(Negative) Polarity Items in Catalan and Other Trans-Pyrenean Romance Languages

: This paper identiﬁes the set of properties that polarity items (PI), negative polarity items (NPI) and negative concord items (NCI) satisfy in Catalan, Aragonese, Benasquese and Occitan. It shows that in Catalan, gaire ‘much, many’ is a PI, pas ‘at all’ is an NPI with enriched meanings and indeﬁnites such as ning ú ‘anybody, n-body’ come in two sets, as both PIs and NCIs. It further shows that in the trans-Pyrenean Romance varieties studied here, a distinction must be made between indeﬁnite pronouns such as deng ú n ‘anyone, n-one’ and scalar minimizers such as gota (lit. drop), which come in parallel PI and NCI sets, whereas NPIs such as brenca (lit. crumb) have developed an enriched meaning. This paper reveals the different status of pas across Catalan and other trans-Pyrenean Romance varieties either as an NPI at stage II of Jespersen’s cycle, which further constrains conventional implicatures, or as a negative operator.


Introduction
The linguistics literature has often referred to a set of phenomena sometimes labelled (negative) polarity (Ladusaw 1979;Giannakidou 1997 and others) and other times negative concord (Labov et al. 1968;Labov 1972 and others), without the difference between these two labels being clearly defined. Let us assume (following Espinal et al. 2021) that polarity sensitivity is a formal relationship between two constituents X and Y, where X c-commands Y, X is a semantically non-veridical licensor and Y is a licensee semantically sensitive to non-veridicality (Giannakidou 1997(Giannakidou , 1998(Giannakidou , 1999. 1 Negative polarity is a formal relationship between two constituents X and Y, where X c-commands Y, X is a semantically anti-veridical licensor and Y is a licensee semantically sensitive to anti-veridicality. According to these definitions, negative polarity is, strictly speaking, a subtype of polarity sensitivity. By contrast, negative concord is a formal relationship between two constituents X and Y, where X c-commands Y, and both X and Y encode a syntactic formal feature [neg] that enters syntactic Agreement and composes a single negation reading. 2 Therefore, we assume here that a distinction must be made between semantic dependencies (Ladusaw 1980(Ladusaw , 1996Giannakidou 1997 and ff.) and syntactic dependencies (Zeijlstra 2004 and ff.), as recently discussed by Dočekal (2020).
In order to refine our understanding of the distinction between (negative) polarity and negative concord, the present study aims to present an overview of the set of items that behave like Polarity-Sensitive Items (PIs for short) in Catalan and other Romance varieties spoken in the Pyrenees, including Languedoc Occitan, Gascon Occitan and its Aranese variety, Benasquese and Aragonese (Llop 2020). The existing research on the varieties of this area has provided important data not only on the linguistic continuum embodied in these varieties (Bec 1968;Nagore 2001;Suïls 2011;Suïls et al. 2010)  The aim of this article is to present an overview of those items in Catalan and other trans-Pyrenean Romance languages that behave like gaire. The behaviour of the PI gaire 'much, many' contrasts with the behaviour of other indefinites that are characterized by: (a) the possibility of their occurring alone as negative fragment answers, (b) the possibility of their occurring with no c-commanding licensor in preverbal position of declarative sentences, (c) the impossibility of their being licensed long-distance, (d) the licensing of a negative reading when they occur in preverbal position in declarative negative sentences (in the absence of a negative marker in the case of Non-Strict NC structures, and in combination with a negative marker in the case of Strict NC structures 5 ) and (e) the possibility of their contributing to a double negation reading when multiple negative indefinites, distributed preverbally and postverbally, combine with an overt negative marker. 6 Those indefinites that share these properties were originally labelled n-words (Laka 1990). We will refer to them here as Negative Concord Items (NCIs).
Accordingly, in this study we address the following questions: 1.
Are there any other items in Catalan varieties that follow the properties attributed to PIs? Answering this question will require considering a subclass of minimizers (not preceded by ni 'not even') (Tubau 2020), and a series of indefinites (e.g., ningú, cap, res, gens, mai, enlloc) that are shown to behave both as PIs and NCIs ).

2.
Is the distinction between PI, NPI and NCI relevant in Catalan? We consider the need to distinguish between PIs licensed in non-veridical contexts (Giannakidou 1997(Giannakidou , 1998, a reduced subset of PIs that are only licensed in anti-veridical contexts and that present collocational restrictions (Postal 2004)-here referred to as NPIs, and an independent subset of PIs that behave like NCIs and are involved in a syntactic Agreement relationship. 3.
Which PIs satisfy the abovementioned criteria in other Romance varieties spoken in the Pyrenees (Llop 2017(Llop , 2020? Do the varieties that show homophonous PI-NCI sets exhibit behaviours like those seen in Catalan? The structure of the paper is as follows. In Section 2 we focus on the different types of polar sensitive items that can be distinguished in Catalan, namely PIs (gaire 'much, many'), NPIs (pas 'at all'), homophonous PIs-NCIs (ningú 'anybody, n-body', cap 'any, no', res 'anything, n-thing', gens 'anything, n-thing', mai 'ever, n-ever', enlloc 'anywhere, n-where') and minimizers (including those without ni 'not even', which behave like PIs, and those with ni, which behave like NCIs).
In Section 3 we argue that some PIs are strictly dependent on the sentential antimorphic operator no 'not' (van der Wouden 1994;Zwarts 1996;Giannakidou 1998 and ff.) and therefore behave like NPIs (thus showing a diachronic change from a broad semantic dependency to a narrow semantic dependency), while other PIs show NCI-like behaviour (thus reflecting a diachronic change from a semantic to a syntactic dependency). We argue that, on close inspection, no items appear to follow the postulated 'ordered passage from weak NPIs to strong NPIs to n-words' (Larrivée 2021; see also Larrivée and Kallel 2020; among others), but rather either a transition from PIs to NPIs (from requiring non-veridical licensors to anti-morphic licensors), or a transition from PIs to NCIs (from a semantic licensing relationship to a syntactic Agreement dependency), thus supporting the conclusion that NPIs and NCIs are independent subspecies of PIs .
In Section 4 we move to other trans-Pyrenean Romance languages and, on the basis of Llop (2020), we describe the set of items that behave either like PIs/NCIs, like scalar minimizers or like NPIs with additional pragmatic enriched meanings Visconti 2009, 2014;Batllori 2015;Larrivée 2010Larrivée , 2020among others).
