1. The apparently expletive pas

To my knowledge, the existence in Quebec French of examples like (1) was first discussed by La Follette & Valin (1961). As shown by the translation, the negative morpheme pas in the embedded clause apparently carries no negative force. I will gloss this apparently expletive pas as ‘neg_x’ and refer to it as pas_x, the ‘x’ subscript being a convenient label to identify the item whose status is on the table.

Kemp (1982) and Gonzalez & Royer (2022) consider that, in (1), pas_x is expletive. However, the context is not one in which expletive negation is generally found, as we will see in section 2. There are therefore two possibilities: either pas_x is a type of expletive negation having properties distinct from the known cases of expletive negation, or it is not expletive. While Gonzalez & Royer (2022), to be discussed below, opt for the first possibility, the aim of the present paper is to argue that pas_x is not expletive.

A number of recent analyses treat the negative morpheme found in expletive negation contexts as being, not an expletive item, but a negative operator playing a role at a level distinct from that of the standard morpheme of predicate negation; this level could be the level of implicated meanings (Delfitto 2020) or a position within the left periphery of the clause (Greco 2020; Tsiakmakis & Espinal 2022). Instead of following these approaches, I will argue that pas_x in (1) is the standard morpheme of predicate negation, operating in syntax at the TP level. The translation in (1) is an approximate paraphrase of the meaning of the sentence, but it does not reflect its underlying structure and its full meaning. In (1), the embedded clause is negative, and the implicit complement of the verb is a comparative: there is no <more beautiful wheat> on the market. Pas_x is, simply, the negative operator pas scoping over a comparative, and there is no need to consider distinct levels of syntax or of meaning to account for it.

Pas_x is found exclusively in the complement of a superlative (1)-(2), in the restrictor of a universal quantifier (3), and in the constructions in (4) whose relevant properties, I will argue, make them behave, from the point of view of accounting for pas_x, like superlative constructions in the case of the ordinal antecedent in (4a) and the partitive antecedent in (4b), and like universals in the case of the free relative in (4c).

I will defend the view that, in every case, pas_x is a negative operator scoping over an implicit, hence silent, comparative.¹

Kemp (1982), to be discussed below, observed that the presence of pas_x is domain widening, in the sense that, when pas_x is present in the complement of a superlative or of a universal quantifier, the comparison class or the restrictor set is enlarged. I will show how this interpretative effect follows from the analysis.

The main message that I would like to put across in this paper, is that an item that looks expletive at first sight may turn out to be, upon careful scrutiny, not expletive at all, that is, non-distinct, syntactically and semantically, from the clearly non-expletive use of the same item, once silent categories are taken into account.

The general structure of the paper is as follows. After a critical review of the literature, I will turn to the analysis of the superlative construction; then I will focus on the universal one.

2. Pas_x is not a standard expletive negation

The superlative and universal contexts in which pas_x is observed are not the typical expletive negation contexts. In general, expletive negation is found in complements of non-veridical predicates expressing notions like fear, doubt, denial, or in clauses introduced by words like before, unless, without (e.g. Espinal 1992; Espinal 2000; Espinal 2007; Delfitto 2020). For example, in (5), from Catalan, no within the complement of a fear predicate does not reverse the polarity of the embedded clause. In standard French, expletive ne appears in this context (6a), provided it is not accompanied by the predicate negator pas (6b). In Quebec French, (6c), pas has the same negative interpretation as in (6b): the negation is polarity reversing, and not expletive. The same could be shown with a before clause.

The third type of context in which expletive negation is typically found is the complement of comparatives of inequality. Expletive negation is, however, not licensed in the complement of superlatives. That is illustrated in (7), from standard French; Catalan expletive no has the same distribution (M.T. Espinal, p.c.). The contexts in which pas_x is licit in Quebec French are exactly the opposite. Pas_x is found in the complement of superlatives, but not in the complement of a comparative of inequality (8). (8b is Gonzalez & Royer’s 2022 example 17; the left-dislocated subject makes it more colloquial, but it has no bearing on the presence of pas_x.)

Thus, the distribution of pas_x is quite different from that of standard expletive negation. If pas_x is expletive, it requires a specific analysis.

3. Kemp (1982) and domain widening

Kemp (1982) observed that when pas_x is present in a clause, the assertion is strengthened. Both (9a) and (9b) are grammatical in Quebec French, but in (9a), the superlative may be interpreted as limited (unfortunately, it it the best you can have; it is the best one that is available here). That is not the case in (9b), where the superlative is maximal (there is no better one; you cannot find a better one.)

The presence of pas_x seems to widen the domain in which the superlative is interpreted. This raises a number of questions: why would pas_x have this effect? how does it achieve it? and why does it surface in these particular constructions?

