III.4 Speaking in turns: Conversation Analysis

§1. Characters in tragedy and comedy talk in turns. Aeschylus’ Persian queen and the ghost of Darius converse about the army’s defeat through a series of questions and answers. Sophocles’ Oedipus gets angrier at Teiresias with every line he utters. Medea and Jason in Euripides’ play express their feelings both in long argumentative speeches and in more rapid dialogues. In Aristophanes, characters regularly interrupt each other with short comments. Characters advance the plot in their spoken interactions by asking questions, giving orders, expressing opinions, and so on. [1]

§2. This turn-taking is one of tragedies’ and comedies’ formal aspects that distinguishes the genres from most other Greek texts: dramatic texts directly reflect the voices of different speakers who were physically co-present in a theatre, communicating with each other in real time. The dialogic nature of plays influences the use of particles, as argued in this chapter. It is fruitful, therefore, to approach particle use in these texts through a framework that deals with the functioning of dialogic interaction: Conversation Analysis (CA).

§3. Naturally the dialogues of Greek drama are stylized and formalized versions of real spoken conversation, but since they are ultimately based on spoken language, we may reasonably assume that these texts contain remnants of the rules of real conversation. CA can teach us something about these rules. The difference between our material and that of most CA research—written texts rather than recorded spontaneous conversation—means that our goals are accordingly different: we do not aim to understand conversation in general by looking at ancient drama texts. Rather, insights from CA, based on real conversation, can clarify the language used in these plays.

§4. Recent scholarship has begun to apply CA methods to the study of ancient Greek literature: Minchin 2007 on Homer, and Van Emde Boas 2010 and 2017b and Schuren 2015, all on Euripides, make use of the framework. [2] Minchin mainly focuses on how linguistic forms are linked to certain social actions, such as rebuking, declining an invitation, and asking a (specific kind of) question. She also discusses aspects of turn-taking in Homer. Van Emde Boas uses CA alongside several other modern linguistic approaches in his analysis of Euripides Electra. He argues that the characters’ linguistic patterns play an important role in their characterization. Furthermore, he discusses how approaches such as CA can help in the case of textual problems, particularly concerning speaker-line attribution. Schuren applies a broad pragmatic framework, including insights from CA as well as speech-act theory, as well as narratology, in her monograph on Euripidean stichomythia. She focuses in particular on turn-taking, social deixis, and storytelling. These three scholars observe, by adopting a CA perspective, phenomena that would otherwise go unnoticed or remain unexplained. At the same time, they strengthen our awareness of the general similarities between our own everyday conversation and the language use in ancient literature: the same communicative principles are often at work. As Van Emde Boas (2010:8) rightly points out, “for dramatic dialogue to be comprehensible to an audience, it still must use the same linguistic resources that are familiar to them from their own daily conversations.” [3]

§5. Earlier remarks on the structure of conversations in ancient Greek drama can be found in Hancock 1917 on stichomythia in different genres, Gelzer 1960 on the Aristophanic agon, Ireland 1974 on Aeschylean stichomythia, Bain 1977 on asides, Mastronarde 1979 on tragic dialogue, and Dover 1987a on language and character in Aristophanes, among others. However, CA raises, and provides answers for, a number of important questions which these scholars have not addressed, and which still require systematic analysis. Indeed Mastronarde 1979 assumes that real conversation is simply a “chaos” without any regularity. [4] Yet CA has shown that conversation in fact exhibits a great deal of systematic organization, a view supported in the Greek by the recurrence of certain forms in certain turn positions and sequential positions. The framework thus has the potential to offer important explanations about these recurrences.

§6. This chapter uses the CA approach to show that conversational structures and practices influence language production in tragic and comic dialogues, and thus also the selection and use of particles. I begin by introducing CA, its terminology and its various aspects. Textual analysis follows, mainly focused on the dialogic parts of the plays, [5] the communicative environment that CA has the most to say about.

§7. CA focuses on talk-in-interaction, that is, on language used for performing social actions. [6] The approach originated in sociology, with Sacks’ lectures on conversation (1964-1968), [7] Schegloff’s work on conversational openings (1967; 1968), and, most well-known, the seminal article by Sacks, Schegloff, and Jefferson 1974 on the systematics of turn-taking in conversation. Further explorations along these lines followed. [8] Conversation analysts aim to describe and understand the system, rules, and practices of talk-in-interaction. They emphasize that we can better understand utterances if we pay attention to what they are doing rather than to what they are about. [9] Here CA builds on Austin’s 1962 claim that words do things: utterances do not merely describe the world, but perform actions. [10]

§8. The basic unit of conversation is the “turn,” also called “turn of speaking,” or “turn-at-talk.” Conversation, like other forms of coordinated, joint activities, requires some kind of turn-taking to manage the contributions of the different participants. [11] Examples of such joint activities, noted by Sidnell 2010:36, are ballroom dancing, road work, and open heart surgery. Clark 1996:59 also gives analogous examples of other non-verbal joint actions, such as playing music or paddling a canoe; as well as of different activities involving talk as part of them, such as a business transaction. In the case of conversation, participants take turns-at-talk. These turns are themselves composed of one or more turn-constructional units (TCUs), the smallest units that may constitute a turn. [12]

§9. The word “turn,” as I use it here, refers to the linguistic realization of actions, that is, to a string of words uttered by one speaker, rather than to the action(s) performed by these words. [13] In my use “turn” is equivalent to “utterance”; these two terms only differ in the perspective they offer. [14] “Utterance” neutrally refers to everything that is said by one speaker until she stops talking; “turn” refers to the positioning of a stretch of talk by one speaker with respect to other stretches of talk by other speakers. [15] Consider the following example: The first turn (or, more neutrally, utterance) by the slave is only one word long: its function is to address Dicaeopolis and catch his attention. [18] Dicaeopolis reacts immediately, starting a turn in the middle of the verse. [19] This turn performs two related requests for information: who the speaker is (or what is going on; see note 17), and why he addressed Dicaeopolis. To respond to these requests, the slave needs a longer turn (slightly more than three lines), built out of several parts or TCUs. With the first TCU, ὅ τι; “(you ask me) why?”, he projects a relatively long answer. [20] Subsequently, he reports Lamachus’ two orders, in this way indirectly ordering Dicaeopolis to provide the requested items. We can interpret the projection and the two reports as three separate actions. Regardless of the number of actions performed by a turn, however, I speak of one turn when it is continuously uttered by one speaker. [21]

§10. An important part of a turn is its start: the start frequently gives indications as to how a turn fits into a sequence or a series of sequences. [22] Turn-initial items also often project what kind of turn the speaker has just started: they foreshadow a certain syntax and/or a certain action. [23] As Sidnell 2010:143 puts it, “the initial components of a turn can strongly project the type of turn underway.” The role of Greek particles in a conversational structure is therefore most visible when they occur in turn-initial position. For this reason my discussion in this chapter focuses mainly on turn-initial particles.

