III.2 Varying one’s speech: Discourse patterns

§1. Every speaker is familiar with the use of different varieties of her language for different communicative situations. The language we encounter and use varies, for example, when we are talking to friends, reading a newspaper, or writing an article. In other words, many linguistic choices depend on the communicative situation. [1] Linguistic variation based on situational context has been the subject of the fields of register studies and Construction Grammar, which have developed the concepts of “register” and “discourse pattern.” In this chapter I will build on this research to explore particle use in Greek drama.

§2. I will argue that dialogues, monologues, and choral songs in Greek drama may be considered discourse patterns—that is, conventional pairings of certain functional and formal characteristics. Because of the differences in these characteristics, a particle’s distribution across the three discourse patterns illuminates its functions and uses. If a particle is usually more frequent in one discourse pattern rather than another, this means that it is connected to specific pragmatic goals associated with that discourse pattern. Consequently, the non-random distributions of particles contribute to the interpretation of their functions: macro-pragmatics illuminates the micro-pragmatics of particles’ local functions. My analysis in this way refines specific readings of particles, and at the same time highlights pragmatic characteristics of each discourse pattern, which inform, and are informed by, the use of linguistic features other than particles as well. We shall see that the distributions differ widely across the particles, and that each particle has different associations that explain why it is more or less frequent in a certain situation. Accordingly a variety of topics, such as shared knowledge, the speaker’s prominence, and allusions to other genres, will play a role in my explanations.

§3. I begin by introducing the concepts of “discourse pattern” and “register” (§§4-9), and then discuss previous research on linguistic variation in Greek drama (§§10-15), and situate my analytical method within the scholarship (§§16-21). The main part of the chapter (§§22-89) discusses the distributions of eleven particles (ἀλλά, γάρ, γε, δέ, δή, δῆτα, ἦ, καί, μέν, οὖν, and τε) in order of their frequencies, and uses these distributions as input for interpretation. In other words, I will argue that analyzing dialogues, monologues, and choral songs as discourse patterns sheds light on the use of linguistic features such as particles. §§90-95 present the chapter’s conclusions. Finally, an appendix (§96) gives the particle distributions that were found to be statistically non-significant.

§4. Scholars who study discourse patterns or registers aim to learn about the interrelation between situational and linguistic variation. The situational variation includes, in the case of different formats within tragedy or comedy, sub-situations within one larger situation. The term “discourse pattern” was developed in Construction Grammar (CxG), a cognitive approach to language use. This approach assumes that all the linguistic knowledge of speakers and writers is stored in symbolic pairings of form and meaning, called “constructions.” [2] The form can be anything, from a morpheme to a word, a phrase, a sentence, or an intonation pattern. Like the form, the meaning may also be as specific or abstract as the construction in question requires. Examples of constructions that have been studied from this perspective are the classical Greek potential optative, and pragmatic particles in several languages. [3]

§5. Östman 2005 argues that even whole texts, such as recipes, novels, or conversations, should be considered constructions, because their general make-up is part of the linguistic knowledge of speakers and writers. [4] That is, speakers and writers know which specific forms are allowed and appropriate in which kinds of text. An entire text, Östman argues, can thus be one construction, with a “form” pole and a “meaning” pole. He calls these large constructions “discourse patterns”: they are “conventionalized associations between text type and genre” (132), just as constructions are conventionalized form-meaning pairings. What is acceptable as a whole text, he argues (127), is subject to similar conventionalizations as what is grammatical as a sentence.

§6. The form pole, called “text type” by Östman, is highly general, and much less filled in with specific words and forms than a small construction such as a word or a phrase. It includes information about the relative frequency of linguistic features, in comparison to other discourse patterns. The meaning pole, which Östman calls “genre,” [5] is also general: it can be a designation such as “an academic article,” “a recipe,” or, we may infer, “an Aristophanic lyric song.” The broad, overarching meaning is not derivable from the parts of the construction, such as a high frequency of nouns, imperatives, or a certain particle.

§7. The concept of discourse pattern is very close to the concept of “register,” a language variation depending on the communicative situation. [6] Halliday 1978 argues that registers mainly differ semantically; their lexicogrammatical differences derive from these semantic differences. [7] Similarly, Agha 2001:212 defines a register as “a linguistic repertoire that is associated, culture internally, with particular social practices and with persons who engage in such practices.” A main method developed in register studies, which are part of sociolinguistics, is to study the co-occurrence patterns of linguistic features in different texts. [8] Co-occurrence tendencies tell us something about the functions of the counted linguistic features, for “co-occurrence reflects shared function.” [9]

§8. Furthermore, register studies provide evidence for different dimensions of variation in order to interpret the co-occurrence patterns. These dimensions have both linguistic and functional content: they are sets of co-occurring linguistic features that can be tied to a certain communicative aspect, such as the degree of formality or narrativity. [10] For example, a high frequency of first- and second-person pronouns, imperatives, questions, and a low frequency of nouns typically co-occur in certain English registers, such as informal conversation. These linguistic characteristics belong to the same dimension of variation; together they indicate high involvement of the speaker, and little time for speech production. Other dimensions concern, for example, narrative versus non-narrative discourse, or the presence of an impersonal style.

§9. The terminology used in register studies is not always clearly defined. In particular, researchers use the term “register” to refer to extralinguistic, situational characteristics, or to linguistic co-occurrence patterns, or to the combination of both. In other words, it is ambiguous whether form, meaning, or both are included in “register.” As Lee 2001 and Sampson 1997 point out, there is much terminological confusion in the scholarly literature about terms such as “genre,” “register,” and “text type”: different authors use these terms in a different way, without clearly distinguishing between them. [11] Lee 2001 attempts a clarification: he notes that “genre” is sometimes set in opposition to “text type,” with “genre” referring to criteria external to the text (such as the audience and nonverbal activity), and “text type” referring to internal criteria (such as lexical and grammatical co-occurring features). Additionally, he observes, “genre” tends to be associated with social purposes of language, while “register” tends to be associated with the situation or immediate context. Despite Lee’s efforts to clarify, however, “genre,” “register,” and “text type” are often used interchangeably. Östman’s term “discourse pattern” is problematic as well, since he uses the poorly defined terms “text type” and “genre” to define it. However, it is clear that “discourse pattern” includes both form and meaning. Because of this clarity, I employ the term “discourse pattern” rather than “register” in this chapter. [12]

§10. Linguistic variation in Greek drama has been studied from different perspectives using different terminology. For example, Earp discusses several aspects of style in two monographs, one written in 1944 on Sophocles, and the other in 1948 on Aeschylus. [13] The major stylistic difference he identifies is between “elevated” and “colloquial” style. [14] Earp describes several factors that influence differences in frequency and use of associated linguistic features. One factor is time: because Aeschylus and Sophocles developed their styles over the course of their careers, fewer ornamental epithets appear in the later Aeschylean plays than in the earlier ones, and similarly the use of amplification declines in Sophocles. Two other factors influencing stylistic difference are a play’s subject matter and the purpose of each individual scene. Earp finds that “intense emotion leaves little room for conventional ornament” in Aeschylus (1948:57), and that Sophocles uses antithesis mainly in argumentative contexts (1944:95). [15] Perhaps the most important factor is the difference between lyric parts and what Earp calls “dialogue,” i.e. all non-sung parts. “Ornament” and other “elevated” features are more frequent in lyric, “for the same mood which impels us to sing impels us also to use heightened language” (1948:78). Even though Earp does not pay attention to particle use, here and in many other places he shows great sensitivity to the influence of external factors, such as communicative situation and a speaker’s goals, upon language variation. [16]

§11. Other scholars note similar linguistic tendencies, though without analyzing as many details as Earp does. Examples are Ruijgh 1971:988-989, H.J. Dik 2007:6, and Rutherford 2010:441, 443-444, who draw attention to the linguistic differences between lyric and iambic parts of tragedies. According to Ruijgh, the non-sung parts are mainly based on contemporary Attic, whereas the sung parts are strongly modeled on choral lyric, usually from different dialects. Baechle 2007 provides metrical and prosodic evidence for the similar claim that “the tragedians had a very highly developed sense of what style was appropriate to dialogue, as opposed to tragic lyric” (4). Rutherford 2010 juxtaposes a lyric passage, an excerpt from a long rhesis, and a stichomythic passage, in order to illustrate the main “modes” (448) of tragic language. Barlow 1971, a study on imagery in Euripides, emphasizes different distinctions: this author includes separate discussions of (1) choral odes, (2) monody and lyric dialogue, (3) messenger speeches, and (4) rheseis and iambic dialogues. She notes for example that the two lyric “modes” employ an “elevated tone” (43).

§12. With a different perspective, Sideras 1971 demonstrates that Aeschylus employs many Homerisms, in lexicon, syntax, morphology, word order, and rhetorical figures. [17] Sideras does not mention particle use, but we will see in this chapter that also in that domain Aeschylus resembles Homer more than the other tragedians. Stevens 1976, furthermore, discusses “colloquialisms” in Euripides, using other genres, mainly Aristophanes, as evidence that certain expressions, including specific particle constructions, were felt as “colloquial.” [18] With this term he means less suited for “poetic,” “prosaic,” or “neutral” language (2). However, as he admits (64), most of these expressions are not frequent enough to provide statistically significant information about their distributions, and the effect of different expressions varies. On top of that, any classification of certain features as colloquial involves a certain degree of subjectivity, as e.g. Collard 2005:358 points out.

