Negation in Modern Greek revisited: selecting between two speaker-based accounts

2.1 Speaker’s interaction with the world

According to Givόn (1979: 107), negative utterances do not appear from nowhere:

Negative assertions are used in language in contexts where the corresponding affirmative has been mentioned, deemed likely, or where the speaker assumes that the hearer – erroneously – holds to a belief in the truth of that affirmative.

This opinion is also held by Foolen (1991: 219):

In a way, descriptive negation also has a contextual or discourse aspect, insofar as the use of a negative sentence always presupposes more or less strongly the contextual relevance of the positive counterpart. Saying The sun isn’t shining today suggests that I or the addressee had the expectation or wish, or did not exclude the possibility, that the sun would shine today.

Similarly, Langacker (1991: 132) suggests that

something is not only in response to some evocation (perhaps implicit) of the positive situation (e.g. I would hardly announce We’re not having pizza for supper unless there were some expectation that we were). In the terminology of cognitive grammar, NEG is conceptually dependent, for it makes salient (though schematic) internal reference to the situation whose existence it denies.

Against this background, consider the following examples, assuming that their “is not” is actually “in response to some evocation (perhaps implicit) of the positive situation” (DPD stands for Holton et al.’s [1997] ‘dependent’^[7]):

(1a) belongs to the last instantiation (i.e. érhete) or to earlier instantiations of the known reality (írthe, erhótan). What can we say of (1b–c), though, which apparently belong to instantiations subsequent to the last instantiation of present reality? Evidently, Langacker’s elaborated epistemic model (see Figure 1) is not much help in this instance. How can we explain the occurrence of dén on the right of the last instantiation of reality, i.e. in the domain of ‘non-reality’?

Langacker’s dynamic evolutionary model, tailored to the future-time epistemic uses of will, provides the basis for a cogent explanation. See Figure 2.

Figure 2:

The dynamic evolutionary model (Figure 6.8 in Langacker [1991: 277]).

Reality, Langacker (1991: 276) comments,^[8]

never holds still. It relentlessly evolves through time, “growing” toward the future as each instantiation of present reality gives rise to the next […] This evolution exhibits a kind of organic continuity: successive instantiations of reality cannot represent totally distinct and unrelated conceptions. Instead, one instantiation bequeaths most of its organization to its successor, which diverges from it only in limited ways, and only as permitted by the world’s structure (emphasis mine).

In this light, the employment of negator dén + tha ^[9] (the future indicative) in (1b–c) comes as no surprise. The established reality lends veridicality to the future reality, as “one instantiation bequeaths most of its organization to its successor”, and this is reflected in the grammar: dén + (future) indicative, instead of mín + subjunctive.

Perhaps the protasis in (1c) deserves special mention in this connection: its conditional an ‘if’ typically reflects speaker uncertainty, whereas the choice of negator and mood, i.e. dé(n) and future indicative,^[10] conveys speaker certainty, in the sense of the previous paragraph. This apparent inconsistency can easily be resolved if we assign the dominant role to the speaker, not to logical entailment. As independently maintained in Nikiforidou and Katis (2000), the conditional marker an is neutral, compared with its rival conditional na, which reveals the speaker’s personal interest and emotional involvement (see Subsection 3.2 below). As is to be expected, therefore, an functions as an “objective” ‘if’ of the Projected Reality, whereas na functions as a subjective ‘if’ of the Potential Reality, as depicted in Figure 2.

On the other hand, (1b–c) mean business for the (non)veridicality approach. To deal with cases such as (1b), as well as with the apodosis in (1c), Chatzopoulou (2019: 73) maintains that “the potential veridicality contribution that can be assigned to NEG1 [dé(n)] in the future tense comes from the notion of assertion, which indicates speaker commitment and is thus veridical”; or, in syntactic terms, that NEG1 being “an unmarked negator in terms of nonveridicality” can, in these cases, be formally “represented as being in the scope of an assert-operator ASSERT(p), while the indication of future tense is below the negator. As a result, NEG1 with future indicative is not in the scope of a nonveridical operator” (2019: 73, fn. 12). But where does this “notion of assertion” come from? Is it also involved in cases like (1a), where the “unmarked” dé(n) is unexceptionally veridical? To the extent that such questions are not anticipated, the assert-operator ASSERT turns out to be a lot like an ad hoc mechanism. An analysis of (1b–c) inspired by Langacker’s ‘dynamic evolutionary model’, as above, explains much more naturally what it is that makes dé(n) welcome in references to the future.