§11. I define turn-initial particles as the first particle occurring in its earliest possible position in the first discourse act of a turn. [24] For second-position particles (in the classical sense), I will still speak of turn-initial position when they are found directly after the first constituent of a turn. Thus, for instance, the “postponed” δέ in Aeschylus Persians 719 (see (t8) with note 68) is considered turn-initial, but for example particles occurring in the discourse act after the one constituted by a vocative are not. [25] If a turn starts with a swearing expression in Aristophanes, a particle following that is not considered turn-initial, because swearing expressions can constitute separate discourse acts, just as vocatives and interjections do; for example, καὶ μήν after νὴ τὸν Δία in Aristophanes Frogs 285 is therefore not considered turn-initial. However, since establishing the start and end of discourse acts involves interpretation, the determination of a particle’s turn-initial position can be subjective. The chapter focuses on the clearest instances of turn-initial particles. [26]

§12. Most turns without turn-initial particles are explicitly connected to their co-text and context by other turn-initial expressions. I call these expressions “contextualization cues,” a term coined by Cook-Gumperz and Gumperz, [27] and I include the following forms apart from particles. First, a turn is immediately situated if it starts with a reference to the speaker or addressee(s), which can be realized by first- or second-person verb forms, vocatives, and pronouns. Second, subordinating conjunctions and demonstrative pronouns and adverbs also make it clear at the outset how a new turn is responding to the preceding one. Third, lexical repetitions of an element from the preceding turn clarify the response’s focus. Fourth, primary interjections and (only in comedy) swearing expressions indicate a reaction to a previous turn or nonverbal action. [28] Fifth, turn-initial question words and negations usually project part of the nature of the new turn. We will see below that turns without turn-initial particles or any of these other turn-initial contextualization cues tend to be found in particular contexts, and that the conversational structure is crucial for situating these turns. [29]

§13. One area of research in CA is how turn-taking is organized, that is, when exactly speakers start and end their turns. Sacks, Schegloff, and Jefferson 1974 show that usually one party talks at a time; pauses and overlaps in talk tend to be brief. Participants in an interaction monitor when there is a “transition relevance place” (TRP) in a turn of speaking, and pay attention to these TRPs in their conversational behavior. [30] Speakers make sure, for example, to leave no pause at a TRP if they want to hold the floor. Listeners tend to start a new turn exactly at the moment of a TRP, such as when a syntactic unit is complete. In English this means, for example, that tag questions are relatively frequently overlapped by the first part of a new turn. In the formalized speech of Greek drama, the end of a verse line is typically a TRP.

§14. CA has also developed the study of what it calls “sequence organization.” Sequences are “courses of action implemented through talk.” [31] For example, a sequence may consist of one speaker asking a question or requesting a certain action, and another speaker’s response. The organization of sequences involves “the ways in which turns-at-talk are ordered and combined to make actions take place in conversation.” [32] Research on sequence organization looks at how speakers make their turns coherent with prior turns. [33] This research thus focuses on the specific actions performed by turns, and the structuring of those actions, rather than the moments at which turns may start or end.

§15. Crucially, sequences are built around “adjacency pairs.” An adjacency pair is a unit of two turns by different speakers that are placed next to each other, are relatively ordered, and are of the same pair type. [34] That is, the order of the two turns matters, and the actions they perform belong together. The first turn, called the “first pair part,” makes only certain responses relevant; this second turn is termed “second pair part.” Some examples of adjacency pairs are greeting-greeting, question-answer, assessment-(dis)agreement, offer-acceptance/rejection. If the expected second pair part is absent, this is an “official absence” for the participants. Speakers often indicate in their response that they notice this absence, as in this example from Aristophanes: The second slave’s question in 121 is the first part of an adjacency pair; an answer to it is the expected second pair part. The first slave does not answer the question, however, but instead orders his friend to pour him another glass of wine. But since, in general, answers are normatively expected [35] after questions, the second slave takes the order as an answer—the oracle says, “pour me a refill”—or pretends to do so for the sake of the joke.

§16. Although a sequence in principle consists of a single adjacency pair, the pair can also be expanded by other pairs placed before, after, or in between it. These other pairs are called pre-, insert, and post-expansions: Which turns are considered the base pair and whether other turns are seen as expansions on it depends on one’s interpretation of the whole sequence. In the case of our ancient plays, we can of course rely only on indications in the texts, as we do not have access to nonverbal cues such as pauses, intonation, or gestures. In addition, the expansions themselves may be subject to further expansions, which can lead to highly complex sequences.

§17. Pre-expansions, as Sidnell 2010:103 puts it, “are recognizably preliminary to some other action whose production they project.” For example, a question about availability typically precedes an invitation sequence. An insertion sequence delays the second pair part to deal with issues that need to be resolved before the second pair part can be produced. Such insert expansion may occur, for example, because a participant has misunderstood the first pair part. [36] Finally, a post-sequence expands on the base sequence after the second pair part. Post-expansions can be “sequence-closing thirds,” with which the speaker intends to close a sequence. [37] Common forms of sequence-closing thirds in English are oh, okay, and assessments, as in the following excerpt. It starts with person A asking his friends B and C, who are a couple, a question. Schegloff 2007:124-125 points out that the assessment “good” by speaker A function as a sequence-closing third. After uttering this post-expansion, speaker A launches a new sequence with “so listen…”.

§18. Adjacency pairs are also structured according to what is known as a “preference organization.” Usually a certain type of second pair part is interactionally preferred over some other type, namely the kind of response that “promotes the accomplishment of the activity underway.” [38] For example, an acceptance is a “preferred response” to an invitation, a rejection a “dispreferred” one. Note that this is a structural, interactional preference that speakers orient to and that deals with normative expectations; it is independent from the speaker’s actual, psychological preference. (A speaker may, after all, be relieved if her invitation is rejected.) As part of preference organization, dispreferred responses are marked, both in form and in delivery. They often contain explanations about why they are produced, for example why an offer is declined or a request refused. As Levinson 2006:48 writes, “[r]esponses that are in the expected direction are immediate and brief, responses that are in the opposite direction are typically delayed, marked with hesitations and particles like well, [39] and accompanied by explanations.”