Finally, Section 5 concludes the paper. The sources of the data discussed in this article are examples from the authors unless otherwise stated.

Types of Polarity-Sensitive Items in Catalan
In Central Catalan 7 gaire 'much, many' is the only item that behaves like a full PI, since it follows the whole set of properties illustrated in (1). This item may correspond to an indefinite determiner, in which case it shows number agreement with the plural count noun it specifies (2), or it may correspond to an adverb, in which case it is either the specifier of a second adverb (3a) or a verbal modifier (3b).
( In this section we review the set of items that behave either partially or entirely like gaire. On the basis of this behaviour we will be able to conclude whether these items are PIs, NPIs or NCIs. We first consider items such as ningú, second minimizers and finally pas. Firstly, indefinite pronouns including ningú 'anyone', res 'anything', gens 'anything', mai 'ever', enlloc 'anywhere' and indefinite determiners such as cap 'any' and gens de 'anything of' behave like PIs with respect to: (a) the impossibility of being licensed unless they appear under the scope of a non-veridical operator such as the interrogative operator (Giannakidou 1997(Giannakidou , 1998, (b) the possibility of being licensed long-distance and (c) the licensing of an existential reading in non-declarative non-negative contexts.
See the examples in (4) to (6) with ningú, the only item in this group that is negative in its origin (from Latin necūnu 'not even one'). We refer to the PI set made up of ningú 'anyone', cap 'any', res 'anything', gens (de) 'anything (of)', mai 'ever' and enlloc 'anywhere' when these indefinites behave as in examples (4) to (6). We also refer the reader to Table A1 in Appendix A for the etymology of the various PIs studied here and in the remainder of the paper. By contrast, we refer to the NCI set made up of ningú 'n-one', cap 'no', res 'n-thing', gens (de) 'n-thing (of)', mai 'n-ever' and enlloc 'n-where' when these indefinites show: (a) the possibility of occurring alone as negative fragment answers; and (b) the possibility of occurring with no c-commanding licensor in preverbal position in declarative negative sentences, as exemplified in (7) and (8)  Secondly, we consider the case of minimizers (Pott 1857;Wagenaar 1930;Bolinger 1972;Horn 1989;and others). These items are minimal measure-denoting expressions such as English a word, a wink, an inch or Catalan un duro (lit. a small denomination coin) 'a penny', una ànima (lit. a soul) 'a person', una mosca (lit. a fly) 'a person', which give rise to an even reading. Catalan minimizers can be optionally preceded by ni 'not even' (Tubau 2020). However, as the following examples illustrate, only when ni is present does the minimizer exhibit the distribution characteristic of an NCI, as exemplified in (10)  The third item we must consider in Central Catalan is pas (lit. step). An experimental study carried out by Tubau et al. (2018) has demonstrated that pas increases neither the amount of negative interpretation of no in the context of fear predicates that trigger expletive negation, nor the amount of double negation in the context of a preverbal NCI. This study calls into question the alleged strengthening role of Central Catalan pas since it does not reverse by itself the truth or falsity of a proposition. In other words, unlike its French counterpart, in Central Catalan, pas is unable to reverse the polarity of a sentence and hence to perform on its own the role of a sentential negation marker. 10 (12) Tubau et al. 2018, p. 125, exs. (6a) and (6c)) Example (12b) illustrates an interesting phenomenon that deserves special attention. Central Catalan has been characterized elsewhere (Espinal 2000(Espinal , 2002Espinal and Tubau 2016; and others) as a Non-Strict NC language (Giannakidou 1998) since, when a preverbal NCI is present the negative marker can be omitted, as is also the case in other Romance languages such as Italian, Portuguese and Spanish. 11 However, when pas occurs in postverbal position, its only possible licenser is an overt negative marker no. In other words, pas differs from the other PIs we have considered so far in that it requires an anti-morphic licensing operator, and from this we conclude that pas is an NPI. Furthermore, pas differs from NCIs in that in Central Catalan it can appear neither as a fragment answer nor in preverbal position. Finally, the licensing of pas in the context of a raising predicate is illustrated in (15). As we saw in (1c), this behaviour indicates that pas is an NPI, since its licenser is restricted to the negative marker no.  To sum up, in this section we have provided support for a distinction between PIs, NPIs and NCIs in Contemporary Central Catalan. Gaire is exclusively a PI, pas is exclusively an NPI, while indefinites of the sort exemplified by ningú and minimizers come in two sets, as PIs and as NCIs.

Diachronic Changes
In this section we distinguish the emergence of NPIs from the emergence of NCIs, which will in turn support the distinction between the diachronic development of the contemporary negative marker pas and the diachronic development of negative indefinites (Zeijlstra 2016).
The former is characterized by a diachronic change according to which a nominal item ends up strictly constrained by a semantic dependency with respect to the anti-morphic operator no 'not', while PIs in general are licensed under the scope of various non-veridical licensers.
We hypothesize (in accordance with Pérez Saldanya and Torrent 2021) that a reanalysis from the nominal expression passu(m) 'step' to the minimizer denoting the minimal distance covered by walking may be the origin of the NPI variant. 12 As illustrated in (16), a nominal passum 'step' in combination with pedis 'foot' appears under the scope of the negative marker nec already in Late Latin (nec passum pedis lit. no step of foot). This use alternates already in the 13th century with other uses in which there is no indefinite marker and a bare pas is supposed to strengthen the negative marker either from an immediate postverbal position or from a close position, but without a quantificational meaning. 13 The possibility of an optional pas, also characteristic of Old and Middle French ne . . . (pas) (Labelle and Espinal 2014), is the core behaviour of this item in Contemporary Central Catalan, from which we conclude that this discontinuous negative marker is at stage II of Jespersen's (1917) cycle (Jespersen 1917;Dahl 1979;van der Auwera 2009;Larrivée and Ingham 2011), with the proviso that-unlike French-pas in Catalan introduces additional semantic and pragmatic constraints. Tubau et al. (2018) experimentally support an analysis of pas according to which it has two components: on the one hand, it is an NPI, but on the other hand it either semantically constrains the implicatures that can be drawn from an accessible discourse context (Espinal 1993(Espinal , 1996, or contributes a denial of a salient discourse-old proposition (Schwenter 2006) or adds emphatic content (Batllori 2015). This enforced pragmatic meaning could be thought of as analogous to English 'at all'. 15 Note, furthermore, that pas in Contemporary Catalan is no longer a nominal expression (as it was in its etymological origin), nor an indefinite one (as characteristic of most PIs), but rather an adverb (see DIEC2), thus indicating that NPIs are the output of both a semantic and a categorial change. This means that, whereas as a minimizer pas is associated with a minimal quantificational meaning, as a so-called strengthener of negation (in appropriate contextual settings) it has lost its quantificational meaning and has developed new meanings that are enriched at the time of utterance interpretation. 16 Let us now consider the evolution from PIs to NCIs in Catalan. In the previous section we already noted that in Contemporary Catalan, items such as ningú 'anybody, n-body', cap 'a, no', res 'anything, n-thing', gens (de) 'anything (of), n-thing (of)', mai 'ever, n-ever' and enlloc 'anywhere, n-where' come in two sets. As suggested by the word translations just provided, these items are used both as PIs-licensed under the scope of all sorts of non-veridical semantic operators (e.g., conditionals, questions, comparatives of inequality, downward entailing operators, the negative operator, etc.)-and as NCIs-licensed by a syntactic Agree relation. 17 What is constant in their diachronic evolution from PIs to NCIs is that these items have maintained their indefinite meaning; what is different is that, once these indefinites acquire a syntactic feature [neg] this feature must be checked at syntax.
