§1. The present chapter builds on the discussion of discourse segmentation set out in IV.3, so to facilitate understanding I briefly summarize the ideas set out there, where the reader may find a fuller discussion and references. In ancient philosophy and rhetoric, language was described in terms of períodoi, kôla, and kómmata. All three terms are hard to define, and our understanding of them is often strongly determined by English translations. It appears that the term períodos initially described an intonation curve between two prosodically similar moments. The kôlon is a subdivision of the períodos: either a complete thought or a complete part of a thought, with distinct parts, and easy to repeat in a breath. Kómma, finally, is a term applied to particularly short kôla. What underlies all three terms is a focus on performance rather than on syntactical shape (e.g. sentence, clause, phrase), and in this chapter I trace modern and contemporary research that takes the same perspective on discourse.

§2. To ancient scholars, discourse consisted of different-sized units not based primarily on syntactical division, but rather on the sense of completion on the one hand and the speaker’s physical limitations on the other. From the nineteenth century onwards, conversely, the approach to discourse segmentation in ancient Greek texts reveals a strong tendency to regard syntax as the primary structure. This exclusive focus on syntax obscures the important role of particles in the articulation of discourse. A greater sensitivity to the linear presentation of both epic and lyric is needed for an understanding of the function of particles and their host units. The present chapter gathers the relevant evidence, both cross-linguistic and specific to ancient Greek, for the contention that not the sentence or clause, but the discourse act is the basic unit of language use. It is with regard to this smallest subdivsion of discourse that the function of particles is to be understood.

§3. In speech, language is more malleable than the grammatical handbooks would have us believe, as we may note from the loose adherence to grammar even academics reveal in their day-to-day conversations. [1] This is not to say that a grammatical analysis is not a valid approach to such performed texts; both the Homeric and Pindaric corpora show a distinct tendency to obey a set of rules that largely conforms to what the handbooks call Greek grammar. A thorough knowledge of the grammar is indispensable for understanding the texts and allows the reader to note deviations from the standard and to hypothesize explanations for them. [2] Anyone interested in gaining a better understanding of the text’s impact in performance, however, cannot stop at contemplating grammar alone. [3] What we wish to challenge here is not the relevance of a prescriptive syntax for the study of Homer and Pindar, but its primacy.

§4. Consider the following example from the Odyssey, followed by Murray’s translation:

Murray’s pleasant translation tells the story adequately, but it obscures the original word order. In the translation the reader is given a neat sentence, which begins before this excerpt and continues after it. It is worth comparing the punctuation in the Greek (from Von der Mühll’s edition) to that in the English translation. Although the placement of some commas corresponds in the two texts, there are discrepancies. First of all, in the translation there is a comma after “the minstrel,” which allows the reader to focus on the new source of the upcoming discourse, but this comma is left out in the Greek. [4] Conversely, Von der Mühll prints a comma before ὡς, but Murray leaves it out before the corresponding “where” in the translation.

§5. These differences may be partly (or even wholly) due to the fact that English has more or less established rules for comma placement in written texts, whereas such a system did not exist for ancient Greek. [5] However, the choices in the Greek edition are revealing: they show that Von der Mühll feels there should be a comma before the “first-position word” ὡς, while he regards the participial construction ὁρμηθείς θεοῦ as too closely connected to the pronoun ὁ to permit a comma to intervene. Commas in written English are largely syntactic markers, since they mark the segments they divide as parts of the same superstructure: the sentence, which is in turn bounded by full stops. In (t1) above this convention explains the placement of a comma rather than a period after βαλόντες: the following nominative Ἀργεῖοι is taken as the subject of a long subclause, itself part of the sentence of which ὁ is the subject. However, it is open to discussion whether or not the syntactical unit of the sentence has true analytical value in a study of Homer, Pindar, or any other discourse produced in, or meant for, performance.

§6. The reception of a performed text is at least partly linear: the hearer cannot hear the second word before the first word, and likewise the second verse only after the first verse. [6] In cognitive studies it has been established that the working memory used for processing aurally received text is not all that large. [7] Still, we find numerous instances in Greek epic and lyric where the grammatical subject of a clause is not given until two verses after the verb, as in the Odyssey passage above. A close translation looks as follows: [8]

If the Greek audience truly needed the subject from line 502 to complete the construction begun in line 500, one would have to conclude that in performance this text would become rather unsuccessful. In fact, the nominative follows at a point where the audience would have been expected to already know who the story was about; it must therefore be regarded as serving another function. My translation focuses on the articulation of discourse in separate acts, each of which serves at least one purpose in furthering the discourse.

§7. This chapter traces the history of this approach to discourse articulation in Greek in general and Homeric epic in particular. I first outline the ideas presented by Wackernagel and refined by Fraenkel and later scholars (§§9-15), which several contemporary scholars have recently revisited (§16). Then I trace contemporary work on discourse articulation, especially in English, which has focused on spoken discourse (§§17-20). The earlier scholarship on ancient Greek, together with contemporary scholarship on modern languages, provides the foundation for the study of discourse acts (§§21-23).

§8. There are several ways in which we regard the study of discourse articulation as relevant to the analysis of the functions of particles. First of all, particles are important boundary markers in ancient Greek, revealing the production of discourse in manageable chunks in Homer (§§24-30) and Pindar (§§37-45). [9] Second, establishing act boundaries allows for a better appreciation of the host units of different particles (δέ §§31-36 and μέν §§46-62). Third, a close examination of discourse segmentation in Homer and Pindar reveals that particles, particle clusters, or a particle in concert with one word (group) function as separate acts. Closely examining the way particles work on the small scale allows for a more precise understanding of the role of some particles in navigating larger discontinuities in discourse, especially in Homeric narrative (§§64-71) and Pindaric performance (§§72-79).

§9. In 1892 Wackernagel published a seminal article about a rule in the word order of Indo-European languages that he felt was followed so strictly that it deserved the predicate “law.” [10] As Goldstein rightly observes, Wackernagel 1892 presented neither the beginning nor the final form of the argument, but it has become the reference point for the large body of literature that has built upon its ideas. [11] The gist of Wackernagel’s argument is that a group of words, including particles, tend to occur in the second position of a clause, be it a main or a subclause. [12] His argument initially excludes enclitic particles that join sentences, [13] as their role of conjoining the host sentence to the preceding one gives them a natural reason to be in second position. His analysis consequently focuses on κε, which does not function on the level of Satzverbindung. [14] For this enclitic particle, as well as for θην, νυ, and τοι, Wackernagel finds that they have a strong tendency to occur in second position of the sentence. Finally, he notes that a number of non-enclitic particles, called postpositive Partikeln by Krüger, follow the same pattern of eschewing initial position. [15]

§10. Some forty years later, Eduard Fraenkel refined and extended Wackernagel’s ideas [16] in two publications, with the intention of better explaining the many apparent exceptions to the law in extant Latin and Greek literature. [17] Fraenkel’s contention is that the postpositives, such as particles, come in the second position not necessarily only of sentences, but also of smaller syntactical units. [18] He calls these smaller units Kola (henceforth kôlon, kôla), a term inherited from ancient scholarship on the units that make up prose and poetry. [19] Fraenkel analyzes the use of ἄν in Greek classical prose, [20] and concludes that it can generally be considered to occupy the second position of a kôlon, when it occurs later in a clause. [21] Having established this principle, Fraenkel revisited the topic in 1965 in order to show that vocatives generally occur at kôlon boundaries. [22]

§11. We might describe Fraenkel’s approach as a shift from a “map view” to a “route view” of language. [23] That is to say, he regards the text as first and foremost a syntactic construct, but rather than contemplating every sentence as an architectonic whole, he assumes that reception was realized in smaller units, which he calls “syntaktische Kola.” [24] These kôla the listener or reader receives in sequence, and they make sense in their linear order. Speakers group their words and thoughts in these smaller units, and it is the existence of these units that explains supposed enjambment in Roman elegy, “abnormal” placement of ἄν in Greek prose, and improves our understanding of how vocatives are used. Fraenkel’s work would form the basis for a series of studies in Greek linguistics in the following decades; his intuitions have pointed the way for the approach I elaborate in this chapter.

§12. Now with regard to second, or peninitial, position, some explanation is warranted. In the majority of cases, especially in classical prose, a “postpositive” particle occurs in second position in a clause. However, second position does not necessarily mean the second lexical item in a clause. If the preceding word is part of a tight group, such as article and noun or preposition and noun, it is possible for the whole word group to precede the particle. In such cases this positioning of the particle is unproblematic, although it is worth considering why it occurs on some but not all occasions. A first-position particle, like ἀλλά and καί, [25] may similarly cause a particle that comes after it to appear as the third rather than second word. Diagnostically most relevant, then, are those cases where a postposed particle occurs later than in second position in a clause even when none of the two abovementioned situations apply. In these instances it is probably the case that what comes before the peninitial particle and its preceding word is not a full syntactical clause, but a separate kôlon nonetheless.