Kemp (1982: 277–281) establishes a parallelism between the negated modal in (10) and the adjective impossible in (11):

In (11), impossible is not interpreted literally, otherwise the sentence would be contradictory (it would assert that I did something impossible to do). It means something like: everything I could do, even unlikely things that could be considered impossible. According to Kemp, that is what the negated modal does in (10). The parallelism is interesting and relevant, but it is not sufficiently precise to constitute an analysis of the construction.

4. The NPI analysis of Gonzalez & Royer (2022)

Gonzalez & Royer (2022) develop an analysis in which pas_x is a hitherto undiscovered type of expletive item, namely one that forms a complex negative polarity item (NPI) in association with an existential predicate. Neither pas_x nor the existential predicate are NPIs by themselves. Therefore, the authors also claim to have discovered a new type of NPI, one that is built out of two syntactic parts. Their analysis, involving the activation of subset alternatives exhaustified by an operator, is based on the one proposed for the NPI any by Chierchia (2013).

Gonzalez & Royer (2022) define as follows the environments where pas_x (ExN for them) appears:

The sentences containing pas_x would involve an exhaustification operator EXH at the head of the sentence, a universal quantifier in the antecedent, and an existential quantifier in the relative clause:

The authors propose that, when pas is adjacent to one of the four predicates conveying existential quantification listed in (12ii), it may form a complex NPI with it. In that case, pas has no negative force. Its sole contribution is to force the existential expression it co-occurs with to activate a set ALT of subset alternatives of the relevant quantificational domain (G&R 2022: 10). EXH is defined as in (14):

“Given a sentence φ and a set of alternatives (ALT) to φ, EXH(φ) asserts φ and negates the alternatives that are not entailed by the assertion” (G&R 2022: 10). The sentence is grammatical if the EXH operator gives rise to a semantically coherent meaning.

The analysis crucially relies on the idea that a universal quantifier is downward-entailing in its first argument, which means that it licenses inferences from sets to subsets (Every boy that ate pizza got sick => Every boy that ate pizza with anchovies got sick). In the configuration in (13), every subset alternative activated by the association of pas with the existential operator in the restrictor of tout is entailed by the assertion. Therefore, in that configuration, “exhaustification is vacuous and simply returns the assertion” (G&R 2022: 11–12), as should be expected since pas_x is considered expletive. Sentence (15) is analyzed as in (16). It has the LF in (16a) and asserts (16b). The set of alternatives triggered by the association of pas with the modal involves subsets of alternative worlds (16c). Since every subset alternative is entailed by the assertion, exhaustification returns the assertion (16d).

For a criticism of Chierchia’s approach to NPIs, see Giannakidou (2018). Here, I will concentrate on the application of the account to pas_x. The analysis faces a number of theoretical and empirical problems.

The authors admit that they cannot explain why the NPI created by pas_x would be licensed only in universal contexts, and not in the other contexts licensing NPIs (e.g. negative, interrogative and conditional contexts). More generally, the fact that no relationship is established between the negative morpheme, its purported role in the construction, and the limitation to universal contexts makes us suspicious of the analysis. Why would a dialect of French select the negative operator to function as an expletive triggering the activation of subset alternatives, and that, only in universal clauses?

Consider the antecedent of the embedded clause. According to (12)–(13), it can only be a universal quantifier, superlatives being considered as universally quantifying. While most of the embedded clauses where one finds pas_x are introduced either by a universal or by a superlative, there are examples that do not conform to the model, and that show that the double implicature in (12) is incorrect. In (4b), for instance, the antecedent one of the good players is neither a universal quantifier nor a superlative:

In (4c), repeated below, pas_x is in a free relative, introduced by the WH expression qu’est-ce que ‘what’ functioning as a unit (Munaro & Pollock 2005; Abeillé & Godard 2021: 1524). The adjective incroyable has a superlative flavor, but it does not count as a universal antecedent: the free relative is anaphoric to the subject in a right dislocation type of structure. (C_i’est incroyable ça_i! ‘It’s incredible that!’).

Hence, the requirement that all antecedents be universal quantifiers or superlatives is too strong. Moreover, the authors assume in (12i) that superlatives are universal quantifiers over sets of degrees, but they do not provide a definition of superlatives, referring the reader to Heim (1999). However, in that paper, superlatives are existential quantifiers over degrees.² The authors do not provide a precise account of how superlatives fit into their model, and, a fortiori, they do not discuss the problem of Strawson downward-entailment. Contrary to universal quantifiers, superlatives do not directly license inferences from sets to subsets. (19a) does not entail (19b), since Max could live and play in San Francisco.