An equally questionable syntactic solution is suggested regarding the protasis of conditionals such as (1c). Chatzopoulou (2019: 166) argues that the “unmarked” negator NEG1 is the only possibility left in Modern Greek, as the – marked in terms of nonveridicality – NEG2 [mí(n)], typical negator in the protasis of conditionals up to the late Medieval stage,

became an element that correlates to the C position, in contributing illocutionary force (as in the case of prohibition, interrogation, and introducing verba timendi complements). The conditional protasis does not offer such a position for NEG2. The C position in conditionals is filled by the ἄν/an/ or the εἰ/i/ conditional particles that now compete with NEG2 and are in complementary contribution.

(See also Chatzopoulou 2013: 28–37). Things are more complicated, however. First, dé(n)’s ancient predecessor u:(k), οὐ(κ) started replacing mí(n)’s ancient predecessor, me:, μή, not after the Medieval stage, but as far back as the time of Christ, cf. τί οὖν βαπτίζεις, εἰ σὺ οὐκ εἶ ὁ Χριστὸς οὔτε Ἡλίας οὔτε ὁ προφήτης; ‘why then do you baptize if you are not the Messiah or Elijah or the Prophet?’(John 1:25). Second, οὐ(κ) may take the place of μή in Attic Greek conditionals introduced with εἰ, provided that the verb in the apodosis has emotional meaning: μὴ θαυμάσῃς, εἰ πολλὰ τῶν εἰρημένων οὐ πρέπει σοι ‘you should not be puzzled if you do not deserve much of what has been said’ (Isocr. 11D). Third, oblique questions introduced with εἰ employ either οὐ(κ) or μή, depending on the negator of the corresponding direct question: compare ἐνετέλλετο […] εἰρωτᾶν, εἰ οὔ τι ἐπαισχύνεται ‘ordered […] to ask him if he is not ashamed’ (Herod. I, 90), with οὐ τοῦτο ἐρωτῶ, ἀλλ’ εἰ τοῦ μὲν δικαίου μὴ ἀξιοῖ πλέον ἔχειν μηδὲ βούλεται ó δίκαιος, τοῦ δὲ ἀδίκου ‘I am not asking you that, but merely if a just person would not have neither the determination nor the will to do this against another just person, but he would against an unjust person’ (in which case the direct question would have been μὴ ἀξιοῖ μηδὲ βούλεται;) (Plat. Pol. 349B).^[11] Fourth, mí(n)’s ancient predecessor, μή, has been contributing illocutionary force (prohibition, interrogation, etc.) since the time of Homer – a fact that nevertheless did not preclude its cooccurrence with conditional εἰ for centuries; cf. μή μ’ ἐρέθιζε ‘don’t irritate me’ (Il. Α. 33), δέδοικα μή σε παρείπῃ ‘I’m afraid that he might fool you’ (Il. Α. 555). Fifth, as examples such as Mána ke na mín agαpái to pedí tis! ‘What a mother! She doesn’t even love her own child!’ in Section 1 reveal, mí(n) cannot unquestionably be considered “marked” in terms of nonveridicality; despite the presence of mí(n) + subjunctive, Mána ke na mín agαpái to pedí tis! does not differ in truth from its counterpart with dén + indicative, Mána ke dén agαpái to pedí tis! ‘A mother who does not love her own child!’. (What really is happening here was hinted at in Section 1 and is more analytically explained in Subsection 3.1.3.)

To end this excursion to (non)veridicality, what is more problematic is the alleged unmarkedness of dén. There is ample evidence to support the conclusion that dé(n) is not less marked than mí(n).

2.2 Speaker’s interaction with the addressee

If the examples in the preceding subsection feature the speaker in interaction with the world, the examples in the present subsection highlight her interaction with the addressee. The discussion, below, draws heavily on Veloudis (2017a, 2017b and the references therein).

