The Grammar of Discourse Referents in Homer Chiara Bozzone, University of California, Los Angeles

chiarabozzone@gmail.com

26th Annual UCLA Indo-European Conference, 24 October 2014

1 . Introduction: The Study of Homeric Discourse

[1.1] While many areas of Homeric grammar (especially phonology and morphology) have been explored in substantial detail, the study of Homeric syntax and discourse has been rather unsystematic in its coverage.

• Much attention has been devoted to the syntax of “second position” clitics, after Hale 1987’s work on Indo-Iranian; see Taylor 1990, Goldstein 2010.

• For a study of discourse particles, see Bakker 2007. [1.2] In particular, the topic of referent realization in Homeric discourse remains largely untouched (apart from general observations concerning the article and particle usage, see Chantraine 1953:159; Bakker 1997:108-11). This lacuna is, in a way, paradoxical, since many of the most well explored areas of formulaic technique (Parry’s noun-epithet formulas) precisely serve to introduce (or re-introduce) referents to the discourse, and are mostly employed in larger constructions that serve to manage the transition from one referent to another. [1.3] The aim of this paper is to outline this area of Homeric grammar (namely, how Homer introduces, maintains, and re-introduces referents in the discourse), and to highlight some interesting features thereof. Given the expanse of the topic, it won't be possible to talk about everything. We are going to limit ourselves to:

• Particular features of ellipsis. • Strong vs. weak anaphoric pronouns. • Discourse status of noun-epithet formulas.

[1.4] This contribution is a first step in the treatment of information structure in Homer, which we need to establish questions of word and constituent order (since Greek is a discourse-configurational language, see Kiss 1995:4). Paying attention to the realization of discourse referents is also a way of getting a sense of narrative breaks and how the poet organizes his thought into units.

2 . Basics of Discourse Studies

[2.1] When we study discourse, we study units of language larger than the sentence, and linguistic phenomena that only become conspicuous at that scale. It’s really a study of how people transmit information. [2.2] If we look at an actual specimen of speech, perhaps the first feature that stands out to us is that it is segmented in a number of small units (intonation units).

2

Table 1. Pear story (see Chafe 1980) by J.P., recorded on April 16, 2013.

a. there was A FARMER, b. (.1) who was picking c. (.3) PEARS. d. (.3) he put the pears into the basket, e. .. and then he went up into the tree, f. (.6) but then A KID came along, g. (.4) and he took'em. h. (.88) the kid put'em on his BIKE i. which was also probably stolen, j. (.6) and he was RIDING down - k. (.4) a path - l. .. and then m. (1.17) he went to, n. (.6) past a girl o. (.46) he was checking her out - p. (.1) but WHILE checking her out q. (.69) he HIT a rock r. (.22) and s. (.63) fell over. t. (.72) and then he CRIED. u. (.66) uh no he didn't really cry.

v. (.39) [laughs] and w. (.6) mmm x. (.57) then then some KIDS came along y. and helped him back up, z. they picked up his PEARS. aa. (.5) for him(?) the pears had fallen over, bb. (.84) mm cc. (.16) cause he fell over dd. (.45) and then the kids kept walking along the

way, ee. (.13) the kid went off into the DISTANCE. ff. .. and then at the VERY end, gg. (.9) the KIDS hh. (1.44) passed by the, ii. (1.38) farmer, jj. (.18) again kk. and the farmer sees them. ll. (.21) with the pears. mm. (1.101) and he nn. (.53) and that’s where it ends.

[2.3] Discourse is segmented in this way because our working memory is limited. There is only so much that we can process at one time.

• One-new idea constraint (Chafe 1994:108-19): each intonation unit can contain one and only one new idea. • In a corpus of spoken language, you’ll never find (1), but you’ll find (2):

(1) a tall man saw the red cat (2) there was a man,

he was very tall; one day, he saw a red cat.

• This, by the way, is why people can have such a hard time at conferences. There’s just too much information to transmit and process at one time.

[2.4] If we look at how the narrative is organized, another feature stands out. In the story above, all of the statements are either about the farmer, the boy, or the boys. We can call these referents, which function as organizing principles for the narrative, Discourse Topics.

• Discourse Topics tend to be human and animate. • Note that the narrator could have picked the pears as a discourse topic, but didn’t.

[2.5] Languages have a number of ways of regulating their discourse topics (and topics in general).

3

• English has a preference for realizing its discourse topics as subjects (so does IE, as far as I can tell). [2.6] Discourse topics help streamline the transmission of information.

• they work as known addresses under which the audience can file the new information that comes in.

3 . Referential Chains

[3.1] We can study the behavior of Discourse Topics more closely if we look at referential chains.

