On the non-homogeneity of expletive negation in Greek and beyond

2.1 The negative uses of dhen and min

The negative marker dhen is used in Greek to negate the verbal predicate of an indicative sentence (Holton et al. 2012 [1997]: 509–515). It can thus be regarded as the exponence of standard negation in the sense of Dahl (2010).

In Example (6), dhen syntactically embeds a Tense Phrase (TP) denoting the proposition It is raining in Lisbon and reverses its truth conditions. The denotation of the whole sentence in (6) is the complementary proposition corresponding to It is not raining in Lisbon. In this sense, dhen is interpreted as canonical sentential negation, formally defined as a function that takes a proposition p as its argument and returns its polar propositional alternative ¬p:

As for the negative marker min, its distribution is complementary to the one of dhen. For this reason, min is often labeled as the subjunctive negative marker (Holton et al. 2012 [1997]). Given the fact that Greek lacks morphological marking of subjunctive mood though (Roussou and Tsangalidis 2010), min is best described as a polar variant of dhen, i.e., as a variant dependent on the presence of a semantic licensor (Chatzopoulou 2018). Specifically, min emerges only in contexts involving a non-veridical operator (Giannakidou 1998), such as the directive speech act operator in (8) or the operator introduced by the want -predicate in (9).

The English translations show that min is interpreted as a run-of-the-mill sentential negation marker; the content of the order in (8) and the speaker’s desire in (9) involve the addressee’s not going to Lisbon. In principle, then, one could attribute to min the same denotation as dhen:

When comparing the negative uses of dhen with those of min, however, one can easily see that (10) cannot be entirely correct. The dhen-sentence in (6) denotes a proposition of negative polarity ¬p, which is true under a state of affairs such that there is no rain falling in Lisbon at the moment of utterance. The min-sentences in (8–9), on the other hand, cannot be characterized as true or false; they both express the speaker’s preference for possible worlds where the negative proposition corresponding to The addressee does not go to Lisbon is true (Condoravdi and Lauer 2012; Giannakidou and Mari 2021; Kratzer 1981, 1991; Portner 2009, among others). Reference to possible worlds is essential for the interpretation of the negative occurrences of min. Consequently, min can be dubbed as the Greek modal negative marker.

For our current purposes, the claim above can be formally translated using the standard theory of modality by Kratzer (1981, 1991). First, I adopt the simplified definition of an ordering source shown in (11).

Then, I adopt the definition of Best worlds as derived by the ordering source g(w):

Now that the relevant terms have been defined, the negative marker min can be claimed to introduce an ordering source such that Best includes only worlds where ¬p is true; worlds where the addressee does not go to Lisbon, in the case of our Examples (8) and (9) above. Min, then, is formally defined as a function that is always interpreted with respect to an ordering source, it takes a proposition p as its argument and ensures that the complementary proposition ¬p is true in all the possible worlds that are Best worlds.

Summing up, Greek uses both the negative marker dhen and the negative marker min to negate a proposition. The main difference between these two, which for the most part determines their distribution, is that the latter further involves a modal component.

2.2 The non-negative uses of dhen and min

Several researchers have identified allegedly non-negative, that is, expletive, uses of the Greek sentential negative markers, mostly extending insights from Romance or Germanic languages. Espinal (1997) cites the following example from Greek:

Under the indicated reading, the sentence in (14) corresponds to an exclamative conveying the speaker’s surprise at the number of people that the person talked about cheated when he was young, not the number of people that he did not cheat. In other words, it appears that the negative exclamative in (14) features an expletive instance of the negative marker dhen.

Romero and Han (2004) report that the negative marker dhen can receive a non-canonical interpretation also in the context of Greek polar questions.

According to the authors, the preposed negation question in (15) encodes the speaker’s previous belief that Yanis actually did drink coffee. Abstracting away from the specifics of the example, Romero and Han (2004) claim that negative polar questions with preposed dhen encode the speaker’s bias towards the positive propositional alternative p, not its negated complement ¬p. In this sense, Greek preposed negation questions can be argued to be expletive dhen hosts.

Negative rhetorical questions could also be added to this list:

Greek rhetorical questions can be identical in form to genuine information-seeking questions (Holton et al. 2012 [1997]: 503–511; Veloudis 2018). In Example (16), the Greek equivalent of after all is meant to strongly favor the rhetorical reading (Sadock 1971), according to which the speaker in a way asserts that rain is useful (Reese and Asher 2009; Sadock 1971). Crucially, under this interpretation, dhen is arguably expletive in (16) as well.

Intriguingly, Greek root (17a) and embedded polar questions (17b) may feature expletive instances of the negative marker min, too (Chatzopoulou 2018; Makri 2013; Tsiakmakis et al. 2023).

A felicitous use of the questions in (17) does not require the existence of compelling evidence against the positive proposition corresponding to The kid is sleeping. In fact, chances are there is evidence in favor of this positive propositional alternative whenever either question is used (see Tsiakmakis et al. 2023). Therefore, (17a–b) do not behave like standard negative polar questions in the sense of Büring and Gunlogson (2000); they involve occurrences of min that are interpreted non-negatively.

The negative marker min receives the same expletive interpretation when it introduces complements of predicates denoting fear (Chatzopoulou 2018; Makri 2013; Tsiakmakis and Espinal 2022):

Notice that, despite the presence of min, Example (18) conveys the speaker’s fear that she will be late for class, not her fear that she will not, as would be expected if min were interpreted as a canonical sentential negation marker.

