§1. In the present chapter I discuss language that refers to the level of interaction not overtly, but through indirect means. This language is not self-referential, but rather marks the relation of the performer to the content in a manner that reveals his expectations about the knowledge of the audience. We are concerned here with the dimension of shared experience, shared knowledge, shared beliefs, and shared discourse.

§2. Tradition is an important part of this dimension, for both Homeric and Pindaric discourse. Consider what Foley says about the function of tradition in the generation and reception of Homeric epic:

Any performance of Homeric epic represents a moment in the continuum of tradition. Whichever part of the tradition is in current focus, the whole remains constantly relevant and accessible as a body of knowledge shared between performer and audience. [1]

§3. Pindar’s engagement with tradition is complex, and his songs must be regarded as both a product and a producer of traditional knowledge. That is to say, his songs stand in a continuum, at once forming and being formed by tradition. [2] The difference between lyric song and epic poetry lies in the fact that lyric has to take into account both local and Panhellenic tradition. [3] As Wells puts it:

In the Panhellenic culture of the fifth century these three temporal dimensions are inextricably linked, and therefore always intertwined in Pindaric song.

§4. The present chapter studies the linguistic reflection in Homer and Pindar of engagement with tradition and other kinds of shared knowledge, a phenomenon that has been termed “discourse memory.” I argue that apparently random use of certain particles – γάρ, ἄρα, and τε in particular – can be understood as metalanguage reflecting the performer’s assumptions about a shared body of knowledge. First, I introduce the term “discourse memory” and locate it within the larger framework of this study (4.1). Then I discuss how the Homeric performer monitors the discourse memory, supplementing it with pieces of knowledge introduced by γάρ (4.2.1). In Pindar’s Victory Odes we find a similar use of γάρ, but, beyond introducing information about the storyworld, Pindar uses γάρ to insert gnômai (4.2.2). A special kind of traditional discourse is that of the Homeric simile, of which a linguistic analysis is presented in the third section, with a focus on τε and ἄρα (4.3). The fourth section of the chapter studies recurrent or common event sequences (“typical scenes”) in Homer, in which ἄρα is important again (4.4). In the final section , I present an analysis of τε in Pindar (4.5).

§5. Before going on to discuss how epic and lyric discourse reflect the process of negotiating shared knowledge, it is necessary to consider what form this body of information takes in our minds. In a 1981 article, Lord describes the role of memory in the performance of epic:

Lord highlights two aspects of memory relevant to the epic performance: knowledge of the narrative sequence on the one hand (the “overall story”), and the crucial elements to every segment of the story on the other. Both of these elements are to a large extent prescribed by tradition, a fact that is not limited to epic. In both lyric and epic, the content and structure of stories is therefore to a certain extent shared by performer and audience. [4] Besides the narratives within epic and lyric, the genres themselves are traditional too: the performer is aware of the audience’s expectations about both form and content. More generally, in a performance at a certain time and place, there exists a body of shared experiences and beliefs, resulting from a world and a culture shared by performer and audience. Finally, the performer can build upon what was said before within the same performance, the shared discourse: he can assume, rightly or not, that the audience remembers the most important events of the preceding discourse.

§6. This pool of information that constantly serves as a background to the unfolding discourse has been discussed in several forms, but I will follow Berrendonner and Roulet in calling it the discourse memory. [5] In II.2 I note that Roulet eventually defines the discourse act as every “update to the discourse memory,” building on the work done by Berrendonner. [6] Cornish, conversely, speaks of “discourse model” rather than “discourse memory,” but he equates his term to Berrendonner’s “mémoire discursive.” In this chapter I follow Berrendonner and Roulet in using the term “discourse memory,” but will also refer to Cornish’s “discourse model,” using the two terms to denote two different things.

§7. The discourse model (Cornish) is “a coherent [mental] representation of the discourse,” [7] while the discourse memory (Berrendonner) contains “the information that, at every moment, is valid for the two interlocutors and shared among them.” [8] Building on Berrendonner’s brief description, I believe discourse memory can be defined more fully. The work by the cognitive linguist Langacker is especially helpful in this respect. He defines the “current discourse space (CDS)” as follows:

I understand the difference between discourse memory and the discourse model as follows: the discourse memory is the whole body of shared knowledge that underlies the current discourse, [9] whereas the discourse model is that part of the discourse memory that has been activated to create a mental representation of the current discourse. In other words, the discourse model is part of the working memory, while the rest of the discourse memory is that part of the long term memory that is shared between performer and audience. [10]

§8. For Homeric and Pindaric discourse, the three components of the discourse memory that exert influence on the linguistic formation and subsequent interpretation of any discourse act in Homer and Pindar are (1) the tradition, (2) the shared knowledge of the world, and (3) the preceding discourse. Every discourse act may thus engage in an interaction with the current discourse model as well as with any other relevant information in the discourse memory.

§9. The world evoked in the narrative of epic or lyric is up to a point removed from the here and now, yet the composer constantly counts on his audience’s basic knowledge of the world. The world created to serve as the theater for a narrative has been called a storyworld. [11] The form of a storyworld is subject to what Ryan calls the “principle of minimal departure,” which states that the storyworld is the same as the “real” world (i.e. the world in which performer and audience live) except for those aspects explicitly mentioned. [12] So in the storyworld of the Iliad, the Greek heroes are taller and stronger than the men of Archaic Greece, but wood still burns, the sea has tides, the sun goes down and comes up again, and so on. Unless the audience receives an instruction to adapt the image, they will project the world they know onto the storyworld.

§10. The discourse memory covers the entire body of relevant knowledge that the performer assumes to be shared between him and the audience at any particular point in the discourse. This cognitive process is a matter of assumption and prediction, since the performer cannot know exactly to what extent the audience shares his knowledge. For that very reason, it is only of limited relevance to know what knowledge exactly is actually shared between performer and audience at a certain point in time. We may compare the use of “of course” in academic papers to indicate that some piece of information is expected to be shared, even if in practice only part of the audience may actually know it. Expressing such assumptions linguistically may thus be as much a rhetorical device as a reflection of reality. In the following sections I trace different linguistic elements that to my mind reflect these assumptions and predictions on the part of the performer.

§11. As a story progresses, every discourse act is an implicit instruction to update the storyworld. However, narration of the action is not the only way for the performer to develop the discourse model; he can also choose to engage with the discourse memory more explicitly. [13] At any moment in the narrative the performer may feel the need to explain something, which he cannot always do while staying in the background. [14] To discuss this phenomenon of narrative discourse in particular, I introduce the concepts of framed and unframed discourse.