Consider the utterances in (2):

(2a), the only example of ‘asymmetric negation’ (Miestamo 2007; van der Auwera and Krasnoukhova 2018) in Section 2, displays a constellation of emphatic Negative Polarity Items (NPIs): the minimizers KÁN, kanÉNAS, to paraMIKRÓ, the generalizers ΤÍPΟΤΕ, POTÉ, and the maximizer kathÓLU. What is the contribution of these emphatic NPIs to the meaning of the utterance? Why do we need them in the first place, given that (2a)’s symmetrical counterpart Dén írthe ‘S/he didn’t come’ is also available? Such questions cannot be answered satisfactorily without resorting to the speaker’s interaction with her addressee. All one needs to do to bring this out is (i) to place a randomly selected version of (2a) in what would be considered its natural context, and (ii) to take into account the etymological origin of its grammaticalized NPI. Consider, for instance, Dén írthe KÁN:

A: – Parémine méhri télus sto sinédrio? ‘Did he stay until the end of the convention?’

B: – Dén írthe KÁN. ‘He didn’t EVEN come.’

As Veloudis (1996, 1998, 2003, 2017a, 2022 points out, KÁN ‘even if’ comes from the conflation of an intensifying KE ‘and’ with the conditional ÁN ‘if’: KÁN is an emphatic, not to say emblematic, concessive NPI. In what way does the concession it expresses benefit the negative reaction of Speaker B? The latter, not being content with a simple No, he didn’t, chooses a circular, estranging path, involving both ‘pragmatic scalarity’ and ‘logical processing’; crucially, on the part of the addressee, as he is essentially invited to participate in a logical ‘game’ (cf. Ahern and Clark’s [2017] ‘signaling games’).

More analytically, KÁN carries out a specific function: as a kind of meta-message, it signals the following theatrical strategy of Speaker B: Even if I do you the favor to check, evaluate, count, mind, bother (to consider), etc. ^[12] the least, most elementary, most rudimentary, etc. condition regarding his staying until the end of the convention (i.e. the fact that he came to the convention), I can assure you this: he did not come (see Veloudis [2022] and the references therein).

Veloudis’ analysis does not end here, however; it goes even deeper. Counting on the current context, he explains, Speaker B subtly furnishes her addressee with all he needs to conclude, deductively, ‘no, he didn’t (stay until the end of the convention)’:

If he stayed until the end of the convention, he came to the venue (where this convention was taking place), at least.

He did not come.

Therefore, he did not stay until the end of the convention.

(modus tollens). It is obvious what this rhetorical strategy amounts to: the negative reaction of speaker B Dén írthe KÁN ‘He didn’t EVEN come’ induces a logical, not simply contingent, truth; crucially, with the involvement of the addressee.

The scalar background of this upgrade should not pass unnoticed. As Israel (2001: 325) more generally points out regarding NPIs and Positive Polarity Items (PPIs):

The need to modulate one’s utterances to achieve particular rhetorical effects is a basic feature of linguistic interaction. The conceptual structure of scalar reasoning turns out to be particularly useful in this respect. By carefully manipulating scalar inferences a speaker can modulate the rhetorical strength of her utterances. The use and distribution of polarity items thus turns out to be the grammatical reflex of a much more general cognitive phenomenon.

There is a whole host of minimizers, maximizers, and generalizers (see Veloudis 2017a) that allow these “rhetorical effects” to emerge. Minimizers, for example, are not a small stock of items; even verbs such as ‘come’ in the logical sequence above may function as minimizers.

Utterances such as (2a) have been widely discussed in the (non)veridicality literature. With a poor outcome, however: (non)veridicality can tell us no more than that NPIs such as KÁN, kanÉNAS, etc., are licensed only in the environment of an antiveridical operator such as dén. This, however, comes nowhere near to what the preceding paragraphs reveal.

Turning now to (2b), Dén érhese?., and (2b′), Dén érhese!?, a negative ‘queclarative’ and a negative ‘whimperative,’ to recall Sadock’s (1970, 1971 long-forgotten terms, are employed for the expression of rhetorical assertions and kind requests,^[13] respectively. Both terms differ from the homomorphic negative question Dén érhese? ‘Aren’t you coming?’ only in the disposition of stress and the intonational contour (the underlining and the symbols ‘?.’ or ‘?!’, adopted here in line with Veloudis [2017b], are intended to represent, respectively, heavier stress and the half-rising intonation contour). What is it, then, that renders formally interrogative utterances suitable, if not indispensable, for the crosslinguistic expression of rhetorical assertions or of kind requests, whose “essential import resides in facets of the speaker-hearer interaction,” to put it in Langacker’s (2008: 475) terms? This question, raised in Veloudis (2017b), was not left without an answer. Assuming that ‘eliciting a yes/no answer on the part of the addressee’ is the plain “story” of polar questions, he suggests that

queclaratives and whimperatives are two adjacent semantic categories which in fact recount this plain story: a yes/no-like response on the part of the addressee is actually at issue in both cases, as queclaratives call for the addressee’s agreement/disagreement with respect to what is being emphatically stated, whilst whimperatives call for the addressee’s compliance, allowing at the same time his non-compliance, with respect to what is being kindly requested. (Veloudis 2017b: 120)