Table 2. Referential Chain for ‘farmer’ 1 a FARMER (a) 2 who (b) 3 he (d) 4 he (e) 5 by the … farmer (hh-ii) 6 the farmer (kk) 7 he (mm)

[3.2] For every link of a referential chain, your options are: ellipsis, a pronoun, or some kind of NP. • Normally, when you introduce a new referent, you start with an indefinite expression, then (as a referent is

established in working memory) you can use pronouns or even ellipsis. • Sometimes you’ll use an NP (this time definite). The choice has to do with discourse continuity.

[3.3] We can arrange referent realizations on a scale from implicit to explicit:

ellipsis à pronouns à NPs HIGH CONTINUITY LOW CONTINUITY

Options on the left are correlated with high discourse continuity, options on the right with low discourse

continuity.

[3.4] What determines discourse continuity? Givón (1983) talks about three factors that contribute to how the speaker groups events (thus clauses) together:

• thematic continuity • action continuity • referential continuity (which can mean grammatical continuity, but not necessarily)

Table 3. Referential Chain for ‘kid’ 1 a KID (f) 2 he (g) 3 the kid (h) 4 his (h) 5 he (j) 6 he (m) 7 he (o) 8 he (q) 9 ø (s) 10 he (t) 11 he (u) 12 him (y) 13 his (z) 14 he (cc) 15 the kid (ee)

4

4 . Ell ipsis

4.1 Ellipsis of the Subject/Agent

[4.1.1] Homer is quite generous in its usage of ellipsis (Greek is a pro drop language). Ellipsis of a topical subject is regular:

• most evident in the 1st and 2nd person, where usage of a pronoun will obtain a contrastive reading. • true for 3rd person too.

[4.1.2] Subject ellipsis (or pro drop) is common in languages where the verb takes subject agreement (so Greek and Italian, but not English).

• Subject pro drop seems in fact much more frequent cross-linguistically than the English-style obligatory pronoun in subject position (WALS gives 82 languages like English, and 437 like Greek and Italian). (Dryer 2013)

[4.1.3] There are some remarkable stretches of ellipsis in Homer. This is one of the longest ones. (3) αὐτὰρ ὅ γ’ ἀμφ’ ὤμοισιν ἐδύσετο τεύχεα καλὰ

δῖος Ἀλέξανδρος Ἑλένης πόσις ἠϋκόμοιο. κνημῖδας μὲν πρῶτα περὶ κνήμῃσιν ἔθηκε (ø) καλάς, ἀργυρέοισιν ἐπισφυρίοις ἀραρυίας· δεύτερον αὖ θώρηκα περὶ στήθεσσιν ἔδυνεν(ø) ο ἷο (G) κασιγνήτοιο Λυκάονος· ἥρμοσε (ø) δ’ αὐτῷ. ἀμφὶ δ’ ἄρ’ ὤμοισιν βάλετο (ø) ξίφος ἀργυρόηλον χάλκεον, αὐτὰρ ἔπειτα σάκος μέγα τε στιβαρόν τε· κρατὶ δ’ ἐπ’ ἰφθίμῳ κυνέην εὔτυκτον ἔθηκεν (ø) ἵππουριν· δεινὸν δὲ λόφος καθύπερθεν ἔνευεν· ε ἵλετο (ø) δ’ ἄλκιμον ἔγχος, ὅ οἱ (D) παλάμηφιν ἀρήρει. (Il. 3.328-38)

And he put on his shoulders the beautiful armor, Alexandros, the husband of lovely-haired Helen. First he placed on his legs the beautiful greaves, linked with silver fastenings at the ankles. Second he wore the corslet around his chest, the corslet of his brother Lucaon. It f itted him also (or he fit it to himself). Across the shoulders he slung the sword with nails of silver, a bronze sword, and above it the great shield, huge and heavy. Over his powerful head he set the well-fashioned helmet, with the hourse-hair crest, and the plumes nodded terribly above it. He took up a strong-shafted spear that fitted his hand’s grip. (Translations here and below adapted from Lattimore).

[4.1.4] Long blocks of ellipsis happen in areas of great discourse continuity:

• We are inside a type-scene. • We have a series of chronologically ordered actions leading to one single clear goal. • We have only one Discourse Topic, Paris.

[4.1.5] This tight-knit relationship between clauses, evidenced by extensive ellipsis, reminds one of clause-chaining, a syntactic strategy common in Papuan Languages, where events that are perceived as ‘leading up’ to a goal form a chain in which only the verb expressing the most salient action (the goal) receives complete morphological marking (see Aikhenvald 2008a).

• This is not unlike Kiparsky’s take on the Vedic injunctive as a case of conjunction reduction (Kiparsky 2005).

4.2 Ellipsis of the Agent, Subject, and Object

[4.2.1] A somewhat less common feature of Homeric Greek is that you can elide the direct object too. In the examples below, we see both subjects and objects pro-dropped, after having been established as topical in the previous clause.