Tsiakmakis et al. (2023) mention in passing the last context that licenses expletive occurrences of min in Greek, namely conditionals.

While this use is understated and understudied in the linguistics literature, min is very productive and unambiguously non-negative in conditionals. According to (19), the addressee’s whining is conditioned on the speaker’s talking, not their silence.

Putting together all the above, the Greek expletive negation hosts form a subset of the general host inventory compiled in the Introduction. The list in (20) summarizes the distribution of Greek expletive negation, displaying its morphological exponence in each case.

Building on the Greek sentential negation landscape described in the present section, I will firstly proceed to motivate two general claims:

1. Greek expletive negation hosts can be classified into two groups, depending on whether they feature the negative marker dhen or the negative marker min.

2. The interpretative contribution of the expletive occurrences of dhen and min retains at least parts of the meaning of their respective non-expletive counterparts.

Then, I will present some first empirical arguments to motivate the following hypothesis as a future research agenda:

1. The bi-partite distinction between dhen-expletive negation hosts and min-expletive negation hosts extends to languages other than Greek.

With this triple goal in mind, the non-negative uses of dhen and the non-negative uses of min will now be studied separately and in depth.

3.1 Greek negative exclamatives

The exclamative sentence type in Greek prototypically involves an initial wh-element optionally followed by the complementizer pu ‘that’ (Holton et al. 2012 [1997]: 516).

Assuming the standard view on exclamatives (Michaelis 2001; see also Castroviejo 2006, 2021; Rett 2011, among others), Example (21) conveys the speaker’s surprise at the extreme degree of niceness that Petros’s clothes exhibit. In principle, Greek grammar allows for negative exclamatives, too:

Villalba (2004) claims that exclamatives do not easily admit sentential negation cross-linguistically, because the negative semantics is not compatible with the factivity and the extreme degree reading contributed by the exclamative operator. Indeed (22), under the reading according to which the speaker expresses their surprise at the extreme degree of niceness of the clothes that Petros did not buy, is not very productive.

Interestingly, if the complementizer pu is dropped, the negative exclamative above becomes ambiguous, allowing both the interpretation given in (22) and the one shown in (23) below; a speaker is definitely more likely to intend the latter.

The English translation of (23) does not involve negation, despite the presence of the negative marker dhen in the Greek version of the example. In other words, (23) instantiates the expletive negation host that has so far been described as negative exclamatives. But why do Greek negative exclamatives differ from positive exclamatives in rejecting the characteristic complementizer pu altogether?

Another puzzling asymmetry between negative exclamatives and their positive counterparts is revealed when looking at sentences that feature singular wh-phrases.

In Example (24a), which is introduced by the singular unmodified wh-phrase ti bluza (lit. ‘what T-shirt’), an extreme degree reading is coerced, as is expected in an exclamative (Rett 2011). Example (24b) on the other hand, which is minimally different in further including an instance of dhen, receives a quantity reading.

Combining the last two observations, Greek negative exclamatives differ from their positive equivalents in rejecting the complementizer pu and receiving a quantity instead of a quality reading. On a first note, such syntactic and interpretative asymmetries cannot be attributed simply to the presence of an (expletive) negative marker; in this case, dhen. On a second, more worrisome note, the two properties that negative exclamatives fail to display make up the definition of the Greek exclamative sentence type. In view of the above, I propose that the Greek expletive negation host previously dubbed as a negative exclamative actually corresponds to negative rhetorical questions uttered in situations where the speaker happens to be surprised (see Rohde 2006, for a similar view inspired in other languages). Notice that Espinal’s (1997) example from Section 2.2, repeated below for convenience, is fully compatible with a rhetorical interpretation:

Treating negative exclamatives as rhetorical questions straightforwardly explains the data presented: the complementizer pu is out because it is not appropriate for the interrogative sentence type (Holton et al. 2012 [1997]: 503–508), and the degree reading does not arise because there is no exclamative operator in the first place. But does this analytical alternative have anything to say about why or how the negative marker dhen ends up being interpreted non-negatively in such environments? In effect, it does not; it simply defers an answer to this question until the discussion of rhetorical questions in the immediately following subsection. This line of analysis is enlightening, though, to the extent that it shows the expletive interpretation of dhen to be independent of the exclamative force of a sentence.

3.2 Greek negative rhetorical questions

If negative exclamatives exemplify the same sentence type as negative rhetorical questions, this is exactly where one needs to look next. To further highlight the affinity between the two, let us stick to Example (25Q) from Espinal (1997), this time without exclamative force:

In the absence of context, (26) is ambiguous. On the one hand, it can correspond to an information-seeking question via which the speaker seeks to know the number of people that the person talked about did not cheat in his youth. On the other hand, (26) can be meant as a rhetorical question, used by a speaker who wants to assert (Reese and Asher 2009; Sadock 1971) or remind the addressee of their shared knowledge (Caponigro and Sprouse 2007; Rohde 2006; Veloudis 2018): that person cheated many people in his youth. Despite the presence of the negative marker dhen, and irrespective of the specific account of rhetorical questions that one prefers, the question in (26) is associated with a positive polarity proposition under its rhetorical reading. Therefore, Greek negative rhetorical questions have been reported as expletive negation hosts (Makri 2013).

One would be tempted to attribute the asymmetry between the form (negative proposition ¬p) and the meaning (positive proposition p) of negative rhetorical questions to the expletive negative marker itself; see for example Delfitto and Fiorin (2014). However, such an analytical choice stumbles upon two empirical observations brought forth by Rohde (2006). The first one is that negative rhetorical questions do not always convey a proposition of positive polarity.