§13. Concerning Aristophanes, likewise, Dover 1972 notes the linguistic differences between lyric and iambic parts. Lyric passages share vocabulary with “the serious lyrics of tragedy,” but are still more similar to comic dialogue, even though “the expression is more concentrated” (71). [19] In other words, we may infer, lyric parts are generally more concerned with what is said, and how, than iambic parts, and less with who is saying it to whom. Comedy as a whole, Dover argues, “combines all the registers of Greek utterance which are known to us: at one extreme a solemnity evocative of heroic warfare and gorgeous processionals, at the other a vulgarity inadmissible in polite intercourse” (72).

§14. Willi’s exploration of “registers” in classical Greek (2010a) is an important precedent to the current study. [20] The author builds upon register studies in order to analyze linguistic differences across Greek (parts of) texts in a detailed manner. He undertakes a sample study of three registers: that of forensic oratory, that of historiography, and that of (the approximation of) casual conversation. He gives the number of occurrences of 23 linguistic features in six roughly contemporaneous text samples of 1,000 words each. [21] Several of these features indeed seem to be influenced by the register that is appropriate to the text in question. For example, nouns are more common in historiography than in the other genres, whereas the inverse distribution is found for first-person verbs. [22]

§15. The underlying assumption of all the preceding scholarship is that situational differences within plays are reflected in linguistic differences. Singing a song, for example, constitutes a different communicative situation in drama than having a dialogue: the meter, music, speakers, length of the turns, and communicative goals are different. The assumption that these differences influence language also informs the current work.

§16. This chapter builds on Willi’s study and the other research described above, focusing on the main discourse patterns of Greek drama, and specifically using the results as input for the interpretation of particles. Because of the specific focus on particle use, information about the distributions of 14 individual particles is taken into account. In scope and detail, my study fills a gap in previous work, as Earp does not discuss particle use at all, and Willi mentions particles only as a group, without considering the distribution of individual items. [23] I also refine earlier work by distinguishing three different discourse patterns, rather than only a iambic and lyric part. [24]

§17. The frequencies of 25 linguistic features have been collected in three different communicative formats within tragedy and comedy: (1) iambic dialogues with short turns of speaking, (2) iambic monologues, and (3) lyric choral songs. These settings constitute the bulk of tragedy and comedy, and have a clearly distinct linguistic shape. Delving into the three discourse patterns associated with these communicative formats therefore forms a natural starting point for this type of investigation in drama. Indirectly, my study may shed light on the linguistic tendencies of rarer formats as well, such as anapaests and lyric parts sung by individual characters, which I leave out of consideration here. For my data collection, turns in the dialogue patterns are no more than 4 lines long; monologues are at least 15 lines long in Aristophanes, and at least 25 lines long in tragedy. The corpus used for this collection is made up of a collection of passages, rather than entire plays. [25] The passages have the following total size:

§18. The account includes a selection of 14 different particles: ἀλλά, ἄρα/ἆρα, [27] γάρ, γε, [28] δέ, δή, δῆτα, ἦ, καί, [29] μέν, μέντοι, μήν, οὖν, [30] and τε. Other linguistic features selected are nouns, participles, finite verbs, imperfects, present indicatives, future finite verbs, finite passives, first-person references, second-person references, [31] negations, and swearing expressions. [32] These can be compared to features of modern languages that are analyzed in register studies. [33]

§19. Note that it is not my goal to describe all characteristics of the three discourse patterns. Differences in lexical semantics, for example, are not taken into account here, despite their relevance for situationally-motivated linguistic variation; these differences involve subjective judgments, and are therefore harder to quantify. [34] The current study uses the frequencies of a limited number of linguistic features; [35] my objective is to use these distributions in order to understand particle use better. The features chosen have a relatively clear general function, and can be compared to modern-language features that have been analyzed in register studies. For example, Biber 1995:94-104 mentions nouns, past tense (markers), present tense (markers), passives, first- and second-person pronouns, and negations as among the features that distinguish registers in English, Tuvaluan, Korean, and Somali. I use these comparisons from modern languages on the assumption that situational variation functions similarly across languages. That is, I assume that the use of, for example, first-person references or negations in ancient Greek drama is so similar to their use in different modern languages that a high or low frequency of these features reflect similar fundamental aspects of verbal communication, such as a high level of interactiveness, as in modern-language registers that have been investigated.

§20. Biber 1995 also includes English discourse particles and Tuvaluan discourse linkers (in his terminology) in his register analysis, but only as a group. In his comparison of registers in ancient Greek, Willi 2010a similarly reports the combined frequency of 21 particles (see note 22 above). In my analysis, on the contrary, the distributions of each particle have been collected separately, which has allowed me to perceive distributional differences across particles. [36]

§21. To test the statistical significance of the attested distributions, I used the software of Preacher 2001 to calculate a chi-squared (χ²) test of independence. [37] A chi-squared test calculates the probability that a certain attested distribution of a feature is found in samples from groups with the same average frequency of that feature, in which case that distribution would not be meaningful (“significant”). In our case, the “groups” are texts, that is, groups of words. The features are particles and other linguistic items with a certain number of occurrences in the different samples. An attested distribution yields a chi-squared value; from this value Preacher’s software calculates the probability, called p-value, that the given distribution would be attested in two random samples from the same text. If the p-value is very low (it may range between 0 and 1), then this attestation is very unlikely, and we may assume that the text A and B in fact do differ in their overall frequency of the feature in question. Normally a distribution with a p-value below 0.05 is considered statistically significant; this means that there is a less than 5% chance that the distribution is due to chance. For example, if a sample of 1,000 words from text A contains 22 δέ instances, and a sample of the same size from text B contains 45 δέ instances, the chi-squared value is 8.169; the p-value is 0.00426 (or 0.426%). This is smaller than 0.05 (or 5%), so the distribution is significant. Once significance has been established, we may then turn to explaining the difference in frequency of δέ between text A and B.

§22. In this main part of the chapter I discuss particle distributions across the three discourse patterns, and use this information as input for interpreting the particles’ functions. If several playwrights consistently choose a certain particle more often in one of the discourse patterns, this suggests that that particle is associated with some communicative purposes or circumstances of the situation related to that discourse pattern. For example, a consistently higher frequency in dialogues may reflect the particle’s association with more interactive speech. If no such consistent frequency differences can be detected for a certain particle, then this particle has several functions, each of which is associated with different communicative aspects, or its functions are equally relevant to all situations. In either case, tendencies on a macro-level can be illuminating.

§23. The analyses focus on different particles in order of their average frequency, starting from the most frequent ones. I will only discuss the eleven particles that have a statistically significant distribution in at least two out of four authors. [38]

§24. The most frequent particles overall are δέ and καί. [39] The distribution of δέ is as follows: Aeschylus uses δέ especially often in choral songs, such as the one in Agamemnon 367-474: The several instances of δέ in this song mark boundaries between discourse acts, short communicative steps that may or may not coincide with a clause. [43] The particle has a relatively neutral function, signaling that a new step in the discourse has begun. The step can correspond to anything: a new event in a narrative, an argumentative point, a vocative, a contrastive noun phrase, an apposition, and so on. In this passage, for example, the act φθονερὸν δ’ ὑπ’ ἄλγος ἕρπει προδίκοις Ἀτρείδαις (450-451) describes one aspect of the scene “Greeks grieving over men lost in Troy” that the singing elders are depicting. οἱ δ’ in 452 constitutes a new step, signaling a switch from the grieving family at home to the deceased warriors themselves. [44] The next δέ, in 455, introduces a new act again, adding another facet to the “deceased warriors at Troy” picture. In 456 δέ starts a gnomic act, [45] and simultaneously accompanies the start of a new antistrophe, a larger boundary than just a new discourse act. In fact, throughout this whole first stasimon (367–474) every strophe and antistrophe except for the very first strophe starts with a δέ. The parallels show that this is a typically Aeschylean phenomenon. [46] Admittedly, apart from this pragmatic boundary-marking function, the δέ instances in this passage work to prevent hiatus. However, other words could have been used for that as well, so its metrical function cannot have been the only justification for the use of δέ. In other words, the particle’s non-random distribution proves that it cannot have been mainly a metrical tool.

§25. The relatively high frequency of δέ in all three communicative settings in all four authors can be connected to the particle’s relatively neutral, “minimal” function, which makes it compatible with many different contexts and co-texts. The especially frequent use in Aeschylean choral songs makes their discourse appear less explicitly subjective, since the nature of the connections between most discourse acts is not spelled out. That is, the singers do not make it explicit how exactly they consider the acts to be related. [47] The “neutral” presentation fits the specific context at this moment of the Agamemnon: as Raeburn and Thomas 2011 point out in their commentary on this stasimon, the song contains many “shifts of thought,” but the idea of retribution unites the different topics. From the other instances of punishment in the song, Raeburn and Thomas note, the chorus can draw the obvious conclusion about Agamemnon’s upcoming fate, “but dare not voice it.” The audience, of course well aware of what will happen to the king, will understand what the song is hinting at. [48] In this context with emphasis on implicit meaning, the minimal boundary signals conveyed by the δέ instances are appropriate.

§26. The high frequency of δέ in Aeschylean songs may at the same time lend these passages an epic air, since δέ is very frequent especially in Homeric narrator text. [49] The songs share with epic a concern with famous, traditional stories and a general tendency to avoid explicitly encoding subjectivity. The same holds true for non-dramatic lyric, where δέ is similarly frequent. [50] The allusion to epic and lyric helps establish Aeschylean songs qua songs, that is, underscore their genre affiliations, as well as endow them with an authoritative voice, typically associated with these genres.