5

(4) τὸν δ’ ἐξήρπαξεν Ἀπόλλων

ῥεῖα μάλ’ ὥς τε θεός, ἐκάλυψε δ ’ ἄρ ’ ἠέρι πολλῇ . (Il. 20.443-4) But Apollon caught him up (Hector) easily, since he was a god, and wrapped him in thick mist.

(5) ὃ δ’ Ἀρηΐθοον θεράποντα ἂψ ἵππους στρέψαντα μετάφρενον ὀξέϊ δουρὶ νύξ’, ἀπὸ δ ’ ἅρματος ὦσε · (Il. 20.487-8) But he (Achilles), as Areithoos his servant turned the horses away, stabbed him (Rhigmos) with the sharp spear, and thrust him from the chariot .

Here too, there is high discourse continuity:

• The clause containing subject and object elision expresses a result or an expansion of the action in the previous clause.

• The referents are the same and they occupy the same grammatical roles. [4.2.2] Typologically, object pro-drop is common in languages that have obj. marking on the verb-word as well (highly polysynthethic languages). Like these languages, Greek:

• Is discourse-configurational. • Allows for hyperbaton (Devine and Stevens 2000, Agbayani and Golston 2010).

[4.2.3] Unlike highly polysynthetic languages, Greek does not strictly have object marking on the verb, but:

• In many of these highly-polysynthetic languages - such as K’iche’ (Larsen 1988), Warlpiri (Rice 1970), and Mohawk (Baker 1996) - the marker of the object is ø for the 3rd person singular, so you only see it in the other persons.

• Greek verbs can take personal object clitics, that make it look like it has object agreement on the verb:

(6) οὐδέ με πείσεις (Il. 1.132) You won't convince me.

(7) καί μ’ ἀπάτησε (Il. 9.345)

And he deceived me. [4.2.4] One particular feature of Homer is that you find battle scenes in which you have two discourse topics, and both of them are realized as zeroes. This happens even as they change their argument roles!

(8) πρῶτος δὲ ἄναξ ἀνδρῶν Ἀγαμέμνων ἀρχὸν Ἁλιζώνων Ὀδίον μέγαν ἔκβαλε δίφρου· πρώτῳ γὰρ στρεφθέντι μεταφρένῳ ἐν δόρυ πῆξεν ὤμων μεσσηγύς, διὰ δὲ στήθεσφιν ἔλασσε , δούπησεν δὲ πεσών, ἀράβησε δὲ τεύχε’ ἐπ’ αὐτῷ. (Il. 5.38-42)

And first Agamemnon, lord of men, hurled tall Odion, lord of the Halizones, from his chariot. He (Agamemnon) stuck the spear in the middle of his shoulders, as he (Odios) was turning his back, and pushed it through his chest. He (Odios) made a thud as he fell, and his armor clattered upon him.

6

[4.2.5] What is striking is the transition between ἔλασσε (subject Agamemnon) and δούπησεν (subject Odios). There is no indication of the change of subject! [4.2.6] In order to understand this phenomenon better, we need to introduce a more fine-grained approach to argument roles (terminology from Dixon 1994):

• Agents (A): Subjects of transitive verbs • Subjects (S): Subjects of intransitive verbs • Objects (O): Objects of transitive verbs.

[4.2.7] Languages normally have pivot conditions, which regulate change in argument roles in case of ellipsis. You can either have (Dixon 2012:202):

(9) S/A Pivot (English): The woman (A) saw the man (O) and ø (SHE) ran away (10) S/O Pivot (Dyirbal): The woman (A) saw the man (O) and ø (HE) ran away

Homer seems to be doing both! [4.2.8] Let’s look at our duel scene above, plus two more that immediately follow. Here, I have systematically marked the argument roles of each Discourse Topic as A, S, or O (I indicate Datives (D) and Genitives (G) as well). (11) πρῶτος δὲ ἄναξ ἀνδρῶν Ἀγαμέμνων (A)

ἀρχὸν Ἁλιζώνων Ὀδίον (O) μέγαν ἔκβαλε δίφρου· πρώτῳ γὰρ στρεφθέντι μεταφρένῳ ἐν δόρυ πῆξεν (A) ὤμων μεσσηγύς, διὰ δὲ στήθεσφιν ἔλασσε (A), δούπησεν (S) δὲ πεσών, ἀράβησε δὲ τεύχε’ ἐπ’ αὐτῷ (D). (Il. 5.38-42)

And first Agamemnon, lord of men, hurled tall Odion, lord of the Halizones, from his chariot. He (Agamemnon) stuck the spear in the middle of his shoulders, as he (Odios) was turning his back, and pushed it through his chest. He (Odios) made a thud as he fell, and his armor clattered upon him

