OV 0, const. OV 1 Total VO 15, asc. VO 1, desc. VO 5 [4 dis], const. VO 9 [pronominals 5 Total OV 5, asc. OV 4, desc. OV 0, const. OV 1 Total VO 0, asc. VO 0, desc. VO 0, const. VO 0] Eum. Main 121 nominal objects (out of 152 main objects =80%) Total OV 68, asc. OV 38 [10 des.ph], desc. OV 10 [3 dis], const. OV 20 Total VO 53, asc. VO 6, desc. VO 26 [18 dis, 6 asc.ph], const. VO 21 [pronominals 31 Total OV 21, asc. OV 17, desc. OV 0, const. OV 4 Total VO 10, asc. VO 0, desc. VO 9 [3 dis], const. VO 1] Subordinate 15 nominal objects (out of 21 subordinate objects =71%) Total OV 11, asc. OV 9 [1 des.ph], desc. OV 0, const. OV 2 Total VO 4, asc. VO 1, desc. VO 3 [1 dis, 2 asc.ph], const. VO 0 [pronominals 6 Total OV 6, asc. OV 6, desc. OV 0, const. OV 0 Total VO 0, asc. VO 0, desc. VO 0, const. VO 0] Medea Main 134 nominal objects (out of 208 main objects =64%) Total OV 58, asc. OV 38 [1 des.ph], desc. OV 7 [6 dis], const. OV 13 Total VO 76, asc. VO 11, desc. VO 40 [18 dis, 12 asc.ph], const. VO 25 [pronominals 74 285 Total OV 38, asc. OV 37, desc. OV 0, const. OV 1 Total VO 36, asc. VO 1, desc. VO 28 [3 dis], const. VO 7] Subordinate 46 nominal objects (out of 71 subordinate objects =65%) Total OV 29, asc. OV 18, desc. OV 5 [3 dis], const. OV 6 Total VO 17, asc. VO 3, desc. VO 6 [4 dis], const. VO 8 [pronominals 25 Total OV 20, asc. OV 19, desc. OV 0, const. OV 1 Total VO 5, asc. VO 0, desc. VO 3, const. VO 2] Crito Main 22 nominal objects (out of 56 main objects =39%) Total OV 14, asc. OV 10, desc. OV 3 [2 phrasal O], const. OV 1 Total VO 8, asc. VO 4, desc. VO 2 [1 asc.ph], const. VO 2 [pronominals 34 Total OV 24, asc. OV 14, desc. OV 4, const. OV 6 Total VO 10, asc. VO 2, desc. VO 6, const. VO 2] Subordinate 5 nominal objects (out of 13 subordinate objects =38%) Total OV 3, asc. OV 3, desc. OV 0, const. OV 0 Total VO 2, asc. VO 2, desc. VO 0, const. VO 0 [pronominals 8 Total OV 7, asc. OV 5, desc. OV 1, const. OV 1 Total VO 1, asc. VO 0, desc. VO 1, const. VO 0] Appendix 1H: VS in Crito Pronominal VS and SV VS 51a3: ejavn se ejpiceirw'men hJmei'" ajpolluvnai divkaion hJgouvmenoi ei\nai 51d4: w|/ a]n mh; ajrevskwmen hJmei'" (infinitive following... ejxei'nai) 51d7 is the central of three consecutive conditionals: ejavn te ti" bouvlhtai uJmw'n eij" ajpoikivan ijevnai, eij mh; ajrevskoimen hJmei'" te kai; hJ povli", ejavn te metoikei'n a[llosev poi ejlqwvn.... 286 SV 50e6: kai; a{tt∆ a]n hJmei'" se ejpiceirw'men poiei'n, kai; soi; tau'ta ajntipoiei'n oi[ei divkaion ei\nai… 51c7: eij hJmei'" tau'ta ajlhqh' levgomen, o{ti ouj divkaia hJma'" ejpiceirei'" dra'n a] nu'n ejpiceirei'". 51e2: o}n trovpon hJmei'" tav" te divka" dikavzomen 51e4: a} a]n hJmei'" keleuvwmen poihvsein tau'ta VS with following non-finite phrases Crito 9 VS constructions have following infinitives or participles, 4 in main clauses: 43c2: ajll∆ oujde;n aujtou;" ejpiluvetai hJ hJlikiva [to; mh; oujci; ajganaktei'n th'/ parouvsh/ tuvch. 44a10: ∆Edovkei tiv" moi gunh; [proselqou'sa kalh; kai; eujeidhv", leuka; iJmavtia e[cousa, kalevsai me kai; eijpei'n: 48a10–1: Æ∆Alla; me;n dhv,Æ faivh g∆ a[n ti", Æoi|oiv tev eijsin hJma'" oiJ polloi; [ajpokteinuvnai.Æ 54c8 ajlla; mhv se peivsh/ Krivtwn [poiei'n a} levgei ma'llon h] hJmei'".Æ 5 in subordinates: 43a5: Qaumavzw o{pw" hjqevlhsev soi oJ tou' desmwthrivou fuvlax [uJpakou'sai. 43d3: ejx w|n ajpaggevllousin [h{kontev" tine" ajpo; Sounivou kai; katalipovnte" ejkei' aujtov]. 44d2: o{ti{{{ oi|oiv t∆ eijsi;n oiJ polloi; [ouj ta; smikrovtata tw'n kakw'n ejxergavzesqai ajlla; ta; mevgista scedovn, 45a6: kai; ga;r oujde; polu; tajrguvriovn ejstin o} qevlousi labovnte" tine;" [sw'saiv se kai; ejxagagei'n ejnqevnde. 51a2: ejavn se ejpiceirw'men hJmei'" [ajpolluvnai [divkaion hJgouvmenoi ei\nai 287 Appendix 2 Hyperbaton about the verb in the Oresteia and other texts 2A: OVO phrasal tmesis with demonstrative Page 288 2B: OVO tmesis with adjective or quantifier 289 2C: SVS in the Oresteia 291 2D: Wide-scope and combined hyperbaton in the Oresteia 292 2E: OVO hyperbaton in other Aeschylean works 294 2F: OVO and SVS hyperbaton in Sophocles 295 2G: OVO and SVS hyperbaton in Euripides 298 2H: Phrasal tmesis in Frogs 300 2I: Aeschylean demonstratives without hyperbaton 302 288 2A: OVO phrasal tmesis with demonstrative Stichomythia Ag.934: ei[per ti", eijdwv" g∆ eu\ tovdvvv ∆ ejxei'pen tevlo"vvv Ag.1202: mavnti"vvv m∆ ∆Apovllwnvvv tw/'d/' /'/' ∆ ejpevsthsen tevleivvv . Ag.1248: ajll∆ ou[ti paiw;n tw'/d'/ '/'/ ∆ ejpistatei' lovgwvvv / Choe.114: tivn∆ ou\n e[t∆ a[llon th'/de'/ '/'/ prostiqw' stavseivvv … Choe.488: pavntwn de; prw'ton tovndevvv presbeuvsw tavfonvvv . Choe.500: kai; th'sd∆' '' a[kouson loisqivou boh'"v 'v 'v ' , pavter, Choe.927: patro;" ga;r ai\sa tovndevvv soujrivzei movronvvv Eum.590: ouj keimevnw/ pw tovndevvv kompavzei" lovgonvvv . Rheses Ag.310: ka[peit∆ ∆Atreidw'n ej" tovdevvv skhvptei stevgo"vvv Ag.1070: i[q∆ w\ tavlaina: tovndvvv ∆ ejrhmwvsas∆ o[con[[[ : Ag.1295: ajporruevntwn, o[mma[[[ sumbavlw tovdevvv . Ag.1409: tovdvvv ∆ ejpevqou quvo"vvv , dhmoqrovou" t∆ ajra;" Ag.1614: movno" d∆ e[poikton[[[ tovndevvv bouleu'sai fovnonvvv … Ag.1627: ajndri; strathgw'/ tovndvvv ∆ ejbouvleusa" movronvvv … Choe.91: h] tou'to''' favskw tou[po"[[[ , wJ" novmo" brotoi'", Choe.149: toiai'sd∆ ejp∆ eujcai'" tavsdvvv ∆ ejpispevndw coav"vvv : Choe.187: plovkamonvvv ijdouvsh/ tovndevvv : pw'" ga;r ejlpivsw Choe.197: ajll∆ eu\ savf∆ h[/nei tovndvvv ∆ ajpoptuvsai plovkonvvv , Choe.226: koura;n;;; d∆ ijdou'sa thvndevvv khdeivou trico;" Choe.267–8: glwvssh" cavrin de; pavntvvv ∆ ajpaggelei' tavdevvv pro;" tou;" kratou'nta": ou}" i[doim∆ ejgwv pote Choe.508–511: a[kou∆: uJpe;r sou' toiavdvvv ∆ e[st∆ ojduvrmataj vj vj v , aujto;" de; swv/zh/ tovndevvv timhvsa" lovgonvvv . kai; mh;n ajmemfh' tovndvvv ∆ ejteivnaton lovgonvvv , tivmhma tuvmbou th'"''' ajnoimwvktou tuvch"vvv : Choe.760: knafeu;" trofeuv" te taujto;nj ;j ;j ; eijcevthn tevlo"vvv . Choe.765: oi[kwn, qevlwn de; tovndevvv peuvsetai lovgonvvv . Choe.985–6: oujc ouJmov", ajll∆ oJ pavntvvv ∆ ejpopteuvwn tavde vvv ”Hlio", a[nagna mhtro;" e[rga th'" ejmh'", Choe.991: h{ti" d∆ ejp∆ ajndri; tou't''' ∆ ejmhvsato stuvgo"vvv Choe.1015: patroktovnonvvv g∆ u{fasma{{{ prosfwnw'n tovdevvv : Eum.3–4: h} dh; to; mhtro;" deutevra tovdvvv ∆ e{zeto mantei'on''' , wJ" lovgo" ti": ejn de; tw'/ trivtw/ Eum.78–9: kai; mh; provkamne tovndevvv boukolouvmeno" 289 povnovvv n: molw;n de; Pallavdo" poti; ptovlin Eum.91–2: pompai'o" i[sqi, tovndevvv poimaivnwn ejmo;nj ;j ;j ; iJkevthnJ vJ vJ v . ....... Eum.306: u{mnon{{{ d∆ ajkouvsh// // tovndevvv devsmion sevqen. Eum.405 + 406: [pwvloi" ajkmaivoi" tovndvvv ∆ ejpizeuvxas∆ o[con[[[ ]. kainh;n;;; d∆ oJrw'sa thvndvvv ∆ oJmilivanJ vJ vJ v cqono;" Eum.581: o{pw" ãt∆à ejpivsta/ thvndevvv kuvrwson divkhnvvv . Eum.639: o{sper tevtaktai thvndevvv kurw'sai divkhnvvv . Eum.671: kai; tovnd∆v vv ejpikthvsaio suvmmaconvvv , qeav, Eum.836: e[cous∆ ej" aijeiv tovndvvv ∆ ejpainevsei" lovgonvvv . Lyrics Ag.1409: tovdvvv ∆ ejpevqou quvo"vvv , dhmoqrovou" t∆ ajra;" Choe.411: tovndevvv kluvousan oi\\ \\kton: Eum.940–1: flogmou;" ojmmatosterei'" futw'n to; mh; pera'n o{ron {{{ tovpwn, Appendix 2B: OVO tmesis with adjective or quantifier Stichomythia Ag.937 mhv nun to;n ajnqrwvpeion; j v; j v; j v aijdesqh'/" yovgonvvv . Ag.1665 oujk a]n ∆Argeivwn tovd∆ ei[h[[[ , fw'ta''' prossaivnein kakovnvvv . Rheses Ag.17: u{pnou tovd∆ ajntivmolponj vj vj v ejntevmnwn a[ko"[[[ , Ag.281: ”Hfaisto", “Idh" lampro;n;;; ejkpevmpwn sevla"vvv : Ag.535–6: ........ kai; panwvleqron aujtovcqonon patrw'/on'/ '/'/ e[qrisen dovmonvvv : Ag.582: Dio;" tavd∆ ejkpravxasa. pavntvvv ∆ e[cei" lovgonvvv . Ag.599: a[nakto" aujtou' pavntavvv peuvsomai lovgonvvv . Ag.857–8: oujk a[llwn pavra maqou's∆ ejmauth'" duvsforonvvv levxw bivv vvon Ag.863: polla;";;; kluvousan klhdovna" paligkovtou"v vv vv v , Ag.893–4: ajmfiv soi pavqhvvv oJrw'sa pleivwvvv tou' xuneuvdonto" crovnou. Ag.1067: pri;n aiJmathro;nJ ;J ;J ; ejxafrivzesqai mevno"vvv . Ag.1210: h[dh polivtai" pavntvvv ∆ ejqevspizon pavqhvvv . Ag.1275: kai; nu'n oJ mavnti" mavntinvvv ejkpravxa" ejmejjj ; 290 Ag.1588+9: tlhvmwn Quevsth" moi'ran''' hu{ret∆ ajsfalhjjj ', to; mh; qanw;n patrw'/on'/ '/'/ aiJmavxai pevdonvvv Ag.1609: pa'san''' xunavya" mhcanh;n;;; dusbouliva". Choe.133–4: ..... a[ndra[[[ d∆ ajnthllavxato Ai[gisqon[[[ , o{sper sou' fovnou metaivtio". Choe.283: a[lla"[[[ t∆ ejfwvnei prosbola;";;; ∆Erinuvwn Choe.545: kai; masto;n;;; ajmfevcask∆ ejmo;n qrepthvrionj ; vj ; vj ; v Choe.854: ou[toi frevn∆v vv a]n klevyeien wjmmatwmevnhnj vj vj v . Eum.41: e{dran{{{ e[conta prostrovpaionvvv , ...... Eum.444: tw'n sw'n ejpw'n mevlhmvvv ∆ ajfairhvsw mevgavvv : Eum.466: a[lgh[[[ profwnw'n ajntivkentraj vj vj v kardiva/, Eum.668: to; so;n povlisma kai; strato;n;;; teuvxw mevganvvv : Eum.734: ejmo;n tovd∆ e[rgon, loisqivanvvv kri'nai divkhnvvv : Eum.760: ...o}" patrw'/on'/ '/'/ aijdesqei;" movronvvv Lyrics Ag.124: ∆Atreiv>da" macivmou"vvv ejdavh lagodaivta"vvv , Ag.395: povlei provstrimmavvv qei;" a[ferton[[[ . Ag.1015–6: fhv" te kai; ejx ajlovkwn ejpeteia'n nh'stin''' w[lesen novsonvvv . Ag.1522–3: tw'/de genevsqai ãdovliovnv vv vv v te acei'n movronvvv oujk ajdivkw":à Choe.77: patrwv/wn douvliovnv vv vv v ãm∆à ejsa'gon ai\\ \\san, Choe.163: scevdiavvv v t∆ aujtovkwpaj vj vj v nwmw'n xivfhvvv . Choe.352–3: .... poluvcwstonvvv a]n ei\ce" tavfonvvv diapontivou ga'" Choe.401–2: cumevna" ej" pevdon a[llo[[[ prosaitei'n ai|ma||| : ...... Choe.418–9: ... h] tavpervvv pavqomen a[cea[[[ prov" ge tw'n tekomevnwn… Choe.615: a{t∆ ejcqrw'n u{per fw't''' ∆ ajpwvlesen fivlonvvv , Krhtikoi'" Eum.385: ajtivetaj vj vj v diovmenai lavchvvv Eum.501: pavntvvv ∆ ejfhvsw movronvvv , 291 Appendix 2C: SVS in the Oresteia Stichomythia Ag.547: povqen to; duvsfron tou't''' ∆ ejph'n stuvgo"vvv stratw'/… Choe.773: ejn ajggevlw/ ga;r kupto;";;; ojrqou'tai lovgo"vvv . Eum.676: hJmi'n me;n h[dh pa'n''' tetovxeutai bevlo"vvv Eum.750: gnwvmh" d∆ ajpouvsh" ph'ma''' givgnetai mevgavvv , Eum.751: balou'sa''' v d∆ oi\kon yh'fo"''' w[rqwsen mivavvv . Rheses Ag.20: nu'n d∆ eujtuch;"j ;j ;j ; gevnoit∆ ajpallaghjjj ; povnwn Ag.347: gevnoit∆ a[n, eij provspaiavvv v mh; tuvcoi kakav. Ag.420–1: ojneirovfantoij vj vj v de; penqhvmone"vvv pavreisi dovxaivvv ..... Ag.487: gunaikoghvrutonvvv o[llutai klevo"vvv . Ag.653: ejn nukti; duskuvmantavvv d∆ wjrwvrei kakavv vv: Ag.766–7: .... o{te to; kuv; v; v; v –– ––rion movlh/ favo"vvv tovkou, Ag.917: aijnei'n, par∆ a[llwn crh;; ;; tovdvvv ∆ e[rcesqai gevra"vvv . Ag.1560: o[neido"[[[ h{kei tovd∆v vv ajnt∆ ojneivdou", Choe.13: povtera dovmoisi ph'ma''' proskurei' nevonvvv , Choe.207: kai; ga;r duvv vv∆ ejsto;n twvde perigrafavvv ; podoi'n, Choe.249–250: ........tou;" d∆ ajpwrfanismevnou" nh'sti"''' pievzei limov"vvv : ouj ga;r ejntelei'" Choe.260: ou[t∆ ajrcikov" soi pa'" o{d{{{ ∆ aujanqei;" puqmh;n;;; Choe.550: kteivnw nin, wJ" tou[neiron[[[ ejnnevpei tovdevvv . Choe.580: o{pw" a]n ajrtivkollaj vj vj v sumbaivnh/ tavdevvv : Choe.660–1: tavcune d∆, wJ" kai; nukto;" a{rm{{{ ∆ ejpeivgetai skoteinovnvvv , w{ra d∆ ejmpovrou" meqievnai Eum.192: stevrghqr∆ e[cousai… pa'"''' d∆ uJfhgei'tai trovpo"vvv morfh'": ......... Eum.249: splavgcnon: cqono;" ga;r pa'"''' pepoivmantai tovpo"vvv , Eum.636: ajndro;" me;n uJmi'n ou|to"||| ei[rhtai movro"vvv Eum.664: mavrtu"vvv pavresti pai'"''' ∆Olumpivou Diov", Eum.742–3: ejkbavlleq∆ wJ" tavcista teucevwn pavlou" o{soi" dikastw'n tou't''' ∆ ejpevstaltai tevlo"vvv . Eum.864: qurai'o"''' e[stw povlemo"vvv , ouj movli" parwvn, 292 Lyrics Choe.65: tou;" d∆ a[kranto"[[[ e[cei nuvxvvv . Choe.961–2: ...mevgavvv t∆ ajfh/revqh yavlionvvv oi[kwn. Eum.313–4: to;n me;n kaqara;" cei'ra" pronevmont∆ ou[ti"[[[ ejfevrpei mh'ni"''' ajf∆ hJmw'n, Eum.336–7: qnatw'n toi'sin aujtourgivaij vj vj v xumpevswsin mavtaioivvv , Eum.544: kuvrionvvv mevnei tevlo"vvv . Eum.942–3: mhd∆ a[karpo" aija[ j[ j[ j - nh;";;; ejferpevtw novso"vvv , Appendix 2D: Wide-scope and combined hyperbaton in the Oresteia Wide-scope hyperbaton (both object and subject) Ag.279: th'"''' nu'n tekouvsh" fw'" tovd∆ eujfrovnh"j vj vj v levgw. Ag.890: ta;";;; ajmfiv soi klaivousa lampthrouciva"vvv Ag.947: mhv ti" provswqen o[mmato" bavloi fqovno"vvv . Ag.1056–7: ........ ta; me;n ga;r eJstiva" mesomfavlou e{sthken h[dh mh'la''' Êpro;" sfaga;"Ê purov", Ag.1062: eJrmhnevw"J vJ vJ v e[oiken hJ xevnh torou'' '' Ag.1102: mevgvvv ∆ ejn dovmoisi toi'sde mhvdetai kakovnvvv , Ag.1137: to;; ;; ga;r ejmo;nj ;j ;j ; qrow' pavqo"vvv ejpegcevai. Ag.1309: fovnonvvv dovmoi pnevousin aiJmatostagh'J 'J 'J '. Ag.1400: h{ti" toiovndvvv ∆ ejp∆ ajndri; kompavzei" lovgonvvv . Ag.1431: kai; thvnd∆v vv ajkouvei" oJrkivwn ejmw'n qevminvvv : Ag.1460–1: h\ ti" h\n tovt∆ ejn dovmoi". “Eri" ejrivdmato"j vj vj v ajndro;" oijzuv". Choe.299: polloi; ga;r eij" e}n xumpivtnousin i{meroi,{ {{ Choe.406: polukratei'"''' i[desqe fqeimevnwn ∆Araiv, Choe.591–2: ..... kajnemoventj vj vj v ∆ a]n aijgivdwn fravsai kovtonvvv . Choe.731: trofo;n;;; d∆ ∆Orevstou thvnd∆v vv oJrw' keklaumevnhnvvv : Choe.798–9: tou't''' ∆ ijdei'n, davpedon aJnovmenonJ vJ vJ v Ê bhmavtwn Ê o[regma[[[ . Choe.896–7: ejpivsce", w\ pai', tovndevvv d∆ ai[desai, tevknon, mastovnvvv , pro;" w|/ su; polla; dh; brivzwn a{ma Choe.911 kai; tovndevvv toivnun Moi'r∆ ejpovrsunen movronvvv . 293 Choe.952: ojlevqrionj vj vj v pnevous∆ ejn ejcqroi'" kovtonvvv . Choe.975: semnoi;; ;; me;n h\san ejn qrovnoi" tovq∆ h{menoi,{ {{ Choe.1015: patroktovnonvvv g∆ u{fasma{{{ prosfwnw'n tovdevvv : Choe.1017: a[zhla[[[ nivkh" th'sd∆ e[cwn miavsmatavvv . Eum.20: touvtou"vvv e;n eujcai'" froimiavzomai qeouv"vvv . Eum.80: i{zou palaio;n;;; a[gkaqen labw;n brevta"vvv : Eum.110: kai; pavnta tau'tav 'v 'v ' la;x oJrw' patouvmenavvv , Eum.224: divka" de; Palla;";;; tw'nd∆ ejpopteuvsei qeav. Eum.227: tima;";;; su; mh; xuvntemne ta;" ejma;"; j ;; j ;; j ; lovgw/. Eum.268: ãi{n∆à ajntipoivnou"j vj vj v tivnh/" matrofovnou duva"vvv : Eum.284: polu;";;; dev moi gevnoit∆ a]n ejx ajrch'" lovgo"vvv , Eum.419: timav"vvv ge me;n dh; ta;" ejma;"; j ;; j ;; j ; peuvsh/ tavca Eum.424: h\ kai; toiauvta"vvv tw'/d∆ ejpirroizei'" fugav"vvv … Eum.447: tekmhvrionvvv de; tw'ndev soi levxw mevgavvv : Eum.455–6: .... patevravvv d∆ iJstorei'" kalw'", ∆Agevmnonvvv ∆, ........ Eum.482: ejpei; de; pra'gma''' deu'r∆ ejpevskhyen tovdevvv , Eum.494–5: pavnta"vvv h[dh tovd∆ e[rgon eujcereiv< a/ sunarmovsei brotouv"vvv : Eum.669: kai; tovndvvv ∆ e[pemya sw'n dovmwn ejfevstionj vj vj v , Eum.700: toiovndevvv toi tarbou'nte" ejndivkw" sevba"vvv Eum.865: ejn w|/ ti" e[stai deino;";;; eujkleiva"j vj vj v e[rw"[[[ : Eum.892: a[nass∆ ∆Aqavna, tivnavvv me fh;/" e[cein e{dran{{{ … Eum.898: kaiv moi pro; panto;";;; ejgguvhn qhvsh/ crovnouvvv … Eum.915: thvndvvv ∆ ajstuvnikon ejn brotoi'" tima'n povlinvvv . Eum.938: dendrophvmwnvvv de; mh; pnevoi blavbavvv , Combined patterns in the Oresteia Choe.508ff. SVS, two OVO constructions, then a double nominal: a[kou∆: uJpe;r sou' toiavdvvv ∆ e[st∆ ojduvrmataj vj vj v , aujto;" de; swv/zh/ tovndevvv timhvsa" lovgonvvv . kai; mh;n ajmemfh' tovndvvv ∆ ejteivnaton lovgonvvv , tivmhma tuvmbou th'"''' ajnoimwvktou tuvch"vvv : Choe.985–6: Nested subject and object hyperbaton: oujc ouJmov", ajll∆ oJ pavntvvv ∆ ejpopteuvwn tavde vvv ”Hlio", a[nagna mhtro;" e[rga th'" ejmh'", 294 Ag.537: VS with object hyperbaton: dipla'' '' d∆ e[teisan Priamivdai qajmavrtiaj vj vj v . Ag.563: ceimw'na''' d∆ eij levgoi ti" oijwnoktovnonj vj vj v , Eum.707–8. Double hyperbaton of object and indirect object: tauvthnvvv me;n ejxevtein∆ ejmoi'" paraivnesivvv n ajstoi'sin ej" to; loipovn:....... Ag.280. Subject and adverbial : kai; tiv"vvv tovd∆ ejxivkoit∆ a]n ajggevlwnj vj vj v tavco"… Eum.668–671. Narrow and wide-scope: to; so;n povlisma kai; strato;n;;; teuvxw mevganvvv : kai; tovndvvv ∆ e[pemya sw'n dovmwn ejfjjj evstionvvv , o{pw" gevnoito pisto;" ej" to; pa'n crovnou, kai; tovndvvv ∆ ejpikthvsaio suvmmaconvvv , qeav, Choe.744–748. Wide-scope hyperbaton: w{" moi ta; me;n palaia; sugkekramevna a[lgh[[[ duvsoista toi'sd∆ ejn ∆Atrevw" dovmoi" tucovnt∆ ejmh;nj ;j ;j ; h[lgunen ejn stevrnoi" frevnavvv : ajll∆ ou[tiv pw toiovnde ph'm∆ ajnescovmhn: ta; me;n ga;r a[lla[[[ tlhmovnw" h[ntloun kakav, Eum.845–6. Wide-scope subject hyperbaton with verb tmesis and factitive adjunct separated from its pronoun: ajpo;j ;j ;j ; me ga;r tima'n danaia'n qew'n duspavlamoi par∆ oujde;n h\\ \\ran dovloi. Appendix 2E: OVO hyperbaton in other Aeschylean works Canonical demonstrative-noun constructions Pers.214: swqei;" d∆ oJmoivw" thvsdevvv koiranei' cqonov"vvv Pers.363: pa'sin profwnei' tovndevvv nauavrcoi" lovgonvvv , Sept.638: fugh"/ to;n aujton tovnde; j v; j v; j v teivsasqai trovponvvv . Sept.717: oujk a[ndr∆ oJplivthn tou'to''' crh; stevrgein e[po"[[[ Supp.252: gevno" Pelasgw'n thvndevvv karpou'tai cqovnavvv . Supp.325–6: dokei'tev ãtoivà moi th'sde''' koinwnei'n cqono;";;; tajrcai'on: Supp.378: oujd∆ au\ tovd∆ eu\fron, tavsdvvv ∆ ajtimavsai litav"vvv . Supp.461: eij mhv ti pisto;n tw'/d'/ '/'/ ∆ uJposthvsei" stovlwvvv / – 295 Supp.607–8: ...........cersi; dexiwnuvmoi" e[frixen aijqh;r tovndevvv krainovntwn lovgonvvv : Supp.790: pri;n a[ndr∆ ajpeukto;n tw'/de'/ '/'/ crimfqh'nai croi?? ?? Prom.31: ajnq∆ w|n ajterph' thvndevvv frourhvsei" pevtranvvv Prom.87: o{tw/ trovpw/ th'sd''' ∆ ejkkulisqhvsh/ tevcnh"vvv . Prom.738: crh/vzwn migh'nai tavsdvvv ∆ ejpevrriyen plavna"vvv . Other constructions Pers.140: ajll∆ a[ge, Pevrsai, tovdvvv ∆ ejnezovmenoi stevgo" ajrcai'onv j 'v j 'v j ' Pers.227: paidi; kai; dovmoi" ejmoi'si thvndevvv ejkuvrwsa" favtinvvv Pers.233: ajlla; mh;n i{meir∆ ejmo;" pai'" thvndevvv qhra'sai povlinvvv … Pers.246: ajll∆ ejmoi; dokei'n tavc∆ ei[sh/ panta namerth' lovgonvvv : Pers.777: ...... oi|" tovdvvv ∆ h\n crevo"vvv , Sept.48: h] gh'n''' qanovnte" thvndevvv furavsein fovnw. Supp.189: pavgonvvv prosivzein tovndvvv ∆ ajgwnivwn qew'n. Supp.233 [SVS]: o{pw" a]n uJmi'n pra'go"''' eu\ nika'/ tovdevvv . Supp.278 [SVS:] o{pw" tovdvvv ∆ uJmi'n ejstin ∆Argei'on gevno"vvv . Supp.508: leuro;n kat∆ a[lso"[[[ nu'n ejpistrevfou tovdevvv . Supp.1029: tovdevvv meilivssonte" ou\\ \\da". Prom.386: ejmo;nj ;j ;j ; dokhvsei tajmplavkhmj vj vj v ∆ ei\nai tovdevvv . Prom.766: tiv d∆ o{ntin∆… ouj ga;r rJhto;nJ ;J ;J ; aujda'sqai tovdvvv e. Prom.975: aJplw'/ lovgw/ tou;" pavnta"; v; v; v ejcqaivrw qeouv"vvv , Prom.980: ..tovdevvv Zeu;" tou[po"[[[ oujk ejpivstatai. Appendix 2F: OVO and SVS hyperbaton in Sophocles OT. OVO 13: ei[hn toiavndevvv mh; ouj katoiktivrwn e{dran{{{ . 43: fhvmhn ajkouvsa" ei[t∆ ajp∆ ajndro;"j ;j ;j ; oi\sqav tou: 72: drw'n h] tiv fwnw'n thvndevvv rJusaivmhn povlinvvv . 86: tivnv vv ∆ hJmi;n h{kei" tou' qeou' fhvmhnvvv fevrwn… 102: Poivou ga;r ajndro;" thvndevvv mhnuvei tuvchnvvv … 104: gh'" th'sde, pri;n se; thvndvvv ∆ ajpeuquvnein povlinvvv . 296 110: ∆En th'/d'/ '/'/ ∆ e[faske gh'/: to; de; zhtouvmenon 113: h] gh'" ejp∆ a[llh" tw'/de'/ '/'/ sumpivptei fovnw/v /v /v /… 134: pro;" tou' qanovto" thvndvvv ∆ e[qesq∆ ejpistrofhvnj vj vj v : 137–8: ÔUpe;r ga;r oujci; tw'n''' ajpwtevrw fivlwnvvv , ajll∆ aujto;" auJtou' tou't''' ∆ ajposkedw' muvso"vvv . 143: i{stasqe, touvsdvvv ∆ a[rante" iJkth'ra" klavdou"J ' vJ ' vJ ' v , 168–9: «W povpoi, ajnavriqmaj vj vj v ga;r fevrw phvmatavvv : nosei' dev moi provpa" 203: w\ Zeu' pavter, uJpo; sw'/'/ '/'/ fqivson keraunw'/. 210: ta'sd''' ∆ ejpwvnumon ga'"''' , 248: kako;n kakw'" nin a[[ [[moron ejktri'yai bivonvvv : 291: Ta; poi'a tau'ta… pavntavvv ga;r skopw' lovgonvvv . 311: mhvt∆ ei[ tin∆ a[llhn[[[ mantikh'" e[cei" oJdovnJ vJ vJ v , 323: th'/d∆ h{ s∆ e[qreye, thvndvvv ∆ ajposterw'n favtinvvv . 