§19. Finally, CA scholars stress that turns perform actions. The term “action,” used in CA, is different from the term “(discourse) act” used in Discourse Analysis. The two concepts are not mutually exclusive; rather, the terms reflect different perspectives on the same idea, that language is used for doing things.

§20. A discourse “act,” on the one hand, is viewed in relation to the surrounding whole: an act is a small step within a larger discourse. [40] Each act has a certain function contributing to the main goal of the discourse. Acts are often described as prosodic or orthographic units. [41] Much research on such acts looks at which segments of discourse can be said to form a small step, and where these segments start and end. [42] Thus, in (t3) above, the second turn by speaker A, “Good. So listen, are you, uh, this thing is still off?”, could be described as consisting of five acts, based on its prosodic realization as reflected in the punctuation of the transcription. In written language, act boundaries manifest themselves through certain linguistic indications; particles constitute one set of important signs in Greek. [43] The concept in CA which is closest to the discourse act is the turn-constructional unit (TCU). [44]

§21. “Action” in CA, on the other hand, focuses on what the speaker wants to accomplish in a social situation: actions are “things that people do in their talking in interaction.” [45] Sidnell and Enfield 2012:328 list “requesting, inviting, offering, complaining, excusing, agreeing, and disagreeing” as examples of what they consider actions. As Levinson 2013:104 points out, the assignment of a certain action to a turn tends to be revealed by the response of a next speaker. Accordingly, Levinson considers the primary action performed by a turn to be “what the response must deal with in order to count as an adequate next turn” (107). To put it in more general terms, an action needs to be something recognizable to the participants, an identifiable communicative doing, for which they hold the speaker accountable. [46] Note that “action” does not refer to the words and nonverbal signs used to accomplish communicative goals, but rather to the thing accomplished itself—the invitation, the summoning, the questioning, and so on. Thus CA scholars are more interested in understanding what a turn or part of a turn is doing in a social situation than they are in identifying the boundaries of an action’s realization in words.

§22. We can combine the two concepts to say that (discourse) acts perform (social) actions—short segments of talk, alone or in groups, perform questions, invitations, summonings, and so on. Actions can be realized in single acts—such as a vocative, which performs the action of addressing someone—as well as in multi-act moves—such as a series of acts performing an invitation. [47] In what follows I will use both terms in the way described above: “act” will refer to short stretches of discourse, which have an arguable start and end, and “action” to the social doings performed by talk.

§23. To sum up: this section has sketched the main concepts that CA scholars use to describe how people interact by means of turns-at-talk. We can now move on to the application of those concepts to the study of particle use in Greek drama.

§24. Greek particles often signal something on a conversational-structural level, in the process of performing other functions, such as signaling contrast between entities or marking discourse boundaries. CA therefore enriches our understanding of particles. Examining the role that particles play in the organization of turns, sequences, preference, and action helps us understand why and how particles are used in the contexts in which they are used.

§25. This chapter applies four concepts of CA: turn-taking organization (§§26-31), sequence organization (§§32-48), preference organization (§§49-56), and the actions performed by turns (§§57-70). The main focus is on question-answer pairs, because they appear frequently in the corpus and are clearly recognizable as pairs. That is to say, a question as a first pair part sets up strong expectations about the relevant second pair part: the norm is that this second part is an answer (see (t2)). These sections are followed by concluding remarks on what we can learn from CA about ancient Greek particle use (§§71-72), and an appendix with quantitative observations on turn-taking and on turn-initial expressions in the corpus (§§73-75).

§26. This section looks at the interaction between turn-taking organization and particle use. Particles play a role in the turn-taking process by indicating the speaker’s lack of acknowledgement of a previous turn (τε), or by helping her hold the floor for a turn of multiple lines (μέν).

§27. The following passage from Euripides Hippolytus illustrates the use of turn-initial τε. In this scene Phaedra is suffering heavily from being in love with her stepson; her nurse, unaware that Phaedra is lovesick, tries to find out the cause of Phaedra’s illness. Finally Phaedra consents to the questioning, and starts to give hints about her trouble. In 339, we find τε in the first act of Phaedra’s turn. The particle marks that she continues her own previous turn, in this case by adding another vocative. She thus does not answer the nurse’s question of 338. Indeed, Phaedra’s σύ does not refer to the nurse, her interlocutor present on stage. When τε occurs in turn-initial position, the speaker is ignoring (or pretending to ignore) the turn just uttered by the interlocutor—in other words, the speaker is continuing her own previous turn. [49] The particle is however infrequent in this position. [50]

§28. In multi-line turns, we can often identify certain expressions that help these speakers hold the floor beyond the TRPs [51] constituted by line-ends. The particle μέν is one of these floor-holding devices. Because this particle’s function is to project upcoming discourse acts, [52] it can effectively signal that the speaker wants to hold the floor for some time. [53] At the same time, μέν carries out its general projecting function, marking for example an upcoming change in addressee [54] or a juxtaposition of several items (with δέ following). [55]

§29. The following passage from Aeschylus Libation Bearers features this floor-holding use of μέν: the particle suggests that Electra’s turn will not be over after one line. [56] During the preceding choral song (152-163), Electra has poured libations on Agamemnon’s grave. She makes it explicit (164) that this ritual has been performed, and that her father has received the libations. However, μέν implies that this was not all she wanted to say. Within the same turn, she switches to a new action (with the act νέου δὲ μύθου τοῦδε in 166): announcing an upcoming piece of news to the chorus. [58] Subsequently, a stichomythic exchange ensues until line 180.

§30. μέν is often found together with other floor-holding markers. In this example from Aristophanes, Peisetaerus manages to speak for more than six lines in a row (164bis-170)—remarkably long in this rapid conversation, and in comedy in general. [59] μέν alone is therefore not enough as a floor-holding device in this case, but is used together with syntactic and semantic incompleteness at the ends of lines: How does Peisetaerus make sure in 164bis-170 that his interlocutor, the Hoopoe, does not interrupt whenever he reaches line-ends, typical TRPs in drama? First, he repeats the Hoopoe’s question in indirect form (ὅ τι;), producing the expectation that he intends to answer it elaborately. [60] At the start of this answer, and notably at the very end of the line, we find πρῶτα μέν: a strong sign that (much) more will definitely follow. [61] Then, in 166, Peisetaerus again starts a new sentence just before line-end. [62] The next line similarly ends with an incomplete subordinate clause: ἔρῃ lacks its object, and the ἤν-clause as a whole lacks a main clause. [63] At the end of 168, it is semantics rather than syntax that is incomplete: the verse-final demonstrative ταδί cannot refer to anything preceding it, so it must refer to something that follows.