Assuming that the EXH operator negates every alternative like (19b) yields a contradiction: Max is the best player in California, but Max is not the best player in Los Angeles, nor in San Francisco, nor in Sacramento, and so on for every subset of California. Because that is incoherent, (19a) should be ungrammatical, contrary to fact. The problem arises because superlatives are Strawson downward-entailing: the downward-entailment follows only if we factor in a presupposition (von Fintel 1999 section 4.2; Gajewski 2010). If Max is the best player in California and if Max plays in Los Angeles, then it follows that Max is the best player in Los Angeles. In sum, it is far from clear how the analysis is supposed to carry over to superlatives. If it does not, at least half of the examples containing pas_x are unaccounted for.

Conversely, not every universal quantifier licenses pas_x. Free-choice tout does not, even though that is a downward-entailing, NPI licensing context (e.g. the indefinite un could be felicitously replaced by the NPI quelque…que ce soit ‘any’—Tovena et al. 2004).

Chaque ‘each’ does not license pas_x either: (21a) is ambiguous, but (21b) only has a reading with a regular negation.

Turning to the embedded predicate, it is claimed that it can only be one of the four predicates of the language assumed to convey existential quantification, those listed in (12ii). While many of the examples contain one of these predicates, it is not difficult to find examples where the embedded predicate is not a member of the list in (12ii), and does not convey existential quantification. There is, as far as I can see, no restriction on the type of predicate allowed in the embedded clause: it may denote a transitive event (22a voir ‘see’, 22b vendre ‘sell’, 22e réaliser ‘realize’, 22h donner ‘give’), an unaccusative event taking the être auxiliary (22c sortir ‘go-out’), an intransitive event constructed with the avoir auxiliary (22d éclater ‘burst’), or a state (22g souffrir ‘suffer’); it can also be a complex predicate, as in (22f) (rêvé de faire ‘dream of doing’), where the constituent quantified by tout is not an object of the negative verb, denoting an activity, but of an infinitival complement of that verb. (Examples b and c are adapted from examples collected by Kemp 1982, ex. 99 and 62):³

We conclude that the contexts in (12) do not constitute necessary and sufficient conditions for pas_x to be licensed. This puts in question the analysis, crucially based on the presence of a universal quantifier in the antecedent and on the association of pas_x with an existential predicate in the embedded clause.

Gonzalez & Royer (2022) argue that pas_x has no negative force. One of their arguments is that the expression du tout ‘at all’ (also pantoute ‘at all’), which must figure under the scope of negation (23), is not licensed by pas in (24b).

The conclusion that pas_x has no negative force is contradicted by (25b) and (26b). (25b) is a variant of (25a) obtained by adding encore to pas_x. The ‘yet’ meaning of encore is licensed by the negation: pas encore = ‘not yet’. If pas is removed, the clause is rejected (25c). (In positive contexts, encore has other meanings—still, again, more—, but none of them is compatible with 25c). Similarly, in (26b), même added to pas_x yields the negative expression même pas ‘not even’ containing the negative operator pas; if pas is deleted, the clause is rejected (26c). The translations in parentheses correspond to the analysis that will be developed in section 5.

Since, in (25) and (26), the presence of pas_x is required to give rise to the ‘not yet, ‘not even’ readings, we are led to conclude that pas_x has negative force. In consequence, the infelicitousness of (24b) must have another explanation.⁴ We will come back to (25b) in section 5.

Another issue is that Gonzalez & Royer do not account for the domain widening effect of pas_x illustrated in section 3. Gonzalez & Royer (2022: 14 fn 11) assume that NPIs do not need to yield widened domains, and that eventual domain widening effects are tied to the activation of sub-domain alternatives, but they do not spell-out how the effect emerges in context in the case of pas_x. Notice, that, contrary to what is the case with any, the domain widening effect illustrated in (9b) and (10) is not associated with focus or stress on pas_x. Pas_x is unstressed in these examples (cf. also 21a). Because there is no attempt to explain the domain widening effect of pas_x, we are left to wonder why a speaker would go to the trouble of producing the complex but vacuous operation of exhaustifying a set of alternatives activated by a syntactically constructed NPI, an operation that simply returns the assertion, given that there is a simpler construction without pas_x that asserts exactly the same thing.

In conclusion, there are reasons to be skeptical of Gonzalez & Royer’s (2022) NPI analysis. In the following sections, I endeavor to defend the idea that pas_x is a true negative operator. If that is correct, there is no need to postulate a new type of expletive negation or a new type of NPI to account for examples like (1). That is a welcome conclusion since it avoids the proliferation of theoretical constructs.

5. Superlatives and comparative NP movement

Superlatives constitute the majority of the examples containing an apparently expletive pas_x, and they are, I think, the clue to understanding the other examples. A superlative identifies the unique member of a comparison class C that has a higher degree of the quality considered than all the other elements of C (Heim 1999; Farkas & Kiss 2000). The biggest book denotes the unique book that is bigger than all the other ones in a given comparison class:

Thus, a superlative involves a comprehensive set of comparisons. In French, the relation between a superlative and the comparatives that it subsumes is explicit. A superlative is composed of a comparative preceded by a definite determiner introducing a uniqueness interpretation. When the adjective or adverb has no morphologically distinct comparative, the comparative morpheme is plus ‘more’.⁵

With this in mind, let us come back to the contrast between a sentence without pas_x and one that contains it, and ask ourselves what is the structure of these constructions.