§12. In her 1997 monograph, Emmott explains narrative as being built around spaces, contexts that contain certain characters at certain times, and are placed relatively or absolutely within the storyworld. Such a space is a “contextual frame” in Emmott’s terms, [15] and it plays a central role in how a narrative is managed. [16] One may imagine a contextual frame as a space in the mental representation of the discourse, which functions as a receptacle for specific characters, items, and events. A character is “bound” to a contextual frame until the performer provides explicit information to the contrary (“covert continuity”). [17] Once characters are bound to a location, activating that character gives both performer and audience access to the entire contextual frame. [18] Emmott demonstrates that the largest part of conventional narratives occurs within contextual frames: this she calls “framed” discourse. [19]

§13. Framed discourse consists of those acts that are temporally and spatially positioned within a contextual frame in the storyworld. They tell of the events that occur in a certain time and place within the frame of the narrative. To these framed acts, Emmott contrasts “unframed” discourse: acts that do not present narrative events, but rather inform the audience about the storyworld and its inhabitants. Consider the following example:

The narrator shifts from reporting direct speech to describing a character. The italicized passage is unframed, since it is not concerned with what is happening at a certain point in time. In English unframed discourse is marked by its form as generalizations, often a description or backstory, as here. [20] The information given in unframed text is true beyond the current scene in the storyworld: it is not happening in a narrative “here and now.” Since unframed discourse often takes this form, scholars have described it in terms of “descriptive mode” or “background discourse,” [21] but unframed is a more inclusive term: it covers not only descriptions, but also expressions of stance, and it avoids the hierarchical implications of foreground versus background. Most importantly, the status of the discourse as framed or unframed is not dependent on content (i.e. descriptions) but on the attitude of the performer towards the discourse. In many cases, the performer can choose between presenting discourse as framed or unframed: compare “they saw her stride in through the gate. She was a tall woman…” (framed) against “She strode in through the gate. She had always been the tallest of three sisters…” (unframed). [22]

§14. The unframed nature of a piece of discourse is reflected through language in multiple ways. First of all, there is often a tense shift after the transition to unframed discourse, and again when the framed text is resumed. Second, the occurrence of unframed discourse has an effect on the flow of discourse. Although unframed discourse typically has direct relevance to the surrounding framed discourse, the change in perspective means that other discourse processes are interrupted. Anaphoric reference in particular is affected: when a character is introduced into a frame, that character remains available in that frame until we are informed otherwise (“covert continuity”). Characters introduced in unframed discourse, however, do not remain available when framed discourse is resumed. [23] In the example above, the “sisters” are not normally available when the “striding through the gate” frame is resumed. [24] Third, metalinguistic markers occur at the transitions into and out of unframed discourse. Emmott observes that in English written narrative temporal markers often serve to mark such frame transitions, [25] and she concludes:

In addition to the linguistic marking mentioned above, Ancient Greek likewise employs metalinguistic markers at the transitions between framed and unframed discourse. I now turn to the most important of these markers in Homer and Pindar.

§15. The Homeric performer is omniscient, so he can report not only the observable actions of the protagonists, but also their backstories and their thoughts. Moreover, the performer can reflect upon the narrative situation himself, either as a narrating persona or as a performer in the here and now. These steps out of the storyworld into unframed discourse are often introduced by γάρ. They reveal the performance’s nature as an interactive activity, [26] acting as a sign of the performer assessing the knowledge shared between him and audience, and providing crucial information when needed. In Homeric narrative γάρ has three functions: (1) to introduce information about characters or the storyworld, (2) to blend the perspectives of the performer with that of a character, and (3) to introduce evaluative comments about the ongoing narrative. My contention throughout is that it is unproductive to link γάρ to the idea of background, since from a narrative and discourse perspective the acts and moves introduced by γάρ are important. [27] They supply information indispensable for the narrative, heighten suspense, and invite audience involvement.

§16. As discussed in II.3, the performer can use γάρ to unfold certain story paths. The mention of a place, item, or character may trigger an association with another narrative, which is then introduced by γάρ. In the following passage Odysseus, a first-person internal narrator, introduces his plan (βουλή) to defeat the Cyclops. Before he can outline his plan to his audience, however, he has to explain to them that there is a club in the Cyclops’ cave, a crucial element in the plan:

In lines 319-324 a passage is inserted to give information that the audience needs to understand the point of the narrative. The information that Odysseus offers is in the form of a piece of knowledge about the storyworld of his Cyclops narrative. This shift to the unframed move is marked by γάρ and a change to the imperfect tense (ἔκειτο etc), as typically occurs in such situations. The return to framed discourse is marked by μέν and three verbs in the aorist. [28]

§17. It is clear that in passages like (t7) there is no sense of causality inherent in the γάρ clause. [29] Not even if we take into account the possibility of “anticipatory γάρ” – i.e. the γάρ clause providing the cause before the result – can we explain this instance as causal. [30] Nor can we describe this as “background information,” since in narrative terms the unframed discourse occurs exactly because it is indispensable to understanding the ongoing narrative. The nature of the discourse changes for a limited time, but its importance does not. Consider one more example from the Cyclops narrative:

Again, γάρ is used to introduce a piece of knowledge that the speaker thinks the hearer may not, or cannot, know. In both (t7) and (t8) the piece of unframed discourse interrupts the flow of the narrative, in (t8) actually interfering with a syntactical construction. [31]

§18. Compare the following excerpt from an informal conversation recorded by Chafe, about an old-fashioned Swiss professor:

There is no clear marking of the beginning of the unframed discourse, but the return to the narrative frame is marked with “so”; Emmott calls this “frame recall.” [32] The similarity with (t7) and (t8) is striking: as the speaker comes to the pointe of his story, he introduces a crucial piece of information: the presence of the club in (t7), the ram in (t8), and the fact that Wesleyan was still a men’s school in (t9). The speaker then proceeds to tell the rest of the story so that Odysseus has a club to make a weapon, an animal to cling to, and so that the “[g]entlemen” in (t9) makes sense. There may be another motivation behind this strategy: by inserting the information at this point in the narrative, in this form, the climax of the story is briefly postponed. Thus introducing this kind of unframed discourse may serve to build up tension in the unfolding of the framed discourse.