§31. This section has argued that paying attention to turn-taking organization improves our understanding of linguistic forms found in tragic and comic dialogues. In particular, the use of turn-initial τε and μέν has been illuminated. In the next section, we will look at the structuring of several turns in a row.

§32. Sequence organization explains how each turn responds to the previous one and points forward to further talk. Particles help indicate how a speaker intends her current turn to fit in the ongoing sequence and series of sequences. This section discusses how the concepts of (series of) adjacency pairs and pair expansions help to understand particle use. The first subsection will discuss the role of particles in series of adjacency pairs, in first pair parts, and in second pair parts. The second subsection will present observations on particle use in pre-, insert, and post-expansions.

§33. Several adjacency pairs may follow each other in a series. [64] In such series, the particle μέν may signal that the speaker intends to perform several actions that are similar to the one marked by μέν. The particle may, in other words, not only project more within a single turn (§§28-30), but also on a larger scale. A question containing a turn-initial μέν, for example, can set up the expectation of more questions: [65] μέν in 316 does not function as a floor-holding device, because the nurse first needs an answer to her question. It is more likely signaling that the nurse is just asking the first of a series of questions, as indeed she proceeds to do. [66] The particle also hints at other possible troubles, left implicit for now, from which Phaedra might be suffering. [67] This suggestion of implicit alternatives in fact helps to create the sequential expectation: that the nurse may go on with asking about those other possibilities. In contrast, the expectations raised by the two other instances of μέν in this passage (313 and 317) are fulfilled within the same turn, through the addition of a complementary δέ-act.

§34. In a series of questions and answers, turn-initial δέ is often employed to signal a new question, that is, a new first pair part within the series. Several examples are found in (t8) from a dialogue between the ghost of the Persian king Darius and queen Atossa. The queen has just told her dead husband, in a turn of six lines, that the Persian kingdom has been ruined. The dialogue then goes on as follows: Prompted by the general news of a disaster, Darius asks in 715 in which way Persia has been ruined. The queen’s answer in 716 starts without a turn-initial particle: we will see below that this is no coincidence. Since an answer forms the expected second pair part of a question-answer pair, it does not need an explicit signal clarifying how the turn is linked to the previous one. The queen’s other answers in this passage (718, 720, 722, 724) similarly lack any turn-initial particle.

§35. Darius’ second, third, and fourth questions (717, 719, 721) contain a turn-initial δέ. [68] In each case the particle marks the turn as the next first pair part, here a new question, in the same series of sequences. [69] Within this series of questions, each δέ question is somehow “new,” that is, there is a change of topic. Turn-initial δέ thus helps to locate a turn within the series of pairs—a frequent use of the particle in my corpus. [70]

§36. Notice that the question in 723 starts with καί: an indication that this question fits differently in the series from those marked with δέ. Here Darius does not simply accept the answer from the previous turn and go on to the next question, but lingers on the current topic, the news he has just received. That is, whereas δέ marks discourse discontinuity (such as a change of topic), καί indicates continuity. The fact that the καί turn is a question suggests that the speaker is surprised or indignant about the previous statement, or has doubts about it. [71] Darius here asks for confirmation of the answer given in 722: did Xerxes really yoke the Hellespont? [72] Because the queen has just provided this information, Darius’ request for repetition indicates his surprise. Hancock 1917:29 describes καί questions similarly: they “[leap] spontaneously from the lips as the significance of the other speaker’s words reaches the mind. Sometimes they merely serve to repeat the amazing fact just stated, sometimes they raise a fresh point arising from the other.” With a καί question, in other words, a speaker “zooms in” on a previous utterance, implying surprise, doubt, or even indignation about that utterance. [73] The particle δέ, by contrast, cannot give the signal that the speaker is further pursuing some element of the previous turn. [74]

§37. Sometimes the playwrights exploit the sequential signal conveyed by turn-initial δέ to indicate something more than just a next question in a series. For example, a δέ-turn after receiving dreadful news implies the speaker’s absence of (emotional) reaction to the news. An example from Euripides Bacchae is Agaue’s response upon hearing that she and her sisters have killed her son Pentheus: The particle δέ in 1290 marks the turn as a new step in Agaue’s series of questions, rather than a reaction to the terrible news. Asking “where did he die” after “you killed your own son” is an unexpected and striking response. It shows that the speaker has not (yet) fully understood the disastrous message, or somehow wants to refrain from giving a reaction to it, such as an emotional assessment. [75] This use of δέ is a specific exploitation of its more general function to mark new or different discourse acts. [76]

§38. Similarly, turn-initial δέ may signal that a turn following a question is not an answer to that question; the speaker starts a new action in her δέ-turn instead. [77] Consider Chremylus’ turn after Blepsidemus’ question in the following dialogue from Aristophanes Wealth: [78] Instead of answering the question, Chremylus adds more acts to his own previous turn, which was marked with μέν as projecting something more. The two turns roughly amount to: “you (σὺ μέν), I know what you think. But (δέ) the situation is not like that.” Because Chremylus treats Blepsidemus’ question as an irrelevant interruption, the position of δέ was not intended to be turn-initial. Therefore δέ does not mark the start of a new adjacency pair within a series here, but only the start of a new act, in this case an act that semantically contrasts with Chremylus’ previous acts. [79] In discourse-analytic terms, we can say that δέ always marks an act boundary, while intentionally turn-initial δέ even marks the boundary of a move, a larger unit that probably corresponds to at least an adjacency pair in dialogue. [80]

§39. Another particle that may signal the start of a new adjacency pair is ἀτάρ. It differs, however, from δέ in that it is almost never found in turn-initial position. Out of the total number of 86 instances of ἀτάρ, 83 occur later in a turn. [81] In this mid-turn position, the particle signals that the speaker moves on to a new sequence after she herself ended the previous one. ἀτάρ may for instance occur after an answer to a question, [82] or after an assessment of the previous turn. [83] An example from Euripides Hecuba is shown in (t11). In this scene a servant comes to Hecuba, bringing her the corpse of her son Polydorus. Hecuba does not know that he is dead, however, and thinks she sees the body of her daughter Polyxena. Hecuba gives an assessment of the servant’s words, and then goes on to ask her a question—a first pair part that opens a new sequence. ἀτάρ signals that the upcoming words will not be part of the assessment anymore, but the start of a new action. With this ἀτάρ question Hecuba turns her attention towards the tableau she sees in front of her. This quick switch from her brief and dismissive assessment to her naive question underlines her ignorance about Polydorus’ fate.