Under its rhetorical reading, the question in (27) may indeed be used as a conversational move parallel to asserting that everybody likes meatball soup. However, if in the context of utterance it is shared knowledge that dad does not like this type of soup, the question can be used simply to refer to him:

Notice that the negative marker dhen is interpreted canonically in (28A), even though the question receives a clearly rhetorical reading.

The second empirical observation of crucial importance to the present discussion is that positive rhetorical questions show an effect parallel to their negative counterparts, obviously in the absence of an expletive negative marker that could be charged with such an effect (Rohde 2006).

Under its rhetorical use, (29) is conversationally similar to the assertion that the person talked about did not cheat anybody in his youth. Is it feasible and/or economical to assume that asymmetries in polarity are derived via expletive negative markers in negative rhetorical questions but via different mechanics in the case of positive rhetorical questions?

I take Rohde’s (2006) observations to show (i) that the polarity reversal effect is not always present in rhetorical questions, and (ii) that the rhetorical effect is independent of the presence of an (expletive) negative marker. In other words, the non-negative interpretation of dhen in Greek negative rhetorical questions and – consequently – negative exclamatives is an epiphenomenon produced by the polarity reversal that most often accompanies rhetoricity. In light of this conclusion, there is neither motivation nor reason to assume that the occurrences of dhen found in negative exclamatives and negative rhetorical questions are different from the standard negative marker dhen, which was defined in Section 2.1 as a function that takes a proposition p as its argument and returns the complementary proposition ¬p.

The canonical negative meaning of dhen in these two expletive negation hosts is masked due to the coincidental presence of an independent factor affecting interpretation, namely rhetoricity. Whether rhetoricity in its turn is best captured in terms of informativity reversal (Delfitto and Fiorin 2014) or as an independent epistemic operator (Tsiakmakis 2023) lies beyond the scope of the present paper.

3.3 Greek preposed negation questions

The last expletive dhen host identified was what Romero and Han (2004) dub as preposed negation questions. Concretely, the authors compare pairs of negative polar questions similar to the following:

According to Romero and Han (2004), Example (31a) would correspond to a standard negative polar question featuring a canonically interpreted negative marker dhen. A speaker is expected to ask (31a) when they want to find out whether dad ate meatball soup and they have evidence that he did not (Büring and Gunlogson 2000); they might have a previous belief favoring the positive answer to this question, but this is not reflected in the form of the question. Example (31b) is allegedly rather different. A speaker may use this preposed negation question in order to double check either the positive proposition corresponding to Dad ate meatball soup or its negative counterpart corresponding to Dad didn’t eat meatball soup (see also Ladd 1981). In either case, the speaker necessarily holds a prior belief in favor of the positive propositional alternative.

If Romero and Han’s (2004) empirical claim above is true, and if the preposing of dhen can be systematically linked to the expression of positive speaker bias, negative polar questions in the form of (31b) constitute an expletive negation host. However, the validity of this claim relies heavily on the reality of the distinction between non-preposed negation questions (31a) and preposed negation questions (31b) in Greek, solid evidence for which is extremely hard to come by.

To begin with, the distinction proposed by Romero and Han (2004) is based on the relative position of negation with respect to the syntactic subject. But Greek is a pro-drop language (Holton et al. 2012 [1997]: 575–577). If the subject is omitted, as is most frequently the case in languages that allow this, the questions in (31) end up being string-identical:

Moreover, the preposed negation variant can appear as part of alternative questions.

Given the well-known incompatibility of bias with alternative questions (Pope 1976; Sailor 2013, among others), Example (33) suggests that Greek preposed negation questions are not necessarily interpreted as biased.

Looking deeper, there are reasons to doubt the very stipulation that the negative marker dhen in questions like (31b) is preposed, that is syntactically higher than its counterpart in (31a). On the one hand, Greek preposed negation questions are perfectly compatible with Negative Polarity Items, such as akoma ‘yet’. This can be explained only if dhen is interpreted as canonical propositional negation (Ladd 1981).

The truly negative status of dhen in preposed negation questions is corroborated further by again-modification (see Goodhue 2022):

Again-modifiers trigger a presupposition that the content of the proposition over which they scope has happened before (von Stechow 1996). In (35), the presupposition is that dad has refrained from eating meatball soup at least once before. This means that dhen is interpreted as part of the proposition in this case and, thus, has a canonical sentential negation status.

In the absence of evidence supporting the distinction between Greek preposed negation questions and non-preposed negation questions, and actually in the presence of evidence undermining it, I propose that the two types of questions can be collapsed into one (see van Rooy and Safarova 2003, for a similar claim regarding English negative polar questions). They all feature occurrences of dhen interpreted as standard sentential negation:

The truly negative meaning of dhen may be obscured by the conversational goal of the speaker (see van Rooy and Safarova 2003); if, in a given context, the positive answer has higher utility value for the speaker than its negative counterpart, dhen is more likely to be interpreted non-negatively. The negative meaning of dhen may also be masked by the speaker’s epistemic bias (Büring and Gunlogson 2000; Sudo 2013); a strong belief in favor of the positive propositional alternative can lead to a non-negative reading of the question. It is highly probable though that speaker bias does not interact with the form of Greek negative polar questions.^[6]

3.4 On the expletiveness of dhen

Let us put together the insights gathered via the individual study of Greek negative exclamatives, negative rhetorical questions, and preposed negation questions. Greek expletive negation hosts that feature dhen arguably involve a standard sentential negation marker, whose negative meaning is masked by factors independent of the intrinsic properties of this particular negative marker (see Abels 2002, for a similar claim on instances of Russian expletive negation). In other words, expletive occurrences of dhen maintain the full interpretative load of their non-expletive counterparts; both instantiate the very same negative marker.

