Zakowski ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 2014 Samuel Zakowski εἰπέ µοι as a Parenthetical: A Structural and Functional Analysis, from Homer to Menander Samuel Zakowski N RECENT YEARS a number of important papers have ap- peared on the grammaticalized Ancient Greek imperatives φέρε, ἄγε, and ἴθι.1 ‘Grammaticalized’ in this sense usually refers mainly to the fact that these expressions have become fixed in form and that their original meaning has been bleached in favor of another, more abstract meaning. There are, however, similar expressions which have not received the same kind of attention. In this article I analyze the expression εἰπέ µοι, which, we shall see, checks the same boxes as, for example, ἄγε as a grammaticalized imperative. Although εἰπέ µοι appears far less frequently than these other grammati- calized imperatives in the corpus under consideration, there is still enough material to provide an intriguing window on its structural and functional characteristics. This corpus consists of a representative sample of Classical Greek, both prose and poetry: the extant works of Plato, Demosthenes, Aristophanes, and Menander. Homer will be referred to as well, as a starting point for the development of εἰπέ µοι. In the first section, I introduce the problem, somewhat pre- theoretically, by indicating why εἰπέ µοι could be considered a 1 E.g. M. Biraud, Les interjections du théâtre grec antique: Etude sémantique et pragmatique (Louvain 2010) 25–42, on φέρε, ἄγε, and ἴθι; C. Fedriani, C. Ghezzi, and C. Molinelli, “Constraining Pragmaticalization: Paths of De- velopment from Lexicon to Discourse,” paper presented at the conference New Reflections on Grammaticalization 5 (Edinburgh, 16–19 July 2012), on ἄγε. I 158 ΕΙΠΕ ΜΟΙ AS A PARENTHETICAL ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 grammaticalized expression. In section two, these features are treated more in detail: εἰπέ µοι’s structural and syntactic traits are analyzed, and I propose, tentatively, a possible diachronic path of how εἰπέ µοι came to function as it does in the authors under consideration. The third section focuses on the semantics and pragmatics of εἰπέ µοι, i.e. on what εἰπέ µοι actually contributes to the discourse. In the fourth section I present my conclusions. 1. Outline of the issue Consider the following example: (1) Dicaeopolis is threatening to ‘kill’ his ‘hostages’, which are actually lumps of charcoal; the chorus reacts (Ar. Ach. 328–329): εἰπέ µοι, τί τοῦτ’ ἀπειλεῖ τοὔπος, ἄνδρες δηµόται, τοῖς Ἀχαρνικοῖσιν ἡµῖν; Tell me, what does this threat mean, citizens, to us Acharnians? There is obviously something strange going on here. While εἰπέ is a second person singular imperative, the leader of the chorus is talking to multiple persons (ἄνδρες δηµόται): we would have expected to see εἴπετέ µοι. Like φέρε, ἄγε, and ἴθι, then, εἰπέ seems to occur almost by default in the second person singular. This is only one of a number of ostensibly anomalous char- acteristics of εἰπέ µοι: (2) Glycera has fled from Polemon’s house. Polemon’s slave, Sosias, is accusing the neighbours of having taken her (Men. Peric. 387–388): πρὸς τίν’ οἴεσθ’, εἰπέ µοι, παίζειν; Who do you think, tell me, you’re trifling with? As in (1), the speaker is again talking to multiple persons (οἴεσθε) using the second person singular εἰπέ. Yet there is an additional wrinkle here: εἰπέ µοι is located clause-internally without being a syntactic part of that clause. This is odd, to say the least: under normal circumstances, separate (non-em- bedded) clauses occur consecutively and do not violate each SAMUEL ZAKOWSKI 159 ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 other’s linear order.2 If this were not peculiar enough, there is one more aspect which immediately catches the eye in some instances: (3) Demosthenes is presenting his fellow Athenians with a stark choice between taking up arms against Philip and the following (4.10): ἢ βούλεσθ’, εἰπέ µοι, περιιόντες αὑτῶν πυνθάνεσθαι, “λέγεταί τι καινόν;” Or are you happy, tell me, while you are running around, to in- quire from each other, “What news is there?” In this case, the speaker is addressing more than one hearer (βούλεσθε); in addition, εἰπέ µοι is located clause-internally. Moreover, there is no interrogative present, as there is in both (1) and (2). In (1), one could make the argument that εἰπέ µοι should be considered the main clause which introduces an indirect question starting with τί. In (2), this is also a possible analysis, although εἰπέ µοι’s clause-internal position seems to render this view extremely unlikely. In (3), however, there is no doubt about εἰπέ µοι’s position in the syntax of the sentence: there is no interrogative, which means that it is entirely sep- arate from the clause in whose linear order it intervenes. These three examples raise many questions. What are the implications of εἰπέ µοι occurring exclusively in the second person singular? Is there a functional difference between its appearance in clause-initial, clause-internal, and clause-final (which occurs as well) position? What are we to make of this positional alternation? What are the ramifications of its being able to appear asyndetically with clauses from which it is syn- tactically detached? These considerations form the basis of the following inquiry. 2. A structural-syntactic analysis of εἰπέ µοι In this section, εἰπέ µοι will be analyzed from a structural- syntactic point of view. I will focus on the structural gram- 2 S. C. Dik, The Theory of Functional Grammar I The Structure of the Clause (Berlin 1997) 402. 160 ΕΙΠΕ ΜΟΙ AS A PARENTHETICAL ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 maticalization criteria it complies with; how it came to be a syntactically independent expression; and its positional mobility in the corpus under consideration. Since the latter two aspects are closely intertwined, they will be grouped together. 2.1 Grammaticalization and εἰπέ µοι: the structural criteria We have seen that εἰπέ µοι only occurs in the second person singular, even when the speaker is addressing multiple persons. This is a common tendency for grammaticalized elements, and is known more technically as ossification of form.3 This raises the question, which other grammaticalization criteria may εἰπέ µοι adhere to. The following seem to be indisputably appli- cable: 2.1.1 Decategorialization Grammaticalized items lose their ability to take complements or to be modified by other elements.4 Contrast (4) with (5): (4) Crito is urging Socrates to flee his prison and not take into ac- count the risks involved (Pl. Crt. 44E2): τάδε δέ, ὦ Σώκρατες, εἰπέ µοι. ἆρά γε µὴ ἐµοῦ προµηθῇ καὶ τῶν ἄλλων … This, Socrates, you have to tell me. You are not considering me and the others, are you …? (5) Socrates is questioning Euthydemus and Dionysodorus on knowing and not-knowing (Pl. Euthyd. 293E5–6): 3 See L. J. Brinton, “Processes Underlying the Development of Pragmatic Markers: The Case of (I) say,” in J. Skaffari et al. (eds.), Opening Windows on Texts and Discourse of the Past (Amsterdam/Philadelphia 2005) 279–299, at 291; E. C. Traugott, “The Role of the Development of Discourse Markers in a Theory of Grammaticalization,” paper presented at ICHL XII, Man- chester 1995 (updated version of 1997: www.wata.cc/forums/uploaded/ 136_1165014660.pdf) 2. Cf. also Biraud, Les interjections 27; and Fedriani et al., “Constraining Pragmaticalization,” on ossification with other verbal ex- pressions in Ancient Greek. 4 Brinton, in Opening Windows 291, and The Comment Clause in English: Syntactic Origins and Pragmatic Development (Cambridge 2008) 51; C. Company Company, “Subjectification of Verbs into Discourse Markers: Semantic- pragmatic Change Only?” Belgian Journal of Linguistics 20 (2006) 97–121, at 100. SAMUEL ZAKOWSKI 161 ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 εἰπέ µοι, σφὼ οὐχὶ τὰ µὲν ἐπίστασθον τῶν ὄντων, τὰ δὲ οὐκ ἐπίστασθον; Tell me, do you not both know some things of reality, and others things you do not? In (4), τάδε is a direct object to εἰπέ—as opposed to (5), where εἰπέ (µοι) is syntactically detached from the rest of the clause and does not take a direct object. The contrast between (4) and (5) also applies to the following criterion. 2.1.2 Divergence Grammaticalized items coexist with non-grammaticalized forms of the expression.5 In (5), εἰπέ µοι is syntactically de- tached from the clause proper and pertains to multiple persons as a singular form (i.e., is ossified). In (4), by contrast, εἰπέ µοι takes a direct object and pertains to one person (Socrates). In other words, εἰπέ µοι functions as a proper main verb of the clause in (4), whereas in (5) it is an extra-clausal, syntactically optional item which is asyndetically connected to the main verb of the clause. Both examples are from Plato, which points to their coexistence. There are other grammaticalization criteria which pertain to εἰπέ µοι, but they are more semantic than structural in nature. Hence they will be treated in §3, where the semantics of εἰπέ µοι are discussed. In this sense, then, εἰπέ µοι can be regarded as grammati- calized: its form is ossified, it is decategorialized, and it coexists with other, non-grammaticalized variants. However, this does not explain how εἰπέ µοι came to be syntactically independent from the clause with which it occurs; nor does it shine any light on the reasons behind its positional mobility. These two features are closely connected,6 and so are considered jointly in 2.2. 