The Attic particle μήν Intersubjectivity, contrast and polysemy

The paper examines the various usages of the Attic particle μήν and proposes a unified analysis of its main function. I argue that the prevalent analysis of Wakker (1997) needs some important reconsideration when instances of μήν in Platonic dialogue are concerned. First, the particle can target not only the propositional content of a discourse act, but also its illocution (felicity conditions). Second, I propose ‘countering expectations or assumptions of the addressee’ as the basic value of the particle. Functions in terms of commitment are better seen as secondary side effects. Third, I argue that differences in the origin of the countered assumptions or expectations are a natural basis for distinguishing between attitudinal μήν (extra-linguistic context and/or previous words of the addressee) and discourse connective μήν (previous words of the same speaker). It follows from my analysis that strict categorical boundaries between these usages are not to be expected.


Introduction
It is a well-known cross-linguistic fact that discourse particles typically occur in many different contexts and in various kinds of speech acts.As such, they characteristically display a broad array of different usages, which makes a unified descriptive account of their meaning or function into an interesting puzzle.This situation is no different for the Attic particle μήν, the topic of the present paper.To give a first impression of the diversity of the different contexts in which μήν occurs, consider the following typical examples and their English translations.All examples are taken from the dialogues of Plato, the main research corpus of the present paper.1 (1) [Then Agathon said, "It turns out, Socrates, I didn't know what I was talking about in that speech."]-Καὶ μὴν καλῶς γε εἶπες, φάναι, ὦ Ἀγάθων.(Pl.Smp.201b-c) "It was a beautiful speech, anyway (μήν), Agathon," said Socrates.
-Μὴ ὄντι μὴν (b) ἄγνοιαν ἐξ ἀνάγκης ἀπέδομεν, ὄντι δὲ γνῶσιν; -Ὀρθῶς, ἔφη.(Pl.r. 478c) "But surely (μήν) the most accurate word for that which is not isn't 'one thing' but 'nothing'?"-"Certainly."-"But(μήν) we had to set ignorance over what is not and knowledge over what is?"-"That's right." valuable comments on earlier drafts of the paper.The research for this paper is supported by the eu under fp7, erc Starting Grant 338421-perspective. 1 The Greek is that of the most recent oct-editions.The English translations are-if not noted otherwise-those of the volume edited by Cooper (1997).Within these translations I have added bracketed μήν behind the expression that seems to be the rendering of μήν.If no such expression seems to be present, bracketed μήν is placed at the beginning of the particular discourse unit.
(5) [ath.… After these remarks, our law (νόμος) on the subject should run like this: description of the law, about 16 oct-lines.]δεύτερος μὴν νόμος· … (Pl.Leg.919d-920a) Now (μήν) for a second law: … Two important observations can already be made on the distribution of μήν.First, we find the particle both in discourse acts that express an assertion, as in (1), (4a) and ( 5), and in discourse acts that have a non-assertive illocution, such as the specifying question in (2) and the yes/no-questions in (3).Second, μήν can occur both at the beginning of a new speech turn, the previous words being spoken by a different speaker (turn-initial use of μήν, as in examples ( 1)-( 4a)), as well as somewhere within a speech turn, the previous words being spoken by the same speaker (turn-internal use of μήν).In the latter case, μήν can occur either more globally at the beginning of a new discourse move, as in (5), or more locally within such a move, as in (4b).2These distributional facts, which seem to allow for distinguishing different usages of μήν, make a unified descriptive account of the particle a complex puzzle.Why do we find one and the same particle in so many different contexts 2 I will use the concepts of 'move' and 'discourse act' here as developed in the work of, for instance, Kroon (1995) and Hengeveld & Mackenzie (2008).A move is defined by the latter as "an autonomous contribution in ongoing interaction", its most important characteristic being that "it either is, or opens up the possibility of, a reaction" (2008:50).Most typically, a move coincides with a speech turn in conversation.In argumentative and narrative genres, however, a speech turn may also be more complex, consisting of multiple (linearly or hierarchically) ordered moves.Typically then, a move corresponds to a paragraph or an episode.A discourse act is defined by Kroon (1995:65) as "the smallest unit of communicative behaviour", which unlike a move does "not necessarily further the communication in terms of approaching a conversational goal."A move consists of at least one act, but typically consists of multiple (linearly or hierarchically) ordered acts.Crucially, it is the discourse act-which does not need to coincide with the traditional clause or sentence-that serves as the basic linguistic unit in the analysis of discourse particles.Cf.Mosegaard Hansen (1998;2006); Pons Bordería (2006) and Waltereit (2006)  (with quite different English translations)?First, we would like to see that, within a unified account, all of the facts mentioned are explained in a natural way.Second, we also want to explain how the various different usages of the particle are interrelated either on a synchronic or a diachronic level.
In my view, the currently prevailing accounts of μήν in Classical Greek, notably those of Denniston (1954) and Wakker (1997), are not fully able to deal with these two issues and account for all of the distributional observations made above.I discuss their analyses as well as their shortcomings in section 2. In sections 3 and 4, I propose a revisited unified account of the particle and examine in detail its most typical usages, focusing on instances in Platonic dialogues.
I believe that the Platonic corpus is indispensable when examining the function of μήν, because it is one of our main sources of Greek dialogue, the text type in which μήν is found most frequently.In addition, the Platonic dialogues-especially the later ones-also consist of considerable stretches of argumentative monologue, in which one speaker holds the floor and unfolds a particular argument, but an addressee is still present in the discourse situation.As has been recognised earlier, μήν is also found in this particular text type.3Furthermore, as far as I know, the function of μήν in Plato has not been the object of a detailed study since Denniston's (1954) important handbook.These two facts certainly make the present study a desideratum.4 3 This is based on the work of George (2009), who gives ample discussion of the degree of 'dialogicity' of particles in relation to text type (with pp.158-164 more specifically on μήν).He clearly demonstrates that μήν can also occur in monological texts with a high degree of 'diaphony' , i.e., passages where an addressee is implicitly 'present' in the text.George points out that it is telling that in a narrative work like that of Thucydides-where an addressee/narratee is at the background most of the time-μήν occurs only 19 times in total.Moreover, they only occur in stretches of (in)direct speech or in evaluative passages where Thucydides addresses "his more explicitly personal observations about the history he is writing" (2009:163), i.e., in passages with a high degree of diaphony.Similar conclusions are presented by Cuypers (2005) on μάν/μήν in Homeric narrative.It would be interesting to see whether the same conclusions can be reached with respect to the narrative works of Xenophon, where we also find an overt narrator that frequently intervenes in the narrative and μήν occurs more frequently than in Thucydides (26× in the Anabasis and even 70× in the Hellenica).I leave this to further research (see also fn.57 below).4 Sicking (1993:51-55), in a short appendix to his study on particle use in Lysias, does discuss a few examples of μήν in Plato, but his observations on the particle do not seem to be based on a fully-fledged and detailed study of the particle in this author.