§40. My analysis of ἀτάρ as marking a switch to a new sequence within a turn suggests that a widely accepted reading in Aristophanes Birds should be revised: Almost all editions read ἀτάρ in 69 in turn-initial position, [84] but as we have seen this is an unlikely choice in view of the fact that ἀτάρ hardly ever occurs in this position. Moreover, manuscript R, the only one which transmits the form ἀτάρ instead of ἀλλά [85] —all editors rightly adopt the former as lectio difficilior—is also the only one which has no speaker change here. [86] It is therefore better to keep the speaker in 69 who also uttered 68 (whether this is Peisetaerus or Euelpides). [87] ἀτάρ would then mark, as it normally does, a turn-internal switch from a second pair part (in this case an answer) to a new first pair part (in this case a question).

§41. Going on to a different adjacency pair, that of the summons and answer, we find a construction in its first pair part that is never accompanied by a particle. This is οὗτος in its use as a summoning expression: [88] With οὗτος Peisetaerus demands the attention of Euelpides, who gives a reaction to indicate that he is listening. While οὗτος in its summoning function does not always constitute a turn all by itself, and is sometimes followed by σύ in the same discourse act, [89] vocative οὗτος is never accompanied by a second-position particle. [90] It thus differs from the vocative pronoun σύ, which regularly precedes a δέ. [91] If a particle such as γε or δέ does follow οὗτος, then οὗτος is always used in its more common function as a nominative third-person demonstrative pronoun, rather than as a vocative expression. [92]

§42. Let us move on to a particular second pair part: the answer to a question. Numerous answers in the corpus share a distinctive trait: they lack any turn-initial particles or other turn-initial contextualization cues. [93] One of the numerous instances is the messenger’s answer in this scene from Sophocles Antigone: After the messenger’s announcement of bad news, Creon asks for clarification of the disaster, to which the messenger gives a straightforward answer (1282). [94] Even without a turn-initial contextualization cue such as a particle, the connection between the two turns of the adjacency pair is clear because of the function of the second one as answer to the first. In other words, the build-up of a dialogue in adjacency pairs has an influence on the linguistic form of this second pair part, i.e. without turn-initial contextualization cues. [95]

§43. As described in §§16-17 above, adjacency pairs may be expanded with pre-, insert, and/or post-expansions to form a complex sequence. Pre-expansions, which project specific adjacency pairs, regularly feature the particle oὖν: Agamemnon has been arguing with Teucer about the possible burial of Ajax, and Odysseus has just arrived to help settle the argument. After the sequence consisting of Odysseus’ question and Agamemnon’s answer in 1325-1327, Odysseus intends to give his friend advice. He does not give his suggestion directly, however, but first inquires about his right to speak (1328-1329). Since the conversation’s further development depends on Agamemnon’s answer to this inquiry, this question and its answer can be called a pre-expansion. [96] The particle οὖν in general marks an inferential link to the preceding discourse, [97] as well as the start of a new move, a discourse unit larger than one act. [98] In this case, the interrogative nature of the turn is already enough to signal the start of a new adjacency pair, as questions are always first pair parts. The presence of οὖν therefore projects an even bigger move than just a new sequence, that is, a new expanded sequence. οὖν implies that the speaker’s current action requires more words than the current turn, which can consequently be interpreted as a pre-expansion (or sometimes an insert expansion). [99]

§44. The first pair part of Odysseus’ base adjacency pair—i.e. his advice—contains enclitic νυν in turn-initial position (1332). This particle does not mark the start of a new move, as οὖν, but has a backward-oriented force only. νυν marks a logical connection between its host act and the preceding discourse: the previous utterances, in this case the pre-expansion, form the necessary background for uttering the current turn. [100] As Finglass 2011 ad loc. remarks, with νυν “Odysseus emphasises that he will hold Agamemnon to his word.” This “word” is what Agamemnon has just given in his response to Odysseus’ preliminary question. Pre-expansions and insert expansions by definition deal with matters on which the decision to utter or continue the base sequence depends; νυν is therefore particularly at home at the start of such subsequent base sequences. [101]

§45. A common way to start a pre-expansion is to ask the addressee about certain knowledge (compare the English “you know what just happened?” and similar pre-expansions). If he turns out already to know, the speaker will not start the base adjacency pair in the way she had planned it. [102] The associations of both “do you know” questions and the particle oὖν with pre-expansions lead to the regular occurrence of οἶσθ’ οὖν in such environments. [103] In the following passage from Aristophanes Knights, the playwright makes fun of the conventional sequential structure: The character Demos immediately understands that the sausage seller’s “question” in 1158 is actually an announcement of a directive. [104] “Do you know what you should do?” is a rather petrified form of pre-expansion, since the speaker does not really expect the addressee to already know the upcoming advice. [105] Demos’ reaction humorously makes this discrepancy between the pre-expansion’s form and function explicit: he does not answer the question (note the turn-initial δέ), but dryly remarks that the sausage seller will tell him the advice anyway.

§46. Insert expansions are most easily discerned when a question or order (a first pair part) is followed by another question instead of the expected response (a second pair part). [106] The insert question does not necessarily contain a turn-initial particle, since its nature as a question [107] already makes it clear that it forms a new first pair part. [108] However, turn-initial δέ helps clarify the signal that an insert expansion has started, especially in the construction τί δέ. We have already seen (§38) that a turn-initial δέ after a question makes it clear that the turn is not a straightforward answer to that question. With τί δέ in such a context, a speaker indicates that she cannot yet answer because some preliminary issue first needs to be clarified. [109] Here is an example. [110] At 478-480, Socrates produces a request for information: a first pair part making relevant the provision of that information by Strepsiades. Instead of giving the expected answer, Strepsiades responds by asking a question of his own, thereby beginning an insert sequence. The particle δέ in 481 marks the turn as a new step and not the expected response. The construction τί δέ as a whole suggests that it is a request for certain additional information, the lack of which motivates the refusal to answer the question.

§47. Since an insert expansion often starts with a question, also a question particle such as ἦ may begin insert expansions, as in (t18). Ajax does not answer Athena’s question in 103, but instead asks for clarification. According to Garvie 1998 ad loc., this is a “predictable” reaction the hero would have “to the name of his enemy.” Finglass 2011 ad loc. similarly remarks that “Ajax’s counter-question indicates his contempt for Odysseus.” Ajax has clearly understood to whom Athena is referring, but does not agree on her manner of referring to him: because of his hatred for Odysseus, Ajax avoids direct mentioning of Odysseus’ name or father. [111] The particle ἦ marks the turn as a request for clarification, and simultaneously reflects Ajax’ emotional involvement. [112] If such a question is found after another question, where an answer is expected, the ἦ turn can be interpreted as starting an insert expansion. [113]

§48. Post-expansions, finally, seem generally to lack a turn-initial particle. Such turns are often sequence-closing thirds, in the form of assessments of the second pair part just received, such as an answer to a question. [114] As argued in section 4.5 below, there is a connection between such evaluating turns and their starts without a particle or other linguistic contextualization cue. Post-sequences can also be opened with a question that is prompted by a preceding answer. [115]

§49. Earlier I noted that most first pair parts have a preferred and a dispreferred response. The latter is usually marked in some way, whereas preferred responses tend to be more straightforward in form. [116] In tragic and comic dialogue, certain turn-initial particles and particle combinations fit the context of preferred responses, others that of dispreferred ones.