5 P. J. Hopper, “On Some Principles of Grammaticalization,” in E. C. Traugott and B. Heine (eds.), Approaches to Grammaticalization (Amsterdam/ Philadelphia 1991) I 17–36, at 22; Brinton, in Opening Windows 292. 6 Brinton, The Comment Clause 8. 162 ΕΙΠΕ ΜΟΙ AS A PARENTHETICAL ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 2.2 Syntactic independence and positional mobility: the matrix clause hypothesis In (3), εἰπέ µοι was unquestionably syntactically independent of the rest of the clause. In (1) and (2), however, matters were not so clear-cut: it could be argued (especially for (1)) that εἰπέ µοι was a matrix clause to which the question introduced by τί formed a complement. The difference between (1), (2), and (3) forms the basis for the diachronic development of grammati- calized εἰπέ µοι which is traced in this section. This ties in nicely with the question of its positional mobility, as we shall see: its position can furnish hints as to whether or not εἰπέ µοι is a matrix clause in a given situation. Describing diachronic developments for Ancient Greek is risky at best, and often downright impossible. A host of ma- terial is unavailable to us that could either offer support to or completely invalidate many claims. The only route available to us is to work with the evidence we do have and attempt to sketch—tentatively—trajectories of linguistic evolution. We can see, synchronically, that εἰπέ µοι had come to be grammati- calized by Aristophanes’ time. Information about its diachronic development into a grammaticalized element, however, is highly obscure. The same goes for the syntactic detachability7 of εἰπέ µοι—there is simply not enough data to form any justifiable conclusions about its evolution. That said, there is some interesting information to be gleaned from the data which is available to us.8 As pointed out above, εἰπέ µοι is syntactically non-obliga- tory, or, more specifically, optional. Consider (3) again, re- peated here as (6): (6) Demosthenes is presenting his fellow Athenians with a stark choice between taking up arms against Philip and the following 7 I use ‘syntactic independence’ and ‘syntactic detachability’ as synonyms. 8 Cf. S. A. Thompson and A. Mulac, “A Quantitative Perspective on the Grammaticization of Epistemic Parentheticals in English,” in Approaches to Grammaticalization II 313–329, at 324, for similar problems with the analysis of, for instance, syntactically detached I think. SAMUEL ZAKOWSKI 163 ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 (Dem. 4.10): ἢ βούλεσθ’, εἰπέ µοι, περιιόντες αὑτῶν πυνθάνεσθαι, “λέγεταί τι καινόν;” Or are you happy, tell me, while you are running around, to in- quire from each other, “What news is there?” The sentence in (6) would be syntactically complete without εἰπέ µοι; more specifically, εἰπέ µοι is syntactically detached from the rest of the sentence. It is an example of what Kaltenböck calls the syntactic “dilemma” presented by these types of ex- pressions, which “do not enter into any syntactic hierarchical relation with their host clause but intervene in its linear or- der.”9 The question now becomes: how did εἰπέ µοι become syntactically ‘stranded’, as it is in (6)? There are no definitive diachronic conclusions to be drawn, but both earlier stages of Ancient Greek (i.e. Homer) and synchronically alternating con- structions with εἰπέ µοι afford some clues. In Homer, εἰπέ µοι appears 8 times (all in the Odyssey). In some cases (5 of the 8), it functions as the matrix clause for embedded clauses: (7) Penelope is questioning a man who claims to have seen Odysseus twenty years earlier—of course, the man is himself the disguised Odysseus (Od. 19.218): εἰπέ µοι ὁπποῖ’ ἄσσα περὶ χροῒ εἵµατα ἕστο Tell me what sort the clothes were which he wore around his skin In (7), ὁπποῖα introduces an indirect question dependent on εἰπέ µοι: it would be ποῖα if it introduced a direct question. In the other 3 instances, µοι is not decategorialized, and hence cannot be considered grammaticalized: it takes a parti- ciple, as in (8): (8) Athena is asking Zeus if the bloodshed on Ithaca will stop after Odysseus’ massacre (Od. 24.474): εἰπέ µοι εἰροµένῃ· τί νύ τοι νόος ἔνδοθι κεύθει; 9 Brinton, The Comment Clause 10, paraphrasing G. Kaltenböck, “Charting the Boundaries of Syntax: A Taxonomy of Spoken Parenthetical Clauses,” View[z]: Vienna Working Papers 14.1 (2005) 21–53, at 22. 164 ΕΙΠΕ ΜΟΙ AS A PARENTHETICAL ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 Tell me, who am asking, what lies hidden within your mind. This last example is particularly interesting, as it marks the transition to the next point of interest: the narrow line between matrix clause and syntactically detached clause in the case of embedded questions introduced by an interrogative. Indeed, the editor’s punctuation in (8) implies that s/he assumed that the question was direct, and not indirect: it would then be translated as “Tell me, who am asking: what lies hidden within your mind?’ The sentence in (8) does not furnish any hints as to which possibility is preferable—both direct question and in- direct are entirely legitimate interpretations.10 Whereas εἰπέ µοι was undeniably a matrix clause in (7), then, matters are not quite so clear-cut for (8). The same ambiguity arises with many instances in the other authors of our corpus: (9) Apollodorus is recounting how Glaucon asked him to tell the story about the party at Agathon’s (Pl. Symp. 173A3–5): καὶ ὅς, “µὴ σκῶπτ’,” ἔφη, “ἀλλ’ εἰπέ µοι πότε ἐγένετο ἡ συν- ουσία αὕτη.” And he [Glaucon] said, “Don’t jest, but tell me when this party took place.” (10) Demosthenes has asked that a law be read that states that acts and judgments from the time of the Thirty have no legal force (Dem. 24.57): ἐπίσχες. εἰπέ µοι, τί δεινότατον πάντες ἂν ἀκούσαντες φήσαι- τε, καὶ τί µάλιστ’ ἂν ἀπεύξαισθε; Stop. Tell me, what would you all say is most frightening, after having heard this [law being read], and what would you reject above all? (11) Sostratos has just encountered his slave Getas (Men. Dysc. 553): (Σω) τί ποιεῖτ’ ἐνθάδ’, εἰπέ µοι; (Γε) τί γάρ; 10 Although the question, if indirect, could have been introduced by ὅτι in (8), τί is an equally valid option. SAMUEL ZAKOWSKI 165 ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 (So.) What are you doing here? Tell me. (Ge.) What I’m doing here? In all these examples, εἰπέ µοι either introduces (9–10) or finishes (11) an interrogative clause which starts with an in- terrogative. The question here is, again, whether those inter- rogative clauses are embedded clauses dependent on εἰπέ µοι as its matrix clause, or are syntactically main clauses, with εἰπέ µοι a syntactically detached addition. To judge from the punctuation of the several editors, εἰπέ µοι is regarded as syn- tactically detached in (10) and (11), and as a matrix clause in (9): the translations reflect this distinction, as only the clause in (9) does not end with a question mark. However, these choices seem to be somewhat random: what is the fundamental difference between (9) and (10), for in- stance? Both have εἰπέ µοι at the beginning of the sentence, followed by an interrogative and an interrogative clause. In (10), εἰπέ µοι is certainly grammaticalized (the verb φήσαιτε is plural); in (9), it is employed to address one person. However, that does not have any bearing on the interrogative clause, which is structurally near-identical in both cases. We have now encountered three possible syntactic roles for εἰπέ µοι. It can be an unambiguous matrix clause (example 7); somewhere between matrix clause and syntactically indepen- dent (9–11); and unambiguously syntactically detached (5–6). It is certainly not impossible to infer a path of development for εἰπέ µοι on the basis of extension: the inherent ambiguity be- tween matrix clause and syntactically detached expression in clauses with an embedded interrogative clause introduced by an interrogative (9–11) would then instigate the innovation of phasing in syntactically detached εἰπέ µοι with interrogative clauses which are not introduced by an interrogative (5–6). Schematically, the extension would work as follows: (12) (a) No ambiguity (as in (7)): [εἰπέ µοι]matrix clause + [indirect interrogative + indirect question]embedded clause 166 ΕΙΠΕ ΜΟΙ AS A PARENTHETICAL ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 (b) Ambiguity between (as in (9–11)) i. [εἰπέ µοι]matrix clause + [interrogative + interrogative clause]embedded clause ii. [εἰπέ µοι]syntactically detached + [interrogative + interrogative clause]main