(12) Ἰδομενεὺς (A) δ’ ἄρα Φαῖστον (O) ἐνήρατο Μῄονος υἱὸν Βώρου, ὃς (S) ἐκ Τάρνης ἐριβώλακος εἰληλούθει. τὸν (O) μὲν ἄρ’ Ἰδομενεὺς (A) δουρικλυτὸς ἔγχεϊ μακρῷ (45) νύξ’ ἵππων ἐπιβησόμενον κατὰ δεξιὸν ὦμον· ἤριπε (S) δ ’ ἐξ ὀχέων , στυγερὸς δ’ ἄρα μιν (O) σκότος εἷλε. Τὸν (O) μὲν ἄρ’ Ἰδομενῆος (G) ἐσύλευον θεράποντες·

(Il. 5.43-8) Idomeneus killed Phaistos, the son of Maionian Boros, who had come out of Tarne of deep soil. Idomeneus, spear-renowed, stabbed him with the long spear as he was mounting his chariot, in his right shoulder. He dropped from the chariot, and the hateful darkness took hold of him. Idomeneus’ servants stripped him (Phaistos) of the armour.

(13) υἱὸν δὲ Στροφίοιο Σκαμάνδριον αἵμονα θήρης (O)

Ἀτρεΐδης Μενέλαος (A) ἕλ’ ἔγχεϊ ὀξυόεντι (50) ἐσθλὸν θηρητῆρα· δίδαξε (O) γὰρ Ἄρτεμις αὐτὴ (A) βάλλειν ἄγρια πάντα, τά τε τρέφει οὔρεσιν ὕλη· ἀλλ’ οὔ οἱ (D) τότε γε χραῖσμ’ Ἄρτεμις ἰοχέαιρα, οὐδὲ ἑκηβολίαι ᾗσιν τὸ πρίν γε κέκαστο (S) · ἀλλά μιν (O) Ἀτρεΐδης δουρικλειτὸς Μενέλαος (A) (55) πρόσθεν ἕθεν φεύγοντα μετάφρενον οὔτασε δουρὶ ὤμων μεσσηγύς, διὰ δὲ στήθεσφιν ἔλασσεν (A), ἤριπε δὲ πρηνής (S) , ἀράβησε δὲ τεύχε’ ἐπ’ αὐτῷ (D).

(Il. 5.49-58) Menelaus, son of Atreus, killed Strophios’ son, Scamandrios, a fine huntsman, with his sharp spear. Artemis herself had taught him to strike down every wild thing that the mountain forest grows. Yet neither Artemis of the showering arrows could hep him, nor his own long spearcasts in which he used to excel. Instead, Menelaus, spear-renowed, son of Atreus, stabbed him with a spear as he fled before him, in the middle of his shoulders, and pushed it through his chest. He dropped forward on his face, and his armour clattered upon him.

[4.2.9] There is a pattern here: the two warriors are kept apart in terms of their argument role assignment.

7

• The attacker only takes A argument roles. • The victim only takes O and S argument roles.

[4.2.10] This looks like some kind of ergative two-track “pivot”: the attacker is always an A (Ergative), the victim is always an S/O (Absolutive). As long as that is maintained, extensive ellipsis can also be maintanied, without risk for ambiguity.

• You can have two discourse topics being highly active at the same time, but they are not going to take the same argument roles.

• There are more important duels that feature both players as As. But these are not most duels.

5 . Pronouns

[5.1] Pronominal realization of discourse topics (in Homer and beyond) is quite varied. • personal pronouns (1-2nd person) • possessive pronouns • anaphoric pronouns (3rd person) • demonstrative pronouns • relative pronouns

For this talk, wee are going to focus on anaphoric pronouns exclusively, and we are going to discuss two accusative forms in particular. [5.2] Kiparsky 2012 has discussed one 3rd person accusative anaphoric pronoun in Homeric Greek: ἑ / ἕ (which can also have reflexive meaning).

• This occurs quite infrequently in the epics (83x). [5.3] There are two other kinds of 3rd person singular accusative anaphoric pronoun, which are actually doing the heavy lifting in Homeric Greek. We can contrast them in these two (otherwise rather similar) speech introduction constructions:

(14) τὸν δ’ ἄρ’ ὑπόδρα ἰδὼν προσέφη πολύμητις Ὀδυσσεύς· (Od. 18.14) To him, with a scowl, replied clever Odysseus.

(15) ὣς ἔφατ’, Ἀντίνοος δὲ χολώσατο κηρόθι μᾶλλον καί μιν ὑπόδρα ἰδὼν ἔπεα πτερόεντα προσηύδα· (Od. 17.458-9) So he spoke, and Antinoos became more enraged at heart, and with a scowl, spoke to him winged words.

[5.4] These constructions seem largely equivalent: • The new speaker is expressed with an NP • Both τόν and μιν are used to refer to the previous speaker, which is the addressee of the new speech that is

being introduced. • What is the difference?