§50. A common case of preferred response is an information-providing answer to an information-seeking question. It has been argued in §42 above that such answers often lack turn-initial particles or other linguistic contextualization cues. Since an answer is the normatively expected response to a question, no specific signal is needed to mark the upcoming turn as such. Similarly, preferred responses to requests and offers, i.e. turns expressing compliance and acceptance, may also start without a particle. Such responses often signal their connection to the speech situation by starting with a verb in the first person, which indicates that the speaker is obeying the request or accepting the offer. [117]

§51. In some cases preferred responses do contain turn-initial particles. Descriptions of compliance after a command can be preceded by the particle combination καὶ δή, as in (t19), in which the chorus of suppliants expresses their obedience to Pelasgus, the king of Argos, who has asked them to leave their boughs on the altar: [118] καὶ δή in the chorus’ turn can be connected to their (immediate) obedience, which they explicitly describe with the first-person present form λείπω. [119] The particle καί indicates a link to the preceding co-text or context, and a zooming in on something specific, whereas δή in drama is regularly associated with referring to perceivable elements. [120] Together the particles work as a cluster, marking a specific event or place as clearly and immediately perceivable. [121] If a speaker starts to carry out a requested action, καὶ δή is thus a fitting signal to draw attention to this obedience. Wecklein 1902 ad loc. adds that καὶ δή is like ἰδού. This latter expression, sometimes similar to English okay, is another explicit indication of a preferred response to a directive, besides explicitly describing that the request is being carried out. [122]

§52. The position of καὶ δή in the turn and sequence makes a difference in its function. In modern languages, too, certain words work differently depending on their placement in a conversation—English oh, for example, does one thing when uttered at the start of an answer (a second pair part), another in an expansion after an answer (a first pair part). [123] Likewise, the function of the polyfunctional discourse marker okay depends partly on its position in a turn. [124] The particular function of καὶ δή connected to expressing compliance thus applies only when the combination appears in turn-initial position in a second pair part. [125]

§53. If a speaker cannot or does not want to provide an answer to a question, grant a request, or otherwise utter a preferred response to a certain first pair part, the response tends to be formally marked. Dispreferred responses are less straightforward in form in English conversation, as pointed out in §18 above. Speakers of dispreferred responses tend to start speaking after a pause, use turn-initial discourse markers, be indirect in their formulation, and give accounts for why they do not answer, grant, accept, or obey.

§54. In the stylized discourse of tragic and comic dialogues, we cannot identify pauses between turns, but we do find dispreferred responses with justifying accounts. An example is the servant’s reply to Peisetaerus’ request in (t20). The hoopoe’s servant feels compelled to explain why he does not want to comply with Peisetaerus’ request: he does not want to wake up his master. The turn-initial ἀλλά signals a correction or switch concerning the explicit or implicit content of the previous turn. [126] In this case, the imperative κάλεσον implies that the servant would carry out the calling, or that there would be no obstacles to do so. The turn starting with ἀλλά corrects such implications, thereby conveying that the servant does not obey. [127]

§55. μέν can signal a dispreferred response in answers to questions, by suggesting that the answer is not straightforward. Consider the start of Ismene’s answer to Antigone’s question: Ismene could simply have answered “no, I don’t know,” but instead she elaborates on what she does know. As Jebb 1888 and Griffith 1999 ad loc. point out, the emphatic position of ἐμοί and the presence of μέν imply a contrast between Ismene herself and unspecified others. While this implication may be present, μέν at the same time conveys that its own discourse act is not all that the speaker wants to say. [128] In this way the particle indicates that the answer, or at least its first act, is insufficient, incomplete, or different from expected. We may compare this use of μέν to English turn-initial well in answers. [129]

§56. Different linguistic forms, then, tend to introduce preferred and dispreferred responses. Preferred responses to questions, directives, and offers often start without any contextualization cue. A directive may elicit a turn-initial καὶ δή or ἰδού from its addressee, who thereby draws attention to his visible compliance. Dispreferred responses to various first pair parts may start with ἀλλά; those reacting to a question sometimes contain turn-initial μέν.

§57. According to CA, it is possible to identify linguistic constructions that regularly perform particular actions. [130] We can also correlate certain actions with the presence or absence of certain particles.

§58. The particle τοι, for example, works to further a speaker’s persuasive ends. [131] The particle’s function is to signal an appeal to the addressee, who is strongly encouraged to take note of, and believe, the statement being uttered. [132] For this function the particle’s position in a turn does not make a difference. Bäumlein 1861:239 and Denniston 1950:539 note that τοι is mainly used in assertions, and give as paraphrases German “sag’ ich dir,” “darfst du glauben,” and English “you know,” “I tell you.” [133] τοι has a frequency of about 0.1% of all words in the four dramatists: less than, for example, δή or oὖν (roughly 0.2% on average), but more than for example δῆτα (roughly 0.05% on average). [134] The quite specific pragmatic meaning of τοι makes it a relatively marked, and therefore infrequent form of expression. [135]

§59. Appealing an addressee is particularly appropriate to turns that are meant to persuade. In (t22) the slave Andromache tries to persuade Hermione, the wife of her master Neoptolemus, that Andromache herself is not the cause of Hermione’s childlessness, as Hermione had angrily claimed. Rather, Andromache argues, Neoptolemus does not like Hermione because of her arrogance and temper. Although Andromache knows that she does not have much chance to persuade this angry woman, she nevertheless urges Hermione, with τοι, to believe her statement in 212. [136] “It is because of this that your husband hates you” is one of the most important points in Andromache’s argument: if it is really Hermione’s own fault that Neoptolemus dislikes her, then she has no reason to hate his concubine. The use of the particle, then, can be connected to the action that the turn is intended to perform.

§60. Aristophanes also uses τοι in this addressee-appealing way in assertions with persuasive purposes: The use of τοι in Lysistrata’s utterance of 3,5 lines (56-59) is similar to the tragic examples: the speaker encourages the addressee, her neighbor Calonice, to accept her statement. [137] In this case the speaker’s persuasive purposes do not reach beyond the single statement, as they may do in a long tragic monologue.