Both sentences are movement constructions: they contain a gap in the complement position of the embedded verb; this gap results from a movement to the head of the embedded clause. I would like to argue that pas looks expletive in (29b) because the construction is implicitly assumed to have the same structure and interpretation as (29a), but, in fact, distinct constructions are at play.

Two different structures are assumed to be possible for relative clauses: the raising structure and the matching structure (e.g. Kayne 1994; Bhatt 2002; Hulsey & Sauerland 2006; Harris 2008; Cinque 2015). The structures vary from author to author. I will follow Hulsey & Sauerland’s (2006) structures illustrated below. Assuming the copy theory of movement, the raising structure of relative clauses is as in (30). The head NP moves to the specifier of the complementizer, and it is interpreted only in the embedded clause. In the matching structure (31), the head NP is interpreted outside of the relative clause, and there is an elided NP inside the relative structure. In (31), the moved element is striked-through and the elided one is respresented with a subscripted Ø. The matching NP must be similar enough to the head NP to license ellipsis.

Bhatt (2002) and Hulsey & Sauerland (2006) argue that both structures are available for English relative clauses and superlatives.

Coming back to French, for superlatives without pas_x, like (29a), I will assume that both the raising structure in (32a) and the matching structure in (32b) are possible. In both structures, the comparative adjective is generated within the matrix DP, and the complement of the embedded verb is the noun livre.⁶ In these constructions, the embedded clause defines the comparison class of the superlative. These constructions are grammatical both in standard French and in Quebec French.

For the construction in (29b), specific to Quebec French, I would like to defend the hypothesis that it is interpreted as in (33a), and that it has the matching structure in (33b):

Here, the complement of the embedded verb is the comparative NP meilleur livre ‘better book’, which is moved to the head of the clause, and elided under phonological identity with the comparative NP meilleur livre that complements the determiner as part of the matrix superlative. Within the embedded clause, the elided comparative NP is phonologically silent, and it is interpreted at LF (like the elided NPs in 31 and 32b).⁷

Under the negation, comparatives may be bare NPs in French, as shown in (34), which is the source of the embedded clause in (33) above.

Pas_x in (33) is exactly the same entity as pas in (34). It is the regular morpheme of predicate negation scoping over a comparative. In (33), if the negation is deleted, the interpretation becomes incoherent, since, without pas, the sentence would mean: it’s the best book such that there is a better book. Thus, pas_x is not expletive in that construction and it is not optional. Comparing (29a) with (29b), one has the impression that pas_x is optional, but that is incorrect. Distinct constructions are involved: in (29a), without negation, the comparative adjective belongs to the antecedent (32), but in (29b), it is part of the moved NP, within the embedded clause, which is necessarily negative (33b).

It can be seen in (34) that the comparative NP has an implicit complement corresponding to the standard of comparison. The standard of comparison is also implicit in (33), and the embedded clause means there is no better book <than x>, with a variable corresponding to the standard of comparison:

When a comparative NP is displaced, it pied-pipes the standard of comparison:

Therefore, I assume that, in (33), the implicit standard of comparison is pied-piped to CP with the comparative NP, and that the CP of the embedded clause contains a variable.

That allows the embedded clause to combine with the superlative antecedent, yielding (38), where the x variable of the embedded CP is bound by the uniqueness operator of the superlative.

I will come back in the next section to the apparent redundancy of (38), and will argue that the pragmatic role of the embedded clause is to enlarge the comparison class, but before getting there, I wish to provide arguments supporting (33b). In the syntactic representations below, I will not show the null standard of comparison forming part of the elided comparative NP, as it would only complexify the representations without contributing to the argumentation.

To start with, note that (33) applies, not only to clauses where the embedded predicate is a member of the four predicates in (12ii), but also to clauses headed by a different verb, like voir in (22a).

Sentence (25b), repeated below, where the ‘yet’ meaning of encore is licensed by the negation, is accounted for by (41).

The meaning of (40) requires the implicit complement of the verb to be a comparative NP. That is because the focus of pas is the comparative adjective: the clause presupposes that I have known some evenings before, and states that none was more beautiful than this one. If, instead of (41), we assumed one of the structures in (32) with the comparative adjective generated in the antecedent, e.g. la plus belle [veillée que j’ai pas encore connu veillée], the comparison class would be a set of future evenings: the most beautiful among the evenings that I have not yet known. That would not reflect the meaning of (40).