§13. In his 1959 work on word order in Pindar, Lauer retrieves Fraenkel’s work on kôla to apply it to Pindar’s songs. [26] Like Fraenkel before him, he uses the idea of kôla to demonstrate that apparently divergent and problematic word order in fact obeys the rules of word order with regard to kôla, if not with regard to sentences. Lauer argues that the Pindaric corpus offers more insight into the performative reality than prose does, because of its metrical form. That is to say, whereas prose is only divided into syntactical units (which may or may not have a specific relation to the discourse units it was realized in), poetry is divided into both syntactical and metrical units, which provides slightly more handholds for establishing possible and even probable discourse division. [27]

§14. Throughout his study Lauer adheres closely to Fraenkel’s method, establishing kôlon boundaries by using criteria such as the placement of postpositives, quasi-independent grammatical constructions (such as participia coniuncta), and stylistic practices (such as parallel constructions). Having established these boundaries, he goes on to describe the kôla that emerge. Most of them fall under the open-ended list of types offered by Fraenkel, [28] but Lauer makes a few additional observations. Most importantly, Lauer observes that quite regularly there is a very short kôlon at the beginning of sentences, which he calls Kurzkola, short kôla, but does not further define them. [29] He is similarly taciturn about the term potentielle Kola, potential kôla, which he uses to describe these same short opening kôla, but which might be much more broadly applicable, on which see below. [30] Sometimes these initial kôla are so short that they consist of nothing more than a conjunction, which leads Lauer to conclude that conjunctions can stand outside the kôlon. [31] Finally, Lauer notes that in case of hyperbaton, the two (or more) members, typically an adjective (or a participle in attributive position) and a noun, can be Kolonbildend, opening and closing them. [32]

§15. In his analysis of vocatives, Fraenkel hints at the phonetic realization of kôla and kôlon boundaries, namely that the insertion of a vocative would result in a pause, with different possible effects. [33] For Stinton 1977, phonetic realization is the crucial question: can we establish where pauses might have been heard in Greek discourse? [34] His corpus, lyric passages of Greek tragedy, is quite different from Fraenkel’s material, but “the categories he [sc. Fraenkel] establishes for Greek prose can be readily adapted to the more condensed language of lyric poetry.” [35] Using Fraenkel’s categories on the one hand, and metrical responsion on the other, he makes the argument that both brevis in longo and hiatus are generally allowed only where there is a syntactical pause. His research into the intricate language of lyric shows that meter and sense stand in a somewhat natural relationship. [36]

§16. Recently, the concept of the kôlon has resurfaced in Scheppers’ The Colon Hypothesis. [37] His work is an important step forward because he includes literature from contemporary linguistics. Scheppers builds on the idea of kôla with help from research into modern languages, for which a spoken corpus is available. Like Bakker before him, whose work on Homer I discuss below, Scheppers incorporates the work done on “intonation units.” As the term suggests, the units are parts of spoken discourse that cohere by virtue of forming a single intonational contour, generally bounded by pauses. Scheppers combines the concept of kôlon and that of intonation unit to analyze the prose discourse of Lysias and Plato “as literary representations of ‘spoken’ Greek.” [38] Goldstein likewise turns to intonation in order to further our understanding of the position of enclitics and postpositives. In a thorough study of fifth-century Greek prose, he demonstrates that Wackernagel’s law is best understood as not primarily a syntactic phenomenon, but rather bound to intonational phrases. [39] Scheppers, Goldstein, and Bakker urge us to attempt to hear the intonational division of Greek discourse, and acknowledge its importance to word order and construction of meaning. In our work on particles in poetry and prose we subscribe to this approach to Greek discourse and hope to contribute to its further development. Their work has a solid basis in research on contemporary spoken discourse.

§17. Research on intonation units starts with Halliday’s concept of the “tone group,” which can be distinguished as a separate unit by virtue of having only one “tonic element.” In later scholarship the focus shifts to the “intonational contour,” to fit better with actual practice. [40] Speakers portion their discourse by proceeding in units that are marked by an independent intonational contour, and are often bounded by pauses. This segmentation, it appears, is not (or not only) a function of the physical need to breathe at regular intervals, but rather of the cognitive effort involved in planning speech. [41] Brown and Yule 1983 follow this line of reasoning and choose to call the small segments “information units,” since they reflect the piecemeal addition of information to the discourse. Chafe proposes to use the term “intonation units,” and adds that they are generally a combination of given and new information. [42] The combination of given and new information in intonation units is only a trend: some intonation units in fact contain only new, or, more rarely, only given information. Similarly, he notes that in his corpus of spoken English the correspondence between intonation units and syntactical clauses is in fact quite low. [43]

§18. It is one thing to observe that a certain kind of unit exists in discourse, but quite another to explain why it occurs. There is a range of explanations available, each reflecting the particular approach used in establishing the smallest discourse unit. One group of scholars, represented by Roulet on the one hand (Geneva School) and Sinclair and Coulthard on the other (Birmingham School), approach discourse as a strategic construct with a certain aim, and smaller discourse units as having a certain function in reaching that goal. In their seminal 1975 article on classroom discourse, Sinclair and Coulthard establish the term “discourse act,” analogous to Austin’s speech act. [44] They define discourse acts as the smallest step toward reaching a (sub)goal of the discourse, hierarchically ordered on a scale of act – move – exchange – transaction – lesson. [45] Roulet 1984 adopts part of this terminology (act – move – exchange), [46] and proceeds to apply the ideas to different kinds of material, including both spoken discourse and written texts. Over two decades of work, Roulet redefines his conceptualization of the discourse act several times, but in Roulet et al. 2001 he settles on the discourse act as every “update to the discourse memory,” [47] basing his idea on the work done by Berrendonner. [48]

§19. More recently, cognitive linguists have engaged with the smallest steps in discourse. Chafe regards the intonation unit as a focus of consciousness, “typically expressed with four words of English.” [49] Langacker talks about the same process in terms of “attentional frames,” which function “as instructions to modify the current discourse space in particular ways.” [50] As regards form, attentional frames may well coincide with grammatical constituents, but this is “only a tendency, not an inviolable principle.” [51] The relation between the empirically observable intonation unit and the cognitive process of language production may be linked productively to the work of the Birmingham and Geneva schools. The new focus of consciousness forms an update to the discourse memory (on which see II.4), which is verbalized in a new intonation unit. [52]

§20. There is one more element that contributes to our understanding of the discourse act. In the field of functional discourse grammar, scholars like Kroon, Hannay, and Hengeveld have approached language in a way largely similar to that of the Geneva and Birmingham schools. They have, however, a slightly different perspective: they define acts specifically as the smallest steps toward the main goal of the discourse. Hannay and Kroon, looking back to Kroon’s earlier work on Latin discourse particles, define discourse acts as “the smallest identifiable unit of communicative behaviour,” [53] but they expand it by stipulating that all acts are “strategic steps which the speaker wishes to make.” [54] Hannay and Kroon conclude that focus should shift from a link between discourse units and grammatical units to a link between discourse units and prosodic/orthographic units. [55] To bring all the above elements together: an intonation unit verbalizes a focus of consciousness, which is both an update to the discourse memory and a new step toward the overarching discourse goal.

§21. Discussions of the non-grammatically defined smallest units of discourse have followed divergent paths, but their results converge. The conclusions that contemporary linguists have come to share are worth revisiting briefly. First, it is generally agreed that there is more than one kind of organization at work at the same time in any discourse. This position accords with the view that different kinds of subdivision are possible, whether based on syntax, content, discourse steps, performance, or other criteria. [56] Second, discourse units are not building blocks of a grammatical structure, but the verbalization of frames or foci of consciousness. Third, discourse acts function in terms of updates to discourse memory. Fourth, the smallest discourse unit tends to align with the strategic function of the smallest step toward a discourse goal. By and large, the smallest subdivisions of discourse are no larger than clauses (often smaller), do not consistently map onto syntactical units, represent some elemental progression within a discourse strategy, represent the current cognitive focus of the speaker, and generally manifest in prosodically independent units. [57]

§22. Building on these separate but connected bodies of research, we will henceforth call the smallest subdivisions of discourse “discourse acts.” This choice is not only based on the belief that we should where possible avoid adding to the plethora of terminology used in linguistics, but also on the conviction that the term captures two important features of these units. First, it categorizes the unit as a subdivision of discourse, in whatever form, rather than a subdivision of a text or a sentence. Second, it characterizes the unit as an action, a word or word group used by the speaker to do something. This resonates with the concept of “action” in Conversation Analysis, which describes what speakers do with their utterances. [58]

§23. Establishing where divisions lie between discourse acts is inevitably tentative, as may be concluded from the outline of the literature on contemporary languages above. In particular Roulet and the Functional Discourse Grammar researchers have discussed the problem of establishing discourse act boundaries in sentences like:

The only apparent difference between these two examples is that in (1) the concessive is expressed through an adverbial phrase, whereas in (2) it takes the form of a subclause. The Geneva School regards both of these discourses as consisting of two acts, whereas Hannay and Kroon agree with them only if the prepositional phrase is realized as a separate intonation or punctuation unit. [59] The deciding factor, then, is performance: whatever the linguistic form of the text, the speaker decides what to present as separate intonation units, and thus as separate discourse acts. [60] In the following analysis of discourse acts in Homer and Pindar, any marking of discourse division necessarily represents a conjecture about the work’s realization in performance: we cannot fully reconstruct how the text was intended to be realized, and we can never establish how it was actually realized. Any division we propose, then, can only be into potential discourse acts, parallel to Lauer’s potentielle Kola. [61] To this we would add one caveat, posited by Bright: factum valet “an action otherwise prohibited by rule is to be treated as correct if it happens nevertheless.” [62] Bright’s general observation may be extended to the idea of a prescriptive definition: whatever definition of discourse act we establish, actual discourse will always prove it inadequate—there is simply no limit to linguistic creativity.