§27. Aeschylus uses δέ especially often in songs, but in monologues too he employs it more frequently than the other dramatists. As for Euripides, he uses δέ most in his monologues. [51] Consider the following passage from a messenger speech, which describes how Dionysus bent down a fir tree: Again, δέ marks the boundaries of discourse acts, which in this context are part of narrative steps. [52] Each narrative step in this excerpt is in fact constituted by several discourse acts; because of its neutral function, δέ may also appear at the boundaries of such moves (multi-act units). [53] First the messenger describes the bending of the fir tree (1066-1069). Next (δέ 1070) Dionysus places Pentheus on top of the tree and sets it straight again. Subsequently, the acts starting with δέ in 1073 describe what happened to the tree as a consequence. In 1075, finally, the particle introduces the next narrative event: the bacchants see Pentheus on the tree. This next step is in fact the climax of the story, as reflected in the change from imperfects to aorist (ὤφθη, 1075). δέ thus helps to move the story forward from one event to the next, marking discrete steps in continuous discourse, in a manner similar to its use in Homeric or Pindaric narrative.

§28. Although commentators do not remark on the high frequency of δέ in Euripidean speeches, they often note that messenger speeches in general resemble epic language. For example, Page 1938 ad Euripides Medea 1141 explains that messenger speeches are modeled upon epic. [54] Similarly, Palmer 1980 notes that in tragedy, “Homerisms are particularly frequent in messenger speeches” (133). J. Barrett, in his 2002 study of messenger speeches, argues that “the messengers’ narrative voice typically resembles that of epic” (xvi). In such epic environments, then, a higher frequency of δέ is also appropriate. [55]

§29. In the next paragraphs I will discuss the distribution of other linguistic features than particles, in order to establish an interpretive link between these other features and the distribution of δέ. Ι will focus on imperfect forms, present indicatives, and participles. Imperfects are relatively frequent in monologues in all four authors. In (t2), for example, we find κυκλοῦτο (1066), ἔκαμπτεν (1069), μεθίει (1071; this form could also be present tense), and ἐστηρίζετο (1073). This high frequency of imperfects can be connected to a frequent occurrence of narratives in monologues. The distribution of imperfects across the discourse patterns is as follows: On imperfects, Rijksbaron writes that it is “crucially connected with historical narrative” (1988:254): the imperfect “unequivocally locates a state of affairs in the past.” Likewise, Bakker 1997e:20-21 considers the imperfect, but not the aorist, to be a true past tense. Rijksbaron observes (248) that in Herodotus, the imperfect is much more frequent in narrative than in direct speech. [56] In all four Greek dramatists, the imperfect’s distribution is in line with these observations: it is most common in monologues, the best setting for telling a story. [57] This distribution indirectly confirms the affinity of δέ for narrative contexts.

§30. The distribution of present indicatives is useful in so far as we can interpret it in light of general, cross-linguistic functions of these forms and of other tense-mood combinations, that is, bound to characteristics of ancient Greek as well as other languages. Sophocles uses present indicatives most in choral songs, Euripides most in dialogues: In dialogues, the present tense tends to be communicatively appropriate because of the implied attention to the current speech situation. [58] In choral songs, conversely, the present tense fits references to general or timeless states. I relate the lower frequency of present indicatives in monologues to analogously general tendencies: this communicative setting is especially used in Classical drama for narratives about the past, and for argumentative purposes. Both scenarios typically require other verb forms besides present indicatives. The distribution of present indicatives, then, on a general level confirms my claim that the communicative situation determines linguistic choices, and more specifically strengthens the connection that I have drawn between monologues and narrative purposes. This indirectly supports my interpretation of δέ’s distribution as reflecting, among other things, an affinity for narrative contexts.

§31. The distribution of participles provides input for our interpretation of δέ as well. This distribution is similar to that of δέ: participles are most frequent in Aeschylean choral songs and Euripidean monologues: In the register study by Willi 2010a, participles are found to be more frequent in oratory and historiography, and less frequent in texts that represent conversation. More participles, he observes, seem to be related to a greater average sentence length. Indeed, participles help to segment the discourse into small chunks, just as δέ, but they achieve this syntactically; [59] we can imagine that such syntactic chunking would be especially helpful in a long, complex turn of speaking. [60]

§32. To sum up the findings of this subsection, δέ is very frequent (compared to other particles) throughout the discourse patterns in all authors, and especially so in Aeschylean choral songs. In this communicative environment, I have argued, the high frequency of δέ reflects a seemingly neutral presentation of the discourse in which the speaking “I” does not come to the fore, and evokes a Homeric style. Euripides’ preference for the particle in monologues may be explained by δέ’s affinity with narratives, and from the well-known link between messenger speeches and epic story telling. Other linguistic characteristics of the discourse patterns, such as the imperfect and present tenses and participles, also reflect such pragmatic associations in their distributions.

§33. The other frequently occurring particle, καί, shows a striking distribution across the three discourse patterns. Aeschylus uses καί more often in dialogues, but Sophocles and Euripides prefer to use it in monologues. Aristophanes, however, uses it more often in songs: The distribution of καί, then, reflects differences in particle use across the playwrights. In fact καί has multiple functions in all authors, but each author tends to exploit this variety in different contexts.

§34. Aeschylean dialogues are a case in point: they contain various καί constructions. Consider the following excerpt, from a scene in which the messenger is telling the Persian queen about the army’s defeat: Each of the three instances of καί in this passage is part of a different construction. The first has a small scope and marks a close connection, together with τε, [61] between Πέρσαις, “the Persians,” and πρόπαντι βαρβάρων γένει, “the whole race of the barbarians.” The two items are semantically and morphologically similar, so there can be no doubt about the function of the particle combination. The second instance of καί, in 437, is not surrounded by two items that could be connected. It must be interpreted as pinning down and highlighting the adverb δίς: “even twice” or “really twice.” [62] The third instance is found in turn-initial position at the start of a question, and has a large scope over that entire question. In this construction the use of καί may imply the speaker’s surprise. [63] All of these specific functions interact with the dialogic situation at hand. The close connection marked by τε καί in 434 indicates that the two interlocutors share certain knowledge (see the subsection on τε below), drawing them closely into the interaction. The “even twice” highlight in 437 fits the messenger’s current dialogic task: to announce a further shocking narrative in addition to his earlier speeches. Finally, the implied surprise in 438 relates to the high interactiveness of this situation in which the speakers immediately react to each other after short utterances. [64] In other words, the larger context of the discourse pattern throws light on the interpretation of καί’s local functions that depend on the direct co-text.

§35. In Sophocles we find the highest frequency of this particle in monologues. [65] The following passage from a speech by Tecmessa, relating Ajax’ actions in his madness, illustrates this tendency: Again, we see the multifunctionality of the particle. The different καί instances have different scopes: the first three introduce whole clauses, the fourth one only one verb form. According to Stanford 1963 ad line 294, this καί indicates a consequence, and at the same time emphasizes the pronoun ἐγώ. The “consequence” Stanford refers to is a pragmatic enrichment of the connection marked by καί: “and therefore”; we can infer this enrichment from the co-text. [66] Jebb 2004 [1896] translates this καί with “and.” The instances in 295 and 298 (both untranslated by Jebb) each form a fronted discourse act on their own—we can infer an act boundary from the μέν instances in the following acts. These two καί instances each project an upcoming multi-act move, encompassing a μέν act, a δέ act, and more. [67] Each καί connects the upcoming move as a whole to the preceding discourse. [68] Finally, καί in 299 connects two items that are morphologically and semantically similar, just as we had seen for τε καί in the Persians example. [69] Jebb’s choice of “or” to translate this καί reflects that the particle here may work differently from the other three instances. [70] In the καί constructions in this monologue, then, the particle has slightly different pragmatic functions. These functions relate less to interactiveness, as we found in the dialogic example, and more to a speaker taking her time to formulate her utterance.

§36. In some cases, mainly found in monologues, the two conjuncts connected by καί are so close in meaning that the second one can be interpreted as a specification or reformulation of the first one. καί then receives the enrichment “that is,” “better to say,” “in other words,” or “to be more precise.” Here is an example of this construction from a monologue by Pentheus: ἀγχόνης ἄξια (“worthy of death by hanging”) can be considered a specification of the first conjunct (“terrible”). This use of καί is described by Hartung 1832, who speaks of “eine nähere Bestimmung des Vorangehenden” (145), and by Humbert 1960: καί marks a “meilleure approximation” (412). [72] All four playwrights employ this construction: other examples are κατεῖχε κἀπράυνεν (“he tried to restrain and, that is to say, to calm them,” Aeschylus Persians 190, in a monologue), ὑπερμαχοῦμαι κἀπὶ πάντ’ ἀφίξομαι (“I will fight and, that is, go to all lengths,” Sophocles Oedipus King 265, in a monologue), and ἐχθρῶν κοὐ φίλων (“from enemies, I mean, not from friends,” Aristophanes Birds 378, in an utterance of six lines). [73] This use of καί shows again how a specific co-text leads to a certain enrichment of the particle’s force. Just as καί may mark a large-scope “zooming in” at the start of questions, likewise it may specify, with smaller scope, a noun phrase or verb phrase when it connects two semantically similar items. The tendency for this function to occur mainly in monologues can be connected to a general pragmatic need of monologues, which is to hold the floor: only when a speaker has ample time to put her message into words, she can “afford” to describe a certain action or concept in two slightly different ways. In addition, using such specifications may support an argumentative goal, which speakers often pursue in uttering a long speech. [74]

§37. Unlike the tragedians, Aristophanes exploits the multifunctionality of καί most in choral songs. The following example from Lysistrata contains many instances; the singers are the united choruses of both men and women, who have finally decided to make peace. In καὶ λέγειν καὶ δρᾶν (1046-1047), the repetition of the particle emphasizes the addition of the two items: “both to say and to do.” The καί in τὰ κακὰ καὶ τὰ παρακείμενα (1047) has a small scope, like the first two instances, but in this case the element before it and the one after it refer to the same entity: the current troubles. That is why the translation “and” is less appropriate here; Henderson 1987 ad loc. calls this καί “emphatic.” In fact, its function is not really distinct from a καί that we do translate as “and”: it binds the two aspects of the troubles closely together, creating a hendiadys. [75] At the same time, it contributes to the κ-alliteration here. The instance in 1049 (ἀνὴρ καὶ γυνή) has a small-scope connecting force; those in 1052, 1053, 1061, 1062, and 1063 have a larger scope, since they connect entire verb phrases or clauses. The καί in 1052 may trigger an enrichment as “even,” as Henderson’s translation shows. The instance in 1060, finally, is found in the fixed phrase καλούς τε κἀγαθούς, another hendiadys as in 1047; also καί in ἁπαλὰ καὶ καλά (1064) works in a similar way. All in all, the particle binds together discourse segments of different size and nature. In this specific song, the staggering number of καί instances may iconically underline that the two semi-choruses are now united. As Wilamowitz 1927 ad loc. remarks, the male and female chorus members possibly form pairs while singing this song. This would be especially apt, we can imagine, for the phrase ἀνὴρ καὶ γυνή in 1049. Thus, the local functions of these καί instances are either appropriate to any communicative situation (1049, 1052, 1053, 1061, 1062, 1063) or reflect the general poetic attention—high in choral language—to how things are said (1046, 1047 (twice), 1060, 1064). [76] The particle’s repetition is especially related to the song’s overarching pragmatic goals, namely to underline and celebrate the union of the two semi-choruses.