We can see in (42) that the definite determiner forming the superlative is generated in the main clause and the embedded clause contains an indefinite complement. Y avoir is the infinitive of the existential expression il y a (lit.: it.loc.has), meaning ‘there is’, which, like its English equivalent, is constructed with an indefinite complement. The indefinite complement of il y a is the bare comparative NP plus beau carosse (42b).

The following example is interpreted with the comparative NP below the belief: that is the biggest stupidity such that he thinks that I haven’t done a bigger stupidity than that. That follows from (44), with successive cyclic movement of the comparative NP.⁸

An argument for the matching structure is the possibility of extraposing the relative clause past a temporal adverb (Hulsey & Sauerland 2006):

Also supporting the matching structure is the following example, quoted by Kemp (1982 ex. 103), where an it-cleft intervenes between the antecedent and pas_x. Extraction out of an it-cleft is also possible in relative clauses (47).⁹

The interpretation of (46) follows from (48), where only the initial and final copies of the comparative NP are represented.

The fact that the elided constituent is not identical to the antecedent, but is similar enough for ellipsis to be licensed, provides support for the matching structure. Pas in (48) is the same as pas_x in (33): a negative operator scoping over an implicit comparative NP. The sentence has no equivalent without pas; the structure in (32b) would not have the correct interpretation: the most beautiful of the parties among the parties that many spend.

To summarize, a number of facts support the matching structure in (33), in which pas_x is the regular negative operator, and the gap in the embedded clause corresponds to a moved and elided comparative NP. The focus of pas_x is the comparative adjective forming part of the silent complement of the verb. Pas_x is not expletive, and there is not always a sentence without it that has a similar interpretation.

6. The domain widening effect of superlatives containing pas_x

I argued that the superlative in (49a) has the matching structure in (b), with the interpretation in (c).

In a matching structure like (49b), the superlative is interpreted in the main clause, and the comparative in the embedded clause. Recall that a superlative is interpreted with respect to a contextual comparison class C. If (49c) means: this is the best book in the comparison class C such that there is no better book than it in C, it is redundant, and the embedded clause is uninformative. However, we saw, in section 3, that the presence of pas_x widens the domain in the sense that, while the comparison class of a superlative without pas_x may be limited, when pas_x is present, the context is maximal. I would like to propose that the pragmatic strengthening effect of the presence of pas_x comes from the fact that it negates a comparative. There is no better book invites us to consider the largest possible domain. While the superlative in the antecedent is interpreted with respect to a contextual comparison class C, the embedded clause tells us that we will not find in a larger domain C’ an element that is higher than the one in the antecedent on the scale denoted by the adjective. If my interpretation is correct, the sentence is not redundant. The embedded clause containing pas_x yields a stronger statement by forcing us to go for a larger domain. This domain-widening effect results from combining the negation with a comparative.

Consider (50), integrating the comparison classes into (49c). The main clause superlative is interpreted with respect to a contextual comparison class C, and the embedded clause tells us that there is no better book in a C’ larger than C:

The embedded negated comparative activates larger sets. The interpretation is absolute because, for any C’⊇C that we may pick, there is no better book than x in C’. Therefore, x is the best book in the set of all books.

We can compare that to the sentence without pas_x, where the embedded clause does not contain a negated comparative:

In that construction, we are free to interpret as we wish the domain in which qu’il y a ‘that there is’ is valid: …qu’il y a sur la table ‘on this table’ / dans ce magasin ‘in this store / dans cette ville ‘in this town’…. The interpretation may be restricted to a limited contextual domain.

When the embedded clause contains a specification of a comparison class, that comparison class is the one over which the comparative is claimed to be valid. Take (52), for instance. The contextual comparison class C of the main clause superlative could be limited to the wheat varieties on a particular market stand, but if the owner of the stand utters (52), he places emphasis on the fact that there is no better wheat on the whole trading market.¹⁰

The comparison class is therefore enlarged from C to C’, defined as being the trading market, and the negated comparative states that no wheat is better than this one in this enlarged domain. From that, we conclude that the superlative holds in the enlarged domain. The focus of the negation is the comparative adjective. The embedded clause presupposes that there is wheat on the market and pas_x, associating with the comparative, states that none is better than this one. A reviewer observes that while in (49) the noun was a count noun, the analysis is applied here to a mass noun; true, but distinct, thus countable, varieties of wheat are compared.

In the equivalent sentence without pas_x, sur le marché ‘on the market’ defines C, and the sentence does not have the domain widening effect of the embedded clause containing a negated comparative:

The following sentence illustrates the domain widening effect of the negated comparative with a set-denoting NP. Here the comparison class is enlarged to all the players that we had.

In (55), the main clause speaks of the conditions in which this child was raised, and the embedded clause enlarges the comparison class to the conditions that you can imagine.