§19. There are countless examples of γάρ introducing pieces of important knowledge in narrative. Beyond the instances discussed above, I simply note examples throughout the other chapters where they occur. For now, I turn to another kind of unframed discourse commonly introduced by γάρ. In II.3 I discuss δὴ γάρ with regard to the function of δή, but now I focus on the function of γάρ in the combination. I have noted before that δή in narrator text occurs in the great majority of cases with a temporal marker, signaling larger steps in the narrative. Much less often, it occurs in narrator text in its intensifying function, either with large or with small scope. One example of this latter category is δὴ γάρ in narrator text: in act-initial position, δή does not appear to have its discourse-segmenting function. [33]

§20. Consider the following example about Deïphobus:

In this prototypical example, it seems almost like we get a glimpse of Deiphobus’ thoughts at the moment that he attacks Idomeneus “I have really always hated him!” However, the addition of ἐμμενὲς αἰεί, as well as the imperfect (ἔχεν) within the narrative in the aorist, strongly suggest the omniscient perspective of the narrator. In contrast to the unframed discourse discussed above, the kind of information introduced by δὴ γάρ here is not about the external world, but about the thoughts of a character. [35] The performer uses γάρ to introduce a little insight into a character’s psyche, creating the impression of what we might call free indirect thought. In this form of representing thought, there is no cue like “he thought: …” (direct thought) or “he thought that …” (indirect thought), but a character’s thoughts are represented more obliquely. [36] The use of intensification through δή within the narrator text may also be a sign of empathy of the narrator with that character: there is a blurring of perspectives. [37] In the parallel instances of δὴ γάρ there is always some similar insight into the feelings of a character. [38]  Unframed discourse in narrative, in short, concerns not only information about the storyworld, but also about the thoughts and feelings of characters. [39]

§21. The expression of omnitemporality ἐμμενὲς αἰεί in (t10) brings us to the final kind of unframed discourse in Homer introduced by γάρ: gnômai, or generalizing statements about the world, which function as comments on the ongoing narrative. Lardinois has noted the possibility of connecting “a gnomic statement to the particular situation to which [the poet] wants the saying to apply, by inserting a particle or conjunction, such as γάρ or ἐπεί, at the beginning of the gnomic expression.” [40] He discusses the phenomenon in Sophocles, but it also occurs in Homer, especially in the combination γάρ τε. [41] Consider the following example from the Odyssey, where Nestor is speaking of Agamemnon’s decision to stay in Troy and attempt to appease Athena:

Nestor is acting like an internal narrator, and his use of νήπιος clearly shows his non-neutral stance. [42] Here and in other examples of γάρ τε moves in Homer, the pattern is clear: after the two adjacent particles there is a shift to present tense (or an elided ἐστι), [43] and this is often accompanied by a form of αἰεί. [44] Finally, this kind of gnomic unframed discourse typically occurs in direct speech, and is practically limited to the Odyssey.

§22. It is telling that we find the combination γάρ τε in these gnomic statements by characters. The discourse function of γάρ can now be explained as introducing unframed discourse, while its logical function here is that of signalling the reason or cause for the preceding. What, then, is the function of τε here? I argue that the addition of τε in this construction serves to mark the statement as referring to a large body of shared knowledge, which we might term tradition. The particle’s use in Homeric gnômai can be compared to that in similes, other instances of timeless discourse, and to its use in Pindaric song. Finally, it may be brought into connection with the occurrence of τε in proverbs in archaic epic. [45] I cannot, therefore, agree with Denniston’s contention that in combinations τε’s “association with particular particles” (i.e. γάρ in this case) is rather “loose and fortuitous.” [46]

§23. There is a correlation in Homer between the occurrence of unframed discourse and the particle γάρ. Whether to introduce additional information that is crucial to the narrative, or to reflect upon the situation, the performer turns to γάρ to mark the transition and to signal the associative link. Any transition between framed and unframed discourse is neutral with regard to the importance of the unframed move. The correlation emerges clearly from the material, but the performer does not have to use the particle when he introduces unframed discourse. [47]

§24. The examples of gnomic unframed discourse provide a natural bridge to Pindaric song: their tendency to occur in direct speech in Homer brings us closer to the communicative situation in Pindar. Pindar’s movements between framed and unframed discourse are both more numerous and more explicit than Homer’s, and he too avails himself of γάρ to add pieces of information about the storyworld where needed. [48] His songs differ from Homeric epic in that Pindar explicitly engages with the real world. Consider the following example:

After the gnomic statement of line 11, Pindar turns to Thrasyboulus, son of the laudandus and addressee for most of this song. [50] The act ἐσσὶ γὰρ ὦν σοφός· marks a discursive discontinuity: it directs attention away from a general statement to Thrasyboulus in particular. The implication is that the preceding expression is not applicable to Pindar’s addressee. Pindar, addressing Thrasyboulus directly and calling upon his authority as poet, seems to say: “You are wise, do not try to deny it, I just proclaimed it.” [51] Here, and in several other instances where γάρ introduces unframed information about the victor or his family, some kind of extralinguistic reference to them would also be appropriate; perhaps in the form of a gesture or a gaze, just before or along with the γάρ act. [52]

§25. Just as the Homeric performer does, but much more often, Pindar provides reflections on the story or the praise event with interjected pieces of wisdom. These gnômai are a common way in Pindaric song to put the rest of the discourse in perspective, and like the γάρ moves in Homer serve to create or recall a shared ground between performer and audience. [53] τε fits gnomic contexts in Homer very well, and it is to some extent surprising that τε does not frequently occur in Pindaric gnômai. [54] Consider the following gnṓmē from the seventh Isthmian ode:

Here the gnṓmē occurs in first-person discourse, it puts Pindar’s statement about his old age in relief. The transition between different moves of the performer becomes even clearer in the following example, called “parenthetical” by Hummel: [55]

Pindar advises Hieron not to be too humble, it seems, and supports his advice with reference to a common piece of wisdom. [56] As in (t8), the γάρ act (and move) interrupts a syntactical whole, and for this reason Hummel calls it a parenthetical construction. She adds: “In this type of structure, the particle serves to signal the change of syntactical and logical level of the utterance” [my italics]. [57] Despite the difference in terminology, Hummel’s assessment resonates with my analysis of γάρ as marking that the speaker is (however briefly) performing a different kind of move. Gnômai introduce knowledge necessary to understand the workings of the world of the discourse. [58] In Homer this world is generally the storyworld, sometimes extended to include the world at large, while in Pindar the emphasis is more on the latter.

§26. The particle γάρ has been called the “most typical PUSH particle,” marking displacement to a new frame of reference. [59] This characterization is important, since in most of its functions γάρ aids in negotiating different strands of discourse, but both in II.3 and here I have described in detail its function in different contexts: II.3 focused on γάρ’s role at the beginning of new moves, especially embedded narratives; here I have described its use in transitions between framed and unframed discourse. From a certain perspective, γάρ’s function of introducing embedded narraratives may be described in terms of starting unframed discourse. [60] An embedded narrative takes the form of framed discourse, but is “unframed” with regard to the ongoing narrative. [61]

§27. Through an unframed move introduced by γάρ, the performer introduces a piece of information (in the form of a motivation, narrative, gnṓmē, or description) into the discourse model, and thus it becomes part of the discourse memory. Another important factor is that a performer can use γάρ to introduce this different kind of discourse exactly when he means to. By choosing the right moment, the performer manages the knowledge shared between him and the audience, but can also manipulate the flow of the narrative and the performance. Below and in II.5.3.3 I discuss ἄρα in unframed narrative, where the performer assumes that the following information is already part of the discourse memory, but activates it again in the current discourse model. Both ἄρα and γάρ can thus serve to manage the discourse memory.