§38. The above passages show that καί participates in several different but related constructions, and that these constructions tend to occur mostly in one of the discourse patterns. [77] Aeschylus’ preference for using καί especially in dialogues may additionally reveal influence of Homeric particle use: in Homer καί is more frequent in direct speech than in narrator text. [78] Sophocles and Euripides employ a broad range of καί constructions especially in monologues, where some of its functions are related to holding the floor for an extended period of time. Aristophanes uses καί mostly in choral songs. The multifunctionality of this particle seems to make it especially suitable for exaggerated repetition, a strategy that fits the mocking tone of many comic songs.

§39. Like καί, τε marks connections, but it differs in its implications, possible enrichments, and specific constructions. [79] τε also has a distribution unlike that of any other particle. The use called “epic,” discussed in IV.2 §§47-69, does not occur in tragedy and comedy, except for the idiomatic construction οἷός τε “able to”: all other instances of τε here have a connective function. [80] However, that does not mean that these instances in drama are in no way related to the “epic” ones. The distributions of τε prove that it does more than merely connect; after all, the action of connecting items is not tied to a particular communicative situation, and τε in fact displays relatively large frequency differences across the three discourse patterns. Most authors prefer to use it in choral songs: This distribution can be explained from several related associations of the particle. First and most fundamentally, τε marks certain knowledge as shared between speaker and addressee. In arguing this I follow the descriptions of τε in Wentzel 1847:2, Bloch 1955:147, and Gonda 1954:207, and build on this monograph’s analyses of τε in epic, lyric, and historiography. [81] Second, what is shared is often traditional or even part of a traditional performance or ritual: τε may imply a link to tradition and rituality as well. Third, a high frequency of the particle may trigger an allusion to epic or lyric, genres where τε is naturally frequent because of the importance of traditional knowledge. Fourth, because of these associations, τε can convey a generally solemn and formal tone. I will discuss each of these implications in turn.

§40. First consider the many τε instances in this excerpt from a speech by the messenger in Aeschylus Persians, shortly after he brought the news of the Persian defeat: The messenger here sums up the names of commanders who died in the recent battle, connecting them with τε. These names are well-known to the speaker and his addressee, the queen (as well as possibly the audience). That is, they represent shared knowledge. [82] At the same time, the high frequency of τε in this speech (as many as 10 instances out of 147 words total) lends a Homeric note to the messenger’s voice. [83] As J. Barrett 2002 shows, tragic messengers in general, and this one in Persians in particular (23-54), are consistently portrayed as resembling an epic storyteller. [84]

§41. The prototypical kind of shared knowledge is the knowledge associated with traditions and rituals. Since the singing of choral songs is a traditional, ritual activity, the use of τε is especially suitable for this environment, regardless of the specific elements that the particle connects. [85] This is a further connection, then, between a particle’s local function, depending on the immediate co-text, and a co-existing global function, depending on the discourse pattern. An example of a lyric passage with many instances is the following part of a Euripidean song: At this point in the play it is becoming clear how disastrous it is that Neoptolemus has “a double marriage,” as the chorus calls it: his jealous wife Hermione is threatening to kill his concubine Andromache and her child. In the song the chorus expands on the topic of disastrous rivalry, generalizing it to other spheres of private and public life. As Stevens 1971 and Lloyd 1994 ad loc. point out, this song fits into “a common pattern in Greek lyric” (Lloyd) through generalization, development, and then applying the general themes to the current situation. Indeed, the τε instances in 476 and 481 are found in gnômai. [86] In general the traditional context is a fitting environment for the particle. In 475 it is combined with καί, closely connecting two noun phrases; in 476 τε has a larger scope, introducing a whole clause; [87] in 581 it has a small scope again. All these instances of τε, then, have a connecting function, but the particle’s distribution over the discourse patterns shows that there is more to the use of τε than this local function. More globally, the many τε occurrences here strengthen the ode’s link to tradition. [88]

§42. There are only five tragic songs without τε in my corpus (which contains 55 tragic choral songs in total). [89] These songs happen to be exceptional in other ways as well: they contain, for example, more first- or second-person references than the average choral ode, and fewer nouns. Their style is in certain respects closer to that of the average dialogue than to that of the average song. In terms of content, they are all directly tied to the immediate context of the play, and do not take the audience out of the ongoing story as many other songs do. [90] One of them is the parodos by the Salaminian sailors in Sophocles Ajax: The song as a whole (lines 172-200) contains 12 references to a “you” (about 10% of all 126 words), a much higher frequency than the average of about 5% for Sophoclean songs. [91] Indeed, three out of four songs in Ajax have a higher frequency of references to either first or second person. [92] These chorus members, in other words, advance themselves as communicators more than is usual for a tragic chorus. This is also reflected in the occurrence of γε in two of the songs, such as here in line 182—this particle is normally absent from tragic choral songs. [93] Moreover, these linguistic features accompany atypical content: the songs concern Ajax himself, or the influence of his troubles on the lives of the chorus members, rather than general or timeless considerations. [94] In different ways, then, the singing sailors show their personal involvement in what they are singing about, [95] that is, the fate of their leader Ajax, and their emotional nearness to him. This fits well with their identity as his followers and friends. [96] Such a chorus of Greek men of military age was unusual in tragedy. [97] In this case, Sophocles has matched the unusual choral identity with unusual language use. The chorus’ characterization is therefore strengthened by the linguistic choices in the odes.

§43. τε tends to appear also in ritual contexts other than choral lyric, such as prayers, prophecies, or the swearing of an oath. This includes instances in dialogues, even though the particle is less frequent there than in monologues and choral songs. In tragic dialogues the τε frequency is significantly higher than in comic dialogues. Consider the following instances: Mastronarde 2002 ad loc. notes that the “generalizing formula” in Medea’s list of gods is typical of “ritual contexts.” Similarly, Page 1938 comments that Euripides “is using conventional language” in 747, and Mossman 2011 observes “the solemnity of the ritual.” A solemn tone fits the situation, because Medea’s life might later depend on the oath that she is now proposing. The double τε is repeated by Aegeus when he indeed swears the oath in lines 752-753. In addition to emphasizing the ritual activity of swearing an oath, τε here marks the entities it connects (the Earth, the Sun, and all gods) as belonging to shared encyclopedic knowledge. [98]

§44. Besides these associations with shared knowledge, tradition, and rituality, a high τε frequency may remind the audience of epic, of non-dramatic lyric, and, in Aristophanes, of tragic lyric. [99] The allusion to other genres is especially apparent in Aristophanic choral songs, where the average frequency of τε is more than eleven times as high as in dialogues. Consider the following example: Aristophanes’ overuse of the particle in this choral song (six instances in 56 words in total) mockingly mirrors τε’s traditional implications, and at the same time parodies epic and tragic lyric. Echoes from epic are also apparent in the lexical choices: words like ὑψερεφεῖς (“high-roofed,” 306) and εὐστέφανοι (“well- garlanded,” 308) are taken from epic vocabulary. The clouds, who form the chorus here, are thus presented as exaggeratedly solemn and divine creatures. [100] τε contributes to this image, beyond its local connecting function in the specific instances. [101]

§45. This example of parody also illustrates the last, related association of τε in my corpus: the particle’s connection to a generally formal or solemn tone. Not only τε’s frequent occurrence in contexts of shared knowledge, tradition, and rituality suggests this formal or solemn implication, but also the distribution of other linguistic features over the discourse patterns, such as nouns and finite verbs. The distributions of these features are illuminating, because they are connected to levels of formality in modern-language registers. For example, in the four languages analyzed in Biber 1995 (English, Nukulaelae Tuvaluan, Korean, and Somali), a higher frequency of nouns is associated with discourse that is less involved, less interactive, and has a less overt expression of personal stance than other texts. [102] This means, for example, that English academic prose, a relatively formal register, has a noun-verb ratio twice as high as spoken conversation, which is relatively informal. [103] Biber, Conrad, and Reppen 1998:69 explain that nouns typically convey something more abstract and more information-focused than verbs. Verbs, in contrast, refer to actions, and also often to thoughts and feelings of the speaker. [104] This makes verbs more suitable for highly interactive situations. [105]