§24. Contemporary scholarship on non-syntactic units in Homeric discourse starts with Egbert Bakker. In his extensive work on Homer from the end of the eighties onward, Bakker has engaged with Chafe’s ideas about intonation units. [63] Transcripts of descriptions of Chafe’s film, The Pear Film, by individuals who just watched it show language use that is remarkably similar to that in Homer. Bakker urges us to consider this spoken narrative style as the blueprint for epic. Let us consider a passage discussed by Bakker, with his division into chunks, or intonation units:

One of the contentions in Bakker’s Poetry of Speech is that what Chafe and others call intonation units have their stylized counterpart in metrical kôla in Homer, generally about half a line long. Like intonation units, the metrical kôla that make up the hexameter are a few words long, probably reflecting “the amount of information that is active at one time in a speaker’s consciousness.” [64] Working from this view, Bakker divides each line in the examples above into two “chunks,” which he assumes to have been “a prosodic, intonational reality.” [65] Each line above represents one intonation unit, and the slash (/) marks verse end. In positing a boundary Bakker takes two factors into account: first the meter, second the content of phrases.

§25. The Homeric hexameter is a metrical form of remarkable consistency. Since it is a particularly long verse form, it generally consists of two or more metrical cola. [66] In ninety-nine percent of verses there is a caesura (called the B caesura) after the first heavy syllable of the third foot (masculine) or between its two light syllables (feminine). [67] Here is an example of each from (t4):

Beyond this mid-verse metrical break, there are two further frequently occurring breaks: A caesura (ninety percent of verses), in the first foot or after the first syllable of the second foot, and C caesura (eighty-six to eighty-seven percent), either between the fourth and fifth foot (called “bucolic diaeresis”) or after the first syllable of the fourth foot. The common metrical breaks are also points where syntactical and sense breaks typically occur, which brings us back to Bakker’s intonation units, or chunks, and our discourse acts. Bakker’s analyses demonstrate that the mid-verse caesura in particular often serves as a place for a sense boundary. On the other hand, this need not be the case, and even the strong metrical break of verse end does not always coincide with the end of a discourse act. [68] Edwards, who anticipated some of Bakker’s points, bases his analysis of the Homeric verse into four units on Hermann Fraenkel’s observations. [69] Edwards argues for a strict correlation between metrical boundaries and sense units, clear linguistic evidence to the contrary notwithstanding, leading to divisions such as | ὃ δὲ τὸν μέν | ἔασε |. [70]

§26. Homeric meter is an important guide in suggesting discourse act boundaries, along with syntax, sense, and boundary markers. Building on Bakker’s approach I consider all these factors in establishing discourse act boundaries in Homer. There are often several possibilities of dividing the verse, depending on which criteria one gives priority to. [71] Consider once more the following line from (t5), first according to Bakker’s division and then ours, with a vertical bar to mark a discourse act boundary: [72]

Bakker keeps οἱ δ᾽ ὅτε δή together, despite the position of δή, yet divides after δή, even though it means dividing the temporal conjunction ὅτε from its clause. [73] Bakker’s reason for dividing after δή in (t7) is that the particle occurs right before a common position of the A caesura, the trithemimeris. [74] However, οἱ δ᾽ is an important separate cognitive act, and therefore I propose to divide directly after. [75] This boundary coincides with another variant of the A caesura, and similar breaks occurs elsewhere in Homer; see also (t8). [76] In (t7) the feminine B caesura (marked by ¦) does not coincide with a discourse act boundary: a metrical break is an attractive place for an act boundary, but it does not entail an act boundary. In the excerpts from Homer given here and in other chapters, discourse act boundaries will be seen to regularly coincide with one of the three common metrical breaks in the hexameter, or with verse end.

§27. A longer excerpt better illustrates the many possible positions of discourse act boundaries, and the many forms acts take. At the end of the Catalogue of Ships there is a description of Achilles and his men loitering near the ships while the other Greeks are advancing on Troy. Taking into account meter and all the other factors outlined above, I divide the fifteen lines into potential discourse acts marked by vertical bars; boundary markers are underlined.

This is a very turbulent passage from the Iliad: [78] out of fifteen verses, only lines 774 and 781 can be divided roughly into two half lines. [79] Most of the discourse act boundaries are based on the words in bold that tend toward peninitial or initial position in the act. Furthermore, I take adverbial phrases as separate, as well as participial phrases. [80] In the first part of the passage (771-779, the description of Achilles’ camp) these participial and adverbial phrases separate the subject from its verb (an adverbial phrase intervenes in 771 and 773, a participial phrase in 776 and 778-779) with only one exception (ἅρματα 777). [81] The verbs, moreover, are isolated to such an extent that they appear only in the following line. Fifteen out of thirty acts are introduced by a boundary marker, thirteen of which are particles. In line 779 οὐδέ begins an act, illustrating the strong tendency for negatives to be act-initial. [82] It becomes clear from this passage why Bakker would characterize δέ as a “boundary marker,” as it accompanies many small and large steps in the narrative. [83]

§28. The description of Achilles’ camp, placed just after the catalogue of ships and just before the return to the battlefield, reveals the full extent of its mastery only when read as a sequence of small steps. A discourse analysis of any stretch of text must take into account semantics, syntax, pragmatics, meter, and cognition—discourse analysis is a holistic approach to the text. [84] The narrative presents us with a “wide” shot of the camp with Achilles and the Myrmidons. It brings us first to the ships, where Achilles indulges in his wrath; then shows us his people, by the sea, keeping themselves busy; then the horses, which stand idle by their chariots, which in turn are parked in the tents of their owners. My reading reveals the clustering of subjects + δέ throughout the passage. The arrangement is not uncommon in Homer, but the symmetry here is striking. Moreover, the (pro)noun + δέ combinations are consistently postponed until the middle of the verse, whereas typically δέ follows a (pro)noun immediately at the head of the verse. [85] Finally, in three places (lines 773, 775, and 778) we find the subject plus particle isolated as a separate discourse act. This is complemented by the isolation of the finite verbs, also consistently produced as separate acts in this passage. In lines 772, 777, and 779 the verb stands almost alone, and up to 779 even the ones that are less independent are separated from their subjects by at least a participle. The position of the verbs, either at line beginning (772 and 777) or at line end (779), suggests that they were prosodically set apart from their context by pauses or other discontinuities. [86] Their isolation will have lent emphasis to the words, putting the spotlight on their sense—which is one of inactivity: κεῖτο, ἕστασαν, οὐδὲ μάχοντο. [87]

§29. The narrative and visual path from one subject to the next, each highlighted in turn, and the overall sense of inactivity bring us to the reversal in 780, where the audience is presented with the act οἳ δ’ ἄρ’ ἴσαν. The beginning of line 780 is a heavy transition, which supports the choice by most editors to indent before this line – it is natural to print it as a new paragraph. The pronoun οἵ here is completely ambiguous, since it comes after a section containing a large selection of referents, but in fact it refers to none of the subjects in the description of Achilles’ camp. Rather, it marks the return to the battlefield before Troy, and functions as the cap of the entire Catalogue. The surprise effected by the narrative turn would not have had as much impact if it were not for the careful build-up of the preceding ten lines. After dwelling on the picture of soldiers lounging and killing time, the performer brings our attention back to the advancing army with three constituents: the pronoun οἵ, the particle combination δ᾽ἄρα, and the verb ἴσαν. The use of a particle combination to follow the pronoun instead of just a simple δέ provides some warning that we are being presented with a different kind of transition from the preceding ones, [88] but undoubtedly the verb here is the crucial factor. For the audience to be abruptly confronted with “they went”, after encountering a series of static verbs, requires quite a bit of cognitive processing on their part, as they search for the group of appropriate referents.

