Roussou (2024) draws attention to a peculiar construction exemplified by the Greek sentences in (1) and (2) below:

The peculiarity of these two sentences and the grammatical construction they represent consists in their four characteristic properties: (i) the licensing of wh-words in the context of a non-interrogative sentence type, (ii) the presence of the conjunction ce ‘and’ in the absence of a second (syntactic) conjunct, (iii) the presence of the sentential negation marker ðen interpreted –at first sight at least– non-negatively, and (iv) the universal quantification reading in the apparent absence of a universal quantifier. Picking up the thread from Roussou (2024), this paper seeks to uncover the syntactic structure of the construction under study –let us dub it the and-what-not-construction– and explore how its relatively unexpected interpretation is derived.1

The paper is structured as follows: Section 2 gives the detailed syntactic and interpretative profile of and-what-not-sentences. Section 3 defines the and-what-not-construction negatively, showing that, despite appearances, it cannot be treated as a subcategory of wh-questions, rhetorical questions, exclamatives or free-choice constructions. Section 4 provides the syntactic analysis of and-what-not-sentences as declaratives involving a fronted wh-phrase. Section 5 shows how the universal quantifier reading is derived based on the wh-phrase being an indefinite, ce being an additive morpheme, and negation licensed via modality. Finally, Section 6 considers how the and-what-not structure fits in the more general typological pattern of universal quantifiers, while Section 7 concludes the paper.

Before we try to determine the exact grammatical status of Greek and-what-not-sentences, let us take a deeper look at what they look like. The first property highlighted regarding the phenomenon is the involvement of a wh-word. Intriguingly, not all wh-words make good candidates for the and-what-not-construction. Wh-words denoting degree variables for example, such as poso ‘how much’ or posa ‘how many’, yield ungrammaticality in and-what-not-sentences. Observe the contrast between (3), on the one hand, and (4) and (5), on the other:

Furthermore, the wh-phrase must be bare and cannot have an NP-restrictor. In other words, D-linked wh-phrases are at best deviant. This is illustrated in (6) and (7) below, modifying (1) and (2) respectively:

Moving on to what was listed in Section 1 as the second characteristic property of the Greek and-what-not-construction, namely the presence of the coordinator ce ‘and’ and the lack of an overt second conjunct, two complementary statements are in order. The first one is that ce cannot be dropped without compromising grammaticality:2

The second is that the absence of an overtly expressed second conjunct is not simply possible but obligatory. In example (9) below, where ce ‘and’ is meant to coordinate two answers to two distinct questions, the and-what-not-interpretation of the second answer is not available.

Regarding the presence of the negative marker ðen in and-what-not-sentences, it should be highlighted that this is also obligatory:

At the same time, ðen does not seem to behave like a sentential negative operator given that it excludes the licensing of both Negative Polarity Items (NPIs) such as kanenas ‘anyone’ and Negative Concord Items (NCIs) such as kanenas ‘no one’ (see Tsimpli & Roussou 1996; Giannakidou 1998) This is shown in (11).

Failure of polarity item licensing seems to indicate that the negative marker lacks the required negative import in this case.

Moreover, ðen is the only sentential negation marker that is allowed in the and-what-not-construction; the complementary negator min (Holton et al. 2012) found in modalized sentences is out:

It is worth noting that min does have non-negatively interpreted uses (see Makri 2013; Roussou 2015; Chatzopoulou 2018; Tsiakmakis et al. 2023) so this cannot be the reason it is excluded from and-what-not-sentences.3 Anticipating the discussion in Section 3, given that min is excluded from root declarative sentences, the ungrammaticality of (12) is an argument for treating and-what-not-sentences as declaratives.

The last characteristic property of the and-what-not-construction anticipated in the introductory section is the universal quantification reading with an unclear source, reflected in the English translations of all the examples used so far. Notice that this reading arises regardless of whether the wh-phrase is specified for singular or plural number:4

In light of the above, the grammatical profile of and-what-not-sentences can now be made specific:

With this in place, we can now determine the place of and-what-not-sentences in Greek grammar specifically, and natural language grammar more broadly.

Although understudied in itself, the and-what-not-construction shares major properties with other well-known types of sentences exhibited by Greek or other natural languages. In this section we show that and-what-not-sentences cannot be subsumed under any of these seemingly affinitive sentence types.