8

[5.5] Etymologically, • μιν has been linked to the *h1i- anaphoric base we find in Lat. is, ea, id. It is found in Homer, Herodotus, and

Mycenean. Cognate Greek forms are: Cypriot ἴν (from Hesychius), Doric and Tragic νιν. (Weiss 2009:340) • τόν is the beloved *so- /*to- demonstrative base, which works anaphorically in Vedic as well, and will become

the definite article in Greek. [5.6] Formally, their most notable difference their accentuation (which also affects their position in the clause). Table 4 below sums up the possibilities of elliptical and pronominal realization for 3rd person singular discourse topics in Homer. Table 4. Tonic and Atonic Anaphoric Accusative Pronouns in Homeric Greek

Atonic Tonic Accusative ø , μιν τόν, τήν, τό

[5.7] Note that this is similar to the well-known opposition between atonic and tonic pronouns for the 1st and 2nd persons in a number of IE branches (Fortson 2010:142-3). [5.8] In fact, many languages distinguish between strong (tonic) and weak (atonic) anaphoric pronouns. These behave differently in terms of their role in discourse (Walker et al. 1998).

• Weak pronouns signal discourse continuity, while strong pronouns signal discontinuity (e.g. start of a new thematic paragraph).

[5.9] Among modern IE languges, we have studies that observe this for English (Cahn 1995), Italian (Di Eugenio 1998), and Modern Greek (Miltsakaki 2001).

• Centering Theory (Walker et al. 1998). • Work by Givón on Topic Continuity (Givón 1983).

[5.10] Both approaches agree that weak pronouns are more like agreement (they can be cliticized directly to the verb), while strong pronouns have an element of contrast to them. [5.11] In English usage, we can see the complete gradation between ø, weak pronoun, and strong pronoun:

(16) Achilles spoke to his boss and ø was not happy. (Same subject, sequential actions). (17) Achilles spoke to his boss. He was not happy. (Same subject, actions not necessarily sequential). (18) Achilles spoke to his boss. HE was not happy. (Different subject, actions not necessarily sequential).

[5.12] We can compare the English examples with the Homer discourse introductions we have seen above, where μιν (like ø in English) appears in a coordinated clause (high continuity), while τόν appears in an independent clause (lower continuity).

(19) ὣς ἔφατ’, Ἀντίνοος δὲ χολώσατο κηρόθι μᾶλλον καί μιν ὑπόδρα ἰδὼν ἔπεα πτερόεντα προσηύδα· (Od. 17.549)

(20) τὸν δ’ ἄρ’ ὑπόδρα ἰδὼν προσέφη πολύμητις Ὀδυσσεύς· (Od. 18.14)

9

[5.13] More specifically: • μιν signals that, within the same multi-clausal sentence, a former subject (S) has changed its role to object (O). • τόν adds an element of contrast: this is often found in situations where you have two discourse referents being

opposed to each other (like when discussing something!). It can also be used to mark the beginning of a new thematic paragraph.

[5.14] In terms of referents:

• τόν can pick up referents previously realized as verbal arguments and local referents. • μιν can only pick up referents that were previously realized as verbal arguments.

[5.15] In terms of word-order, following Matić’s (2003) framework and terminology:

• τόν is an Exclusive Contrastive Topic; Exclusive Contrastive Topics are sentence initial (in several languages they are also phonologically separate from the rest of the clause).

• μιν is a Continued Topic (which moves to the front of the clause looking for a prosodic host). [5.16] Summing up: while μιν is a weak anaphoric pronoun, τόν is a strong anaphoric pronoun. In English, this would amount to the difference between:

(21) Weak pronoun: Odysseus replied to him. (22) Strong pronoun: to HIM Odysseus replied.

Or in Italian:

(23) Weak pronoun: Gli rispose Odisseo (24) Strong pronoun: A LUI rispose Odisseo

[5.18] An important observation is that discourse continuity is a choice of the speaker, not something that is dictated by outside factors.

• This is why, even within something as regulated as Homeric technique, there are different options for grouping chains of events that apparently achieve rather similar goals (like transitioning from one speaker to another in the examples above, or narrating a duel).

• Chafe (1980:38) observed that when having the same speaker recount the same story multiple times, intonation units would remain similar, but their grouping in larger units (sentences, paragraphs etc.) would be much more variable.

10

6 . Noun Phrases and Noun-Epithet Formulas

[6.1] Just like pronouns can be subdivided in several subcategories with respect to signaling discourse continuity, so can expressions featuring an Noun Phrase. [6.2] Givón (1983:56) gives the following global gradation, spanning from zero anaphora to cleft constructions:

Table 5. The coding of discourse continuity on referents (from Givón 1983:56). zero anaphora (ellipsis)

unstressed pronouns or grammatical agreement (weak pronouns)

stressed pronouns (strong pronouns)

right-dislocated definite NPs

simple definite NPs

left-dislocated definite NPs

contrastive topicalization

cleft/focus constructions

HIGHEST

CONTINUITY LOWEST

CONTINUITY [6.3] Among these many possibilities, we are only going to look at only one thing: the position in the line of noun-epithet formulas and their informational status. [6.4] Noun-epithet formulas are most deceptive among referent realizations in Homer. We often claim that they introduce a new referent. They are said to stage the entrance or epiphany of a god or hero (Bakker 1997:156-83).