In (56), the contextual enlargement of C is defined by what you can possibly have (which is pragmatically larger than the present contextual domain).

We can summarize the present hypothesis as follows:

The embedded clause strengthens the assertion by stating that no entity is ADJ-er than the antecedent in a set larger than the present contextual domain.

Gonzalez & Royer (2022: 5 ex. 10b) claim that the inability of pas_x to cooccur with the negative coordination ni…non plus (literally ‘nor…neither’) in (58a) proves that pas_x is expletive. However, we see in (58b), that a continuation with ni…d’ailleurs ‘nor…for that matter’ is perfectly acceptable.

The fact that a conjunct introduced by ni ‘nor’ is possible confirms that pas_x in the embedded clause has negative force. The problem with (58a) is that non plus ‘neither’ following the locative PP à Montréal implies that the focus of the negation in the first conjunct is the locative PP à Québec, but that is incorrect: as we saw, the focus of the negation is the silent comparative adjective within the elided complement; the location defines C’. In (58b), the conjunct ni au Canada d’ailleurs strengthens the assertion by enlarging C’ from Quebec City to an even more encompassing set (Canada ⊃ Quebec City): we won’t even find a better restaurant in the whole country.

7. Other silent comparatives

I would like to argue that the role of pas_x that we just saw for superlatives is also the one that is at play in (4a) and (4b), where the antecedent does not overtly contain a comparative.

Consider the case where the antecedent contains the ordinal premier ‘first’ in (4a), repeated below. The implicit complement of the embedded verb cannot be premier anglais: (59b) is contradictory (see 33a):

In standard superlatives, the determiner is followed by an adjectival comparative identical to the elided comparative in the embedded clause, but the ordinal premier ‘first’ is not a comparative, but a superlative; that is why it cannot occupy the gap position. If individuals are ranked on an ordinal scale ⟨first,last⟩ defined by the order of arrival in the neighborhood (where first>next>last), le premier designates the individual occupying the highest position on the scale, and it entails an exhaustive list of comparisons: Joe came first in the neighborhood entails Joe came first with respect to Tom, and Joe came first with respect to Sam, and Joe came first with respect to Bob, etc. The presence of pas_x in (59a) is accounted for if we assume that the implicit object of the embedded verb is a silent comparative NP corresponding to a y that is higher than x on the ordinal scale (60a). After raising the implicit comparative to the specifier of CP (60b), the structure of the embedded clause is that of (33).

Assume that we are talking of Joe. The sentence is interpreted as in (61a), schematized in (b):

The interpretation is parallel to that of standard superlatives: for any larger set of Englishmen that we may activate, no Englishman came first in the neighborhood with respect to Joe. Like the other superlatives, (59a) is Strawson downward-entailing: He was the first Englishman that came to the neighborhood does not entail that he was the first left-handed Englishman that came to the neighborhood, unless we presuppose that he was left-handed. The difference between superlatives and ordinals is that, with ordinals, the meaning of the embedded clause cannot be spelled out with an overt comparative adjective: x was the first Englishman s.t. there was no y *first-er than x. But since, in the matching structure, the comparative NP is elided, therefore interpreted at LF, it is not a problem that, here, the embedded comparative has no phonological realization, as long as it is interpreted at LF in the same way as the elided comparative NP is interpreted in superlatives: the implicit comparative in (61b) has the same ordinal scale as the antecedent, and its second argument is bound by the antecedent.

In (62), heard on the radio, the partitive antecedent un des meilleurs joueurs ‘one of the best players’ contains a superlative, and the example may be derived as in the preceding sections. That is shown in (62b), where the portmanteau partitive des of (62a) is separated into its components de les ‘of the’.

But in (4b), repeated below, the expression un des bons joueurs ‘one of the good players’ does not contain a superlative, and the embedded clause could not possibly mean that we did not have good players in Canada.

Rather, the embedded clause is interpreted with a comparative plus in the implicit object of the embedded verb.

Since plus bon ‘more good’ = meilleur ‘better’, it turns out that the embedded clause is (64) is interpreted like that of (62b) having a superlative in the antecedent.¹¹

That may seem surprising until one realizes that the interpretation of one of the good is exactly the same as that of one of the best, except for the size of the set considered. Both one of the good and one of the best entail a ranking of players along a scale of goodness and refer to the top set containing the best players. But the set of ‘good’ players is larger than the set of ‘best’ players. Let us arbitrarily suppose that, to be considered one of the best players, one must belong to the set composed of the five players at the top of the scale of goodness, but to be considered one of the good players, one must belong to the set composed of the twenty best players. Under that assumption, (65) is interpreted as in (66). Because of the partitive, the standard of comparison is the individual identified in the antecedent as being a member of the set of good players.