2
Previous literature on μήν Denniston (1954), as the culmination of the preceding tradition of particle research, distinguishes three distinct uses of μήν: (i) emphatic, (ii) adversative connecting and (iii) progressive connecting.5 On emphatic μήν Denniston gives little more semantic information than that "it is difficult to grasp the exact difference in sense between μήν and the far commoner δή" (330).Adversative connecting μήν is explained by Denniston as 'balancing' in nature, i.e., connecting two coexisting, but opposed facts.English 'yet' and 'however' are proposed as apt translations of this use.Finally, progressive connecting μήν "either adds a fresh point ('again' , 'further'), or marks a fresh stage in the march of thought ('well' , 'now')" (336).When used solitarily (i.e.not in collocation with other particles), progressive μήν is almost completely confined to the Platonic corpus.
Although this categorisation gives an extremely useful overview of a large part of the data, Denniston is less clear and convincing when it comes to the exact semantico-pragmatic function(s) of μήν, as well as the interrelationship between these three widely divergent uses.In particular, his discussionor rather lack of discussion-of the semantics of emphatic μήν gives rise to problems in this respect.6 These issues are addressed in Wakker's (1996;1997) account of μάν/μήν, which is based upon the functionally oriented discourse-pragmatic framework that Kroon (1995) has developed for the description of Latin particles.As her research corpus she uses the dialogues of Theocritus (1996) and tragedy (1997).7I believe that her account is a huge step in the right direction, but it still needs some important reconsideration.This is particularly due to the fact that some usages of the particle do not occur-or only very infrequently-in her research corpora.My readjustments concern two main topics: (i) the basic semantics of the particle, and (ii) the connective usages of the particle.reflects the fact that an utterance is always part of a particular communicative situation with two or more discourse participants, i.e., speaker and hearer(s).The particles in this class specify the speaker's attitude with regard to (the truth value of) the expressed information or the supposed knowledge, expectations and presuppositions of his hearer(s).Wakker generally analyses μήν as a marker of speaker commitment, giving the following definition: In using μήν the speaker expresses his positive commitment to the truth of the proposition; he indicates that he as it were personally guarantees its truth: 'in truth' , 'really' .This insisting on the truth of the proposition is not a mere sign of 'emphasis' , rather the speaker in this way anticipates a possible reaction of disbelief on the part of the addressee.
wakker 1997:213 The latter part of this description involves the supposed beliefs and expectations of the addressee: the speaker assumes that the addressee might not expect his proposition to be true.This expectation, Wakker argues, can be based upon information in the previous conversation or upon the speech situation in general.9Example (6), in which Tiresias rejects (the implications of) the previous words of Creon, clearly illustrates the proposed analysis of μήν as marking personal insistence on the truth of the proposition and anticipation of possible disbelief or counter-expectation. to avoid confusion with particles like ἄν, which are also modal but in a totally different way, having a close relationship with the verbal category of mood.9 Wakker's analysis overlaps with Sicking's view on μήν.He argues that "[w]ith μήν … the speaker shows himself aware that his audience may not be inclined to accept the statement, and indicates that he will nevertheless uphold it.It thus implies the possibility of a distance between the two" (1993:52, italics original).Further on he adds that "[t]he particle μήν seems to be at home in expressing the contrary of what the person addressed might either (1) suppose or (2) wish" (54).Sicking's view is also followed by Cuypers (2005), who examines μάν/μήν in the epic genre (Homer and Apollonius).
On the basis of this semantico-pragmatic description Wakker is able to explain some striking distributional properties of μήν, viz.that μήν has a strong tendency to occur (i) in the dialogical text type10 and (ii) in declarative clauses, i.e., in assertions.Now, this second point constitutes my main objection.Wakker focuses on μήν in discourse acts that express assertions and correspondingly takes μήν as targeting the proposition.However, we have seen that in Plato μήν also occurs in acts with a non-assertive nature, e.g. in interrogatives (cf.( 2) and (3) above) or directives as in the following example (note the imperative σκόπει 'take care'):11 (7) Καὶ ὁ Διονυσόδωρος, Σκόπει μήν, ἔφη, ὦ Σώκρατες, ὅπως μὴ ἔξαρνος ἔσει ἃ νῦν λέγεις.(Pl.Euthd.283c) And Dionysodoros said: "(μήν) Take care, Socrates, that you don't find yourself denying these words." Since specifying questions and directives are typically understood as not expressing full propositions, Wakker's analysis runs into trouble here.Since there is no proposition present, the speaker cannot express his commitment to its truth value.12Thus, it seems that the particle in these instances does not pertain to the propositional content of the discourse act in question, but rather to (aspects of) its illocution.Furthermore, in yes/no-questions, which ask for confirmation, it seems to be at least out of place that a speaker strongly commits himself to the questioned proposition.These distributional facts entail that commitment to, or insistence upon, propositional truth cannot be the right concept for a unified description of μήν's basic semantics.Furthermore, in some instances a reading in terms of personal commitment to the truth of the proposition may be possible technically, but does not naturally harmonise with particular contextual clues.One such a clue is the cooccurrence of μήν with expressions that indicate the speaker's doubt or uncertainty, i.e., lack of full commitment to the truth of the proposition expressed.As noted by Denniston (1954:331-332), we already find μήν/μάν in wishes and directives in Homeric epic (e.g.Il. 1.302,5.765,7.459).For the few directive examples in tragedy, cf.Wakker (1997: fn. 13 Here, by using που ('I suppose'), Socrates indicates that he is not fully committed to the information he provides-or at least he presents it in this way, leaving room for possible disagreement on the part of Hippias.If we take μήν as a marker of speaker commitment to propositional truth, the collocation with such expressions of uncertainty seems to be out of place.Finally, a more theoretical argument may be adduced in addition.Even if one would argue that the notion of personal speaker commitment in itself could still be saved-viz.on the assumption that it essentially targets illocutionary acts rather than propositional content15-the general usefulness of this concept in ascribing a basic semantic value to μήν (or any other attitudinal particle) is rather questionable, at least if it is not defined more exactly in terms of e.g.epistemic certainty, evidentiality, speaker's belief or common ground.

13
See for instance example (10) below.14 See e.g.Sicking (1993:59) on που: "with που a speaker presents his statement as a surmise whose accuracy he does not vouch for so that disputing it need not impair the basis for an understanding between the two partners in the conversation."Sicking also shows that που can be used to mark potentially face-threatening acts, i.e., as a marker of politeness on the part of the speaker.He argues that this analysis fits the question-based character of Platonic dialogue, which, in turn, explains the high frequency of που in Plato (as opposed to e.g.tragic dialogue or oratory).I do not agree with Koier's (2013) recent analysis of που as primarily a marker of obvious or accessible information.In my view, this lacks the important conversational aspect of Sicking's analysis. 15 This was suggested by one of my anonymous reviewers.Cf. also Kroon's (1995) description of Latin vero.Note, by the way, that the collocation of μήν with που would still be problematic on this account.Theoretically, this does not seem to be very satisfactory.In case of assertions, for instance, the question arises what is it exactly that μήν conventionally 'adds' to the 'basic' commitment or speaker belief that is typically taken to be present in every assertion as a prerequisite for a cooperative way of communicationas reflected in, for instance, the sincerity condition for assertions in speech act theory: 's believes p' (Searle 1969), and the Gricean maxim of quality: 'Try to make your contribution one that is true' (Grice 1975).In other words, to what extent would an utterance containing μήν differ in speaker commitment from the same utterance without μήν?I believe that the real difference between such utterances lies not so much in (a higher degree of) speaker commitment, but should rather be sought within the domain of interpersonal relations between speaker and hearer.In other words, I think it is better to focus on the aspect of counter-expectation.
(ii) My second point of reconsideration concerns the discourse connective usages of μήν.Whereas Wakker convincingly shows that μήν develops a progressive discourse connective use,16 she strongly doubts the existence of an adversative connective use.She does concede that for many instances of μήν an adversative nuance is present, as in (6) above or in the following example: (  (1997:226).From here, Wakker argues that μήν acquires a progressive nuance, introducing a new point in the discourse-which the addressee probably did not expect-or a sudden turn in the course of events.So the attitudinal value is not wholly lost in this type of instance.However, elsewhere Wakker (1996:257-259) argues that in the 3rd-century Greek of Theocritus the development to a progressive connective particle has been fully completed, since here examples can be found where the attitudinal value of μήν (in Wakker's view, speaker commitment and anticipation of disbelief, see above) is completely absent.Thus, Wakker concludes, in tragedy the particle never loses its primary attitudinal value and must never be taken "as just an adversative connector" (226).