4.1 Greek fear-predicate complements

When a Greek predicate denoting fear takes a whole sentence as its complement, this complement may be introduced by min:^[7]

The negative marker min is unambiguously interpreted non-negatively in this context (Chatzopoulou 2018; Makri 2013; Tsiakmakis and Espinal 2022, among others). This is shown not only by the English translation provided, but also by the fact that min can co-occur with the complementary negative marker dhen in this position, in which case the speaker’s fear is about the negative propositional alternative:

Greek is a Strict Negative Concord language (Giannakidou 1998; Zeijlstra 2004). This means that canonical sentential negative markers can license NCIs such as the emphatic tipota ‘nothing’, without giving rise to a double negation reading. The non-negative status of min in fear-predicate complements is further witnessed by its inability to license tipota (see Chatzopoulou 2018; Makri 2013; Tsiakmakis and Espinal 2022).

Notice that tipota is licensed if the canonical negative marker is also added:

The data above clearly show that Greek fear predicates are expletive min hosts but tell us very little about the exact status of min in such contexts. Digging deeper in this direction, Makri (2013) observes that min is incompatible with epistemic adverbs, such as isos ‘maybe’ for example.

Makri (2013) takes this incompatibility as an indication that expletive min in fear-predicate complements is interpreted as an epistemic modal itself. Tsiakmakis et al. (2023) explore this hypothesis even further and offer experimental evidence that non-negative min is indeed interpreted as an epistemic modal conveying positive speaker bias. The speaker is not clueless as to whether dad ate the meatball soup; they consider it more likely that he did. Formally, expletive min is defined as a function interpreted with respect to an ordering source that takes a proposition p as its argument and ensures that this proposition p is true in all the possible worlds that are Best worlds – that is in all the possible worlds that are ranked higher by the respective ordering source.

Notice that min in fear-predicate complements shares the modal component with its canonical sentential negation counterpart but differs from the latter in that it introduces no negation. It is worth noting that fear-min differs from canonical sentential negation not only semantically but also syntactically. Tsiakmakis and Espinal (2022) bring attention to the fact that, in fear-predicate complementation, informationally unmarked subjects cannot intervene between the matrix predicate and min; they always appear post-verbally.

The authors consider this as evidence that expletive min is merged in the structure outside the TP-domain, somewhere in the left periphery of the clause (see also Chatzopoulou 2018; Lekakou 2024; Roussou 2015).

4.2 Greek conditionals

The expletive occurrences of min in Greek fear-predicate complements were shown to (i) potentially co-occur with the complementary negative marker dhen, (ii) fail to license NCIs, (iii) make an epistemic interpretative contribution, and (iv) merge syntactically in the left periphery of the sentence in which they occur. Let us see whether min displays the same properties in the context of a second expletive negation host, namely Greek conditionals.

The English translation provided for (45) already shows why Tsiakmakis et al. (2023) included conditionals in the list of Greek expletive min hosts. The non-negative status of min is further witnessed by the fact that replacing the indefinite kati ‘something’ with the NCI tipota ‘nothing’ in the example above leads to ungrammaticality:

Moreover, conditional min is compatible with dhen, in which case a single negation reading arises:

Interestingly, the non-negative interpretation of min seems to go hand in hand with a high merge position in the case of conditionals, too. The subject of the conditional antecedent cannot precede min, or even intervene between min and the verb; it has to be postposed in the way shown earlier for fear-predicate complementation.

As already mentioned, the presence of expletive min in conditionals is understated and, consequently, understudied. This means that systematic work of any kind concerning its interpretation is lacking. However, its distribution and use are informative enough. Notice that conditional min, as all other negative and non-negative occurrences of min that we have seen so far, occurs in the scope of a non-veridical operator (Giannakidou 1998), in this case the conditional operator. This means that worlds where the positive proposition p is true and worlds where its propositional alternative ¬p is true are relevant for its interpretation (Giannakidou and Mari 2021). The question to ask next is whether these alternative worlds are ordered, in the sense of, e.g., Kratzer (1981, 1991).

I pursue an answer to this question by comparing a conditional with min to its closest min-free counterpart that Greek grammar allows for, which is introduced by the conditional conjunction an ‘if’. I postulate that, while an ‘if’ is the realization of the conditional operator itself, the conditional operator is phonologically null in the case of min-conditionals.

As suggested by the unique English translation given, (49a) and (49b) are interpreted roughly in the same way. However, the variant with min is felicitous only if whatever is described by both the antecedent and the consequent of the conditional has happened before. In fact, (49a) describes something repetitive, a habit or a norm. Example (49b) is unmarked in this respect; the speaker simply considers the possibility of the addressee hearing about some gossip and, consequently, approaching. I propose that the asymmetry above can be explained if, unlike what happens in neutral conditionals featuring an ‘if’, the presence of min in a conditional orders the alternative possible worlds in such a way that p worlds are Best.^[8] The speaker’s bias in this case would stem exactly from the fact that what is described by the min-clause has happened before.

Notice that, if the reasoning above is on the right track, Greek expletive conditional min displays exactly the same properties as expletive fear-predicate min. Under this view, an example like (49a) is best translated in English as ‘In the possible scenario that you hear about some gossip, you will immediately run.’