§61. In a few cases in tragedy and comedy τοι is used as as a general intensifier and attention-getting device, unrelated to making a statement credible. An example from Aristophanes Wealth may clarify the difference. [138] In this other, less frequent use τοι is also connected to the action performed by the host turn: the interpretation is appropriate when τοι is used outside of assertions, or when the speaker describes her own action of calling someone. In (t24) the god Hermes has just secretly knocked on the door; Cario opens it and does not immediately see him. Then Hermes makes himself visible: Hermes’ utterance at 1099bis-1100 is a multi-act move that aims to attract Cario’s attention. That is, the move, including the discourse act σέ τοι λέγω, does not state anything that the addressee is encouraged to “believe.” σέ τοι λέγω approximates the attention-getting use of οὗτος, found in 1100bis. [139] In a move such as the one by Hermes the force of τοι needs to be interpreted as weaker than in assertions. It does still appeal to the addressee: not to believe anything, but simply to pay attention.

§62. Another particle that is connected to the actions performed by its host turn is γε. It is often found at the beginning of answers to questions. The questioner has implied, by the very action of asking, that he is not yet fully aware of the answer; therefore the answerer may feel the need to highlight the most important part of her answer with γε. Consider the following example from Euripides Medea: The first discourse act of Medea’s answer in 698, μέγαν γ’ ἔρωτα, picks up the construction with ἐρασθείς from the previous turn. [140] As Mastronarde 2002 ad loc. notes, γε is common in such resonating answers. [141] With the particle, Medea emphasizes μέγαν ἔρωτα, thereby inviting the addressee (Aegeus) to infer a contrast between this element and others she does not explicitly state (such as a better reason her husband might have had to leave her). [142] In this case, the emphasis thus created leads to a sarcastic nuance: later in the dialogue Medea shows her conviction that Jason’s great “passion” did not concern the other woman, but the royal power he would receive through his new marriage. [143] The function of γε to emphasize one element by invoking implied others makes it suitable for the start of answering turns.

§63. Turn-initial γε is also regularly found in turns performing a stancetaking. This common action in talk-in-interaction involves a speaker evaluating something, and thereby positioning herself and (dis)aligning with others. [144] Stancetaking is pragmatically close to answering: after all, when asked a question, a speaker gets the opportunity to express her own view on something. [145] In the following Aristophanic passage, Dionysus assesses Heracles’ suggestion of a route to the Underworld: By using γε, Dionysus commits himself emphatically to the assessment ψυχράν, “cold” (note the exclamation mark in Henderson’s translation). This stancetaking turn functions as a second pair part, indirectly rejecting Heracles’ suggestion in 123-124 after the insert expansion in 124bis-125. γε at the same time implies a contrast between its host act and an implicit alternative, such as, in this case, a positive adjective that would convey acceptance of the suggested route. [146] The function of γε, then, is not only appropriate to answers, but also to assessments.

§64. It is not surprising that we often find γε in the context of stancetaking: this communicative action makes explicit a subjective—and sometimes particularly emotional—view of the speaker. [147] γε is comparable to the prosodic prominence rendered by an exclamation mark: a verbal equivalent, we can say, of banging one’s fist on the table, or of stamping one’s feet. [148] It is noteworthy in this respect that turn-initial γε is very rare in questions. [149] Such “stamping” of one’s words is appropriate when expressing an opinion or giving an answer that is deemed highly relevant, but less so when asking a question, where the speaker tends to primarily ask the addressee for a certain response.

§65. Turn-initial interjections are also related to stancetaking: they indicate an emotional reaction to the previous turn. [150] Interjections usually form a separate discourse act on their own, without any accompanying particles. [151] An example is οἴμοι τάλας by Oedipus: The turn starting in 744 is a post-expansion after a question-answer pair: a natural position for expressing a stance about the received answer. With the interjection, Oedipus directly expresses his emotional reaction to Iocaste’s answer. As opposed to a stancetaking response containing an adjective or adverb assessing the evaluated object (possibly followed by γε), an interjection indicates more focus on its speaker. [152]

§66. Stancetaking turns can also start without any particles or other linguistic contextualization cues. In these cases the speaker expresses an opinion without highlighting one specific part of it (as with γε) and without making her emotions explicit (as with an interjection). An example is the last turn by Medea in the following passage: [153] Medea’s request for the oath and Aegeus’ compliance (after an insert expansion in which he asks for more details) are followed by Medea’s assessment of Aegeus’ oath in a sequence-closing third: [155] ἀρκεῖ, “it is good.” Because the assessment is sequentially dependent on what preceded it, its linguistic form does not need to tie the turn explicitly to its co-text and context right at the beginning.

§67. We also find several borderline cases of stancetaking turns that contain a contextualization cue as their second or third word, but still within the first discourse act. Since these cues often form a syntactic construction with the turn’s very first word, their turns appear very similar to ones that are immediately contextualized. That is to say, whereas I employ a narrow definition of “turn-initial position” in this chapter, [156] I am aware that the linguistic reality is more flexible. Consider the following example: Chremylus’ assessment καλῶς λέγεις of Poverty’s answer (a sequence-closing third) is contextualized by referring explicitly to the addressee. The verb phrase as a whole can probably be felt as a turn-initial contextualization cue, even though the second-person verb form is not the turn’s very first word. [157]

§68. Turns without turn-initial contextualization cues may take on a particularly important or formal character. A turn’s gnomic nature, content of special significance for the speaker, or uncommon words or constructions may help create this impression. [158] The five quotes from tragedy in Aristophanes Frogs without turn-initial contextualization cues probably have a formal ring to them as well. [159] Consider the following example from Aeschylus Agamemnon: [160] Clytaemnestra reacts to Agamemnon’s worries about spoiling the purple fabrics (expressed in lines 948-949): the sea will never cease to abundantly supply purple, and the royal family is rich enough. It is not immediately clear, however, that this turn responds to Agamemnon’s earlier words, since other remarks by him came in between, and the relation between the turns is not made explicit. Instead, Clytaemnestra “begins in a tone of magnificent emphasis,” as Fraenkel 1950 ad loc. notes. He does not specify which features of her utterance convey this tone; I argue that the start without contextualization cues plays a role. Besides conveying a sense of formality, ἔστιν here also alludes to the use of this word at move beginnings in Homer. [161]