Leg. 663e)
Truth is a fine thing, Stranger, and it is sure to prevail, but (μήν) to persuade men of it seems no easy task.
Here too, to be sure, the adversative relation between the two discourse acts would already have been present without the particles, i.e., due to the context and the content of the two acts.As such, it is rather hard to definitely prove that μήν is an adversative connective here.18However, μήν occurs here in reaction to the 'preparatory' discourse connective particle μέν, i.e., in the same position in 17 Wakker (1996:255-257) also discusses the μέν-μήν-pattern in Theocritus.Here, she maintains that μήν never loses its main attitudinal value of commitment, but she admits that the particle might have developed an additional adversative connective function in this pattern.In my view, as argued in this section, this connective feature is already present in (late) classical Greek.18 I thank one of my anonymous reviewers for this important caveat.We are dealing here with the notorious difficulty of assessing whether a meaning aspect (in this case adversative connection) is part of the coded or conventional meaning of the particle itself or part of the contextual meaning (as a side-effect of the particle).However, it has also been shown cross-linguistically that on a diachronic level such contextual meanings or frequently occurring side-effects tend to become part of the conventional meaning of the particle itself (cf.Traugott & Dasher (2002:34f.)on the Invited Inferencing Theory of Semantic Change).This seems to be the case for μήν too (cf.also section 5 on diachronic change and Allan forthc.).which we typically find other discourse connective particles like δέ (a marker of thematic discontinuity) or ἀλλά 'but' .19This fact gives us a rather strong indication, in my view, that in this pattern μήν should be regarded not as strictly attitudinal in nature (expressing counter-expectation), but also as connective in a discourse-structural sense, just as the two other particles.Additional support for this view is the conspicuous absence-in the entire body of Classical Greek literature-of the collocation of μήν with δέ.20 If μήν were to be strictly attitudinal in nature, collocation with the attitudinally neutral connective δέ theoretically would not be impossible.However, the distributional evidence shows that μήν behaves as a competitor of δέ in instances like these.21 In view of these facts then, I am strongly inclined (following Denniston) to regard μήν as an adversative discourse connective particle in these instances: it indicates a contrastive coherence relation between two discourse units of the same speaker (denial-of-expectation contrast, see section 4).In sections 4 and 5 I will examine the differences between the attitudinal and discourse connective usages as well as issues of categorisation in more detail.
To sum up, the reasons for reconsideration are twofold.First, the notion of commitment to the truth of the proposition as part of μήν's basic semantics cannot be aligned with all of the available Platonic data, either because of its occurrence in non-assertive speech acts or because of other incompatible clues in the context.Second, in Plato there is strong evidence that μήν has an additional use as a discourse connective particle of an adversative nature.
In order to deal with these issues, I propose that (i) Wakker's secondary aspect of μήν's basic value-countering supposed expectations or assumptions of the addressee-constitutes the primary meaning aspect coded by μήν; (ii) commitment or affirmation is a secondary element and is at best seen as a pragmatic side-effect in particular contexts (i.e.assertions).In table 1, the   19 Cf.Bakker (1993) for an analysis of the particle δέ and the μέν-δέ-pattern in Ancient Greek.20 There is only one example where δέ and μήν take the same postpositive position: Pl.
Leg. 782c Τὸ δὲ μὴν θύειν … There are a few other instances, but these all have variae lectiones without δέ.For this reason, it is probably justified to doubt this instance as well, as Denniston (1954:341)

already proposes. 21
Another interesting distributional fact is that it is only within the μέν-μήν-pattern that μήν is found in complements that are embedded under verbs of speech or thought, as in Leg.860e: ἄκοντας μὲν ἀδίκους εἶναί φησιν, ἀδικεῖν μὴν ἑκόντας πολλούς 'he states (φησιν) that μέν there are those who are unjust against their will, but (μήν) that many men do commit unjust acts voluntarily.'Cf. also Leg. 723a, r. 529e and Sph.216b.This might be taken as further evidence that μήν in this pattern behaves differently from the strictly attitudinal usages and has an additional discourse connective feature.main differences between my proposal and those of Denniston and Wakker are shown.In the remainder of the paper, I discuss my own proposal in more detail.First, I discuss the notion of counter-expectation or contrast in general (section 3); second, I review and exemplify the various different usages of μήν, both the attitudinal and the discourse-connective ones (section 4).I end with some theoretical remarks on the categorization of μήν (section 5).

Intersubjective coordination and contrast
A useful way of looking at the concepts of counter-expectation and contrast in general is the cognitive-linguistic notion of intersubjective coordination as used by Verhagen (2005;2015).Verhagen assumes that every linguistic utterance involves not only an objective content level, but also a subjective level, which comprises (at least) two conceptualisers, i.e., the speaker and the addressee(s), as well as the mutually shared knowledge between them: their common ground.22The common ground comprises (i) knowledge of the present communicative event, both 22 In the functional discourse approach of Kroon (1995)-followed by Wakker (1997) physically and linguistically (the preceding discourse), (ii) the common personal history of the interlocutors and (iii) general culturally-oriented knowledge about (regularities in) the world, which includes shared cultural models of nationality, ethnicity, religion etc., particular cognitive schemata, frames and generalized topoi23 (of the form: 'normally if x, then y' or 'generally, x because of y').24 Crucially, the conceptualisers may well entertain different perspectives or points of view (i.e.knowledge states, assumptions, expectations etc.) with regard to a particular object of conceptualisation.Consequently, negotiation must take place as to whether a particular piece of information is agreed upon and accepted as being part of the common ground.Moreover, a speaker needs to manage the coordination between the perspectives sufficiently in order to produce utterances that the addressee can interpret felicitously at a given point within a conversation.The relationship between the conceptualisers' perspectives is called intersubjective coordination.Now, one of the main tenets of Verhagen's work is that the conventional meaning of many natural language expressions and grammatical constructions (e.g.sentence negation, complementation constructions or discourse connectives) should be placed not only on the objective level, where it would be described in terms of truth-conditional representation of the world, but rather on the subjective level of speaker and hearer.Of course, both attitudinal and discourse connective markers are good examples of this type of expression: they have no truth-conditional or referential meaning and are typically seen as either expressions of speaker attitude or stance (subjective use) or as instructions from speaker to hearer on how the information within an utterance should be integrated within the common ground between them in a coherent way (intersubjective use).25 marking of rhetorical coherence relations between discourse units, as well as the 'interactional level' , where interpersonal relations and the evaluation of information within discourse units have a central place.Cf. also Aijmer & Simon-Vandenbergen (2004), Foolen (2006) and White (2003) on the importance of the concepts of 'heteroglossia' and 'engagement' in linguistic analysis.These concepts reflect "the ways in which speakers position themselves vis-à-vis other, real or imagined, voices.Bakhtin's notion of heteroglossia is thus evoked to account for the fact that linguistic choices reflect either speakers' explicit recognition or denial of the existence of other, possibly divergent discourse" (Aijmer & Simon-Vandenbergen (2004:1784)).23 This term originates from the influential theory on argumentativity in language by Anscombre & Ducrot (1983), which Verhagen's work is strongly influenced by.24 See also Clark (1996: ch. 4), on which Verhagen's (2015) account is based.25 See for instance the definition given by Mosegaard Hansen (2006:25): "The role of mark- In view of this framework and Wakker's earlier analysis, I propose to analyse μήν as an intersubjective particle that is at work in the management of the intersubjective coordination between the perspectives of speaker and addressee(s).26I take it that a speaker using μήν links the information within its host unit with the assumptions, attitudes or expectations that are assumed to be present in the perspective of the addressee(s).Crucially, these are taken to be not in complete coordination with-typically opposite to-those of the speaker.This means that μήν is involved in the negotiation of whether particular information is agreed upon and accepted as being common ground information or not.It signals a discrepancy in the individual sets of assumptions or expectations of each interlocutor at the moment of utterance.On this account, contrast or adversativity is part of μήν's basic semantic meaning (pace Wakker): it always reacts to and points at an alternative view or perspective.In other words, the contrast is of an intersubjective nature.
Thus, by using μήν the speaker indicates that he counters the supposed expectations or assumptions of his addressee(s).An apt paraphrase of μήν is 'unlike what you may expect or assume now' .27The divergent expectations or assumptions may involve the objective (propositional) content that is communicated in the discourse, but-crucially-also the felicity of the speech acts ers, in my view, is to provide instructions to the hearer on how to integrate their host utterance into a developing mental model of the discourse in such a way as to make that utterance optimally coherent."Or Bazzanella (2006:456): "Discourse markers are items external to propositional content which are useful in locating the utterance in an interpersonal and interactive dimension, in connecting and structuring phrasal, interphrasal and extra-phrasal elements in discourse, and in marking some on-going cognitive processes and attitudes."See also section 5 on the categorisation of discourse markers.