§28. In order to identify the type of move that γάρ introduces, the tense that follows γάρ offers a cue: (1) when γάρ introduces embedded narratives it is commonly followed by an aorist, (2) when γάρ introduces unframed information about the (story)world it is commonly followed by an imperfect, and (3) when γάρ (τε) introduces an unframed gnṓmē it is followed by a verb in the present tense. [62] Of course, patterns involving γάρ are tendencies rather than rules, and since the different uses represent aspects of the same particle we must keep in mind that boundaries will be fuzzy. Notwithstanding this fact, an analysis of γάρ in terms of introducing different kinds of unframed discourse reveals that the particle is involved in at least three different constructions. [63] This discourse-sensitive approach may serve as a step toward explaining γάρ’s multifunctionality across register, genre, and time.

§29. The simile is one of the most recognizable kinds of Homeric discourse, [64] and it touches upon discourse memory in at least two ways. On the one hand it presupposes shared knowledge about the world, while on the other hand it presents itself as a body of tradition within epic discourse recognized by the audience. The simile aids in visualizing a complex scene by imposing a somehow different (if not always simpler) image upon it. [65] In a simile such as “Achilles faught like a lioness who, when she sees her cubs attacked, comes snarling at her prey. Just so Achilles attacked…” “Achilles” is the tenor, “lioness” is the vehicle, and the whole desciption of the vehicle (“like a lioness…prey”) is the “vehicle portion.” [66] By its very nature the simile presupposes that the image in the vehicle portion is recognizable, even if in reality the audience may never have seen a lion. [67] The recognizability derives from the fact that the image is part of the tradition familiar to both performer and audience. In this section I argue that the simile’s dependence on tradition is one of the determinants of the Homeric performer’s use of particles. [68]

§30. My results are based on a study of the language of the similes listed in the appendix of Scott’s study: 341 in the Iliad and 134 in the Odyssey. [69] A number of these similes take the form of only a marker of comparison (ὡς, ἠύτε, (ε)ἴκελος, etc.) and a noun phrase, and are therefore too short for the study of particle use in the simile. The corpus that remains is made up of all those instances where the vehicle portion of the simile is expressed with a clause containing at least one finite verb. This kind of construction occurs 215 times in the Iliad (making up 708 lines) and 47 times in the Odyssey (148 lines). Of this group, the tenor is repeated in the majority of cases, 182 times in the Iliad and 38 times in the Odyssey.

§31. Typically, the simile takes the form of the following two examples, one drawn from the Iliad and the other from the Odyssey: [70]

Both these similes follow the basic pattern of the Homeric simile: (1) the vehicle is introduced by a marker, here ὡς, as in the large majority; (2) a combination of a relative pronoun and a particle introduces the information in the vehicle relevant to the comparison; (3) cohesion in the first part of the simile is attained by the recurrence of τε; [71] (4) the simile is rounded off with another marker – here ὥς, which is the rule – in (t15) followed by ἄρα. The particles τε and ἄρα are particularly frequent in similes: τε most typically occurs in the first part of the simile, the vehicle portion, while ἄρα generally occurs at the start of the simile’s resolution, with the re-introduction of the tenor. The first element in the simile, the introductory marker of comparison, varies in form but is always present. [72] The latter three elements deserve more discussion.

§32. τε is inherently bound to the vehicle portion of the simile, where it occurs on average once every two lines in the Iliad, and once every three lines in the Odyssey. [73] The high frequency of τε is remarkable, and suggests that it is connected to the simile’s strong dependence on discourse memory. Scholarship has sought to link the value and function of τε, especially when it is not copulative, to the permanent truth of the content of the clause in which it appears. [74] Rather, the particle should be linked to the interaction between performer and audience: τε marks not a state of the propositional content of its host act, but an assumption by the performer that this piece of knowledge is shared between him and the audience. [75] In other words, the performer uses τε to mark its host act as accessible in the discourse memory. To be more precise, τε refers either to the “tradition” part of the discourse memory (consider especially its use with names and places) or to the “shared experience” part. In similes, the performer accesses on the one hand shared daily experience (as in shepherd or weather similes), and on the other hand the shared experience of epic: it is unlikely that the audience learned the image of the bloody lion from anywhere other than epic. In the end, this use of τε is a discourse strategy: regardless of whether a piece of information is “permanently true,” the performer presents it as something shared between him and the audience.

§33. Aside from this general high frequency of τε in the vehicle portion, there is a further interesting element to the use of τε in similes. The vehicles introduced in similes, gusts of wind in (t15) and a lion in (t16), do not carry the comparison per se – the armies and Odysseus are not compared to winds or a lion, but rather to these things in a specific state. [76] This closer approximation is introduced by a relative pronoun accompanied by τε – οἵ τε in (t15), and by ὅς ῥά τε in (t16). In this way the expanded vehicle portion makes the scene more vivid for the audience while at the same time enabling the poet to convey what he views as the crux of the comparison. [77] In (t15) it is the combination of dust and wind that creates the necessary image, and likewise in (t16) it is the lion after feeding that provides the image of Odysseus that the poet wants to invoke. [78]

§34. The following example from the beginning of Iliad 3 illustrates this use of τε:

The entry of the Greek and Trojan armies on the battlefield is accompanied by a series of similes. The second, given in (t17), is used to describe the advance of the Greeks. The advance of the Trojans has just been compared to that of a flock of twittering birds, noisy and without order, and now the performer turns to the Greeks. They come on in silence, and the dust their feet throw up provides a cover for them as if they were thieves. Since we are in a situation of war rather than peace, the fact that the mist is a nuisance to shepherds suggests the perspective of the Trojans: the Greeks are hidden from their view, just as a mist may hide a flock from shepherds. The image of the thief, in turn, matches the Greeks who benefit from the dust that hides them. The crucial link proposed by the simile is that between the advancing Greeks and the thief taking advantage of the mist, and this is the element introduced by τε.

§35. Despite the frequency of τε in similes, it does sometimes occur that the particle is absent from the vehicle portion of a simile. In the following example from Iliad 9, Achilles compares himself to a mother bird: [79]

Here we find a whole simile without τε, the absence of which is especially remarkable in the second half of 324, since this act provides the simile’s key as in the examples above. A closer look at the editions reveals that most manuscripts as well as a third-century papyrus read κακῶς δέ τέ οἱ. [81] It appears to be mainly the testimony of Aristarchus, transmitted in the scholia, that has led to most editions giving δ᾽ἄρα. [82] Considering the frequency of τε in the vehicle portion of the simile, the frequent link between τε and the key of the simile, and finally the manuscript evidence, I would propose that δέ τέ οἱ is the more attractive reading.