§30. To seal the return, Homer introduces a simile, comparing the army to an advancing, all-consuming forest fire (780). This image effectively redirects the audience’s attention to the battlefield, for it recalls another simile at the beginning of the Catalogue (lines 2.455-458) in which the shine of the Greeks’ armor is likened to the blaze of a forest fire on a mountain. [89] The simile serves to accommodate the cognitive effort required to travel back to the army advancing, that was last referred to in 2.455-473. The similes before and after the Catalogue are further connected in the sense that the one just before it describes a glinting army seen far off while the simile at line 780 gives an image of a Greek horde about to engulf the Trojans. The echo of the imagery must have helped the audience negotiate a considerable narrative and spatial discontinuity. [90]

§31. The scene at Achilles’ camp is a linguistically turbulent passage, produced in a stream of short acts that stand in no fixed relation to syntax or meter. The acts direct attention to different aspects of the scene, and allow the audience to take everything in step by step. δέ is instrumental in accomplishing this compartmentalizing effect, and in fact is essentially omnipresent in Homer. Yet the particle, despite its prevalence in the text, has not been the object of as much close analysis as might have been expected. The particle’s fundamental function in Homer has most recently been described by Bakker:

Before Bakker, δέ had been viewed as a primarily “connective” particle, sometimes with adversative force. [91] However, especially in Homer, δέ can occur at the boundary between two clauses, two phrases, between main clause and subclause, or vice versa, or between a vocative and what follows. In other words, δέ is not in the first place a syntactic marker. [92] Bakker, however, argues that it reflects the production of discourse in small steps. As such, δέ has little to do with content, and everything with form: the term connective thus is useful only if it concerns discourse rather than content.

§32. One aspect of the previous scholarship on δέ that Bakker preserves is the idea of δέ as in some way weak or “bleached. ” From the earliest studies onward δέ has been regarded as etymologically connected to δή; Bakker, like others before him, views δέ as a bleached version of δή. [93] This sense of δέ as a bleached form of δή, however, focuses overmuch on the semantic load of the particle.

§33. Indeed, Homeric δέ often goes (and should go) untranslated in English, but that is because its function does not have its equivalent in English in a lexical item, but in punctuation or prosody. Rather than anything bleached, as a boundary marker δέ is in fact the strongest of all second-position words: the only one that is practically never moved from its position in Homer and to a lesser extent in lyric. [94] The particle’s strength on a purely functional level has seldom been discussed, yet it is crucial to a number of other questions involving δέ. For instance, the fact that δή (or at least one of its aspects) seems to have been a mobile in early Greek, of which some trace remains in Homer, [95] speaks against the close relation envisioned between the two words by Bakker and most earlier scholars. Indeed δή’s mobility all but eliminates the possibility that in Homer δέ functioned as a bleached version of δή. [96]

§34. To better understand the differences between δέ and δή, consider first the instances where both particles occur, as in the following haunting scene from the Odyssey:

Here and in all thirty-one other instances of the combination δέ precedes δή. [97] The same thing holds for the combination of δέ with ἄν/κε(ν); remember that Fraenkel actually chose ἄν as a case study for his research into kôlon boundaries. The reason must quite simply be that δέ, at least in Homer, cannot leave its peninitial position for another second-position word. This tenacity suggests that its function is tightly connected to its position, which supports Bakker’s description of δέ as the quintessential boundary marker.

§35. The boundary that δέ marks can be that between a main and a subclause, between two parallel phrases, or after a vocative in direct speech, but the particle can also occur at stronger discontinuities in the discourse, as in the following example:

After describing how the sun comes up and the ship carrying the sleeping Odysseus reaches Ithaca, the performer devotes some attention to the place where Odysseus will be left by the Phaeacians. [99] This new section is set apart by the present tense and starts with δέ but is not otherwise marked, yet it represents a significant redirection from action to image and from movement to stasis. The two parts of discourse are divided by nothing more than δέ. [100]

§36. This passage is one example of how the particle can occur in any kind of transition, because of its ability to occupy the second position of any discourse act. Δέ in Homer is the simplest metalanguage: its syntactic and semantic flexibility shows that its function does not primarily lie on those levels, but is concentrated on marking a discourse boundary. [101] In Homer, δέ is markedly more frequent in narration than in direct speech. The imbalance suggests that there is something in the steps of narrative discourse that allows or requires more neutral boundaries than in direct speech. [102] Over time δέ develops from being an omnipresent boundary marker to a particle used to start new períodoi, and eventually to a marker of adversativity. This development is accompanied by a steady decline in the particle’s frequency in later texts, as well as a looser adherence to its peninitial position. [103]

§37. Pindaric language is widely regarded as difficult. This view is implicitly reflected in the punctuation practices found in modern editions, which give the impression that Pindar sometimes wrote monstrously long sentences. [104] This syntactical approach does not do justice to how the songs were received by a listening audience. Like Homeric epic, Pindaric song proceeds in small acts. The most important difference between the two corpora is the music that was an integral part of Pindar’s songs.

§38. The melodies of Pindaric song are all lost to us, and we must keep in mind that the melody may have mitigated the apparent complexity of Pindaric discourse. The melodic dimension may have clarified constructions, created breaks and links, and overall made linguistic construction quite secondary. Moreover, music changes the way the lyric language is received and understood on a level above the discourse act. As opposed to the repetitive rhythm of Homer, the division of Pindaric song into strophe, antistrophe, and epode creates an intermittent recurrence of rhythmic (and melodic) units. Songs exploit the audience’s “melodic memory” to create resonance beyond the linear production of verse after verse, in a phenomenon called “tautometric responsion.” [105] In the example of Pythian 6 (t12 and t13 below), for instance, the phrases ἄγεις ἐφημοσύναν (20, about the victor’s son) and νόημα τοῦτο φέρων (29, about Antilochus), occupy the same metrical position in strophe and antistrophe. In this case, the resonance reinforces the comparison that Pindar proposes between the victor’s son and Antilochus.

§39. The one aspect of Pindaric song that is still accessible to us is its meter. Therefore, in dividing Pindar’s songs into discourse acts, metrical considerations play an important role. Since the verses are decidedly shorter than the Homeric hexameter, they are not often broken up into separate metrical kôla. [106] Although verse end does not coincide with act boundary quite as regularly as in Homer, this is still a strong tendency. Moreover, the strong metrical boundary after strophe, antistrophe, and epode always coincides with act boundary.

§40. Consider the third strophe of Pythian 6. After two strophai introducing the event (the chariot race in the Pythian games), the winner (Xenocrates), and his clan, Pindar focuses on Xenocrates’ son, who is addressed in line 15 with the vocative Θρασύβουλε, and again in line 19 with σύ τοι: [107]

In the passage above, I mark discourse act boundaries on the basis of syntax and sense (e.g. the apposition βαρύοπα…πρύτανιν), but also considering postpositives (νιν, ποτε, μέν, and δέ) and hyperbaton constructions (ὀρθὰν…ἐφημοσύναν). After σύ τοι [108] follows a participial phrase with the enclitic νιν in second position. [109] Τhe first two lines of this third strophe thus proceed in three acts, the first to direct attention to the focus of the upcoming two strophai (“You then”), the second to provide a link to the preceding (“keeping him [sc. Thrasyboulos’ father] at your right hand”), [110] and the third to look ahead to the upcoming discourse (“straight you keep the command”).

§41. The next act starts with a neuter pronoun (τά, “those things”) that retrieves the referent of ἐφημοσύναν “command.” At the same time, the pronoun functions as a transition into a little mythical reference (23-26). [111] The clause that provides an orientation for this narrative neatly demonstrates the difference between Pindar and Homer. Line 21 ends with the dative adjective μεγαλοσθενεῖ, which can in no natural way be construed with what preceded; it thus projects [112] an indirect object (dative) of an upcoming new action that will fit the epithet. The first words that follow are Φιλύρας υἱόν, a circuitous way of naming Cheiron, who after φαντί will naturally be taken as the agent in the emerging construction. At this point the audience may already have the information they need: Pindar is elaborating on a saying that Cheiron once taught someone in the mountains (ἐν οὔρεσι), someone as yet unnamed to whom the (rare) epithet μεγαλοσθενής can suitably apply. The combination of cues will have activated the figure of Achilles in the audience’s minds. The suspicion gains strength in line 22 when the audience encounters a second epithet, ὀρφανιζομένῳ, and finally is confirmed with the patronymic in line 23 (Πηλεΐδᾳ). Because of the interweaving construction and it is hard to establish where this relatively long act (21-23) may have been subdivided. The adjective in the dative (μεγαλοσθενεῖ) creates the expectation of a noun, and the thought is not yet complete at the end of line 22. Despite its length, however, the audience would still have been able to semantically process the construction on-line.