26
Other attitudinal particles, e.g.Greek ἦ, only indicate the speaker's stance or epistemic certainty with respect to an object of conceptualisation, i.e., these particles are subjective in nature.Cf. the terminology of Cuypers (2005), who consistently distinguishes between speaker-oriented and addressee-oriented attitudinal particles.27 Note that I have excluded Wakker's term 'anticipation of possible disbelief' from my description, since this term seems to presuppose that μήν only occurs in assertive speech acts, which is not the case, as we have seen.The present analysis also neatly explains the observed high frequency of μήν in reaction to and collocation with sentential negation (e.g.Wakker (1997:219)).As is explained by for instance Verhagen (2005), sentential negation also works on the level of intersubjective coordination, rather than only on the objective content level.It contradicts an alternative view, which is assumed to be (possibly) entertained by the addressee or someone else.It might not come as a surprise then to find μήν and negation in each other's vicinity quite frequently.
made in the given discourse context as well as the proper and cooperative continuation of this discourse.In the next section, I examine in more detail how both the attitudinal and the discourse connective usages of μήν could comply with this general description.

4
The different usages of μήν I have proposed that μήν always indicates a discrepancy in perspectives, i.e., points to a possible alternative view of the addressee.Now, to look ahead a bit, I believe that the variation in the origin of this alternative view is a natural basis for explaining the different usages of the particle.In short, if the alternative view originates from either the extra-linguistic discourse situation (e.g. common ground information) or the previous words of the addressee (turninitial use), μήν is characteristically purely attitudinal in nature (section 4.1).If, on the other hand, this alternative view is evoked by the previous words of the speaker himself (turn-internal use), μήν has an additional discourse connective function, indicating an adversative or contrastive coherence relation between two discourse units of one and the same speaker (section 4.2).

4.1
Attitudinal μήν We have seen in section 2 that attitudinal μήν occurs in various kinds of speech acts.Consequently, it does not always relate to the propositional content of the discourse act that is μήν's host unit, but rather to aspects of its illocution.This is in line with the way the group of modal or attitudinal particles in Germanic languages are typically analysed.Waltereit & Detges (2007:78), for instance, state that they involve answering the question "What do I believe that you believe concerning the felicity of my speech act?"In this way, modal particles are a means to modify the felicity conditions, and particularly the preparatory conditions, of a particular speech act.28It might be helpful at this point to look briefly at two examples of contrastive modal particles in Dutch and German as discussed by Foolen (2006) As Foolen explains, in (11a) the particle toch/doch targets a preparatory condition of a directive speech act, viz.that the addressee does not know that the speaker wants him to do something.The contrastive toch/doch signals that this usual condition does not hold for the present speech act, i.e., that (the speaker assumes that) the addressee does know that the speaker wants him to do this (for instance, because the request has been made earlier).The result is that the request has an indulgent character.In (11b), a preparatory condition of assertions is targeted, viz.that the addressee does not already know the information conveyed by the speaker.Again, the particle signals that this usual condition is cancelled: the speaker assumes that the addressee did know this information.Thus, Foolen points out, utterances like these function as a check: the speaker checks whether the addressee has indeed taken this information for granted.29It is interesting to see whether a similar analysis in terms of felicity conditions can be helpful for a description of attitudinal μήν, at least when it occurs in non-assertive speech acts.In the following, I discuss attitudinal μήν in (i) assertions, (ii) directives, (iii) specifying questions and (iv) yes/no-questions.
(i) In assertions,30 attitudinal μήν can accompany propositional information that has a surprising, unexpected, unprecedented or abnormal character-that is, in view of general extra-textual knowledge about regularities in the world.I give one example:31 29 It is interesting that in Dutch it has even become common to use 'checking' toch also in an iconic way at the end of Dutch utterances (like tag-questions or the colloquial 'right?' in English).30 The present account of μήν in assertions is in line with most of Wakker's (1997) actual analyses of examples in tragedy.Cf. also Kroon's (1995: ch. 11) analysis of Latin vero.