§36. The following two similes feature similar textual variants:

In line 523 of (t19), several manuscripts and papyri give τε as a variant reading for ἑ. Editors appear to prefer ἑ as the lectio difficilior, but the statistics for τε in similes support reading τε. The same goes for Odyssey 4.338 in (t20), where a group of manuscripts reads ὁ δέ τ᾽ ὦκα rather than ὁ δ᾽ ἔπειτα. [84] I cannot establish here what has led editors to prefer the latter reading, but nothing textual or metrical speaks against reading the former. The more common ὁ δ᾽ ἔπειτα might be argued to be the lectio facilior, while at the same time this construction never occurs in a simile. [85] A further argument for reading ὁ δέ τ᾽ ὦκα is that in Iliad 21.261 we find τὸ δέ τ᾽ ὦκα, in the same metrical position, in a simile describing how the river Scamander overtakes Achilles. From the discourse perspective, in both (t19) and (t20) the acts that have τε in a variant reading introduce the simile’s salient element: the gods who drive on the smoke is compared to Achilles driving on the Trojans, and the lion who enters his lair is compared to Odysseus returning to his palace.

§37. To sum up, the function of τε in the vehicle portion of the Homeric simile is roughly twofold. On the one hand the particle creates cohesion on the level of language and on the level of knowledge shared between performer and audience. [86] With τε, the performer marks a certain piece of knowledge as available in the discourse memory. [87] On the other hand, when it occurs in the crucial act like in (t14) and (t15) – especially when it occurs only there in the vehicle portion – τε appears to mark its host act as the pointe, the salient element of the simile. [88]

§38. Now let us consider the end of the simile. As mentioned above, the comparison is always capped by a marker, generally ὥς, sometimes followed by ἄρα. The occurrence of ἄρα in the final part of the simile accompanies the return to the narrative frame. Scholars have noted that across languages conclusive particles (e.g. “so” in English (t9), also in German, and igitur in Latin) often serve to pick up or recall the main thread. [89] ἄρα in the return to the tenor of the simile does precisely that. [90] This function of ἄρα can also be seen in its occurrence directly after direct speech, as well as after different kinds of unframed discourse.

§39. Earlier scholarship on ἄρα vacillates between two extremes: some scholars view it as marking the upcoming sentence as expected from the preceding, [91] while others assert that the particle marks the new sentence as something unexpected and noteworthy. [92] Denniston’s analysis, following Hartung’s position that ἄρα indicates surprise, is still most commonly held to be right. He explains ἄρα as expressing a “lively feeling of interest,” since “[f]or Homer, as for a child, the most ordinary things in daily life are profoundly interesting.” The main reason for Denniston’s choice is that the particle occurs so often in Homer that it is unlikely to confer an idea of connection (as held by other scholars), “except in so far as some kind of connexion must be present in all speech or action.” [93] Ironically, in his attempt to refute connection as the central value of ἄρα, Denniston may have pointed exactly to how the particle works.

§40. Whereas τε is used to refer to knowledge shared beyond the current discourse, ἄρα introduces information that is assumed to be available or inferrable from the preceding discourse. It is thus used in rephrasings, quasi-repetitions, and recaps, after similes, direct speech, and excursus. Moreover, ἄρα is used when the action or event in its host act follows naturally from the preceding. [94] That is to say, it occurs with expected actions in typical sequences (see below §§45-52), and with other logical consequences of preceding discourse. The use of ἄρα in similes matches perfectly the fact that it marks something readily accessible in the mental representation of the discourse. Consider the following examples:

In (t21) ἄρα accompanies the return to the narrative frame in a verse that recapitulates the line just preceding the simile. There is clear resonance on the level of content as well as on the lexical level, as the bold words show. In (t22), conversely, the link established is more tenuous, but the image of the Achaeans advancing is retrieved with the expression “under their feet”:

In both instances, ἄρα marks the return to the narrative frame, and the projection of the simile’s image on the situation in the storyworld.

§41. ἄρα’s presence is not required after a simile, however: it occurs 30 times in the second half of the simile out of a total 182 in the Iliad, versus 9 times out of 38 in the Odyssey. [95] When ἄρα is absent, no other particle follows, except for μέν on five occasions. [96] Since the second half of the simile signals the return to the narrative and thus the beginning of a new move, the occurrence of μέν is not surprising, [97] nor is the high frequency of asyndeton.

§42. Out of 262 full similes in Homeric epic, there is really only one example that diverges significantly in its particle use, from Iliad 23. After Diomedes has won the chariot race at Patroclus’ funeral games, Menelaus and Antilochus battle for second place, which is finally taken by Antilochus – though apparently not entirely fairly. The minuscule distance between the two chariots is described by the following simile:

This little-discussed simile seems, at first glance, to have all the characteristics of a Homeric simile, but it is remarkable in several ways. Rather than illustrating the scene by means of unrelated imagery, the simile retains the image of a horse and chariot in order to establish quite precisely the distance between Antilochus and Menelaus. [98] The mention of the horse drawing its lord (ἄναξ) through the plain does not remove the image from the current scene, for Menelaus qualifies as an ἄναξ, and the venue for the games is the Trojan plain. Beyond making an analogy, then, the passage is an attempt to more precisely establish the physical position of the characters in the storyworld. From a narrative perspective, moreover, the tension evoked by the image seems to miss its mark since the outcome of the scene has already been reached: Antilochus has beaten Menelaus.

§43. The language adds to these peculiarities, in the form of τόσσον δή just after the vehicle portion. This is the only occurrence of δή in the resolution of a simile, and the particle does not fit the context as readily as τε and ἄρα. The function of δή in this situation is more likely to be one of intensification rather than of marking a narrative boundary. [99] The only comparable instance is in Odyssey 21.253, ἀλλ᾽ εἰ δὴ τοσσόνδε βίης ἐπιδευέες εἰμὲν / ἀντιθέου Ὀδυσῆος…, where δή appears to have a small-scope intensifying force over τοσσόνδε: “but if we are really that much weaker in might than god-like Odysseus….” As noted in II.3, δή in peninitial position of the act may have scope over the following word, as possibly in the Odyssey example, [100] or over the entire act, as appears to be the case here in Iliad 23.522. It cannot be excluded that δή has a small scope over the preceding τόσσον, which gives the most natural reading, but this use of δή is rare in Homer, and only becomes more frequent in later Greek. [101]

§44. Now, the nature of δή does not exclude its use after the vehicle portion of the simile, but the question remains why it is employed only here. The answer may lie in the key of the simile, which is the distance between two competing chariots. In all other instances of ὅσσος – τόσσος similes, [102] the comparison is one of approximation, whereas in this case the image is very specific: as closely as a chariot follows the horse – so close that its tail can touch the wheel – exactly as closely behind Antilochus was Menelaus. With scope over the entire act, δή places emphasis on the exactness of the entire act – they were that close. This emphasis serves moreover a narrative purpose, since it vindicates Menelaus’ claim that if Antilochus had not cheated, the former would certainly have beaten Nestor’s son.