After the extensive negative definition of and-what-not-sentences in the previous section, let us now try to define this peculiar construction positively. Having discarded all other alternatives, we revert to the last alternative, namely that and-what-not-sentences are declaratives and, specifically, root declaratives; recall that they cannot be embedded. But how exactly are they syntactically derived? Beyond being a label facilitating reference, and-what-not also summarizes the syntactic identity of the grammatical construction under study here, which consists in the combination of ce ‘and’, a wh-word, and the negative marker ðen ‘not’, all in the same sentence.

With a fully-fledged proposal regarding the syntactic derivation of and-what-not-sentences in place, let us address their meaning composition. We stick to our reference example about Fanis’ cooking, repeated below as (58), and focus again on the constitutive parts of the specific construction under study: the wh-phrase, the coordinator ce ‘and’ and the negator ðen.

Following Roussou and Vlachos (2024), we assume that wh-words in Greek are indefinite pronouns (cf. also the typology of Haspelmath 2001) which introduce a variable. Consequently, the denotation of the wh-word ti ‘what’ in (58) can be represented roughly as in (59):

Recall that the wh-phrase in and-what-not-sentences is focused. Adopting the main insight in Delfitto and Fiorin (2014) regarding wh-phrases in interrogatives and exclamatives going back to Han (2002), we take the focus value of a focused wh-phrase to be a structured Boolean algebra of alternatives. For ease of exposition, we assume that in the context of our example (58) Fanis could have only cooked pizza, pasta and shepherd’s pie. In this case, the focus value of focused ti ‘what’ can be represented as in the algebra below:

Consider next how the presence of ce ‘and’ affects the above picture. Chatzikyriakidis et al. (2015) analyze Greek additive ce as a Focus Association Operator that interacts with the focus value of its associate in the following way:11 it introduces a presupposition that at least one focus alternative distinct from the one denoted by the associate also makes the corresponding proposition true. Consequently, the interpretation of the complex ce ti ‘and what’ from our example can be represented as follows:

Once the ce ti wh-phrase –merged as an internal argument– composes with the verb majirepse ‘cooked’ and the external argument Fanis, the interpretation of the derived structure amounts to the following:

Importantly, at this point in the derivation, the focus value of the derived structure can be represented also as in the structured Boolean algebra of alternative propositions in (60ˈ) below (see also Delfitto & Fiorin 2014):

Note that the alternative propositions in the algebra above are ordered by entailment. Consequently, and following standard assumptions (see Han 2002; Delfitto & Fiorin 2014), informativity increases and probability decreases, as one moves upwards in the algebra, with the top element being the one with lowest probability to make the corresponding proposition true.

It is finally time to move to the role of the negative marker ðen. We have anticipated that ðen is construed as a polarity term which yields a negative reading when the negative operator is imported at the point of interpretation or not, if modality for example is at play. We saw that in and-what-not-sentences ðen does not yield a sentential negation reading, since it does not reverse the truth value of the proposition. This leaves us with the second option, namely the non-canonical (expletive) one where the interpretation that arises relates to modality. Interestingly, according to Veloudis (2023), “utterances [like those we dub as and-what-not-sentences] 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.” Also enlightening is Veloudis’ (2023) association of such utterances with “scale reversal”.13 We depart from Veloudis (2023) in distinguishing and-what-not-sentences from negative rhetorical questions (see Section 3), but side with him in reserving a scale-reversal interpretation for the negator ðen in this construction. Note that the polarity property of ðen is consistent with the notion of ‘reversal’. The latter is not restricted to propositional content but can hold for ordering as is the case with alternatives, inducing a modal reading. We formally capture Veloudis’ insight by extending the spirit of Delfitto & Fiorin’s (2014) analysis of non-canonically interpreted negators in Romance and Germanic rhetorical questions and exclamatives to Greek and-what-not-sentences. Concretely, we propose that ðen does not operate on the propositional meaning of the declarative and-what-not-sentence but on the focus value of the structure featuring the focused wh-phrase (see (60ˈ)). Instead of reversing the truth conditions under which the sentence is true, as would be the case of a canonically interpreted sentential negator, ðen reverses the probability scale in the algebra of focus alternatives in such a way that the top element of the algebra becomes now the least informative and, thus, most probable.14 This interpretative import associated with ðen can be considered as a non-at-issue modal statement, in line with Delfitto (2020) and Veloudis (2023).