31
The translations in sections 4 and 5 are modified as far as the rendering of μήν is concerned.In English-a language that lacks the group of attitudinal particles-μήν can often be felicitously rendered by means of the contrastive adverb yet and/or by stressing And unlike what you may expect (μήν) I had strange emotions when I was there.For I was not filled with pity as I might naturally be when present at the death of a friend.
Here, the function of μήν can be analysed as signalling that the proposition that Phaedo had strange emotions (θαυμάσια) comes as unexpected for Echecrates.This is based on the general regularity (present in the cultural common ground of the interlocutors) that one typically is filled with pity when a friend dies, as is explicitly spelled out in the following γάρ-act.Thus, μήν signals a discrepancy between the presented information and some general knowledge that is assumed to be present in the cultural common ground (and thus also entertained by the addressee).More typically however, μήν is more closely tied up with the reasoning in the preceding conversation, as it points at an (argumentative) implication or conclusion that has somehow arisen out of the preceding words of the addressee (i.e. a combination of extra-linguistic and linguistic context; see section 5 the verb with a polar do-construction.This also holds for (6) above: 'and yet, you do speak rudely.'In Dutch, which does have attitudinal particles, the particles wel or toch seem to be good candidates, as these particles also involve a denial of some negation or an assumed negative view (cf.Hogeweg (2009) and Foolen (2006)  In (1'), Agathon says that he knew nothing of what he was talking about.From this utterance Socrates infers that Agathon might assume now that he has not spoken καλῶς 'beautifully' either.32This however, is not in accordance with Socrates' own view, which warrants the use of μήν in Socrates' following utterance: 'and unlike what you (might) think now, you did speak καλῶς' .In (13), the Visitor's rather cautious statement-notice ὡς φαίνεται 'as it seems'-that the explanation needed is difficult, makes Young Socrates draw the conclusion that the Visitor will not give the explanation at all.This is in contrast to Socrates' own view and μήν is used to signal this.33 Attitudinal μήν also occurs in what seem to be confirmative assertions, indicating agreement or compliance on the part of the speaker.34In most of these cases, this discourse act is a reaction to utterances that display some degree of what Kroon (1995: ch. 11) sium), in which it is the relation between knowledge and beauty that is at stake continually.33 These two examples of μήν are very much related to the discourse-connective use discussed below (section 4.2.1).See also section 5 and especially fn.67 below.34 This use is called the 'assentient' use of ἀλλὰ μήν and καὶ μήν by Denniston (1954:342-344, 353), but as Wakker (1997) and van Erp Taalman Kip ( 2009) have already pointed out, the countering of expectations or assumptions of the addressee is at stake in these instances as well.
Journal  14).In (4a'), the εἰ-clause implies Critias' lack of full certainty and inherently leaves room for the possibility of a negative answer on Timaeus' part.In Timaeus' view however, the approval for further discussion was already implicitly given and, correspondingly, should not have been questioned by Critias.36In other words, there is a discrepancy in perspectives and this allows the use of μήν in Timaeus' confirmative act.A similar analysis can be given for ( 14), where we find the elliptical rhetorical question τί μήν; used as a confirmative assertion (i.e. as an indirect speech act).37Finally, yes/no-questions in Plato characteristically elicit a preferred reaction of agreement or confirmation (note ἆρ᾿ οὐ 'isn't it the case that' in ( 15)): but the very fact that the speaker uses a yes/no-question-to which inherently both 'yes' and 'no' are possible answers-points out that an opposite view might also be possible, i.e., that the speaker is not (yet) certain whether his addressee is 'on the same page' as he.Thus, it does not need to surprise us that we find μήν also in reaction to yes/no-questions.
(ii) In case of imperatives, the particle pertains to the illocution of the discourse act and could be described in terms of modifying felicity conditions (as exemplified above).I repeat example (7) here with some more context: (7') [This gave me the idea that they must have thought we were joking earlier … When this idea occurred to me, I insisted all the more that we were in dead earnest.]35 Denniston (1954:343-344, 353) also has a category for μήν "substantiating a condition".36 In speech act-theoretical terms, we could argue that μήν signals here that (in Timaeus' view) the preparatory condition of his assertion (i.e. that Critias does not know the information conveyed) does not hold.37 The exact linguistic analysis of τί μήν; is an interesting topic but goes beyond the scope of the present paper.Now, it is quite conspicuous that all of the instances of μήν in specifying questions are found in very similar discourse contexts.The main characteristic is that the μήν-question is typically preceded by an utterance of the addressee (speaker b), in which a possible assumption of the speaker of the μήν-question (a) is rejected.39Note the use of sentential negation οὐ and the explicit ὡς σὺ οἴει 'as you think' in (16).Then, after this rejection, speaker a uses a specifying μήνquestion to ask which assumption should replace the rejected one, i.e., which assumption should be added to the common ground.
An analysis of μήν in these contexts seems to involve Gricean reasoning on cooperative communication (Grice (1975)).Since speaker b only rejects an assumption of speaker a and does not explicitly state which information should be added to the common ground in its place, his utterance is not very informative-the maxim of quantity is violated-and thus not very cooperative.On the basis of the cooperative principle-a regularity present in the cultural common ground between the speakers-speaker b is expected to unfold his own view on the matter now, but he does not comply with this implicit cooperative principle.Thus, speaker A needs to explicitly request him to do so (with the μήν-question).There is a discrepancy in views on the cooperative and proper continuation of the conversation here and it is μήν that signals this.40Note that, again, the alternative view arises out of the previous words of the addressee.This time, however, it involves inferences and assumptions about cooperativeness and informativeness in conversation (the Gricean maxim of quantity specifically).
In speech-act theoretic terms, we could say that μήν signals the cancellation of a preparatory condition of a request again, viz.that (the speaker assumes that) the addressee does not know that the speaker wants him to do something, in this case providing particular information.With μήν the speaker signals that he assumes that his hearer does already know or should have known this (based on common ground information about cooperative behaviour in discourse).The fact that he nevertheless has to ask the question gives the utterance a bit of an indulgent character.
(iv) In Plato, μήν also occurs quite frequently in moves that seem to be yes/noquestions or rising declaratives, as in (3').41Typically, such eliciting moves with 39 See also Denniston (1954:333) and Rijksbaron (2007:143) for this observation.40 The absence or presence of ἀλλά in these utterances might have something to do with the question whether the information (i.e. the rejection) in the previous utterance is immediately accepted by speaker a, as seems to be the case in (2), but not in ( 16).41 An important qualification on the status of yes/no-questions is appropriate here.In As argued in section 2 above, it is unnatural to analyse μήν here in terms of (a high degree of) speaker commitment to the truth of the proposition expressed by the question.But is an analysis in terms of countering expectations and assumptions of the addressee possible here?After all, in the present context there are no clues that the propositional content presented in the discourse act with μήν is in contrast with the view of the addressee (Glaucon): rather, Glaucon continually confirms the reasoning by Socrates.
Ancient Greek, these can be indisputably marked as such by means of introductory question-particles like ἆρα, ἆρ' οὐ (cf.( 15) above) and οὐκοῦν or juxtaposed illocutionary expressions like εἰπέ μοι 'tell me' or ἀποκρίνου 'answer (me)' (cf.Shalev 2001).However, when such indicators are not present; the difference between a yes/no-question (a rising declarative) and an assertion cannot be made on formal grounds in the text, but only on interpretation of the surrounding context.Thus we need to keep in mind that, in these cases, it depends only on editor's choice whether a period or a question mark is printed in a particular text edition (cf.Sicking (1997) on editorial issues like these).At any rate, utterances as exemplified in (3')-which are, of course, very typical of the inquisitive character of Platonic dialogue-clearly elicit a response (preferably agreement) from the addressee; as such, they do not seem to be regular assertions.42 Denniston (1954:337, 344, 351-352) already makes the observation that μήν occurs frequently in premises of an argument in Plato.He assigns the progressive connecting function to the particle in these contexts (but cf.my section 5 below for some qualifications on the alleged connection here).Cf. van Ophuijsen (1993) for the use of ἄρα in Plato.brave-here it is checked whether Clinias is still aware of the fact that they accepted this a little earlier (ἐλέγομεν νυνδή 'we just now said')-and a brave leader, in turn, implies not being easily thrown off balance by fears.
In general, I believe that this analysis of 'checking' μήν can be aligned very well with the inquisitive character of the Platonic dialogue.The printing of the question marks in example (3)-i.e. an interpretation as a yes/no-question or a rising declarative-corroborates my interpretation of μήν and, as far as I am concerned, this is a good thing to do in example ( 17) too.45 Now, if this analysis is correct, we can also explain why μήν in this particular use can occur in collocation with attitudinal που, as in example (8) in section 2 above.I think που has its typical function here: indicating lack of full speaker commitment to the expressed information as to leave room for possible disagreement on the part of the addressee-note also ἢ οὐ συγχωρεῖς; 'or don't you agree?'-and to imply a degree of politeness on the part of the speaker.46Μήν, on the other hand, signals that this information, trivial as it might be, is as yet not activated by the addressee and not taken into account at the present moment in the conversation, most typically because it is in some kind of contrast with the view of the addressee as laid out in the preceding discourse.In other words, μήν targets the illocution here and has a checking function, while που is used to tone down the force of the utterance and save the addressee's face.47 To sum up, I believe that the notion of countering expectations or assumptions is present in all instances of attitudinal μήν.It always indicates a (presumed) discrepancy between the assumptions of the speaker and the addressee, either with respect to the propositional content of the discourse act in question or with regard to its illocutionary force and felicity as such.The possibility of a discrepancy typically arises out of the previous discourse context (spoken by the addressee) in combination with extra-linguistic knowledge about regularities in the world (including communicative patterns).Now, in view of a unified analysis of the particle, it might even be attractive to argue that all instances of attitudinal μήν, even when they occur in assertive 45 At least an editor should be consequent, as Sicking (1997)  speech acts, target the illocution, rather than the propositional content.48This would be possible for ( 1), ( 4a) and ( 13) for instance.However, there are also instances where it seems to be only the propositional content of the discourse act that is relevant for μήν's analysis, e.g. ( 12).As such, it seems better to allow variability in the linguistic entities that μήν can target.

4.2
Discourse-connective μήν I believe that discourse-connective μήν can also be analysed in terms of countering expectations or assumptions, but has an additional structural function: it marks contrastive coherence relations between two discourse units spoken by one and the same speaker and thus connects these units (also in a more structural way, see section 5).It indicates either what in the literature is often called a denial-of-expectation contrast (4.2.1) or a more global discourse contrast (4.2.2).The former is reminiscent of Denniston's 'adversative connecting μήν' , whereas the latter covers his 'progressive connecting μήν' .