§46. The distributions of nouns and finite verbs over the discourse patterns in my corpus are as follows: Dialogues tend to have a higher frequency of finite verbs and a lower frequency of nouns than monologues and choral songs. Stichomythic conventions may play a role in the high frequency of verbs in dialogues: the tight schema of turns ensures that a speaker has only a short time to utter her turn. Since a finite verb is often needed for that, this increases the frequency of verbs in such an environment. [106] This conventional “pressure” may mirror a similar situation for speakers in face-to-face conversation, to which dramatic dialogues are closer than monologues or songs. [107] More generally, a lower verb frequency and a higher noun frequency are found in formal, less interactive, and less explicitly involved discourse in the modern languages analyzed in register studies. [108] The language used in tragic and comic monologues, as compared to that used in dialogues, thus reflects these situational characteristics, and choral lyric language even more so. Not only nouns and verbs indicate these correlations, but also the distributions of first- and second-person references, negations, and swearing expressions. [109]

§47. In formal or solemn communicative environments, we find a corresponding high frequency of τε. Consider the following passage from an Aeschylean monologue, the environment in which this author uses τε most: Electra is here performing a prayer to Hermes and the dead Agamemnon: a solemn tone can be expected in such a context. Praying is also a ritual activity; the solemn-tone association of τε in this case overlaps with this other implication. The entire speech (124-151) contains 46 nouns (26.6% of its 173 words), 17 finite verbs (a low frequency of 9.8%), and as many as seven τε instances (4.0%). [112] Together these features reflect a formal style, appropriate to the praying activity. The two τε in 124a (=165) and 128 signify a tight connection of their conjuncts as well as an allusion to traditional knowledge. In 130-131, the first two instances establish a close link between the elements “me” and “Orestes”; the τε after φῶς [113] has a larger scope, adding a new verb phrase. [114]

§48. Choral songs exceed the spoken parts in their formal and ritual nature. In addition to all the features mentioned, other characteristics of lyric language, such as more semantically obscure words and Doric coloring, contribute to the defamiliarizing style of choral songs, since they remove the language even more from daily spoken Attic than the general stylization throughout tragedy already does. [115] The characteristics of the songs’ melodic patterns and accompanying dances will have played a role in the overall defamiliarization of this communicative situation as well. [116]

§49. The distributions of τε as well as of other linguistic features over the three discourse patterns make it apparent that this particle contributes more to the discourse than only its local connecting force. I associate τε with several implications deriving from their performative context: shared knowledge, tradition, rituality, an allusion to epic, a parody of tragic lyric, and a generally solemn or official tone.

§50. The distribution of γάρ in my corpus is as follows: In Aeschylus and Euripides γάρ is roughly equally distributed across the three discourse patterns, whereas in Sophocles and Aristophanes it is used more frequently in dialogues. [117] Compared to τε, γάρ has a more complex distribution pattern, a probable result of its use in several constructions, in which it has varying functions.

§51. γάρ in drama often marks its host act as the cause, explanation, or clarification of a preceding (or sometimes following) act. [118] This cause or explanation may relate to the other act in two ways, either alone or simultaneously: de re (concerning content) or de dicto (concerning the discourse acts). [119] An example of a content relation is “I did this because my master ordered me to”: the “because” clause introduces a state of affairs that plays a role in the realization of the main clause. Here is an instance of this kind of γάρ in Aeschylus: Now that Troy has been captured, the chorus has decided to address the gods appropriately. The γάρ act explains why the speakers are getting ready (παρασκευάζομαι) to sing the upcoming song. [120]

§52. As represented in the above example, γάρ typically expresses a de re relation (signaling a cause on the level of content) when a speaker has just referred to her own actions or feelings, a natural consequence of the fact that one normally knows the reason(s) for one’s own actions or feelings. [121] When the discourse concerns agents other than the speaker herself, it is more likely that γάρ signals a de dicto relation. For example: “what time is it?—because you have a watch.” Here the “because” clause explains why the speaker felt she could direct her question to this specific addressee. The following excerpt contains an instance of γάρ in this use: [122] In this case the particle does not refer to a causal relation in the reported world, but marks its host act as clarifying why the preceding interrogative act was uttered. The speaker can ask this question, he assumes, because the Hoopoe knows the answer.

§53. A γάρ act may also signal a de re and a de dicto relation at the same time. This dual signification regularly occurs when a speaker explains why she has just used a certain evaluative expression. On the de re level, the γάρ act explains why the evaluated object of thought really fits the given characterization; simultaneously, the γάρ act clarifies, de dicto, the speaker’s use of that evaluative expression in the preceding act. Here is an example from Euripides: [124] The γάρ act in 236 explains, de re, why it is the ἀγὼν μέγιστος (“the greatest issue”) for women whether their husband is good or bad: that is, women have no choice but to get married and stay with their husbands. At the same time, Medea explains why she has just uttered that qualification (“I said that because…”). [125] This reason may well contain another evaluative expression, [126] such as here οὐ εὐκλεεῖς (“not reputable”), which may subsequently trigger more γάρ acts clarifying the use of that assessment.

§54. Just as in Homer, Pindar, and historiography, γάρ in drama is also regularly found in gnomic contexts. [127] Since gnômai concern general matters, they represent a stepping out of the surrounding discourse: they are “unframed.” [128] γάρ does not necessarily imply any causality when it introduces unframed discourse (see II.4). A gnôme in drama may however imply a de dicto causal relation that can be paraphrased as “(I know that this is the case) because that is how things always go.” An example of γάρ introducing a gnôme occurs in this song by the frightened Danaids: As Wecklein 1902 and Friis Johansen and Whittle 1980 ad loc. point out, the idea that the dead are free from misery is proverbial in tragedy. γάρ marks this gnôme as backing up the maidens’ death wish. [130] Most other gnomic statements with γάρ are preceded by a discourse act with a third person. The speaker in those cases has to infer why a certain situation is or was as she perceives it, as she cannot know it from her own experience.

§55. At the beginning of utterances in dialogues γάρ usually does not signal any cause or explanation, whether de re or de dicto. Rather, it signals explicitly that the speaker is expanding on the preceding utterance, and that she infers something from it. For example, after Euelpides in Aristophanes Birds 109bis-110 has clarified that he and his friend are no jurors, but jurophobes from Athens, the Hoopoe asks σπείρεται γὰρ τοῦτ’ ἐκεῖ /τὸ σπέρμ’; (110bis-111), “(are you saying that) that seed sprouts there?” The Hoopoe, accepting the two men’s claim to be jurophobes, infers that jurophobes exist in Athens; the question with γάρ indicates his inference, and implies his surprise at the realization.

§56. The expansion or inference signaled by turn-initial γάρ often implies indignation or anger. [131] Examples are mainly found in Sophocles and Aristophanes, who use γάρ especially often in dialogues, such as in this altercation between Menelaus and Teucer: In 1125 Teucer expands on what Menelaus has just said, angrily implying that this reproach about his own “terrible temper” (θυμὸν ὡς δεινὸν, 1124) can in fact be turned into an argument for Teucer’s side: “yes, (it is right for me to have a terrible temper) for with justice it is allowed to be high-minded.” [132] The γάρ instances in questions, in 1126 and 1130, suggest that Menelaus has made an inference from the preceding utterance: “are you really saying that…?” [133] The γάρ in the answer in 1128 signals an expansion on the preceding question (“did you die and yet stay alive?”), as well as a de dicto explanation: “yes, I said that, because…” [134] The inferences that are communicated in all cases signal the speakers’ strong indignation. [135] γάρ thus indirectly contributes to the expression of this emotion.

§57. The above examples show that γάρ’s distribution may help to distinguish among several different uses of the particle. Sophocles and Aristophanes are fond of a use of γάρ in questions or answers, where the particle indicates that an inference has been drawn, which may arouse indignation in the speaker. This use is highly interactional and fits dialogues best. In contrast, the constructions in which γάρ marks a de re and/or de dicto cause or explanation occur in all three discourse patterns. Finally, the use of γάρ to introduce gnômai, which occurs less frequently than the causal uses, seems to occur especially in choral songs, where one of the general communicative goals is to recall and reinforce communal values. For the understanding of γάρ’s function in a certain tragic or comic passage, then, it is crucial to take into account the global pragmatic functions of the specific discourse pattern.

§58. γε and δῆτα have consistent and significant distributions across the three discourse patterns. I discuss these two particles together, because their distributions are very similar. The consistent distribution of these two particles suggests that their pragmatic functions are most suitable for the dialogue setting. Indeed, we will see below that they are connected to interactiveness. First I will discuss the distribution of several other features in order to strengthen my claim that linguistic choices reflect the degree of interactiveness of the different discourse patterns. Then I will focus on the particles themselves.

§59. We already saw in §§45-46 above that the distributions of nouns and finite verbs reflect that dialogues are less formal, but more interactive and explicitly involved, than the settings of monologues and choral songs. The distributions of first- and second-person references and of swearing expressions strengthen this impression as well. They all have their highest frequency in dialogues. [136] Scholars connect functionally similar features in modern languages to the speaker’s explicit involvement. First- and second-person references, in this count including personal and possessive pronouns as well as verb forms, make the speaker more present as a communicator. That is, explicitly mentioning an “I” or “you” emphasizes that an interaction involves not only a message, but also a speaker communicating it to an addressee. Biber, Conrad, and Reppen 1998:147 and Quaglio 2009:126 similarly find that English first- and second-person pronouns are related to interactiveness and involvement. Aristophanic swearing expressions, in the form of νή or μά followed by the name of one or more gods, display a similar distribution. Werres 1936:20-22 notes that swearing expressions often co-occur with first-person pronouns and the particle γε. They can to a certain extent be functionally compared to English expletives and taboo words, which are associated with informality and emotionally-loaded language. [137] Werres (11) describes Greek swearing expressions as colloquial expressions serving to emphasize utterances, or parts of them. The term “colloquial” in scholarly literature tends to refer to interactional discourse situations, mostly in drama (see §12 above). Regardless of whether the Aristophanic swearing expressions were actually common in fifth-century spoken Attic, then, they do seem to reflect the high interactiveness of a communicative situation.