§45. The linguistic make-up of the Homeric simile is thus rather uniform, and a good illustration for the relation of τε and ἄρα with discourse memory. The difference between τε and ἄρα may be illustrated by the fact that ἄρα can occur within the first half of the simile, while τε never occurs after the final ὥς in the simile. In Homeric discourse, ἄρα is used especially to mark the assumption that the preceding discourse provides the background for understanding what follows. More restricted than ἄρα, τε is used to introduce pieces of knowledge that the performer treats as shared between him and the audience, on the basis of shared experience and tradition. In his similes, the Homeric performer adduces evocative images to both clarify and intensify particular moments in the narrative. The simile illustrates these functions of τε and ἄρα particularly well, but they are relevant to the interaction between performer and audience in all Homeric discourse.

§46. The discourse memory contains any kind of information: not just facts, descriptions, names, and events, but also experience-based knowledge of event sequences. This latter kind of knowledge is called a scenario or a script. [103] Scenarios or scripts are packages of associated knowledge about certain activities, such as a sacrifice or an assembly, that are activated as soon as the relevant activity or event is evoked. “[T]he basic structural tenet of scenario theory is that much of the information that we store about the world is stored as situation-specific representations.” [104] When a scenario becomes relevant, the knowledge shared between performer and audience creates expectations that can either be met or frustrated in the following discourse. As with the storyworld, the natural reaction to the activation of a scenario is to assume that it follows the known path, unless it is stated explicitly that it does not. [105] In epic, scenarios or scripts are most clearly visible in passages known as type scenes or themes.

§47. The heroes of the Iliad and Odyssey typically engage in activities that the audience may not be familiar with, such as debating with kings or laying siege to a city; in addition, they engage with gods and monsters. Besides these special activities, however, Homeric heroes must still perform the mundane acts of eating, washing, and travelling from place to place. These activities may seem ordinary, but they are not in traditional epic, and accordingly have received a considerable amount of attention. [106]

§48. The first to study this corpus of recurrent scenes was Arend, who coined the term typical scenes (“typische Szenen” ). His intent was to show what connects them and how Homer uses variation for rhetorical and stylistic purposes. [107] In his review of Arend’s work, Parry makes one important addition. [108] Parry argues that the origin of the type scene must be sought in the nature of oral poetry. Born of tradition, the type scene is a resource in composition, a pattern that the singer learns in his training and can access at will. Lord expands on Parry’s claims, but he speaks of “themes” rather than of type scenes. In a comparative study of Homeric and Yugoslav epic, Lord defines theme more broadly than Arend’s typische Szene, and speaks of “a process of composition by theme among oral poets.” [109] In Lord’s analysis, themes are not a special kind of discourse, but the main building blocks of epic composition, important aids to the composer’s memory. Nagler pushes this final idea a bit further when he speaks of a theme as a “preverbal Gestalt for the spontaneous generation of a “family” of meaningful details.” [110]

§49. Minchin draws a direct link between Nagler’s idea and the cognitive concept of scripts. [111] Her adaptation of Arend’s type scene and Lord’s theme is that she does not (only) regard the epic theme as a product of epic tradition, but (also) of human experience. [112] The poet does not narrate the epic theme of saddling a horse, he merely activates his own (and depends on the audience’s) memory of preparing a horse for riding. In my terms, then, Minchin proposes that the performer taps into a different part of memory: that of experiential knowledge in addition to that of epic tradition. [113] Minchin describes the relation between formula and theme as follows:

That is to say, whereas the formulaic language is part of what the epic singer learns in his craft, the use of themes in composition follows naturally from human experience. I concur with Minchin that a natural connection is to be drawn between the epic phenomenon of typical scene or theme and the cognitive concept of the script. However, the origin of the script need not always lie in common human experience, but may well be the specific epic experience of the performer.

§50. In the following analysis of two recurrent scenes (bathing and arming) I address the question whether typical scenes are instantiations of elements inherent to epic (traditional) or whether they may also be regarded as verbalizations of common scripts (universal). I study the linguistic make-up of the scenes with special attention to particles. The typical scenes of “bathing” and “arming/clothing” demonstrate how the scripts in the performer’s mind project a sequence known to the audience, the fulfillment of which is marked by ἄρα.

§51. The typical scene of bathing occurs only in the Odyssey, but it serves as a good starting point because of its brevity and relative uniformity throughout its different instantiations. Consider the fullest example from 17.87-90:

The parallels are largely the same, though in some instances the scene is shorter. [114] As here, the particles that appear most frequently in these scenes are δέ and ἄρα. The frequency of δέ roughly matches its average in narrator text, but the same is not true of ἄρα. Especially in the second part of the scene, the particle recurs consistently. Its presence thus marks the bathing scene as a little narrative (the progression of which is marked by δέ) that happens to be predictable. Washing, clothing, and returning to the public space are details that can always be expected. The underlying script explains why the performer marks the later narrative steps with ἄρα. To put it in different terms, the activation of the “bathing” script projects a sequence of actions, and the fulfilment of this projection is marked with ἄρα.

§52. Another recurrent scene is that of clothing the hero. Consider this clothing scene from Iliad 2, when Agamemnon has just woken from his prophetic dream:

The clothing [115] or arming [116] scene has numerous parallels, mostly in the Iliad, that all follow the same pattern. As in the bathing scene in (t25), the sequential actions are separated by δέ. Even more than the bathing scene, the linguistic make-up of the clothing scene is indistinguishable from the surrounding narrative.

§53. An important implication of this conclusion is that the use of ἄρα throughout Homeric narrative may be connected to the existence of knowledge stored in the form of scripts. [117] This supports Lord’s idea that themes are not a special kind of discourse, but the main building blocks of epic. This knowledge available in the discourse memory includes those scripts learned through the experience of epic. That is to say, the origin of the scripts that underlie the building blocks of oral epic is not just daily experience, but also the daily experience of the epic poet as epic poet. Likewise, the knowledge that the performer assumes that his audience possesses derives from their experience as an audience at epic performances. Let me give one example, often adduced as proof that ἄρα marks “surprising” information:

If we imagine being one of the characters, we may agree with Ruijgh’s claim that “the fact marked by ἄρα is surprising.” [118] However, the audience of an epic performance knows that the linguistic construction καί νύ κε always projects its opposite. [119] In other words, the narratively surprising event is rendered in a projected and thus expected discourse act. ἄρα, as often, functions as metalanguage to mark its host act as expected, irrespective of the discourse act’s content.