4.2.1
Denial-of-expectation contrast A denial-of-expectation contrast typically consists of three elements.49First, there is a contextual issue, the topic under discussion.Second, there is a concession, the information contributed by the first part of the contrastive pair.This contains a partial answer to a contextual question and a confirmation of some information.Thirdly, there is the correction, the information contributed by the second part.The correction initiates a search process for conflicting implications, which is typically an implication of the first conjunct, interpreted with 48 Additional support for this view would be the distributional fact that μήν never occurs in subordinate adverbial clauses expressing a proposition-e.g.temporal, conditional or relative clauses-whereas other Greek particles that are typically taken as belonging to the group of attitudinal particles, e.g.που and δή, do occur in these environments.But admittedly, to come up with conclusive evidence in this matter is quite difficult in a dead language.49 This section is based upon the very similar analyses of e.g.Anscombre & Ducrot (1977;1983), Kroon (1995: ch. 9), Verhagen (2005: especially ch. 2 and 4) and also the more formal work of Spenader & Maier (2009).Cf.Allan forthc.with respect to Ancient Greek.I believe discourse connective μήν behaves very similar to μέντοι in Lysias and Herodotus as described by Sicking (1993:33-35) and Slings (1997) respectively.This should not come as a surprise in view of the etymological relations between these two particles, μέντοι being a combination of original μήν and τοι ('I tell/assure you' , developed from the second person dative pronoun).My analysis would also be in line with that of Oréal (1997) on μήν in Demosthenes' speeches.She compares μήν with the French discourse marker 'pourtant' .respect to the issue.Thus, it involves the denial of an implication or expectation that the addressee might infer from the first part of the contrastive pair (i.e. a denial of an alternative view).Typically, this implication or expectation is derived from generalised (but defeasible) topoi, which are assumed to be present in our world knowledge, i.e., in the cultural common ground (see section 3).The contrastive pair is used to indicate that the speaker is aware of this topos, but that it does not hold in this particular case.An example, taken from Spenader & Maier ( 2009): (18) [Issue: Will Mary and John get married?]Mary is in love with John, but she turned down his proposal for marriage.
Given the contextual issue in ( 18), the first part might be seen as evidence for the conclusion that Mary and John will get married, based on a topos that love is typically a prerequisite for getting married.However, this implication is denied by means of the following but-clause.In other words, the but-clause always cancels the argumentative load of the concessive part.
In Greek this same analysis could be given for the μέν-μήν-pattern (see section 2 above).Here the discourse act with μέν is the concession, the one with μήν the correction or denial.This means that the μήν-act denies a possible implication or expectation that is derivable from the μέν-act and related somehow with the issue at stake.I discuss two earlier examples here: (10') [Ath.Suppose that the truth had been different from what the argument has now shown it to be … could a lawgiver have told a more useful lie than this, or one more effective making everyone practice justice in everything they do, willingly and without pressure?]κλ.Καλὸν μὲν ἡ ἀλήθεια, ὦ ξένε, καὶ μόνιμον· ἔοικε μὴν οὐ ῥᾴδιον εἶναι πείθειν.(Pl.Leg.663e) cli.(μέν) Truth is a fine thing, Stranger, and it is sure to prevail, yet/but (μήν) to persuade men of it seems no easy task.
In example (10'), there is a contextual issue of the desirability of persuading people of something that is true.The μέν-act, the concession, confirms some contextual information, viz.that truth is something beautiful and prevailing.This part is accepted into the common ground.Then the μήν-act, which contributes the information that it seems difficult to persuade people of the truth (in this particular case), corrects a possible implication of the information in the μέν-act.The implication arises out of a topos that normally if something is beautiful and prevailing, it would be easy to persuade people of it.The μήν-act denies or corrects the implication that is raised by this topos.In (4b'), the information in the μέν-act (the discourse act μάλα μὲν ἀτόπου 'very strange') raises an expectation or a possible implication (in this context), which is rejected by the μήν-act (παντάπασί γε μὴν ἀληθοῦς 'in all respects true').The implication is based on a topos that normally very strange things are not regarded as (very) true ones.Now, I believe that the analysis given can easily be extended to instances in which we have no preceding μέν-act, but where an analysis of denial-ofexpectation contrast fits the context very well.50Consider ( 19): (19) [vis.Now it seems that there are two routes to be seen stretching out in the direction of the part towards which our argument has hurried, one of them (τὴν μέν) quicker, dividing a small part off against a large one,] τὴν δέ, ὅπερ ἐν τῷ πρόσθεν ἐλέγομεν ὅτι δεῖ μεσοτομεῖν ὡς μάλιστα, τοῦτ᾿ ἔχουσαν μᾶλλον, μακροτέραν γε μήν (a).ἔξεστιν οὖν ὁποτέραν ἂν βουληθῶμεν, ταύτην πορευθῆναι.νε.σω.Τί δέ; ἀμφοτέρας ἀδύνατον; ξε.Ἅμα γ᾿, ὦ θαυμαστέ· ἐν μέρει γε μὴν (b) δῆλον ὅτι δυνατόν.(Pl.Plt.265ab) while the other more closely observes the principle we were talking about earlier, that one should cut in the middle as much as possible, but/yet (μήν) is longer.We can go down whichever of the two routes we like.y.soc.What?Is it impossible to follow both?-vis.By both at once it is, you strange boy; but (μήν) clearly it is possible to take each in turn.
In (19a), μήν again scopes over an adjective that forms a separate discourse act (μακροτέραν γε μήν).Here, the second road is first described as being more in accord with earlier statements of the Visitor.This might lead Young Socrates to the conclusion that this is the best path to choose.The μήν-act however, provides a counterargument for this conclusion-the fact that this road will be longer-and cancels it.In (19b), the first part of the Visitor's reply most thijs Journal of Greek Linguistics 17 (2017) 73-112 probably triggers Young Socrates to infer that it is indeed impossible to take both ways (note also the use of the rather offensive vocative ὦ θαυμαστέ 'you strange boy').However, this implication is corrected by the following discourse act with μήν, in which the possibility of taking both paths is preserved in a modified version.I end this section with a nice example from a more narrative, rather than a purely argumentative, passage, in which (assumed) expectations on the proceeding of the story are cancelled by means of μήν:51 (20) [Having emptied this great flood of words into our ears all at once like a bath attendant, Thrasymachus intended to leave (ἐν νῷ εἶχεν ἀπιέναι):] οὐ μὴν εἴασάν γε αὐτὸν οἱ παρόντες, ἀλλ' ἠνάγκασαν ὑπομεῖναί τε καὶ παρασχεῖν τῶν εἰρημένων λόγον.(Pl.r. 344d) But/yet (μήν) those present did not let him, but made him stay to give an account of what he said.
Here, the previous discourse could create the expectation that Thrasymachus would indeed leave; at least, this is what he plans to do-and most typically people act in accordance with their plans (a topos in the common ground).However, this possible expectation is denied in the next discourse unit, which has μήν.52

4.2.2
Global discourse contrast Μήν also seems to have a function on a more global level of discourse, viz.marking relations between communicative moves rather than discourse acts within a move.Thus, it seems to involve "the negotiation of coherence and joint coordination of interaction: what are we going to do next?"(Waltereit & Detges (2007:78)).In other words, we could say that μήν in this use is not so much involved in indicating relations between the objective content of discourse 51 Cf. example (26) in Wakker (1997).Note that we are dealing with a narrative spoken by Socrates here.Although there are no other interlocutors in the Republic, I think it is safe to assume that his addressees are still very much present, since we are dealing with a spoken narrative here.Cf. fn. 3 on the notion of diaphony and for some figures on μήν in narrative texts.52 Note also that after the denial, a corrective ἀλλά follows which introduces the things that did happen, i.e., the things that should replace the denied alternative view.I believe that from instances like these (οὐ μήν … ἀλλά …) the particles have become grammaticalised into the fixed contrastive combination οὐ μὴν ἀλλά (cf.Denniston (1954:28-30)).Cf. also fn.18 above and section 5 on diachronic change.
Journal of Greek Linguistics 17 (2017) 73-112 units (as in the examples in section 4.2.1).Rather, it pertains to expectations and assumptions about the argumentative or thematic structure of the discourse itself.53I repeat (5) here as a typical example: (5') [ath.… After these remarks, our law (νόμος) on the subject should run like this: description of the law, about 16 oct-lines.]δεύτερος μὴν νόμος· … (Pl.Leg.919d-920a) But/yet (μήν) there is second law: … Here it seems to be the case that μήν just marks a next move in the discourse, signalling a discourse boundary and a transition to a following (sub-)topic (like δέ also does).This function is called 'discourse contrast' by Kroon (1995).54One might conclude then that the notion of countering expectations or assumptions is not or no longer present in this use.55However, in accordance with Wakker's (1997:226-229) ideas on the progressive connective use, I believe μήν typically does more than signalling a discourse boundary.In (5'), it signals that the addressee probably did not expect this next step at this point: since the first law has been introduced as a singular law (νόμος) and takes a considerable amount of space, the addressee might draw the conclusion that this law suffices.However, a second law is needed as well.
Here is another example: (  Slings (1997) in his description of the discourse connective μέντοι in Herodotus.In the terminology of Kroon (1995) and Wakker (1997), this function of μήν would be on the presentational level of discourse; cf.fn.22 and section 5 below.54 Cf. the view of van Erp Taalman Kip (2009), who uses the notion of 'shift in the focus of attention' to describe some usages of καὶ μήν.Cf. also Allan forthc.on discoursecontrastive μήν.55 Cf. fns.57 and 66 as well as section 5 on the possibility of the diachronic development of a semantically bleached discourse function.But (μήν) consider, both you and Clinias, whether these remarks of ours have any relevance at all to legislation.After all, this is the object of the exercise-I'm not going through all this simply for the story.
In ( 21), we also find μήν at the start of a new step or move within a speech turn.
Whereas the first part of this turn constitutes a reaction to the previous turn of Megillus, the second part, starting with μήν, shifts the focus of attention to a new point about the story just told, viz.its purpose for the present discussion.
We can add that this move of the Stranger might be somewhat unexpected for Megillus and Clinias at this point in the conversation.56I conclude that, as far as Plato is concerned, even the discourse-contrastive use of μήν seems to indicate a possible discrepancy between two perspectives, countering the possible expectations or assumptions of the addressee.Consequently, I think the progressive use should not be seen as a completely separate function or category.The use is somewhat different from the others, however, as it is oriented towards assumptions about the thematic and argumentative structure of a complex move.It signals that a next related issue or topic might not (yet) have been cognitively activated in the common ground between speaker and hearer, but is relevant for the complete discussion of the current, more general theme.Thus, μήν is also of importance for the thematic organisation within argumentative discourse.5756 In my view, ἀλλά could have been used here as well to indicate a discourse contrast (cf.Slings (1997) and Allan forthc.).The difference to μήν seems to be that ἀλλά is more strongly corrective or replacing in character, i.e., it replaces the preceding discourse topic as being completely unimportant or irrelevant.This connotation does not necessarily need to be conveyed by μήν.Some other clear examples of discourse-contrastive μήν are Leg. 729d,778b,828a;Phdr. 224b. 57 This conclusion is in line with observations on the collocation γε μήν by Revuelta Puigdollers (2009:106-109) and George (2009:166).The former, however, also discusses examples from Xenophon's Hellenica, i.e., from a narrative text (see fn. 3 above): I leave it to future research to examine whether there is evidence in Xenophon too for maintaining the original attitudinal value of μήν (countering expectations/assumptions of the addressee) in the discourse-contrastive examples, or that the particle has definitely developed a semantically bleached, i.e. attitudinally neutral, discourse function only (similar to δέ); cf.section 5 and fn.66.I do not think that γε μήν should be seen as a fixed combination of particles, since γε may also occur later on in the μήν-utterance (cf.examples (1) and (3a) above).In other words, γε-a demarcating scope particle functioning at the level of information structure (cf.Wakker 1994:308f.)-andμήν might sometimes just end up in the same postpositive position.In my view, it does not come as a surprise that a parti-Downloaded from Brill.com09/16/2023 01:32:25AM via free access