4.3 Greek polar questions

The last Greek expletive min host identified in the literature includes matrix (51a) and embedded (51b) polar questions (Chatzopoulou 2018; Makri 2013; Tsiakmakis et al. 2023).

Intriguingly, question expletive min behaves in exactly the same way as its equivalents in fear-predicate complementation and conditionals. To begin with, it is unambiguously non-negative, as shown by its inability to license NCIs (52) and its potential co-occurrence with the complementary negative marker dhen (53).

Moreover, these instances of expletive min are also merged in the left periphery of the sentence, as suggested by the fact that they cannot be preceded by informationally unmarked subjects:^[9]

As far as interpretation is concerned, Tsiakmakis et al. (2023) compare matrix questions with min to their min-free counterparts:

The authors present experimental evidence that min-free questions like (55b) are usually attributed to fully ignorant speakers (in the sense of Farkas 2020), who consider both the positive proposition p and its propositional alternative ¬p as equally probable. Questions featuring expletive min, on the other hand, are interpreted as encoding positive speaker bias. In the case of (55a) for instance, the speaker considers the proposition corresponding to Dad ate the meatball soup as more probable than its alternative corresponding to Dad didn’t eat the meatball soup. In light of these results, Tsiakmakis et al. (2023) claim that min in Greek polar questions is interpreted as a biased epistemic modal (see Giannakidou and Mari 2017, 2021). They formally define it as a function interpreted with respect to an ordering source that takes a proposition p as its argument and ensures that Best worlds, that is the worlds that are ranked higher by the relevant ordering source, are p worlds:

Tsiakmakis et al. (2023) do not test the interpretation of embedded polar questions. In the absence of contradicting evidence, I see no reason to doubt that min encodes positive speaker bias when introducing embedded questions too.

A comment is in order before closing off this subsection. Descriptively speaking, Greek polar questions with min feature a negative marker in initial position and systematically convey positive speaker bias. Could these be actual instantiations of the preposed negation question type that Romero and Han (2004) postulate? A positive answer to this question is tempting since it would also explain why no evidence could be found in support of a preposed versus non-preposed negation distinction in polar questions with dhen. However, I am hesitant to endorse this view based on a robust syntactic asymmetry between the two types of questions. Min-questions have already been shown to be embeddable. Prototypical examples of preposed negation questions from languages other than Greek though, such as English, cannot be embedded while preserving negation in initial position (see also Sailor 2013):

Therefore, I take Greek min-questions to exemplify a polar question type different from the one illustrated by (57a) from English.

4.4 On the expletiveness of min

Greek expletive min is licensed by three distinct non-veridical operators: predicates denoting fear, the conditional operator, and the question operator. In all three types of structures, min fails to license NCIs, potentially co-occurs with the complementary negative marker dhen, is merged outside the TP-domain of the sentence, and is interpreted as a biased epistemic modal. In other words, expletive min corresponds to a function that is interpreted with respect to an ordering source; it takes a proposition p as its argument and ensures that p is true in Best worlds.

Comparing the meaning of expletive min to the meaning of negative min, the former is missing the polarity reversal effect but maintains the modal component, that is the consideration and ordering of alternative worlds. In contrast to expletive dhen, the non-negative interpretation of min cannot be attributed to independent peripheral factors kicking in at the utterance level. The expletiveness of min must be pinned onto intrinsic properties of the marker itself. In other words, the min that appears in the Greek expletive negation hosts discussed in the present section is not a negative marker at all; it is a biased epistemic modal.

5.1 Emotive doxastic predicates

Greek emotive doxastic predicates and, specifically, predicates denoting fear were shown to license expletive occurrences of neg₂, interpreted as epistemic modals that rank p worlds over ¬p worlds. Can this view be extended to similar constructions in other languages (see Espinal 1992, 2000; Horn 2010; Jin and Koenig 2019, 2020, among others)?

Tahar (2021, 2022) studies expletive negative markers in French fear-predicate complementation:

Based on a rich diachronic dataset, the author concludes that the expletive negative marker ne in such examples is primarily related to the expression of dispreference (Tahar 2022). In (59), the speaker does not want the addressee to get stung. In other words, French fear-predicate ne is interpreted as a deontic modal, flagging p worlds as more preferred than ¬p worlds. This conclusion can be regarded as some initial evidence that the neg₂ status of the fear-predicate expletive marker extends beyond Greek:

Intriguingly, the syntactically high merge position postulated for fear-predicate neg₂ based on evidence from Greek also finds crosslinguistic support. In Hebrew, for example, the expletive negative marker emerging in the complement of a predicate denoting fear attaches to the complementizer:

Similarly to what was shown for Greek, the canonical negative marker – the independent lo in (61) – appears lower in the structure than its expletive counterpart.

5.2 Negative predicates

In languages such as Old French, Arabic, and Mandarin, negative predicates like forbid also license expletive negative markers (Horn 2010; Jin and Koenig 2019; Makri 2013; Tahar 2022, among others).

Despite the presence of the French negative marker ne in the complement of the forbid-predicate in Example (62), the subject stops the humiliation of another person, he does not cause it. In other words, ne in (62) is not interpreted as a polarity reversal operator; it is expletive.

Makri (2013) suggests that negative predicates licensing expletive negative markers are similar to emotive doxastic licensors, in the sense that both groups of predicates share a doxastic semantic component: the speaker considers both p and ¬p worlds. If this is on the right track, French negative predicates such as défendre ‘forbid’ in (62) probably feature non-negative instances of neg₂, too.