§54. Out of all archaic and classical Greek authors of whom a significant corpus is extant, Pindar’s songs have the highest frequency of τε, higher even than Homeric epic. [120] In the great majority of cases, τε in Pindar is copulative. Keeping in mind the strong influence of Homer on Pindaric language and its overall archaic nature, it is surprising that so-called “epic” τε is rare in Pindar. In this section I first discuss these few instances of “epic” τε in terms of activating shared, traditional knowledge in the discourse memory. Second, I examine copulative τε’s range of uses, on the one hand to conjoin constituents and on the other hand to connect phrases or clauses. Then I argue that there is a clear common element in the function of τε across copulative and non-copulative uses. Even in its copulative function, τε signals that a fact or a relationship between facts or constituents is available in the discourse memory. Finally, I take the instances of τε in Olympian 1 as a case study to demonstrate that the choice for τε over καί is not arbitrary.

§55. Ruijgh defines “epic” τε as those instances of the particle where τε is not copulative, that is, where it cannot be substituted by καί. [121] There are around sixteen instances (out of 526 [122] total instances in the Victory Odes) where τε is not copulative; in each case after a relative or demonstrative pronoun. [123] Ruijgh divides the cases into two categories: τε introducing digressions containing “permanent facts” and τε introducing relative clauses containing “temporary facts.” Based on my argument in the present chapter about τε accompanying knowledge in the discourse memory, and particularly knowledge shared beyond the present discourse, the distinction between “permanent” and “temporary” facts is perhaps not productive. Consider an example from Ruijgh’s category of permanent facts:

Themis speaks to Zeus and Poseidon, who have been quarreling over Thetis. Themis prophesies that if Thetis will indeed lie with Zeus or his brother, the offspring of that union will be more powerful than his father. Instead, she proposes, Thetis should be married to the mortal Peleus. The relative clause introduced by ὅν τε adds that he is “said to be” the most pious man in Iolkos. Most of Ruijgh’s permanent facts concern gods, but he explains the use of τε here because it concerns the fame (“renommée”) of a hero. [125]

§56. This complication can be avoided altogether if we consider the two relevant contexts of this utterance. On the one hand Themis makes a claim about Peleus’ piety among the gods, and on the other hand Pindar declares Peleus the most pious man from Iolkos. In the first context, it is the gods who judge who is most pious, which explains why Themis seeks to create intersubjective agreement through the use of τε. In the second context, Pindar appeals to the audience’s knowledge of the story of Peleus, who came to Aegina from Iolkos. The fact of Peleus’ piety may be traditionally accepted or not, but what is important is that the particle marks the fact as shared in nature, and by extension possessing intersubjective truth.

§57. Compare the following example of a “temporary” fact quoted by Ruijgh:

The only explanation Ruijgh offers for this passage is “fait mythique,” “ mythical fact.” [126] The effectiveness of Ruijgh’s dichotomy between permanent and temporary facts appears to break down here. After all, if a mythical fact is not permanent, then what is? The passage introduced by ἅν τε appears to be the salient element in the mythical narrative about Ixion’s attempt on Hera (announced in lines 26-33). [127] The reason he was caught is that Zeus tricked him. This is a fact that Pindar takes from tradition, which he conveys with the addition of τε—whether the entire audience actually knows this exact detail or not. To come back to Ruijgh’s definition, it is neither important whether τε marks a fact or not, nor whether this fact is permanent or temporary. All that matters is the social contract of the lyric performance: performer and audience partake in the same world, culture, tradition, and event: τε appeals to and creates exactly this shared knowledge.

§58. So far the use of τε in Pindar is largely consistent with its use in Homer. However, there is a basic quantitive difference between use of τε in Homer and in Pindar: whereas in Homer non-copulative τε makes up a fifth of the total, in Pindar τε is copulative in the absolute majority of instances. [128] These numbers reflect the tendency of “epic” τε’s decline after Homer, except in hexameter and elegiac poetry. [129] This quantitive development notwithstanding, the copulative and non-copulative uses of τε may be more closely connected synchronically than Ruijgh and most other grammarians concede. In fact, I believe that τε’s pragmatic function of marking knowledge as shared is common to both the copulative and non-copulative uses. The difference between the two uses of τε lies only in their syntactic function. [130]

§59. To support this claim, let me now turn to copulative τε in Pindar. Hummel distinguishes the following kinds of conjuncts connected by τε (with my numbering): [131] (1) τε can connect clauses (“sentential τε”); (2) τε can connect phrases; (3) τε can connect “paires idiomatiques”: two items that exist as a pair not only in Pindar, but independently in language; [132] (4) τε conjoins pairs unique to Pindar; (5) τε conjoins complementary pairs, sometimes in the form of a hendiadys.

§60. The last three entries in the list all concern τε connecting constituents. Except for the examples of hendiadys, all these instances may be understood better when we consider the cross-linguistic notion of “natural coordination,” as proposed by Viti. [133] Viti’s point of departure is Wälchli’s definition of “natural coordination” as opposed to “accidental coordination”:

In two recent studies, Viti has argued that in Homer copulative τε is used more for natural coordination, and καί more for accidental coordination. For both Homer and Pindar, I believe we may apply the idea of natural coordination even more productively to the use of τε if we bring it in connection with the concept of discourse memory.

§61. Wälchli speaks of pairs that are “closely related in meaning, and (…) form conceptual units.” However, I submit that the natural association of any pair depends on the cultural, performative, and discursive context. In the right situation, almost any two items may be regarded as naturally connected. The factor that makes any coordination natural is shared knowledge: the whole of the discourse memory. Consider the pairs “earth and sky,” and “blood, sweat, and tears.” The perceived natural association between these terms has its origin in different parts of the discourse memory: “earth and sky” is shared human experience, “blood, sweat, and tears” depends on a shared culture, and will only be a natural tricolon for a limited group of people. I will apply natural coordination as a relative term, depending on the context, language, and the participants in the discourse. [134]

§62. In the context of Pindaric performance, many items can be presented as naturally associated. Consider an example from each of Hummel’s last three categories:

Nectar and ambrosia are related as the drink and food of the gods, but this link is limited to a context of people who share this tradition. Hummel lists it among the examples of “paires idiomatiques.” Slightly more specific is the relation between the following two items:

Hummel lists this passage as an example of a pair that is unique to Pindar. However, the relation between athletic games and battle is inherent in the ritual dimension of the games. It is generally assumed that in essence athletic contests are mock battles that allow participants to win honour outside of actual war. [135] In the epinician genre, we may say that it is a natural pair from the perspective of the generic conventions. Finally, there are pairs that are related on a more ad hoc basis:

In the final lines of Pythian 4, Pindar asks Arkesilas to allow the exiled Damophilos to return to Cyrene, since he has been away “from his fatherland and his possessions.” The two are no natural pair (Hummel lists the passage under complementary pairs) but they presuppose some specific knowledge about the exile, presumably shared between performer and at least a part of the audience.