What this amounts to is obvious: according to Veloudis (2017b), negative queclaratives and whimperatives such as (2b) and (2b′) function as figurative products of the source category ‘yes/no question’. In particular, (2b), heavily based on the encyclopedic knowledge the speaker shares (or assumes she shares) with her addressee, becomes a short-circuited yes/no question, “homonymous” with its yes-answer only, “so that it can figuratively stand for it” (Veloudis 2017b: 123). On the other hand, (2b′), seriously considering the addressee’s negative ‘face’, invites him to see a request, actually a Face Threatening Act, “as a question that allows for a quick, as well as foreseen, exit” (Veloudis 2017b: 124).

Two conventional expressions, namely dén mu lés?! ‘tell me!’ (lit. ‘you are not telling me’) and mí(n) mu pís?! ‘tell me more!’ (lit. ‘don’t tell me’), should be mentioned in this connection. What makes these expressions distinct, in addition to their specific intonational contour, is their more or less metalinguistic character. Τhe expression dé(n) mu lés!? functions as a prelude to, or is embedded in, or even at the end of, questions in Modern Greek, coercing the addressee to answer the speaker – an interesting, if not curious, variety of “on-record off-recordness,” to use Brown and Levinson’s (1987: 212) term. Its twin expression, mí(n) mu pís!?, on the other hand, pushes the speaker toward saying more, rather than silencing her. Undoubtedly, dén mu lés?! and mí(n) mu pís?!, as well as queclaratives and whimperatives such as (2b) and (2b′), pose a serious challenge to the (non)veridicality approach, assuming that its adherents were ever willing to look at this section of the relevant data.

This point applies also to utterances such as (2c), Ke pjós dén írthe/erhótan ?!,^[14] because they deepen the speaker-addressee interaction. To begin with, such utterances point basically to the upper value of a pragmatic scale that the speaker shares (or considers shared) with her addressee. Hence, the ‘even’ aura of these utterances. (2c), for example, implies ‘even the most unexpected visitor came,’ which is equivalent to ‘everyone (in the scale) came’. It is no exaggeration to state that the construction at hand functions as a rhetorical maximizer. At the same time, if its negator dén is omitted, the rhetorical character is retained but, crucially, the relevant scale is reversed. Ke pjós írthe?. lit. ‘and who came?’, for instance, implies ‘not even the most expected visitor came’, which is equivalent to ‘nobody came’. (In the same line, the examples in fn. 14 become minimizers, provided their negator is omitted: Ke pjón rótise?. ‘He didn’t ask anybody’, Ke pú épsaksan?. ‘They looked nowhere’, Ke póte ítan lojikós o Yánis?. ‘Yannis has never been reasonable’, Ke tí ídame ekí?. ‘We saw absolutely nothing there’, and so on.) Obviously, this behavior subsumes these utterances under the phenomenon of ‘scale reversing’ (Fauconnier 1975a, 1975b, 1979, 1980). Ke pjós dén írthe?! and its counterpart Ke pjós írthe?! function as rhetorical ways of eliciting what Fauconnier described as ‘quantificational readings’ of superlatives, grammatical, cf. the most difficult, the simplest, etc., and pragmatic, cf. Einstein, Onassis, etc.

2.3 Speaker’s interaction with the text

‘Metalinguistic’ negation^[15] is a “by no means marginal or inconsequential in communication” phenomenon (Horn 1989: 373), the clear characterization of which is pending (see Foolen 1991). In this subsection, the term ‘metalinguistic’ applies to the use of negation in utterances in which the speaker “may contradict what has just been said by another, [s]he may contradict some assumption or implication of what has been said or done by another, or [s]he may contradict [her]self” (Liberman and Sag 1974: 421). The brief presentation of this phenomenon, below, draws heavily on Veloudis (1982, 1986, 2019.