This speculative claim is independently supported by Tahar (2022). The author’s diachronic study leads her to conclude that expletive occurrences of ne in the complement of both forbid predicates and fear predicates can be associated with the expression of speaker’s dispreference. Put differently, in both environments ne does not bear negative truth-conditional semantics but makes a deontically modal interpretative contribution.

While not available in Modern Greek, the expletive negation host described as negative predicates is relevant for Classical Greek (Makri 2013). Importantly, the negative marker emerging in this case is me:, which is diachronically linked to Modern Greek min (Chatzopoulou 2018), that is neg₂.

Example (64), then, can be considered as additional evidence in support of associating negative predicates with expletive occurrences of neg₂.

5.3 Dubitative predicates

The next expletive negation host, namely dubitative predicates, is readily grouped together with the previous two.

In the Spanish example above, the negative marker no that appears in the complement of the verb dudar ‘doubt’ is interpreted non-negatively; ignoring the potentially rhetorical flavor of the question, the speaker asks if anyone doubts the falsehood of the stories told, not their veracity. Based on such data, Spanish dubitative predicates have been identified as expletive negation hosts (Espinal 2000; Makri 2013).

I know of no studies specifically focusing on the interpretative import of Spanish expletive no in the complement of dudar. However, like emotive doxastics and negative predicates (Anand and Hacquard 2013; Makri 2013), dubitative predicates involve a doxastic semantic component (Uegaki 2023). This means that they are compatible with the speaker considering both p and ¬p worlds and even having an epistemic preference. Capitalizing on this, expletive negative markers emerging in the complement of Spanish dubitative predicates are likely non-negative instances of neg₂:

Once again, Classical Greek independently points towards a similar generalization. The dubitative predicate apisto: ‘doubt’ in (67) below licenses expletive instances of the negative marker me:. Notice that the complementary negative marker ou(k), that arguably had neg₁ status (Chatzopoulou 2018), follows me: and is interpreted as canonical sentential negation.

5.4 Temporal expressions

Languages like Catalan, Italian, German, French, and Mandarin allow for expletive negative markers in the scope of temporal expressions such as before or until (Cépeda 2018; Espinal 1992; Jin and Koenig 2019, 2020; Makri 2013; Tahar 2022; Tovena 1996).

The translation provided for (68) from Italian suggests that the negative marker non is interpreted non-negatively in this case, justifying the designation of Italian finché as an expletive negation host. In fact, (68) seems in principle interchangeable with the minimally different (68′):

Crucially, Tovena (1996) argues that there is actually a difference between the variant with non and the one without, which is obscured by the fact that arrivare ‘arrive’ denotes an achievement. If the verb of the temporal sentence is replaced with parlare ‘talk’ for example, which can be perceived as an activity, then the difference is revealed:

Notice that the finché-clause is translated as a positive until-clause in (69a), but as a positive while-clause in (69b). Importantly, both readings can be derived compositionally (see Tovena 1996). In the absence of negation (69b), Italian finché demands that my staying and your speaking finish at the same time. If the temporal clause is negated though (69a), finché conveys that the ending point of my staying coincides with the ending point of your non-speaking, that is the starting point of your speaking.

In light of the empirical observation above, Tovena (1996) proposes that the negative marker non emerging in the scope of finché is a true negative marker bringing about the expected reversal. If the verb of the finché-sentence denotes an achievement, which starts and ends at the same point in time, then the interpretative contribution of non is imperceivable, making non expletive on the surface. Mind, however, that this expletiveness is an epiphenomenon stemming from the lexical aspect of the finché-verb.

Tovena’s (1996) analysis of finché-non is strikingly reminiscent of the conclusions reached concerning expletive instances of the Greek negative marker dhen: a canonical negative marker whose negative semantics is obscured by independent factors. Consequently, temporal expressions in Italian, and possibly other languages manifesting similar constructions (see Abels 2002 on Russian poka ‘until’; Cépeda 2018 on Spanish hasta ‘until’ and German bevor ‘before’), are likely to feature expletive occurrences of neg₁.

5.5 Negative adverbials

Adverbials encoding some negative meaning such as without or almost license expletive negative markers, too (Horn 2010; Jin and Koenig 2020; Makri 2013, among others).

In the French example above, the negative marker ne is interpreted as expletive since its co-occurrence with sans ‘without’ does not bring about a double negation reading. The question of immediate interest to the present study is whether ne is more similar to an expletive instance of neg₁ or neg₂. Tahar (2022) proposes that French without-ne is similar to its fear- and negative predicate-counterparts in conveying some kind of speaker dispreference. If this is correct, then without-ne could be captured as an occurrence of expletive neg₂.

I am hesitant to endorse Tahar’s (2022) claim in this case for two main reasons. The first one is empirical and comes from Jamaican English:

The Jamaican English word for without includes the non-negatively interpreted negative marker -n as a bound morpheme. In other words, Jamaican without is also an expletive negation host. Importantly, though, the presence of -n cannot be associated with speaker dispreference; the singer would actually prefer that “they” were tried before found guilty.

Example (72) suggests that without expletive negation licensors cannot be cross-linguistically linked to the expression of dispreference. Looking closer, I believe that they are not to be associated with neg₂ altogether. Note that the expletive neg₂ hosts identified so far share a modal semantic component. Assuming that without-words also involve a modal component seems far-fetched in the absence of solid supporting evidence. In light of the above, I propose that the expletive negation host dubbed as negative adverbials in languages such as French and Jamaican English involves negative markers of the neg₁ type.