With all the bits in place, our analysis predicts the complete interpretation of our reference example (58) to be as follows:

Unpacking the notation above, the and-what-not-sentence in (58) is predicted to (i) assert the existence of something that Fanis cooked, (ii) carry the presupposition that Fanis cooked more than one thing, and (iii) convey in a non-at-issue manner that it is more probable that Fanis cooked more things than less.

Considering the above, our account derives the interpretation of Greek and-what-not-sentences. Importantly, however, it predicts that non-uniqueness is part of the semantic core of the and-what-not-construction, while the universal quantification arises as part of an additional, non-at-issue modal statement. The gradually increasing felicity of the follow-ups in the following three examples can be considered as evidence that this aspect of the proposed account is on the right track:

Taking stock, the analysis of the and-what-not-construction proposed here correctly derives its interpretation as a result of the interaction of its constitutive parts. Importantly, it is not forced to make any unmotivated assumptions regarding the meaning contribution of each of these parts. Greek wh-words have been independently motivated to be interpreted as indefinites (Roussou & Vlachos 2024), additive uses of the coordinator ce ‘and’ have independently been linked to a non-uniqueness presupposition and non-canonically interpreted negative markers have been independently related to modal interpretative import both in Greek (Makri 2013; Tsiakmakis et al. 2023) and in other languages (Yoon 2011; Delfitto & Fiorin 2014; Choi & Lee 2017; Tsiakmakis & Espinal 2022, among many others).

Before closing off this section, we need to briefly address two other properties of and-what-not-sentences that we saw in Section 2 in the context of the above account. Specifically, wh-phrases with an NP restrictor, that is D-linked phrases in the sense of Pesetsky (1987), and degree wh-phrases are excluded from and-what-not-structures. Our analysis predicts that D-linked and degree wh-phrases may be allowed to enter and-what-not-sentences in syntax but are filtered out at the level of interpretation because their focus alternatives are not organized in a structured Boolean algebra with inherent probability on which ðen can operate. Concretely, the alternative focus values of a degree wh-phrase are ordered on a continuous scale and the probability of the focus alternatives of a D-linked phrase heavily relies on the context. Both conditions would render the application of the scale-reversal effect of ðen problematic.

The universal quantification-like reading of and-what-not-sentences was derived earlier as the result of the combination of three elements, namely ce ‘and’ as an additive particle, the wh-phrase (an indefinite) and the negator ðen ‘not’. In this section we focus on the combination of the first two elements, since the participation of conjunction in universal quantification constructions is a well-documented typological tendency (Bhat 2000; Haspelmath 2001; Jayaseelan 2001; Gil 2013; Szabolsci 2013; 2015; among others). In his typological review, Gil (2013) adopts a wide definition of ‘conjunctions’ as “taken to include not only forms with meanings similar to that of and, but in addition expressions that are sometimes characterized as conjunctive operators or focus particles, with meanings resembling those of also, even, another, again, and in addition the restrictive only.” Similarly, for ‘universal quantifiers’, “these are assumed to encompass not only forms with meanings such as those of every, each and all, but also expressions that are sometimes referred to as free-choice, with meanings corresponding to that of any”. The and-what-not-construction seems to fall under this pattern, given that ce ‘and’ has the additive ‘also’ meaning and the reading that arises falls within what Gil (2013) includes in the universal quantifier set.

The second important typological generalization refers to the role of the wh-phrase which combines with the conjunction (of any of the sorts above). In languages where such combinations are available, the wh-phrase is taken to be an ‘indeterminate pronoun’ in the sense of Nishigauchi (1990), that is an indefinite pronoun subject to (selective or unselective) binding. In Japanese for example, the particle -mo ‘and/also’ can attach to the indeterminate pronoun and yield different readings which fall under universal quantification in Gil’s (2013) sense. Consider the examples below, cited in Gil & Tsoulas (2013: ex. (2), ex. (10a,b)):

In all cases the item mo is involved but the reading changes depending on the attachment site of mo in combination with negation and/or modality, ranging from a negative quantifier (no-x) to a polarity item (any-x) to a free choice item (any-x).