Focusing on ningú 'anybody, n-body', Pérez Saldanya and Torrent (2021, Table 56.5) point out the existence in the history of Catalan of several alternative forms, namely negú/un (the predominant form in written texts from the second half of the 11th century to the second half of the 15th century), together with nengú/-un, degú/-un and dengú/un (rarely used from the second half of the 13th century to the 16th century). Ningú becomes the most frequent form starting in the first half of the 16th century. However, most important for our purposes is the fact that the forms negú/-un, nengú/-un, degú/-un and dengú/un, regardless of whether they occur in preverbal or postverbal position, always coappear with a licensing overt negative marker in a scope position. Furthermore, negú/-un, as a PI in a non-veridical licensing context, is associated with an existential meaning. It is only in the 15th century that it becomes possible to associate a negative meaning with a preverbal ningú even in the absence of the negative marker no. Consider (21) Similarly, an NCI use of cap 'no', illustrated by a preverbal occurrence with no overt negative marker that conveys a negative meaning, is also documented for the 15th century. Concerning ningú and cap, it is worth noting that the emergence of an NCI use with a plain negative meaning does not exclude a PI use. Thus, Rigau (1998, p. 66) points to sequences such as those exemplified in (23) and (24) in the Contemporary Catalan variety spoken in Ports de Morella (see also Hualde 1992, p. 160;Morant 1993, p. 127 Concerning res, Pérez Saldanya and Torrent (2021, p. 13) point to the simultaneous existence since the 14th century of two associated meanings for the pronominal form no res, 'negation of any thing' and 'insignificant thing'. Example (25) illustrates one of the diachronic developments commonly manifested in natural languages, by which a nominal expression res 'thing' turns into a generalizer when used as a PI (Condoravdi and Kiparsky 2006 Gens is used as a nominal specifier (27), a pronoun (28)  In (29) the use of gens is close to pas, with the meaning 'at all, absolutely not'. In fact, it can even appear with pas, which it may precede (30). In Contemporary Central Catalan this indefinite maintains its use both as a PI (in the context of a non-veridical operator such as the interrogative or negative operator) and as an NCI (in fragment answers, when an overt no is omitted). Consider (31). 'None.' (Espinal 2002(Espinal , p. 2761 Mai is a temporal expression, the first written evidence of which appears in the 14th century and whose use predominates over other temporal expressions (e.g., anc, jamés, jamai, nunca, null(s) temps, ne(n)gu(n) temps) during the first half of the 16th century. It is used as an NCI already in the 14th century, as the following examples illustrate. On the other hand, a PI use of mai is also common with an existential interpretation, as illustrated in (34) Finally, we consider the locative indefinite enlloc (<Latin IN LOCU)-and its forms in Old Catalan, en loch and enloch. In the 14th century, the positive meaning of a modified bare nominal (36) alternates with a negative meaning in the scope of an overt negative marker (37a). The latter is at the origin of a grammaticalized PI, as exemplified in (37b) As a PI, enlloc requires an overt scope marker in fragment answers already in the 14th century, as illustrated in (38), and today continues its polar use in the form of en no cap lloc (lit. 'in not any place') in the variety of Ports de Morella (see above, ex. (23a)). By contrast, as a preverbal NCI, enlloc licenses other indefinites (PIs and NCIs) in postverbal position, while conveying a single negation interpretation (39a). Furthermore, as an NCI, enlloc occurs alone as a fragment answer, as illustrated in (39b) To sum up, in this section we have argued that, whereas pas in Central Catalan has evolved as an NPI, only licensed nowadays in the context of an overt negative marker, items such as ningú 'anybody, n-body', cap 'a, no', res 'anything, n-thing', gens (de) 'anything (of), n-thing (of)', mai 'ever, n-ever' and enlloc 'anywhere, n-where' come in two sets. In the PI set, the items are characterized by their dependence on a non-veridical operator (including the anti-morphic no 'not'); we associate the above indefinites with PIs whenever an overt negative scope marker shows up preceding any of these indefinite expressions in fragment answers, when an overt negative marker head of NegP in preverbal position licenses an indefinite expression in postverbal position, and when licensed in non-declarative nonnegative contexts with an existential reading. In the NCI set, the items are characterized by the possibility of being self-licensed as negative. In other words, we associate the above indefinites with NCIs whenever these indefinites occur isolated as fragment answers, when they occur in preverbal position and may license single negation in combination with postverbal indefinites (additional PIs or NCIs) without the need for an overt negative marker and when-while in preverbal position-they co-occur with no, the overt Spell-Out of a [neg] feature that is copied in a scope position with respect to TP . Recall that postverbal indefinites in Catalan may correspond either to PIs or to NCIs, since in this language they are not distinguished by stress. 20 If they are PIs, they are licensed by a c-commanding non-veridical operator. If they are NCIs, we assume that the negative feature these items have is first copied in a scope position (with respect to Tense) while the negative feature of the original NCI is deleted; furthermore, the copied feature may be instantiated as no at the time of lexical insertion and NCIs may still move to left-peripheral positions.