I take the extra step to speculatively suggest that the canonical status of the negative marker emerging in the scope of the negative adverbials we saw above is obscured by a Negative Concord relation formed between the licensor and the licensee (see Espinal 2000, for a first attempt to relate expletive negation to Negative Concord, and Espinal et al. 2023, for a fresh analysis of the Negative Concord phenomenon).^[11] While further research is needed to evaluate the validity or strength of this specific proposal, the link between negative adverbials and neg₁-type expletive markers finds independent support in other studies. Fortuin (2023), for example, argues that the expletive negative markers that emerge in the scope of Dutch without-constructions do contribute negative semantics but only at the level of presuppositions. Moreover, Olguín Martínez (2024) shows that without-expletive negative markers in a language such as Huasteca Nahuatl may encode mirativity. See Tsiakmakis and Espinal (2022) for a proposal on how to derive mirativity from a canonical negative marker merged outside the TP-domain of the clause.

5.6 Conditionals

Greek conditionals were earlier shown to allow for expletive instantiations of neg₂. Recall that expletive min occurs in the absence of any (overt) negative licensor in this case. Example (45) is repeated below for ease of reference.

Crucially, Greek is the only language reported so far to license expletive negative markers in the positive antecedent of conditionals (Tsiakmakis et al. 2023). In the rest of languages for which conditionals have been identified as an expletive negation host, the presence of unless is obligatory (Greco 2019; Jin and Koenig 2019, 2020; Makri 2013; Tahar 2022, among others).

Italian non in (75) above, for instance, is interpreted non-negatively in the scope of the conditional conjunction a meno che ‘unless’; the co-occurrence of these two elements does not lead to a double negation reading. Replacing a meno che with se ‘if’, however, forces non to be interpreted as standard negation:

If, in a given language, negative conditionals host expletive negative markers but positive ones do not, then the licensing element is the negative component of unless , not the conditional component. Under this perspective, unless -words are better viewed as a subcategory of negative adverbials that, as was shown in the previous subsection, appear to host instances of neg₁. Importantly, the status of conditionals as an expletive negation host then depends on future research: Do positive conditionals license expletive negative markers in languages other than Greek?

5.7 Comparatives

Comparative constructions are one of the most studied expletive negation hosts (Belletti 2001; Delfitto 2018; Espinal 1992; Giannakidou and Yoon 2008; Labelle 2023; Makri 2018; Napoli and Nespor 1976, among others). Let us take a look at the following example from Italian.

The Italian negative marker non appears to be interpreted non-negatively in (77). In fact, the minimally different (77′) is considered to mean the same.

Napoli and Nespor (1976) claim that non in (77) has an epistemic interpretative import: what the speaker says contradicts somebody’s previous belief. This would suggest that comparatives license expletive instances of neg₂.

Crucially, Makri (2018) argues convincingly that the interpretative effect highlighted by Napoli and Nespor (1976) stems from subjunctive mood, not from the presence of expletive non. Moreover, she supports empirically and theoretically the claim that expletive negative markers in comparatives are the optional spell-out of the negation incorporated in comparative operators (Seuren 1984), independently motivated by Delfitto (2018). This view is based on the intuitive observation that the interpretation of both (77) and (77′) can be represented as in (78):

If Delfitto (2018) and Makri (2018) are on the right track, Italian comparatives are more likely to feature instances of neg₁ (see also Labelle 2023). The truly negative semantics of neg₁ are in this case obscured by the fact that the latter appears to be spelled out separately from the comparative operator (see Espinal et al. 2023 for independent evidence for the existence of ‘disembodied’ negative markers), possibly in order to mark the exact scope of negation.

Additional evidence in favor of associating comparatives with expletive neg₁ markers can be sought in Classical Greek, where comparatives licensed expletive instances of ou(k) ‘not’, that is neg₁ (Chatzopoulou 2018).

5.8 Polar questions

Polar questions were shown to be one of the most puzzling expletive negation hosts in Greek since they license instances of both neg₁, dhen, and neg₂, min. Examples (31b) and (51a) are repeated below for reference.

Both types of questions are found also in other languages. Spanish preposed negation questions (83) are reminiscent of dhen-questions in Greek, while Latin ne-questions (84) are similar to Greek min-questions.

Notice that in (84) the expletive negative marker -ne co-occurs with the canonical negative marker non, just like we saw earlier for Greek.

Based on the above, polar questions in different languages appear to host expletive instances of either neg₁, potentially coinciding with some bias, or neg₂, which triggers a bias interpretative effect itself.

5.9 Rhetorical questions

Greek negative rhetorical questions were shown to feature canonical instances of neg₁, whose negative semantics is obscured by the independent rhetorical effect. The main arguments were that (i) not all negative rhetorical questions convey a proposition of positive polarity, and (ii) the polarity reversal effect is present also in positive rhetorical questions, which obviously lack (expletive) negative markers.

Importantly, the arguments above were anticipated by Rohde (2006) – see also Caponigro and Sprouse (2007) – based on data from languages other than Greek, such as English. Let us take a look at the following examples:

Starting from the former, (86b) under its rhetorical reading in this specific context is roughly equivalent to “I haven’t hurt your feelings, so it’s me you should trust”. The conveyed proposition in this case is of the same polarity as the proposition expressed by the question, suggesting that negative markers in rhetorical questions are not always interpreted as expletive. Moving on to Example (87), the rhetorical question in (87b) corresponds roughly to the assertion “I haven’t hurt your feelings”. In other words, it triggers a polarity reversal in the absence of any negative marker, showing that there is no systematic link between these two.