§42. In Homer, this act would probably have been broken in two, presented with the different components clustered together. Note, however, that such syntactical and discursive flourishes in Pindar are designed precisely to be flourishes; they function as verbal (and perhaps musical) tours-de-force that stand out against the general piecemeal progression of the discourse in his songs. [113] The saying itself is introduced at μάλιστα μέν and illustrates how so-called μέν – δέ constructions are often not symmetrical in early Greek poetry (more on this in §56-§59). Finally, line 24 (βαρύοπα στεροπᾶν κεραυνῶν τε πρύτανιν) contains another idiosyncrasy of Pindar’s. Lauer first observed that Pindar has a tendency to demarcate his kôla using hyperbata, which Lauer called Kolonbildend. [114] The suggestion is that a lone adjective (βαρύοπα) at the beginning of an act (or Kolon in Lauer’s terms) creates the expectation of a noun, and the eventual occurrence of the noun (πρύτανιν) then marks the end of the discourse act. [115] Markovic remarks that throughout early Greek literature, hyperbaton is a tool frequently used especially for “signaling or reinforcing the end of syntactical and semantic units” [my emphasis], and regards it as a feature “of oral tradition.” [116]

§43. The wisdom that Pindar offers in Pythian 6.23-25 is that one should honor Zeus most out of all the gods, and that one should honor one’s parents. The second part of the gnṓmē (“honor your parents”) introduces an important theme in the ode which provides the starting point for the following strophe: [117]

The theme established by the end of the third stanza is explored in the fourth, which here takes the form of a little narrative. The transition to the past (ἔγεντο καὶ πρότερον) and the introduction of the story are clear. The pronoun in νόημα τοῦτο secures the connection to the preceding discourse, and the next act, starting with the personal pronoun ὅς referring to Antilochus, begins a little abstract of the entire story. [118] Μέμνονα in apposition at the head of line 32 is semantically superfluous, but the name here serves to close the ring begun at the naming of Antilochus, a ring from victim to killer. [119] Furthermore, placing the name here enables Pindar to juxtapose Memnon with Nestor: a reflection in the language of the situation on the battlefield. The narrative proper follows, beginning with the particle γάρ, [120] and proceeds in six subsequent acts, no fewer than four of which start with δέ. [121] Use of δέ in Pindar is noticeably different from Homer’s usage: the particle occurs much less frequently, and is more often used to mark more significant boundaries in the discourse. [122] It is in Pindar’s narrative sections that we find a distribution of δέ closer to that found in Homer, but even in Pindaric narrative the particle is not used quite as frequently as in Homer. As for the particle’s position, in line 37 δέ is postponed until after the word group αὐτοῦ μένων, a license that does not occur in Homer. [123]

§44. The function of μέν in line 39 is not immediately obvious. It occurs in the line that appears to function as the resolution of the narrative (see II.3 §14), the end of the story. The sense of resolution is confirmed when we consider how the strophe continues:

Line 39, then, functions as the end of the narrative proper, while 40-42 are a reflection on the story’s outcome. If we take the classical approach to μέν as a “preparatory” particle, we find no satisfactory explanation for its occurrence here, in the act πρίατο μὲν θανάτοιο κομιδὰν πατρός. On a semantic level there is no parallellism or contrast with the following ἐδόκησέν τε τῶν πάλαι γενεᾷ. Syntactically, the verbs πρίατο and ἐδόκησεν do resonate with each other, as both have Antilochus as their subject. Scholars have consequently argued that μέν in such occurrences is “answered” by τε instead of δέ in Pindar. [124] But the two acts that would here be connected through μέν – τε diverge in their discursive nature. The first (39) rounds off the narrative, while the second (40) functions as a postscript.

§45. How, then, should we understand this μέν? In the following section I argue that μέν must be taken in instances like this as a metalinguistic cue: its host act rounds off one move, while μέν projects that a new move is coming up. The particle μέν possesses this function regardless of what introduces the upcoming act, be it δέ, τε, καί or no particle at all.

§46. In the passages from Pythian 6 above (t12-t14), there are two instances of the particle μέν, both of which may offer insight into the workings of the particle. In Attic Greek prose, μέν mostly has a narrow function, that is, it marks the first part in a μέν – δέ construction. In such contexts it may be translated “on the one hand,” or remain untranslated. Schömann first argued that μέν developed from μήν, and that its original force was affirmative. [125] This view is compatible with the argument put forward by Spitzner, and soon promoted to communis opinio, that μέν in Homer is simply a dialectal variant (Ionic) of μήν (Attic) and μάν (Doric). [126] Spitzner’s observation led to the argument that μέν tout court is descended from μήν and its variants, both on a formal and on a functional level. Leumann further hypothesizes that the μέν – δέ construction preserves the Ionic variant of μάν because the construction found its way into Greek discourse through the scientific works from Ionia. At the same time, all dialects kept their own variant of μάν for all other instances, that is, where it affirms its clause. [127] Among these scholars, some also regard the fact that μέν is a shorter form of μήν as a sign of a bleaching or weakening of the function performed by its ancestor. [128]

§47. Thus the particle’s origin in μήν and its more common Attic use in a μέν – δέ construction have been linked. [129] One thing that has been lost in this development of scholarship is the forward-looking force of μέν. If μέν is from affirmative μήν, it implies contrast with the following by strongly affirming the present clause, phrase, or constituent. [130] I follow the majority of scholars in assuming a formal (i.e. diachronic) relation between μέν on the one hand and μήν and μάν on the other. However, unlike the majority I do not believe that μέν is always functionally connected to μήν and μάν by the time of Homer (i.e. synchronically), since to describe the use of μέν in Homer as “affirmative” is insufficient.

§48. Before the link between μήν and μέν became central to the analysis of μέν, scholars paid more attention to how μέν was actually used by Greek authors. This led Hoogeveen to say the following: “when a speaker uses μέν, he warns the reader or hearer that he should not agree to this first part, but should wait for the ἀπόδοσις, which resolves the utterance.” [131] In other words, with μέν the speaker promises to add something after the current clause, phrase, or constituent. [132] Rather than a development from the particle’s putative original affirmative force, this “anticipatory” sense is primary in μέν. The question now remains how to read μέν in Homer, since it can either be the Ionic form of Attic μήν, that is with the function of μήν, or the μέν shared by all dialects. In practice, this question would have to be answered per instance, but we may follow the generalization offered by Hartung that μήν is only shortened to μέν when it is used with other particles. [133] In my analysis of μέν in Homer and Pindar, I have found no instances of μέν on its own (i.e. except in particle combinations and after οὐ) where it cannot be explained as forward-pointing.

§49. The next step is to come to a better understanding of this “anticipatory” function of μέν. The words of Hoogeveen may be linked to the idea of metalanguage: μέν is first and foremost an instruction to the hearer. Discourse acts are not strung together at random, but rather stand in some kind of logical relationship to each other. As I discuss in more detail in II.3, different kinds of multi-act discourse units (“moves”) cohere for different reasons. The acts within a move may share a common topic, or in narrative they may share a place of action and set of characters (in what we call “contextual frames”). As a result of the coherence that acts possess in relation to each other, acts create expectations in the discourse, in a process called projection. [134] This phenomenon may help explain the workings of μέν, the particle that appears to have specialized in cueing projection. [135]

§50. In his work on projection, Auer describes the pragmatic process of how one word, utterance, or conversational action projects another. In conversation, speakers are remarkably capable of predicting what is coming next, often even beyond the interlocutor’s upcoming turn. This predictability is a result of experience, but more basically it depends on relevance. Any conversational action restricts the range of possible consequent actions:

An act can project several others, and in practice these possibilities form a limited set. Beyond conversation, projection can become inherent in syntactical constructions, such as “the thing is,…” [137] Therefore, Auer speaks of pragmatic projection: “one act or action projects another” and syntactic projection: “one syntactical constituent projects another.” Projection can work on the microlevel (syntax: a preposition projects a noun to follow) [138] and on the macrolevel (genre: the first part of a priamel projects at least one more). [139] Moreover, projection can be more specific (preposition projects noun) or less specific (“and” after a syntactic closure “leave[s] all options open apart from not continuing”). [140]

§51. The concept of projection can offer better understanding of μέν in its entire range of pragmatic functions in ancient Greek, including the Homeric corpus and Pindar’s Victory Odes. In the present section I discuss the functions of μέν in terms of projection. First, the focus is on μέν creating pragmatic projection across discourse transitions. Second comes a discussion of the ultimate pragmatic projection: μέν at the beginning of discourse units, performative units, and even whole works. Finally, I consider the μέν – δέ construction, which works differently when μέν has large scope than when it has small scope.

§52. One aspect of μέν that has suffered in particular from the scholarly focus on μήν is its use in discourse transitions. In his list of functions of μέν – δέ constructions, the first function Hoogeveen lists is μεταβατικός (“transitional”). [141] In recent scholarship this has been retrieved by Bakker: rather than signaling a juxtaposition in states of affairs, μέν – δέ signals a juxtaposition of two discourse units. [142] As Devarius noted (see fn. 141), μέν can have this function on its own as well, without a following δέ. Consider the beginning of book ten of the Iliad:

After ending the scene in book nine with all the leaders of the Greeks returning to their tents for the night, at the beginning of book ten the focus shifts to Agamemnon. In contexts like this, Apollonius Dyscolus describes μέν as marking a περιγραφή, a conclusion or summary. [143] Devarius took up this same term, calling μέν περιφραγικός, whereas Hoogeveen speaks of μεταβατικός (“transitional”); I believe Hoogeveen’s description is more helpful. In discourse terms, μέν does not function so much to mark a conclusion, since that is contained within its host act, but to point ahead. In instances like (t16) μέν pragmatically projects a new move, even as its host act rounds off the preceding one. As (t16) shows, the continuation of the discourse does not have to start with δέ: in line 3 ἀλλά introduces the new focus of discourse.