Types of PIs in the Trans-Pyrenean Varieties
In this section we present the set of PIs found today in the trans-Pyrenean varieties under study, and we investigate whether they behave like PIs, NPIs or NCIs.
In Section 1 we showed that in Central Catalan gaire 'much, many' is the only item that behaves like a full PI, according to the set of properties identified earlier, namely: (a) the impossibility of occurring alone as negative fragment answers, (b) the impossibility of occurring with no c-commanding licensor, (c) the possibility of being licensed longdistance, (d) the licensing of an existential reading in non-declarative non-negative contexts and (e) the impossibility of contributing to a double negation reading when multiple PIs combine with a c-commanding negative marker. Likewise, the item equivalent to gaire in all the Pyrenean varieties under study (guaire) behaves like a PI, as can be seen in the following samples; replies in (40A) and (40A') illustrate properties (a) and (b), (41) illustrates property (c) and (42)  In what follows we describe the set of items available in the Romance varieties spoken in the Pyrenees on the basis of the set of these properties that characterize full PIs. In Section 4.1 we focus on a group of indefinite pronouns that can be analysed as items of two homophonous but separate PI-NCI sets; we exemplify this case with degun/dengún 'anybody/n-body'. In Section 4.2 we deal with a specific group of originally nominal minimizers which have been reanalysed and incorporated into the negative system of the trans-Pyrenean varieties under study as scalar minimizers (Israel 1995) and are used as PIs or NCIs; we exemplify this with gota (lit. drop). Finally, in Section 4.3 we investigate the properties of a group of PIs that are available as NPIs, but with some additional semantic and pragmatic constraints, similar to pas in Contemporary Catalan.

Indefinite Pronouns: PIs and NCIs
We here consider a set of indefinite pronouns of the trans-Pyrenean Romance varieties under study that include the indefinite cap (Benasquese, Languedocian and Gascon Occitan), nat/nada/nats/nades (Languedocian and Gascon Occitan) 'any, no'; degús (Languedocian Occitan), degun (Languedocian and Gascon Occitan), digun (Gascon Occitan), dengún (Aragonese), degú/digú (Benasquese) 'anybody, n-body'; arrés (Gascon Occitan) 'anybody, n-body'; arrén (Gascon Occitan), res (Languedocian Occitan), cosa (Aragonese) 'anything, nthing'; jamai/jamei (Occitan), mai (Benasquese), nunca (Aragonese) 'ever, n-ever'; and enluòc, enlòc (Languedocian and Gascon Occitan) 'anywhere, n-where' (see Table A2 in Appendix A for a summary of the distribution and meaning of these items). 22 These elements behave like PIs with respect to (a) the impossibility of being licensed unless they appear under the scope of a non-veridical operator (e.g., conditionals, questions, comparatives of inequality, downward entailing operators, the negative operator, etc.), (b) the possibility of being licensed long-distance and (c) the licensing of an existential reading in non-declarative non-negative contexts. 23  PI items of the trans-Pyrenean Romance varieties also have homophonous NCI sets, which can occur either alone as negative fragment answers 24 or in subject or topic position of declarative negative sentences with no c-commanding licensor. This is exemplified in (46) and (48) for dengún 'nobody' in Aragonese, and in (47) and (49)  According to the analysis presented in Section 2, the items dengún/degun in preverbal position in (48) and (49) are considered NCIs because they are the only items contributing a single negation reading to the sentence. In our terms, optional non in (48) should be considered the overt Spell-Out of a [neg] formal feature that adjoins to TP for scope reasons before dengún is moved to a sentence-peripheral position.
In example (50), two preverbal NCIs combine with and license a postverbal PI/NCI in an NC structure in Languedocian Occitan.
(50) Degun jamai demanda res. n-body n-ever asks anything/n-thing 'Nobody ever asks anything.' (Gascon Occitan;BaTelÒc n.d.: Bodon 1964) The reason why this example with two preverbal NCIs yields a single negation interpretation rather than a double negation reading is accounted for in the following way. Let us assume that NCIs are inherently negative (Larrivée 2021), which we interpret as encoding a syntactic [neg] formal feature. In the absence of a negative head out-scoping Tense, when a sentence contains various NCIs, at least one of them (with its [neg] feature) must move from its sentence-internal position to a position from which it takes scope over the event predicated by the verb, thus rendering the sentence negative. In (50), jamai is the NCI responsible for this reading, since it is the constituent that is closest to Tense. In the output configuration following this movement, any other [neg] feature-be it the one corresponding to the postverbal res (if it also corresponds to an NCI) or the left peripheral degun-has no effect on the interpretation of the sentence after an operation of Concord, and therefore does not render a double negation reading .
In contrast to this example, (51) contains an instantiation of the negative operator pas. Recall here the difference between the anti-morphic operator pas in Languedocian Occitan (and Gascon Occitan) and the NPI pas in those Catalan varieties presented in Section 2 (as well as in Aragonese, Benasquese and Aranese Occitan, as discussed in Section 4.3). Given the fact that a sentence such as (51) conveys a single negation meaning, we postulate that degús, jamai, res and enluòc are PIs under the scope of a postverbal negative operator pas that modifies the event denoted by the predicate. For this to take place, the PI degús must be reconstructed to a sentence-internal position at LF (Chomsky 1977 On the PI/NCI distinction, it is of interest to point out that in the Aragonese prescriptive grammar (2021, p. 55) the possibility of using preverbal negative indefinites without a licensing overt negative marker is accepted. However, the use of indefinite pronouns with an overt marker no is recommended 'because this was the historically preferred option in Aragonese.' A similar situation holds for parallel items in Languedocian Occitan, where nowadays jamai 'ever', res 'anything', degun 'anybody', nat/nada or cap 'any' and enlòc 'anywhere' appear in preverbal position either with the overt licensing negative marker pas (which corresponds to their use as PIs) (51), or without it (which corresponds to their use as NCIs) (50).