Contrast the negative utterances in (3a) and (3a′):

Apparently, (3a)’s ‘ordinary’ dén differs in function from its ‘contradiction’ twin in (3a′) (adopting Karttunen and Peters’ 1979 terminology; see also Veloudis 1982). Unlike (3a)’s “informative” (Veloudis 1986: 207–8) dén (: dén íhe éna pódi ‘[it is the case that] it lacked a leg’), dén in (3a′) is relatively uninformative as it exercises a rejecting, even erasing, action. Speaker B employs this metalinguistic aspect of dé(n) “to put things in order: a wrong conversational entry is wiped out leaving room to what the speaker considers as the right entry” (Veloudis 2019: 72; 1986: 207–8). That is, before passing on to her own contribution (i.e. ítan kremastí ‘[it] was a swinging one’), Speaker B quotes Speaker A’s contribution just to delete it from the context (dén “íhe éna pódi”).^[17] To recall Horn’s (1989: 397) eloquent depiction of metalinguistic negation, A’s contribution is in B’s response but not of it. This action is figuratively reflected in the ‘rise in pitch’ that characterizes Speaker B’s negative utterance. Since Bolinger (1965), it is widely acknowledged that a rise in pitch, his so-called ‘accent B,’ indicates incompleteness. See Veloudis (1982: 362–7) for the employment of this ‘marked’ pitch contour in Modern Greek context-bound denials.

Note that, as a rule, metalinguistic and ordinary negators do not differ in form across different languages. This is quite natural, provided that we espouse Langacker’s conjecture that “negation is absence from a space”, e.g. from present reality (1991: 133 and fn. 14; emphasis mine). Metalinguistic negation, as outlined in the preceding paragraph, can be understood as absence from a conversational space.

Apparently, then, we are dealing with two sides of the same coin, as this is also testified by the second negator of the internal cylinder. Similar to dé(n), the non-verbal negator óhi also has both an ordinary and a metalinguistic function;^[18] cf. (3bB), (3b′) and (3cB), (3c′), respectively, below (adapted from Veloudis [1982], examples [36a′], [41a] and [57]):

Óhi to Yáni in (3b) is the elliptic version of the sentence negation Dén épjasan to Yáni ‘They didn’t arrest Yannis’. The same holds for alá/ke óhi to Yáni in (3b′), which can be considered a verbose version of (3b.B).

On the other hand, (3c) exemplifies a corrective function of óhi: it acts as an a posteriori eraser of the wrong information (óhi to Y á ni); see Veloudis (1982: 339–44, Appendix) for this context-bound, erasing function of óhi – ‘metalinguistic’, according to the later terminology (Horn 1985; Veloudis 1986). Giannakidou (1998: 50) rediscovered the metalinguistic function of óhi; however, she does not at all mention the metalinguistic function of dén. To the best of my knowledge, metalinguistic negation has not been a subject of attention in the (non)veridicality literature.

Finally, (3c′) presents an interesting case of constituent negation. As pointed out in Veloudis (2019), the speaker appears to distance herself from her own characterization of the visitors’ comings: according to her alethic calculations, líjes sounds too strong and should be toned down. Αs with metalinguistic negation, the negated material is again a priori available, only this time not as a quotation of someone else’s saying: both líjes and its negation belong to the same person, i.e. the speaker. A contradiction, in all appearances! What rescues her is ‘litotes’, a figure of speech known for a long time: the speaker exploits the correcting action of her óhi to leave room rhetorically for negotiation of what her líjes is expressing in rather absolute terms. For the similarities, and dissimilarities, between metalinguistic negation and litotes, see Veloudis (2019).

2.4 Summary

It is well known in the relevant literature that a peculiar asymmetry characterizes the negative side of the Square of Opposition. See Figure 3.

Figure 3:

The square of opposition (see Veloudis [2019] and the references therein).

According to Horn (2012: 419), “[i]n many, probably most, languages, neither of the negative positions (E or O) is lexicalized. But the generalization remains sound: A, I, and often E values may lexicalize, O values may not”. The O values, in other words, resist univerbation (cf. *nall), in contrast to the E values (cf. none). Veloudis (2019) argues that this mysterious and, as it turns out, universal restriction is not inexplicable once we admit that it is the expressive speaker’s interaction with her addressee that drives this – otherwise logical – structure.