§53. Compare one further example, from the Catalogue of Ships:

The Catalogue consists of a list of entries that describe a people and their leaders, one by one. [144] The entries are linguistically quite consistent, giving first the people and the leader in the first line, then often a little narrative, then naming the leader again, and giving the number of ships that he brings. In the renaming of the leader, which starts the final element of the entry, we often find μέν, as in (t17). Just like in (t16) μέν in (t17) prepares the hearer for the transition to a new focus, in this case a new entry. Here μέν is followed by δέ (line 653), but there is no syntactical symmetry between the two constructions. Rather, the μέν act looks like a conclusion, with the anaphoric pronoun retrieving the people named in the preceding discourse, whereas the δέ act looks like a new beginning, since it introduces a new name in first position.

§54. The function of μέν, however, is not itself “summarizing,” nor does it only juxtapose two list items or states of affairs. [145] The particle provides a metalinguistic cue for the upcoming entry even while its host act rounds off the present one: μέν projects an upcoming move. [146] In (t17) the information about the ships intervenes after the projection (652), but, as Auer shows, the “trajectory” of a projection may be quite long. [147] The audience will not have regarded the projection as resolved until the start of a new entry; they are able to mentally carry the projection over the intervening discourse. [148]

§55. In (t16) and (t17) μέν is not attached to one particular word in the act. That is to say, I do not take μέν as having small scope over the word preceding it. Its scope extends over the entire act, and its influence in fact reaches beyond. Compare this similar example from Pindar, where δέ marks the resolution of the pragmatic projection triggered by μέν:

In Nemean 10 Pindar tells the story of Castor and Pollux, who have here just been introduced as the stewards of the agones in Sparta. In line 54 this is expanded in an additional participial phrase, introduced by μέν. Now, this participial phrase rounds off the topic of Castor and Pollux as stewards of the Games. The final act of line 54 introduces a new grammatical subject and contains a gnomic thought. Then the epode ends, and at the beginning of the new strophe Castor and Pollux are retrieved as subject (μεταμειβόμενοι), but as part of a new topic: their mortal/immortal status. In this case, then, μέν serves to project a new move, yet it also projects the current discourse topic (Castor and Pollux) across a gnṓmē and the strong performative discontinuity between the epode and a new strophe. [149]

§56. In early Greek poetry, the relation between a μέν act and the following discourse is decidedly loose: μέν projects pragmatically, so it does not determine how discourse continues, only that it does. This touches upon another function of the particle often addressed in earlier scholarship: its so-called “inceptive” function which in turn is inherently connected to the “asseverative” function of the particle. [150] Bakker links the transitional function of μέν to its inceptive use as follows: “[μέν] is often used to mark a statement that clears the ground, establishing a framework for discourse to come, and as such it tends to be used at the beginning of a speech.” [151] Consider the following statement by Denniston about the frequent occurrence of μέν at the beginning of speeches:

However difficult we may find it to resist this impression, Denniston here presents as a special use of μέν something that represents a central and original function of the particle. Although any initial act may be said to naturally project by virtue of being initial (projection on a macrolevel), this use of μέν should be regarded as pragmatic projection. In practice, instances of μέν at the beginning of a new discourse rarely represent a balanced combination of a μέν and δέ act. Consider the famous beginning of Olympian 1:

The priamel to Olympian 1 starts with a juxtaposition of water and gold, or so it appears. The particles μέν and δέ in the second position of the first two clauses suggests that we are faced with a classic parallel construction, but the reality is more complex. Gildersleeve’s translation nicely brings out the asymmetry in the construction: “…but there is another—gold—a blazing fire like it loometh—a night fire far above all proud wealth.” [152] Rather than just an article, ὁ in Pindar is “still largely deictic”; [153] because of that and for metrical reasons Gildersleeve translates ὁ δέ as a separate step (“but there is another”), a separate act in our terms. The first act with μέν grounds the song and performance, upon which the next act starting with δέ builds. Then follows another foil before the priamel’s climax is reached in line 7. μέν thus marks a discursive beginning, not one half of a syntactically or propositionally symmetrical construction. The fact that μέν so often occurs at the beginning of strophes, antistrophes, epodes, and songs is an extension of this same function. [154]

§57. Until this point, I have focussed on the function of μέν on its own, with large scope, projecting the progression of discourse. We must also consider the construction μέν – δέ, however, which occurs from the earliest Greek literature onward. Since the most basic cue for discursive progress in Homer is δέ, [155] it comes as no surprise that at some stage the frequent act projection of μέν and act introduction of δέ led to a grammaticalization of the μέν – δέ construction. [156] In this construction, μέν does not project “a continuation of discourse” but it grammatically projects “a δέ act.” [157] The following two examples from Pindar illustrate the difference between μέν on its own marking pragmatic projection, and μέν in a μέν – δέ construction:

In a typical Pindaric transition, the Isthmian ode proceeds from the manifold possible topics to the one the composer wants to focus on. [158] μέν creates a ground, yet points ahead, [159] and is followed by καὶ νῦν, which pins down what Pindar will in fact focus on. [160] Some view the relation between the μέν act and what follows as that of general to particular, or of secondary to primary. [161] However, in practice this is often clearly not the case, and moreover when it does occur, it may quite simply reflect the presentation of discourse in order of increasing relevance or importance, more than any particular function of the μέν – δέ construction.

§58. With an apparently similar turn of phrase, Pindar rounds off one of his Pythian odes:

Here full symmetry is achieved through the verbal resonance of πολλὰ μέν – πολλὰ δέ, and this parallellism may have been marked prosodically, as happens with juxtapositions in spoken English. [162] Rather than read (t21) as an instance where the δέ component of the juxtaposition is “omitted,” one might say conversely that Pindar creates a beautiful symmetry in (t22) by adding a parallel δέ act. In this construction, one can also see how both μέν and δέ have small scope, to emphasize the symmetry. The particle μέν lends itself to this kind of construction, but the occurrence of μέν does not – or at least not in archaic and early classical Greek – entail the construction in every instance. [163]

§59. In its use with small scope, which may have been rendered prosodically, μέν has another pragmatic function. As in the μέν – δέ construction, small scope μέν can project something very specific, dependent on the word preceding the particle; one can think for example of ἄλλοτε μέν (which projects ἄλλοτε δέ). [164] A special use of this small-scope μέν is with personal pronouns in the nominative, as in the following example from the Iliad. Personified Sleep reminds Hera of a time that Hera had commanded Sleep to distract Zeus:

Examples like this are rare in Homer, but in this instance there appears to be a close relation between ἐγὼ μέν and σὺ δέ. That is to say, uttering ἐγὼ μέν, when the particle has small scope, limits the possibilities of the following to σὺ δέ, ὑμεῖς δέ, ὁ δέ, or similar. [165] This projection can even reach across utterances, see this example from Aeschylus (— marks a new speaker):

The chorus hear Agamemnon’s cries as he is attacked, and consider what to do. The deliberations proceed chorus member by chorus member, each sharing his thoughts. The first chorus member in the excerpt starts his utterance with ἐγὼ μέν, which implicitly juxtaposes his opinion to the others to follow. [166] Two lines later, a second speaker starts with ἐμοὶ δέ, reacting to the first speaker and continuing the discussion of possible courses of action that runs on until line 1371. That is to say, μέν serves to project an act or in this case utterance that is in a significant way parallel to the present act: it has small scope, which limits the possibilities of what follows. The projection is fulfilled with δέ in the following utterance. [167]

§60. If μέν can thus project a δέ act across speakers, it follows that μέν can be used to exploit this function even when the projection is never fulfilled. [168] In (t25) μέν projects a specific act, which remains unspoken:

Athena tries to convince Odysseus to gloat over the downfall of Ajax, but Odysseus is hesitant, and uses ἐμοὶ μέν to disalign himself with Athena: “it suffices for me if he stays in the house. [But you do not seem to agree.]” Quite specifically, in instances like (t25) μέν locates the referent over which it has scope (ἐμοί) with respect to other available (textually explicit or implicit) referents. [169] This use of μέν is especially frequent in drama. [170]

§61. Small-scope μέν following a pronoun or name can thus create a contrast by projecting a viewpoint that differs at least in part from the one held by the first referent. As we explore in a later chapter (II.5), γε after names or pronouns can have the same pragmatic effect of creating contrast, but for a different reason. Whereas μέν creates contrast through projection of another referent (who is often later expressed), γε creates it through emphasis on the current referent. [171]

§62. Projection can account for a range of pragmatic functions that μέν has in Homer, Pindar, and beyond. First, the particle serves as metalanguage to guide the hearer through the discourse, often foreshadowing transitions to new moves within the discourse. In this function, its scope extends over its entire host act, and there is no particular relation between μέν and the word that precedes it. Second, μέν can have scope over the preceding word, with a range of possible effects. In Homer and Pindar not every μέν entails a δέ: when μέν has large scope, the projection can be fulfilled with any particle that can continue the discourse, or no particle at all. If μέν has small scope, it most typically forms part of a μέν – δέ construction, which in later literature covers the majority of μέν instances.