Some data confirm the status of such elements as PIs in Old Occitan, where they could be licensed by any non-veridical operator, such as for example the conditional operator (52); see Jensen (1986, pp. 176-81), also Medina (1999 When used in preverbal position, it was the former Occitan negative marker no(n) that secured the proper licensing of the PI forms, as stated by Jensen (1986, p. 168 Interestingly, Contemporary Languedocian Occitan offers some evidence in favour of an independent negative quantifier set (parallel to French personne, rien), members of which hold a negative meaning in the absence of the negative marker. Thus, (54) and (55) illustrate that cap and jamai are self-licensed as negative in postverbal position, a possibility that is excluded from PIs and NCIs (see Llop 2017, p. 317 In contrast to Languedocian Occitan, special note should be made of the distinction between PI and NCI sets for indefinites in Gascon Occitan, a language that shows a discontinuous ne/non . . . pas to express unmarked negation (Bernini and Ramat 1996;Romieu and Bianch 2005;Olivieri and Sauzet 2016). 25 In those cases in which degun 'none', arrés 'n-body/n-thing', arren 'n-thing', jamès 'n-ever', cap 'no' or enlòc 'n-where' are placed in preverbal position, the presence of one of the elements of the bipartite discontinuous negation (ne/non . . . pas) is compulsory. Consider first arren 'n-thing' in (56), which illustrates the NCI status of this item, whereas ne/non is conceived as the overt Spell-Out of a [neg] formal feature copied to an adjunct position to TP for scope reasons before arren reaches the final left-peripheral position.  To sum up, according to the analysis of Gascon discontinuous negation in Llop (2017Llop ( , 2020, ne/non is not considered to be the sentential negative marker, whilst pas is. Non/ne is assumed to act as a marker of the scope of negation, and we therefore hold that non/ne is the morphological realization of a morphosyntactic [neg] feature. Thus, the examples in Gascon can be considered to be parallel to those in the other Pyrenean varieties, where the indefinite pronouns are self-licensed NCIs. However, in Gascon varieties in contact with Languedocian Occitan, where indefinite pronouns are combined with a postverbal pas, the head of NegP, these indefinite pronouns must be considered PIs.

Scalar Minimizers: PIs and NCIs
The second set of items in the trans-Pyrenean continuum we are studying is made up of nominal minimizers which are used as either PIs or as NCIs depending on the variety. Minimizers are items originally denoting a minimal part or amount of something. We exclude from our discussion minimizers with overt ni 'not even' which, as we saw for Catalan in Section 2, behave like NCIs. Instead we focus here on bric/brica (lit. shard/shred, broken piece), found in Gascon and Languedocian Occitan; brenca (lit. 'twig'), found in Aragonese, Benasquese and Ribagorçan Catalan); gota (lit. drop), used in Aragonese, Benasquese and in some Catalan varieties, namely in Northern Central Catalan varieties, as well as in North-Western, Ribagorçan and Pallarese Catalan; molla and mica/miaja (lit. crumb), used in Pallarese, Ribagorçan and North-Western Catalan; and pon (lit. point), used by some Aragonese and Benasquese speakers. The etymological origin of these items is presented in Table A1 in the Appendix A.
The salient characteristic of minimizers is that they are scalar items, whose interpretation depends on the availability of a scalar model (Fauconnier 1975), with a set of propositions ordered so as to support inferences between them (Kay 1990). According to Israel (1995, p. 164), minimizers 'mark a phantom minimal element on the scale', which means that, in order to refer, they entail other instances within a set and encode a low quantitative value within the scalar ordering of propositions. At the same time, as emphatic items, they do not obey the default interpretation-that of a minimal element on the scale-but instead pragmatically entail an ordered set of alternatives. Given that they are restricted to scale-reversing contexts, the pragmatic inferences they trigger go from low to high values within the scalar model. 26 Crucially for our study, the difference between minimizers and the indefinite pronouns presented in Section 4.1 is that the latter are not inherently emphatic and, thus, do not denote the existence of an ordered scale along which alternative values are ordered, as scalar minimizers do. For Israel (1995, p. 164), instead, 'indefinites are pure phantoms encoding only an arbitrary instance randomly selected from an array of possible instances'.
We seek to determine whether scalar minimizers in trans-Pyrenean Romance varieties are PIs or NCIs. Consequently, we review how far they conform to the set of properties that identify PIs (listed in Section 1) and summarize their behaviour in the varieties under study. A full list of these items can be seen in Table A3 of Appendix A, although for space reasons we will not be able to give full detail of the properties analysed. 27 Firstly, the data reveal the existence of homophonous PI-NCI sets in both Aragonese and Benasquese for minimizers such as brenca, gota, mica/miaja and pon. As NCIs, these minimizers can be used as negative fragment answers (58), and can appear in the preverbal position of negative sentences (with or without a negative marker) (59). We here exemplify this behaviour with the minimizer gota in Aragonese and Benasquese. In all these contexts the scalar nature of minimizers is inferred from the fact that, by referring to a minimal scalar degree, they trigger the expression of a maximally emphatic proposition (Israel 2001, p. 3 Crucially, the contrast between examples (58)-(59) and (60)-(61) proves the existence in Aragonese and Benasquese of an NCI set for scalar minimizers. As illustrated below, the scalar minimizer NCI set coexists in Aragonese and Benasquese simultaneously with a PI set. This PI use of scalar minimizers is also possible in all the other varieties studied here: all scalar minimizers in trans-Pyrenean Romance varieties contribute a single negation reading when licensed by a preverbal negative marker (62), and can be licensed both long-distance (63) and in non-declarative non-negative contexts (64)-(65). Again, we exemplify these uses with gota in Aragonese. Regarding the licensing conditions of all trans-Pyrenenan Romance scalar PIs, grammaticality judgements regarding an alleged hierarchy from more to less restrictive PI licensing operators (that is anti-morphic ⊂ anti-additive ⊂ downward entailing ⊂ non-veridical) reveal that speakers firmly accept the use of scalar PIs in anti-veridical contexts. The licensing in interrogative and conditional sentences is accepted by the majority of speakers but not all of them. The reasons for this (non)-acceptability seem quite idiosyncratic. 28 Conversely, speakers adamantly deny the licensing of PIs in non-veridical contexts that used to license PIs in standard varieties-mainly comparative sentences as well as free-relative clauses-but no longer do so (see Camus 2007).