clause (c) Extension of (b.ii) (as in (5–6)) [εἰπέ µοι]syntactically detached + [direct question]main clause What happens here is that the ambiguity of instances with an interrogative clause introduced by an interrogative, like (9–11), where εἰπέ µοι can be processed as a syntactically detached item, ‘spill over’ into instances where it could not otherwise have been used. Through this extension, εἰπέ µοι can now be used asyndetically, i.e. as a syntactically detached expression, with direct questions which are not introduced by an inter- rogative—there is no longer any need to mark these questions as indirect (as in (12a)) since εἰπέ µοι is no longer felt to be a main clause. Instead, it can be considered syntactically inde- pendent. Conversely, syntactic roles are switched: the question becomes the main clause, and εἰπέ µοι becomes a syntactically stranded, optional addition.11 Moreover, εἰπέ µοι, as a syn- tactically detached item, gains in positional mobility: it can be implemented clause-initially, clause-internally, or clause-finally. As stated above, we lack the necessary diachronic under- pinnings for the hypothesis that the three syntactic roles of εἰπέ µοι are consecutive steps in its development. It is clear that εἰπέ µοι does not function as an unambiguously syntactically de- tached expression in Homer, but that it does function as such in Aristophanes. Obviously, some innovation must have taken hold in the intervening time. It is also clear that the three roles appear in tandem synchronically in Aristophanes:12 this is an 11 See Thompson and Mulac, in Approaches II 313 and 323, and Brinton, The Comment Clause 35–36, for comments on the similar development of I think; cf. J. R. Ross, “Where to Do Things with Words,” in P. Cole and J. L. Morgan (eds.), Syntax and Semantics III Speech Acts (New York 1975) 233–256, on ‘slifting’. 12 For εἰπέ µοι as an unambiguous matrix clause in Aristophanes, see e.g. Nub. 759 and Thesm. 628. SAMUEL ZAKOWSKI 167 ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 example of divergence (cf. §2.1.2), which strengthens the argu- ment that εἰπέ µοι had become grammaticalized by Classical times. But this is as far as we can go from a diachronic per- spective. There is one factor which can help us gauge whether εἰπέ µοι is a matrix clause or a syntactically detached expression in the ambiguous cases. That factor is based on its positional mobility. Contrast the following examples: (13 = 1) Dicaeopolis is threatening to ‘kill’ his ‘hostages’, which are actually lumps of charcoal; the chorus reacts (Ar. Ach. 328–329): εἰπέ µοι, τί τοῦτ’ ἀπειλεῖ τοὔπος, ἄνδρες δηµόται, τοῖς Ἀχαρνικοῖσιν ἡµῖν; Tell me, what does this threat mean, citizens, to us Acharnians? (14) Cnemon has found his slave Simice, who had dropped a bucket into a well; she is fearful of her punishment (Men. Dysc. 590): τί ποιεῖν δ’, εἰπέ µοι, µέλλεις; What, tell me, are you going to do? (15) Strepsiades has just discovered some mathematical instru- ments (Ar. Nub. 200): πρὸς τῶν θεῶν, τί γὰρ τάδ’ ἐστίν; εἰπέ µοι. By the gods, what are these things then? Tell me. All these instances involve an ambiguous εἰπέ µοι: it can be considered either a matrix clause or syntactically detached vis- à-vis the interrogative clause introduced by τί. However, if εἰπέ µοι’s position can be statistically correlated with the nature of the interrogative clause, this could aid in establishing the dis- tinction between εἰπέ µοι as a matrix clause and εἰπέ µοι as a syntactically detached expression in examples such as (13–15). Say that there is a significant correspondence between clause- initial εἰπέ µοι and interrogative clauses introduced by an inter- rogative on the one hand, and between non-clause-initial εἰπέ µοι and interrogative clauses without an interrogative on the other. In the latter case, εἰπέ µοι is unequivocally syntactically detached. If εἰπέ µοι then appeared in non-clause-initial posi- tion with interrogative clauses introduced by an interrogative (as in e.g. (14) and (15)), this would provide evidence for the 168 ΕΙΠΕ ΜΟΙ AS A PARENTHETICAL ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 hypothesis that it is syntactically detached in these instances. If εἰπέ µοι occurs in clause-initial position with interrogative clauses introduced by an interrogative (as in e.g. (13)), this would point towards a matrix clause interpretation. Author + INTERROGATIVE – INTERROGATIVE Initial Internal Final Total Initial Internal Final Total Hom. 1 (100%) Ø Ø 1 Ø Ø Ø Ø Ar. 21* (70%) Ø 9 (30%) 30 6 (75%) 1 (12.5%) 1 (12.5%) 8 Pl. 4 (100%) Ø Ø 4 12 (92.31%) 1** (7.69%) Ø 13 Dem. 5 (100%) Ø Ø 5 3 (75%) 1 (25%) Ø 4 Men. Ø 5 (55.56%) 4 (44.44%) 9 1*** (8.33%) 9 (75%) 2 (16.67%) 12 Table 1: Instances of syntactically ambiguous and detached εἰπέ µοι with interrogative clauses *There is one example in Aristophanes (Lys. 830) which is inconclusive: τί δ’ ἐστίν; εἰπέ µοι, τίς ἡ βοή; “What is this? Tell me, what is this yelling?” The question here is whether εἰπέ µοι is located clause-final (following the first question) or clause-initial (preceding the second question). I have chosen the latter option here simply because it is statistically more prevalent (and is in accordance with the editor’s punctuation), although I realize that this amounts to somewhat circular reasoning. **This instance is quite interesting (Pl. Spuria 395E7; cf. 174 below): σὺ γὰρ εἰπέ µοι, ὦ Ἐρυξία, ἀγαθὸν ἡγῇ τὸ πλουτεῖν; “You tell me, Eruxia, do you think it’s good to be rich?” The only item preceding εἰπέ µοι is the topicalized subject σύ (γάρ); both elements are separated from the question proper by the vocative. Although εἰπέ µοι is, then, located in clause-internal position, strictly speaking, this should be qualified: both σύ (as a topic which is not located in the clause proper) and εἰπέ µοι (as a syntactically detached expression) can be regarded as extraclausal, preceding the clause itself (cf. D. Matic, “Topic, Focus, and Discourse Structure,” Studies in Language 27 [2003] 573–633, at 580). Hence, εἰπέ µοι could be regarded as clause- initial in this case as well. The same could be true for the only internal –INT instance in Aristophanes (Plut. 1100). *** It is possible that this (unique) example of initial position in Menander is due to the meter (Asp. 310): µέλλει γαµεῖν γὰρ αὐτός. {Δ∆α} εἰπέ µοι, γαµεῖν; The quantity of the relevant vowel and syllable (short -ᾰ- in γαµεῖν; diphthong ει- in εἰπέ) necessitates placing εἰπέ µοι in initial position here. If this is correct, Asp. 310 cannot be regarded as a representative example for this discussion. SAMUEL ZAKOWSKI 169 ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 With the caveat that sample sizes are fairly limited, Table 1 provides an overview of the relevant numbers.13 The table should be read as follows: in Demosthenes, there are 5 in- stances of εἰπέ µοι with an interrogative clause introduced by an interrogative (+INT). In all 5 instances, εἰπέ µοι precedes the question. There are also 4 examples of εἰπέ µοι with inter- rogative clauses which are not introduced by an interrogative (–INT) in Demosthenes: in 3 instances, it precedes the question; in 1 instance, it is located question-internally. Drawing conclusions from this table is, as stated, problematic because of inadequate sample sizes. However, there are some interesting tendencies to consider. Before comparing the in- dividual authors, I will first focus on the different possible con- structions. Let us begin by looking at +INT initial position (the interrogative is in bold type): (16) Dicaeopolis is defending the Spartans against the chorus’ pro- testations—things are getting violent (Ar. Ach. 319): εἰπέ µοι, τί φειδόµεσθα τῶν λίθων, ὦ δηµόται …; Tell me, why are we sparing our rocks, fellow citizens? Although the editor has opted to regard εἰπέ µοι as syntacti- cally detached in this case (as indicated by the comma and the question mark) and, therefore, the question as direct, it is more reasonable to assume that εἰπέ µοι here is instead a matrix clause. As Houben, Fraser, and Allan argue, complement clauses usually follow main clauses.14 The question can be regarded as a complement here (it takes the place of the direct object) if it is considered an indirect question. It is more plausible to regard εἰπέ µοι as syntactically de- 13 Percentages apply to the +INT or –INT totals, not to both combined. The fragments attributed to Menander have not been taken into account here as they are often unclear. 14 J. L. Houben, “Word-order Change and Subordination in Homeric Greek,” JIES 5 (1977) 1–8, at 5; B. L. Fraser, Word Order, Focus, and Clause Linking in Tragic Greek Poetry (diss. Cambridge 1999) 253; R. J. Allan, “Clause Intertwining and Word Order in Ancient Greek,” JGL 12 (2012) 5–28, at 6. 