§60. In short, several linguistic characteristics of tragic and comic dialogues suggest speakers who come to the foreground, and a strong focus on the ongoing interaction. Since γε and δῆτα are highly preferred in dialogues, we may conclude that they are functionally related to this explicit presence of the speaker and this interactiveness. In general, the particle γε can be compared to prosodic emphasis: it highlights a specific part of an utterance, and it often contrasts this part implicitly to something else. [138] This highlighting reflects the speaker’s view or attitudes, and implies that the γε utterance is reacting to certain previous discourse. [139] If speakers using γε react to the view or the utterance of others, the subjectivity and the sense of contrast implied by γε acquire an interactional character. This connection to interactiveness is what makes γε especially compatible with a dialogic environment. [140] Consider the following passage from an Aristophanic dialogue containing both γε and δῆτα: [141] The speakers, Xanthias and Dionysus, use several γε instances to single out certain elements of their short utterances. In 25bis γε qualifies the whole discourse act ὅς ὀχεῖ (“[you] who are riding”). By highlighting this piece of information, Dionysus implies that Xanthias’ claim to be bearing a load cannot hold because he is himself being carried. Xanthias in turn underlines his answer in 26 with γε—the luggage he is carrying (ταυτί) is obviously the most relevant part of the discussion to him. With the γε element in 28 he refers to his luggage again, [142] whereas Dionysus’ γε act in 29 repeats his point about the donkey. The highlighting in each case implies a contrast between the speaker’s view and that of his interlocutor.

§61. For an example of γε from a completely different context, consider this passage from Aeschylus Libation Bearers: γε is infrequent in monologues, and generally in Aeschylus: he uses γε less than the other authors in all three discourse patterns. These distributional facts warn us that the particle’s occurrence here is especially striking. γε appears in an emotionally highly involved, even agitated observation by Electra, a context that makes the speaker’s own experience come to the foreground. Electra had inferred that Orestes had sent, not personally brought, his lock of hair, which is the first sign she had found on her father’s tomb (lines 168-180). The sudden discovery of the footprints—καὶ μήν marks this observation as an unexpected appearance [143] —is a highly emotional moment for her. [144] She stresses the word στίβοι with γε, marking the discovered footprints as bearing special importance. [145] Again, we may compare γε to prosodic emphasis. [146] In this case, Electra does not react to someone else’s utterance, but rather to her own earlier remarks on the first sign of Orestes.

§62. As for the less frequent particle δῆτα, it is almost completely restricted to dialogues. [147] In the quoted passage from Aristophanes Frogs (line 28) δῆτα indicates Xanthias’ strong personal involvement with the item to which it is attached, in this case the negation. He emphasizes that the donkey is “absolutely not” carrying his luggage; later in the line he does so again with the swearing expression μὰ τὸν Δία. This denial is of emotional importance to Xanthias, as he is (apparently) suffering from his heavy load. [148] δῆτα always conveys such an emphatic signal of personal involvement when it occurs in answers or other reactions. [149] This function is linked to the alternation of speakers, as is δῆτα’s use in questions. [150]

§63. To recap: the distributions of γε and δῆτα have led me to establish a connection between these particles and the communicative activities that characterize dialogue in general. My interpretations are strengthened by the distributions of nouns, finite verbs, first- and second-person references, and swearing expressions. These features do not only imply that the speaker explicitly comes to the foreground in dialogues, but also enhance, in different ways, the interactiveness of the discourse. The particles γε and δῆτα achieve this by juxtaposing the views of different speakers (γε), and by marking the emotional charge of answers and questions (δῆτα).

§64. The particle ἀλλά shows the following distributions in the passages analyzed: Sophocles shows a significant preference for ἀλλά compared to the other two tragic poets throughout the three patterns. [151] He and Aeschylus use the particle most often in dialogues; Euripides most often in monologues.

§65. The general function of ἀλλά can be described as the substitution of one alternative with another, which can include the correction of an explicit element, an implicit element, and the switch to a different topic. [152] That is, the substitution can be de re or de dicto. Consider the following examples from Sophocles: Philoctetes is suffering a painful attack from his sickness. He does not want Neoptolemus to notice, however, so tries to deny that anything serious is happening. In 733bis, after a short answer to Neoptolemus’ question, Philoctetes quickly shifts to a directive discourse act: ἀλλ’ ἴθ’. [154] In this case, ἀλλά and the preceding negation οὐδέν are not part of the same construction; rather, the particle marks a switch to a different kind of act. That is, ἀλλά here marks a de dicto rather than a de re substitution. In 735 and 740, conversely, one element is first negated, and then its substitution is explicitly mentioned. Sophoclean characters frequently bring up such substitutions, especially in dialogues. [155] The speaker first utters an act (often including a negation) that is to be substituted, either de re or de dicto; the following act effects the substitution, and the speaker marks this by means of ἀλλά. She thereby emphasizse the ἀλλά act. That is, a formulation such as “will you not tell, but remain silent?” is more emphatic than just “will you remain silent?” In other words, I consider ἀλλά to contribute to emphasizing the subjective character of a speaker’s substitutions in front of the interlocutor. In Greek drama such emphasis is most often needed in dialogues, and also more at home in monologues where there is an individual speaker, than in choral songs. [156]

§66. ἀλλά in drama is often combined with a negation, another sign of this subjective emphasis. The distributions of negations (both οὐ and μή forms) across the discourse patterns are as follows: Sophocles, who is fond of ἀλλά, has the highest frequency of negations in all discourse patterns. In all authors negation is the most frequent in dialogues. [158] This can be explained because negation is both functionally and formally marked across languages. [159] One of the reasons is that negation is cognitively more demanding than affirmation: it requires more cognitive effort to process negated information than to process positive information. [160] An utterance such as “I am not in pain at all” (see (t20) above) inevitably refers to a situation in which the speaker is in pain. As Miestamo 2009 puts it, negation “is a mental process added by language users” (211). A hearer needs to imagine both the situation with pain, in this case, and its negation. This markedness, in turn, helps the speaker to emphasize part of her utterance: she is not just describing a situation, for example, but also defining it in terms of what it is not.

§67. ἀλλά, then, verbally encodes and emphasizes the substitutions (de re or de dicto) that speakers perceive or construct in their discourse. This function makes the particle suitable for argumentative situations as well: speakers in such contexts want to substitute certain views with certain others. I interpret the high frequency of ἀλλά in Euripidean monologues along this line. Whereas Aeschylus and Sophocles tend to use monologues for narrative or descriptive purposes, Euripidean speeches tend to serve an argumentative function, as they are often integrated within disputes between characters. [161] Euripidean monologues may thus be considered to be more interactive, more strongly oriented toward a specific interlocutor, and hence closer to both dialogue within drama and rhetoric speeches beyond drama, for example in oratory, philosophy, and historiography. [162] An example of this kind of argumentative speech is Jason’s monologue in Euripides Medea 522-575, which contains as many as six instances of ἀλλά (in 339 words). Two of these instances are shown here: In this monologue, Jason is arguing with Medea, carefully answering her reproaches against him. Jason’s style is reminiscent of the sophists. [163] The first ἀλλά in the passage (529) marks a correction of an implied element from the preceding act. Medea may have a delicate or clever mind (νοῦς λεπτός), [164] but she is telling an invidious story about saving Jason; it was in fact Eros who had forced her to save him. The second instance (532) marks a larger shift or “switch” in the discourse. Although the remark about Eros in 530-531 may have given rise to an expectation that Jason would go on about this point, with ἀλλ᾽ οὐκ he signals a different direction in the discourse: other arguments against Medea’s position are more important to him. The particle thus helps in shifting to different points, and so in marking out the steps in one’s argument more precisely. [165] As Mossman 2011 ad loc. remarks, there is a “play of assertion and partial withdrawal” in these lines.

§68. As we have seen, in all of its constructions ἀλλά marks some kind of substitution or correction. The tendency of ἀλλά to appear most often in Aeschylean and Sophoclean dialogues, as well as its frequent co-occurrence with negations, support a connection to the speaker’s subjectivity and to interactiveness. Euripides’ preference for ἀλλά in monologues can be explained with reference to the rhetorical purposes of many of these speeches.

§69. μέν is generally described as setting up an expectation for some part to follow. [166] That is, in pragmatic terms, it projects another discourse act, in fifth-century Attic often a δέ act. [167] In drama μέν occurs most frequently in monologues: Besides indicating that more narrative steps will follow in story-telling monologues, [168] speakers can also use μέν, for example, to mark a juxtaposition of (parts of) conditions, arguments, or points of view. [169] μέν therefore does not seem to be especially associated with narratives, unlike δέ (see §§27-30 above).

§70. An example of a monologue with many μέν instances is the priest’s speech in Sophocles Oedipus King 14-57. It contains six occurrences of the particle in 267 words in total (a frequency of more than two percent). Here is the beginning of this monologue: These μέν occurrences have different scopes. The first one, in line 15, helps to establish that the priest is starting a long, elaborate answer to Oedipus’ questions. [170] The second one (16) falls within the scope of this first instance. As part of a οἱ μέν… οἱ δέ construction, μέν simultaneously marks one discourse act as preliminary, and projects another act. [171] The μέν in 18 signals an implicit contrast between the priest himself and other persons. [172] Though this last use of the particle could work in short utterances just as well, the former two are especially appropriate to utterances with a certain length of speaking.