All the scalar minimizers considered in this section contribute to a single negation reading when combined with non-scalar PIs under the licensing condition of a preverbal negative marker, as illustrated in (66) Satué 2001, p. 54) To sum up, in this section we have argued that trans-Pyrenean Romance varieties display a set of scalar minimizers that require a model with ordered alternative values. This model predicts that in scale-reversing contexts (including negative contexts) inferences derived from the use of minimizers go from low to high quantitative values, because they are inherently emphatic lexical items. Herein lies the difference with those indefinites studied in Sections 2 and 4.1, which are not inherently emphatic. Crucially for our study, the difference between minimizers and the indefinite pronouns presented in the previous section is that the latter are not inherently emphatic and, thus, do not entail the scalar norm pragmatically. Additionally, even if indefinites trigger entailments over a full set of alternative propositions, they do not need a quantitative scalar ordering of them. Relevant for our purposes is the fact that only in Aragonese and Benasquese are these items available as NCIs, since they can self-license a negative reading in fragment answers and in a leftperipheral position. In all trans-Pyrenean Romance varieties-Aragonese and Benasquese included-scalar minimizers show the distribution of PIs.

NPIs with Enriched Meaning
In this section we focus on a set of items that behave like NPIs, since they require an anti-morphic licensing operator, similarly to what we saw for pas in Central Catalan. In Aragonese and Benasquese this is the case for pas (and brenca, mica and its variants to a lesser extent). In Aranese Occitan and Pallarese and Ribagorçan Catalan this is the case for the NPI cap (and pas to a lesser extent). 29 In Gascon and Languedocian Occitan bric/brica have an NPI version, although the NPI ges is also available. A salient characteristic of this set of items is that, contrary to their PI-NCI homophones and to scalar minimizers, these NPIs introduce additional conventional implicatures that restrict the inferable meanings that these items have. Consider the examples in (67), where the postverbal NPIs are licensed by different negative operators: in (67a-c), for Aragonese, Benasquese, Pallarese and Ribagorçan Catalan, and Aranese Gascon Occitan, the operator is the preverbal no(n), whereas in (67d-e), for Gascon and Languedocian Occitan), the licensor is the postverbal negative operator pas. 30 The minimizers in italics introduce a non-descriptive use of negation, according to which some contextual proposition must be accessible and some conventional implicatures must be inferred. In other words, these examples are associated with some contextually enriched meaning that is absent in the case of indefinite PIs/NCIs and in the case of scalar minimizers.
Of particular relevance to our research on the different uses of originally scalar minimizers is the fact that, as in French, the item pas no longer holds this enriched meaning in any of the Occitan varieties studied here (except for Aranese Gascon Occitan). Today, in Gascon Occitan pas is used as the expression of the logical negative operator, whereas non/ne remains as a marker of the scope of negation. Contemporary Languedocian Occitan (alongside Roussillonese Catalan) displays the last stage of Jespersen's cycle: postverbal pas is used alone as the logical negative operator (see Olivieri and Sauzet 2016, p. 346 for Languedocian; and Gómez 2011 for Roussillonese). By contrast, non was the negative operator in Old Occitan Romieu and Bianch (2005, p. 245). See Schwegler (1988, p. 163), who shows that the use of ne as the sentence negative operator could still be found in Languedocian texts of the 16th century (68) Lafont 1970, p. 88) The first occurrences of pas in Paoli and Bach's (2020) corpus of Old Occitan are from the end of the 12th century, with the 15th century being the moment when the use of pas (with an enriched pragmatic meaning linked to previous discourse) begins to increase exponentially, particularly in rhetorical questions (Paoli and Bach 2020, pp. 119, 124;Paoli 2020Paoli , p. 1022. The combination of pas with PI scalar minimizers is already attested in Occitan during the period between the 13th and 15th centuries (Paoli and Bach 2020, p. 124; see also Jensen 1986, p. 309;Romieu and Bianch 2005, p. 245). These data also support that pas was not used as a scalar minimizer, but as an item with an enriched meaning. According to Paoli (2020Paoli ( , pp. 1039Paoli ( , 1043, pas was used in negative rhetorical questions in Old Occitan to deny a counter-expectation (changing the initial negative question to a positive assertion). This use contributed to pas progressively being associated with 'polarity reversing properties', first at the speech-act level (referring back to discourse-old content) and, later on, at the propositional level. According to Paoli (2020Paoli ( , p. 1043, this process would have led to the use of pas as the sentential negator. Olivieri and Sauzet (2016, p. 346) assert that unmarked discontinuous negation emerged in Occitan varieties in the 17th century (see also Schwegler (1988, p. 308). At the same time, however, instances of pas as a negative operator can already be found in some 17th century Languedocian Occitan texts (69). 31 (69) Debe pas ieu prendre une corde? Should not I take a rope 'Shouldn't I take a rope?' (Old Occitan, 17th century; example from Pansier 1973, p. 262 [apud Schwegler 1988 In the 19th century the use of the postverbal marker becomes widespread as the unmarked strategy for negative declaratives in Languedocian Occitan. Note, though, that the switch to a single postverbal negation does not seem to have occurred uniformly in Languedocian Occitan, given that several instances of an unmarked discontinuous negation have been identified in Languedocian texts from the 19th century (Schwegler 1988, p. 308 and 19th century texts in BaTelÒc). Paoli andBach (2020, p. 119, following Schwegler 1988, p. 163ff.) specify that the co-occurrence of pas with the preverbal negative marker was largely restricted to the written language in that variety in the 19th century. 32 Overall, despite the difficulties involved in determining a clear chronology for the evolution of sentential negation in the varieties studied here, the data suggest that in Gascon and Languedocian Occitan the minimizer pas has undergone a further reanalysis process: not only does the literature on Old Gascon and Languedocian Occitan present diachronic evidence of its use as an NPI with an enriched pragmatic meaning, but we also have evidence of the use of pas as an anti-morphic operator capable of licensing other PIs.

Conclusions
In this article we have described the set of items available in Catalan and other trans-Pyrenean Romance languages, namely Aragonese, Benasquese and Gascon and Languedocian Occitan, which behave like PIs, NPIs and NCIs.
With respect to the question of whether, besides gaire, there are other items that share the properties attributed to PIs, we have argued that in Catalan indefinite pronouns ningú, cap, res, gens, mai and enlloc actually constitute two homophonous but distinct PI-NCI sets. On the one hand, we have shown that the PI variants of such indefinites appear when they are used as fragment answers preceded by an overt negative scope marker, as well as when they occur licensed by a non-veridical operator, one possibility being an overt negative scope marker. On the other hand, the counterpart NCI set corresponds to the self-licensed negative indefinites that occur isolated as fragment answers, as well as in preverbal position in declarative negative sentences. We have shown that minimizers introduced by ni 'not even' behave like NCIs, while minimizers not preceded by ni behave like PIs. Moreover, pas has been shown to behave like an NPI, only licensed in the context of an overt negative marker.