Our discussion in the preceding three subsections revealed that Modern Greek belongs to the languages that lexicalize A, I, and E values; and, more importantly, that it falls nicely under the rubric of Veloudis’ (2019) speaker-based conception of the square. This fact allows one to consider the examples so far discussed from a different, more general, perspective. Indeed, (apart from the interrogative in form [2b], [2b′], and [2c], of course) they echo the right-side corners, E and O. Specifically, as Veloudis (2019: 68) remarks, the E-corner (i.e. no, none, nothing, etc.), contradictory to the I- corner (some, someone, somewhere, etc.), hosts the crosslinguistic reinforcement of negation through the engagement of – the open class of – minimizers (no, no one, nowhere, etc.), on both a synchronic and a diachronic basis. Following the Jespersenian history of Modern Greek dé(n), this negator is the remnant of u:dén, οὐδέν, i.e., of the intensified form of the ancient indicative negator (<οὐ ‘not’ + δέ [INTENSIFIER] + ἕν ‘one’) meaning ‘not even one’.^[19] This fact obviously suggests that not only (2a) but also (3a), (3b.B), (3b′), and the examples in (1) above are reflections of the E-corner.

On the other hand, the O- corner (not all, not every, etc.), contradictory to the A-corner (all, every, etc.), is the habitat of the counteractive function of negation. As Veloudis (2019: 72) puts it,

due to its contradictoriness, O-corner’s not suggests that everything is true but what falls within its scope, i.e., but the inhabitant of the A-corner. This is in fact what makes this corner’s negation special: it carries out an essentially correcting function, which however leaves us halfway: we know what is to be modified, rejected, revised, disallowed, etc., but not what its appropriate replacement is. In my view, it is this correcting function that makes O-corner′s – but not E-corner′s – not an ideal candidate for rhetorical devices like litotes, as well as for the expression of metalinguistic negation […].

This means that the O-corner is the base of the operations of the metalinguistic and litotic ‘not’, hosting examples like our (3a′B), (3c.B) and (3c′).

Evidently, there is no place for the other two negators, mí(n) and mí, in the Square of Opposition. The next section considers why this is the case.

3.1 Negative mí(n)

According to Liddell and Scott (1996), the ancient μή, me: displays subjectivity, compared with its objective rival οὐ, u:; moreover, it is “used in cases where the Negation depends on some previous Condition, either expressed or implied” (A Lexicon Abridged from Liddell and Scott’s Greek-English Lexicon [1889: 442]). The functions of me:’s modern descendant, mí(n) μή(ν), are quite similar. The subsections, below, display (albeit non-exhaustively) a spectrum of the speaker’s emotional involvement with something “either expressed or implied” (see the quotation above). More specifically, they deal with the feeling speaker’s interaction with the world (3.1.1), with the addressee (3.1.2), and with the text (3.1.3) – a tripartite interaction, again, as in the case of the thinking speaker, Section 2.

3.2 Affective mí(n)

The preceding subsections did not exhaust the spectrum of a speaker’s personal involvement regarding “some Condition, either expressed or implied”, to recall the quotation from A Lexicon Abridged from Liddell and Scott’s Greek-English Lexicon (1879) in Section 3.1 above. The present subsection develops this personal involvement further, reflecting to a large extent the functions of ancient me.

The questions below display an emotional speaker; she is genuinely worrying (10a), she is anything but indifferent (10b), she rhetorically paves the way for her dramatic announcement (10c):^[31]

Crucially, mí(n) has not entirely abandoned its original negativity in this new role (i.e., as a head of questions), with or without na: what the speaker refers to is essentially either meant for rejection, as in (10a) and (10c),^[32] or considered as rather impossible, as in (10b). More specifically, (10a) would never express something that the speaker wishes; hence, the unacceptability of (10a′), below:

Similarly, in the (rhetorical) speaker’s view, (10c) deserves nothing but a negative answer. In general, in folk poetry, na mí(n) occurs in reverse order, mína ‘lest’, to form rhetorical questions intended for negative answers;^[33] crucially, for negative answers introduced with midé (<ancient μηδέ), which in this case acts as a veridical operator.