§69. The pragmatic analysis of linguistically uncontextualized utterance starts can throw light on textual problems involving the addition or removal of a turn-initial particle. In the following passage from Aristophanes Lysistrata, several editors add a turn-initial δέ, while others retain the reading of the manuscripts, which show that the turn starts without a contextualization cue. [162] The desperate Athenian Cinesias tries to convince his wife Myrrhine in this scene to end her sex strike, which was contrived by the women in order to force the men to make peace. Cinesias introduces his most important point in lines 898-899. While a turn-initial δέ would be appropriate here—it often introduces new questions in a series (see §§34-35 above), and it is very frequent in turn-initial position in Aristophanes (see Table 1 in the appendix below)—certain linguistic features better support the reading of the manuscripts. The turn contains the unusual word ἀνοργίαστα, “uncelebrated,” in its first occurrence in extant Greek literature. [163] The utterance would have a more humorous effect if a husband’s complaint about “Aphrodite’s holy rites” is presented as something formal, as the remark would exploit the incongruity of speaking in a “high” style in a “low” context. To add δέ to this line would spoil an opportunity for a joke. [164]

§70. In general, this section on connections between particle use and the actions performed by turns has shown that turns starting with γε are pragmatically similar to those without any turn-initial contextualization cue: both are mainly used for answering and stancetaking. Indeed, γε does not directly mark a relation between its turn and the previous discourse—its scope does not exceed its host act [165] —but highlights a specific element by implicitly contrasting it to a different element. The actions of answering and stancetaking both tend to be uttered in non-initial positions in a conversational sequence: the preceding discourse sets up the relevance of an answer or (usually) a stancetaking; therefore their relationship to the previous turns does not need to be made explicit.

§71. This chapter has argued that CA illuminates particle use in tragedy and comedy. Despite their stylized language, the plays still reflect many practices of everyday spoken conversation. Just as in conversation, in Greek drama speakers often use special signals to hold the floor (§§28-30). Furthermore, turns relate to each other either by initiating (first pair parts) or by reacting (second pair parts); the pairs of turns thus formed can be structured in a series, and expanded with preliminary, intervening, or appending material (§§32-48). Again, just as in conversation, responses that fit the goal of the preceding turn best (preferred responses) tend to start with different constructions than responses that were not called for (dispreferred responses; §§49-56). Finally, speakers habitually perform certain actions such as answering and stancetaking by using certain linguistic constructions (§§57-70). Most important, it has been shown that the use of particles is sensitive to the interactional aspects that are clarified by CA. That is to say, particles reveal how turns relate to each other and to the structure of an ongoing interaction.

§72. By taking into account the interactional context surrounding every turn, a CA approach helps us understand why different particles are appropriate in different communicative contexts. For example, one of the known functions of τε is to mark two elements as closely connected; paying attention to the structure of a conversation makes us understand why this particle is highly infrequent in turn-initial position, and why it does occur there in rare cases (§27). The particle μέν may project an upcoming move of multiple discourse acts; speakers exploit this projecting function within turns (§§28-30) as well as within series of similar turns (§33). δέ generally indicates a new step in the discourse; CA makes it clear why this kind of signal is relevant in a series of questions, in response to news, or when a speaker ignores a previous turn (§§34-38). Similarly, we understand better how καί’s zooming-in force can interact with its turn-initial position (§36). ἀτάρ has a specialized function in drama, confined to non-initial position in turns (§§39-40). ἦ, οὖν, and καὶ δή all have their own pragmatic functions; CA allows us to clarify how these functions interact with certain positions within a sequence or within a turn (§43, §47, and §§51-52). τοι and turn-initial γε have been shown to be connected to the communicative actions that their turns perform (§§58-70).

§73. The following is an overview of all turn-initial expressions in the twelve plays of my corpus. [166] The table documents the frequency of several particles at the start of turns, and the differences in distribution among the authors. It also indicates what other linguistic forms tend to be found in turn-initial position, along with their frequencies. In all four authors turns tend to be explicitly contextualized at their start in some way. Most often speakers do this by referring to the speaker or addressee with a verb form, vocative, or pronoun, by projecting the nature of the upcoming turn with a question word or negation, or by indicating an emotional reaction with an interjection. As for turn-initial particles, δέ and καί are the most frequent ones, not surprising in view of their overall high frequency. [169] For ἀλλά, γε, γάρ, δῆτα, and μέν, the frequencies vary quite widely across the authors. As also observed in III.2 on discourse patterns, Aeschylus tends to avoid γε and δῆτα. He is, however, relatively fond of turn-initial μέν. [170] Euripides is the exception in his avoidance of turn-initial ἀλλά. [171] Furthermore, in his plays overall fewer turns start with a particle, and we find a higher frequency of non-contextualized starts. A more detailed comparison of these turn starts in Euripides versus the other authors might reveal whether he uses them in different ways. [172] As for turn-initial interjections: they are typical of tragedy, while comic characters more often start speaking with a swearing expression. [173] The uses of and differences across these turn-initial expressions other than particles can, I expect, be fruitfully explored in a CA approach as well.

§74. Concerning the total number of turns per play, comedies feature more turns than tragedies, and the plays of Aeschylus contain the fewest turns. The following diagram shows that this holds true for all individual plays. In my corpus, there are many more turns in each play of Aristophanes than in any individual tragedy. These quantitative differences reflect different conversational styles. The three comedies are on average longer than the tragedies, [174] but comic turns also tend to be shorter than tragic turns. The average turn lengths are 6.7 lines per turn in Aeschylus; 4.3 lines per turn in Sophocles; 5.2 lines per turn in Euripides; and only 2.3 lines per turn in Aristophanes. [175] Comic speakers more often “interrupt” each other: a conversational style leading to Slings’ remark (2002:101) that most Aristophanic speakers are characterized by “aggressiveness.”

§75. In tragedy, the plays of Aeschylus have the fewest speaking turns: only 190 on average. This is primarily because the long lyric sections make up a large part of Aeschylean plays. These three tragedies also contain fewer and shorter stichomythic passages than those by Sophocles and Euripides. The higher relative frequency of turn-initial interjections in Aeschylus is due to the extremely high number in Persians, where lamenting—with which many interjections tend to be associated—is one of the main communicative actions of the play. Furthermore, in this sample Sophocles Oedipus King, with 441 turns, appears as the tragedy of dialogue and of stichomythia par excellence: it has many more turns of speaking than any of the other eight tragedies. We can connect the high number of turns to the play’s plot: Oedipus’ quest for information is enacted in his many stichomythic exchanges with other characters. [176] Not surprisingly, this play also has a higher frequency of turn-initial question words than the other tragedies (51, whereas the other tragedies have no more than 34 per play). [177]