• If this were so, though, Homeric Greek would be continuously infringing Preferred Argument Structure. [6.5] Preferred Argument Structure (PAS) is a universal statistical tendency that has been verified in a number of non-related languages (Du Bois 1987, 2003), whereby, in spoken discourse:

• If you have to introduce a new referent (thus a lexical argument), you DO NOT introduce it as an Agent (A). • If you need to introduce a new referent, you introduce it as a Subject (S) or as an Object (O).

[6.6] Homeric noun-epithet formulas are constantly in agent role. This is true for expressions at different levels of formularity, and for noun-epithet formulas of different sizes:

(25) Τὸν δὲ μέγ’ ὀχθήσας προσέφη πόδας ὠκὺς Ἀχιλλεύς· (Il. 22.14) And to him, greatly enraged, spoke swift-footed Achilles.

(26) πάντας γὰρ κατέπεφνε ποδάρκης δῖος Ἀχιλλεὺς (Il. 6.423) For noble swift-footed Achilles killed them all.

(27) Ὣς ἄρα φωνήσας προτέρω ἄγε δῖος Ἀχιλλεύς (Il. 9.199) And thus speaking noble Achilles guided them forward.

• If noun-epithet formulas were introducing new referents, this would be a continous breach of Preferred Argument Structure.

[6.7] If we look closely at the usage of noun-epithet formulas, however, we can see that they are not actually

11

introducing anything new. They are always old information (continued topics) – which explains their position in the clause (see below).

[6.8] The observation that noun-epithet formulas convey old information is true in a general sense:

• heroes/gods that have noun-epithet formulas are well-known members of a closed cast of characters that is expected to appear in epic narrative.

• expecially given the particular medium of epic, where the audience is already familiar with all of the stories, the poet can treat all of the main characters as ‘constantly available’ or ‘situationally available’.

[6.9] But this is also true in a more local sense. Let us take, once again, some speech transitions:

(28) ὣς ἐφάμην, ὁ δέ μ’ αὐτίκ’ ἀμειβόμενος προσέειπεν· (Od. 4.471) Thus I spoke, and he immediately said to me in reply:

(29) ὣς ἐφάμην, ἡ δ’ αὐτίκ’ ἀμείβετο δῖα θεάων· (Od. 4.382) Thus I spoke, and she immediately replied, the noble goddess:

• In the first case, there is no noun-epithet formula. Menelaos is dialoguing with Proteus, the old man of the sea. This is their third exchange, they are the only characters talking, and the audience clearly does not have any trouble recovering who the referent of the pronoun ὁ is, so no nominal realization is necessary.

• In the second case, Menelaus is talking to Proteus’ daughter, Eidothea; this is also their third exchange, they are also the only two characters talking, and there is clearly no need to identify her for the audience. The pronoun ἡ, which starts the sentence, would have been entirely sufficient to resume her as a subject and marking some discourse discontinuity (new paragraph). Yet, this line also presents the epithet δῖα θεάων in right-dislocated (post verbal) position.

[6.10] Within these constructions, noun-epithet formulas seem to be in free variation with ø. I take this as an indication that they are continued topics (part of the presupposition). They are not conveying any new information.

• The referent is either already active, and just mentioned a few lines before. • Or is situationally active.

[6.11] In terms of their linear position, noun-epithet formulas regularly appear at the end of the clause (which is also the end of the line), following the finite verb.

• In terms of Matić (2003) scheme of word order in Classical Greek, this is where Continued Topics go. [6.12] Because of their position in the clause (final) and their informational role (topical), I regard noun-epithet formulas as right-dislocated definite NPs:

• In Givón’s table above, right-dislocated definite NPs are just one step above anaphoric pronouns. • Using a noun-epithet formula instead of a simple anaphoric pronoun just conveys a slightly stronger discourse

boundary. [6.13] A particular keenness to use right-dislocated Continued Topics may be a stylistic device of Homeric narration. [6.14] In fact, other narrative traditions do this too! In a number of unrelated discourse-configurational languages,

12

stylized oral narratives favor “clarificatory” nominal realization of Continued Topics in a right-dislocated position. • Mention of this narrative strategy is found in the grammar of Manambu, a Papuan language (Aikhenvald

2008:538), where it is said to be more frequent with older women “accustomed to explaining and spelling out things for their grandchildren”.