Some qualifications concerning categorisation
Since μήν executes both strictly attitudinal and discourse connective functions, it is an interesting case study with respect to the issue of particle categorisation.I will make some remarks here about the categorisation of the different usages of μήν and similar particles, which in my view should allow flexibility.I will also relate my data to questions about connectivity and diachronic change of particle meanings or functions.First, it is important to note that, very generally speaking, both attitudinal and discourse-connective particles have a relational function.58Their function is always concerned with integrating or situating their host unit into a wider context, linking two elements with each other.In this sense, all of these pragmatic or discourse markers may be said to be 'connective' , albeit in a very loose way.Thus they are all cohesion devices, contributing to the coherence in a mental or cognitive representation of a particular discourse.59However, as has been recognised by many scholars, these markers typically pertain to particular functional domains: some execute epistemic, modal or interpersonal functions, relating their host units to the extra-linguistic context of the interlocutors' perspectives, stance, (common) knowledge and cognitive reasoning processes, whereas others execute primarily presentational, metatextual or structuring functions, relating their host units to (mental representations of) other discourse units, i.e., to linguistically evoked context.60I take it that the subgroup cle indicating the exact scope of a message (γε) frequently collocates with a particle that indicates a contrast between views or perspectives (μήν): we might say that γε explicitly points at the locus of the contrastive view.58 In a similar way, Aijmer & Simon-Vandenbergen (2004) talk about 'indexicality' as a crucial feature of pragmatic markers.59 I follow Mosegaard Hansen (1998;2006) here; see fn. 25 for her definition of discourse markers.Crucially, the label 'discourse marker' refers to a very broad functional category that contains non-propositional and non-truth-conditional expressions, which may belong to various syntactic categories, e.g., adverbial expressions, prepositional phrases, coordinators or modal particles.The group of particles, I take it, should be defined only in formal and not in functional terms (e.g.small, monosyllabic, uninflectable words).Thus, discourse particles are a strictly formal subgroup of discourse markers.Cf.Kroon (1995) and Pons Bordería (2006) for similar ideas.60 Cf. for instance the categorisation of discourse marker functions by Bazzanella (2006), who discerns cognitive (epistemic, inferential) functions, conversational functions and metatextual functions.The first two correspond with Kroon's (1995)  of discourse connectives is mainly concerned with this second type of function and are connective in a stricter, discourse-structural sense.61Now, we have seen that μήν typically executes functions that belong to the former group, relating discourse units to extra-linguistic assumptions of speaker and addressee ('the alternative view'; cf. the examples in section 4.1).But of course, such an assumption can also be inferred from or evoked by-and as such is more closely related to-the preceding linguistic context.For μήν this is clearly the case in the discourse-connective examples given in section 4.2.1 on denial-of-expectation contrast.Therefore, I think the boundary between marking relations to strictly extra-linguistic and to linguistic context should not be seen as rigid, but rather as a scalar phenomenon.Consequently, the difference between strictly attitudinal and discourse connective particles is also of a scalar nature, which means that particles (in some of its usages) may execute functions belonging to different functional domains at the same time.62Such a conclusion, I believe, is especially relevant for languages that do not feature specific syntactic criteria that delineate a separate formal class of attitudinal particles such as in the Germanic languages.When we look at Dutch toch for instance, we can easily tell the difference between the modal or attitudinal use and the discourse-connective use:63 (22) a. Maar ik kan toch thuis koffie zetten.
And yet, I would wait a bit before doing so.(= Foolen 2006: ex.9) 61 Thus, I understand discourse connectives in a strict sense-as defined by e.g.Fraser (1998;1999), who confusingly labels them 'discourse markers'-as a functional subgroup of the general class of discourse markers (cf.Pons Bordería ( 2006)).In Greek, discourse connective particles like δέ and γάρ belong to this class, as well as the coordinators καί and ἀλλά when they connect discourse acts or moves.See also George (2009:164-169) for the issue of connectivity in relation to Greek particles.62 Cf. van Ophuijsen (1993:79), Sicking (1993:45), Kroon (1995:63-64) for a similar idea.It is also in line with the work of Schoonjans (2013) in which he argues that in view of the categorisation of modal particles and discourse markers it is best to make use of the notion of categorical prototypes.This means that some particles may be a prototypical instance of a particular class, but others reside more at the periphery of a class and may even have typical features of more than one class.63 See Foolen (2006) for an analysis of these data.As a modal particle, as in (22a), toch is unstressed and occurs in the socalled 'middle field' of the sentence, two defining characteristics of modal particles in Dutch (and German), whereas toch as an adversative discourseconnective, as in (22b), is stressed and occurs in the sentence-initial position.For Ancient Greek-as far as we can possibly know-criteria like these are absent: both discourse-connective particles and attitudinal particles occur in the left periphery of discourse units, either at the initial prepositive position (e.g.connective ἀλλά or καί and attitudinal ἦ) or as unstressed postpositives in 'second' position (e.g.connective γάρ or δέ and attitudinal μήν or τοι).Given these considerations, it does not seem to be appropriate to maintain a very strict division between the classes of attitudinal and discourse-connective particles in Ancient Greek.64This conclusion pertains all the more to μήν, since it can combine attitudinal functions with discourse connective ones in some of its usages.
Additional support for this conclusion involves regularities in diachronic change.Cross-linguistically, function words such as particles are known to be prone to easily undergoing formal (phonological) or functional change: over time, a form might acquire new additional meanings or extend its basic meaning into new linguistic contexts, so that a polysemous picture arises.65In my view, this is what has happened with μήν too: whereas the basic meaning of μήν (countering expectations or assumptions of the addressee) remains quite stable over time, I take it that this meaning gradually extends its application to new contexts.More specifically, the discourse-connective use, where μήν behaves as a competitor of the more frequent δέ, is clearly a later extension of the purely attitudinal use.The μέν-μήν-pattern, for instance, is not found in early Greek and the earliest instances are found in Euripides, Aristophanes and 64 See Diewald (2013) for the argument that questions of categorisation of and subdivision between attitudinal and discourse connective particles should be entirely a languagespecific issue.She argues that it has been generally accepted that almost every language has connective items and consequently there seems to be a language-universal class of (discourse) connectives, which is based on functional (onomasiological) criteria.However, attitudinal particles are language-specific and are typically defined as a natural word class on the basis of strictly formal (semasiological) criteria.65 This is the reason why, in my view, a polysemic-rather than a purely monosemicapproach to the description of these words is a priori most promising.See Fischer (2006b) for discussion of the two approaches.See Mosegaard Hansen (1998), Foolen (2006), Waltereit (2006) and Waltereit & Detges (2007) for good examples of polysemous particle studies.For Ancient Greek, cf.Koier (2013)  more frequently in Plato.Thus, it seems safe to conclude that these are (late) Classical Greek extensions of the original attitudinal use of the particle.66This also means that a rigid boundary is not to be expected.With these considerations in mind, I would like to look in more detail at a group of instances that seem to be ambiguous in view of categorisation, viz. the instances of turn-initial μήν in eliciting moves (see also section (iv) in 4.1 above).First, consider ( 23): (23) -οὐκοῦν ἓν μὲν οὐκ ἔσται τἆλλα.
-πῶς γάρ; -οὐδὲ μὴν πολλά γε· ἐν γὰρ πολλοῖς οὖσιν ἐνείη ἂν καὶ ἕν.(Pl.Prm.165e) "(μέν) Well, the others won't be one?"-"Obviouslynot."-"But/yet(μήν) they won't be many either, since oneness would also be present in things that are many." In this example, the previous move of Parmenides has μέν.As such, it seems as if we have a μέν-μήν-pattern here-indicating denial-of-expectation contrast, see section 4.2.1-thatsurpasses the boundaries of speech turns.The addressee, Aristotle, has agreed with the information in the μέν-part ("the others are not one").From this he might draw the conclusion-based on general knowledge-that the others will be many.However, this is not the case either, as explained in the next turn, which has μήν.It seems safe then to state that μήν is again used here as an adversative discourse connective, indicating a contrastive coherence relation between the two moves of Parmenides, which are, however, interrupted by a signal of agreement on the part of the addressee.
In other turn-initial instances, however, such as (3') and ( 15) above, such an analysis is less straightforward.Here it might be questioned whether μήν has 66 This conclusion is very much in line with one of the important cross-linguistic regularities of semantic change distinguished by Traugott & Dasher (2002): meanings tend to evolve from the internal (objective, representational) or external (subjective) level to the textual level of language, gradually adopting a (possibly semantically bleached) discourseoriented function, marking coherence relations between (larger) units of discourse or communicative moves.Correspondingly, the scope of the meanings tends to extend from proposition or illocution to larger stretches of discourse (cf.also Waltereit (2006) on this).This seems to have happened with μήν too: in the discourse-contrastive use, a later development, it seems to relate longer stretches of discourse (i.e.moves instead of acts).Cf. also Traugott (1999) on counter-expectative in fact and Simon-Vandenbergen & Aijmer (2007) for other examples of expressions in the semantic field of epistemic certainty that comply to these clines, as well as fn.18 above.
Journal of Greek Linguistics 17 (2017) 73-112 discourse connective-i.e. more structurally connective-features in addition to the attitudinal use described above.One could adduce the argument that μήν in this use is very frequently combined with coordinators like ἀλλά, καί or οὐδέ, so that it is most plausibly not μήν that structurally connects the units of discourse here.67On the other hand, it also occurs quite frequently without these connectives (as in the given examples).In that case, it is on the one hand possible to argue for asyndeton, i.e., the absence of an explicit connection device between moves within a conversational exchange.68On the other hand, it might very well be the case that μήν also marks structural connection here, indicating the (thematic) coherence between the consecutive steps or moves within the encompassing argument and-also in a structural wayconnecting these moves (like the discourse-contrastive use without interruptions of the addressee; see section 4.2.2).Furthermore, recall that μήν and discourse-connective δέ are never found in collocation in instances like these (see section 2 above).Thus it seems that μήν, when used turn-initially in eliciting moves, prefers to be used in collocation with a coordinator, but has also gained discourse-connective features on its own.Again, it seems best to see it as a scalar phenomenon.This gives us the schematic overview in Figure 1, which represents the different usages of μήν based on the origin of the alternative view.It might originate from purely extra-linguistic context as in example (12) above, but most typically, as we have seen in the other examples in section 4, μήν reacts to an implication evoked by the preceding linguistic context, which could be more loosely related with it, or more strictly tied to a particular expression or construction (e.g. an εἰ-clause or a yes/no-question).The arrow indicates that this should be seen as a scale, as explained above.Note that μήν is never used to deny what is explicitly said in the previous discourse.It always targets implications of some kind.
In my view, the more the origin of the alternative view is tied up with the preceding linguistic context, the more μήν may be said to have an additional discourse-connective function, both marking a contrastive coherence relation between two discourse units and connecting these units in a structural way.