170 ΕΙΠΕ ΜΟΙ AS A PARENTHETICAL ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 tached in +INT instances if it is located in final position: (17) Chremylus reacts to the story of an old woman about a man who is no longer friendly to her (Ar. Plut. 998): τί σ’ ἔδρασ’; εἰπέ µοι. What did he do to you? Tell me. Preposing complement clauses can be pragmatically motivated —the complement clause is highlighted vis-à-vis the main clause.15 On this view, εἰπέ µοι would seem to lose its value as a matrix clause in (17) and, instead, move towards a semantically and syntactically less important function. Note that the elision suggests that εἰπέ µοι is phonologically integrated with the question—in other words, that it is not a fully separate phono- logical sequence.16 Internal position indicates full-on parenthetical εἰπέ µοι, even in +INT examples. Example (14) is repeated as (18) here: (18) Cnemon has found his slave Simice, who had dropped a bucket into a well; she is fearful of her punishment (Men. Dysc. 590): τί ποιεῖν δ’, εἰπέ µοι, µέλλεις; What, tell me, are you going to do? Elision occurs here as well: just as in the clause-final instances, εἰπέ µοι is phonologically incorporated into the question.17 Moreover, it now violates the principle of domain integrity, which states that “[c]onstituents prefer to remain within their proper domain; domains prefer not to be interrupted by con- stituents from other domains.”18 It is a clear instance of an expression which intervenes in the linear order of a clause of 15 Fraser, Word Order 253; Allan, JGL 12 (2012) 6. 16 Interestingly, in this regard, all final +INT instances of εἰπέ µοι in both Aristophanes and Menander contain elision wherever it is possible (Ar. Thesm. 89; Men. Dysc. 553, Sam. 453). 17 The other internal +INT instances in Menander contain elision as well, wherever it is possible: Dysc. 466, Peric. 387. 18 Dik, Functional Grammar 402 (cited in Allan, JGL 12 [2012] 6). SAMUEL ZAKOWSKI 171 ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 which it does not form a syntactic part.19 In other words, it is syntactically detached from the clause in which it occurs. The –INT instances of εἰπέ µοι pose less of a problem. While the positional mobility evident in the +INT examples comes to the fore here as well, –INT εἰπέ µοι is unambiguously syn- tactically detached from the question: it does not form a syntactic part of the question, yet it is asyndetically attached to the question all the same ((3) repeated as (19)): (19) Demosthenes is presenting his fellow Athenians with a stark choice between taking up arms against Philip and the following (4.10): ἢ βούλεσθ’, εἰπέ µοι, περιιόντες αὑτῶν πυνθάνεσθαι, “λέγεταί τι καινόν;” Or are you happy, tell me, while you are running around, to in- quire from each other, “What news is there?” It should not surprise us that εἰπέ µοι demonstrates positional movability in –INT instances. In fact, this mobility can be regarded as “a consequence” of its “syntactic independence.”20 As it does not belong to the clause from a syntactic point of view, it does not occupy a set position in that clause. Con- versely, it is not a priori liable to appear in clause-internal position—it occurs clause-initially and clause-finally as well.21 What has become clear is that εἰπέ µοι is characterized by a high degree of positional mobility across different authors. In measuring those same authors against one another, however, that positional mobility should be qualified: there are con- spicuous discrepancies, especially between Menander and the other authors. Initial position occurs only once in Menander, and then probably metri causa (see 168 above); for the other, earlier, writers, it is the most common pattern. Although the 19 Kaltenböck, View[z]: Vienna Working Papers 14.1 (2005) 22; cf. Brinton, The Comment Clause 12. 20 Brinton, The Comment Clause 8. 21 Brinton, The Comment Clause 12. See e.g. Pl. Euthyd. 302B4–5 for initial position, Ar. Plut. 172 for final position. 172 ΕΙΠΕ ΜΟΙ AS A PARENTHETICAL ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 limited sample sizes do not allow for any sweeping generali- zations, the numbers do bear out that something must have happened to εἰπέ µοι somewhere between Aristophanes and Menander. In post-Menander Greek, the tendency to avoid clause-initial position seems to have become even more pro- nounced. In John Chrysostom’s homilies Contra Anomoeos, Adv. Iudaeos, and Ad populum Antiochenum, for example, where εἰπέ µοι occurs 114 times in all, it does not appear clause-initially even once. The preceding discussion indicates that a clear distinction has to be made between +INT and –INT instances of εἰπέ µοι. The –INT instances are relatively unproblematic: in these cases, εἰπέ µοι is syntactically detached from yet asyndetically connected to a direct question. In +INT instances, however, there is a certain ambiguity: because of a linguistic idiosyncracy of Ancient Greek, interrogatives can introduce both direct and indirect questions. What was needed was a yardstick against which these occurrences could be measured; εἰπέ µοι’s posi- tional mobility was considered the basis for such a yardstick. For +INT instances, I took initial position to designate unam- biguous matrix clause status; final position to designate ambig- uous status between matrix clause and syntactically detached item; and internal position to designate unambiguous syntacti- cally detached status. This distinction also ties in nicely with Biraud’s observation that the ‘interjective’, i.e. syntactically de- tached, forms of φέρε, ἄγε, and ἴθι occur clause-internally (“en incise”), and the non-interjective forms clause-initially.22 In keeping with this division, the numbers bear out that εἰπέ µοι is simply more syntactically detached in Menander than it is in the other writers under consideration. The positional mobility and increasing syntactic detachability of εἰπέ µοι suggests that it belongs to a functional class of words commonly described as ‘parentheticals’.23 As will become clear in the fol- 22 Biraud, Les interjections 27. 23 Brinton, The Comment Clause 18. SAMUEL ZAKOWSKI 173 ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 lowing section, εἰπέ µοι is parenthetical in a semantic, truth- conditional sense as well. 3. εἰπέ µοι as a parenthetical: non-truth conditionality Following Brinton,24 we can say that a parenthetical is char- acterized by at least three different aspects: • It is syntactically detached from the rest of the clause. • It is positionally mobile. • It lacks referential meaning; instead, it functions at the pragmatic plane of discourse. The first two features were discussed in the preceding section; the third will provide a springboard for the semantic analysis of εἰπέ µοι. At that point, we can also finally consider what εἰπέ µοι contributes to the utterance specifically. First, however, some preliminary remarks are in order on the term ‘parentheti- cal’. ‘Parenthetical’ in the technical sense employed here refers to those expressions which are, in Kaltenböck’s words, “a clausal category, i.e., contain a verb.”25 As opposed to Déhé and Kavalova’s more inclusive notion of the concept,26 then, and with Brinton and Urmson, who first coined the term, a paren- thetical as it is understood in this paper consists of, minimally, a verb in some shape or form.27 Examples in English include I suppose, as in “Your house is, I suppose, very old”;28 I think, as in “John came, I think, later than Sue”;29 say, as in “Say, you pro- 24 Brinton, The Comment Clause. 25 G. Kaltenböck, “Spoken Parenthetical Clauses in English: A Taxon- omy,” in N. Déhé and Y. Kavalova (eds.), Parentheticals (Amsterdam/Phila- delphia 2007) 25–52, at 47. 26 They also regard what and like as possible parentheticals, for instance: N. Déhé and Y. Kavalova, “Parentheticals: An Introduction,” in Parentheti- cals 1–22, at 1–2. 27 Brinton, The Comment Clause; J. O. Urmson, “Parenthetical Verbs,” Mind 61 (1952) 480–496. 28 Urmson, Mind 61 (1952) 481. 29 N. Déhé and A. Wichmann, “The Multifunctionality of Epistemic 174 ΕΙΠΕ ΜΟΙ AS A PARENTHETICAL ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 nounce Kenya funny”;30 and if you will, as in “But it was kind of an issue that was brushed under the carpet if you will.”31 I now turn to one of the three defining aspects of paren- theticals which still requires some explanation: their lack of referential meaning. Consider the following example (Pl. Spuria 395E7): (20) σὺ γὰρ εἰπέ µοι, ὦ Ἐρυξία, ἀγαθὸν ἡγῇ τὸ πλουτεῖν; You tell me, Eruxia, do you think it’s good being rich? This is a straightforward question, one which can be answered with a simple yes or no. Let us say that Eruxia answers no. This implies that she will answer yes to (21), unless she refrains from making a moral judgment on the matter: (21) σὺ γὰρ εἰπέ µοι, ὦ Ἐρυξία, κακὸν ἡγῇ τὸ πλουτεῖν; You tell me, Eruxia, do you think it’s bad being rich? This means that ἀγαθόν is truth-conditional in (20): it bears on the answer of the hearer. If ἀγαθόν is replaced by a non- synonymous word, the hearer will adjust his or her response accordingly. More importantly, if ἀγαθόν is left out, we simply do not have a complete question—we do not have a complete proposition to which truth conditions can apply. If εἰπέ µοι is deleted, however, the truth conditions of the question are not altered: (22) σὺ γὰρ, ὦ Ἐρυξία, ἀγαθὸν ἡγῇ τὸ πλουτεῖν; You, Eruxia, do you think it’s good being rich? We are left with a complete proposition to which Eruxia will reply with the same answer she provided for (20). As such, εἰπέ µοι is non-truth-conditional. The question of truth-conditionality is often bound up with that of procedurality. Procedural items are opposed to con- ceptual items: while the latter map onto concepts which form the backbone of the content of the utterance, the former can be ___ Parentheticals in Discourse: Prosodic Cues to the Semantic-pragmatic Boundary,” Functions of Language 17 (2010) 1–28, at 2. 