§71. When an Aristophanic character manages to utter a substantial monologue, μέν also appears regularly. In the following passage from Birds a herald has arrived in the bird city to inform Peisetaerus of the bird mania in Athens: The herald’s entire speech runs from 1277 to 1307 (180 words in total), an impressive length for a comic utterance. As Dunbar 1995 ad 1298 remarks, such “lengthy narrative speeches” are rare in Aristophanes. The speaker therefore needs linguistic signals in order to clarify that he still has more to say. In the quoted passage the two μέν instances are among these signals. [173] The one in 1286 creates the expectation, together with πρῶτον, that the first action described will not be the only one. μέν in 1292 projects more entries in a list of bird nicknames, after the announcement of this topic in 1291. The communicative setting also helps the herald to hold the floor: his addressee, Peisetaerus, actually wants to hear the message, which makes interruptions less likely. The herald’s speech, as Dunbar points out, is a parody of a tragic messenger speech, a dramatic subgenre that by definition requires some length.

§72. In general, then, μέν’s projective function is more appropriate when the speaker goes on talking than in short utterances. Though the projected next act or entity may be left implicit, in my corpus usually the projection is indeed fulfilled in the discourse acts following a μέν act. In this way I connect μέν’s relatively high frequency in dramatic monologues to its basic pragmatic function, which is to project one or more discourse acts.

§73. Aeschylus uses δή most in dialogues, Euripides most in monologues. Interestingly, in Homer δή is more frequent in direct speech than in narrator text: the Aeschylean distribution of this particle may thus form another link to Homeric particle use. [174] δή’s relatively low frequency in songs mirrors its rarity in lyric in general. [175] Various uses of δή, found mainly in dialogues or in somehow dialogic contexts (see e.g. (t25) below), tend to be connected to the speaker’s expression of stance, and the interaction between speaker and hearer. Sometimes it intensifies part of an utterance; in other contexts it marks the event referred to as perceivable, obvious, or expected, or it marks the discourse act as such. The use of δή to mark narrative steps, which is common in Homeric narrator text and in Herodotus, occurs as well, but is rare in drama; this use requires relatively long stretches of narrative. [176]

§74. Just as in Homer and historiography, δή in drama may occur in stancetaking expressions, where it is used in order to intensify these expressions. The co-text in such cases contains an adjective or an adverb that expresses an extreme of some kind, such as superlatives or words like “all” or “alone.” [177] Here is an example from a 6-line utterance from Aeschylus Persians: δή here emphasizes the subjective judgement κακῶν ὕψιστα. [178] The particle’s late position in the act (the discourse act arguably starts directly after the interjection αἰαῖ) suggests that δή here has small scope, over only the superlative preceding it. This subjectivizing function can be connected to descriptions of δή as expressing affirmation or the speaker’s certainty. [179] The particle’s higher frequency in Aeschylean dialogues supports a connection to the speaker’s own view or attitude. However, the use of this particle does not in itself imply a contrast with different views, as γε does. [180]

§75. When it has scope over an entire act, δή may have a number of functions that are related to each other. One is to mark the content of its act as visible or otherwise perceptible to speaker and addressee. Many scholars in fact argue for “clearly,” “obviously,” or referring to something known or visible as the basic meaning of δή. [181] While I do not claim that this nuance is present in every instance of the particle in drama, I do consider it part of its functions. The following Euripidean example stems from a 12-line utterance in which Hippolytus says goodbye to several addressees, among whom the goddess Artemis: δή is in peninitial position in its act here, since an act boundary after a vocative is likely. [182] Hippolytus addresses Artemis at the moment of his exile. He assumes that the goddess sees or hears what is happening to him (if she can hear this very utterance, she will also have heard Theseus’ preceding order to Hippolytus to leave the country). Thus the particle δή accompanies the description of an action that is perceptible to speaker and addressee alike, and marks the act as such. [183] Though there is no quick dialogic exchange here, the δή act does convey the speaker’s attention to his addressee.

§76. A similar situation, where δή also has scope over an entire act, is when a speaker refers back to what the addressee has just said. An example is found in the magistrate’s reaction (1.5 lines long, see 571-572) to Lysistrata’s utterance in the following passage: δή in 571 co-occurs with a lexical echo from the preceding utterance (κλωστήρων καὶ ἀτράκτων, echoing κλωστῆρ’ (…) τοῖσιν ἀτράκτοις from 567-568). A heavy paraphrase of the particle in this context would be “as you say.” The information referred to in such quoting contexts may not be directly perceptible, but will nevertheless be obvious to both of the interlocutors. This use of is δή close to the one found in historiography to resume a narrative thread after an interruption or another topic (“as I was saying”). [184] The difference, however, is that in those cases δή refers back to the narrator’s own previous discourse, whereas in drama speakers refer to utterances by their addressees. [185]

§77. If certain information or a certain event is perceptible, or has just been mentioned, this means, more generally, that it is evident for speaker and addressee, or at least it can be presented as such. The content of δή acts can also be evident, obvious, or expected for reasons other than being directly perceptible or quoted. [186] In this use the particle is sometimes combined with καί; together the two particles have become a cluster to mark the noticing of a new character on stage, [187] as well as the speaker’s obedience to a directive. [188] These uses, again, directly concern the interaction among the characters on stage.

§78. Related to this use is δή’s ability to mark not the content of its act, but the very utterance of an act as expected or obvious. [189] This is the most natural interpretation of many δή instances in questions, such as in the Lysistrata dialogue directly after the lines cited in (t26): [190] I interpret the δή in 574 as marking its host question as an expected action in its context. After Lysistrata’s vague statement about applying wool strategies to politics, the magistrate considers it a logical communicative action to ask what she means exactly. That is, the particle does not concern logical relations between the content of different utterances (as δῆτα signals in questions), but indicates in this case that the action of asking is an obvious one. This signal is most appropriate in dialogues, the setting in which the focus on the ongoing interaction is the highest.

§79. To sum up: δή in drama, is most frequent in situations with high interactiveness: that is, in relatively short turns of speaking, or those that otherwise betray attention to the interaction, for example by using vocatives (see (t25)). In choral songs, a setting where a face-to-face interaction on the stage is hardly present, δή has a relatively low frequency. Based on this distributional input as well as co-textual patterns and the behavior of δή in other genres, I interpret its functions as follows. When δή has small scope, over an adjective or adverb that describes a quality in the extreme, it intensifies an expression of the speaker’s stance, a function that is also common in Homer and historiography. With act scope δή either marks the content of its act or the uttering of the act as perceptible, evident, or expected to the speaker. These functions reflect the speaker’s attention to her addressee: she presents the content or the action as evident to both herself and a “you.” They do not, however, require an immediate verbal reaction from this addressee, as is likely in the case of γε and δῆτα (see §§58-63 above), particles that favor dialogues in an even stronger way. Finally, monologues display some uses of to δή to mark narrative progression (frequent in Homer and Herodotus), but because of the relatively small proportion of narrative in drama, this concerns a minority of instances.

§80. This particle is the most frequent in dialogues, in Sophocles and Euripides significantly so: As for οὖν’s low frequency in choral songs, at least in Sophocles and Euripides, it is relevant that the particle is rare in epic and Pindar. [191] As I have argued in my discussion of δέ and τε above, we can see once more that choral songs tend to be linguistically closer to epic and lyric than monologues and dialogues. [192] In general, the playwrights’ preference to use οὖν in dialogues can be explained from the greater explicit presence of the speaker in this situation, and the high level of interactiveness among the characters. [193]

§81. With οὖν a speaker may present an utterance as an inference or conclusion, as in this excerpt from a monologue by Hippolytus: Hippolytus here expresses his shock and disgust at the proposal of Phaedra’s nurse that he start an affair with Phaedra. οὖν occurs in a rhetorical question that implies “I could never be (so) base.” The rejected idea is based on the following premises: (1) someone who would violate his own father’s bed would have a strong desire to do so, and (2) Hippolytus himself is too pure to ever have such a desire. Premise (1) is left implicit; (2) is mentioned in the following relative clause. οὖν, by marking its host act as an inference or conclusion, shows the speaker’s choice to present his discourse as such, in pursuing his goal to communicate his view to the nurse. That is, instead of leaving it to the addressee to detect a specific relation between parts of the utterance, the particle explicitly marks the kind of connection that the speaker wants to be understood. [194]

§82. οὖν may also be combined with μέν. In some instances, as in (t29), each particle retains its own separate function; in others it is used as a cluster (see (t30) below). Both uses are slightly different from those of μὲν οὖν in Thucydides, discussed in IV.3 §§144-146. While the historian uses μὲν οὖν to introduce new discourse threads within extended narratives, in drama the combination contributes to the interpersonal level of communication, in several ways. What remains the same across these genres, however, is the contribution to the structuring of the discourse: μὲν οὖν connects different acts to each other in a specific way. Consider this example from a 20-line utterance by Ismene to Antigone: In this case, as commentators point out, the two particles work separately. [195] μέν, on the one hand, has a projecting function. It therefore places emphasis on ἐγώ in a way that sets it in contrast against something implied. Here the speaker Ismene, in outlining her intention to obey Creon’s injunction against burying her brother, suggests that her plan of action may diverge from Antigone’s, her addressee. [196] οὖν, on the other hand, marks the upcoming acts as a conclusion from the preceding discourse. Ismene should bury her brother, but has to consent to the king; therefore she will ask the deceased for understanding, and obey. That is, μέν helps to indicate potential disagreement; οὖν underlines how the speaker interprets a certain situation—in Ismene’s view one has to conclude that obedience is inevitable. Both particles thus relate to both the discourse organization (connections among different acts) and the interaction between the speaker and addressee (disagreement, and Ismene’s responsibility for her conclusion).