§63. Projection is omnipresent in ancient Greek discourse, reaching far beyond the specialized lexical item μέν. It manifests itself in syntactic constructions, semantic interlinking, and discourse articulation. In contemporary texts, an indentation creates certain expectations in the reader about the relation between the upcoming discourse and the preceding discourse (some kind of discontinuity) and about the nature of the upcoming discourse (in some way coherent). In spoken discourse other tools are available to obtain similar effects, first and foremost utterance- and sentence-initial discourse markers, which are generally intonationally independent from what follows (such as “First, I would like to welcome you all” or “Sadly, Sarah could not make it”). [172] This section focuses on short acts in epic and lyric that share important characteristics with such discourse markers, especially their syntactic and presumably prosodic independence, which creates a projecting effect over the following discourse. The acts under examination are those consisting of nothing more than a (pro)noun and a particle, which reflect ad hoc cognitive processing while projecting the referent’s relevance for the upcoming discourse. I propose to call such phenomena priming acts. [173] In the section on Homeric discourse the focus is on the narrative function of these priming acts, while the section on Pindaric song considers the performative effect of priming acts involving a second-person pronoun.

§64. To illustrate the nature of priming acts, consider again Bakker’s reading of Iliad 7.313-314 (see t5), repeated here. Each line represents, in Bakker’s terms, one chunk or intonation unit, forward slash (/) marks hexameter end:

The boundary marker δή is, without further explanation, put here in fourth position of an intonation unit. [174] In performance, it is not inconceivable that οἱ δ᾽ ὅτε δὴ was performed as part of one intonation unit. [175] Cognitively, however, the first intonation unit in (t26) contains two acts, or rather one act and the beginning of a second, which concludes with the second chunk in Bakker’s division. Consider again my alternative presentation of line 313:

When we read these two lines as four acts – with the nominative ἄναξ ἀνδρῶν Ἀγαμέμνων taken as a separate act expanding the subject contained within ἱέρευσεν – οἱ δ᾽ turns out to be a subject without a verb, a phenomenon generally called a pendant nominative. [176] Most translations, like Murray’s above, solve this problem by taking the pronoun as the subject of the following act (a subclause). As a result, the independence of the pronoun is lost in the translation, obscuring the emergent discourse articulation. [177] Consider an alternative translation:

Syntactically, the isolation of οἱ δ᾽ is troubling, which is probably why instances such as this one are generally read as Murray does, without further discussion.

§65. Fraenkel had already noted the existence of such short initial kôla containing particles like μέν, δέ, τε, and others in second position. [178] He notes that when a constituent is brought to the front of the sentence, it receives strong emphasis:

As Fraenkel saw, the act containing the pronoun + particle is a normal and productive rather than anacolouthic construction. [180] Classicists have more recently described the phenomenon in Greek as a sign of “oral syntax, ” [181] a characterization that, like earlier views, marks this kind of construction as divergent from a supposed standard syntax.

§66. At certain points in narratives, but also in other kinds of discourse, we find such short discourse acts, comprising a (pro)noun and often a particle, set apart from their main verb by a participle, adverbial phrase, or temporal subclause, or even lacking a main verb, as in (t28). These acts make sense if we consider their possible function not at the level of the sentence, but of the larger discourse. From that perspective Iliad 7.313, for example, reveals itself as a narrative transition. In fact, Murray indents the line, which suggests that he interprets it as the start of a new scene or episode. At the beginning of this new scene, its main characters are activated and primed first of all, with the pronoun, before moving on to a new action: the priming act aids the redirection of the audience’s attention.

§67. In contemporary linguistic studies scholars have observed a similar phenomenon in spoken language, generally called left dislocation. [182] More specifically relevant is Ochs Keenan and Schieffelin’s discussion of a phenomenon that they call referent + proposition. [183] In these constructions, the subject of a proposition is presented as a separate intonation unit, followed by a proposition with the subject repeated (or changed). Consider the following typical example:

As οἱ δέ in (t27), the noun phrase “My sister” stands on it own: it does not form part of a clause finished at a later point. Chafe discusses the same kind of construction, but he calls the initial unit an “isolated referent.” In his analysis of oral narratives he comes to the conclusion that “[i]t is not unusual for an intonation unit to verbalize little or nothing more than a referent.” [184] Left-dislocated elements in contemporary discourse are typically associated with topic status: the terms found in the left periphery of the sentence often reflect what a significant part of the upcoming discourse is about. [185]

§68. Greek is a Pro-Drop language, which means that its verb forms do not require the addition of a grammatical subject in the nominative. Since a nominative is often not strictly speaking necessary, its addition can be regarded as marked. This brings us to examples like the following passage, where the referent in the priming act takes the form of a name, and the main clause has a coreferential verb:

In this example the name in the priming act (Τηλέμαχος) cannot be constructed with the verb in the immediately following act (δέδμητο), since it is not coreferential: it supports the validity of likewise regarding οἱ δέ as independent in (t27). [186] Here, Telemachus is singled out from a larger group and remains the main referent in the last lines of book one of the Odyssey. Despite the slight variations in form, in all incarnations the priming act reflects a cognitive process: the reorientation of the mind’s eye. From a discourse perspective, these priming acts typically occur at narrative transitions, redirecting the joint attention of the performer and audience to a character that has been out of focus for a time. When a referent is retrieved in a priming act, it creates an expectation of (i.e. projects) an upcoming action done by (or to, if the (pro)noun is in an oblique case) [187] the character.

§69. In a long and complex narrative, neither the audience nor the performer can keep track of every detail or character in the imagined world. Rather, a traditional storyworld is a construction based on the discourse as well as the whole body of knowledge and assumptions about that world shared by audience and performer. [188] A great deal of cognitive processing is involved in keeping track of who is where doing what at any particular point in the narrative. For this complex task the mind is well equipped, and Emmott has studied this cognitive process for readers of English literature. She speaks of “contextual frames,” a concept that regards people and places in a narrative as interconnected in networks.

§70. Emmott describes how characters are bound to certain places in the storyworld unless we are cued to change our knowledge of their location. The storyworld is only theoretically a whole world: in practice it is a string of discrete spaces containing certain characters and thus not others. [189] “Thus a narrative has regions, filled by landmarks (=reference objects) through which a path is established.” [190] Just as one word or construction will activate a whole semantic frame, the mention of one referent will make the entire contextual frame accessible. Thus a priming act can accomplish more than just re-orient the performer’s and audience’s attention on a specific character. The reference to a character activates the entire contextual frame to which the character is bound. [191] That is to say, the priming act has a potential double relevance: it is quite likely that it had a performative relevance in allowing the audience to follow the complex narrative of long epics, but it is also quite possible that the language accommodates the cognitive processing of the Homeric performer as he produces his instantiation of, for example, the story of how Telemachus reaches the decision to stand up to the suitors, in the first book of the Odyssey (t31). They might be signs of the performer navigating his memory by navigating the storyworld, moving from place to place and focusing on different characters as they become relevant. [192]

§71. In the ongoing narrative, priming acts may also serve to prime a character that has been covertly present in the ongoing contextual frame. [193] This is what happens in example (t32): after a long speech by Telemachus, at which Odysseus and the suitors are present, Odysseus is set apart from the others (πάντες) and primed as the character whom we will be following in the upcoming stretch of narrative. [194]

Here the act αὐτὰρ Ὀδυσσεύς is divided from what follows by verse end, as well as the boundary marker μέν in the following line. [195] Denniston says the following about such constructions:

This clearly holds for (t32), where αὐτὰρ Ὀδυσσεύς provides the “subject” of both the μέν and the following δέ clause. However, Denniston fails to address the question of why this construction occurs, nor does he allow for the fact that the act preceding the μέν – δέ construction may project beyond it (i.e. “apply” to more than “the whole μέν…δέ complex”). [196] The priming act αὐτὰρ Ὀδυσσεύς redirects the performer’s and audience’s focus, and projects the character’s relevance for the upcoming discourse. [197] It may not be a coincidence that this kind of short act, which occurs in a myriad of contexts, has been incorporated in the quasi-fixed constructions of the form X δ᾽ | ὅτε δή, Χ δ᾽ | ἐπεὶ οὖν (and 3x τ᾽ ἐπεὶ οὖν), and Χ δ᾽ | ὥς οὖν. The beginnings of new scenes, often marked by indentation in the editions, typically require extra attention to be placed on particular referents. In such contexts, priming acts have become part of the verse that concerns itself with framing the upcoming scene. [198]