Concerning the question of whether the distinction between PIs, NPIs and NCIs appears to be relevant at all in Catalan, we have argued-on the basis of data from Contemporary Catalan (as well as additional diachronic evidence)-that NPIs have emerged from PIs that are semantically restricted to the scope of anti-morphic operators, while NCIs have emerged from PIs by being syntactically restricted by a [neg] formal feature. We therefore conclude that the distinction between NPIs and NCIs is best conceived as referring to two different subclasses of PIs, the former semantically dependent on an anti-morphic operator, the latter syntactically constrained by an Agreement relationship.
As for which PIs are present in the other Romance varieties spoken in the Pyrenees, we have demonstrated the availability of synchronic homophonous PI-NCI sets of indefinite pronouns and specifiers. Of particular interest to our research are data from Languedocian Occitan showing that when indefinite pronouns combine in this variety with postverbal pas, the head of NegP, they are not NCIs but must instead be considered PIs, under the scope of an anti-morphic operator. Moreover, Contemporary Languedocian Occitan offers some evidence for the emergence of negative quantifiers that license a negative reading for the whole sentence in the absence of a preverbal negative marker or a [neg] feature.
As for the characterization of scalar minimizers in trans-Pyrenean Romance varieties, we have shown that they refer to a minimal scalar degree and trigger the expression of a maximally emphatic proposition. Scalar minimizers have also been proved to appear in homophonous PI-NCI sets in Aragonese and Benasquese, but not in Catalan and Occitan varieties, where they behave like PIs. Interestingly, we have further shown that some of these PI scalar minimizers have homophonous items without degree, which have been proven to be NPIs whose meaning can be enriched depending on the context of utterance.
Concerning the diachrony of NPIs in trans-Pyrenean Romance varieties, it can be concluded that the change observed in all the varieties studied implies the transition from a broad semantic dependency to a narrow semantic dependency (i.e., items with an enriched pragmatic meaning are used in all trans-Pyrenean Romance varieties, exclusively licensed by anti-morphic operators). We have presented additional evidence to illustrate the different status of pas across these Romance varieties, not only as an NPI at stage II of Jespersen's cycle that further constrains conventional implicatures, but as a negative operator that licenses the presence of PIs/NPIs/NCIs.
Beyond the languages studied here, this paper contributes to a better understanding of the distribution and meaning of PIs, NPIs and NCIs. In particular, the overview provided here of the polar items identified in Catalan and other trans-Pyrenean Romance languages sheds considerable light on the types of PIs available in natural languages and the likely tendencies in their diachronic evolution.  Let us assume the notion of (non-)veridicality postulated by Zwarts (1995, p. 287):  Giannakidou (1998) in substituting Zwarts's (1995) a-veridicality for anti-veridicality, since the meaning intended is 'opposite to veridicality' not 'without veridicality properties'. 3 We thus use the terms PIs and NPIs instead of weak and strong NPIs (Ladusaw 1992(Ladusaw , 1996van der Wouden 1994). 4 Notice that gaire can occur as a fragment answer and in preverbal position in declarative sentences only if it appears in the immediate scope of a c-commanding licensor such as the negative marker no, as shown in (i). These examples contrast with (1aR) and (1b) Giannakidou (1997Giannakidou ( , 2000 postulates that NC languages are either Strict or Non-Strict, which differ with respect to the possibility of having a negative subject followed by a negative marker in a single negation reading. Strict NC languages (e.g., Greek, Russian) allow such structures. Non-Strict NC languages (e.g., Italian, Portuguese) do not. See footnote 10 for a reference to Catalan. 6 We acknowledge that it is unclear whether the possibility of licensing a double negation reading is a property of items participating in NC structures (as suggested for Strict NC languages such as Hungarian and Romanian; see Puskás 2012; and Fălăuş and Nicolae 2016, respectively) or rather a property of emerging negative quantifiers (as suggested for a Non-Strict NC language such as Catalan; see Déprez et al. 2015). 7 Central Catalan is the dialect of Catalan with greatest demographic weight, since it is spoken in the whole province of Barcelona, half of Tarragona province and most of Girona province. 8 See (Déprez et al. 2015) for an experimental investigation that supports the existence of an emerging negative quantifier set among Catalan university students for sequences with indefinite DPs and indefinite pronouns in subject position. These indefinites in combination with an overt negative marker lead to a non-negligible double negation reading, as one of the possible English translations suggests.
(i) a. Cap dels alumnes no llegeix cap llibre. no of.the students not reads no book 'None of the students read any books' and 'None of the students don't read any books at all' (i.e., 'All the students read at least a few books.') b. Ningú no neteja alguna cosa. n-body not cleans something 'Nobody cleans something' and 'Nobody is not cleaning something' (i.e., 'Everybody is cleaning something.') 9 Note that two NCIs may also occur preverbally, conveying a single negation reading: (i) Des de l'atemptat, mai res serà com abans. since the.attack n-ever n-thing be.FUT as before 'Since the terrorist attack, nothing will ever be as it used to be.' Since we assume that NCIs have a formal [neg] feature, without being negative quantifiers, a single [neg] feature immediately c-commanding Tense is the only item responsible for the single negation reading. All remaining potential [neg] features that may occur in a sentence are removed by means of an operation of Concord ). We will come back to this issue in Section 4.1. 10 The Catalan variety spoken in Roussillon is the only one that uses pas as the expression of the logical negative operator (Gómez 2011), as is also the case in Occitan (Alibèrt 1976) and French (Grevisse and Goosse 2007). See Section 4.3. Notice that for the varieties where pas is an NPI, we translate the sentences where it appears as sentences with a standard negation. We set aside for the moment the enriched meaning its use conveys, depending on the context of utterance, but will explore this issue in Section 4.3.

11
Central Catalan differs from Italian, Portuguese and Spanish in the fact that in the context of preverbal NCIs the negative marker no 'not' appears to be optional (Fabra 1956;Espinal 2002).