§63. The combination τε καί serves even more consistently than τε alone to mark “natural” coordination. [136] In the great majority of instances, the construction X τε καί Y in Pindar conjoins two constituents that are closely connected, for several possible reasons. Closest to Wälchli’s idea of “natural coordination” come examples like the following: [137]

Combinations like land and sea appear to be expressions of shared human experience, but even they are context-based: this one will not be natural to people from a land that knows no coastline. Other pairs conjoined by τε καί are even more clearly bound together by a relation that is natural within a specific context. [138] Finally, τε καί occurs particularly often with geographical locations and names. [139] In the remaining cases, τε καί either does not conjoin two syntactically symmetrical constituents, or τε and καί work separately. [140]

§64. Beyond τε καί, in the two semantic fields of geography and proper names, Pindar prefers τε coordination over καί coordination. Overall, τε is directly adjacent to a place or name in almost a third of instances. [141] Often, the relation between the conjuncts is expressed as natural for contextual reasons. Consider the following passage from Pythian 11:

The venues for the two most important Panhellenic games are listed with the use of τε only. [142] The relation between Olympia and Pytho is obvious in the context of an epinician ode. The same natural connection exists between the siblings Castor, Pollux, and Helen:

This kind of connection between people or gods that are naturally associated in a specific context recurs throughout the Odes. [143]

§65. In all these cases, it is the knowledge shared between performer and audience that makes two conjuncts a natural pair. From this perspective, the copulative use of τε and so-called “epic” τε are no longer so far apart. Whether τε introduces a relative clause or conjoins two naturally connected items, τε consistently denotes a known relation: either between a referent and something about him or between a number of items, referents, or places.

§66. In addition to marking constituents as a natural pair, τε signals this kind of coordination on a macro-discursive level. In the final paragraphs of this section, I discuss the use of τε in the narrative of Pindar’s first Olympian. This most well-known of Pindar’s songs in part owes its fame to the intriguing aetiology it presents of the Olympic Games. In his telling of the myth of Pelops, Pindar ostensibly rejects one version of the tradition and substitutes a more politically correct one. However, Pindar does not actually substitute a new story for an old one, but places two available stories in a hierarchy. [144] The first aetiology mentioned in Olympian 1 is the story of Pelops being cooked and eaten. This particular narrative correlates with the footrace, the earliest event in the Games. This footrace ended at the ash heap where the thigh pieces of a slaughtered ram would then be burnt by the victor. [145] By the fifth century, however, the chariot race had become the most important event in the Olympic Games. It is this shift in popularity that explains why Pindar gives priority to another version of the story. [146]

§67. The foregrounded story in Olympian 1 is the episode where Pelops has to race Oinomaos in a chariot to win his daughter Hippodamia. Pindar does not treat this as a new version of the myth, but the “Pelops is eaten” episode precedes the “chariot race” in time. In Nagy’s analysis, Olympian 1 presupposes the following elements of the story:

Now let us consider the linguistic realization of these elements (/ marks line end):

Out of Nagy’s nine elements of the story that are expressed in the text, six contain τε (one shared between (c3) and (a’)). These passages account for eight out of thirteen instances of τε in Olympian 1; the other five instances are outside the narrative.

§68. In these narrative acts, τε almost completely supplants καί (one instance in the acts above, out of nine total instances in the ode). Moreover, τε highlights specifically the essential elements of the tradition, a phenomenon that we might compare to the high frequency of τε in the salient parts of a simile. Whether in specific pairs of people, places, items, and ideas or in crucial elements of a traditional narrative, the distribution of τε follows a consistent pattern. It may not always be clear to us what determines the choice of τε over καί, but it cannot be a coincidence that in contexts of shared tradition and shared knowledge τε is preferred.

§69. The language of any discourse is to a large extent determined by what came before, and what comes before a current discourse act reaches far beyond what has been said before: “The meaning of a text is more than the sum of the meanings of the individual sentences that comprise it.” [148] The reason behind this claim by Schank and Abelson is that every discourse act interacts with the discourse memory to create a fuller meaning, and a fuller representation in the discourse model than the words alone add up to. In this chapter I have described some of the many possible interactions between current discourse and the larger discourse memory, and its marking through metalanguage.

§70. The particle γάρ serves to introduce acts that ensure a shared ground, a shared mental representation of the discourse (discourse model) between the performer and the audience. Engagement with the discourse memory is a question of attitude rather than of fact, so in the case of γάρ it is the performer’s belief that a piece of knowledge is missing from the discourse memory that determines its use. Both in Homer and Pindar γάρ is thus used to introduce information about the storyworld, but also to insert comments on the ongoing discourse. One subset of the latter, gnômai, generally occur in direct speech in Homer, and are introduced by γάρ τε. In Pindar, gnomic acts account for a much larger proportion of the instances of γάρ than in Homer.

§71. In the Homeric simile, interaction between discourse and discourse memory is constant and particularly visible. Whereas γάρ serves to introduce additional information into the discourse memory, ἄρα and τε accompany knowledge already shared. Both particles do this in their own way, and again they are relevant to the performer’s expectations. ἄρα occurs in practically every context in Homeric discourse, and I align with the scholarship that links ἄρα to “expectedness.” Rather than focus on the audience’s perspective, however, I propose that ἄρα reflects the performer’s stance toward his discourse. By uttering ἄρα, the performer metalinguistically marks the current discourse act as either known or naturally expected from what comes before. Since ἄρα works as metalanguage, this value of ἄρα does not necessarily mean that the propositional content of an act is expected, but typically it is.

§72. As emerges from its use in the Homeric simile, τε occurs in more specific contexts. In terms of discourse memory, the difference between ἄρα and τε concerns the part of the discourse memory that is accessed. ἄρα typically – but not exclusively – refers to the current or past discourse, whereas τε refers to the discourse memory beyond the preceding discourse. An analysis of τε in the Homeric simile and Pindar has revealed that τε marks both facts and relations between facts, concepts, places, and people as shared between performer and audience beyond the present discourse. Thus it typically co-occurs with names, places, or actions that are part of the shared experience or tradition. In Homeric epic, this pattern holds for those instances where τε is copulative as well as for those where it is not. In Pindar τε has specialized in its copulative function, but its use still shows clear traces of interaction with the discourse memory.