§72. Narrative priming acts are especially suited to the performance situation of the Homeric epics, but they can be found even in Pindar’s songs. [199] Examples of the kind described above are few, however, and the units are syntactically fully integrated, as opposed to many instances in Homer. The following passage may function as an exemplar of the narrative priming act in Pindar:

In this passage in Olympian 6, where king Aipytus returns from Pytho to search for Iamos, the offspring of Euadne and Apollo, the priming act | βασιλεὺς δέ | effects a transition back to the king, who is subject and protagonist for the next five lines. The isolated | τοὶ δ᾽ | in line 52 effects another transition, this time to the king’s listeners, who had been passive bystanders until then. [200] In both cases the nominatives fit neatly into the syntactic structure, but as in the Homeric examples their function is not to be sought on a syntactic level only. Here, too, the priming acts function as cognitive pivots, accommodating the redirection of attention. [201]

§73. The Pindaric corpus features a different category of priming act, which may serve to illustrate a generic difference between epic and lyric. As songs composed for an occasion and for a specific audience, Pindar’s Victory Odes display a more direct interaction between performer and audience than the Homeric epics do. As a result, priming acts in Pindar typically do not shift perspective or focus solely within the storyworld, but shift from one world to the other, such as from the storyworld to the hic et nunc. [202] Second-person discourse is especially effective for moving out of the storyworld and into the here and now. Even in the interactive performance situation of Pindaric song, a singular or plural second-person reference has a strong effect. In the following paragraphs I show how the disruptive effect of a transition to second-person discourse can be linked to the occurrence of second-person pronouns in priming acts. [203] These priming acts can take different forms, and can be syntactically integrated to greater or lesser extent into the following discourse. Finally I discuss the frequent co-occurrence of second-person priming acts and names in the vocative.

§74. Pindaric discourse is a complex negotiation between speakers and addressees, who can be physically present (the performer(s), the audience, the victor, and his clan), vicariously present (as the composer through his song), or treated as present (ancestors, heroes, and gods). That is to say, not every “you” has to refer to someone physically present at the performance. [204] When a second-person pronoun or verb occurs at the beginning of a new act or move, moreover, the audience cannot know who is being addressed unless it is made clear through extra-linguistic means, or until the addressee’s identity is specified in what follows. [205] Therefore, when a you-reference occurs (especially outside direct speech), the audience cannot but be highly involved. [206] More than third-person forms, and even more than first-person references, second-person forms create an immediacy in the performance that lends itself well to transitions in the discourse, and especially to marking an upcoming passage as significant.

§75. Taking into account this performative effect of you-references, it is not surprising that we often find a second-person pronoun occurring in a priming act, as at the end of Nemean 3:

The last epode of the song starts with a metaphor that sets up a comparison between the eagle, best of birds, and the victor. [207] Then in line 83 we find a priming act that consists of the dative τίν and the particles γε and μέν. Unlike the nominatives in Homer, discussed above, the pronoun here is fully part of the syntactical construction, but the verb and subject that complete the construction are postponed to the very end of the song (δέδορκεν φάος). A priming act in this form, consisting of a second-person pronoun and a particle, is comparatively rare in Homer. [208]

§76. In (t35), I take μέν as the Ionic variant of μάν in a cluster with γε. The function of γε μέν is not to mark an adversative relation (Denniston 1950²:387, with Bowra’s reading of the passage), but to mark a conceptual connection. As in the cluster νῦν γε μάν, the first word (τίν) introduces a concept (“the present” for νῦν and “the victor” for τίν) that is emphatically juxtaposed with the preceding. The juxtaposition is in itself neutral, but in the majority of instances of γε μέν/γε μάν the relation between the preceding and the following tends to be one of similarity rather than one of difference. [209] Here, eagle and victor are united by the conceit of distant sight; the eagle sees things from afar (τηλόθε), while the victor has light gleaming from afar on his behalf. Despite the syntactic integration of τίν here, the act τίν γε μέν has a force of its own. Irrespective of what follows, the second-person reference at once redirects attention to the “you,” linking the victor to the image of the eagle. By emphatically directing attention to the victor at the end of the song, the most important person of the event is primed and will remain in the audience’s mind even when the music dies down.

§77. The cognitive and performative usefulness of second-person priming acts like the one in (t35) readily suggests itself: just like priming acts that occur in narrative, these redirect joint attention to a new referent. This referent may be physically present at the performance, such as the victor or a member of his clan, [210] or not. [211] Since even deceased people, heroes, and gods can be made present through Pindar’s use of “you,” the potential absence of the referent does not significantly influence the performative impact of uttering the second-person pronoun. [212] The performative effect, in turn, leads to a higher level of audience involvement directly following the priming act. As a result, second-person pronouns have a natural place right before discursive peaks, such as the final line of the song in (t35). The position of the priming act has a particularly poignant peak effect in the second epode of Isthmian 7:

In the ode for Strepsiades of Thebes, Pindar reserves most of his praise for Strepsiades’ uncle and namesake who had died fighting for Thebes. In these lines, the song transitions from praising this ancestor’s virtues to addressing him directly. [213] Consider especially the stark contrast between the third person imperative ἴστω in 27 and the second-person pronoun followed by the vocative in 31. Lines 27-30 are the expression of a wish (ἴστω, “may he know”), and line 31 appears to immediately fulfill that wish. In performing 31-36, the performer(s), especially if it was a chorus of citizens, enact the praise of the uncle Strepsiades. [214] The transition from the expression of the wish to its fulfillment is reflected in a change from the third person to the second person, and the pivotal act is τὺ δέ, with the pronoun in the vocative or the nominative. [215]

§78. Considering the effect of the second-person priming act, it does not come as a surprise that it often co-occurs with a name in the vocative. Just as a second-person reference does, a vocative turns attention to a new addressee, [216] and in combination the two are complementary. As noted above, at the moment of utterance the referent of a second-person pronoun is underdefined. The most natural way to resolve the ambiguity through verbal means is to provide a name, which after a second-person pronoun inevitably takes the form of a vocative. [217] Vocatives in Pindar have been decribed as adding “liveliness” to the songs, [218] or as reminiscences of epic and hymns, [219] but other scholars have noted their discursive importance. [220] Names in the vocative in Pindar may be compared to apostrophes of characters in Homer. Characters in the epic narrative tend to be addressed at or just before important moments. [221] In Pindar the vocatives likewise occur at discursive transitions, both within the mythical narrative and between myth and hic et nunc.

§79. The effect of second-person references in priming acts in a Pindaric performance must have been profound. By devoting a separate discourse step to redirecting attention, Pindar made sure that the audience would be able to follow the path of his song. Moreover, the second-person pronoun primes the audience to focus on the directly upcoming discourse as an important new action. The difference between second-person priming acts and priming acts in narrative is that they do not only effect a cognitive redirection of attention, but may also trigger a physical shift of gaze, especially when Pindar turns to the victor, as in (t35). The cognitive usefulness of the priming act always goes hand in hand with its discursive importance. The fact that a discourse act is entirely devoted to directing attention to a new referent naturally creates anticipation about the upcoming discourse.

§80. The first sections of this chapter have sketched the scholarship on the smallest subdivisions of discourse in ancient Greek and contemporary languages. The importance of the concept of the discourse act reaches beyond this chapter and beyond the study of particles. In language, the act – consisting of a few words that verbalize the focus of consciousness – has more claim to the status of the basic linguistic unit than the clause or the sentence. Based on the argument that the discourse act is the most basic unit in the language-producing mind, the functions of particles are to be understood primarily with regard to discourse acts rather than to clauses or sentences. Discourse acts and particles exist independently from each other, but each is relevant to the other: discourse acts are the domain over which most particles exercise their force.

§81. δέ in Homer is inherently linked to the discourse act, by consistently marking boundaries between many different kinds of act. Its function is analogous to the use of “stop” to mark boundaries in telegrams written in English. The other particle discussed in terms of discourse acts is μέν. I describe this particle’s pragmatic functions in terms of projection: μέν projects that another act is to be expected. This force of the particle is further exploited in the μέν – δέ construction, when μέν has small scope and specifically projects a contrasting or complementary δέ act. Finally, the priming act, a strongly “incomplete” chunk of discourse in syntactic and semantic terms, illustrates an important discourse function of discourse acts. Priming acts typically occur at the beginning of new scenes, a type of transition that demands significant cognitive processing. To accommodate the speaker and audience’s cognitive needs, the “who” precedes the “what”. This reading of syntactically “dislocated” or “pendant” units recognizes their cognitive efficiency rather than focusing on their grammatical deficiency.

§82. An analysis of Homeric and Pindaric discourse demonstrates that the language of the two performative genres is more similar than one might have expected. Despite the fact that the syntax of Pindar’s songs is occasionally more intricate than that of Homeric epic, the discourse still generally progresses in small-ish acts. Both epic and lyric discourse proceed in small strategic steps, discourse acts, each with their own function in reaching the discourse goals.