The examples in (11), on the other hand, imply negative emotions experienced by a stressed subject:

([11a–c] are adapted from Veloudis [1982: 12–4]). According to Liddell and Scott (1996: 1124), me: occurs “after verbs of fear or apprehension (μή οὐ)”. Μodern Greek mí(n), “a virtual rather than an explicit negative” (Landsman 1988–1989: 28), obviously continues this transition of its ancestor, with or without the companion of na. In (11a) and (11b), as commented on by Veloudis (1982), the speaker expresses her concerns about somebody coming and not coming, respectively, as she wishes neither to happen. The same holds as far as (11c) and (11d) are concerned: again, the speaker expresses her concerns about accepting and not accepting a gift, respectively, as she wishes for neither to happen. There is one interesting difference, however: as Veloudis (1982: 14) remarks, the na mí version of (11c) is liable to a second reading because the subject of the main verb is identical to the subject of the complement. Thus, (11c) is ambiguous between the readings ‘I fear that at the end I will be compelled to accept the gift’ (affective mi(n)) and ‘I am afraid not to accept the gift’ (negative mi(n)). In the first reading (heavy stress on déhtó), the speaker appears to have a deep, unstated desire to accept the gift, whereas in the second reading (heavy stress on fováme and on mí), she feels obliged to accept an unwanted gift (see Veloudis 1982: 14, fn.1). This ambiguity is removed when na is absent, hosted in the unstressed, affective mi. On the other hand, (11d) is unambiguous, for mi is liable to only one reading: it is unavoidably affective (otherwise, dehtó would be negated twice).

Arguably, direct questions and verbs of fear have something in common: a positive or a negative answer may equally follow the former, whereas reality may or may not finally put flesh on the bones of the fear expressed. Therefore, we can legitimately suppose that this indeterminacy (actually, the speaker’s uncertainty) has possibly left its trace on uses such as those above, turning mi(n) into a degenerated negator. What, however, caused mi(n) to be inserted into such contexts in the first place? In the background of uses such as those in (10) and (11), lies something that the speaker disapproves of, finds difficult to foresee, or even wishes to block out. ‘Let nothing bad happened to him!’ is, for instance, what the speaker feels when asking (10a), and an unwanted arrival is what she anguishes over in (11a). For a parallel explanation, see Alexiou (2005: 63–5) and Theodoropoulou (2001: 254).

Similarly we could account for the use of mi(n) (i) in rhetorical questions like Na mi milúsa?. ‘shouldn’t I speak? = What was I supposed to do, keep my mouth shut? Of course, I didn’t!’ (the speaker pretends she responds to an interlocutor’s claim she should have kept her mouth shut) and (ii) in subordinate clauses introduced by verbs other than verbs of fear like Éhete to nu sas mi sas ksefíji ‘Be careful so he does not slip away’ (two of the relevant examples cited in the Dictionary of Standard Modern Greek 1998). For a more general discussion on the function of mí(n) in rhetorical questions, see Alexiou (2005).

3.3 Negative mí

Mí is extensively discussed as a distinct negative particle of Modern Greek in Veloudis (1982: 4–83): it is considered a fourth negator that does not share the morphosyntactic and phonological characteristics of mí(n) and “should not be confused with the forms of the latter that lack (under certain conditions) their final/n/” (1982: 4). We can roughly think of mi as óhi’s counterpart in the external cylinder: it occurs in elliptical utterances counting on the context, linguistic and otherwise. Thus, in uttering (12a–b), below, the speaker discourages her addressee from – or asks him to stop – doing something she does not approve, whereas the common exclamation in (12c), “an elliptical expression of disapproval, suggesting, ‘may I never live to see worse than this’” (Mackridge 1987: 244), is a comment on what the annoyed speaker has just heard or seen:

Mí may also act as a negator of adjectives, adjectival participles, or even nouns of mainly verbal origin such as apódosi (<apodído ‘attribute, ascribe’), katáthesi (<katathéto ‘deposit, testify’), aksiopíisi (<aksiopió ‘make good use of’) etc.^[34] Crucially, only as a negator of adjectives is this mí interchangeable with óhi (maybe because adjectives display an enduring quality, indicating the established reality of the internal cylinder, which is the typical locus of óhi’s activity):

Note that all examples in (13) radiate the speaker’s disapproval/frustration to some degree. For instance, in (13a), borrowed from Veloudis (1982: 43, ex. [47c]), she appears to imply that it would have been better if the subject had married a famous actor.