30 Brinton, in Opening Windows 283. 31 Brinton, The Comment Clause 165. SAMUEL ZAKOWSKI 175 ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 regarded as ‘signal posts’ which “guide the process of re- covering various aspects” of the utterance.32 Words such as cat, think, and maliciously are conceptual; words such as therefore, wow, and but are procedural. Procedural items, then, help the hearer in decoding the message while not being part of the message itself; in Carston’s words, they are “effort-saving devices” which are geared towards facilitating hearer comprehension of the utterance.33 But, for instance, can encode a contrast between two different clauses: it is not part of the message, but indicates how the two clauses should be interpreted vis-à-vis each other, and, hence, how the message should be construed as a whole.34 Although procedural items are often non-truth-conditional (wow and but do not influence whether a proposition is parsed as true or false), and conceptual items are often truth-con- ditional, there are some notable exceptions: there is no a priori overlap, and both categories (truth-conditionality and nature of semantic contribution) cross-cut each other.35 But since most non-truth-conditional linguistic items are procedural as well, the case for a conceptual analysis must be quite strong: if it cannot be decisively proved that εἰπέ µοι is conceptual, it should be regarded as procedural.36 It is unclear what εἰπέ µοι would contribute to the propo- sition conceptually. Obviously, it is semantically superfluous: asking a question entails an expectation that the hearer will provide some answer to that question. A deletion test brings 32 A. Hall, “Do Discourse Connectives Encode Concepts or Pro- cedures?” Lingua 117 (2007) 149–174, at 158. 33 R. Carston, Thoughts and Utterances: The Pragmatics of Explicit Communica- tion (Oxford 2002) 162. 34 Hall, Lingua 117 (2007) 154. 35 Cf. B. Clark, Relevance Theory (Cambridge 2013) 318–322, giving examples. 36 Cf. e.g. Clark, Relevance Theory 326; D. Wilson and Dan Sperber, “Linguistic Form and Relevance,” Lingua 90 (1993) 1–25, at 11; D. Blake- more, Relevance and Linguistic Meaning: The Semantics and Pragmatics of Discourse Markers (Cambridge 2002). 176 ΕΙΠΕ ΜΟΙ AS A PARENTHETICAL ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 this out most plainly:37 altering the clause in (20) by leaving out εἰπέ µοι in (22) had no discernible influence on Eruxia’s answer. Thus εἰπέ µοι cannot be said to contribute to the semantic content of the utterance. The same is true of (23): (23) The cook Sicon has asked Cnemon for a pot, to which Cne- mon replied that he does not have one (Men. Dysc. 510–511): οὐδ’ ὁπόθεν ἄν τις, εἰπέ µοι, ἐλθὼν λάβοι φράσαις ἄν; Could you not even say, tell me, where someone could go find [a pot]? In this example, εἰπέ µοι is clearly unnecessary from a semantic point of view—not only is Sicon asking a question, but he has also added φράσαις ἄν, which already encodes the idea of ‘saying’ or ‘telling’. How, then, can the inclusion of εἰπέ µοι be explained? As Brinton has pointed out for say (very similar, of course, to εἰπέ µοι), seemingly superfluous additions to the utterance can encode meanings which are based on, but go beyond, their original, conceptual sense. Put differently, if a speaker adds εἰπέ µοι to a question, which already entails that the speaker expects an answer from the hearer, the speaker invites the inference that “something more is intended” than the basic, conceptual meaning of εἰπέ µοι.38 Cognitive resources are val- uable, and tying them up with unnecessary prolixity amounts to poor judgment on the part of the speaker.39 The question now, of course, becomes what exactly that something more would be. Traugott and Dasher argue that semantic change can be cross-linguistically correlated with (a) a change from semantic to procedural meaning, and (b) non-subjectivity to 37 B. Frank-Job, “A Dynamic-interactional Approach to Discourse Markers,” in K. Fischer (ed.), Approaches to Discourse Particles (Amsterdam 2006) 359–374, at 366. 38 Brinton, in Opening Windows 293. 39 Cf. E. Ifantidou, “Parentheticals and Relevance,” UCL Working Papers in Linguistics 5 (1993) 193–210, at 207; Clark, Relevance Theory 108. SAMUEL ZAKOWSKI 177 ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 (inter)subjectivity.40 These two facets will form a stepping stone for the subsequent discussion of the function of εἰπέ µοι. 3.1 Semantic > procedural Many studies of parentheticals have demonstrated that they serve procedural functions. Urmson refers to a pre-theoretical notion of procedurality by stating that they “function as signals guiding the hearer to a proper appreciation of the statement in its context” and “help the understanding and assessment of what is said rather than being a part of what is said.”41 Brinton, too, argues that parentheticals serve “procedural functions,” as does Rouchota;42 Fedriani et al. conclude that parenthetical ἄγε functions procedurally as well.43 Some examples will bring out how procedurality is under- stood. Brinton’s analysis of say emphasizes the desemanti- cization of the original meaning in favour of more abstract, procedural meanings: it gradually comes to express an “emotional response” on the part of the speaker, or is used to “evoke the hearer’s attention.”44 Frank-Job argues that Italian dimmi (which is obviously almost identical to εἰπέ µοι) functions as a turn-yielding device for the speaker.45 Fedriani et al. make the same case for some instances of ἄγε; other occurrences can be regarded as having transitioned from an exhortation to pro- ceed “across space” to an exhortation to proceed “in doing something,” the latter being more abstract and procedural.46 40 E. C. Traugott and R. B. Dasher, Regularity in Semantic Change (Oxford 2005) 40. 41 Urmson, Mind 61 (1952) 495 and 496. 42 Brinton, The Comment Clause 241; V. Rouchota, “Procedural Meaning and Parenthetical Discourse Markers,” in A. H. Jucker and Y. Ziv (eds.), Discourse Markers: Descriptions and Theory (Amsterdam/Philadelphia 1998) 97– 126. 43 Fedriani et al., “Constraining Pragmaticalization.” 44 Brinton, in Opening Windows 284, also The Comment Clause 73–110. 45 Frank-Job, in Discourse Particles 369. 46 Fedriani et al., “Constraining Pragmaticalization.” 178 ΕΙΠΕ ΜΟΙ AS A PARENTHETICAL ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 Italian guarda, finally, transitions from conceptual (“look!”) to procedural meaning (the speaker has something important to say).47 3.2 Non-subjective > (inter)subjective Subjective expressions encode the speaker’s perspective and point of view; intersubjective expressions encode the relation- ship between speaker and hearer.48 Among the functions noted for subjective parentheticals are the ability to express “re- sponses, reactions, attitudes, understanding, tentativeness, or continued attention”; intersubjective parentheticals can, for instance, express “intimacy, cooperation, shared knowledge, deference, or face-saving (politeness).”49 Examples of subjective parentheticals are I mean, which “is used when the speaker focuses attention on him/herself”;50 and say, as in “ ‘Say, that’s our City’, bubbles Dolores,” which can “express a (rather weak) emotional response, such as surprise, regret, anger, dis- belief, delight, etc., to what is (about to be) said.”51 Examples of intersubjective parentheticals are you know, which can mark information “shared by the speaker and his addressee”;52 and dimmi, which facilitates the transition from speaker to hearer and vice versa.53 Parenthetical ἄγε is intersubjective as well. In sum, then, there is cross-linguistic evidence that paren- 47 R. Waltereit, “Imperatives, Interruption in Discourse, and the Rise of Discourse Markers: A Study of Italian guarda,” Linguistics 40 (2002) 987– 1010. 48 Brinton, The Comment Clause 70; Traugott and Dasher, Regularity 6; Company Company, Belgian Journal of Linguistics 20 (2006) 98. 49 Brinton, The Comment Clause 18. 50 D. Schiffrin, Discourse Markers (Cambridge 1987) 299. 51 Brinton, in Opening Windows 283–284. 52 S. Vincent, S. Darbaky, and A. Mattouchi, “The Grammaticalization of you know: From Shared Knowledge to Control over the Co-speaker,” English Text Construction 2 (2009) 209–227, at 214. 