337: ∆Orgh;n;;; ejmevmyw th;n ejmhvn; j v; j v; j v , th;n sh;n d∆ oJmou' 340: kluvwn a} nu'n su; thvndvvv ∆ ajtimavzei" povlinvvv … 374: Mia'"''' trevfh/ pro;" nuktov"vvv , w{ste mhvt∆ ejme; 538: h] tou[rgon[[[ wJ" ouj gnwrivsoimiv sou tovdevvv 582: ∆Entau'qa ga;r dh; kai; kako;";;; faivnh/ fivlo"vvv . 595: w{st∆ a[lla crhv/zein h] ta;; ;; su;n kevrdei kalav. 606–7: koinh'/ ti bouleuvsanta, mhv m∆ aJplh'/J '/J '/J '/ ktavnh/" yhvfwvvv /, diplh'/ de;, th'/ t∆ ejmh'/ kai; sh'/, labwvn, 615: kako;n de; ka]n ejn hJmevraJ vJ vJ v / gnoivh" mia'/. 640: dra'sai dikaioi', duoi'n' '' ajpokrivna" kakoi'n' '' , 664: ojloivman, frovnhsinvvv eij tavndvvv ∆ e[cw. 670: h] gh'"''' a[timon th'sd''' ∆ ajpwsqh'nai biva/: 688: toujmo;nj ;j ;j ; pariei;" kai; katambluvnwn kevarvvv … 709: brovteion oujde;n mantikh'"''' e[con tevcnh"vvv : 841: Poi'on dev mou perisso;n;;; h[kousa" lovgonvvv … 901: pa'sin''' aJrmovsei brotoi'"''' . 1070: tauvthn d∆ eja'te plousivwvvv / caivrein gevneivvv . 1148: dei'tai kolastou' ma'llon h] ta; tou'd∆ e[ph[[[ . 1226: tw'n Labdakeivwn' v' v' v ejntrevpesqe dwmavtwnvvv . 1272: ou[q∆ oi|∆ e[pascen ou[q∆ oJpoiJJJ '∆ e[dra kakav, 1280: Tavdvvv ∆ ejk duoi'n e[rrwgen, ouj movnou, kakav, 1320: dipla' se penqei'n kai; dipla' forei'n kakav. 297 SVS OT.281: a}n mh; qevlwsin oujd∆ a]n ei|"||| duvnait∆ ajnhvrj vj vj v . 327: pavnte"vvv se proskunou'men oi{d∆ iJkthvrioi{ J v{ J v{ J v . 391: pw'" oujc, o{q∆ hJ rJayw/do;"J J / ;J J / ;J J / ; ejnqavd∆ h\n kuvv vvwn, 417: Kaiv s∆ ajmfiplh;xj ;j ;j ; mhtrov" te kai; tou' sou' patro;" ejla'/ pot∆ ejk gh'" th'sde deinovpou" ajrav jv jv j v, 508: Fanera; ga;r ejp∆ aujtw'/ pterovess∆vvv h\lqe kovravvv 561: Makroi; palaioi;;; v t∆ a]n metrhqei'en crovnoi.v vv 732: Kai; pou' ∆sq∆ oJ cw'ro" ou|to", ou| tovdvvv ∆ h\n pavqo"vvv … 753: kh'rux: ajphvnhj vj vj v d∆ h\ge Lavi>on mivavvv . 816: tiv" ejcqrodaivmwnv j vv j vv j v ma'llon a]n gevnoit∆ ajnhvrj vj vj v … 1440: ∆All∆ h{ g∆ ejkeivnou pa's''' ∆ ejdhlwvqh favti"vvv , Combined pattern 1032: øAG.Ø Podw'n a]n a[rqra[[[ marturhvseien ta; sa;;; v. øOI.Ø Oi[moi, tiv tou't∆ ajrcai'on' j '' j '' j ' ejnnevpei" kakovnvvv … Hyperbaton elsewhere in Sophocles Examples: Aj. 545: ....... tarbhvsei ga;r ou] neosfagh' pou tovndevvv prosleuvsswn fovnonvvv , Aj. 907: biva/ politw'n tovndvvv ∆ a]n h/jrovmhn povnonvvv Aj. 1103: oujd∆ e[sq∆ o{pou soi; tovndevvv kosmh'sai plevonvvv Aj. 1126: divkaia ga;r tovndvvv ∆ eujtucei'n kteivnantavv vv vv v me… Aj. 1179: au[tw" o{pwsper tovndvvv ∆ ejgw; tevmnw plovkonvvv . Trach.774: poivai" ejnevgkoi tovndevvv mhcanai'" pevplonvvv : El.388: tivn∆, w\ tavlaina, tovndvvv ∆ ejphravsw lovgonvvv … El.1216: ei[per g∆ ∆Orevstou sw'ma''' bastavzw tovdevvv … Ph.37: keivnou to; qhsauvrisma; v; v; v shmaivnei" tovdevvv . Ph.1299: ajll∆ ou[ ti caivrwn, h]n tovdvvv ∆ ojrqwqh/' bevlo"vvv . OC.712–3: w\ pai' Krovnou, su; gavr nin eij" tovdvvv ∆ ei|sa" au[chm[[[ ∆, a[nax Poseidavn, 298 Formulaic constructions in other Sophoclean works: pavnt∆ vvv ..... lovgonvvv Aj.: 480 (ajkhvkoa"), Tr.: 484 (ejpivstasai), OT.: 291(ga;r skopw'), Phil.: 1240 (ajkhvkow"). thvnde vvv ....... povlinvvv Ant.994 (ejnauklhvrei"), 1058 (e[cei" swvsa"), OT.72 (ejrusaivmhn), 104 (ajpeuquvnein), 340 (ajtimavzei"), OC. 1533(ejnoikhvsei") thvnde vvv ..... novsonvvv (Aj. 66 perifanh'), bohvnvvv (Aj. 335 qwuvssei), tuvchnvvv (OT.102 mhnuvei), favtinvvv (OT.323 ajposterw'n, El.1213 prosfwnei'n), carin (OC.586 m∆ ejxaith/'), e{dran{{{ (OC.1166 oJ prosqakw'n), tevcnhnvvv (Tr.620 pompeuvw). Appendix 2G: OVO and SVS hyperbaton in Euripides Medea OVO 166: w\ pavter, w\ povli", w|n ajpenavsqhn aijscrw'" to;n ejmo;n; j ;; j ;; j ; kteivnasa kavsinvvv . 201: dai'te", tiv mavthnvvv teivnousi bohvnvvv … 340: mivanvvv me mei'nai thvndvvv ∆ e[ason hJmevranJ vJ vJ v 373: gh'" ejkbalovnti, thvndvvv ∆ ajfh'ken hJmevranJ vJ vJ v 391: dovlw/ mevteimi tovndevvv kai; sigh'/ fovnonvvv : 433–5: mainomevna/ kradiva/, diduvmou"vvv oJrivsasa povntou pevtra"vvv : ejpi; de; xevna/ naivei" cqoniv, ta'" ajnavndrou 462–3: mhvt∆ ejndehv" tou: povllvvv ∆ ejfevlketai fugh; kaka; xu;n auJth'/. kai; ga;r eij suv me stugei'", 487–91: paivdwn uJp∆ aujtou', pavntavvv t∆ ejxei'lon dovmonvvv . kai; tau'q∆ uJf∆ hJmw'n, w\ kavkist∆ ajndrw'n, paqw;n prouvdwka" hJma'", kaina; d∆ ejkthvsw levchvvv , paivdwn gegwvtwn: eij ga;r h\sq∆ a[pai" e[ti, suggnwvst∆ a]n h\n soi tou'd''' ∆ ejrasqh'nai levcou"vvv . 510–1: e[qhka" ajnti; tw'nde: qaumasto;n;;; dev se e[cw povsinvvv kai; pisto;n;;; hJ tavlain∆ ejgwv, 531: tovxoi" ajfuvktoisÊ toujmo;nj ;j ;j ; ejksw'/sai devma"vvv . 576: ∆Ia'son, eu\ me;n touvsdvvv ∆ ejkovsmhsa" lovgou"vvv : 299 588: kalw'" g∆ a[n, oi\mai, tw'/d'/ '/'/ ∆ uJphrevtei" lovgw/,v /v /v / 599: mhd∆ o[lbo" o{sti" th;n ejmh;n; j ;; j ;; j ; knivzoi frevnavvv . 604: ejgw; d∆ e[rhmo" thvndevvv feuxou'mai cqovnavvv . 612: levg∆: wJ" e{toimo" ajfqovnw/ j v /j v /j v / dou'nai ceri;; ;; 667: Foivbou palaio;n;;; ejklipw;n crhsthvrionvvv . 677: mavlist∆, ejpeiv toi kai; sofh'"''' dei'tai frenov"vvv . 682: su; d∆ wJ" tiv crhv/zwn thvndevvv naustolei'" cqovnavvv … 691: tiv fhv/"… safw'" moi sa;";;; fravson dusqumiva"vvv . 697: povteron ejrasqei;" h] so;n;;; ejcqaivrwn levco"vvv … 705: pro;" tou'… tovdvvv ∆ a[llo kaino;n;;; au\ levgei" kakovnvvv . 727: aujth; d∆ ejavnper eij" ejmou;"j ;j ;j ; e[lqh/" dovmou"vvv , 776: molovnti d∆ aujtw'/ malqakou;";;; levxw lovgou"vvv , 780: pai'da"''' de; mei'nai tou;" ejmou;"; j ;; j ;; j ; aijthvsomai, 789–90: toioi'sde''' crivsw farmavkoi"vvv dwrhvmata. ejntau'qa mevntoi tovndvvv ∆ ajpallavssw lovgonvvv : 811: ejpeivper hJmi'n tovndvvv ∆ ejkoivnwsa" lovgonvvv , 901–2: a\r∆, w\ tevkn∆, ou{tw kai; polu;n;;; zw'nte" crovnonvvv fivlhn ojrevxet∆ wjlevnhn… tavlain∆ ejgwv, 927: dravsw tavd∆: ou[toi soi'"''' ajpisthvsw lovgoi"vvv : 929: tiv dh'ta livan toi'sd''' ∆ ejpistevnei" tevknoi"vvv … 932: ajll∆ w|nper ou{nek∆ eij" ejmou;"j ;j ;j ; h{kei" lovgou"vvv , 959: tiv d∆, w\ mataiva, tw'nde sa;";;; kenoi'" cevra"vvv … 973: ej" cei'r∆ ejkeivnh" dw'ra''' devxasqai tavdevvv . 1010: aijai' mavl∆ au\qi". øPa.Ø mw'n tin∆ ajggevllwn tuvchnvvv 1083: kai; pro;" aJmivlla"J vJ vJ v h\lqon meivzou"vvv 1120: deivknusin w{" ti kaino;n;;; ajggelei' kakovnvvv . 1124: tiv d∆ a[xiovn moi th'sde''' tugcavnei fugh'"''' … 1127: kavllistonvvv ei\pa" mu'qon''' , ejn d∆ eujergevtai" 1138: h{sqhmen oi{per soi'"''' ejkavmnomen kakoi'"''' 1167: toujnqevnde mevntoi deino;n;;; h\n qevamvvv ∆ ijdei'n: 1180: stevgh puknoi'sin''' ejktuvpei dromhvmasinvvv . 1189: lepth;n;;; e[dapton savrkavvv th'" dusdaivmono". 1212: crhv/zwn geraio;n;;; ejxanasth'sai devma"vvv 1300: ajqw'/o" aujth; tw'nde''' feuvxesqai dovmwnvvv … 1307: ∆Ia'son: ouj ga;r touvsdvvv ∆ a]n ejfqevgxw lovgou"vvv . 1317: tiv tavsdevvv kinei'" kajnamocleuvei" puvla"vvv , 1355: terpno;n;;; diavxein bivotonvvv ejggelw'n ejmoiv: 300 SVS Med.386–8 (double pattern): ei\eJn: kai; dh; teqna'si: tiv"vvv me devxetai povli"vvv … tiv"vvv gh'n a[sulon kai; dovmou" ejcegguvou" xevno"vvv parascw;n rJuvsetai toujmo;n devma"… 539: pavnte"vvv dev s∆ h[/sqont∆ ou\san ”Ellhne" sofh;n 871: eijkov" s∆, ejpei; nw'/n povllvvv ∆ uJpeivrgastai fivlavvv . 906: øCo.Ø kajmoi; kat∆ o[sswn clwro;n;;; wJrmhvqh davkruvvv : 1214: leptoi'si pevploi", deina; d∆ h\n palaivsmatavvv : 1345: davkoimi: toiovndvvv ∆ ejmpevfukev soi qravso"vvv : Combined pattern 911–3: ajll∆ ej" to; lw'/on so;n;;; meqevsthken kevarvvv , e[gnw" de; th;n;;; nikw'san''' , ajlla; tw'/ crovnw/, boulhvnvvv : gunaiko;" e[rga tau'ta swvfrono". Appendix 2H: phrasal tmesis in Frogs OVO and non-finite object constructions 105: (Euripidean quote, probably from Andromeda): Mh; to;n ejmo;nj ;j ;j ; oi[kei nou'n''' : e[cei" ga;r oijkivan. 143: Meta; tau't∆ o[fei" kai; qhriv∆ o[yei murivavvv deinovtatavvv . 170: Kai; gavr tin∆ ejkfevrousi toutoni; nekrovnvvv . 333: carivtwn plei'st''' on e[cousan mevro"vvv , aJgnhvn, iJera;n 371: kai; pannucivda" ta;" hJmetevra" ai} th'/de'/ '/'/ prevpousin eJorthJJJ '/. 478: ejf∆ a}" ejgw; dromai'on''' oJrmhvsw povdavvv . 502: Fevre nun, ejgw; ta; strwvmat; v; v; v ∆ ai[rwmai tadiv. 506: e[tnou" duv∆ h] trei'", bou'n''' ajphnqravkiz∆ o{lon{{{ , 511: e[fruge, kw\\ \\ /non /// ajnekeravnnu glukuvtatonvvv . 676: to;n polu;n; ;; ;; ; ojyomevnh law'n o[clon[[[ , ou| sofivai 682: ejpi; bavrbaronvvv eJzomevnh pevtalonvvv : 747–8: Tiv de; tonqoruvzwn, hJnivk∆ a]n plhga;";;; labw;n polla;";;; ajpivh/" quvraze… øOI.Ø Kai; tou'q∆ h{domai. 808: Pollou;";;; i[sw" ejnovmize tou;" toicwruvcou"; v; v; v . 814: «H pou deino;n;;; ejribremevta" covlonvvv e[ndoqen e{xei, 815–6: hJnivk∆ a]n ojxuvlalovnj v vj v vj v v per i[dh/ qhvgonto" ojdovntaj vj vj v ajntitevcnou: 301 827: glw'ss∆ ajnelissomevnh fqonerou;";;; kinou'sa calinouv"vvv , 889 (Spoken by Euripides): e{teroi gavr eijsin oi|sin||| eu[comai qeoi'"''' . 909–10 (Spoken by Euripides): wJ" h\n ajlazw;n kai; fevnax oi{oi" te tou;" qeata;"; ;; ;; ; ejxhpavta mwvrou"vvv labw;n para; Frunivcw/ trafevnta"vvv . 951 (Spoken by Aeschylus): oujk ajpoqanei'n se tau't''' ∆ ejcrh'n tolmw'nta''' … øEU.Ø Ma; to;n ∆Apovllw: 1044 (Spoken by Aeschylus): oujd∆ oi\d∆ oujdei;" h{ntin∆ ejrw'san{ j '{ j '{ j ' pwvpot∆ ejpoivhsa gunai'ka''' . 1121 (Euripides of Aeschylus): prwvtiston aujtoujjj ' basaniw' tou' dexiou'.' '' '' ' 1126 (Aeschylean quotation): ÆÔErmh' cqovnie, patrw'/∆'/ '/'/ ejpopteuvwn kravthvvv 1146 (Spoken by Aeschylus): oJtih; patrw'/on tou'to''' kevkthtai gevra"vvv . 1178–9 (Spoken by Euripides): Ka[n pou di;" ei[pw taujtovn, h] stoibh;n;;; i[dh/" ejnou'san e[xw tou' lovgou, katavptuson. 1240 (Euripidean quote: Lecythian): Oijneuv" pot∆ ejk gh'" poluvmetronvvv labw;n stavcunvvv quvwn ajparcav" 1299–1300 (Spoken by Aeschylus): h[negkon au[q∆, i{na mh; to;n aujto;n; j ;; j ;; j ; Frunivcw/ leimw'na''' Mousw'n iJero;nJ ;J ;J ; ojfqeivhn drevpwn: 1301 (Spoken by Aeschylus): ou|to" d∆ ajpo; pavv vvntwn me;n fevrei pornw/diw'n/ '/ '/ ' , 1436 (Spoken by Dionysus): peri; th'" povlew" h{ntin{{{ ∆ e[ceton swthrivanvvv SVS 154: ∆Enteu'qen aujlw'n tiv"vvv se periveisin pnohv, 314: au[ra ti"[[[ eijsevpneuse mustikwtavthvvv . 503: «W fivltaqvvv ∆ h{kei" ÔHravklei"vvv … Deu'r∆ ei[siqi. 302 948 (Spoken by Euripides): “Epeit∆ ajpo; tw'n prwvtwn ejpw'n oujde;nj ;j ;j ; parh'k∆ a]n ajrgovnj vj vj v , 1206 (Spoken by Euripides): ÆAi[gupto", wJ" oJ plei'sto"J 'J 'J ' e[spartai lovgo"vvv , 1307 (Spoken by Aeschylus): pro;" h{nper ejpithvdeia tavdvvv ∆ e[st∆ a[/dein mevlhvvv . Appendix 2I: Aeschylean demonstratives without hyperbaton Noun > Demonstrative (35) i) 22 at the penthemimeral caesura: Ag.18: klaivw tovt∆ oi[kou tou'de[ '[ '[ ' sumfora;n stevnwn - - v - |- : - v : - | v - v - Ag.906: e[kbain∆ ajphjjj vnh" th'sdev 'v 'v ' , mh; camai; tiqeiv" Ag.1039: e[kbain∆ ajphvnh" th'sdej v 'j v 'j v ' , mhd∆ uJperfrovnei: Ag.1071: ei[kous∆ ajnavgkh/ th'/dej v / '/j v / '/j v / '/ kaivnison zugovn. Ag.1283: kavteisin a[ta" tavsde[ v[ v[ v qrigkwvswn fivloi". Ag.1419: ouj tou'ton ejk gh'" th'sde' '' '' ' crh'n s∆ ajndrhlatei'n Ag.1438: kei'tai gunaiko;" th'sde; '; '; ' oJ lumanthvrio", Ag.1441: kai; koinovlektro" tou'dev 'v 'v ' , qesfathlovgo", Choe.142: hJmi'n me;n eujca;" tavsdej ; vj ; vj ; v , toi'" d∆ ejnantivoi" Choe.200: a[galma tuvmbou tou'dev 'v 'v ' kai; timh;n patrov". Choe.231: ijdou' d∆ u{fasma tou'to{ '{ '{ ' , sh'" e[rgon cerov", Choe.540: ajll∆ eu[comai gh'/ th'/de'/ '/'/ '/'/ '/ kai; patro;" tavfw/ Choe.740: keivnh/, dovmoi"vvv de; toi'sde''' pagkavkw" e[cein Eum.1: Prw'ton me;n eujch'/ th'/dej '/ '/j '/ '/j '/ '/ presbeuvw qew'n Eum.16: Delfov" te cwvra" th'sdev 'v 'v ' prumnhvth" a[nax: Eum.103: o{ra de; plhga;" tavsde; v; v; v kardiva/ sevqen: Eum.185: ou[toi dovmoisi toi'sdev 'v 'v ' crivmptesqai prevpei, Eum.207: ouj ga;r dovmoisi toi'sdev 'v 'v ' provsforoi molei'n. Eum.623: fravzein ∆Orevsth/ tw'/dev / '/v / '/v / '/ , to;n patro;" fovnon Eum.762: ejgw; de; cwvra/ th'/dev / '/v / '/v / '/ kai; tw'/ sw'/ stratw'/ Eum.834: pollh'" de; cwvra" th'sdev 'v 'v ' tajkroqivnia Eum.854: e[stai polivtai" toi'sdev 'v 'v ' , kai; su; timivan 303 13 Noun > Demonstrative, positioned elsewhere: Ag.311: favo" tovdv vv vv v ∆ oujk a[pappon ∆Idaivou purov". Ag.504: dekavtou se fevggei tw'/dv '/v '/v '/ ∆ ajfikovmhn e[tou", Ag.829: qeoi'" me;n ejxevteina froivmion tovdev vv vv v : Ag.867: ajnh;r o{dj ; {j ; {j ; { ∆ wJ" pro;" oi\kon wjceteuveto Ag.1186: th;n ga;r stevghn thvndv vv vv v ∆ ou[pot∆ ejkleivpei coro;" Ag.1603: ejk tw'ndev soi pesovnvvv ta tovndvvv ∆ ijdei'n pavra: Ag.1613: su; d∆ a[ndra tovnde[ v[ v[ v fh;/" eJkw;n kataktanei'n, Choe.256: patro;" neossou;" touvsd; v; v; v ∆ ajpofqeivra" povqen Choe.561: h{xw su;n ajndri; tw'/dj ; '/j ; '/j ; '/ ∆ ejf∆ eJrkeivou" puvla" Choe.669: oJpoi'avper dovmoisi toi'sdv 'v 'v ' ∆ ejpeikovta, Choe.685: qavptein, ejfetma;" tavsdej ; vj ; vj ; v povrqmeuson pavlin. Choe.1011: fa'ro" tovd' v' v' v ∆, wJ" e[bayen Aijgivsqou xivfo": Choe.1017: a[zhla nivkh" th'sdv 'v 'v ' ∆ e[cwn miavsmata. Eum.852: gh'" th'sd' '' '' ' ∆ ejrasqhvsesqe. prounnevpw tavde: Demonstrative > Noun 32: ii) 22 at the penthemimeral caesura: Ag.24: pollw'n ejn “Argei th'sde sumfora'"' '' '' ' cavrin. Ag.33: tri;" e}x balouvsh" th'sde''' v moi fruktwriva"vvv : Ag.619: h{kei su;n uJmi'n, th'sde gh'"' '' '' ' fivlon kravto". Ag.1182: kluvzein pro;" aujga;" tou'de ''' phvmato"vvv polu; Ag.1282: fuga;" d∆ ajlhvth" th'sde gh'"' '' '' ' ajpovxeno" Ag.1405: povsi", nekro;" dev, th'sde dexia'"' '' '' ' cero;";;; Ag.1583: ∆Atreu;" ga;r a[rcwn th'sde gh'"' '' '' ' , touvtou pathvr, Choe.85: ejpei; pavreste th'sde prostroph'"' '' '' ' ejmoi; Choe.129: kajgw; cevousa tavsde cevrnibav vv vv v " nekroi'" Choe.246: Zeu' Zeu', qewro;" tw'nde pragmavtwn' v' v' v genou', Choe.555: aijnw' de; kruvptein tavsde sunqhvka"v vv vv v ejmav", Choe.692: w\ duspavlaiste tw'nde dwmavtwn' v' v' v ∆Arav, Choe.718: bouleusovmesqa th'sde sumfora'"' '' '' ' pevri. Choe.761: ejgw; dipla'" de; tavsde ceirwnaxiva"v vv vv v Choe.1042: ejgw; d∆ ajlhvth" th'sde gh'"' '' '' ' ajpovxeno" Choe.1043: zw'n kai; teqnhkw;" tavsde klhdovna"v vv vv v lipw;n Eum.142: ijdwvmeq∆ ei[ ti tou'de froimivou' v' v' v mata'/. Eum.179: e[xw, keleuvw, tw'nde dwmavtwn' v' v' v tavco" 304 Eum.622: Zeuv", wJ" levgei" suv, tovnde crhsmo;nv ;v ;v ; w[pase Eum.630: yhvfw/ diairei'n tou'de pravgmato"' v' v' v pevri. Eum.761: swv/zei me, mhtro;" tavsde sundivkou"v vv vv v oJrw'n. Eum.800: uJmei'" d∆ mhvte th'/de gh'/'/ '/'/ '/'/ '/ baru;n kovton 10 positioned elsewhere Ag.17: u{pnou tovd∆ ajntivmolv j vv j vv j v pon ejntevmnwn a[ko", Ag.28: ojlolugmo;n eujfhmou'nta th'/de lampavdi'/ v'/ v'/ v Ag.1271: kajn toi'sde kovsmoi"' v' v' v katagelwmevnhn ÊmevtaÊ Ag.1635: dra'sai tovd∆ e[rgonv [v [v [ oujk e[tlh" aujtoktovnw". Choe.92: cevousa tovnde pelano;nv ;v ;v ; ejn tuvmbw/ patrov". Choe.1035: xu;n tw'/de qallw'/ '/'/ '/ kai; stevfei prosivxomai Choe.1038: feuvgwn tovd∆ ai|mav |v |v | koinovn: oujd∆ ejf∆ eJstivan Eum.195: ejn toi'sde plhsivoisi' v' v' v trivbesqai muvso". Eum.278–9: siga'n q∆ oJmoivw": ejn de; tw'/de pravgmati'/ v'/ v'/ v fwnei'n ejtavcqhn pro;" sofou' didaskavlou. Eum.613: dokei', tovd∆ ai|mav |v |v | kri'non, wJ" touvtoi" fravsw. 305 Appendix 3 Complementation 3A: Homeric o{ti and wJ" Page 306 3B: Complements in tragedy and comedy 319 3C: Complements in prose 332 306 3A: Homeric o{ti{{{ and wJ"JJJ ”Oti Specifying Il.2.254–6: tw; nu'n ∆Atrei?dh/ ∆Agevmnoni poimevni law'n h|sai ojneidivzwn, o{ti{{{ oiJ mavla polla; didou'sin h{rwe" Danaoiv: ........ Il.9.75–7: ........mavla de; crew; pavnta" ∆Acaiou;" ejsqlh'" kai; pukinh'", o{ti{{{ dhvi>oi ejgguvqi nhw'n kaivousin pura; pollav: ...... Il.16.34–5 : ..........glaukh; dev se tivkte qavlassa pevtrai tæ hjlivbatoi, o{ti{{{ toi novo" ejsti;n ajphnhv". Il.21.410–1: ......oujdev nuv pwv per ejpefravsw o{sson ajreivwn eu[comæ ejgw;n e[menai, o{ti{{{ moi mevno" ijsofarivzei". Il.21.487–8: ................... o[fræ eju÷ eijdh'/" o{sson fertevrh ei[mæ, o{ti{{{ moi mevno" ajntiferivzei". Il.23.483–4: Ai\an nei'ko" a[riste kakofrade;" a[llav te pavnta deuveai ∆Argeivwn, o{ti{{{ toi novo" ejsti;n ajphnhv". Il.24.239–40: ........... ou[ nu kai; uJmi'n oi[koi e[nesti govo", o{ti{{{ mæ h[lqete khdhvsonte"… Il.24.241: h\ ojnovsasq∆ o{ti{{{ moi Kronivdh" Zeu;" a[lge∆ e[dwke Od.14.53–4: ÆZeuv" toi doivh, xei'ne, kai; ajqavnatoi qeoi; a[lloi, o{tti mavlistæ ejqevlei", o{ti {{{ me provfrwn uJpevdexo.Æ Od.18.333=393: h\ ajluvei" o{ti{{{ «Iron ejnivkhsa" to;n ajlhvthn… Od.22.35–6: Æw\ kuvne", ou[ mæ e[tæ ejfavskeqæ uJpovtropon oi[kade nei'sqai dhvmou a[po Trwvwn, o{ti{{{ moi katekeivrete oi\kon 307 Verbs of emotion Il.1.56: khvdeto ga;r Danaw'n, o{ti{{{ rJa qnhvskonta" oJra'to. Il.5.325–6: dw'ke de; Dhi>puvlw/ eJtavrw/ fivlw/, o}n peri; pavsh" ti'en oJmhlikivh" o{ti{{{ oiJ fresi;n a[rtia h[/dh, Il.23.555–6: ’W" favto, meivdhsen de; podavrkh" di'o" ∆Acilleu;" caivrwn ∆Antilovcw/, o{ti{{{ oiJ fivlo" h\en eJtai'ro": Il.24.113–5=134–6: ..... ejme; d∆ e[xoca pavntwn ajqanavtwn kecolw'sqai, o{ti{{{ fresi; mainomevnh/sin ”Ektor∆ e[cei para; nhusi; korwnivsin oujd∆ ajpevlusen Od.5.339–40: kavmmore, tivpte toi w|de Poseidavwn ejnosivcqwn wjduvsatæ ejkpavglw", o{ti{{{ toi kaka; polla; futeuvei… Od.8.237–9: ajllæ ejqevlei" ajreth;n sh;n fainevmen, h{ toi ojphdei', cwovmeno", o{ti{{{ sæ ou|to" ajnh;r ejn ajgw'ni parasta;" neivkesen, ...... Od.11.103=13.343: cwovmeno" o{ti{{{ oiJ uiJo;n fivlon ejxalavwsa". Od.17.377–8: h\ o[nosai, o{ti{{{ toi bivoton katevdousin a[nakto" ejnqavd∆ ajgeirovmenoi, su; de; kaiv poqi tovnd∆ ejkavlessa"… Od.19.71–2: daimonivh, tiv moi w|d∆ ejpevcei" kekothovti qumw'/… h\ o{ti{{{ dh; rJupovw, kaka; de; croi÷ ei{mata ei|mai, Od.19.247–8: ............. tiven dev min e[xocon a[llwn w|n eJtavrwn ∆Oduseuv", o{ti{{{ oiJ fresi;n a[rtia h[/dh. 308 Cognitive verbs Il.1.536–8: wJ" o} me;n e[nqa kaqevzet∆ ejpi; qrovnou: oujdev min {Hrh hjgnoivhsen ijdou's∆ o{ti{{{ oiJ sumfravssato boula;" ∆Argurovpeza Qevti", .... Il.6.230–1: ... o[fra kai; oi}de gnw'sin o{ti{{{ xei'noi patrwvi>oi eucovmeq∆ ei\nai. Il.7.448: oujc oJrava/" o{ti{{{ d∆ au\te kavrh komovwnte" ∆Acaioi; tei'co" ejteicivssanto Il.8.175: gignwvskw d∆ o{ti{{{ moi provfrwn katevneuse Kronivwn nivkhn kai; mevga ku'do" ... Il.15.217: i[stw tou'q' '' ∆ o{ti{{{ nw'i>n ajnhvkesto" covlo" e[stai. Il.17.630: gnoivh o{ti{{{ Trwvessi path;r Zeu;" aujto;" ajrhvgei. Il.17.688: h[dh me;n se; kai; aujto;n oji?omai eijsorovwnta gignwvskein o{ti{{{ ph'ma qeo;" Danaoi'si kulivndei, Il.17.641–2: ..... ejpei; ou[ min oji?omai oujde; pepuvsqai lugrh'" ajggelivh", o{ti{{{ oiJ fivlo" w[leqæ eJtai'ro". Il.20.434: oi\da d∆ o{ti{{{ su; me;n ejsqlov", ejgw; de; sevqen polu; ceivrwn. Il.23. 545: ta; fronevwn o{ti {{{ oiJ blavben a{rmata kai; tacev∆ i{ppw Il.23.576–8: ∆Antivlocon yeuvdessi bihsavmeno" Menevlao" oi[cetai i{ppon a[gwn, o{ti{{{ oiJ polu; ceivrone" h\san i{ppoi, ...... Il.24.592–4: mhv moi Pavtrokle skudmainevmen, ai[ ke puvqhai eijn “Aidov" per ejw;n o{ti{{{ ”Ektora di'on e[lusa patri; fivlw/, ejpei; ou[ moi ajeikeva dw'ken a[poina. 