§83. When the two particles work together as a cluster, μὲν οὖν may mark a correction of a preceding element. The corrected element is often from an utterance spoken by the addressee, which clearly connects this function to the ongoing interaction. An example is found in the following Aristophanic scene. Hermes and Trygaeus are talking about the Spartans: As Platnauer 1964 and Olson 1998 ad loc. note, μὲν οὖν indicates that Trygaeus does not agree with the previous utterance, but corrects part of it: in his view the Spartan farmers were not innocent in the war. Since the Spartans were responsible for Trygaeus’ loss of his fig tree, in his view they deserved to lose some figs themselves. Olson interprets the combination as “no, to the contrary.” [197] Since this corrective use cannot be inferred from the normal pragmatic contributions of each of the two particles, I speak of a cluster in this case. [198] It occurs especially in turn-initial position, which makes it more frequent in the discourse pattern connected to dialogues. [199] The cluster’s predominance in this communicative setting is understandable in view of its interactional value.

§84. Just as several other particles that tend to be most frequent in dialogues, then, οὖν fits situations with high interactiveness. More specifically, with οὖν a speaker may mark a discourse act as her own conclusion or inference from the preceding. Such a relation is not inherent in the semantic content of an utterance, but, crucially, betrays the speaker’s interpretation of that content, and the way she wants to present it to the addressee. The cluster μὲν οὖν is even more strongly subjective: the speaker presents her discourse act as a correction, usually of an act uttered by the addressee.

§85. ἦ is used the most in tragic dialogues, less often in songs, and hardly ever in monologues: In general, ἦ is considered to have two functions, an interrogative and an affirmative one. [200] In fact, the particle’s two different functions are divided across the two dramatic communicative settings in which it is mainly used: we tend to find ἦ in questions in dialogues, and in assertions in choral songs. [201] However, we shall see that its uses in these two contexts are related.

§86. For a few examples in questions, consider the following passage from Sophocles Oedipus King. It is part of the dialogue with the highest number of ἦ instances in my corpus (six instances in 517 words in total). The speakers are Oedipus and a Corinthian messenger. In each case ἦ starts a question. Since none of these questions contains a question word, I interpret ἦ as one of the signals that the upcoming utterance is a question. Several indications lead to this interpretation. First and most important, there are no elements that would point to ἦ’s other use, which is connected to stancetaking (see on (t32) below). Second, the replies to the turns in 1039, 1043, and 1045 start with οὐκ (1040), μάλιστα (1044) and a turn-initial γε (1046): since these words are typical for answers to questions, they retrospectively suggest that the earlier turns were questions. Third, the ἦ in 1041 occurs after τίς οὗτος, a clear question; therefore it is logical to take the ἦ act as an appendix to that question.

§87. The question in 1039 is marked with turn-initial γάρ as an inference from the preceding utterance, [202] and asks for confirmation of this inference. ἦ γάρ together can be translated as, “are you really (ἦ) saying that…?” Oedipus’ other three questions in this passage likewise ask for confirmation of their suggested statements. However, unlike ἦ questions or ἦ statements in Homer, which always seem to concern an evaluation of a character, these suggestions here concern facts. [203] In general, asking for confirmation is close to the affirmative force that the particle has in assertions. [204] At the same time, the ἦ instances imply that Oedipus is highly emotionally involved in this interrogation. [205] The facts he is asking about are not just plain facts; they have emotional significance to him. This involvement is indeed apparent from his insistence, throughout this scene, to find out everything about his own origin. [206]

§88. The emotional charge of ἦ questions is present in ἦ assertions as well, which are mainly found in choral songs. Such utterances generally do not concern facts, but the speaker’s stance. Consider the following excerpt from a song by the Persian elders, who ask Earth and the underworld gods to send up their dead king Darius: The elders underscore the attitude they are expressing toward Darius with ἦ: it is φίλος, the word of emotion, that receives the emphasis. [207] It is disputed whether there should be one or two instances in this line; [208] in any case their function here is clearly connected to the emotional evaluation. The contribution of ἦ to emotional evaluation blurs the borderline between a particle and an interjection, a connection strengthened by ἦ’s capacity to form a discourse act on its own (as e.g. in 1045 in (t31) above). [209] Several interjections are typically at home in choral songs. Their pragmatic function, like that of ἦ, primarily concerns the expressive level of communication; it is less focused on eliciting a certain reaction from the addressee. I interpret the distribution of ἦ’s use in assertions in light of this connection: choral songs imply less individuality than the other settings, but emotional involvement may still be high.

§89. In short, in both of its uses in drama the particle ἦ conveys the speaker’s involvement. ἦ questions ask for confirmation of the suggested assertion. Unlike in Homer, such questions may concern facts as well as opinions. Since questions are inherently dialogic, this use is most frequent in dialogues. In assertions the particle affirms an assessment, an usually emotionally-laden one. In this use, found mainly in choral songs, ἦ’s function is similar to that of an interjection, by expressing the speaker’s involvement without necessarily asking for a reaction.

§90. In this chapter I have argued that paying attention to the distribution of particles over different communicative situations and their co-occurrence patterns with other features improves our understanding of these particles’ functions. That is to say, exploring dialogues, monologues, and choral songs as discourse patterns with specific formal as well as functional features has proven useful for linguistic analysis, and promises to be a fruitful avenue for further research. One could, within drama, delve into hybrid communicative situations such as lyric dialogues, [210] or into the distributions of linguistic features other than the ones analyzed here, such as adjectives or certain subordinating conjunctions. In other genres, it would be interesting to compare similar distributional patterns across, for example, narrative and direct speech, [211] as well as across authors of the same genres.

§91. The general pictures of the three discourse patterns are the following. Dialogues usually have a relatively high frequency of γε and δῆτα in all authors; ἦ in tragedy; γάρ in Sophocles and Aristophanes; οὖν in Sophocles and Euripides; and καί and δή in Aeschylus. Other features that tend to be relatively frequent in this environment are finite verbs, first- and second-person references, negations, and (in Aristophanes) swearing expressions. Nouns tend to be relatively infrequent. These features reflect a high degree of on-stage interactiveness and individual involvement, and a low degree of formality. In monologues relatively common particles are μέν in tragedy; καί in Sophocles and Euripides; δέ, δή, and ἀλλά in Euripides; and τε in Aeschylus. In this discourse pattern we find the highest relative number of imperfects: a sign of this setting’s affinity to narratives. Aeschylus also uses the most nouns in his monologues. The linguistic shape of monologues reflects a lower degree of involvement and interactiveness than in dialogues, but a higher degree than in choral songs. It can also be connected to floor-holding: the features reflect that speakers take more time to formulate their utterance than in the short turns of dialogues. Choral songs, finally, have the greatest density of the particle τε (except in Aeschylus) and of nouns. Aeschylus tends to use many δέ in these lyric parts, and Aristophanes many καί. The distributions of the selected features reflect that choral songs have a low degree of interactiveness among characters, a low degree of individual involvement, a high degree of formality, and a particular connection to tradition and rituality. Certain features, such as a high frequency of δέ, also help to regularly allude to the style of epic.

§92. In general, this information about relative frequencies of particles and other features may be used as a blueprint, so to speak, for the three discourse patterns. It will for example help us to tell, when reading tragedy and comedy, if a certain particle is marked in its current context (see e.g. (t19)), or if in fact the absence of a particle is striking (see e.g. (t9)). The distributional tendencies also clarify which particles are expected to co-occur more or less often: at least for drama, we do not need to share Denniston’s surprise at the rarity of τε and γε occurring together. [212] On the contrary: the functions of these two particles fit different communicative settings.

§93. The observations also throw light upon differences in particle use across the four authors and two genres. Aeschylus is the king of δέ among these playwrights: especially in his choral songs, the particle is extremely frequent. Sophoclean and Aristophanic dialogues often resemble each other in their particle use: they share a high frequency of γάρ, ἀλλά, οὖν, and δῆτα. Aristophanic dialogues, however, surpass all other contexts in their frequency of γε. [213] Aristophanes further shows an extreme fondness for καί and τε in choral songs, although the latter particle is about as common in Euripidean songs.

§94. By using the distributions of the particles and co-occurring features as input for interpretation, I have connected the particles’ local functions to several global associations invited by each discourse pattern. That is, the local co-text is not enough to interpret a particle’s pragmatic contribution: knowledge about its distribution enhances our understanding of why a certain particle fits a certain context. For example, the distribution of δέ can be connected to its relatively neutral local function, and, more globally, to allusions to epic in specific contexts such as messenger speeches. καί is particularly multifunctional: some of its uses fit the high interactiveness of dialogues, others the longer floor-holding of monologues. τε does not only mark coordination between two items: its higher frequency in certain contexts demonstrates its overarching connection to shared encyclopedic or cultural knowledge and several related associations. This interpretation of τε’s large-scale pragmatics echoes its associations in other genres. Along the same lines, I interpret the contrastive function of γε as related to the interaction between characters, more specifically to their engagement with each other’s varying opinions. For δή, the local co-text is required to distinguish between its several functions; most of these signpost personal and/or interpersonal attitudes that have a bearing on the global communicative situation. All in all, my results serve as a warning not to treat all these particles together as one linguistic feature. Their distributions, local functions, and global associations vary greatly, even within the dramatic corpus.

§95. In other words, then, the small-scale pragmatics of particles can be connected to their large-scale pragmatics. This illustrates the necessity of a discourse perspective on particle use: since particles are used differently in different discourse patterns, and since they carry additional implications beyond their local function, the sentence in which they occur is not sufficient for their interpretation. Large-scale patterns of distribution are crucial for understanding the particles’ functions, uses, and implications.

§96. For the sake of completeness, I give here the distributions of ἄρα/ἆρα, μέντοι, and μήν. They are not discussed in the chapter, because they are not statistically significant in most authors. This has to to do with their generally low frequency, compared to that of other particles.