53 Frank-Job, in Discourse Particles 359–374; see also C. Bazzanella, “Discourse Markers in Italian: Towards a ‘Compositional’ Meaning,” in Discourse Particles 449–464, at 457. SAMUEL ZAKOWSKI 179 ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 theticals are (a) desemanticized, procedural elements, and (b) subjective or intersubjective elements of the discourse. Not coincidentally, desemanticization and (inter)subjectification are also two of the semantic features of grammaticalized expres- sions.54 In what follows I turn to εἰπέ µοι itself, to analyze whether it fits this pattern and what it contributes to the dis- course. 3.3. εἰπέ µοι’s semantics εἰπέ µοι’s exact function will be discussed by first analyzing its core meaning and then the contexts in which it usually occurs. 3.3.1. Directness Consider the following example: (24) Socrates has just denounced sophists as being immoral. Cal- licles reacts to Chaerephon (Pl. Grg. 481B6–7): εἰπέ µοι, ὦ Χαιρεφῶν, σπουδάζει ταῦτα Σωκράτης ἢ παίζει; Tell me, Chaerephon, is Socrates being serious about these things or joking? This is a standard disjunctive question. The text proceeds with Chaerephon admitting that he is not sure about the correct 54 Cf. S. Nicolle, “Conceptual and Procedural Encoding: Criteria for the Identification of Linguistically Encoded Procedural Information,” in M. Groefsema (ed.), Proceedings of the University of Hertfordshire Relevance Theory Workshop (Chelmsford 1997) 47–56, at 54–55, who postulates a link between procedurality and grammaticalization. For desemanticization see Brinton, in Opening Windows 292, and The Comment Clause 91. For intersubjectification see Traugott and Dasher, Regularity 6; Company Company, Belgian Journal of Linguistics 20 (2006) 98; Brinton, The Comment Clause 70. It should be noted that ‘desemanticization’ does not amount to ‘deterioration’. As Sweetser notes, the latter is burdened with a “rather pejorative connotation”: E. Sweetser, “Grammaticalization and Semantic Bleaching,” Proceedings of the Fourteenth Annual Meeting of the Berkeley Linguistics Society (1988) 389–405, at 401. What actually happens is that the original conceptual source of an expression provides the basis for a related yet distinct meaning. Cf. Biraud, Les interjections 247, and Fedriani et al., “Constraining Pragmaticalization,” on desemanticization with other parentheticals in Ancient Greek. 180 ΕΙΠΕ ΜΟΙ AS A PARENTHETICAL ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 answer and that they had better ask Socrates himself. At first blush, nothing too strange seems to be going on here. Callicles is asking a clear question prefaced by εἰπέ µοι, which is clearly parenthetical (syntactically detached yet asyndetically con- nected to the main clause; non-truth-conditional). As argued above, εἰπέ µοι does not add to the semantic content of the utterance. However, its inclusion does affect how the hearer interprets the utterance. Let us say, for argument’s sake, that there was no εἰπέ µοι in (24). In that case, we could paraphrase Callicles’ words: (25) Callicles is asking Chaerephon if Socrates was being serious or joking. Yet now we have left out a part of Callicles’ message. With the addition of εἰπέ µοι, his words could be paraphrased: (26) Callicles is ordering Chaerephon to tell him if Socrates was being serious or joking. The paraphrase in (26) is still only an approximation of Cal- licles’ message. It is more accurate to state that he utters a proposition (ὦ Χαιρεφῶν, σπουδάζει ταῦτα Σωκράτης ἢ παίζει) in the form of a question; by including εἰπέ µοι, the strength of the proposition is enhanced, without that propo- sition being altered as such: Chaerephon’s answer (‘he is being serious’ or ‘he is joking’) is not influenced by the addition of εἰπέ µοι. The point is that εἰπέ µοι “fine tunes the inter- pretation” of the proposition without adding to it.55 A more comprehensive paraphrase of Callicles’ utterance in (24), then, would combine both (25) and (26). By this I mean that Callicles has uttered two separate speech acts with the same illocution: he has uttered a question (ὦ Χαιρεφῶν, σπουδάζει ταῦτα Σω- κράτης ἢ παίζει;), which is a directive speech act;56 he has also 55 D. Blakemore, “Divisions of Labour: The Analysis of Parentheticals,” Lingua 116 (2006) 1670–1687, at 1683; cf. Ifantidou, UCL Working Papers in Linguistics 5 (1993) 206; Urmson, Mind 61 (1952) 495–496. 56 J. R. Searle, “A Classification of Illocutionary Acts,” Language in Society 5 (1976) 1–23, at 11. SAMUEL ZAKOWSKI 181 ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 uttered a speech act with imperative εἰπέ µοι, which attempts to “get the hearer to do something” and hence is directive as well.57 The two can be distinguished in that the question constitutes a perfectly well-formed message on its own, but the imperative does not. In other words, the speech act εἰπέ µοι needs another speech act to which it can apply. Moreover, ὦ Χαιρεφῶν, σπουδάζει ταῦτα Σωκράτης ἢ παίζει is truth- conditional, while εἰπέ µοι is not. εἰπέ µοι, then, only operates (non-truth-conditionally) in function of another, semantically ‘main’ speech act. This leads to the question what exactly εἰπέ µοι contributes to the speaker’s message. It is somewhat peculiar for speakers to employ imperatives in regular “face-to-face interaction”: under normal circumstances, interlocutors “will take into account each other’s desire for freedom of action” and avoid ordering each other around, and while interlocutors will usually hide their intention to get the other to perform some task, imperatives entail the exact oppo- site of that goal—they involve the speaker plainly commanding the hearer to do something, and hence they fall “on the lowest end of the politeness scale.”58 Put differently, imperatives in- crease the ‘directness’ of the interaction between speaker and hearer:59 they usually result in “face-threatening acts,” where the speaker attempts to impose his will on his interlocutor, and, more importantly, does not go to any trouble to disguise this fact.60 In this sense, they can be considered the opposite of 57 J. R. Searle, “A Taxonomy of Illocutionary Acts,” in J. R. Searle, Expression and Meaning: Studies in the Theory of Speech Acts (Oxford 1979) 1–29, at 13. See also Biraud, Les interjections 220, who argues that φέρε, ἄγε, and ἴθι “conservent une valeur illocutoire directive” (cf. Fedriani et al., “Constrain- ing Pragmaticalization,” on ἄγε). 58 K. Pelsmaekers, “Directness and (im)politeness: The Use of Impera- tives in Business Letters,” in G. Tops et al. (eds.), Thinking English Grammar (Louvain 1999) 263–279, at 263. 59 Cf. A. Mehrabian, “Immediacy: An Indicator of Attitudes in Linguistic Communication,” Journal of Personality 34 (1966) 26–34, at 28. 60 H. Haverkate, “Politeness Strategies in Verbal Interaction: An Analy- 182 ΕΙΠΕ ΜΟΙ AS A PARENTHETICAL ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 conversational mitigators such as please, will you?, if you wouldn’t mind, and so on:61 broadly, a speaker who adjoins imperatives to his (direct or indirect) requests is less concerned about saving face and being polite. In (24), this seems to be a plausible in- terpretation of εἰπέ µοι’s contribution: the question gains in directness by the addition of εἰπέ µοι, but nothing else besides seems particularly distinctive about the utterance. Put differ- ently, we are dealing with an entirely ‘neutral’ context here: Callicles is asking a sincere question to which he desires an in- formative answer, without any emotional overtones. 3.3.2. Affectivity It seems intuitively appealing that εἰπέ µοι increases the directness of the main speech act (the question) and functions as an urgent imploration that the hearer provide an answer. However, on delving deeper into the instances where it occurs, it seems that it operates in certain specific contexts. Directness is an attractive starting point from which we can proceed to a more nuanced view for these examples. Take the following in- stances: (27) Trygaeus is trying to stop wedding guests from singing songs of war (Ar. Pax 1300): εἰπέ µοι, ὦ πόσθων, εἰς τὸν σαυτοῦ πατέρ’ ᾄδεις; Tell me, little boy, are you singing this for your own father? (28) Geton has been banging on Cnemon’s door; Cnemon’s re- action (Men. Dysc. 466–467): τί τῆς θύρας ἅπτει, τρισάθλι’, εἰπέ µοι, ἄνθρωπε; Why are you hanging on to the door, tell me, you triply wretched man? In (27), Trygaeus has been infuriated by (sons of ) wedding guests who have been singing war songs for a time; at 1271– ___ sis of Directness and Indirectness in Speech Acts,” Semiotica 71 (1988) 59–71, at 65. 