(64) behaves like a superlative because the set referred to in the antecedent contains the individuals at the top of the scale of goodness (i.e. the 20 best players). Like with superlatives, pas_x scopes over a silent comparative whose scale is that of the antecedent (a scale of goodness), and it negates the existence in an enlarged context of a set of entities higher than the standard of comparison on that scale. I think that this explains why pas_x is found in (63). If, in (66) we replace ‘twenty’ by ‘five’ we get the interpretation of ‘one of the best’, and, mutatis mutandis, (66) reflects how the partitive in (62), with a superlative in the antecedent, is interpreted.

The interesting aspect of (59) and (63) is that, to account for the interpretation, it is essential to postulate that the embedded complement contains a silent comparative that has no overt match in the antecedent.

8. Universal antecedents

Let us finally turn to the case where the relative clause is headed by a universal quantifier.

(67) is interpreted as saying that he invited each and every one of the musicians in the area. Like the redundant expression ‘each and every one’, the presence of pas_x has a strengthening effect on the interpretation. It stresses the fact that every relevant entity is included in the set quantified over by the universal: no entity is overlooked. In that sense, its role is similar to the role of pas_x in superlatives: it maximizes the set over which the universal quantifies.

We saw in the preceding section that the interpretation of sentences containing pas_x requires us to postulate, in some constructions, an implicit comparative not overtly present in the antecedent. With universal antecedents, the pragmatic strengthening effect of pas_x in (67) may similarly be accounted for if we hypothesize that the focus of pas_x is a silent comparative <plus> within the relative clause: there aren’t <more> musicians in the area. <Plus> is the comparative morpheme, and as such it requires a scale. With count nouns like musicians, the comparative scale is defined in terms of numbers (Bale & Barner 2009): if I didn’t eat more bananas, I didn’t eat a larger number of bananas, i.e. the set of bananas that I ate is not larger that this set, ‘this set’ being the standard of comparison. In (68), which is the matching structure for relative clauses, the embedded clause states that there aren’t more musicians <than that> in the area.¹² The relativized NP musiciens moves to the head of the embedded clause where it is elided under identity with the head musiciens in the antecedent. The embedded clause is negative, and it contains a silent comparative <plus> ‘more’ over which pas scopes, as well as its associated standard of comparison <que ça> ‘than that’.¹³ Supporting the matching structure is the fact that extraposition past a temporal adverb is possible (69) (Hulsey & Sauerland 2006).

Let us assume that the relative clause in (68) is interpreted as in (70): pas_x<plus> activates larger sets and states that there is no relevant entity outside of the set X forming the standard of comparison.

Suppose that the definite DP les musiciens ‘the musicians’ in the antecedent introduces a unique set A of musicians into the interpretation:

By combining the relative clause with the antecedent, we end up with (72), where the set variable in (70) is bound by the uniqueness operator in the antecedent. The restrictor set quantified over by the universal quantifier is a set of musicians that are members of the unique set A such that there are no musicians in the area that do not belong to A.

Under that analysis, the negative morpheme pas_x in (68) has nothing peculiar. It is simply the negative operator scoping over a comparative. Within the relative, the affirmative predicate is presupposed (there are musicians in the area), and the focus of pas_x is the silent comparative <plus>. The peculiar aspect of (68) is the silent <plus> that is postulated. That hypothesis accounts for the strengthening effect of pas_x: pas_x scoping over <plus> forces us to postulate the restrictor set A that satisfies (72), that is, the largest possible one containing all the musicians in the area. Like with superlatives, the strengthening effect comes from the fact that the negation scopes over a comparative. The hypothesis that pas_x is a negative operator scoping over a comparative provides a unified account covering both superlative and universal constructions. In both cases, pas_x appears in a movement construction, and its focus is a silent comparative. In both cases, the standard of comparison is bound by a uniqueness operator in the antecedent. In both cases, the strengthening effect of negating a comparative is to widen the domain considered (the comparison set or the restrictor set). We understand why pas_x surfaces in these particular constructions, because they involve a set that may be pragmatically widened. In every case, pas_x appears to be expletive and optional because the constituent over which it scopes is implicit, and because there exists a distinct construction without pas that has a similar meaning, but no domain-widening effect. I conjecture that the silent <plus> in (68) is an extension to universal clauses of the silent comparative at play in superlative constructions containing pas_x. That is, seeing, in superlatives, an apparently expletive pas_x scoping over an implicit comparative and having a domain widening effet, speakers of Quebec French extended to universal constructions the use of pas_x plus a silent comparative to express domain widening.¹⁴ The hypothesis that there is a silent <plus> in the construction, if it is correct, gives support to Kayne’s (2016) claim that there are more silent heads in syntax than what is usually thought.