309 Od.8.461–2: cai're, xei'n∆, i{na kaiv pot∆ ejw;n ejn patrivdi gaivh/ mnhvsh/ ejmei'∆, o{ti{{{ moi prwvth/ zwavgri∆ ojfevllei". Od.13.314: tou'to d∆ ejgw;n eu\ oi\d∆, o{ti{{{ moi pavro" hjpivh h\sqa Od.17.269: gignwvskw d∆, o{ti{{{ polloi; ejn aujtw'/ dai'ta tivqentai Od. 18.11: oujk aji?ei", o{ti{{{ dhv moi ejpillivzousin a{pante" Speech verb Od.16.130–1: a[tta, su; d∆ e[rceo qa'sson, ejcevfroni Phnelopeivh/ ei[f∆, o{ti{{{ oiJ sw'" eijmi kai; ejk Puvlou eijlhvlouqa. ”Otti Specifying Il.5.349: h\ oujc a{li" o{tti{{{ gunai'ka" ajnavlkida" hjperopeuvei"… Il.15.226–8: ajlla; tovdæ hjme;n ejmoi; polu; kevrdion hjdev oiJ aujtw'/ e[pleto, o{tti{{{ pavroiqe nemesshqei;" uJpoveixe cei'ra" ejmav", ....... Il.23.670: h\ oujc a{li" o{tti{{{ mavch" ejpideuvomai… ...... Verbs of emotion Il.5.421: Zeu' pavter h\ rJav tiv moi kecolwvseai o{tti{{{ ken ei[pw… Il.14.406–7: ........... cwvsato dæ ”Ektwr, o{tti{{{ v rJav oiJ bevlo" wjku; ejtwvsion e[kfuge ceirov", Il.15.155–6: .... oujdev sfwi>n ijdw;n ejcolwvsato qumw'/, o{tti{{{ v oiJ w\kæ ejpevessi fivlh" ajlovcoio piqevsqhn. Il.16.530–1: Glau'ko" dæ e[gnw h|/sin ejni; fresi; ghvqhsevn te 310 o{tti{{{ v oiJ w\kæ h[kouse mevga" qeo;" eujxamevnoio. Il.17.567–8: ’W" favto, ghvqhsen de; qea; glaukw'pi" ∆Aqhvnh, o{tti{{{ v rJav oiJ pavmprwta qew'n hjrhvsato pavntwn. Il.22.291–2: .............. cwvsato dæ ”Ektwr o{ttiv{ v{ v{ v rJav oiJ bevlo" wjku; ejtwvsion e[kfuge ceirov" Od.14.51–2 (with both conjunctions): ....... cai're dæ ∆Odusseuv", o{tti{{{ min w}"}}} uJpevdekto, e[po" tæ e[fatæ e[k tæ ojnovmaze: Od.14.366–7: .............. o{ tæ h[cqeto pa'si qeoi'si pavgcu mavlæ, o{tti{{{ min ou[ ti meta; Trwvessi davmassan Od.14.526–7: ............. cai're dæ ∆Odusseuv", o{tti{{{ rJav oiJ biovtou perikhvdeto novsfin ejovnto". Od.21.414–5: ghvqhsevn tæ a[ræ e[peita poluvtla" di'o" ∆Odusseuv", o{tti{{{ rJav oiJ tevra" h|ke Krovnou pavi>" ajgkulomhvtew. Od.23.115–6: nu'n dæ o{tti{{{ rJupovw, kaka; de; croi÷ ei{mata ei|mai, tou[nek[[[ æ ajtimavzei me kai; ou[ pwv fhsi to;n ei\nai. Cognitive verbs Il.5.406–7: nhvpio", oujde; to; oi\de kata; frevna Tudevo" uiJo;" o{tti{{{ mavlæ ouj dhnaio;" o}" ajqanavtoisi mavchtai, Il.11.408: oi\da ga;r o{tti{{{ kakoi; me;n ajpoivcontai polevmoio, Il.13.674–6: ”Ektwr dæ oujk ejpevpusto Dii÷ fivlo", oujdev ti h[/dh o{ttiv{ v{ v{ v rJav oiJ nhw'n ejpæ ajristera; dhi>ovwnto laoi; uJpæ ∆Argeivwn. ..... Il.24.56 3–4: kai; dev se gignwvskw Privame fresivn, oujdev me lhvqei", o{tti{{{ qew'n tiv" sæ h\ge qoa;" ejpi; nh'a" ∆Acaiw'n. Od.10.44: Ai[olo". ajllæ a[ge qa'sson ijdwvmeqa, o{tti{{{ tavdæ ejstivn, 311 Speech verbs Il.17.410–1: dh; tovte gæ ou[ oiJ e[eipe kako;n tovson o{sson ejtuvcqh mhvthr, o{tti{{{ v rJav oiJ polu; fivltato" w[leqæ eJtai'ro". Il.17.654–5: o[trunon dæ ∆Acilh'i> dai?froni qa'sson ijovnta eijpei'n o{tti{{{ rJav oiJ polu; fivltato" w[leqæ eJtai'ro". Il.22.438–9: ”Ektoro": ouj gavr oi{ ti" ejthvtumo" a[ggelo" ejlqw;n h[ggeilæ o{tti{{{ v rJav oiJ povsi" e[ktoqi mivmne pulavwn, Od.4.391–2: kai; dev kev toi ei[ph/si, diotrefev", ai[ kæ ejqevlh/sqa, o{tti{{{ toi ejn megavroisi kakovn tæ ajgaqovn te tevtuktai Od.9.402: iJstavmenoi dæ ei[ronto peri; spevo", o{tti{{{ eJ khvdoi: Od.19.463–4: cai'ron nosthvsanti kai; ejxereveinon e{kasta, oujlh;n o{tti{{{ pavqoi:... Relative use Il.1.294: eij dh; soi; pa'n e[rgon uJpeivxomai o{ttiv{ v{ v{ v ken ei[ph/": Il.1.542–3: ........ oujdev tiv pwv moi provfrwn tevtlhka" eijpei'n e[po" o{tti{{{ nohvsh/". Il.2.361: ou[ toi ajpovblhton e[po" e[ssetai o{tti{{{ v ken ei[pw: Il.6.176–7: kai; tovte min ejreveine kai; h[/tee sh'ma ijdevsqai o{tti{{{ v rJav oiJ gambroi'o pavra Proivtoio fevroito. Il.8.408 (=422): aijei; gavr moi e[wqen ejnikla'n o{ttiv{ v{ v{ v ken ei[pw. Il.14.190: h\ rJav nuv moiv ti pivqoio fivlon tevko" o{ttiv{ v{ v{ v ken ei[pw, Il.15.109: tw; e[ceqæ o{tti{{{ v ken u[mmi kako;n pevmph/sin eJkavstw/. 312 Il15.148: aujta;r ejph;n e[lqhte, Diov" tæ eij" w\pa i[dhsqe, e{rdein o{tti{{{ v ke kei'no" ejpotruvnh/ kai; ajnwvgh/. Il.18.63–4: ajllæ ei\mæ, o[fra i[dwmi fivlon tevko", hjdæ ejpakouvsw o{tti{{{ v min i{keto pevnqo" ajpo; ptolevmoio mevnonta. Il.22.73: kei'sqai: pavnta de; kala; qanovnti per o{tti{{{ fanhvh/: Il.24.92: ei\mi mevn, oujdæ a{lion e[po" e[ssetai o{ttiv{ v{ v{ v ken ei[ph/. Od.1.158: xei'ne fivlæ, h\ kaiv moi nemeshvseai o{tti{{{ ken ei[pw… Od.1.316–7: dw'ron dæ o{tti{{{ kev moi dou'nai fivlon h\tor ajnwvgh/, au\ti" ajnercomevnw/ dovmenai oi\kovnde fevresqai, Od.1.389: ∆Antivnoæ, ei[ pevr moi kai; ajgavsseai o{tti{{{ ken ei[pw, Od.2.25=2.161–2.229=24.454: kevklute dh; nu'n meu, ∆Iqakhvsioi, o{tti{{{ ken ei[pw. Od.4.600: dw'ron dæ, o{tti{{{ kev moi dwvh/", keimhvlion e[stw: Od.8.548–9: tw' nu'n mhde; su; keu'qe nohvmasi kerdalevoisin, o{tti{{{ kev sæ ei[rwmai: favsqai dev se kavlliovn ejstin. Od.8.550: ei[pæ o[nomæ, o{tti{{{ se kei'qi kavleon mhvthr te pathvr te, Od.14.53–4: Zeuv" toi doivh, xei'ne, kai; ajqavnatoi qeoi; a[lloi, o{tti{{{ mavlistæ ejqevlei", o{ti{{{ me provfrwn uJpevdexo. Od.14.444–5: oi|a pavresti: qeo;" de; to; me;n dwvsei, to; dæ ejavsei, o{tti{{{ ken w|/ qumw'/ ejqevlh/: duvnatai ga;r a{panta.Æ Od.15.317: ai\yav ken eu\ drwvoimi meta; sfivsin, o{tti{{{ qevloien. Od.18.112–3: Zeuv" toi doivh, xei'ne, kai; ajqavnatoi qeoi; a[lloi, o{tti{{{ mavlistæ ejqevlei" kaiv toi fivlon e[pleto qumw'/, 313 Od.18.142: ajllæ o{ ge sigh'/ dw'ra qew'n e[coi, o{tti{{{ didoi'en. Od.19.378: ............ ajllæ a[ge nu'n xunivei e[po", o{tti {{{ ken ei[pw: Od.19.403–4: Aujtovlukæ, aujto;" nu'n o[nomæ eu{reo, o{tti{{{ ke qei'o paido;" paidi; fivlw/: Od.19.406: gambro;" ejmo;" quvgatevr te, tivqesqæ o[nomæ, o{tti{{{ ken ei[pw: Od.20.115: krh'non nu'n kai; ejmoi; deilh'/ e[po", o{tti{{{ ken ei[pw: Od.23.139–140: ........................ e[nqa dæ e[peita frassovmeqæ, o{ttiv{ v{ v{ v ke kevrdo" ∆Oluvmpio" ejggualivxh/.Æ Sentence adverbs Il.9.102: eijpei'n eij" ajgaqovn: sevo dæ e{xetai o{tti{{{ v ken a[rch/. Od.17.316–7: ouj me;n gavr ti fuvgeske baqeivh" bevnqesin u{lh" knwvdalon, o{tti{{{ divoito: ... Final/causal use Il.24.538–9: ajllæ ejpi; kai; tw'/ qh'ke qeo;" kakovn, o{tti{{{ v oiJ ou[ ti paivdwn ejn megavroisi gonh; gevneto kreiovntwn, Od.14.440–1: ai[qæ ou{tw", Eu[maie, fivlo" Dii; patri; gevnoio wJ" ejmoiv, o{tti{{{ me toi'on ejovntæ ajgaqoi'si geraivrei". Od.15.341–2: ai[qæ ou{tw", Eu[maie, fivlo" Dii; patri; gevnoio wJ" ejmoiv, o{tti{{{ mæ e[pausa" a[lh" kai; oji>zuvo" aijnh'". 314 Adverbial Il.4.192–3: «H kai; Talquvbion qei'on khvruka proshuvda: Talquvbiæ o{tti {{{ tavcista Macavona deu'ro kavlesson Il.9.658–9: Pavtroklo" dæ eJtavroisin ijde; dmwh'/si kevleuse Foivniki storevsai pukino;n levco" o{tti{{{ tavcista. Il.15.146: Zeu;" sfw; eij" “Idhn kevletæ ejlqevmen o{tti{{{ tavcista: Il.22.129: bevlteron au\tæ e[ridi xunelaunevmen o{tti{{{ tavcista: Il.23.71: qavptev me o{tti{{{ tavcista puvla" ∆Ai?dao perhvsw. Il.23.403: e[mbhton kai; sfw'i>: titaivneton o{tti{{{ tavcista. Il.23.414: ajllæ ejfomartei'ton kai; speuvdeton o{tti{{{ tavcista: Od.5.112: to;n nu'n sæ hjnwvgein ajpopempevmen o{tti{{{ tavcista: Od.8.433–4: w}" e[fatæ, ∆Arhvth de; meta; dmw/h'/sin e[eipen ajmfi; puri; sth'sai trivpoda mevgan o{tti {{{ tavcista. Od.16.151–3: ......... ajta;r pro;" mhtevra eijpei'n ajmfivpolon tamivhn ojtrunevmen o{tti{{{ tavcista kruvbdhn: ... ”O ti Speech verbs Il.1.64: o{" k∆ ei[ph/ o{ ti{{{ tovsson ejcwvsato Foi'bo" ∆Apovllwn Il.1.85: qarshvsa" mavla eijpe; qeoprovpionvvv o{ ti oi\sqa Il.14.195 (=18.426=Od.5.89): au[da o{ ti{{{ fronevei" .... 315 Od.8.577: eijpe; d∆ o{ ti klaivei" ..... Relative use Il.1.527: oujd∆ ajteleuvthtonj vj vj v , o{ ti ken kefalh'/ kataneuvsw Il.2.215: ajll∆ o{ ti oiJ ei[saito geloivi>on ∆Argeivoisin e[mmenai Il.10.142: nuvkta di∆ ajmbrosivhn, o{ ti dh; creiw; tovson i{kei… Il.10.503: Aujta;r o} mermhvrize mevnwn o{ ti kuvntaton e{rdoi Il.14.220–1: .............. oujdev sev fhmi a[prhktovn[ v[ v[ v ge nevesqai, o{ ti{{{ fresi;n h\/si menoina/'. Od.2.33–4: .............. ei[qe oiJ aujtw'/ Zeu;" ajgaqo;nj ;j ;j ; televsaien, o{ ti{{{ fresi;n h\/si menoina/'. Od.7. 150: kthvmat∆ ejni; megavroisi gevra"vvv q∆, o{ ti dh'mo" e[dwken Od.8.147–8: ouj me;n ga;r mei'zon klevo"vvv ajnevro", o[fra ken h\/sin h] o{ ti possivn te rJevzh/ kai; cersi;n eJh'/sin Od.12.331: ijcqu'" o[rniqav" te, fivla" o{ ti cei'ra" i{koito 316 Completive ÔW": constructions cited by Monteil (1963, 354) Verbs of emotion Il.16.17: h\e suv gæ ∆Argeivwn ojlofuvreai, wJ"JJJ ojlevkontai nhusi;n e[pi glafurh'/sin uJperbasivh" e{neka sfh'"… Il.16.599–600: douvphsen de; peswvn: pukino;n dæ a[co" e[llabæ ∆Acaiouv", wJ"JJJ e[pesæ ejsqlo;" ajnhvr: mevga de; Trw'e" kecavronto, Il.23.648: tou'to dæ ejgw; provfrwn devcomai, caivrei dev moi h\tor, w{"{{{ meu ajei; mevmnhsai ejnhevo", oujdev se lhvqw, Od.9.413: w}" a[ræ e[fan ajpiovnte", ejmo;n dæ ejgevlasse fivlon kh'r, wJ"JJJ o[nomæ ejxapavthsen ejmo;n kai; mh'ti" ajmuvmwn. Od.11.418: ajllav ke kei'na mavlista ijdw;n ojlofuvrao qumw'/, wJ"JJJ ajmfi; krhth'ra trapevza" te plhqouvsa" keivmeqæ ejni; megavrw/, Cognitive verbs Il.4.360: oi\da ga;r w{"{{{ toi qumo;" ejni; sthvqessi fivloisin h[pia dhvnea oi\de: ta; ga;r fronevei" a{ tæ ejgwv per. Il.10.160: oujk aji?ei" wJ"JJJ Trw'e" ejpi; qrwsmw'/ pedivoio ei{atai a[gci new'n, ojlivgo" dæ e[ti cw'ro" ejruvkei… Il.14.482: fravzesqæ wJ"JJJ uJmi'n Provmaco" dedmhmevno" eu{dei e[gcei ejmw'/, Il.15.204: oi\sqæ wJ"JJJ presbutevroisin ∆Erinuve" aije;n e{pontai. Il.23.787: eijdovsin u[mmæ ejrevw pa'sin fivloi, wJ"JJJ e[ti kai; nu'n ajqavnatoi timw'si palaiotevrou" ajnqrwvpou". 317 Il.24.662: oi\sqa ga;r wJ"JJJ kata; a[stu ejevlmeqa, thlovqi dæ u{lh ajxevmen ejx o[reo", mavla de; Trw'e" dedivasin. Od.3.193–4: ∆Atrei?dhn de; kai; aujtoi; ajkouvete novsfin ejovnte", w{"{{{ tæ h\lqæ w{"{{{ tæ Ai[gisqo" ejmhvsato lugro;n o[leqron. Speech verbs Od.4.376: ejk mevn toi ejrevw, h{ ti" suv pevr ejssi qeavwn, wJ"JJJ ejgw; ou[ ti eJkw;n kateruvkomai, .... Od.8.75: nei'ko" ∆Odussh'o" kai; Phlei?dew ∆Acilh'o", w{"{{{ pote dhrivsanto qew'n ejn daiti; qaleivh/ ejkpavgloisæ ejpevessin, Od.8.266: aujta;r oJ formivzwn ajnebavlleto kalo;n ajeivdein ajmfæ “Areo" filovthto" eju>stefavnou tæ ∆Afrodivth", wJ"JJJ ta; prw'tæ ejmivghsan ejn ÔHfaivstoio dovmoisi lavqrh/: Od.8.307: deu'qæ, i{na e[rgæ ajgevlasta kai; oujk ejpieikta; i[dhsqe, wJ"JJJ ejme; cwlo;n ejovnta Dio;" qugavthr ∆Afrodivth aije;n ajtimavzei, Od.8.497: aujtivka kai; pa'sin muqhvsomai ajnqrwvpoisin, wJ"JJJ a[ra toi provfrwn qeo;" w[pase qevspin ajoidhvn.Æ Od.15.157: nosthvsa" ∆Iqavkhnde kicw;n ∆Odush'æ ejni; oi[kw/ ei[poimæ, wJ"JJJ para; sei'o tucw;n filovthto" aJpavsh" Od.19.464: oujlh;n o{tti pavqoi: oJ dæ a[ra sfivsin eu\ katevlexen, w{"{{{ min qhreuvontæ e[lasen su'" leukw'/ ojdovnti Parnhsovndæ ejlqovnta su;n uiJavsin Aujtoluvkoio. Od.22.373: o[fra gnw'/" kata; qumovn, ajta;r ei[ph/sqa kai; a[llw/, wJ"JJJ kakoergivh" eujergesivh mevgæ ajmeivnwn. 318 Od.24.236: kuvssai kai; perifu'nai eJo;n patevræ hjde; e{kasta eijpei'n, wJ"JJJ e[lqoi kai; i{koitæ ej" patrivda gai'an, Completive wJ"JJJ in Iliad 9 and Odyssey 9 Verb of emotion Od.9.413–4: ....... ejmo;n d∆ ejgevlasse fivlon kh'r, wJ"JJJ o[nom∆ ejxapavthsen ejmo;n kai; mh'ti" ajmuvmwn. Cognitive verbs Il.9.112: frazwvmesq∆ w{"{{{ kevn min ajressavmenoi pepivqwmen Il.9.527–8: mevmnhmai tovde e[rgonv [v [v [ ejgw; pavlai ou[ ti nevon ge wJ"JJJ h\n: ....... Il.9.647: mnhvsomai w{"{{{ m∆ ajsuvfhlon ejn ∆Argeivoisin e[rexen ∆Atreiv>dh" ....... Il.9.704 : ajll∆ a[geq∆ wJ"JJJ a]n ejgw; ei[pw peiqwvmeqa pavnte": Od.9.442–3: ........... to; de; nhvpio" oujk ejnovhsen, w{"{{{ oiJ uJp∆ eijropovkwn ojiv>wn stevrnoisi devdento. Speech verbs Il.9.103 =314 : aujta;r ejgw;n ejrevw w{"{{{ moi dokei' ei\nai a[rista. Il.9.369: ........ tw'/ pavntvvv ∆ ajgoreuevmen wJ"JJJ ejpitevllw ajmfadovn, ........ 319 Final use Il.9.181: peira'n wJ"JJJ pepivqoien ajmuvmona Phleiv>wna. Il.9.309–311: crh; me;n dh; to;n mu'qon ajphlegevw" ajpoeipei'n, h|/ per dh; fronevw te kai; wJ"JJJ tetelesmevnon e[stai, wJ"JJJ mhv moi truvzhte parhvmenoi a[lloqen a[llo". Od.9.42 =549: dassavmeq∆, wJ"JJJ mhv tiv" moi ajtembovmeno" kivoi i[sh". 3B: Complements in tragedy and comedy Aeschylus ÔW" in the Oresteia 22 instances introduce resultative or final clauses, usually after verbs expressing actions (Ag.358, 575, 665, 911, 1188, 1293, 1381; Choe.20, 556, 735, 767, 771, 984, 987, 1021; Eum.36, 613, 638, 771, 799, 882, 895). There are 11 instances of complementizing use, 7 with antecedents, and two dependent on a nominal. The importance of verbs of witness is discussed in the main text. Most other completives involve cognitive verbs. Ag.494–6: ... marturei' dev moi kavsi" phlou' xuvnouro" diyiva kovni" tavdevvv , wJ"JJJ ou[t∆ a[naudo" ou[tev soi daivwn flovga u{lh" ojreiva" shmanei' kapnw'/ purov": Ag.1367: manteusovmesqa tajndro;"j ;j ;j ; wJ"JJJ ojlwlovto"… Ag.1505: wJ"JJJ me;n ajnaivtio" ei\ tou'de fovnou tiv" oJ marturhvswnJ vJ vJ v … Ag.1619: gnwvsh/ gevrwn w]n wJ"JJJ didavskesqai baru; tw'/ thlikouvtw/, swfronei'n eijrhmevnon: 320 Choe.492 (corrupt): mevmnhso d∆ ajmfivblhstronj vj vj v wJ"JJJ ejkaivnisa". Choe.988–9: wJ" a]n parh'/ moi mavrtu"vvv ejn divkh/ pote; wJ"JJJ tovnd∆ ejgw; meth'lqon ejndivkw" movron Choe.1034: kai; nu'n oJra'tev m∆, wJ"JJJ pareskeuasmevno" xu;n tw'/de qallw'/ kai; stevfei prosivxomai mesovmfalovn q∆ i{druma, Loxivou pevdon, Eum.310–1: levxai te lavchvvv ta; kat∆ ajnqrwvpou" wJ"JJJ ejpinwma'/ stavsi" aJmhv. Eum.454: gevno"vvv de; toujmo;nj ;j ;j ; wJ"JJJ e[cei peuvsh/ tavca. Eum.657: kai; tou'to''' levxw, kai; mavq∆ wJ"JJJ ojrqw'" ejrw': There is one appositive construction, at Ag.1464–6: mhd∆ eij" ÔElevnhn kovton ejktrevyh/" wJ" ajndrolevteir∆, wJ"JJJ miva pollw'n ajndrw'n yuca;" Danaw'n ojlevsas∆ ajxuvstaton a[lgo" e[praxen. 321 ÔW" in other Aeschylean works There is a total of 101 (134 including Prometheus), with 12 (or 21) introducing complements. Of these, 7 follow verbs of speech and 4 of perception, and 1 of showing (deivxaq∆). Supp 390–1: dei' toiv se feuvgein kata; novmou"; v; v; v tou;" oi[koqen, wJ"JJJ oujk e[cousin ku'ro" oujde;n ajmfi; sou' Pers.287–9: memnh'sqaiv toi pavra wJ"JJJ Persivdwn polla;" mavtan eu[nida" e[ktissan hjd∆ ajnavndrou" Pers.356–7: ejlqw;n e[lexe paidi; sw/' Xevrxh/ tavdevvv : wJ"JJJ eij melaivnh" nukto;" i{xetai knevfa" ”Ellhne" ouj menoi'en .... Pers.525: ejpivstamai me;n wJ"JJJ ejp∆ ejxeirgasmevnoi", ajll∆ ej" to; loipo;n ei[ ti dh; lw'/on pevloi. Pers.599–600: ejpivstatai brotoi'sin wJ"JJJ , o{tan kluvdwn kakw'n ejpevlqh/, pavnta deimaivnein fivlei', Pers.754: .. levgousi d∆ wJ"JJJ su; me;n mevgan tevknoi" plou'ton ejkthvsw su'n aijcmh/' .... Pers.819–20: a[fwna shmanou'sin o[mmasin brotw'n wJ"JJJ oujc uJpevrfeu qnhto;n o[nta crh; fronei'n: Sept 176: deivxaq∆ wJ"JJJ filopovlei" Sept 375–6: levgoim∆ a]n eijdw;" eu\ ta;; ;; tw'n ejnantivwn, wJ"JJJ t∆ ejn puvlai" e{kasto" eijlhcen pavlon. Sept 468–9: boa'/ de; cou\to" grammavtwn ejn xullabai'" wJ"JJJ oujd∆ a]n “Arh" sf∆ ejkbavloi purgwmavtwn. Sept 617: ajll∆ oi\den w{"{{{ sfe crh; teleuth'sai mavch/ Sept 922–3: pavresti d∆ eijpei'n ejp∆ ajqlivoisinj j vj j vj j v wJ"JJJ ejrxavthn polla; me;n polivta" Prometheus 5 of the 9 constructions follow verbs of speech. 211–3: to; mevl; v; v; v lon h|/ kranoi'to prouteqespivkei, wJ"JJJ ouj kat∆ ijscu;n oujde; pro;" to; kartero;n creivh 259–61: ........ oujc oJra'/" o{ti h{marte"… wJ"JJJ d∆ h{marte", ou[t∆ ejmoi; levgein kaq∆ hJdonh;n soiv t∆ a[lgo" .... 296–7: ouj gavr pot∆ ejrei'" wJ"JJJ ∆Wkeanou' 322 fivlo" ejsti; bebaiovterov" soi. 442–4: ..... tajn brotoi'" de; phvmatavvv ajkouvsaq∆, w{"{{{ sfa" nhpivou" o[nta" to; pri;n e[nnou" e[qhka kai; frenw'n ejphbovlou" 842–3: shmei'av soi tavd∆ ejsti; th'" ejmh'" frenov", wJ"JJJ devrketai plevon ti tou' pefasmevnou. 889–90: glwvssa/ diemuqolovghsen, wJ"JJJ to; khdeu'sai kaq∆ eJauto;n ajristeuvei makrw'/, 1002–3: eijselqevtw se mhvpoq∆ wJ"JJJ ejgw; Dio;" gnwvmhn fobhqei;" qhluvnou" genhvsomai 1073–5: ..... mhdev pot∆ ei[phq∆ wJ"JJJ Zeu;" uJma'" eij" ajprovopton ph'm∆ eijsevbalen.... 1093: ejsora'/" m∆ wJ"JJJ e[kdika pavscw. Eij jjj in Aeschylus Ag. 618: su; dæ eijpev, kh'rux, Menevlewnvvv de; peuvqomai, eijj jj novstimov" te kai; sesw/mevno" pavlin h{xei su;n uJmi'n, th'sde gh'" fivlon kravto". Ag. 881–4: Strofivo" oJ Fwkeuv", ajmfivlekta phvmata ejmoi; profwnw'n, tovn qæ uJpæ ∆Ilivw/ sevqen kivndunon, ei[[ [[ te dhmovqrou" ajnarciva boulh;n katarrivyeien, Choe.105: levgoi" a[n, ei[[ [[ ti tw'ndæ e[cei" uJpevrteron. Choe.668: xevnoi, levgoitæ a]n ei[[ [[ ti dei':...... Choe.755–7: ouj gavr ti fwnei' pai'" e[tæ w]n ejn spargavnoi", eij limov", h] divyh ti", h] liyouriva e[cei: Choe.851–3: ijdei'n ejlevgxaiv t∆ eu\ qevlw to;n a[ggelon, ei[t[[[ ∆ aujto;" h\n qnhv/skonto" ejgguvqen parwvn ei[t[[[ ∆ ejx ajmaura'" klhdovno" levgei maqwvn: Choe. 890: eijdw'men eijj jj nikw'men, h] nikwvmeqa. Eum.142: ijdwvmeqæ ei[ ti tou'de froimivou mata'/. Eum.269–70: o[yh/ de; kei[[ [[ ti" a[llo" h[liten brotw'n h] qeo;n h] xevnon tin∆ ajsebw'n Eum. 587: th;n mhtevræ eijpe; prw'ton eijj jj katevktona". Eum. 609–10: h[dh su; martuvrhson, ejxhgou' dev moi, “Apollon, ei[ [[[ sfe su;n divkh/ katevktanon. 323 Prom 997: o{ra nun ei[[ [[ soi tau'tæ ajrwga; faivnetai. {Opw"{{{ There are 5 instances of subordinating o{pw" in Aeschylus, always with an adverbial force, and all depending on verbs of knowing or saying. Two have a connotation of possibility: the complex introduction to the Parodos at Ag.105ff., and the idiomatic oujk e[sq∆ o{pw" at Ag.620, where the modality is expressed through an optative subordinate verb. There is a possible adverbial connotation, as also with wJ", which Liddell and Scott (1968: 1243) note at Prom.1001 (ku'm∆ o{pw"). Supp 289–290: ... didacqei;" d∆ a]n tovdvvv ∆ eijdeivhn plevon, o{pw"{{{ gevneqlon spevrma t∆ ∆Argei'on to; sovn. Ag.105–10: ......... e[ti ga;r qeovqen katapneivei Peiqwv, Êmolpa;n ajlka;nÊ suvmfuto" aijwvn: o{pw"{{{ ∆Acaiw'n divqronon kravto", ÔEllavdo" h{ba" xuvmfrona tagavn, pevmpei xu;n dori; kai; ceri; pravktori qouvrio" o[rni" Teukrivd∆ ejp∆ ai\an, Ag.620: oujk e[sq∆ o{pw"{{{ levxaimi ta; yeudh' kalav, Eum.591: eijpei'n ge mevntoi dei' s∆ o{pw"{{{ katevktane". The word order at Ag.1370–1 is described by Denniston and Page (1957: 195) as 'an incoherence of language without parallel or proper explanation.' It might, however, also be interpreted as a regular participle construction, with adverbial o{pw": tauvthn ejpainei'n pavntoqen plhquvnomai, tranw'" ∆Atreivdhn eijdevnai kurou'nq∆ o{pw"{{{ . Prom 640: oujk oi\d∆ o{pw"{{{ uJmi'n ajpisth'saiv me crhv Prom 939–940: dravtw, krateivtw tovnde to;n bracu;n crovnon o{pw"{{{ qevlei: ... 324 Mhvvvv in the Oresteia There are 103 instances of mhv, most of which are with imperative, optative, or non-finite verbs. None are complements. Sophocles ”Oti There are 30 instances with complement function in the extant works, 18 without explicit subject (Tr. 439 and 1110; Ant.61, 98, 276, 311, 779; Aj.678 and 792; OT.59 and 1133; El.44, 332, 988, and possibly 1070; Phil.405; OC.872 and 1039). There are 10 instances of SV (Tr. 464, Ant.188–9, 325–6, 649–50, 1043–4, OT.499–500 and 525–6, Ph.325–6, 649–50, OC.941–2), 1 SVS (OC.666–7) , and 1 of finite VS, at El.426–7: pleivw de; touvtwn ouj kavtoida, plh;n o{ti{{{ pevmpei m∆ ejkeivnh tou'de tou' fovbou cavrin. This mirrors its inspiration, Choe.524–5, in having VS order preceded by focalization, even though the Choephoroi passage is not in a subordinate clause: kai; nuktiplavgktwn deimavtwn pepalmevnh coa;" e[pemye tavsde duvsqeo" gunhv. OT. Completive o{ti{{{ OT.59–60: proshvlqeq∆ iJmeivronte": eu\ ga;r oi\d∆ o{ti{{{ nosei'te pavnte", kai; nosou'nte" wJ" ejgw; OT.499–501: ........ ajndrw'n d∆ o{ti{{{ mavnti" plevon h] ∆gw; fevretai, krivsi" oujk e[stin ajlhqhv": OT.525–526: Tou[po" d∆ ejfavnqh tai'" ejmai'" gnwvmai" o{ti{{{ peisqei;" oJ mavnti" tou;" lovgou" yeudei'" levgoi… OT.1132ff.: ..... eu\ ga;r oi\d∆ o{ti{{{ kavtoiden h\\ \\mo" to;n Kiqairw'no" tovpon oJ me;n diploi'si poimnivoi", ejgw; d∆ eJni; ‹ › ejplhsivazon tw/'de tajndri; trei'" o{lou" 325 OT.1401–3: .......... a\rav mou mevmnhsq∆ o{ti{{{ oi|∆ e[rga dravsa" uJmi;n ei\ta deu'r∆ ijw;n oJpoi'∆ e[prasson au\qi"… ........ Relative o{ ti{{{ OT.71: e[pemya Foivbou dwvmaq∆, wJ" puvqoiq∆ o{ ti{{{ drw'n h] tiv fwnw'n thvnde rJusaivmhn povlin. Adverbial o{ti{{{ OT.1340: ∆Apavget∆ ejktovpion o{ti tavcistav me, ÔW" in OT. In OT., there are 73 instances of wJ", of which 15 have complement function. Most (10) have an antecedent, and 12 follow verbs of speech: in 6 constructions, placed at the line start. Cognitive verbs OT.543: Oi\sq∆ wJ"JJJ povhson… ajnti; tw'n eijrhmevnwn OT.729–30: “Edox∆ ajkou'sai sou' tovdvvv ∆, wJ"JJJ oJ Lavi>o" katasfageivh pro;" triplai'" aJmaxitoi'". OT.848–9: ∆All∆ wJ"JJJ fanevn ge tou[po" w|d∆ ejpivstaso, koujk e[stin aujtw'/ tou'tov g∆ ejkbalei'n pavlin: Speech verbs (Optative subordinate verbs appear in 8 constructions) OT.536–8: Fevr∆ eijpe; pro;" qew'n, deilivan h] mwrivan ijdwvn tin∆ e[n moi tau't∆ ejbouleuvsw poei'n… h] tou[rgon[[[ wJ"JJJ ouj gnwrivsoimiv sou tovdevvv dovlw/ prosevrpon koujk ajlexoivmhnj vj vj v maqwvn… OT.547: Tou't∆ aujto' j' j' j ; nu'n mou prw't∆ a[kouson wJ"JJJ ejrw'. OT.555–6: “Epeiqe" h] oujk e[peiqe" wJ"JJJ creivh m∆ ejpi; to;n semnovmantin a[ndra pevmyasqaiv tina… OT.711–3: crhsmo;";;; ga;r h\lqe Laiv>w/ pot∆, oujk ejrw' Foivbou g∆ ajp∆ aujtou', tw'n d∆ uJphretw'n a[po, wJ"JJJ aujto;n h{xoi moi'ra pro;" paido;" qanei'n 326 OT.780: kalei' par∆ oi[nw/ plasto;";;; wJ"JJJ ei[hn patriv. OT.790–1: kai; deina; kai; duvsthna; ; v; ; v; ; v proujfavnh levgwn, wJ"JJJ mhtri; me;n creivh me micqh'nai, gevno" d∆ OT.842–3: Lh/sta;"/ ;/ ;/ ; e[faske" aujto;n a[ndra"[[[ ejnnevpein w{"{{{ nin katakteivneian. Eij me;n ou\n e[ti OT.955–6: ∆Ek th'" Korivnqou, patevra to;nv ;v ;v ; so;n;;; ajggelw'n wJ"JJJ oujkevt∆ o[nta Povlubon, ajll∆ ojlwlovta. OT.1161: Ouj dh't∆ e[gwg∆, ajll∆ ei\pon wJ"JJJ doivhn pavlai. OT.1172: kavllist∆ a]n ei[poi sh; gunh; tavdvvv ∆ wJ"JJJ e[cei. OT.1289–90: to;n mhtro;": aujdw'n ajnovsij vj vj v ∆ oujde; rJhtaj ; Jj ; Jj ; J v moi, wJ"JJJ ejk cqono;" rJivywn eJautovn, oujd∆ e[ti OT.1369–70: ÔW" me;n tavd∆ oujc w|d∆ e[st∆ a[rist∆ eijrgasmevna, mhv m∆ ejkdivdaske, mhde; sumbouvleu∆ e[ti. ”Opw" There are six instances of subordinating o{pw", three as complement introducers, at OT.548: Tou't''' ∆ aujto;j ;j ;j ; mhv moi fravz∆ o{pw"{{{ oujk ei\ kakov". OT.1058: Oujk a]n gevnoito tou'q' '' ∆, o{pw"{{{ ejgw; labw;n shmei'a toiau't∆ ouj fanw' toujmo;n gevno". OT.1366: Oujk oi\d∆ o{pw"{{{ se fw' bebouleu'sqai kalw'". Eijj jj Almost all instances are conditionals. A conditional following a verb of speech at 703 has a formal identity with a complement structure: Levg∆, eij safw'" to; nei'ko" ejgkalw'n ejrei'". Two instances of eij introduce complements, both with antecedents, at OT.584: Skevyai de; tou'to''' prw'ton, ei[ tin∆ a]n dokei'" a[rcein eJlevsqai xu;n fovboisi ma'llon h] a[treston eu{dont∆, eij tav g∆ au[q∆ e{xei kravth. OT.604: peuvqou ta; crhsqevnt; v; v; v ∆ eijj jj safw'" h[ggeilav soi: 327 Mhvvvv Of 71 constructions in OT., 5 are subordinating. All have the meaning 'lest' following verbs of fearing (747, 767, 948, 1011, 1012). These might be regarded as carrying a causal force rather than a reporting function, and all have a verb in the optative or subjunctive (though at 1074– 5, devdoic∆ o{pw" mh{{{ v is followed by a future indicative). Euripides, Medea ”Oti Med.560: kai; mh; spanizoivmesqa, gignwvskwn o{ti{{{ pevnhta feuvgei pa'" ti" ejkpodw;n fivlo", ÔW" in Medea Cognitive verbs Med.67–72: h[kousav tou levgonto", ouj dokw'n kluvein, pessou;" proselqwvn, e[nqa dh; palaivteroi qavssousi, semno;n ajmfi; peirhvnh" u{dwr, wJ"JJJ touvsde pai'da" gh'" ejla'n Korinqiva" su;n mhtri; mevlloi th'sde koivrano" cqono;" Krevwn... Med.85–6: tiv" d∆ oujci; qnhtw'n… a[rti gignwvskei" tovdevvv , wJ"JJJ pa'" ti" auJto;n tou' pevla" ma'llon filei', Med.446–7: ouj nu'n katei'don prw'ton ajlla; pollavki" tracei'an ojrgh;nj ;j ;j ; wJ"JJJ ajmhvcanon kakovn. Med.600: oi\sq∆ wJ"JJJ mevteuxai, kai; sofwtevra fanh'/… Med.1119–20: ... pneu'ma d∆ hjreqismevnon deivknusin w{"{{{ ti kaino;n ajggelei' kakovn. Med.1246–7: kai; mh; kakisqh'/" mhd∆ ajnamnhsqh'/" tevknwnvvv , wJ"JJJ fivltaq∆, wJ"JJJ e[tikte": ajlla; thvnde ge Med.1311: wJ"JJJ oujkevt∆ o[ntwn sw'n tevknwn frovntize dhv. Med.1405: Zeu', tavdvvv ∆ ajkouvei" wJ"JJJ ajpelaunovmeq∆, 328 Speech verbs Med.248–9: levgousi d∆ hJma'"J 'J 'J ' wJ"JJJ ajkivndunon bivon zw'men kat∆ oi[kou", oi} de; mavrnantai doriv: Med.452: levgous∆ ∆Iavsonvvv ∆ wJ"JJJ kavkistov" ejst∆ ajnhvr: Med.529–31: ... ajll∆ ejpivfqono" lovgo"vvv dielqei'n, wJ"JJJ “Erw" s∆ hjnavgkase Êtovxoi" ajfuvktoisÊ toujmo;n ejksw'/sai devma". Med.619–20: ajll∆ ou\n ejgw; me;n daivmona"vvv martuvromai, wJ"JJJ pavnq∆ uJpourgei'n soiv te kai; tevknoi" qevlw: Med.776–7: molovnti d∆ aujtw'/ malqakou;";;; levxw lovgou"vvv , wJ"JJJ kai; dokei' moi taujtav, kai; kalw'" e[cein Med.1410–2: marturovmeno" daivmona"vvv w{"{{{ moi tevkna kteivnas∆ ajpokwluvei" yau'saiv te ceroi'n qavyai te nekrouv" Eijj jj Med.184–5: dravsw tavd∆: ajta;r fovbo"vvv eij peivsw devspoinan ejmhvn: Med.346: toujmou' ga;r ou[ moi frontiv"vvv , eij jjj feuxouvmeqa, Med.931: ejsh'lqev m∆ oi\\ \\kto" eij genhvsetai tavde. Med.941: oujk oi\d∆ a]n eij peivsaimi, peira'sqai de; crhv. Med.1319–20: .........eij d∆ ejmou' creivan e[cei", levg∆, ei[ [[[ ti bouvlh/.......... A double interrogative construction occurs at Med.492–4: ......... oujd∆ e[cw maqei'n eijj jj qeou;" nomivzei" tou;" tovt∆ oujk a[rcein e[ti, h]] ]] kaina; kei'sqai qevsmi∆ ajnqrwvpoi" ta; nu'n, A preposed construction at Med.1103–4: e[ti d∆ ejk touvtwn ei[t[[[ ∆ ejpi; flauvroi" ei[t[[[ ∆ ejpi; crhstoi'" mocqou'si, tovd∆ ejsti;n a[dhlon. 329 Mhvvvv There are 5 instances with mhv after verbs of fearing, at Med.37, 40, 283, 306, 317, with subjunctive verbs. All except the nominal-dependent 317 have antecedents, mostly proleptic subjects, as at 37: devdoika d∆ aujth;n mhj ;j ;j ; v ti bouleuvsh/ nevon: 39–40: pavscous∆: ejgw\/da thvnde, deimaivnw tev nin mh; qhkto;n w[sh/ favsganon di∆ h{pato" Other Euripidean o{ti{{{ -complements following cognitive and speech verbs Cognitive verbs (3) Cyclops 321–3: oujd∆ oi\d∆ o{ ti{{{ Zeuv" ejst∆ ejmou' kreivsswn qeov". ou[ moi mevlei to; loipovn: wJ"JJJ d∆ ou[ moi mevlei, a[kouson. o{tan a[nwqen o[mbron ejkcevh/, Cyc.421–2: ....gignwvskwn o{ti{{{ trwvsei nin oi\no" kai; divkhn dwvsei tavca. Phoen.1617: h{d∆ hJ qanou'sa… zw'sav g∆ a]n savf∆ oi\d∆ o{ti{{{ Speech verbs (8) Ba.173–4: i[tw ti", eijsavggelle Teiresiva"vvv o{ti zhtei' nin: ... Ba.649: ouvk ei\pon, h] oujk h[kousa", o{ti{{{ luvsei mev ti"… El.171–3: ajggevllei d∆ o{ti nu'n tritaiv- an karuvssousin qusivan ∆Argei'oi, ... Hel.1491–4: karuvxat∆ ajggelivan Eujrwvtan ejfezovmenai, Menevlew"vvv o{ti Dardavnou povlin eJlw;n dovmon h{xei. HF.1417: pw'" ou\n e[t∆ ei[ph/" o{ti{{{ sunevstalmai kakoi'"… IT.1093–4: eujxuvneton xunetoi'" boavnvvv , o{ti{{{ povsin keladei'" ajei; molpai'" Or.8–10: wJ" me;n levgousin, o{ti{{{ qeoi'" a[nqrwpo" w[n koinh'" trapevzh" ajxivwm∆ e[cwn i[son ajkovlaston e[sce glw'ssan, ... 330 Or.892–3: lovgou"vvv eJlivsswn, o{ti{{{ kaqistaivh novmou" ej" tou;" tekovnta" ouj kalouv": ... Aristophanes, Frogs ”Oti Cognitive verbs 584: Oi\d∆ oi\d∆ o{ti{{{ qumoi', kai; dikaivw" aujto; dra'/": 599–600: o{ti{{{ me;n ou\n, h]n crhsto;n h\/ ti, tau't∆ ajfairei'sqai pavlin peiravsetaiv m∆ eu\ oi\d∆ o{ti{{{ . 741–2: To; de; mh; patavxai s∆ ejxelegcqevnt∆ a[ntikru", o{ti{{{ dou'lo" w]n e[faske" ei\nai despovth". 1136: ÔOra'/" o{ti{{{ lhrei'". øEU.Ø ∆All∆ ojlivgon gev moi mevlei. Speech verbs Ra.8 and 9 (without main verbs): metaballovmeno" tajnavforon o{ti{{{ cezhtia'/". Mhd∆ o{ti{{{ tosou'ton a[cqo" ejp∆ ejmautw'/ fevrwn, eij mh; kaqairhvsei ti", ajpopardhvsomai… 19–20: «W triskakodaivmwn a[r∆ oJ travchlo" ouJtosiv, o{ti{{{ qlivbetai mevn, to;; ;; de; gevloionvvv oujk ejrei'. 519–20: “Iqi nun, fravson prwvtista tai'" ojrchstrivsin tai'" e[ndon ou[sai" aujto;"j ;j ;j ; o{ti{{{ eijsevrcomai. The free relative constructions are: Ra.1034–5: .................. oJ de; qei'o" ”Omhro" ajpo; tou' timh;n kai; klevo" e[scen plh;n tou'd∆ o{ti{{{ crhvst∆ ejdivdaxen, Ra.1162: Pw'" dhv… Divdaxon gavr me kaq∆ o{ti{{{ dh; levgei". Ra.1169: Eu\ nh; to;n ÔErmh'n: o{ ti{{{ levgei" d∆ ouj manqavnw. JW"JJJ There are 41 instances of wJ" in Frogs, with 4 complements. Two with antecedents are dependent on verbs of speech, at Ra.683–4: truvzei d∆ ejpivklauton ajhdovnion novmonj v vj v vj v v , wJ"JJJ ajpolei'tai, 331 Ra.908–10: ejn toi'sin uJstavtoi" fravsw: tou'ton''' de; prw't∆ ejlevgxw, wJ"JJJ h\n ajlazw;n kai; fevnax oi{oi" te tou;" qeata;" ejxhpavta... The two others follow cognitive verbs, at Ra.400: kai; dei'xon wJ"JJJ a[neu povnou pollh;n oJdo;n peraivnei". Ra.1030–1: Tau'ta ga;r a[ndra" crh; poihta;" ajskei'n. Skevyai ga;r ajp∆ ajrch'" wJ"JJJ wjfevlimoij vj vj v tw'n poihtw'n oiJ gennai'oi gegevnhntai. At Ra.5, there is an exclamative use which the context suggests may introduce direct speech: Mhd∆ e{teron ajstei'ovn ti… (Dion.) Plhvn g∆ ÔW" qlivbomai. {Opw"{{{ Ra.75: ouj ga;r savf∆ oi\d∆ oujd∆ aujto; tou'qj ; 'j ; 'j ; ' ∆ o{pw"{{{ e[cei. 640: Oujk e[sq[[[ ∆ o{pw" oujk ei\ su; gennavda" ajnhvr: 1349: klwsth'ra poiou's∆, o{pw"{{{ knefai'o" eij" ajgora;n fevrous∆ ajpodoivman. 1520–3: Mevmnhso d∆ o{pw"{{{ oJ panou'rgo" ajnh;r kai; yeudolovgo" kai; bwmolovco" mhdevpot∆ eij" to;n qa'kon to;n ejmo;n mhd∆ a[kwn ejgkaqedei'tai. Eijj jj Complementizing eij is rarer in Frogs than in any other of the texts studied. There is only one instance of eij introducing an indirect question, in a Euripidean quotation (=E.fr.638) at 1477: ÆTiv" d∆ oi\den eijj jj to; zh'n mevn ejsti katqanei'nÆ 332 3C: Complements in prose Thucydides, Melian Dialogue There are 11 complement clauses, 4 with wJ", and 7 with o{ti. Instances other than those cited in the main text (in Chapters 4 and 7): 5.85: gignwvskomen ga;r o{ti{{{ tou'to fronei' hJmw'n hJ ej" tou;" ojlivgou" ajgwghv), 5.105.4: pro;" de; tou;" a[llou" polla; a[n ti" e[cwn eijpei'n wJ"JJJ prosfevrontai, 5.105.4: mavlist∆ a]n dhlwvseien o{ti{{{ ejpifanevstata w|n i[smen ta; me;n hJdeva kala; nomivzousi, 5.111.1: Touvtwn me;n kai; pepeiramevnoi" a[n ti gevnoito kai; uJmi'n kai; oujk ajnepisthvmosin o{ti{{{ oujd∆ ajpo; mia'" pwvpote poliorkiva" ∆Aqhnai'oi di∆ a[llwn fovbon ajpecwvrhsan. ejnqumouvmeqa de; o{ti{{{ fhvsante" peri; swthriva" bouleuvsein oujde;n ejn tosouvtw/ lovgw/ eijrhvkate w|/ a[nqrwpoi a]n pisteuvsante" 5.111.5: ........ kai; ejnqumei'sqe pollavki" o{ti{{{ peri; patrivdo" bouleuvesqe, Plato, Crito ”Oti Specifying 51e4: kai; to;n mh; peiqovmenon trich'/ famen ajdikei'n, o{ti{{{ te gennhtai'" ou\sin hJmi'n ouj peivqetai, kai; o{ti{{{ trofeu'si, kai; o{ti{{{ oJmologhvsa" hJmi'n peivsesqai ou[te peivqetai ou[te peivqei hJma'", Cognitive verbs 43d4: dh'lon''' ou\n ejk touvtwn ªtw'n ajggevlwnº o{ti{{{ h{xei thvmeron, 44.d.1: ∆All∆ oJra'/" dh; o{ti{{{ ajnavgkh, w\ Swvkrate", kai; th'" tw'n pollw'n dovxh" mevlein. aujta; de; dh'la''' ta; parovnta nuni; o{ti{{{ oi|oiv 333 t∆ eijsi;n oiJ polloi; ouj ta; smikrovtata tw'n kakw'n ejxergavzesqai ajlla; ta; mevgista scedovn, 46c1: eu\ i[sqi o{ti{{{ ouj mhv soi sugcwrhvsw, 48b4: kai; tovnde de; au\ skovpei eij e[ti mevnei hJmi'n h] ou[, o{ti{{{ ouj to; zh'n peri; pleiv stou poihtevon ajlla; to; eu\ zh'n. ∆Alla; mevnei. 49d2: oi\da ga;r o{ti{{{ ojlivgoi" tisi; tau'ta kai; dokei' kai; dovxei. 51a7: h] ou{tw" ei\ sofo;" w{ste levlhqevn se o{ti{{{ mhtrov" te kai; patro;" kai; tw'n a[llwn progovnwn aJpavntwn timiwvterovn ejstin patri;" kai; semnovteron kai; aJgiwvteron 54c6: eijdovte" o{ti{{{ kai; hJma'" ejpeceivrhsa" ajpolevsai to; so;n mevro". 54d2: Tau'ta''' , w\ fivle eJtai're Krivtwn, eu\ i[sqi o{ti{{{ ejgw; dokw' ajkouvein, Speech verbs Crito 45b7: mhvte, o}} }} e[lege" ejn tw'/ dikasthrivw/, duscerev" soi genevsqw o{ti{{{ oujk a]n e[coi" ejxelqw;n o{ti{{{ crw'/o sautw'/: Crito 46c8: povteron kalw'" ejlevgeto eJkavstote h] ou[, o{ti{{{ tai'" me;n dei' tw'n doxw'n prosevcein to;n nou'n, tai'" de; ou[… h] pri;n me;n ejme; dei'n ajpoqnhv/skein kalw'" ejlevgeto, nu'n de; katavdhlo"vvv a[ra ejgevneto o{ti{{{ a[llw" e{neka lovgou ejlevgeto, 46d7: ejlevgeto dev pw", wJ" ejgw\/mai, eJkavstote w|de uJpo; tw'n oijomevnwn ti; levgein, w{sper nundh; ejgw; e[legon, o{ti{{{ tw'n doxw'n a}" oiJ a[nqrwpoi doxavzousin devoi ta;" me;n peri; pollou' poiei'sqai, ta;" de; mhv. 47a2: oujc iJkanw'" dokei' soi levgesqai o{ti{{{ ouj pavsa" crh; ta;" dovxa" tw'n ajnqrwvpwn tima'n ajlla; ta;" mevn, 51c6: ÆSkovpei toivnun, w\ Swvkrate",Æ fai'en a]n i[sw" oiJ novmoi, Æeij hJmei'" tau'ta ajlhqh' levgomen, o{ti{{{ ouj divkaia hJma'" ejpiceirei'" dra'n a} nu'n ejpiceirei'". 334 Direct speech 50c1: h] ejrou'men pro;" aujtou;" o{ti{{{ Æ∆Hdivkei ga;r hJma'" hJ povli" kai; oujk ojrqw'" th;n divkhn e[krinen…Æ 50c7: ....i[sw" a]n ei[poien o{ti{{{ Æ«W Swvkrate", mh; qauvmaze ta; legovmena ajll∆ ajpokrivnou, 52a6: i[sw" a[n mou dikaivw" kaqavptointo levgonte" o{ti{{{ ejn toi'" mavlista ∆Aqhnaivwn ejgw; aujtoi'" wJmologhkw;" tugcavnw tauvthn th;n oJmologivan. fai'en ga;r a]n o{ti{{{ Æ«W Swvkrate", megavla hJmi'n touvtwnvvv tekmhvriav ejstin, o{ti{{{ soi kai; hJmei'" hjrevskomen kai; hJ povli": ÔW" in Crito Verbs of emotion Crito 43b4: ajlla; kai; sou' pavlai qaumavzw aijsqanovmeno" wJ"JJJ hJdevw" kaqeuvdei": 44b9: e[ti de; kai; polloi'" dovxw, oi} ejme; kai; se; mh; safw'" i[sasin, wJ"JJJ oi|ov" t∆ w[n se swv/zein eij h[qelon ajnalivskein crhvmata, ajmelh'sai. 44c3: ouj ga;r peivsontai oiJ polloi; wJ"JJJ su; aujto;" oujk hjqevlhsa" ajpievnai ejnqevnde hJmw'n proqumoumevnwn. 45e1: ....... aijscuvnomai mh; dovxh/ a{pan to; pra'gma to; peri; se; ajnandriva/ tini; th'/ hJmetevra/ pepra'cqai, kai; hJ ei[sodo"J [J [J [ th'" divkh" eij" to; dikasthvrion wJ"JJJ eijsh'lqen ejxo;n mh; eijselqei'n, kai; aujto;" oJ ajgw;nj ; J j ;j ; J j ;j ; J j ; th'" divkh" wJ"JJJ ejgevneto, 48e2: pau'sai h[dh, w\ makavrie, pollavki" moi levgwn to;n aujto;n ; j ;; j ;; j ; lovgonvvv , wJ"JJJ crh; ejnqevnde ajkovntwn ∆Aqhnaivwn ejme; ajpievnai: 50d3: fravson ou\n, touvtoi" hJmw'n, toi'" novmoi" toi'" peri; tou;" gavmou", mevmfh/ ti wJ"JJJ ouj kalw'" e[cousin…Æ 50e3: e[coi" a]n eijpei'n prw'ton me;n wJ"JJJ oujci; hJmevtero" h\sqa kai; e[kgono" kai; dou'lo", aujtov" te kai; oiJ soi; provgonoi… 53d4: kai; i[sw" a]n hJdevw" sou ajkouvoien wJ"JJJ geloivw" ejk tou' desmwthrivou ajpedivdraske" skeuhvn tev tina periqevmeno" 335 Eijj jj and ejavnj vj vj v in Crito 46d4ff.: ejpiqumw d∆ e[gwg∆ ejpiskevyasqai, w\ Krivtwn, koinh'/ meta; sou' ei[ tiv moi ajlloiovtero" fanei'tai, ejpeidh; w|de e[cw... 48b4ff.: kai; tovnde de; au\ skovpei eijj jj e[ti mevnei hJmi'n h] ou[, o{ti ouj to; zh'n peri; pleivstou poihtevon ajlla; to; eu\ zh'n. 48d3ff: mh; ouj devh/ uJpologivzesqai ou[t∆ eijj jj ajpoqnhv/skein dei' paramevnonta" kai; hJsucivan a[gonta", ou[te a[llo oJtiou'n pavscein pro; tou' ajdikei'n. 52d3f.: prw'ton me;n ou\n hJmi'n tou't∆ aujto' j' j' j ; ajpovkrinai, eijj jj ajlhqh' levgomen There is one polar question with ejavn, at 48e5: o{ra de; dh; th'" skevyew" th;n ajrch;n ejavnj vj vj v soi iJkanw'" levghtai 336 Bibliography Currently published classical journals are cited using the abbreviations given in the index of L'Année Philologique. Other journals are cited by full title. _________________________________________________________. Abraham, W., ed. 1983. On the Formal Syntax of the Westgermania: Papers from the ‘3rd Groningen Grammar Talks,’ Groningen, January 1981, Amsterdam: Benjamins. — , Givón, T., and Thompson, S.A., eds. 1995. Discourse Grammar and Typology: Papers in Honor of John W.M. Verhaar, Amsterdam: Benjamins. Adams, J.N. 1971. ‘A Type of Hyperbaton in Latin Prose.’ PCPhS, 197: 1–16. — 1994a. ‘Wackernagel’s Law and the Position of Unstressed Personal Pronouns in Classical Latin.’ TPhS, 92: 103–178. — 1994b. Wackernagel’s Law and the Placing of Copula esse in Classical Latin. PCPhS Supplementary Volume 18, Cambridge: Cambridge Philological Society. — 1996. ‘Interpuncts as Evidence for the Enclitic Character of Personal Pronouns in Latin.’ ZPE, 111: 208–210. Aitchison, J. 1979. ‘The Order of Word Order Change.’ TPhS: 43–65. Allén, S., ed. 1982. Text Processing: Text Analysis and Generation, Text Typology and Generation, Proceedings of Nobel Symposium 51, Stockholm: Almqvist and Wiksell. Allen, T.W., ed. 1917–19. Homeri opera: recognovit brevique adnotatione critica instruxit Thomas W. Allen. Odyssey, 2nd Edition, 2 volumes, Oxford: Clarendon. — , and Sikes, E.E., eds. 1904. The Homeric Hymns: Edited with Preface, Apparatus Criticus, Notes and Appendices, London: Macmillan. Allen, W.S. 1973. Accent and Rhythm. Prosodic Features of Latin and Greek: A Study in Theory and Reconstruction, Cambridge: Cambridge University Press. — 1987. Vox Graeca: A Guide to the Pronunciation of Classical Greek, 3rd Edition, Cambridge: Cambridge University Press. Amberber, M. 1996. ‘The Transitivity of Verbs of Saying Revisited’ in Camacho, Choueiri, and Watanabe 1996: 1–15. Amigues, S. 1977. Les subordinées finales par OPWS en attique classique, Paris: Klincksieck. Ammann, H. 1922. Untersuchungen zur homerischen Wortfolge und Satzstruktur: mit besonderer Berücksichtigung der Stellung des Verbums, Teil 1, Freiburg: Boltze. — 1924. ‘Untersuchungen zur homerischen Wortfolge und Satzstruktur. Teil 2: Die Stellung des Verbums, im Einzelnen untersucht.’ IF, 1: 149–178. Anderson, J.M., ed. 1982. Language Form and Linguistic Variation: Papers Dedicated to Angus McIntosh, Amsterdam: Benjamins. — and Jones, C., eds. 1974. Historical Linguistics I: Syntax, Morphology, Internal and Comparative Reconstruction. Proceedings of the First International Conference on 337 Historical Linguistics, Edinburgh 2nd–7th September 1973, Amsterdam: North Holland. Anderson, S.R. 1992. A-Morphous Morphology, Cambridge: Cambridge University Press. — 1993. ‘Wackernagel’s Revenge: Clitics, Morphology, and the Syntax of Second Position.’ Language, 69: 68–98. — 1996, ‘How to Put Your Clitics in their Place, or Why the Best Account of Second position Phenomena May Be Something Like the Optimal One.’ The Linguistic Review, 13: 165–191. — and Chung, S. 1977. ‘On Grammatical Relations and Clause Structure in Verb-Initial Languages’ in Cole and Sadock 1977: 1–25. — and Kiparsky, P., eds. 1973. Festschrift for Morris Halle, New York: Holt, Rinehart, and Winston. Andersson, E. 1985. On Verb Complementation in Written English, Malmö: Gleerup. Andrews, A. 1971. ‘Case Agreement of Predicate Modifiers in Ancient Greek.’ Linguistic Inquiry, 2: 127–151. Aronoff, M. 1994. Morphology by Itself: Stems and Inflectional Classes, Cambridge MA: The MIT Press. Atkinson, J.M. and Heritage, J., eds. 1984. Structures of Social Action: Studies in Conversation Analysis, Cambridge: Cambridge University Press. Austin, J.L. 1962. How to Do Things with Words, Cambridge MA: Harvard University Press. Autentrieth, G. 1984. Homeric Dictionary, tr. Keep, R. first published 1873; English translation first published 1877, London: Duckworth. Auwera, J. van der, and Goossens, L. eds. 1987. Ins and Outs of the Predication, Dordrecht: Foris. Bach, E. 1962. ‘The Order of Elements in a Transformational Grammar of German.’ Language, 38: 263–269. — 1971. ‘Questions.’ Linguistic Inquiry, 2: 153–166. — and Harms, R.T., eds. 1968. Universals in Linguistic Theory, New York: Holt, Rinehart and Winston. Bader, F. 1975. ‘Une isoglosse gréco-tokharienne: *yo affixe casuel et particule d’énumeration.’ BSL, 70: 27–89. — 1992. ‘Formes, fonctions, étymologies pronominales’ in Morel and Danon-Boileau 1992: 27–41. Bahner, W., Schildt, J., and Viehweger, D., eds. 1990. Proceedings of the Fourteenth International Congress of Linguists. Berlin/GDR, August 10–August 15, 1987, Berlin: Academie Verlag. Bain, A. 1887. English Composition and Rhetoric, Part 1, London: Longman. 