61 Pelsmaekers, in Thinking English Grammar 264; B. Fraser, “Conversa- tional Mitigation,” Journal of Pragmatics 4 (1980) 341–350. SAMUEL ZAKOWSKI 183 ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 1272, he had insulted one of them by calling him τρισκακόδαι- µον (“thrice-possessed”), ἀµαθές (“foolish”), and κατάρατον (“accursed”). Accordingly, the use of εἰπέ µοι fits into a pattern of extremely ‘direct’ messages in which the speaker does not take his own ‘face’ or conversational politeness into account. (28) is very similar: Cnemon has just been forced to open his door because there was so much noise, which means that he is not in the best of moods. He does not take his hearer’s feelings or his own ‘face’ into account, but immediately insults Geton. In both cases, the presence of the insult (πόσθων and τρισ- άθλιε) as well indicates that these two are very direct, face- threatening utterances.62 The use of εἰπέ µοι here is certainly warranted from the ‘directness’ point of view. εἰπέ µοι has a penchant for appearing in these types of affec- tive contexts. In the next example, the speaker is expressing surprise: (29) Chaireas is informing Daos that Smicrines is planning to marry the girl he himself loves (Men. Asp. 310–311): (Χα) µέλλει γαµεῖν γὰρ αὐτός. (Δ∆α) εἰπέ µοι, γαµεῖν; δυνήσεται δέ; (Cha.) because he himself is about to marry. (Da.) Tell me, to marry? And will he be able to do that? Daos is taken aback by this new information Chaireas has provided—so much so, in fact, that he echoes Chaireas (γαµεῖν). He then asks if Smicrines’ plans are even possible, which illustrates that this turn of events is completely un- expected to him.63 62 Cf. Ar. Nub. 847, where Strepsiades has just called Phidippides ἀµαθής and παχύς (“thick”); Men. Sam. 677, where Moschion is rebuking his servant Parmenon for a perceived slight, using the insult ἱερόσυλε (“temple rob- ber”). 63 Cf. Ar. Nub. 200 (πρὸς τῶν θεῶν, τί γὰρ τάδ’ ἐστίν; εἰπέ µοι, “By the gods, what is this? Tell me,” which is treated as (15) above), where Strep- siades is utterly confounded by the esoteric mathematical instruments he has discovered. 184 ΕΙΠΕ ΜΟΙ AS A PARENTHETICAL ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 In (30), which expands on (14) and (18), the speaker is caught by a sudden panic: (30) Simice has dropped her master Cnemon’s bucket into a well; he has just barged in, looking for what he assumes is the thief who stole this bucket (Men. Dysc. 589–590): (Σιµ) ἄκουσα, δέσποτ’, ἐνέβαλον. (Κν) βάδιζε δὴ εἴσω. (Σιµ) τί ποιεῖν δ’, εἰπέ µοι, µέλλεις; (Sim.) Listen, master, I dropped it in [a well]. (Cn.) Come, step inside. (Sim.) What, tell me, are you going to do? Simice is fearful of what she presumes is going to be a severe punishment; as a consequence, she forgets all sense of hierar- chy and decorum, and demands to know from her master what is going to happen. This is quite unusual—a slave does not or- der his or her master around, telling him what he must or must not say or do. In this sense, (30) is affective as well: Simice’s fear trumps any considerations of deference and propriety, which results in a very direct utterance. Affectivity forms an important part of the answer to what εἰπέ µοι’s function is. As a very direct expression, it is liable to occur in contexts where the speaker is less concerned with politeness and propriety. As such, εἰπέ µοι seems to have a subjective function: it subtly focuses attention on the speaker by combining with affective, direct utterances, where the speaker expresses irritation, anger, surprise, and fear.64 Of course, this subjective dimension does not preclude an intersubjective component from being present as well. In the previous examples (25–30), the speaker was asking a question of the hearer to which he expected an informative answer; an exception is (29), where the speaker’s echoing of his inter- locutor’s words indicates his surprise.65 The speaker, in the 64 Cf. also Brinton, in Opening Windows 284, on say. 65 Cf. D. Wilson and D. Sperber, “Mood and the Analysis of Non- declarative Sentences,” in J. Dancy et al. (eds.), Human Agency: Language, Duty SAMUEL ZAKOWSKI 185 ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 majority of these instances, is soliciting a response from the hearer—in other words, trying to catch the hearer’s attention and get him to invest in the conversation.66 3.3.2 Rhetorical questions εἰπέ µοι also occurs in non-affective contexts. As a marker of affective utterances, it is more frequent in comedy than it is in the prose authors under consideration (i.e. Plato and Demos- thenes). Take the following instance: (31) By Socrates’ request, Chaerephon is asking questions of Gor- gias (Pl. Grg. 447D6–8): εἰπέ µοι, ὦ Γοργία, ἀληθῆ λέγει Καλλικλῆς ὅδε ὅτι ἐπαγγέλλῃ ἀποκρίνεσθαι ὅτι ἄν τίς σε ἐρωτᾷ; Tell me, Gorgias, is it true what Callicles here says, that you pro- claim to answer whatever someone asks of you? There is no affective component to this utterance—it appears to be a straight-up question. The answer, however, is entirely straightforward: yes, Gorgias will answer any and all questions. This is information which was already provided to Chaerephon by Callicles: why would Callicles lie about it? In a sense, then, the answer to the speaker’s question is embedded into the ques- tion itself. Put differently, the way in which the question is posed implies that the speaker already knows what the hearer will reply.67 Example (31) is contiguous to rhetorical questions. Yet it can- not quite be considered such. Even though the speaker pro- vides the bulk of the answer to the hearer, he is still asking for ___ and Value (Stanford 1988) 77–101. 66 Cf. Brinton, in Opening Windows 284 and The Comment Clause 89, both on say; also Bazzanella, in Discourse Particles 457. Interestingly, Maeschler notes that the Hebrew expression tagidi li, which, like εἰπέ µοι, translates as ‘tell me’, functions intersubjectively as well: Y. Maeschler, “Metalanguaging and Discourse Markers in Bilingual Conversation,” Language in Society 23 (1994) 325–366, at 350. 67 See e.g. Euthyd. 293E5–6, Phdr. 229B4, and Hp. Mai. 292C5 for other examples of these types of questions with εἰπέ µοι in Plato. 186 ΕΙΠΕ ΜΟΙ AS A PARENTHETICAL ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 confirmation or denial. However, (31) is entirely self-evident, i.e. it includes a “particular answer” which the speaker imposes on the hearer—this is one of the main characteristics of a rhetorical question.68 Accordingly, I will label instances like these ‘pseudo-rhetorical questions’. In the corpus under con- sideration, εἰπέ µοι does in fact occur with genuine rhetorical questions. Compare (31) to the following, by now familiar, example: (32) Demosthenes is presenting his fellow Athenians with a stark choice between taking up arms against Philip and the following (4.10): ἢ βούλεσθ’, εἰπέ µοι, περιιόντες αὑτῶν πυνθάνεσθαι, “λέγεταί τι καινόν;” Or are you happy, tell me, while you are running around, to in- quire from each other, “What news is there?” Of course, the implied answer to Demosthenes’ question is a resounding no. More importantly, while (31) elicited a con- firmation or denial from the hearer(s), and hence was not completely superfluous, (32) is not actually a request for in- formation. Instead, it amounts to a rebuke on Demosthenes’ part, an admonition for the Athenians to smell the proverbial coffee and take the threat posed by Philip seriously. Demos- thenes’ implicit message can be paraphrased as follows: (33) You should not be happy to run around and inquire from each other, “What news is there?” The speaker, then, is not asking for confirmation or denial, i.e. is not asking for information, but simply performing an indirect directive speech act, couched in a question. I give one other example of a rhetorical question from De- mosthenes: 68 C. Ilie, What else can I tell you? A Pragmatic Study of English Rhetorical Ques- tions as Discursive and Argumentative Acts (Stockholm 1994) 3. As pointed out to me by a reviewer for GRBS, these ‘pseudo-rhetorical’ instances of εἰπέ µοι “occur always at the beginning of an elenchus or conversation, or introduce a new stage or topic.” SAMUEL ZAKOWSKI 187 ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 (34) Demosthenes is making his case against Philip (19.312): εἰπέ µοι, τῆς νῦν οὔσης Ἑλλάδος ταυτησὶ καὶ οἰκουµένης ἔσθ᾽ ὅ τι ταύτην ἂν τὴν προσηγορίαν εἶχεν ἢ ᾠκεῖθ᾽ ὑπὸ τῶν νῦν ἐχόντων Ἑλλήνων, εἰ µὴ τὰς ἀρετὰς ὑπὲρ αὐτῶν ἐκείνας οἱ Μαραθῶνι κἀν Σαλαµῖνι παρέσχοντο, οἱ ἡµέτεροι πρόγονοι; Tell me, of this Greece which now exists and is inhabited, is there a part which would have this name or would be inhabited by the Greeks who do so now, if they, in their defence, had not applied this courage to Marathon and Salamis—I mean our forefathers. Again, Demosthenes’ (paraphrased) message departs signifi- cantly from what he is saying: (35) There is no part of contemporary Greece which would bear that name or would be inhabited by Greeks if our forefathers had not shown their courage at Marathon and Salamis. In this case, Demosthenes is stating an (indirect) assertion in- stead of a true question. As he himself points out after posing his question in (34), no other answer except no is plausible (19.312): (36) οὐδ᾽ ἂν εἷς εὖ οἶδ᾽ ὅτι φήσειεν, ἀλλὰ πάντα ταῦθ᾽ ὑπὸ τῶν βαρβάρων ἂν ἑαλωκέναι. I know for certain that not even one person would say ‘Yes’, but [he would say] that all these areas would have been taken by the barbarians. Demosthenes, in (36), makes explicit what most rhetorical questions do implicitly: they “emphasize one particular answer by excluding all the other alternative answers.”69 He is not asking for information from his audience in (34), but, instead, makes an assertion which he assumes his audience will share. At first blush, it might seem slightly bizarre that a rhetorical question (to which the speaker does not expect an actual answer) is punctuated by εἰπέ µοι, which, by its imperative nature, presses the interlocutor for an answer. However, the idea that the hearer does not answer because the speaker did not ask an actual question is not entirely right. Not all answers 69 Ilie, What else can I tell you? 3. 