67
I would like to point out here that cross-linguistically the co-occurrence of discourseconnective particles or adverbs and coordinators is very common, e.g., French et/mais néanmoins, English and/but nevertheless, Dutch en/maar toch, etc. Note, however, that the discourse particles may also occur without these coordinators.See respectively Anscombre & Ducrot (1977), Foolen (2006) and Fraser (1998) for these data.Cf. also examples ( 1) and ( 6) given above, where we have καὶ μήν.68 See Sicking (1993:40-44) for asyndeton in Ancient Greek.However, a second relevant parameter comes in here, viz. the speaker of the previous words.This could be either the addressee or the speaker of the μήνunit himself.As argued in sections 2 and 4.2, I think it is safe to conclude that, when the previous words are spoken by the same speaker, μήν is also discourse connective in nature.However, this is less clear when the previous words are spoken by the addressee: μήν prefers to be used in collocation with coordinators and is more purely attitudinal in nature.The use of μήν in eliciting moves, however, seems to take an in-between position, as explained in the present section.

Conclusion
In this paper I have analysed occurrences of μήν in Platonic dialogue and proposed some important reconsiderations to Wakker's analysis of the particle in tragedy in order to arrive at a more generally valid description of μήν in Classical Greek.First, I have shown that the notion of commitment to the truth of the proposition is a problematic notion for a unified description of the semantic value of μήν and that the cognitive notion of countering expectations or assumptions of the addressee (i.e.indicating an alternative view of the addressee) should take its place.This crucially involves the concept of common ground.Second, I have shown that the origin of the contrastive alternative view is a good basis to explain the different usages of the particle, both the Downloaded from Brill.com09/16/2023 01:32:25AM via free access attitudinal usages and the discourse-connective uses.The latter usages are best seen as diachronic extensions of μήν into new contexts.Thus, whereas the basic semantics of μήν remains stable in all of its usages, a polysemous picture arises in which this basic meaning extends to new contexts and new functional domains.It follows that rigid categorisation of different usages is not to be expected and would be rather artificial.
(i) Wakker's most important observation is that Denniston's emphatic μήν belongs to the class of attitudinal or modal particles.8This group crucially thijs Journal of Greek Linguistics 17 (2017) 73-112 figure 1 The different uses of μήν based on the origin of the alternative view ).Other Platonic examples of μήν in directives are: with imperative Plt.263b, Leg.644d, 699d; with hortatory subjunctive Plt.297d and Leg.842a.12 A related speech-act issue is the occurrence of μήν in utterances with a first-person performative verb, e.g., Leg.810e λέγω μὴν ὅτι 'I state that' or Ep.347c ἀξιῶ μήν 'I request to' .Though perhaps not impossible, it seems at least unnatural that a speaker explicitly commits himself to the information that he performs a particular speech act. : Echecrates and Phaedo talk about Socrates' last day in prison and his death.]εχ.… ἀλλὰ πειρῶ ὡς ἂν δύνῃ ἀκριβέστατα διεξελθεῖν πάντα.φα.Καὶ μὴν ἔγωγε θαυμάσια ἔπαθον παραγενόμενος.οὔτε γὰρ ὡς θανάτῳ παρόντα με ἀνδρὸς ἐπιτηδείου ἔλεος εἰσῄει· (Pl.Ph. 58e) echecr.… But try to tell us every detail as exactly as you can.-phaed.