Taking into account the above, as well as the fact that no language has been reported to have only negative rhetorical questions to my knowledge, rhetorical questions across languages are likely to host instances of neg₁, whose non-negative interpretation is only an epiphenomenon due to rhetoricity.

5.10 Exclamatives

Negative exclamative hosts of expletive negation in Greek were earlier argued to have a negative rhetorical question structural basis, based on two fundamental differences between them and their positive counterparts: (i) negative exclamatives do not admit the exclamative complementizer pu ‘that’, and (ii) negative exclamatives trigger an extreme quantity-instead of an extreme degree-reading. Strikingly, the asymmetry in (ii) is reported also for other languages, like Hebrew (Eilam 2007) and Italian (Delfitto and Fiorin 2014).

If negative exclamatives behave similarly to their Greek counterparts in these languages, there is good reason to assume that the expletive negative markers they involve correspond to instances of neg₁, whose negative semantics is, however, masked by rhetoricity, exactly like we saw for rhetorical questions.

5.11 Free relatives

The last expletive negation host identified in the literature includes free relatives and, to my knowledge, it has been reported only for Hebrew (Eilam 2007). Let us compare the following pair:

Example (92b) differs minimally from (92a) in further including the negative marker lo. Crucially, though, lo is interpreted non-negatively in this case, as suggested by the English translation given. Instead, it reportedly makes a modal interpretative contribution that can be described as ignorance or indifference, best captured in English by means of the bound morpheme -ever (Eilam 2007).

The modal effect would in principle suggest that Hebrew free relatives license expletive instances of neg₂. Taking such an analytical step, however, requires further explanation in this case. Up to this point, expletive occurrences of neg₂ have been related to polar propositional alternatives p and ¬p and a ranking of possible worlds such that p worlds are ranked higher than ¬p worlds. Notice, though, that in free relatives the alternative worlds differ not with respect to the polarity of p but with respect to the value assigned to the referent of the wh-word in each case; in this sense, we have {p ₁ , p ₂ , p ₃ … p _n }. Moreover, these worlds are not ranked in any way, as suggested by the ignorance/indifference effects reported by Eilam (2007). In light of the above, I side with Eilam (2007) that free-relative expletive lo is a separate lexical element defined similarly to von Fintel’s (2000) English -ever. I differ from him in speculatively proposing that free-relative lo is derived from neg₂ uses of the standard negative marker lo, with which it shares a modal component. More research is needed to test the validity of this hypothesis. Importantly, however, Hebrew free relatives cannot be established as a canonical expletive negation host under this view.

5.12 Taking stock

The study of individual examples of expletive negation hosts showed that the expletive neg₁ versus expletive neg₂ asymmetry motivated on the basis of Greek data can be extended to other languages. In other words, it appears that one can distinguish two classes of expletive negation licensors even in languages that do not display morphological evidence in support of this distinction.

On the one hand, we have (i) temporal expressions, (ii) negative adverbials, (iii) comparatives, (iv) optionally biased polar questions, (v) rhetorical questions, and (vi) exclamatives. In different languages, these licensors have been argued to feature canonical instances of neg₁ markers, which are interpreted as a function that takes a proposition p as its argument and returns the complementary proposition ¬p. This is represented formally in (93).

The truly negative semantics of the neg₁ occurrences in the environments listed above is masked by peripheral factors, such as the interaction of negation with other negative operators, with verbal aspect, or even with rhetorical strategies factored in at the level of utterance interpretation. Notice that (i) there is nothing special about the negative markers emerging in this group of hosts, and (ii) there is no characteristic property defining all and only these specific hosts.

On the other hand, we have (i) emotive doxastic predicates, (ii) negative predicates, (iii) dubitative predicates, (iv) biased questions, and possibly (v) conditionals and (vi) free relatives. In the languages examined in this section, these expletive negation hosts feature expletive instances of neg₂, which only retain the modal component of their truly negative counterpart. Concretely, expletive negative markers occurring in these environments are interpreted with respect to an ordering source; they take a proposition p as their argument and ensure that p is true in all the possible worlds that are ranked higher by this ordering source. Formally:

This second class of hosts can be designated as the class of expletive negation licensors proper since (i) it features intrinsically non-negative instantiations of neg₂ markers, and (ii) it involves licensors characterized by their compatibility with a modal component; concretely, an epistemic or deontic ordering source (see Kratzer 1981, 1991, among others). In fact, proper expletive negation can be considered as the spell-out of this ordering source in each case (see also Mari and Tahar 2020; Tsiakmakis et al. 2023).

I would like to submit the conclusions of the crosslinguistic excursus that made up this section as some first but promising evidence in support of pursuing the following broad research hypothesis: The bi-partite distinction between dhen-expletive negation hosts and min-expletive negation hosts may be cross-linguistically valid. Under this hypothesis, I propose that the expletive negation host inventory across languages can be broken down to what could be dubbed as Apparent expletive negation hosts and what I have already dubbed as Expletive negation hosts proper:

I acknowledge that the research hypothesis introduced above is inspired in Greek and has been motivated mostly on the basis of data from a very small set of Indo-European languages – the ones in which the phenomenon of expletive negation has been studied the most. Therefore, this hypothesis is to be considered merely as a research agenda, an invitation to study what is broadly understood as expletive negation phenomena under this novel perspective. It would be interesting to see, for example, whether the distinction above is reflected in the morphosyntax of genetically unrelated languages, what the limits of the class of Apparent expletive negation hosts are and whether there is mobility between the two classes.