With respect to Malayalam, Jayaseelan (2001: 66) observes that “the adding of the conjunction marker to a question word yields a universal quantifier”, e.g., aar-um (who+conj) = ‘any’. This reading is also polarity sensitive and falls under Gil’s (2013) notion of universal quantifiers. Jayaseelan (2001: 86, ex. (64)) provides the following statement:

Given the above, the data and the analysis of and-what-not-structures in Greek seems to obey a more general typological pattern. Interestingly, the combinations of wh+conj in the languages under consideration are polarity sensitive (see Gil & Tsoulas 2013; and for a semantics approach Szabolsci 2015). Unlike Japanese and Malayalam, the Greek ce+what combination exhibits a quite restricted distribution, as it arises only in the presence of the negator ðen in a declarative; when ce takes a wh-complement in the absence of negation, the universal quantifier reading does not arise:

As discussed in Section 5, negation in and-what-not sentences contributes towards a modal reading computed outside syntax. Tentatively we could say that this accounts for the fact that other types of modality markers in Greek morphosyntax (e.g., modal verbs or adverbs) do not license this reading. Note that no equivalent of the Japanese (67b) is available in Greek.

Space limitations do not allow us to offer a more thorough typological study at this point or derive theoretically the similarities and differences between the Greek and-what-not-construction and its near-parallels sampled above. Before we conclude this section, we would like to add to this cross-linguistic excursus a structure from Russian which is intriguingly similar to Greek and-what-not-sentences.15 Take a look at example (70) below:

The sentence in (70) includes a wh-word, namely kogo ‘whom’, an instance of the negative marker ne ‘not’ that, at least on the surface, does not contribute sentential negation, and tol’ko ‘only’, which can be considered a conjunction in Gil’s (2013) terms. Moreover, the whole example is interpreted as involving universal quantification in the apparent absence of a typical universal quantifier. These properties bring the Russian construction –let us dub it as the “what-only-not-construction”– strikingly close to the Greek and-what-not-construction.

Other than the surface similarities mentioned above, there are two more empirical reasons to posit a significant parallel between Russian what-only-not-sentences and Greek and-what-not-sentences. First, like its Greek counterpart, the Russian what-only-not-construction is incompatible with wh-words such as kak sil’no or skol’ko ‘how much’:

Second, Russian what-only-not-sentences seem to fall into the declarative sentence type too, as they can be used also for explicit (and exhaustive) listing:

We do not have enough evidence to commit to any strong theoretical claim regarding the syntax and interpretation of Russian what-only-not-sentences at this point. We simply mention in passing that, if one were to extend our analysis of Greek and-what-not-sentences to Russian, they would have to dispense with the non-uniqueness presupposition –which in the Greek case is contributed by ce ‘and’– and possibly take tol’ko ‘only’ to contribute an exhaustivity operator operating on the focus alternatives of the wh-word. At this point we acknowledge the potential similarities and leave the discussion of the Russian data open for future research.

The present paper focused on a peculiar construction exhibited by Greek with the following characteristic properties: (i) the licensing of fronted wh-words that denote variables in the absence of an Inter operator, (ii) the obligatory presence of the coordinator ce ‘and’ in the absence of an overtly realized second conjunct, (iii) the obligatory presence of the Greek negative marker ðen in the absence of sentential negation semantics, and (iv) an interpretation involving universal quantification. After disproving several plausible alternative analytical hypotheses, we concluded that the best way to account for what we dubbed as the Greek and-what-not-sentences is to treat the latter as declaratives featuring a fronted wh-phrase. Regarding their interpretation, Greek and-what-not-sentences are existential statements enriched with a non-uniqueness presupposition and a modal statement giving rise to a universal quantification-like inference. Our analysis derived the syntactic and interpretative behavior of the and-what-not-construction as a function of the interaction of its constitutive parts, while ascribing to the negator ðen in and-what-not-sentences a modal interpretation independently available to other non-canonically interpreted negative markers in Greek and other languages. Finally, we showed how the Greek data fit in the typological picture of universal quantifiers formed by conjunction and indefinite pronouns. In this respect the present discussion enriches the empirical typological base while maintaining parametric variation. It also contributes to a better understanding of the mapping between morphosyntax and semantics regarding what looks like sentential negation.