The research on NPIs has shown that, in downward-entailing contexts, set enlargement typically consists in adding to the set more peripheral members, items that are pragmatically less probable than what would normally be assumed (e.g. Kadmon & Landman 1993; Chierchia 2006). I do not think that that is at play in (67), but we can observe it in (3a), repeated below, which, if my hypothesis is correct, would contain an implicit <plus> in the embedded clause (74a), and would be interpreted as in (74b). We may assume that, in this context, the D head ‘ce’ introduces a definite set into the interpretation,¹⁵ and that the restrictor set of tout is as in (74c).

There are no necessary and sufficient conditions for something to count as an x such that I could do x in a particular context. If I did everything I could to find him, did I go to Madeira? It depends on the context. The things that I could do may be ordered on a scale based on relevance or pragmatic probability in the context. Normally, when we hear everything that I could do, we assume in the restrictor set all the things at the top of the relevance scale up to a certain point where the actions may be deemed less probable. Pas_x, associated with a silent comparative, tells us to pragmatically enlarge the set to include actions figuring lower on the scale (up to a point where the action is deemed definitely irrelevant). That is in line with Kemp’s (1982) idea that, when pas_x is present, the interpretation is similar to ‘I did the impossible’.

Similar to (73) in meaning is the exclamative sentence discussed in (18), where pas_x is within a free relative:

The fact that a free relative denotes a set explains why it may accept pas_x with the same strengthening effect as with a universal antecedent: the set of x such that x is a thing he could do to find him is the pragmatically largest one there is. The acceptability of pas_x in an exclamative construction like (75) is closely related to the existence in French of negative exclamatives like the following ones:

Exclamatives are presuppositional and scalar (Michaelis 2001), like the other constructions discussed here. Formally, the exclamations in (76) are negative WH interrogatives whose expected answer is ‘close to nothing’. For instance, (76a) comments on the fact that there is almost nothing that he didn’t do to find him. Thus, he did everything he could to find him. Here again, the set of x such that x is a thing he could do to find him is maximal.¹⁶

The association of pas_x with a silent comparative activates larger sets and states that the restrictor set is the largest possible one. That requires a domain with an indeterminate number of items. In (67), the restrictor set is defined intensionally, and one does not necessarily know how many musicians there are in the area; some elements could easily be overlooked. In (73), the set of items that would normally be considered in the context is not precisely defined, and it may be pragmatically enlarged by including more marginal members. As observed by Kemp (1982), it is infelicitous to use pas_x when the domain contains a limited and easily quantifiable number of items:

The present approach suggests an answer to why tout ‘all, every’ licenses pas_x, but chaque ‘each’ does not, as we saw above in (21). We assume that tout DP, found in tous les N ‘all the N’ and in tout ce que ‘everything that’, is defined in terms of relations between sets but chaque N ‘each N’ is defined in terms of individuals and their properties (Beghelli & Stowell 1997; Knowlton et al. 2022). (78a) is true if the set A of children that participate is included in the set B of people who get a reward; (78b) is true if each individual that is a child and participates is also an individual that gets a reward.¹⁷

Assuming that, with universal antecedents, pas_x <plus> tells us that the restrictor set is the pragmatically largest relevant set there is, then it is expected to be possible (given the right context) with tous les N, which refers to sets, but not with chaque N, where the relative is construed in individual terms. An individual construal could also be invoked to explain the ungrammaticality of pas_x with free choice tout N in (20).

To summarize, in the present section, I proposed that, when pas_x is present in the universal construction, it scopes over a silent comparative <plus> ‘more’, forcing the restrictor set to be the largest relevant one. If we accept the hypothesis that there is a silent comparative negated by pas_x, the domain-widening interpretation follows. Moreover, that hypothesis provides a unified account of pas_x in both superlative and universal constructions.

9. Conclusion

In Quebec French, an apparently expletive negative morpheme pas may be found in superlative and universal constructions. The contexts in which this pas is licensed are not the ones licensing the usual cases of expletive negation. It was argued that pas is not expletive in these constructions. The characteristics of the constructions fall into place under the hypothesis that pas is a regular morpheme of predicate negation scoping over a comparative. If that is correct, there is no need to postulate a special type of expletive negation to account for it. Superlative and universal constructions require respectively a comparison set and a restrictor set. When the apparently expletive pas is present within the embedded clause in these constructions, it has a domain-widening effect on the comparison or restrictor set. The hypothesis that pas negates a comparative accounts for this domain widening effect.

In order to account for the interpretation of the constructions, it was necessary to postulate an implicit comparative. Whenever the apparently expletive pas surfaces in an embedded clause, the comparative over which it scopes is silent, and interpreted at LF. The negative morpheme appears to be expletive because its focus is silent, and because there exist parallel constructions without it that have a related meaning but no domain widening effect.

The main conclusion of the present paper is that a detailed analysis of an item that appears to be expletive at a superficial level may reveal that it is not expletive after all, and that it may be explained if silent categories are taken into account.

Competing interests

The author has no competing interests to declare.