338 Baker, M. 1985. ‘The Mirror Principle and Morphosyntactic Explanation.’ Linguistic Inquiry, 16: 373–415. Bakker, E.J. 1988a. Linguistics and Formulas in Homer: Scalarity and the Description of the Particle per, Amsterdam: Benjamins. — 1988b. ‘Restrictive Conditionals’ in Rijksbaron et al. 1988: 5–26. — 1993. ‘Boundaries, Topics, and the Structure of Discourse: An Investigation of the Ancient Greek Particle dev.’ Studies in Language, 17: 275–311. Ball, C.N. 1994. ‘Relative Pronouns in It-clefts: The Last Seven Centuries.’ Language Variation and Change, 6: 179–200. Barlow, M. and Ferguson, C.A., eds. 1988. Agreement in Natural Languages: Approaches, Theories, Descriptions, Stanford: Centre for the Study of Language and Information. Basset, L. 1997. jAll j ejxovloisq j aujtw/' koax. Réexamen des emplois de ajllav à la lumière de l’énonciation dans les Grenouilles d’Aristophane’ in Rijksbaron 1997a: 75–99. Battye, A. and Roberts, I., eds. 1995. Clause Structure and Language Change, New York: Oxford University Press. Bauer, B.M.L. 1993a. ‘The Tendency towards Right Branching in the Development and Acquisition of Latin and French’ in Marle 1993: 1–17. — 1993b. ‘The Development of Latin Absolute Constructions: From Stative to Transitive Structures.’ General Linguistics, 33: 64–83. — 1995. The Emergence and Development of SVO Patterning in Latin and French: Diachronic and Psycholinguistic Perspectives, New York: Oxford University Press. — 1996. ‘Residues of Non-Nominative Syntax in Latin: The mihi est Construction.’ HSF, 109: 241–256. — 2000. Archaic Syntax in Indo-European: The Spread of Transitivity in Latin and French, Berlin: Mouton. Bäumlein, W. 1861. Untersuchungen über griechische Partikeln, Stuttgart: Metzlerschen. Beckman, M.E. and Edwards, J. 1994. ‘Articulatory Evidence for Differentiating Stress Categories’ in Keating 1994: 7–33. Behaghel, O. 1909. ‘Beziehungen zwischen Umfang und Reihenfolge von Satzgliedern.’ IF, 25: 110–142. — 1929. ‘Zur Stellung des Verbs im Germanischen und Indogermanischen.’ Zeitschrift fur Verleichende Sprachforschung, 56: 276–281. Belletti, A. 1990. Generalized Verb Movement: Aspects of Verb Syntax, Turin: Rosenberg and Sellier. Ben Chaim, D. 1981. Distance in the Theatre: The Aesthetics of Audience Response, Epping: Bowker. Benveniste, E. 1933. ‘L’Anaphorique prussien din et le système des démonstratif indo- européens.’ Studi Baltici, 3: 121–130. 339 — 1954. ‘La Phrase relative, problème de syntaxe générale.’ BSL, 53: 39–54, reprinted in English translation in Benveniste 1971: 181–192. — 1971. Problems in General Linguistics, tr. Meek, M.E., Coral Gables: University of Miami Press. Berent, G.P. 1973. ‘Absolute Constructions as “Subordinate Clauses”‘ in Corum, Smith-Stark, and Weiser 1973: 147–154. Bergaigne, A. 1877. ‘Essai sur la construction grammaticale: considérée dans son développement historique, en sanskrit, en grec, en latin, dans les langues romanes et dans les langues germaniques.’ Section 3, Mémoires de la Société de Linguistique de Paris, III: 169–186. Berman, A. 1974. ‘On the VSO Hypothesis.’ Linguistic Inquiry, 5: 1–38. Bernert, E. 1940. ‘Die Partikel -que.’ Glotta, 28: 78–88. Bers, V. 1984. Greek Poetic Syntax in the Classical Age, New Haven: Yale University Press. — 1994. ‘Tragedy and Rhetoric’ in Worthington 1994: 176–195. Berwick, R.C. 1985. The Acquisition of Syntactic Knowledge, Cambridge MA: The MIT Press. — and Weinberg, A.S. 1984. The Grammatical Basis of Linguistic Performance: Language Use and Acquisition, Cambridge MA: The MIT Press. — , Abney, S.P., and Tenny, C., eds. 1991. Principle-Based Parsing: Computation and Psycholinguistics, Dordrecht: Kluwer. Besten, H. den 1983. ‘On the Interaction of Root Transformations and Lexical Deletive Rules.’ in Abraham 1983: 47–131. Bichakjian, B.H. 1982. ‘La Genèse de la subordination de l’indo-européen au français’ in Mok, Spiele, and Verhuyck 1982: 5–20. Bierwisch, M. and Heidolph, K.E. 1970. Progress in Linguistics, The Hague: Mouton. Bilmes, J. 1986. Discourse and Behavior, New York: Plenum Press. Biraud, M. 1985. ‘La Syntaxe de wJ" dans la langue homérique: du subordonnant relatif au subordonnant conjonctif.’ Annales de la Faculté des Lettres et Sciences Humaines de Nice, 49: 158–172. Blackburn, S. 1984. Spreading the Word: Groundings in the Philosophy of Language, Oxford: Clarendon. Blomfield, C.J., ed. 1818. AISCULOU AGAMEMNWN. Aeschyli Agamemnon: ad fidem manuscriptorum emendavit, notas et glossarium adjecit, Cambridge: Mawman. — 1824. AISCULOU COHFOROI. Aeschyli Choephoroe: ad fidem manuscriptorum emendavit, notas et glossarium adjecit, Cambridge: Mawman. Bloomfield, L. 1933. Language, New York: Holt. Boegehold, A.L. 1989. ‘A Signifying Gesture: Euripides, Iphigeneia Taurica, 965–66.’ AJA, 93: 81–83. 340 Boel, G. De 1980. ‘Towards a Theory of the Meaning of Complementizers in Classical Attic.’ Lingua, 52: 285–304. — 1988. Goal Accusative and Object Accusative in Homer: A Contribution to the Theory of Transitivity, Brussels: Koninklijke Academie voor Wetenschappen, Letteren en Schone Kunsten van België. Bolinger, D.L. 1952. ‘Linear Modification.’ Publications of the Modern Language Association of America, 67: 1117–1144. Boolos, G.S. and Jeffrey, R.C. 1989. Computability and Logic, 3rd Edition, Cambridge: Cambridge University Press. Booth, W.C. 1974. A Rhetoric of Irony, Chicago: University of Chicago Press. Braunmüller, K. 1978. ‘Remarks on the Formation of Conjunctions in Germanic Languages.’ Nordic Journal of Linguistics, 1, 99–120. Brecht, B. 1940. ‘Short Description of a New Technique of Acting which Produces an Alienation Effect’ in Willett 1964: 136–147. Bremen, K. von 1983. Question-words: A Study in the Syntax of Relativization, Free Relatives, Pseudo-cleft Sentences and Certain Indefinite Pronouns, Angered: Graphic Systems. Brennan, C. 1894. ‘On the Manuscripts of Aeschylus.’ Journal of Philology, 22.43: 49–71. Brentano, F. 1973. Psychology from an Empirical Standpoint, tr. Rancurello, A.C. et al., ed. Kraus, O., English ed. McAllister, L.L., London: Routledge and Kegan Paul. Bresnan, J.W. 1970. ‘On Complementizers: Toward a Syntactic Theory of Complement Types.’ Foundations of Language, 6: 297–321. — and Mchombo, S.A. 1987. ‘Topic, Pronoun, and Agreement in Chichewa.’ Language, 63: 741–782. Britton, B.K. and Pellegrini, A.D., eds. 1990. Narrative Thought and Narrative Language, Hillsdale: Erlbaum. Brown, G., ed. 1994. Language and Understanding, Oxford: Clarendon. — and Yule, G. 1983. Discourse Analysis, Cambridge: Cambridge University Press. Brugmann, K. 1911. Vergleichende Laut-, Stammbildungs- und Flexionslehre der indogermanischen Sprachen, Band 2, Teil 2: Nomina, Strassbourg: Trübner. Bühler, K. 1934. Sprachtheorie: die Darstellungsfunktion der Sprache, Jena: Fischer. — 1982. ‘The Deictic Field of Language and Deictic Words’ in Jarvella and Klein 1982: 9–30. Bullough, E.P. 1912. ‘Psychical Distance as a Factor in Art and an Aesthetic Principle.’ British Journal of Psychology, 5: 87–118. Bunt, H. and Tomita, M. 1996. Recent Advances in Parsing Technology, Dordrecht: Kluwer. Burbank, J. and Steiner, P., trs. and eds. 1977. The Word and Verbal Art: Selected Essays by Jan Mukarovsky, foreword by René Wellek, New Haven: Yale University Press. 341 Burns, E. 1972. Theatricality: A Study of Convention in the Theatre and in Social Life, London: Longman. Burton, D. 1980. Dialogue and Discourse, London: Routledge and Kegan Paul. Butler, H.E., ed. and tr. 1920–22. The Institutio Oratoria of Quintilian, reprinted 1963–68, 4 volumes, Cambridge MA: Harvard University Press. Calboli, G., ed. 1989. Subordination and Other Topics in Latin: Proceedings of the Third Colloquium on Latin Linguistics, Bologna 1–5 April 1985, Amsterdam: Benjamins. Camacho, J., Choueiri, L., and Watanabe, M., eds. 1996. The Proceedings of the Fourteenth West Coast Conference on Formal Linguistics, Stanford: Center for the Study of Language and Information. Campbell, L. 1993. ‘The Explanation of Syntactic Change: A Historical Perspective’ in Marle 1993: 49–69. Canedo, J. 1937. Zur Wort- und Satzstellung in der alt- und mittelindischen Prosa, Göttingen: Vandenhoeck & Ruprecht. Carruba, O. 1969. Die satzeinleitenden Partikeln in den indogermanischen Sprachen Anatoliens, Rome: CNR. Cervin, R. 1988. ‘On the Notion of “Second Position” in Greek.’ Studies in the Linguistic Sciences, 18: 23–39. — 1990. Word Order in Ancient Greek: VSO, SVO, SOV, or All of the Above?, Doctoral Dissertation, University of Illinois. Chafe, W.L. 1976. ‘Givenness, Contrastiveness, Definiteness, Subjects, Topics and Points of View’ in Li 1976: 25–56. — and Nichols, J., eds. 1986. Evidentiality: The Linguistic Coding of Epistemology, Norwood NJ: Ablex Publishing Corporation. Chantraine, P. 1952. ‘Les Recherches sur l’ordre des mots en grec.’ AFC, 5: 71–80. — 1958. Grammaire homérique, tome 1: phonétique et morphologie, Paris: Klincksieck. — 1963. Grammaire homérique, tome 2: syntaxe, Paris: Klincksieck. — 1980–83. Dictionnaire étymologique de la langue grecque: histoire des mots, New Edition, Paris: Klincksieck. Chomsky, N. 1957. Syntactic Structures, The Hague: Mouton. — 1965. Aspects of the Theory of Syntax, Cambridge MA: The MIT Press. — 1971. ‘Deep Structure, Surface Structure, and Semantic Interpretation’ in Steinberg and Jakobovits 1971: 183–216. — 1973. ‘Conditions on Transformations’ in Anderson and Kiparsky 1973: 232–286. — 1976. ‘Conditions on Rules of Grammar.’ Linguistic Analysis, 2: 303–352. — 1981. Lectures on Government and Binding, Dordrecht: Foris Publications. — 1982. Some Concepts and Consequences of the Theory of Government and Binding, Linguistic Inquiry Monograph 6, Cambridge MA: The MIT Press. 342 — 1986. Barriers Linguistic Inquiry Monograph 13, Cambridge MA: The MIT Press. — 1992. ‘A Minimalist Program for Linguistic Theory.’ MIT Occasional Papers in Linguistics 1, republished in Hale and Keyser 1993: 1–52. Christol, A. 1989. ‘Prolepse et syntaxe indo-européenne’ in Calboli 1989: 65–89. Clapp, E.B. 1891. ‘Conditional Sentences in the Greek Tragedians.’ TAPhA, 22: 81–92. Clay, D.M. 1958. A Formal Analysis of the Vocabularies of Aeschylus, Sophocles, and Euripides, Minneapolis: Numerical Studies. Cole, P., ed. 1981. Radical Pragmatics, New York: Academic Press. — and Sadock, J., eds. 1977. Grammatical Relations Syntax and Semantics Series Vol. 8, New York: Academic Press. Cole, T. 1991. The Origins of Rhetoric in Ancient Greece, Baltimore: Johns Hopkins University Press. Coleman, R. 1985. ‘The Indo-European Origins and Latin Development of the Accusative with Infinitive Construction’ in Touratier 1985: 307–342. — 1989. ‘The Rise and Fall of Absolute Constructions: A Latin Case History’ in Calboli 1989: 353–374. Collinge, N.E. 1988. ‘Thoughts on the Pragmatics of Ancient Greek.’ PCPhS, 214: 1–13. Comrie, B. 1980. ‘Morphology and Word Order Reconstruction: Problems and Prospects’ in Fisiak 1980: 83–96. — 1981. Language Universals and Linguistic Typology: Syntax and Morphology, Oxford: Blackwell. — 1986. ‘Conditionals: A Typology’ in Traugott et al. 1986: 77–99. — and Horie, K. 1995. ‘Complement Clauses Versus Relative Clauses: Some Kmer Evidence’ in Abraham, Givón, and Thompson 1995: 65–75. Conacher, D.J. 1987. Aeschylus’ Oresteia: A Literary Commentary, Toronto: University of Toronto Press. Conte, M.E., Petofi, J.S., and Sozer, E., eds. 1989. Text and Discourse Connectedness: Proceedings of the Conference on Connexity of Text, Amsterdam: Benjamins. Conrad, C.W. 1990. From Epic to Lyric: A Study in the History of Traditional Word-Order in Greek and Latin Poetry, New York: Garland. Cook, A. 1992. ‘Particles, Qualification, Ordering, Style, Irony, and Meaning in Plato’s Dialogues.’ QUCC, 40: 111–126. Corum, C., Smith-Stark, T.C., and Weiser, A., eds. 1973. You Take the High Node and I’ll Take the Low Node. Papers from the Comparative Syntax Festival: The Differences between Main and Subordinate Clauses, 12 April 1973, Chicago: Chicago Linguistic Society. Coulthard, M. 1985. An Introduction to Discourse Analysis, London: Longman. — , ed. 1994. Advances in Written Text Analysis, London: Routledge and Kegan Paul. 343 — and Brazil D. 1979. ‘Exchange Structure.’ Discourse Analysis Monograph 5, Birmingham: English Language Research. Craig, R.T. and Tracy, K., eds. 1983. Conversational Coherence: Form, Structure, and Strategy, Beverly Hills: Sage Publications. Crespo, E. 1984. ‘On the System of Substantive Clauses in Ancient Greek.’ Glotta, 62: 1–16. Cristofaro, S. 1995. ‘La complementazione frasale dopo verbi di “dire” e “pensare”‘ in Cuzzolin 1995: 91–112. — 1996. Aspetti sintattici e semantici delle frasi completive in greco antico, Florence: La Nuova Italia Editrice. — 1998. ‘Grammaticalization and Clause Linkeage Strategies: A Typological Approach with Particular Reference to Ancient Greek’ in Ramat and Hopper 1998: 59–88. Croft, W. 1990. Typology and Universals, Cambridge: Cambridge University Press. Cunliffe, R.J. 1924. A Lexicon of the Homeric Dialect, Norman: University of Oklahoma Press. Cuzzolin, P., ed. 1995. Studi di linguistica greca, Pavia: Francoangeli. Dane, J.A. 1991. The Critical Mythology of Irony, Athens GA: University of Georgia Press. Davidson, D. 1968. ‘On Saying That.’ Synthese, 19: 130–146. — 1979. ‘Moods and Performances’ in Margalit 1979: 9–20. Davies, A.M. 1997. ‘Particles in Greek Epigraphic Texts: The Case of Arcadian’ in Rijksbaron 1997a: 49–73. Dawe, R.D., ed. 1982. Sophocles: Oedipus Rex, Cambridge: Cambridge University Press. Delbrück, B. 1878. Die altindische Wortfolge aus dem çatapathabrähmana dargestellt, published as Delbrück, B. and Windisch, E., Syntaktische Forschungen 3, Halle: Waisenhaus. — 1893. Vergleichende Syntax der indogermanischen Sprachen, Vol 1. Vol 3.1 of Brugmann, K. and Delbrück, B., Grundiss der vergleichenden Grammatik der indogermanischen Sprachen. Strasbourg: Trübner. — 1897. Vergleichende Syntax der indogermanischen Sprachen, Vol 2. Vol 4 of Brugmann, K. and Delbrück, B., Grundiss der vergleichenden Grammatik der indogermanischen Sprachen. Strasbourg: Trübner. — 1900. Vergleichende Syntax der indogermanischen Sprachen, Vol 3. Vol 5.3 of Brugmann, K. and Delbrück, B., Grundiss der vergleichenden Grammatik der indogermanischen Sprachen. Strasbourg: Trübner. — 1911. ‘Germanische Syntax II: Zur Stellung des Verbums.’ Abhandlungen der philologisch-historischen Klasse der Königlich Sächsischen, Gesellschaft der Wissenschaften, 28, No 7. Denniston, J.D. 1952. Greek Prose Style, Oxford: Clarendon. — 1954. The Greek Particles, Second Edition, Oxford: Clarendon. — and Page, D., eds. 1957. Aeschylus, Agamemnon, Oxford: Clarendon. 344 Descroix, J. 1931. Le Trimètre iambique: des iambographes a la comédie nouvelle, Macon: Protat. Devarius, M. 1835–42. De Graecae Linguae Particulis, ed. Klotz, R., Leipzig: Baumgaertner. Devine, A.M. and Stephens, L.D. 1984. Language and Metre: Resolution, Porson’s Bridge, and Their Prosodic Basis, Chico CA: Scholars Press. — 1994. The Prosody of Greek Speech, New York: Oxford University Press. — 2002. Discontinuous Syntax: Hyperbaton in Greek, New York: Oxford University Press. Diggle, J., ed. 1981–1994. Euripidis fabulae, 3 volumes, Oxford: Clarendon. Dik, H. 1995. Word Order in Ancient Greek: A Pragmatic Account of Word Order Variation in Herodotus, Amsterdam: Gieben. — 2007. Word Order in Greek Tragic Dialogue, Oxford: Oxford University Press. Dik, S.C. 1978. Functional Grammar, Amsterdam: North-Holland Publishing Company. — 1980. Studies in Functional Grammar, London: Academic Press. — 1989. The Theory of Functional Grammar. Part 1: The Structure of the Clause, Dordrecht: Foris. Dindorf, G., ed. 1851. Aeschyli tragoediae superstites et deperditarum fragmenta, including scholia, 2nd Edition, 3 volumes, Oxford: University Press. — , ed. 1876. Lexicon Aeschyleum, 2 volumes, Leipzig: Teubner. Dionysius Thrax, Ars grammatica, in Uhlig 1883, Volume 1: 1–132 . Dixon, R.M.W., ed. 1976. Grammatical Categories in Australian Languages, Canberra: Humanities Press. Dodd, W.N. 1981. ‘Conversation, Dialogue and Exposition.’ Strumenti Critici, 44: 171–191. Dodds, E.R. 1953. ‘Notes on the Oresteia.’ CQ, 3: 11–21. Dover, K.J. 1960. Greek Word Order, Cambridge: Cambridge University Press. — 1987. Greek and the Greeks. Collected Papers. Volume 1: Language, Poetry, Drama, Oxford: Blackwell. — , ed. 1993. Aristophanes: Frogs, Oxford: Clarendon. — 1996. ‘Prose Rhythm, Greek’ in Hornblower and Spawforth 1996: 1260–1261. — 1997. The Evolution of Greek Prose Style, Oxford: Clarendon. Downing, A. 1991. ‘An Alternative Approach to Theme: A Systemic-functional Perspective.’ Word, 42.2: 119–143. Downing, B. 1973. ‘Corelative Relative Clauses in Universal Grammar.’ Minnesota Working Papers in Linguistics and Philosophy of Language, 2: 1–17. Dowty, D.R., Karttunen, L., and Zwicky, A.M. 1985. Natural Language Parsing: Psychological, Computational, and Theoretical Perspectives, Cambridge: Cambridge University Press. 345 Dressler, W. 1971. ‘Über die Rekonstruktion der indogermanischen Syntax.’ Zeitschrift für vergleichende Sprachforschung =HSF, 85: 5–22. Dryer, M.S. 1992. ‘The Greenbergian Word Order Correlations.’ Language, 68: 81–138. DuBois, J.A. 1985. ‘Competing Motivations’ in Haiman 1985: 343–366. Duhoux, Y. 1997. ‘Grec écrit et grec parlé: une étude contrastive des particules aux Ve-IVe siècles’ in Rijksbaron 1997a: 15–48. Duke, E.A., Hicken, W.F., Nicoll, W.S.M., Robinson, D.B., and Strachan, J.C.G., eds. 1995. Platonis opera. Volume 1, Oxford: Clarendon. Duncan-Rose, C. and Vennemann, T., eds. 1988. On Language: Rhetorica, Phonologica, Syntactica. A Festschrift for Robert P. Stockwell from his Friends and Colleagues, London: Routledge. Dunkel, G.E. 1983. ‘IE Conjunctions.’ Zeitschrift für vergleichende Sprachforschung =HSF, 96, 178–199. Dunn, G. 1985. ‘Ancient Greek Word Order and the Lehmann Hypothesis.’ Te Reo, 28: 81–94. — 1988. ‘Syntactic Word Order in Herodotean Greek.’ Glotta, 66: 63–79. — 1989. ‘Enclitic Pronoun Movement and the Ancient Greek Sentence Accent.’ Glotta, 67: 1–19. Dyson, M. 1988. ‘Euripides, Medea 926–31.’ CQ, 38: 324–327. Earp, F.R. 1948. The Style of Aeschylus, Cambridge: Cambridge University Press. Easterling, P.E. 1973. ‘Presentation of Character in Aeschylus.’ G&R, 20: 3–18. — 1990. ‘Constructing Character in Greek Tragedy’ in Pelling 1990: 83–99. Ehlich, K. 1982. ‘Anaphora and Deixis: Same, Similar, or Different?’ in Jarvella and Klein 1982: 315–338. Eichner, H., and Rix, H., eds. 1990. Sprachwissenschaft und Philologie: Jacob Wackernagel und die Indogermanistik heute. Kolloquium der Indogermanischen Gesellschaft vom 13 bis 15 Oktober 1988 in Basel, Wiesbaden: Reichert. Emonds, J.E. 1976. A Transformational Approach to English Syntax: Root, Structure-Preserving, and Logical Transformations, New York: Academic Press. Erdmann, P. 1988. ‘On the Principle of “Weight” in English’ in Duncan-Rose and Vennemann 1988: 325–339. Erp Taalman Kip, A.M. van 1997. ‘H GAR in Questions’ in Rijksbaron 1997a: 151–156. Erteschik-Shir, N. 1986. ‘Wh-Questions and Focus.’ Linguistics and Philosophy, 9: 117–149. Evans, B. 1960. Shakespeare’s Comedies, Oxford: Clarendon. Ewans, M., ed. and tr. 1995. Aischylos, The Oresteia: Agamemnon, Libation Bearers, Eumenides, London: Dent. Fagles, R., tr. 1977. Aeschylus, The Oresteia: With Introductory Essay, Notes, and Glossary by Robert Fagles and W.B. Stanford, Harmondsworth: Penguin Books. — and Stanford, W.B. 1977. ‘The Serpent and the Eagle’ in Fagles 1977: 13–97. 