188 ΕΙΠΕ ΜΟΙ AS A PARENTHETICAL ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 manifest themselves in actuality, i.e. as a spoken or written string of words. In the case of rhetorical questions, the audience answers the question mentally. More to the point, the speaker’s goal is “to elicit the addressee’s mental agreement with the implication of the rhetorical question.”70 In this sense, then, rhetorical questions are intersubjective. Adding εἰπέ µοι to a rhetorical question increases the ques- tion’s directness and presses the audience either to find fault with the speaker’s argument or to agree with it. The rhetorical questions are constructed so that accepting Demosthenes’ point of view is the only reasonable conclusion: if the hearer is happy to run around in (32), he leaves himself open to the charge of cowardice; if the hearer does not acknowledge the important role of his forebears in protecting Greece in (34), he could be considered unpatriotic and even treasonous. Thus, the rhe- torical question, made more direct by εἰπέ µοι, involves the audience in Demosthenes’ oration by inviting agreement with his arguments. εἰπέ µοι’s basic contribution to the utterance, then, seems to concern conversational directness. It amounts to a secondary speech act, a directive ‘utterance modifier’71 which combines with a primary speech act (the rest of the sentence) that is directive as well. With εἰπέ µοι, the speaker manifestly imposes his/her will on the hearer to tell him/her something. In itself, εἰπέ µοι is not part of the proposition: it is more appropriate to say that εἰπέ µοι encodes an instruction for the hearer on how to interpret the proposition (viz., as a more direct, explicit di- rective), without contributing to the truth-conditional content of the utterance. As such, εἰπέ µοι should be considered pro- cedural in nature: procedural expressions provide instructions for computations about propositions.72 In this sense, εἰπέ µοι 70 Ilie, What else can I tell you? 38. 71 Cf. K. Bach, “The Myth of Conventional Implicature,” Linguistics and Philosophy 22 (1999) 327–366. 72 D. Blakemore, “Performatives and Parentheticals,” PAS 91 (1990) 197–213, at 211–212; K. Aijmer, English Discourse Particles: Evidence from a SAMUEL ZAKOWSKI 189 ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 does “not contribute to the expression of a thought [i.e., to the proposition] but characterize[s] the expression of it” by in- creasing its directness.73 This directness can manifest itself in different ways, as we have seen. At the most basic level, εἰπέ µοι operates (inter- subjectively) as an attention-getting, response-eliciting device.74 This is, of course, also a function of its non-parenthetical coun- terpart. Accordingly, the semantic grammaticalization criterion of ‘persistence’ pertains to εἰπέ µοι: traces of the meaning of the original lexeme are retained in the grammaticalized form.75 With rhetorical questions, εἰπέ µοι is used to press the audience for mental agreement with the point the speaker is making. Subjectively, εἰπέ µοι often combines with affective utterances, where conversational propriety and politeness are less of an issue—accordingly, they harmonize perfectly with εἰπέ µοι. One instance can function both subjectively and intersub- jectively: εἰπέ µοι can point to the speaker’s anger and also elicit a reponse from the hearer, for example. A word of caution to end this section. There are some instances (especially in Menander) which seem to be paren- thetical from a structural-syntactic point of view (cf. §2 above) and have been counted as such in the previous section. Yet these utterances, or the surrounding context, are incomplete, which means that there is very little to be said about their function (e.g. Sam. 170, Dysc. 691). These examples were not included in this section on εἰπέ µοι’s function, since any inter- pretation would be founded on mere speculation. Thus, this account has been predominantly qualitatively-based. ___ Corpus (Philadelphia 2002) 11; Carston, Thoughts and Utterances 162. 73 Bach, Linguistics and Philosophy 22 (1999) 341, cf. 328. 74 Cf. Fedriani et al., “Constraining Pragmaticalization,” for a similar view on ἄγε. 75 Cf. Hopper, in Approaches to Grammaticalization I 22; Brinton, The Com- ment Clause 51; Traugott and Dasher, Regularity 11–12; cf. also Biraud, Les interjections 220, and Fedriani et al., Constraining Pragmaticalization, on per- sistence with other parentheticals in Ancient Greek. 190 ΕΙΠΕ ΜΟΙ AS A PARENTHETICAL ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 4. Conclusions Research into Ancient Greek parentheticals has come a long way since Schwyzer’s (very valuable) pre-theoretical remarks.76 Modern linguistic insights and frameworks have been success- fully applied to these expressions; this paper should be regarded as a contribution to this growing body of work. Although much about parenthetical εἰπέ µοι remains unclear and further (dia- chronic and synchronic) research is certainly required, some tendencies can be pointed out. Semantically, εἰπέ µοι can be regarded as a parenthetical, a non-truth-conditional item which makes a procedural con- tribution to the utterance. This contribution is connected to illocution. εἰπέ µοι, which is itself directive, always occurs with other directive speech acts (i.e., questions): it increases the di- rectness of these speech acts, i.e., makes the speaker’s desire to impose his/her will on the interlocutor more explicit. Thus it can be regarded as an ‘utterance modifier’ in Bach’s sense in that it does not contribute to the proposition, but instead mod- ifies how the proposition’s expression should be interpreted.77 From a diachronic syntactic perspective, it seems that the most plausible evolution from matrix clause to parenthetical is based on the ambiguity when εἰπέ µοι is followed by an inter- rogative. As interrogatives can indicate both direct and indirect questions, two readings are possible in these cases. On the direct question reading, εἰπέ µοι becomes a syntactically de- tached part of the sentence, with the question becoming the main clause; on the indirect question reading, εἰπέ µοι is a matrix clause of which the question forms a complement. This ambiguity could then have spread to instances where the ques- tion was not introduced by an interrogative. εἰπέ µοι would then be syntactically detached from the question, which was now unambiguously direct and the main clause of the sentence. 76 E. Schwyzer, Griechische Grammat II Syntax und syntaktische Stilistik (Munich 1950) 583. 77 Bach, Linguistics and Philosophy 22 (1999) 327–366. SAMUEL ZAKOWSKI 191 ————— Greek, Roman, and Byzantine Studies 54 (2014) 157–191 Synchronically, εἰπέ µοι occurred clause-initially, clause- internally, and clause-finally. This positional mobility was correlated with its syntactic independence, which is, as pointed out above, ambiguous if εἰπέ µοι appears with interrogative- introduced questions. If εἰπέ µοι occurred clause-initially, it was regarded as a matrix clause; clause-internally, a paren- thetical; and clause-finally, still ambiguous but more likely to be parenthetical. In comparing Aristophanes, Menander, Plato, and Demos- thenes, it became obvious that Menander could be dis- tinguished from the other three. In his plays, εἰπέ µοι has a clear predilection for clause-internal and hence unambiguously parenthetical status. There are still gaps in this account. A quick search of John Chrysostom’s works, for instance, seems to indicate that εἰπέ µοι does not occur clause-initially in his sermons even once (cf. 172 above). Does this mean that it could only be used paren- thetically in the fourth century? What about its semantics? Can it still be considered an utterance modifier, or has its range branched out into other distinctive functions? These and other questions remain.78 January, 2014 Ghent University Blandijnberg 2 9000 Ghent, Belgium Samuel.Zakowski@ugent.be 78 This work was funded by the Fund for Scientific Research (FWO; grant number B/10040/02) and the Special Research Fund (BOF; grant number B/13790/01). I am very much indebted to Mark Janse, whose remarks improved an early draft of this paper tremendously. Many thanks to the anonymous GRBS referee for his/her valuable remarks.