• As well as Southern Quechua (Sanchez 2010:186ff), where right-dislocated Continued Topics have a characteristic phonetic realization (with breathy vowels and a specific intonation pattern) in children’s stories (Sánchez 2010:226-8).

[6.15] One should point out that we have actually observed noun-epithet formulas used in two different kinds of constructions:

• with “pronoun doubling” (where the clause contains an anaphoric pronoun that is coreferential with the righ-dislocated NP):

(30) ὣς ἐφάμην, ἡ δ’ αὐτίκ’ ἀμείβετο δῖα θεάων · (Od. 4.382)

• without “pronoun doubling”: (31) Τὸν δὲ μέγ’ ὀχθήσας προσέφη πόδας ὠκὺς Ἀχιλλεύς· (Il. 22.14)

[6.16] Following the reasoning outlined above, I assume that here too it’s a matter of signaling more or less discontinuity in the discourse, with the tonic anaphoric pronoun adding more contrast. It would be interesting to investigate this matter further.

7 . Conclusions

[7.1] The realization of discourse referents in Homer is an area of the grammar that deserves more attention, especially because of its connection with word order issues and its capacity to reflect thematic boundaries in the poems. [7.2] While the results of this investigation are preliminary, and the investigation itself has been partial, I hope to have driven home the following points:

• Εllipsis in Homer can be extensive and sometimes show a two-track “ergative/absolutive” pivot. • τόν and μιν are respectively strong and weak anaphoric pronouns, and are used to encode different levels of

discourse continuity. • Νoun-epithet formulas are, in most of their usages, to be regarded as continued topics (which explains their

clause-final position).

References

Aikhenvald, Alexandra Y. 2008. The Manambu Language of East Sepik, Papua New Guinea. Oxford: OUP. Aikhenvald, Alexandra Y. 2008a. Semantics and Grammar in Clause Linking, in Dixon, Robert MW, and Alexandra Y.

Aikhenvald (eds). The semantics of clause linking: a cross-linguistic typology 380-402. Oxford University Press, 2009. Agbayani, Brian and Christ Golston. 2010. Phonological Movement in Greek. Language 86.1.133-67. Baker, Mark C. 1996. The Polysynthesis Parameter. Oxford: OUP. Bakker, Egbert J. 1997. Poetry in Speech. Ithaca NY, London: Cornell University Press.Chafe, Wallace L. 1980. The

13

Deployment of Consciousness in the Production of a Narrative. In Wallace L. Chafe (ed.), The Pear Stories: Cognitive, Cultural, and Linguistic Aspects of Narrative Production, 9-50. Norwood: Ablex.

Bozzone, Chiara. 2006. Coesione e coerenza in alcuni libri dell’Iliade: un esame di linguistica testuale. Unpublished BA thesis, Università Cattolica del Sacro Cuore, Milan.

Cahn, Janet. 1995. The Effect of Intonation on Pronoun Referent Resolution, Proceedings of the 32nd Annual Meeting of the Association of Computational Linguistics, Stanford, CA, 155-62.

Chafe, Wallace (ed.). 1980. The Pear Stories: Cognitive, Cultural, and Linguistic Aspects of Narrative Production. Norwood, NJ: Ablex.

Chafe, Wallace. 1994. Discourse, Consciousness, and Time: The Flow and Displacement of Conscious Experience in Speaking and Writing. Chicago: The University of Chicago Press.

Chantraine, Pierre. 1953. Grammaire homérique. Tome 2. Syntaxe. Paris: Klincksieck. De Beaugrande, Robert and Wolfgand Ulrich Dressler. 1981. Introduction to Text Linguistics. London and Ner York: Longman. Devine, A.M. and Stephens, Laurence D. 2000. Discontinuous Syntax: Hyperbaton in Greek. Oxford: OUP. Di Eugenio, Barbara. 1998. Centering in Italian. In Marilyn Walker, Aravind Joshi, and Ellen Prince (eds.) Centering Theory

in Discourse, 115-137. Oxford: OUP. Dixon, R. M. W. 1994. Ergativity. Cambridge: Cambridge UP. Dixon, R. M. W. 2012. Basic Linguistic Theory, Volume III: Further Grammatical Topics. Oxford: OUP. Dryer, Matthew S. 2013. Expression of Pronominal Subjects. In Matthew S. Dryer and Martin Haspelmath (eds.) The World

Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://wals.info/chapter/101, Accessed on 2014-10-22.)