346 Fairclough, N. 1989. Language and Power, London: Longman. Fanselow, G. 1990. ‘Scrambling as NP-Movement’ in Grewendorf and Sternefeld 1990: 113– 140. Ferguson, C.A. and Barlow, M. 1988. ‘Introduction’ in Barlow and Ferguson 1988: 1–22. Fillmore, C.J. 1968. ‘The Case for Case’ in Bach and Harms 1968: 1–88. — 1981. ‘Pragmatics and the Description of Discourse’ in Cole 1981: 143–166. Firbas, J. 1964. ‘On Defining the Theme in Functional Sentence Analysis.’ Travaux Linguistiques de Prague, 1: 267–280. Fischer, P. 1924. ‘Zur Stellung des Verbums im Griechischen.’ Glotta, 13: 1–11. Fisiak, J., ed. 1980. Historical Morphology, Trends in Linguistics Studies and Monographs 17, The Hague: Mouton. — , ed. 1984. Historical Syntax Trends in Linguistics Studies and Monographs 23, The Hague: Mouton. Fodor, J.A. 1983. The Modularity of Mind, Cambridge MA: The MIT Press. Foucault, J.A. de 1964. ‘L’Hyperbate du verbe.’ RPh, 38: 59–69. Fournier, H. 1946. Les Verbes ‘dire’ en grec ancien: exemple de conjugaison supplétive, Paris: Klincksieck. Fowler, D. 1990. ‘Deviant Focalisation in Virgil’s Aeneid.’ PCPhS, 36: 42–63. Fraenkel, E. 1932. ‘ Kolon und Satz: Beobachtungen zur Gliederung des antiken Satzes.’ Nachrichten von der Gesellschaft der Wissenschaften zu Göttingen, Philologisch- Historische Klasse, 1932: 197–213. — 1933. ‘Kolon und Satz: Beobachtungen zur Gliederung des antiken Satzes II.’ Nachrichten von der Gesellschaft der Wissenschaften zu Göttingen, Philologisch- Historische Klasse, 1933: 319–354. — , ed. 1950. Aeschylus, Agamemnon, 3 volumes, Oxford: Clarendon. Fraenkel, H. 1955. Wege und Formen frühgriechischen Denkens, Munich: Beck. Fraenkel, J.J. 1947. ‘A Question in Connection with Greek Particles.’ Mnemosyne, 13: 183–201. Fraser, B.L. 1994. ‘A Study of Narrative Irony in the Oresteia: A Literary and Linguistic Approach’. M.Phil. dissertation, Cambridge University. Frazier, L. 1985. ‘Syntactic Complexity’ in Dowty, Karttunen, and Zwicky 1985: 129–189. — and Fodor, J.D. 1978. ‘The Sausage Machine: A New Two-stage Parsing Model.’ Cognition, 6: 291–325. Frege, G. 1952. ‘On Sense and Reference’ (first published 1892), in Geach and Black 1952: 56–78. French, P.A., Uehling, T.E. and Wettstein, H.K., eds. 1989. Contemporary Perspectives in the Philosophy of Language II. Midwest Studies in Philosophy XIV. Notre Dame: University of Notre Dame Press . 347 Friedrich, J. 1960. Hethitisches Elementarbuch, Vol 1: Kurzgefasste Grammatik, 2nd Edition, Heidelberg: Winter. Friedrich, P. 1975. Proto-Indo-European Syntax: The Order of Meaningful Elements, Butte: Journal of Indo-European Studies. Fries, P. H. 1981. ‘On the Status of Theme in English: Arguments from Discourse.’ Forum Linguisticum, 6.1: 1–38. — 1994. ‘On Theme, Rheme and Discourse Goals’ in Coulthard 1994: 229–249. Frisk, H. 1932. Studien zur griechischen Wortstellung, Göteborg: Elanders, published as Göteborgs Högskolas Arsskrift, 39: 1933.1. — 1960–70. Griechisches etymologisches Wörterbuch, 2 volumes, Heidelberg: Winter. Gagarin, M. 1975. ‘The Vote of Athena.’ AJPh, 96: 121–127. Gale, R. 1971. ‘The Fictive Use of Language.’ Philosophy, 46: 324–340. Garnett, R. 1845. ‘On the Antiquity and Import of the Genitive Case.’ Proceedings of the Philological Society, 2.39: 165–176. Garrett, A. 1992. ‘Topics in Lycian Syntax.’ HSF, 105: 200–212. — 1996. ‘Wackernagel’s Law and Unaccusativity in Hittite’ in Halpern and Zwicky 1996: 85–133. Garvie, A.F. 1969. Aeschylus’ Supplices: Play and Trilogy, Cambridge: Cambridge University Press. — , ed. 1986. Aeschylus, Choephori: With Introduction and Commentary, Oxford: Clarendon. Gazdar, G. 1982. ‘Phrase Structure Grammar’ in Jacobson and Pullum 1982: 131–186. — , Klein, E., and Pullum, G.K., eds. 1983. Order, Concord, and Constituency, Dordrecht: Foris Publications. — , Klein, E., Pullum, G.K., and Sag, I. 1985. Generalized Phrase Structure Grammar, Oxford: Blackwell. Geach, P.T., and Black, M., eds. 1952. Translations from the Philosophical Writings of Gottlob Frege, Oxford: Blackwell. Genette, G. 1986. Narrative Discourse, tr. Lewin, J.E., Oxford: Blackwell. Giacalone Ramat and Hopper 1998: see Ramat and Hopper 1998. Gil, D. 1982. ‘Case Marking, Phonological Size, and Linear Order’ in Hopper and Thompson 1982: 117–141. — 1985. ‘A Prosodic Typology of Language.’ Folia Linguistica, 20: 165–231. Gildersleeve, B.L. 1876. ‘On eij with the Future Indicative and ejavn with the Subjunctive in the Tragic Poets.’ TAPhA, 7: 2–23. Gill, C. and McCabe, M.M., eds. 1996. Form and Argument in Late Plato, Oxford: Clarendon. Givón, T. 1976. ‘Topic, Pronoun, and Grammatical Agreement’ in Li 1976, 149–188. 348 — , ed. 1983. Topic Continuity in Discourse: A Quantitative Cross-Language Study, Amsterdam: Benjamins. — 1988. ‘The Pragmatics of Word-Order: Predictability, Importance and Attention’ in Hammond, Moravcsik and Wirth 1988: 243–284. — 1992. ‘On Interpreting Text-Distributional Correlations: Some Methodological Issues’ in Payne 1992: 305–320. — 1995. Functionalism and Grammar, Amsterdam: Benjamins. Goffman, E. 1974, Frame Analysis: An Essay on the Organisation of Experience, Cambridge MA: Harvard University Press. Golston, C. 1989. ‘Clitics in Homeric Greek: Less Evidence that PIE was Head-Final’ in Hall et al. 1989: 344–355. Gonda, J. 1954a. ‘The Original Character of the Indo-European Relative Pronoun io-.’ Lingua, 4: 1–41. — 1954b. ‘The History and Original Function of the Indo-European Particle kue, Especially in Greek and Latin.’ Mnemosyne, 4th Series, 7, 177–214: 265–296. — 1958. ‘On the So-called Proleptic Accusative in Greek.’ Mnemosyne, Series 4, 11: 117– 122. — 1980. The Character of the Indo-European Moods: With Special Regard to Greek and Sanskrit, Wiesbaden: Harrassowitz. Goodell, T.D. 1890. ‘The Order of Words in Greek.’ TAPhA, 21: 5–47. Goodwin, W.W. 1889. Syntax of the Moods and Tenses of the Greek Verb, Revised Edition, London: Macmillan. Green, G.M. 1989. Pragmatics and Natural Language Understanding, Hillsdale NJ: Erlbaum. Green, K., ed. 1995. New Essays in Deixis: Discourse, Narrative, Literature, Amsterdam: Editions Rodopi. Greenbaum, S. and Quirk, R. 1990. A Student’s Grammar of the English Language, Harlow: Longman. Greenberg, J.H., ed. 1963. Universals of Language: Report of a Conference Held at Dobbs Ferry, New York, April 13–15, 1961, Cambridge MA: The MIT Press. — , ed. 1978. Universals of Human Language, Volume 4: Syntax, Stanford: Stanford University Press. — 1986. ‘The Realis-Irrealis Continuum in the Classical Greek Conditional’ in Traugott et al. 1986: 247–264. Greimas, A.J. 1966. Sémantique structurale: recherche de méthode, Paris: Librairie Larousse. Grewendorf, G. and Sternefeld, W., eds. 1990. Scrambling and Barriers, Amsterdam: Benjamins. Grice, P. 1989. Studies in the Way of Words, Cambridge MA: Harvard University Press. 349 Griffith, M. 1977. The Authenticity of Prometheus Bound, Cambridge: Cambridge University Press. Groot, A.W. de 1919. A Handbook of Antique Prose-Rhythm, 1: History of Greek Prose- Metre, Demosthenes, Plato, Philo, Plutarch, and Others. Bibliography, Curves, Index, Groningen: Wolters. — 1935. ‘Wesen und Gesetze der Caesur ein Kapitel der allgemeinen Versbaulehre.’ Mnemosyne, 3rd Series, 2, 81–154. Gruber, J.S. 1976. Lexical Structures in Syntax and Semantics, Amsterdam: North-Holland Publishing Company. Guiraud, C. 1962. La Phrase nominale en grec: d’Homère a Euripide, Paris: Klincksieck. Gumperz, J. J. and Hymes, D., eds. 1972. Directions in Sociolinguistics: The Ethnography of Communication, New York: Holt, Rinehart and Winston. Gundel, J.K. 1988. ‘Universals of Topic-Comment Structure’ in Hammond, Moravcsik and Wirth 1988: 209–239. — , Houlihan, K. and Sanders, G. 1988. ‘On the Functions of Marked and Unmarked Terms’ in Hammond, Moravcsik and Wirth 1988: 285–301. Haegeman, L., ed. 1997. The New Comparative Syntax, London: Addison Wesley Longman. Haider, H. 1984. ‘Was zu haben ist und was zu sein hat.’ Papiere zur Linguistik, 30: 23–36. — 1990. ‘Topicalization and Other Puzzles of German Syntax’ in Grewendorf and Sternefeld 1990: 93–112. Haiman, J. 1978. ‘Conditionals are Topics.’ Language, 54: 564–589. — , ed. 1985. Iconicity in Syntax, Amsterdam: Benjamins. Hale, K. 1976. ‘ The Adjoined Relative Clause in Australia’ in Dixon 1976: 78–105. — 1983. ‘Warlpiri and the Grammar of Non-Configurational Languages.’ Natural Language and Linguistic Theory, 1: 5–47. — 1992. ‘Basic Word Order in Two “Free Word Order” Languages’ in Payne 1992: 63– 82. — and Keyser, S.J., eds. 1993. The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, Cambridge MA: The MIT Press. Hale, M. 1987. ‘Note on Wackernagel’s Law in the Language of the Rigveda’ in Watkins 1987: 38–50. — 1996. ‘Deriving Wackernagel’s Law: Prosodic and Syntactic Factors Determining Clitic Placement in the Language of the Rigveda’ in Halpern and Zwicky 1996: 165– 197. Hall, K., Meacham, M., and Shapiro, R., eds. 1989. Proceedings of the Fifteenth Annual Meeting of the Berkeley Linguistics Society, February 18–20, 1989. General Session and 350 Parassession on Theoretical Issues in Language Reconstruction, Berkeley: Berkeley Linguistics Society. Halliday, M.A.K. 1961. ‘Categories of the Theory of Grammar.’ Word, 17: 241–292. — 1967. ‘Notes on Transitivity and Theme in English Part 2.’ Journal of Linguistics, 3: 199– 244. — 1968. ‘Notes on Transitivity and Theme in English Part 3.’ Journal of Linguistics, 4: 179– 215. — 1970a. ‘Functional Diversity in Language as Seen from a Consideration of Modality and Mood in English.’ Foundations of Language, 6: 322–361. — 1970b. ‘Language Structure and Language Function’ in Lyons 1970: 140–165. — 1982. ‘How is a Text Like a Clause?’ in Allén 1982: 209–247. — and Hasan, R. 1976. Cohesion in English, London: Longman. Halpern, A.L. 1992. Topics in the Placement and Morphology of Clitics, Doctoral Dissertation, Stanford University. Revised version in Halpern 1995. — 1995. On the Placement and Morphology of Clitics, Stanford: Center for the Study of Language and Information. — and Zwicky, A.M., eds. 1996. Approaching Second: Second Position Clitics and Related Phenomena, Stanford: CSLI Publications. Hammond, M., Moravcsik, E. and Wirth, J., eds. 1988. Studies in Syntactic Typology, Typological Studies in Language Series, Vol. 17, Amsterdam: Benjamins. Hancher, M. 1979. ‘The Classification of Cooperative Illocutionary Acts.’ Language in Society, 8: 1–14. Hancock, J.L. 1917. Studies in Stichomythia: A Dissertation Submitted to the Faculty of the Graduate School of Arts and Literature in Candidacy for the Degree of Doctor of Philosophy, Department of Greek, Chicago: University of Chicago Press. Hannay, M. and Vester, E., eds. 1987. ‘Non-Restrictive Relatives and the Representation of Complex Sentences’ in Auwera and Goossens 1987: 39–52. Harris, M.B. 1984. ‘On the Strengths and Weaknesses of a Typological Approach to Historical Syntax’ in Fisiak 1984: 183–197. Hartung, J.A. 1832. Lehre von den Partikeln der griechischen Sprache, Erlangen: Palm und Enke. Hatcher, A.G. 1956. ‘Syntax and the Sentence.’ Word, 12.2: 234–250. Haudry, J. 1973. ‘Parataxe, hypotaxe, et corrélation dans la phrase latine.’ BSL, 68: 147–186. — 1977. L’Emploi des cas en védique: introduction à l’étude des cas en indo-européen, Lyons: Editions l’Hermès. Hausser, R. 1989. Computation of Language: An Essay on Syntax, Semantics, and Pragmatics in Natural Man-Machine Communication, Berlin: Springer. 351 Havelock, E. A. 1978. The Greek Concept of Justice: From its Shadow in Homer to its Substance in Plato, Cambridge MA: Harvard University Press. Havers, W. 1926. ‘Der sog. “Nominativus pendens”.’ IF, 43: 207–257. — 1931. Handbuch der erklärenden Syntax: Ein Versuch zur Erforschung der Bedingungen und Triebkräfte in Syntax und Stilistik, Heidelberg: Winters. Hawkins, J.A. 1978. Definiteness and Indefiniteness: A Study in Reference and Grammaticality Prediction, London: Croom Helm. — 1983. Word Order Universals, New York: Academic Press. — , ed. 1988. Explaining Language Universals, Oxford: Blackwell. — 1990. ‘A Parsing Theory of Word Order Universals.’ Linguistic Inquiry, 21: 223–261. — 1994. A Performance Theory of Order and Constituency, Cambridge: Cambridge University Press. Headlam, W. 1891. On Editing Aeschylus: A Criticism, London: Nutt. — 1938. See Thomson 1938. Heath, M. 1987. The Poetics of Greek Tragedy, London: Duckworth. Helm, J., ed. 1966. Essays on the Verbal and Visual Arts, Seattle: University of Washington Press. Herington, C.J. 1970. The Author of the Prometheus Bound, Austin: University of Texas Press. Herman, V. 1991. ‘Dramatic Dialogue and the Systematics of Turn-taking.’ Semiotica, 83: 97– 121. Hermann, G., ed. 1852. Aeschyli tragoediae, 2 volumes, Leipzig: Weidmann. Hernadi, P., ed. 1982. The Horizon of Literature, Lincoln: University of Nebraska Press. Hester, D.A. 1981. ‘The Casting Vote.’ AJPh, 102: 265–274. Hettrich, H. 1989. Untersuchungen zur Hypotaxe im Vedischen, Berlin: de Gruyter. Hock, H.H. 1982. ‘Clitic Verbs in PIE or Discourse-Based Verb Fronting? Sanskrit sá hova —ca gárgyaḥ and Conjoiners in Avestan and Homeric Greek.’ Studies in the Linguistic Sciences, 12: 1–38. — 1996. ‘Who’s on First? Towards a Prosodic Account of P2 Clitics’ in Halpern and Zwicky 1996: 199–270. Hockett, C.F. 1958. A Course in Modern Linguistics, New York: Macmillan. Holland, J.H., Holyoak, K.J., Nisbett, R.E., and Thagard, P.R. 1986. Induction: Processes of Inference, Learning, and Discovery, Cambridge MA: The MIT Press. Hooper, J.B. and Thompson, S.A. 1973. ‘On the Applicability of Root Transformations.’ Linguistic Inquiry, 4: 465–498. Hopper, P.J., and Thompson, S.A. 1980. ‘Transitivity in Grammar and Discourse.’ Language, 56: 251–299. 352 — — , eds. 1982. Syntax and Semantics, Volume 15: Studies in Transitivity, New York: Academic Press. Horn, G.M. 1985. ‘Raising and Complementation.’ Linguistics, 23: 813–850. Hornblower, S. and Spawforth, A., eds. 1996. The Oxford Classical Dictionary, Third Edition, Oxford: Oxford University Press. Horrocks, G. 1983. ‘The Order of Constituents in Modern Greek’ in Gazdar, Klein, and Pullum 1983: 95–111. — 1990. ‘Clitics in Greek: A Diachronic Review’ in Roussou and Panteli 1990: 35–52 . — 1994. ‘Subjects and Configurationality: Modern Greek Clause Structure.’ Journal of Linguistics, 30: 81–109. — 1995. ‘On Condition. Aspect and Modality in the History of Greek.’ PCPhS, 41: 153– 173. — 1997. Greek: A History of the Language and its Speakers, London: Longman. — and Stavrou, N. 1987. ‘Bounding Theory and Greek Syntax: Evidence for Wh- Movement in NP.’ Journal of Linguistics, 23: 79–108. Horvath, J. 1986. Focus in the Theory of Grammar and the Syntax of Hungarian, Dordrecht: Foris. — 1995. ‘Structural Focus, Structural Case, and the Notion of Feature-Assignment’ in Kiss 1995b: 28–64. Houben, J.L. 1977. ‘Word-order Change and Subordination in Homeric Greek.’ JIES, 5: 1–8. Householder, F.W., ed. 1972. Syntactic Theory 1, Structuralist: Selected Readings, Harmondsworth: Penguin Books. — tr. and ed. 1981. The Syntax of Apollonius Dyscolus Studies in the History of Linguistics Series 23, Amsterdam: Benjamins. Howorth, R.H. 1955. ‘The Origin of the Use of AN and KE in Indefinite Clauses.’ CQ, 5: 72–93. Huck, G.J., and Ojeda, A.E., eds. 1987. Syntax and Semantics, Vol. 20: Discontinuous Constituency, Orlando: Academic Press. Hudson, R.A. 1984. Word Grammar, Oxford: Blackwell. — 1987. ‘Zwicky on Heads.’ Journal of Linguistics, 23: 109–132. Humbert, J. 1960. Syntaxe grecque, 3rd Edition, Paris: Klincksieck. Humboldt, W. von 1988. On Language: The Diversity of Human Language-Structure and its Influence on the Mental Development of Mankind, tr. Heath, P., Cambridge: Cambridge University Press. Hyman, L.M. 1975. ‘On the Change from SOV to SVO: Evidence from Niger-Congo’ in Li 1975: 113–147. Jackendoff, R. 1972. Semantic Interpretation in Generative Grammar, Cambridge MA: The MIT Press. — 1977. X’ Syntax: A Study of Phrase Structure, Cambridge MA: The MIT Press. 353 — 1983. Semantics and Cognition, Cambridge MA: The MIT Press. Jacobs, H. 1997. ‘Latin Enclitic Stress Revisited.’ Linguistic Inquiry, 28: 648–661. Jacobson, P. and Pullum, G.K., eds. 1982. The Nature of Syntactic Representation, Dordrecht: Reidel. Jacquinod, B. 1997. ‘Sur le rôle pragmatique de kaiv toi’ in Rijksbaron 1997a: 131–149. Jakobson, R. 1937. ‘On Ancient Greek Prosody’ in Jakobson 1962: 262–271. — 1956. ‘Two Aspects of Language and Two Types of Aphasic Disturbances’ in Jakobson 1987: 95–114. — 1958. ‘Linguistics and Poetics’ in Jakobson 1987: 62–94. — 1962. Roman Jakobson: Selected Writings, 1: Phonological Studies, The Hague: Mouton. — 1987. Language in Literature, eds. Pomorska, K. and Rudy, S., Cambridge MA: Harvard University Press. Jannaris, A.N. 1897. An Historical Greek Grammar: Chiefly of the Attic Dialect, as Written and Spoken from Classical Antiquity down to the Present Time, London: Macmillan. Janse, M. 1990. ‘ Wackernagel’s Law’ in Bahner et al. 1990: 2645–2649. Jarvella, R. and Klein, W., eds. 1982. Speech, Place, and Action: Studies in Deixis and Related Topics, Chichester: Wiley. Jebb, R.C. 1981. Sophocles, Oedipus Tyrannus, Edited with Introduction and Notes, first published 1885 by Cambridge University Press. London: Bristol Classical Press. Jens, W. 1955. Die Stichomythie in der frühen griechischen Tragödie, Munich: Beck. Jespersen, O. 1922. Language: Its Nature, Development and Origin, London: Allen and Unwin. — 1924. The Philosophy of Grammar, London: Allen and Unwin. — 1937. Analytic Syntax, London: Allen and Unwin. — 1940. A Modern English Grammar on Historical Principles, Volume 5, Syntax 4, Copenhagen: Munksgaard. — 1949. A Modern English Grammar on Historical Principles, Volume 7, Copenhagen: Munksgaard. Jo, M.-J. 1995. ‘The Theory of Syntactic Focalization Based on a Subcategorization Feature of Verbs’ in Kiss 1995b: 335–374. Johnson-Laird, P.N. 1983. Mental Models: Towards a Cognitive Science of Language, Inference, and Consciousness, Cambridge: Cambridge University Press. — 1986. ‘Conditionals and Mental Models’ in Traugott et al. 1986: 55–75. Jones, F. 1991. ‘Subject, Topic, Given and Salient: Sentence-Beginnings in Latin.’ PCPhS, 37: 81–105. Jones, H.L. 1910. The Poetic Plural of Greek Tragedy: In the Light of Homeric Usage, Ithaca: Longman, Green and Company. 354 Jones, H.S., ed. 1902. Thucydidis historiae. 2nd Edition, Volume 2, Oxford: Clarendon. Jong, I.J.F. de 1997. ‘GAR Introducing Embedded Narratives’ in Rijksbaron 1997a: 175–185. Josephson, F. 1972. The Function of the Sentence Particles in Old and Middle Hittite, Uppsala: Skriv Service. Kahn, C.H. 1996. Plato and the Socratic Dialogue: The Philosophical Use of a Literary Form, Cambridge: Cambridge University Press. Kaimio, M. 1970. The Chorus of Greek Drama: Within the Light of the Person and Number Used, Helsinki: Societas Scientiarum Fennica. Kaisse, E.M. 1985. Connected Speech: The Interaction of Syntax and Phonology, Orlando FL: Academic Press. Karttunen, L. 1971. ‘Some Observations on Factivity.’ Papers in Linguistics, 4: 55–69. — 1973. ‘Presuppositions of Compound Sentences.’ Linguistic Inquiry, 4: 169–93. — 1977. ‘Syntax and Semantics of Questions.’ Linguistics and Philosophy, 1: 3–44. Kashket, M.B. 1991. ‘Parsing Warlpiri: A Free Word Order Language’ in Berwick, Abney and Tenny 1991: 123–151. Kassel, R., ed. 1965. Aristotelis de arte poetica liber, Oxford: Clarendon. Katz, J. J. and Fodor, J.A. 1963. ‘Structure of a Semantic Theory.’ Language, 29: 170–210. Kaufer, D. 1977. ‘Irony and Rhetorical Strategy.’ Philosophy and Rhetoric, 10: 90–110. Kayne, R.S. 1984. Connectedness and Binary Branching, Dordrecht: Foris Publications. — 1994. The Antisymmetry of Syntax. Linguistic Inquiry Monograph 25, Cambridge MA: The MIT Press. Keating, P.A., ed. 1994. Phonological Structure and Phonetic Form: Papers in Laboratory Phonology III, Cambridge: Cambridge University Press. Keenan, E.L. 1985. ‘Relative Clauses’ in Shopen 1985: 141–170. — and Hull, R.D. 1973. ‘The Logical Syntax of Direct and Indirect Questions’ in Corum, Smith-Stark, and Weiser 1973: 348–371. Keil, F.C. 1979. Semantic and Conceptual Development: An Ontological Perspective, Cambridge MA: Harvard University Press. Keil, H., ed. 1857–1880. Grammatici latini, 8 volumes, Leipzig: Teubner. — , ed. 1874. Scriptores artis metricae, Volume 6 of Keil 1857–1880, Leipzig: Teubner. Kennedy, A.K. 1983. Dramatic Dialogue: The Duologue of Personal Encounter, Cambridge: Cambridge University Press. Kennedy, G.A. 1963. The Art of Persuasion in Greece, Princeton: Princeton University Press. — 1992. ‘A Hoot in the Dark: The Evolution of General Rhetoric.’ Philosophy and Rhetoric, 25.1: 1–21. Keyser, S. Jay, ed. 1978. Recent Transformational Studies in European Languages, Cambridge MA: The MIT Press. 355 Kieckers, E. 1911. Die Stellung des Verbs im Griechischen und in den verwandten Sprachen, Erster Teil: Die Stellung des Verbs im einfachenHauptsatze und im Nachsatze nach den griechischen Inschriften und der älteren griechischen Prosa, verglichen mit den verwandten Sprachen Vol. 2 of Brugmann, K. and Thumb, A., eds. Untersuchungen zur indogermanischen Sprach- und Kulturwissenschaft, Strassburg: Trübner. Kierkegaard, S. 1968. The Concept of Irony: With Constant Reference to Socrates, tr. Capel, L.M., Bloomington: University of Indiana Press. Kiparsky, P. 1968. ‘Tense and Mood in IE Syntax.’ FL, 4: 30–57. — 1973. ‘The Inflectional Accent of Indo-European.’ Language, 49: 794–849. — 1995. ‘Indo-European Origins of Germanic Syntax’ in Battye and Roberts 1995: 140– 169. — and Kiparsky, C. 1970. ‘Fact’ in Bierwisch and Heidolph 1970: 143–173. Kiss, K.E. 1995a. ‘Discourse Configurational Languages Introduction’ in Kiss 1995b: 3–27. — , ed. 1995b. Discourse Configurational Languages, New York: Oxford University Press. Klaiman, M.H. 1976. ‘Correlative Clauses and IE Syntactic Reconstruction’ in Steever, Walker, and Mufwene 1976: 159–168. Klavans, J.L. 1985. ‘The Independence of Syntax and Phonology in Cliticization.’ Language, 61: 95–120. Klotz, R. 1835–42. See Devarius 1835–42. Knox, A.D. 1932. ‘The Early Iambus.’ Philogogus, 87: 18–39. Koopman, H. and Sportiche, D. 1982. ‘Variables and the Bijection Principle.’ The Linguistic Review, 2: 139–160. — 1991. ‘The Position of Subjects.’ Lingua, 85: 211–258. Koppers, B.T. 1959. Negative Conditional Sentences in Greek and Some Other Indo-European Languages, The Hague: Westerbaan. Korpimies, L. 1983. A Linguistic Approach to the Analysis of a Dramatic Text: A Study in Discourse Analysis and Cohesion with Special Reference to The Birthday Party by Harold Pinter, Jyväskylä: University of Jyväskylä. Koster, J. 1978. ‘Why Subject Sentences Don’t Exist’ in Keyser 1978: 53–64. — 1987. Domains and Dynasties: The Radical Autonomy of Syntax, Dordrecht: Foris Publications. Kovacs, D. 1986. ‘On Medea’s Great Monologue E.Med.1021–80.’ CQ, 36: 343–352. Kranz, W. 1933. Stasimon: Untersuchungen zu Form und Gehalt der griechischen Tragödie, Berlin: Weidmann. Kroon, C.H.M. 1995. Discourse Particles in Latin: A Study of nam, enim, autem, vero and at, Amsterdam: Gieben. Krupa, V. 1982. ‘Syntactic Typology and Linearization.’ Language, 58: 639–645. 356 Kühner, R. 1898. Ausführliche Grammatik der griechischen Sprache II: Satzlehre, Volume 1, revised by Gerth, B., 3rd Edition, Hanover: Hahnsche. — 1904. Ausführliche Grammatik der griechischen Sprache II: Satzlehre, Volume 2, revised by Gerth, B., 3rd edition, Hanover: Hahnsche. — 1914. Ausführliche Grammatik der lateinischen Sprache II: Satzlehre, Volume 2, revised by Stegmann, C., Hanover: Hahnsche. Kuno, S. 1974. ‘The Position of Relative Clauses and Conjunctions.’ Linguistic Inquiry, 5: 117– 136. Kuppevelt, J. van 1995. ‘Discourse Structure, Topicality and Questioning.’ Journal of Linguistics, 31: 109–147. Kuroda, S.-Y. 1988. ‘Whether We Agree or Not: A Comparative Syntax of English and Japanese.’ Linguisticae Investigationes, 12: 1–47. Kurzová, H. 1981. Der Relativsatz in den indoeuropäischen Sprachen, Hamburg: Buske. — 1985. ‘Zum Typus des altgriechischen.’ Listy Filologické, 108: 3–13. Labov, W. 1970. ‘The Study of Language in its Social Context.’ Studium Generale, 23: 30–87. — and Waletsky, J. 1966. ‘Narrative Analysis: Oral Versions of Personal Experience’ in Helm 1966: 12–44. Lakoff, R.T. 1968. Abstract Syntax and Latin Complementation, Cambridge MA: The MIT Press. Lambrecht, K. 1994. Information Structure and Sentence Form: Topic, Focus, and the Mental Representations of Discourse Referents, Cambridge: Cambridge University Press. Lapointe, S.G. 1988. ‘Toward a Unified Theory of Agreement’ in Barlow and Ferguson 1988: 67–87. Lasnik, H. and Stowell, T. 1991. ‘Weakest Crossover.’ Linguistic Inquiry, 22: 687–720. Lebeck, A. 1971. The Oresteia: A Study in Language and Structure, Washington DC: Center for Hellenic Studies, distributed by Harvard University Press. Lehmann, C. 1984. Der Relativsatz: Typologie seiner Struk turten, Theorie seiner Funktionen, Kompendium seiner Grammatik, Tübingen: Narr. — 1988. ‘On the Function of Agreement’ in Barlow and Ferguson 1988: 55–65. — 1989. ‘Latin Subordination in Typological Perspective’ in Calboli 1989: 153–179. Lehmann, W.P. 1973. ‘A Structural Principle of Language and its Implications.’ Language, 49: 47–66. — 1974. Proto-Indo-European Syntax, Austin: University of Texas Press. — 1976a. ‘From Topic to Subject in Indo-European’ in Li 1976: 445–456. — 1976b. ‘A Preface to Diachronic Syntactic Investigation’ in Steever, Walker, and Mufwene 1976: 169–178. — , ed. 1978. Syntactic Typology, Hassocks (Sussex): The Harvester Press. 357 — , ed. 1986. Language Typology 1985: Papers from the Linguistic Typology Symposium, Moscow, 9–13 December 1985, Amsterdam: Benjamins. — 1993. Theoretical Bases of Indo-European Linguistics, London: Routledge. Lepore, E. and Loewer, B. 1989. ‘You can say that again’ in French et al. 1989: 338–356. Lepschy, G., ed, 1994. History of Linguistics. Volume II: Classical and Medieval Linguistics, London: Longman. Lesky, A. 1966. ‘Decision and Responsibility in the Tragedy of Aeschylus.’ JHS, 86: 78–85. — 1983. Greek Tragic Poetry, tr. Dillon, M., New Haven: Yale University Press. Létoublon, F. 1992a. ‘La deixis spatio-temporelle et le système verbal: le cas du grec ancien.’ in Morel and Danon-Boileau 1992: 265–276. — , ed. 1992b. La Langue et les textes en grec ancien: actes du colloque Pierre Chantraine, Grenoble, 5–8 Septembre 1989, Amsterdam: Gieben. Levinson, S.C. 1983. Pragmatics, Cambridge: Cambridge University Press. Lewis, D. 1973. ‘Causation.’ Journal of Philosophy, 70: 556–567. Li, C.N., ed. 1975. Word Order and Word Order Change, Austin: University of Texas Press. — , ed. 1976. Subject and Topic, New York: Academic Press. — , ed. 1977. Mechanisms of Syntactic Change, Austin: University of Texas Press. — and Thompson, S.A. 1976. ‘Subject and Topic: A New Typology of Language’ in Li 1976: 457–490. Liberman, M. and Prince, A. 1977. ‘On Stress and Linguistic Rhythm.’ Linguistic Inquiry, 8: 249–336. Liddell, H.G. and Scott, R. 1968. A Greek Lexicon, 9th Edition 1940, revised by Jones, H.S. and McKenzie, R., with Supplement 1968. ed. Barber, E.A., Oxford: Clarendon. Lightfoot, D. 1971. ‘Natural Logic and the Moods of Classical Greek’ Unpublished PhD Dissertation, University of Michigan, Ann Arbor: University Microfilms. — 1979. Principles of Diachronic Syntax, Cambridge: Cambridge University Press. — and Hornstein, N., eds. 1994a. Verb Movement, Cambridge: Cambridge University Press. — — 1994b. ‘Verb Movement: An Introduction’ in Lightfoot and Hornstein 1994a: 1–17. Lloyd-Jones, H., ed. 1957. ‘Appendix’ in Smyth, H.W. 1957: 523–603. — , tr. 1979. Aeschylus, Oresteia first publ. 1970 Prentice-Hall, London: Duckworth. — and Wilson, N.G., eds. 1990. Sophoclis fabulae, recognoverunt brevique adnotatione critica instruxerunt, Oxford: Clarendon. Lodge, D., ed. 1988. Modern Criticism and Theory: A Reader, London: Longman. Loepfe, A. 1940. Die Wortstellung im griechischen Sprechsatz erklärt an Stücken aus Platon und Menander, Dissertation, Fribourgh: Paulus. 358 Lord, C. 1976. ‘Evidence for Synactic Reanalysis: From Verb to Complementizer in Kwa’ in Steever, Walker, and Mufwene 1976: 179–191. Lowe, N.J. 1996. ‘Tragic and Homeric Ironies: Response to Rosenmeyer’ in Silk 1996: 520–533. Luraghi, S. 1998. ‘The Grammaticalization of the Left Sentence Boundary in Hittite’ in Ramat and Hopper 1998: 189–210. Lyons, J. 1968. Introduction to Theoretical Linguistics, Cambridge: Cambridge University Press. — , ed. 1970. New Horizons in Linguistics, Harmondsworth: Penguin Books. — 1977. Semantics, 2 vols, Cambridge: Cambridge University Press. — 1982. ‘Deixis and Subjectivity: Loquor, ergo sum?’ in Jarvella and Klein 1982: 101–124. Maas, P. 1962. Greek Metre, tr. Lloyd-Jones, H., Oxford: Clarendon. McCarthy, J.J. 1979. ‘On Stress and Syllabification.’ Linguistic Inquiry, 10: 443–466. McCawley, J.D. 1970. ‘English as a VSO Language.’ Language, 46: 286–299. McCloskey, J. 1983. ‘A VP in a VSO Language?’ in Gazdar, Klein, and Pullum 1983: 9–55. McGonigle, B. and Chalmers, M. 1986. ‘Representation and Strategies During Inference’ in Myers, Brown, and McGonigle 1986: 141–164. Macwhinney, B. and Bates, E., eds. 1989. The Crosslinguistic Study of Sentence Processing, Cambridge: Cambridge University Press. Mair, C. 1990. Infinitival Complement Clauses in English: A Study of Syntax in Discourse, Cambridge: Cambridge University Press. Mallinson, G. and Blake, B.J. 1981. Language Typology: Cross-Linguistic Studies in Syntax, Amsterdam: North-Holland Publishing Company. Maranhão, T., ed. 1990. The Interpretation of Dialogue, Chicago: University of Chicago Press. Marcus, M.P. 1980. A Theory of Syntactic Recognition for Natural Language, Cambridge MA: The MIT Press. Margalit, A., ed. 1979. Meaning and Use. Papers Presented at the Second Jerusalem Philosophical Encounter, April 1976, Dordrecht: Reidel. Marle, J. van, ed. 1993. Historical Linguistics 1991: Papers from the 10th International Conference on Historical Linguistics, Amsterdam, 12–16 August 1991, Amsterdam: Benjamins. Marouzeau, J. 1935. Traité de stylistique appliquée au latin, Paris: Belles Lettres. Marshall, M.H.B. 1987. Verbs, Nouns, and Postpositives in Attic Prose, Edinburgh: Scottish Academic Press. Martin, W. 1986. Recent Theories of Narrative, Ithaca: Cornell University Press. Mathesius, V. 1939. ‘O tak zvaném aktuálnim cleneni vetném.’ SaS, 5: 171–174, cited Firbas 1964: 267–268. Matthei, E. 1982. ‘The Acquisition of Prenominal Modifier Sequences.’ Cognition, 11, 301–332. Matthews, P.H. 1981. Syntax, Cambridge: Cambridge University Press. — 1994. ‘Greek and Latin Linguistics’ in Lepschy 1994: 1–133. 359 May, R. 1985. Logical Form: Its Structure and Derivation, Cambridge MA: The MIT Press. Mecke, J. 1990. ‘Dialogue in Narration: The Narrative Principle’ in Maranhão 1990: 195–215. Meier-Brügger, M. 1992. Griechische Sprachwissenschaft, Band 1, Bibliographie, Einleitung, Syntaxe, Berlin: de Gruyter. Meijering, R. 1987. Literary and Rhetorical Theories in Greek Scholia, Groningen: Forsten. Meillet, A. 1898. ‘Note sur lat. plerique.’ Mémoires de la Société de Linguistique de Paris, 10: 271– 273. — 1903. Introduction a l’étude comparative des langues indo-européennes, Paris: Hachette. — 1908. Les Dialectes indo-européennes, Paris: Champion. — 1912. ‘L’Evolution des formes grammaticales.’ Scientia rivista di scienza, 22.26, reprinted in Meillet 1921: 130–148. — 1921. Linguistique historique et linguistic générale, Paris: Champion. — 1967. The Comparative Method in Historical Linguistics, tr. Ford, G.B., Paris: Champion. — and Vendryes, J. 1927. Traité de grammaire comparée des langues classiques, Paris: Champion. Merkel, R., ed. 1871. Aeschyli quae supersunt in Codice Laurentiano veterrimo: quoad effici potuit et ad cognitionem necesse est visum typis descripta, Oxford: Clarendon. Mey, J.L. 1993. Pragmatics: An Introduction, Oxford: Blackwell. Miehle, H.L. 1974. ‘Relative Constructions in the Rig-Veda, Book 5.’ JIES, 2: 407–434. Miller, D.G. 1974. ‘On the History of Infinitive Complementation in Latin and Greek.’ JIES, 2: 223–246. — 1975. ‘Indo-European: VSO, SOV, SVO, or All Three?.’ Lingua, 37: 31–52. Misener, G. 1908. ‘The eij gavr Wishes.’ CPh, 3: 137–144. Mithun, M. 1992. ‘Is Basic Word Order Universal?’ in Payne 1992: 15–61. Mok, Q.I.M., Spiele, I., and Verhuyck, P.E.R., eds. 1982. Mélanges de linguistique, de littérature et de philologie médiévales offerts à J.R. Smeets, Leiden: Brill. Monro, D.B. 1891. A Grammar of the Homeric Dialect, Second Edition, Oxford: Clarendon. — , and Allen, T.W., eds. 1920. Homeri opera: recognoverunt brevique adnotatione critica instruxerunt David B. Monro et Thomas W. Allen. Iliad. 2nd Edition, 2 volumes, Oxford: Clarendon. Monteil, P. 1963. La Phrase relative en grec ancien: sa formation, son développement, sa structure: des origines a la fin du Ve siècle A.C., Paris: Klincksieck. Moorhouse, A.C. 1955. ‘The Origin of the Infinitive in Greek Indirect Statement.’ AJPh, 76: 176–183. — 1959. Studies in the Greek Negatives, Cardiff: University of Wales Press. Moravcsik, E.A. 1978. ‘Agreement’ in Greenberg 1978: 331–374. — 1988. ‘Agreement and Markedness’ in Barlow and Ferguson 1988: 89–106. 360 — and Wirth, J., eds. 1980. Current Approaches to Syntax Syntax and Semantics Series Vol. 12, New York: Academic Press. Morel, M.-A., and Danon-Boileau, L., eds. 1992. La deixis. Colloque en Sorbonne 8–9 juin 1990, Paris: Presses Universitaires de France. Mosegaard, H. M.-B. 1998. The Function of Discourse Particles: A Study with Special Reference to Spoken Standard French, Amsterdam: Benjamins. Moulton, W.G. 1973. ‘Sentence Features in Embedded Sentences’ in Corum, Smith-Stark, and Weiser 1973: 387–396. Muecke, D.C. 1970. Irony, reprinted 1982 as Irony and the Ironic, London: Methuen. — 1980. The Compass of Irony, London: Methuen. Munro, P. 1982. ‘On the Transitivity of “Say” Verbs’ in Hopper and Thompson 1982: 301–318. Mura, S.S. 1983. ‘Licensing Violations: Legitimate Violations of Grice’s Conversational Principle’ in Craig and Tracy 1983: 101–115. Myers, T., Brown, K., and McGonigle, B., eds. 1986. Reasoning and Discourse Processes, London: Academic Press. Napoli, D.J. 1993. Syntax: Theory and Problems, New York: Oxford University Press. Nespor, M. and Vogel, I. 1986. Prosodic Phonology, Dordrecht: Foris. Neuberger-Donath, R. 1982. ‘Der Gebrauch von o{ti und wJ" in Subjekt- und Objekt-Sätzen.’ Reinishes Museum für Philologie, 125: 252–274. Nevis, J.A., Joseph, B.D., Wanner, D., and Zwicky, A.M. 1994. Clitics: A Comprehensive Bibliography 1892–1991, Amsterdam: Benjamins. Nightingale, A.W. 1996. Genres in Dialogue: Plato and the Construct of Philosophy, New York: Cambridge University Press. Noonan, M. 1985. ‘Complementation’ in Shopen 1985: 42–140. Ogden, C.K. and Richards, I.A. 1946. The Meaning of Meaning: A Study of the Influence of Language Upon Thought and of the Science of Symbolism, London: Routledge and Kegan Paul. Ohmann, R. 1971. ‘Speech Acts and the Definition of Literature.’ Philosophy and Rhetoric, 4.1: 1–19. Olson, D.R., Torrance, N., and Hildyard, A., eds. 1985. Literacy, Language, and Learning: The Nature and Consequences of Reading and Writing, Cambridge: Cambridge University Press. Oostdijk, N. 1990. ‘The Language of Dialogue in Fiction.’ Literary and Linguistic Computing, 5.3: 235–241. Page, D.L., ed. 1938. Euripides, Medea. The Text Edited with Introduction and Commentary, Oxford: Clarendon. — , ed. 1972. Aeschyli septem quae supersunt tragoedias, Oxford: Clarendon. Paley, F.A., ed. 1855. The Tragedies of Aeschylus, London: Whittaker. 361 Palmer, L.R. 1978. Descriptive and Comparative Linguistics: A Critical Introduction, 2nd Edition, London: Faber. — 1980. The Greek Language, London: Faber and Faber. Panhuis, D. 1984. ‘Prolepsis in Greek as a Discourse Strategy.’ Glotta, 62: 26–39. Parry, A., ed. 1980. The Making of Homeric Verse: The Collected Papers of Milman Parry, New York: Arno. Parry, M. 1928. L’Epithète traditionnelle dans Homère, Paris: Belles Lettres, republished in translation in Parry 1980: 1–190. Payne, D.L., ed. 1992. Pragmatics of Word Order Flexibility, Amsterdam: Benjamins. Pelling, C., ed. 1990. Characterization and Individuality in Greek Literature, Oxford:Clarendon. Perlmutter, D. 1980. ‘Relational Grammar’ in Moravcsik and Wirth 1980: 195–229. Pfister, M. 1988. The Theory and Analysis of Drama, Cambridge: Cambridge University Press. Philippaki-Warburton, I. 1985. ‘Word Order in Modern Greek.’ TPhS, 1985: 113–143. Pilkington, A. 1992. ‘Poetic Effects.’ Lingua, 87: 29–51. Pittman, R.S. 1948. ‘Nuclear Structures in Modern Linguistics.’ Language, 24: 287–292. Podlecki, A.J., ed. and tr. 1989. Aeschylus, Eumenides: Edited with an Introduction, Translation and Commentary, Warminster: Aris and Phillips. Pollard, C. and Sag, I. 1988. Information-Based Syntax and Semantics. Vol. 1: Fundamentals, Stanford: CSLI. Pollock, J.-Y. 1989. ‘Verb Movement, UG, and the Structure of IP.’ Linguistic Inquiry, 20: 365– 424. Porson, R. 1802. Euripidis Hecuba, ad fidem manuscriptorum emendata et brevibus notis emendationem potissimum rationes reddentibus instructa. In usum studiosae juventutis, 2nd Edition, Cambridge: University Press. Porter, T.E. 1986. ‘Drama as Text: Mythos and Praxis.’ Word, 37.1: 93–110. Porzig, W. 1923. ‘Die Hypotaxe im Rigveda, 1: Die durch das Pronomen ya charakterisierten Sätze und syntaktischen Gruppen in den ältern Büchern des Rigveda.’ IF, 41: 210–303. Powell, J.G.F. 1996. ‘Prose Rhythm, Latin’ in Hornblower and Spawforth 1996: 1261–2. Prince, G. 1982. Narratology: The Form and Functioning of Narrative, New York: Mouton. Pullum, G. 1977. ‘Word Order Universals and Grammatical Relations’ in Cole and Sadock 1977: 249–277. Quicoli, A.C. 1982. The Structure of Complementation, Ghent: Story-Scientia. Quine, W.V. O. 1968. ‘Reply to Davidson.’ Synthese, 19: 303–305. Quirk, R., Greenbaum, S., Leech, G., and Svartvik, J. 1985. A Comprehensive Grammar of the English Language, London: Longman. Raalte, M. van 1986. Rhythm and Metre: Towards a Systematic Description of Greek Stichic Verse, Assen/Maastricht: Van Gorcum. 362 Radford, A. 1988. Transformational Grammar: A First Course, Cambridge: Cambridge University Press. Ramat, A. G. and Hopper, P., eds. 1998. The Limits of Grammaticalization, Amsterdam: Benjamins. Ransom, E.N. 1986. Complementation: Its Meanings and Forms, Amsterdam: Benjamins. Rauh, G., ed. 1983. Essays on Deixis, Tübingen: Narr. Reinhart, T. 1981. ‘Pragmatics and Linguistics: An Analysis of Sentence Topics.’ Philosophica, 27: 53–94. — 1983. ‘Point of View in Language – The Role of Parentheticals’ in Rauh 1983: 169–194. Reuland, E.J. and Meulen, A.G.B., ter, eds. 1987. The Representation of Indefiniteness, Cambridge MA: The MIT Press. Richards, I.A. 1936. The Philosophy of Rhetoric, Oxford: Oxford University Press. Ricoeur, P. 1978. The Rule of Metaphor: Multi-disciplinary Studies in the Creation of Meaning in Language, tr. Czerny, R., London: Routledge and Kegan Paul. Ridgeway, W. 1910. The Origin of Tragedy: With Special Reference to the Greek Tragedians, Cambridge: Cambridge University Press. Riemann, O. and Goelzer, H. 1897. Grammaire comparée du grec et du latin: syntaxe, Paris: Armand Colin. Rijksbaron, A. 1989. Aristotle, Verb Meaning, and Functional Grammar: Towards a New Typology of States of Affairs. With an Appendix on Aristotle’s Distinction between Kinesis and Energeia, Amsterdam: Gieben. — , ed. 1997a. New Approaches to Greek Particles. Proceedings of the Colloquium Held in Amsterdam, January 4–6, 1996, to Honour C.J. Ruijgh on the Occasion of His Retirement, Amsterdam: Gieben. — 1997b. ‘Introduction’ in Rijksbaron 1997a: 1–14. — 1997c. ‘Adverb or Connector? The Case of kai;...devv’ in Rijksbaron 1997a: 187–208. — , Mulder, H.A., and Wakker, G.C., eds. 1988. In the Footsteps of Raphael Kühner: Proceedings of the International Colloquium in Commemoration of the 150th Anniversary of the Publication of Raphael Kühner’s Ausführliche Grammatik der griechischen Sprache, II: Syntaxe, Amsterdam: Gieben. Rizzi, L. 1990. Relativized Minimality, Cambridge MA: The MIT Press. Roberts, W.R., ed. 1910. Dionysius of Halicarnassus, On Literary Composition, London: Macmillan. — 1982. Demetrius, On Style, Cambridge MA: Harvard University Press. Rosén, H.B. 1994. Is a Comparative Indo-European Syntax Possible?, Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft. Rosenbaum, P.S. 1967. The Grammar of English Predicate Complement Constructions, Cambridge MA: The MIT Press. 363 Rosenmeyer, T.G. 1996. ‘Ironies in Serious Drama’ in Silk 1996: 497–519. Rosier, I., ed. 1988. L’Héritage des grammairiens latins de l’antiquité aux lumières. Actes du Colloque de Chantilly, 2–4 septembre 1987, Paris: Société de l’information grammaticale. Ross, J.R. 1967. ‘Constraints on Variables in Syntax.’ MIT Dissertation, distributed in Microfilm, 1971, Cambridge MA: Massachusetts Institute of Technology. — 1970. ‘Gapping and the Order of Constituents’ in Bierwisch and Heidolph 1970, 249– 259. — 1973. ‘The Penthouse Principle and the Order of Constituents’ in Corum, Smith- Stark, and Weiser 1973: 379–422. Ross, W.D. ed. 1959. Aristotelis, Ars rhetorica, Oxford: Clarendon. Rossi, L.E. 1963. ‘Anceps: vocale, sillaba, elemento.’ RFIC, 91: 52–71. Rostagno, E. 1896. ‘Introduction.’ L’Eschilo Laurenziano: facsimile pubblicato sotto gli auspici del Ministero dell’ Istruzione Pubblica, Florence: R. Biblioteca Mediceo Laurenziana. Roussou, M. and Panteli, S., eds. 1990. Greek Outside Greece, Volume 2, Athens: Diaspora Books. Rozik, E. 1986. ‘Theatrical Irony.’ Theatre Research International, 11: 132–151. — 1993. ‘The Functions of Language in the Theatre.’ Theatre Research International, 18: 104–114. Ruijgh, C.J. 1971. Autour de ‘te epique’: études sur la syntaxe grecque, Amsterdam: Hakkert. — 1990. ‘La Place des enclitiques dans l’ordre des mots chez Homère d’après la loi de Wackernagel’ in Eichner and Rix 1990: 213–233. Ryle, G. 1949. The Concept of Mind, London: Hutchinson. Sacks, H., Schegloff, E.A., and Jefferson, G. 1974. ‘A Simplest Systemics for the Organisation of Turn-taking for Conversation.’ Language, 50: 696–735. Sadock, J.M. 1991. Autolexical Syntax: A Theory of Parallel Representations, Chicago: The University of Chicago Press. Sanford, D.H. 1989. If P, then Q: Conditionals and the Foundations of Reasoning, London: Routledge. Sapir, E. 1921. Language: An Introduction to the Study of Speech, New York: Harcourt, Brace, and World. Sauzet, P. 1989. ‘L’Accent du grec ancien et les relations entre structure métrique et représentation autosegmentale.’ Langages 95: 81–113. Scaglione, A. 1972. The Classical Theory of Composition: From its Origins to the Present. A Historical Survey, Chapel Hill: The University of North Carolina Press. Schachter, P. 1973. ‘Focus and Relativization.’ Language, 49: 19–46. 364 Schäufele, S. 1996. ‘Now That We’re All Here, Where Do We Sit? Phonological Ordering in the Vedic Clause-Initial String’ in Halpern and Zwicky 1996: 447–475. Schegloff, E.A. and Sacks, H. 1973. ‘Opening Up Closings.’ Semiotica, 8: 289–327. Schenkeveld, D.M. 1988. ‘From Particula to Particle – The Genesis of a Class of Words’ in Rosier 1988: 81–93. Schlegel, A.W. von 1840. A Course of Lectures on Dramatic Art and Literature, tr. Black, J., 2nd Edition, London: Templeman. Schmidt, C.E. 1885. Parallel-Homer, Göttingen: Vandenhoeck und Ruprecht. Schmidt, G. 1979. ‘Über indogermanische nominale Relativkonstruktionen.’ IF, 82: 61–74. Schwinge, E.K. 1968. Die Verwendung der Stichomythie in den Dramen des Euripides, Heidelberg: Winter. Schwyzer, E. 1950. Griechische Grammatik: auf der Grundlage von Karl Brugmanns griechischer Grammatik, Zweiter Band: Syntax und syntaktische Stilistik, completed and published by Debrunner, A., Munich: Beck. Seaford, R., ed. 1984. Euripides, Cyclops: With Introduction and Commentary, Oxford: Clarendon. — 1994. Reciprocity and Ritual: Homer and Tragedy in the Developing City-State, Oxford: Clarendon. Searle, J.R. 1969. Speech Acts, Cambridge: Cambridge University Press. — 1974. ‘The Logic of Fictive Discourse.’ New Literary History, 6: 319–332. Sedgewick, G.G. 1935. Of Irony, Especially in Drama University of Toronto Studies, Philology and Literature Series, No. 10, Toronto: University of Toronto Press. Seidensticker, B. 1969. Die Gesprächsverdichtung in den Tragödien Senecas, Heidelberg: Winter. Selkirk, E. 1984. Phonology and Syntax: The Relation Between Sound and Structure, Cambridge MA: The MIT Press. Shannon, C.E. and Weaver, W. 1949. The Mathematical Theory of Communication, Urbana: University of Illinois Press. Sharpe, R.B. 1959. Irony in the Drama, Chapel Hill: University of North Carolina Press. Shattuck-Hufnagel, S. 1994. ‘“Stress Shift” as Early Placement of Pitch Accents: A Comment on Beckman and Edwards’ in Keating 1994: 34–42. Sheppard, J.T. 1928. ‘Agamemnon 964’ PCPhS, 139–141: 13–17. Shiro, M. 1994. ‘Inferences in Discourse Comprehension’ in Coulthard 1994: 167–178. Shklovsky, V. 1917. ‘Art as Technique’ in Lodge 1988: 16–30. Shopen, T., ed. 1985. Language Typology and Syntactic Description, Volume II: Complex Constructions, Cambridge: Cambridge University Press. Sicking, C.M.J. 1997. ‘Particles in Questions in Plato’ in Rijksbaron 1997a: 157–174. 365 — and Ophuijsen, J.M. van 1993. Two Studies in Attic Particle Use: Lysias and Plato Mnemosyne Supplement 129, Leiden: Brill. — and Stork, P. 1996. Two Studies in the Semantics of the Verb in Classical Greek, Leiden: Brill. Siewierska, A. 1991. Functional Grammar, London: Routledge. — 1993. ‘Syntactic Weight vs. Information Structure and Word Order Variation in Polish.’ Journal of Linguistics, 29: 233–265. Sihler, A.L. 1995. New Comparative Grammar of Greek and Latin, New York: Oxford University Press. Sikes, E.E. 1931. The Greek View of Poetry, reprinted 1969, London: Methuen. Silk, M.S., ed. 1996. Tragedy and the Tragic: Greek Theatre and Beyond, Oxford: Clarendon. Silverstein, M. 1976. ‘Hierarchy of Features and Ergativity’ in Dixon 1976: 112–171. Simon, H.A. 1962. ‘The Architecture of Complexity.’ PAPhS, 106: 467–482. Simpson, J. 1991. Warlpiri Morpho-Syntax: A Lexicalist Approach, Dordrecht: Kluwer. Slings, S.R. 1992. ‘Written and Spoken Language: An Exercise in the Pragmatics of the Greek Sentence.’ CPh, 87: 95–109. — 1997. ‘Adversative Relators between Push and Pop’ in Rijksbaron 1997a: 101–129. Slobin, D.I. 1986. ‘The Acquisition and Use of Relative Clauses in Turkic and Indo-European Languages’ in Slobin and Zimmer 1986: 273–294. — and Zimmer, K., eds. 1986. Studies in Turkish Linguistics, Amsterdam: Benjamin. Sluiter, I. 1990. Ancient Grammar in Context: Contributions to the Study of Ancient Linguistic Thought, Amsterdam: VU University Press. — 1997. ‘Parapleromatic Lucubrations’ in Rijksbaron 1997a: 233–246. Smith, N.V., ed. 1982. Mutual Knowledge, London: Academic Press. Smyth, H.W. 1956. Greek Grammar first published 1920, revised by Messing, G.M., Cambridge MA: Harvard University Press. — 1957. Aeschylus with an English Translation Loeb Classical Library, Cambridge MA: Harvard University Press. Sommerstein, A.H. 1973. The Sound Pattern of Ancient Greek. Publications of the Philological Society, 23, Oxford: Blackwell. — , ed. 1989. Aeschylus, Eumenides, Cambridge: Cambridge University Press. Speas, M. 1990. Phrase Structure in Natural Language, Dordrecht: Kluwer. Spencer, A. 1991. Morphological Theory: An Introduction to Word Structure in Generative Grammar, Oxford: Blackwell. Sperber, D. and Wilson, D. 1981. ‘Irony and the Use-Mention Distinction’ in Cole 1981: 295– 318. — 1986. Relevance: Communication and Cognition, Oxford: Blackwell. Staal, J.F. 1967. Word Order in Sanskrit and Universal Grammar, Dordrecht: Reidel. 366 Stanford, W.B. 1936. Greek Metaphor: Studies in Theory and Practice, Oxford: Blackwell. — 1939. Ambiguity in Greek Literature: Studies in Theory and Practice, Oxford: Blackwell. — 1942. Aeschylus in his Style: A Study in Language and Personality, Dublin: Dublin University Press. States, B.O. 1971. Irony and Drama, Ithaca: University of Chicago Press. — 1983. ‘The Dog on the Stage: Theater as Phenomenon.’ New Literary History, 14.2: 373– 388. Steele, S. 1977. ‘Clisis and Diachrony’ in Li 1977: 539–579. Steever, S.B., Walker, C.A., and Mufwene, S.S., eds. 1976. Papers from the Parasession on Diachronic Syntax, April 22, 1976, Chicago: Chicago Linguistics Society. Steinberg, D.D. and Jakobovits, L.A., eds. 1971. Semantics: An Interdisciplinary Reader in Philosophy, Linguistics and Psychology, Cambridge: Cambridge University Press. Steriade, D. 1988. ‘Greek Accent: A Case for Preserving Structure.’ Linguistic Inquiry, 19: 271– 314. Stinton, T.C.W. 1977a. ‘Pause and Period in the Lyrics of Greek Tragedy.’ CQ, 27: 27–66. — 1977b. ‘Interlinear Hiatus in Trimeters.’ CQ, 27: 67–72. Stolnitz, J. 1961. ‘On the Origin of Aesthetic Disinterestedness.’ Journal of Aesthetics and Art Criticism, 20: 131–143. Strawson, P.F. 1952. Introduction to Logical Theory, London: Methuen. Strunk, K. 1992. ‘A propos de quelques catégories marquées et non-marquées dans la grammaire du grec et de l’indo-européen’ in Létoublon 1992: 29–42. Swearingen, C. Jan 1990a. ‘Dialogue and Dialectic: The Logic of Conversation and the Interpretation of Logic’ in Maranhão 1990: 47–71. — 1990b. ‘The Narration of Dialogue and Narration Within Dialogue: The Transition from Story to Logic’ in Britton and Pellegrini 1990: 173–197. — 1991. Rhetoric and Irony: Western Literacy and Western Lies, New York: Oxford University Press. Sweetser, E. 1990. From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure, Cambridge: Cambridge University Press. Szabolcsi, A. 1983. ‘Focussing Properties, or the Trap of First Order.’ Theoretical Linguistics, 10: 125–145. Szemerényi, O.J.L. 1985. ‘Summary of Recent Developments in IE Linguistics.’ TPhS, 1985: 1– 71. Tannen, D., ed. 1993. Framing in Discourse, New York: Oxford University Press. Taplin, O. 1977. The Stagecraft of Aeschylus. The Dramatic Use of Exits and Entrances in Greek Tragedy, Oxford: Clarendon. Taylor, A. 1994. ‘The Change from SOV to SVO in Ancient Greek.’ Language Variation and Change, 6: 1–37. 367 — 1996. ‘A Prosodic Account of Clitic Position in Ancient Greek’ in Halpern and Zwicky 1996: 477–503. Tesnière, L. 1959. Eléments de syntaxe structurale, Paris: Klincksieck. Thirlwall, C. 1833. ‘On the Irony of Sophocles.’ Philological Museum, 2: 483–537. Thompson, A.R. 1948. The Dry Mock: A Study of Irony in Drama, Berkeley: University of California Press. Thompson, E.M. and Jebb, R.C., eds. 1885. Facsimile of the Laurentian Manuscript of Sophocles, London: Society for the Promotion of Hellenic Studies. Thomson, G., ed. 1938. The Oresteia of Aeschylus: Edited with Introduction, Translation, and a Commentary, in Which is Included the Work of the Late Walter G. Headlam, 2 vols, Cambridge: Cambridge University Press. — 1939a. ‘On the Order of Words in Plato and Saint Matthew.’ The Link: A Review of Mediaeval and Modern Greek, 2: 7–17. — 1939b. ‘The Postponement of Interrogatives in Attic Drama.’ CQ, 33: 147–152. Thomson, J.A.K. 1926. Irony: An Historical Introduction, London: Allen and Unwin. Tindale, C.W. and Gough, J. 1987. ‘The Use of Irony in Argumentation.’ Philosophy and Rhetoric, 20: 1–17. Tomlin, R.S. 1986. Basic Word Order: Functional Principles, London: Croom Helm. — , ed. 1987. Coherence and Grounding in Discourse, Amsterdam: Benjamins. Touratier, C. 1980. La Relative: essai de théorie syntaxique, à partir de faits latins, français, allemands, grecs, hébreux, etc., Paris: Klincksieck. — , ed. 1985. Syntaxe et latin: actes du IIème Congrès International de Linguistique Latine. Aix-en-Provence, 28–31 Mars 1983, Marseilles: Laffitte. — 1989. ‘La Subordination: essai de définition’ in Calboli 1989: 219–237. Traugott, E.C., Meulen, A. ter, Reilly, J.S., Ferguson, C.A., eds. 1986. On Conditionals, Cambridge: Cambridge University Press. Trubetzkoy, N.S. 1969. Principles of Phonology, tr. Baltaxe, C.A.M., Berkeley: University of California Press. Tsimpli, I.M. 1995. ‘Focusing in Modern Greek’ in Kiss 1995: 176–206. Tsui, A.B. 1991. ‘Sequencing Rules and Coherence in Discourse.’ Journal of Pragmatics, 15.2: 111–129. Turner, V.W. 1974. Drama, Fields, and Metaphors: Symbolic Action in Human Society, Ithaca: Cornell University Press. Tzanidaki, D.I. 1995. ‘Greek Word Order: Towards a New Approach.’ UCL Working Papers in Linguistics, 7: 347–376. Uhlig, G., ed. 1965. Grammatici graeci, originally published 1883–1910, Leipzig: Teubner, 10 volumes, Hildesheim: Olms. 368 Ussher, R.G. 1978. Euripides, Cyclops: Introduction and Commentary, Rome: Ateneo & Bizzarri. Vendryes, J. 1904. Traité d’accentuation grecque, Paris: Klincksieck. Vennemann, T. 1974. ‘Topics, Subjects, and Word Order: From SXV to SVX via TVX’ in Anderson and Jones 1974: 339–376. — 1984. ‘Typology, Universals and Change of Language’ in Fisiak 1984: 339–376. Verrall, A.W., ed. 1889. The ‘Agamemnon’ of Aeschylus, London: Macmillan. — , ed. 1893. The ‘Choephori’ of Aeschylus, London: Macmillan. — 1895. Euripides the Rationalist: A Study in the History of Art and Religion, Cambridge: Cambridge University Press. — , ed. 1908. The ‘Eumenides’ of Aeschylus, London: Macmillan. Vester, E. 1989. ‘Relative Clauses: A Description of the Indicative-Subjunctive Opposition’ in Calboli 1989: 327–350. Vlastos, G. 1983. ‘The Socratic Elenchus.’ OSAPh, 1: 27–58. — 1987. ‘Socratic Irony.’ CQ, 37: 79–86. Vonlaufen, J. 1974. Studien über Stellung und Gebrauch des lateinischen Relativsatzes: unter besonderer Berücksichtigung von Lukrez, Freiburg: Universitätsverlag. Wackernagel, J. 1892. ‘Über ein Gesetz der indogermanischen Wortstellung.’ IF, 1: 333–436, reprinted in Wackernagel 1953: 1–104. — 1953. Kleine Schriften, Volume 1, Göttingen: Vanderhoeck. Wake, W.C. 1957. ‘Sentence Length Distributions of Greek Authors.’ Journal of the Royal Statistical Society, Series A, 120: 332–346. Wakker, G.C. 1992. ‘Les Propositions relatives dites à valeur conditionnelle’ in Létoublon 1992: 47–61. — 1994. Conditions and Conditionals: An Investigation of Ancient Greek, Amsterdam: Gieben. — 1997. ‘Emphasis and Affirmation: Some Aspects of mhvn in Tragedy’ in Rijksbaron 1997a: 209–231. Wanner, D. 1987. The Development of Romance Clitic Pronouns: From Latin to Old Romance, Berlin: de Gruyter. Wasow, T. 1997. ‘Remarks on Grammatical Weight.’ Language Variation and Change, 9: 81–105. Watkins, C. 1963. ‘Preliminaries to a Historical and Comparative Analysis of the Syntax of the Old Irish Verb.’ Celtica, 6: 1–49. — 1964. ‘Preliminaries to the Reconstruction of Indo-European Sentence Structure.’ Proceedings of the Ninth International Congress (1962) of Linguists: 1035–1042. Reprinted in Householder 1972: 124–134. — 1976. ‘Towards Proto-Indo-European Syntax: Problems and Pseudo-Problems’ in Steever, Walker, and Mufwene 1976: 305–326. 369 — , ed. 1987. Studies in Memory of Warren Cowgill, 1929–1985: Papers From the Fourth East Coast Indo-European Conference. Cornell University, June 6–9, 1985, Berlin: de Gruyter. Webelhuth, G. 1984. ‘German is Configurational.’ The Linguistic Review, 4: 203–246. — 1990. ‘Diagnostics for Structure’ in Grewendorf and Sternefeld 1990: 41–75. — , ed. 1995. Government and Binding Theory and the Minimalist Program: Principles and Parameters in Syntactic Theory, Oxford: Blackwell. Webster, T.B.L. 1941. ‘A Study of Greek Sentence Construction.’ AJPh, 62: 385–415. Wecklein, N., ed. 1885. Aeschyli fabulae: cum lectionibus et scholiis Codicis Medicei et in Agamemnonem Codicis Florentini ab Hieronymo Vitelli denuo collatis, 2 volumes, Berlin: Calvary. Weerman, F. 1989. The V2 Conspiracy: A Synchronic and a Diachronic Analysis of Verbal Positions in Germanic Languages, Dordrecht: Foris. Weil, H. 1869. De l’ordre des mots dans les langues anciennes comparées aux langues modernes: question de grammaire générale, 2nd edition in vol. 3, Collection philologique: recueil de travaux originaux ou traduits relatifs a la philologie et a l’histoire littéraire, Paris: Librairie A. Franck. Werth, P. 1984. Focus, Coherence and Emphasis, London: Croom Helm. West, M.L. 1982. Greek Metre, Oxford: Clarendon. — 1987. Introduction to Greek Metre, Oxford: Clarendon. — , ed. 1990a. Aeschyli tragoediae: cum incerti poetae Prometheo, Stuttgart: Teubner. — 1990b. Studies in Aeschylus, Stuttgart: Teubner. Wierzbicka, A. 1986. ‘Introduction: Special Issue on Particles.’ Journal of Pragmatics, 10: 519– 534. Wilamowitz-Moellendorff, U. von, ed. 1914a. Aeschyli tragoediae, Berlin: Weidmann. — 1914b. Aischylos interpretationen, Berlin: Weidmann. Willett, J., ed. and tr. 1964. Brecht on Theatre: The Development of an Aesthetic, London: Methuen. Williams, C.B. 1970. Style and Vocabulary: Numerical Studies, London: Griffin. Williams, E. 1977. ‘Discourse and Logical Form.’ Linguistic Inquiry, 8: 101–139. — 1980. ‘Predication.’ Linguistic Inquiry, 11: 203–238. — 1981. ‘On the notions “Lexically Related” and “Head of a Word”.’ Linguistic Inquiry, 12: 245–274. — 1984. ‘Grammatical Relations.’ Linguistic Inquiry, 15: 639–673. Wills, J. 1993. ‘Homeric Particle Order.’ HSF, 106: 61–81. Wilson, D. 1975. Presuppositions and Non-Truth-Conditional Semantics, London: Academic Press. Winnington-Ingram, R.P. 1983. Studies in Aeschylus, Cambridge: Cambridge University Press. 370 Winograd, T. 1983. Language as a Cognitive Process: Volume 1: Syntax, Reading MA: Addison-Wesley. Winter, W. 1984. ‘Reconstructing Comparative Linguistics and the Reconstruction of the Syntax of Undocumented Stages in the Development of Languages and Language Families’ in Fisiak 1984: 613–625. Worthington, I., ed. 1994. Persuasion: Greek Rhetoric in Action, London: Routledge and Kegan Paul. Yngve, V.H. 1960. ‘A Model and an Hypothesis for Language Structure.’ Proceedings of the American Philosophical Society, 104: 444–466. Yule, G.U. 1944. The Statistical Study of Literary Vocabulary, Cambridge: Cambridge University Press. Zwart, J.-W. 1997. ‘The Germanic SOV Languages and the Universal Base Hypothesis’ in Haegeman 1997: 246–267. Zwicky, A.M. 1969. ‘Phonological Constraints in Syntactic Description.’ Papers in Linguistics, 1: 411–463. — 1977. On Clitics, Bloomington: Indiana University Linguistics Club. — 1985a. ‘Clitics and Particles.’ Language, 61: 283–305. — 1985b. ‘Heads.’ Journal of Linguistics, 21: 1–20.