Du Bois, John W. 1987. The Discourse Basis of Ergativity. Language 63.4.805-55. Du Bois, John W., Lorraine E. Kumpf, and William J. Ashby (eds.). 2003. Preferred Argument Structure: Language as Architecture

for Function. Amsterdam: John Benjamins. Erteschik-Shir, Nomi. 2007. Information Structure. Oxford: OUP. Fortson, Benjamin W. IV. 20102. Indo-European Language and Culture: An Introduction. Oxford: Wiley-Blackwell. Givón, Talmy (ed.) 1983. Topic Continuity in Discourse: A Quantitative Cross-language Study. Amsterdam: John Benjamins. Givón, Talmy. 1983a. Topic Continuity in Discouse: The Functional Domain of Switch Reference. In John Haiman and

Pamela Munro (eds.), Switch Reference and Universal Grammar, 51-82. Amsterdam: John Benjamins. Goldstein, David M. 2010. Wackernagel’s Law in Fifth-Century Greek. Ph.D. Dissertation, University of California, Berkeley. Hale, Mark. 1987. Studies in the Comparative Syntax of the Oldest Indo-Iranian Languages. Ph.D. Dissertation, Harvard

University. Kiparsky, Paul. 2005. The Vedic Injunctive: Historical and Synchronic Implications. The Yearbook of South Asian Languages

and Linguistics 35. Kiparsky, Paul. 2012. Greek Anaphora in Cross-Linguistic Perspective. Journal of Greek Linguistics 12. 84-117. Kiss, Katalin É. 1995. Discourse Configurational Languages Introduction. In Katalin É. Kiss (ed.) Discourse Configurational

Languages, 3-27. Oxford: OUP. Larsen, Thomas Walter. 1988. Manifestations of Ergativity in K’iche’ Grammar. Unpublished Ph.D. dissertation. University of

California, Berkeley. Matić, Dejan. 2003. Topic, Focus, and Discourse Structure: Ancient Greek Word Order. Studies in Language 27.3.573-633. Parry, Milman. 1971. The Making of Homeric Verse: the Collected Papers of Milman Parry. Oxford: Clarendon. Reece, Laurie. 1970. Grammar of the Wailbri Language of Central Australia. Sidney: University of Sidney, Australia. Sánchez, Liliana. 2010. The Morphology and Syntax of Topic and Focus: Minimalist Inquiries in the Quechua Periphery.

Amsterdam: John Benjamins. Taylor, Ann. 1990. Clitics and Configurationalilty in Ancient Greek. Ph.D. Dissertation, University of Pennsylvania.

14

Weiss, Michael. 2009. Outline of the Historical and Comparative Grammar of Latin. Ann Arbor: Beech Stave Press.

Appendix: Quantitative Data on Referential Chains in Homer

[1] In 2006, I wrote my BA thesis on cohesion and coherence in Homer. • Coherence: the invisible (conceptual) means by which a text hangs together. • Cohesion: the visible (linguistic) means by which a text hangs together, like anaphoric phenomena.

[2] As part of the study of cohesion, I systematically tracked the referential chains in books 1-4 of the Iliad. The table below shows the counts for Iliad 3, which are largely representative of the situation in the other books. I have several chains because, in any complex narrative, you’ll have multiple chains going at one time – to a limit. We see here again, some effects of working memory limitations.

Table 6. The Surface Realization of Topic Chains in I l iad 31 (from Bozzone 2006:44) Chain #1 Chain #2 Chain #3 Chain #4 Chain #5 Start NPs 41 22 27 10 5 NPs (copies) 7 (4.9%) 4 (3.3%) 16 (22.5%) 6 (27.2%) 2 (33.3%) NPs (synonyms)

5 (3.5%) 7 (5.7%) 3 (4.2%) 0 0

Pronouns 66 (46.4%) 61 (50.4%) 37 (52.1%) 11 (50%) 2 (33.3%) Ellipsis (null anaphora) 64 (45%) 49 (40.4%) 15 (21.1%) 5 (22.7%) 2 (33.3%)

Total nr. of links 142 121 71 22 6 Average length (in links) 3.4 5.5 2.6 2.2 1.2

[3] A few observations here:

• While pronominal realization seems to account for about half of the links in most chains regardless, nominal realization and ellipsis (null anaphora) are highly sensisive to saliency ranking of their chain.

• Ellipsis predominates in higher-ranked chains, and nominal realization predominating in lower-ranked ones. • Beyond chains #1 and #2, nominal realization increases from around 8% to above 20%; at the same time, ellipsis

sharply decreases from around 40% to around 20%. [4] Saliency ranking of a chain impacts its average length as well: beyond chains #1 and #2, chains become sensibly shorter (with chain #5 being most often composed of just one head and one link). A similar effect is seen in the total number of links per each ranking level, which is steadily decreasing from chain #1 onwards. [5] Overall, there seems to be a sharp divide between chains #1 and #2 and the remaining ones, a divide that we can tentatively map onto the cognitive difference between ideas that are at the ‘center’ of our working memory, and ideas that are at the ‘periphery’ thereof.

1 Note that the NP realizations in each chain are sorted into Starts (the first mention of a referent), Copies (verbatim repetitions of the first referent, including equivalent formulas), and Synonyms (definite descriptions such as ‘the old man’ standing in for ‘Priam’).