(i) Ningú (no) porta maleta. nobody not carries suitcase 'Nobody carries a suitcase. ' Zeijlstra (2004 and ff.) argues for the existence of two different dialects: a Non-Strict NC one (without no) and a Strict one (with no). However, this hypothesis has been strongly argued against in the literature (Espinal and Tubau 2016;Tubau et al. 2018), given that: (a) sequences with the structure NCI + V + NCI without no are accepted by all speakers; (b) sequences with the structure NCI + no + V + NCI are also accepted by all speakers; and (c) double negation in Catalan may also arise in NCI + no sequences-unlike what we see in languages like Romanian, where double negation is only possible in NCI + nu + NCI sequences (Fălăuş 2007;Fălăuş and Nicolae 2016). In Espinal et al. (2021) it is argued that the apparent optionality of no in (i) is better analysed by appealing to two different analyses: one in which the pre-sentential NCI ningú self-licenses (Ladusaw 1992) a negative reading by c-commanding Tense, and another one in which the pre-sentential NCI ningú combines with no, which is not the head of NegP but is merely the overt Spell-Out of a copied [neg] feature (in a scope position) in the movement of ningú to the left-periphery of the sentence. 12 However, as already pointed out by Larrivée (2011, p. 4) for French, Llop (2017, pp. 189-92) for Catalan and Paoli and Bach (2020, pp. 119-20) for Occitan, this hypothesis (i.e., the use of pas as a measure phrase and its reanalysis as an NPI) appears to be speculative for the languages studied because-as pointed out by a reviewer-that move is not unambiguously attested in the earliest textual base. 13 Although there appear not to be diachronic examples that license pas in non-veridical contexts beyond the anti-morphic operator (Pérez- Saldanya, p.c.), this possibility has been attested in Catalan for other minimizers such as gota 'drop'. See Section 4.2. 14 It should be noted that this pre-negative distribution is characteristic of languages such as Basque and Hindi that have been described in the literature as containing only PIs (Etxepare 2003;Etxeberria et al. 2021;Lahiri 1998 Given that the interpretation of utterances varies depending on the context, the translations for the examples in this section do not reflect the enriched pragmatic meaning associated with pas. According to the authors of this article, depending on the contextual information available, the pragmatic meaning of pas, in addition to rejecting a positive accessible proposition, may also consist in strengthening a negative proposition. See Espinal (1993) for details. 16 It is necessary to point out that those NPIs that are exclusively licensed by the negative marker are a very small class in natural languages and show a tendency towards idiomaticity (e.g., English lift a finger; Catalan obrir boca lit. open mouth 'say a word'). See Sedivy (1990), Postal (2004), Sailer (2021) and others. 17 See Zeijlstra (2004Zeijlstra ( , 2012 for the hypothesis that NCIs carry a syntactic [uNeg] feature that probes for a goal with a matching [iNeg] feature with which to Agree. In contexts where the goal (i.e., an anti-morphic operator) is not overt, the [uNeg] feature is assumed to trigger an abstract [iNeg] negative operator as a Last Resort operation. We follow here Espinal et al. (2021) in assuming that NCIs are inherently specified with a syntactic [neg] feature that can render the sentence negative provided it is in a relevant position scoping over TP (Acquaviva 1995;Ladusaw 1996;Zanuttini 1997;Herburger 2001;Penka 2011). According to this analysis, the presence of [neg] in fragment NCIs self-licenses a negative reading (Larrivée 2021); that is, this syntactic formal feature restricts the choice of possible non-veridical operators into a subset of them, with the addition that by being a syntactic feature it must be checked at syntax. When the NCI superficially occurs either in a postverbal or in a preverbal position, a feature-copying mechanism (Chomsky 1995(Chomsky , 1998Watanabe 2004) applies by which a [neg]-chain is formed. This [neg]-chain is solved by eliminating the original [neg] feature of the NCI in sentence-internal position and keeping the highest one (in a scope position). The Vocabulary Item no that combines with postverbal or preverbal NCIs, under this approach, is not the head of NegP but rather the Spell-Out of a disembodied [neg] feature moved to TP for scope reasons. 18 Pérez Saldanya and Torrent (2021, footnote 24) point out that in another 15th century text, Vita Christi, out of 116 cases of a preposed nengú/-un, only three occur without an overt negative marker, again suggesting that it is not until that century that NCI uses of this form begin to emerge in written texts. 19 Although the Catalan spoken in Ports de Morella is not a variety of Central Catalan, what the examples in (23A) and (24A) show is that in Contemporary Catalan, PI readings for cap 'any' and ningú 'anybody' occur simultaneously with NCI readings, as illustrated by the well-formed answers in (iA)  Greek shows the paradigm of non-emphatic PIs side by side with emphatic NCIs. See Giannakidou (1997 and future work).
tipota/TIPOTA 'anything'/'n-thing' 21 For a fine-grained analysis and contrastive description of the distribution and meaning of guaire in Aragonese, Gascon Occitan and Catalan, see Tomás (2016, pp. 128-35). 22 We set aside for the moment the item gens and its equivalents in trans-Pyrenean Romance varieties since we comment on them in Section 4.2 when referring to those items derived from nominal minimizers (and generalizers) that must be considered scalar items. See Table A1 in Appendix A for the etymological origin of PIs used in Catalan and other trans-Pyrenean Romance varieties.
For the licensing of such elements in non-veridical contexts in the varieties studied, see Medina (1999) for Old Occitan and Estudio de Filología Aragonesa (2021, pp. 52-56) for Aragonese. See also Tomás (2016, pp. 61-141) for a comparative study of Aragonese, Gascon Occitan and Catalan. 24 The use of the PI dengún 'anybody' as a fragment answer and licensed by the overt negative marker no (i.e., no dengún lit. not anybody 'nobody') is widespread in Aragonese, side by side with the NCI dengún 'n-body' illustrated in (46) The second element in this discontinuous negation is cap in the region of Couserans and Commenge (Eastern Pyrenean Gascon in contact with Pallarese Catalan), according to Bec (1968, § 102). 26 For Chierchia (2006), the fact that the interpretation of the minimizer corresponds to the pragmatically strongest alternative in the context is secured by the proper licensing of a scalar semantic feature [+σ]. In Tubau (2015) this interpretation is attributed to the presence of a Focus even particle in the structure of minimizers. See also Chierchia (2013) for the hypothesis that nonemphatic polarity items are exhaustified under silent ONLY operators, while emphatic minimizers are exhaustified under silent EVEN operators. 27 We claim that a similar distribution to scalar minimizers (as either PIs or NCIs) has been observed for the generalizer ge(n)s (lit. kind), used in Catalan and Occitan (but not in Aranese or Benasquese), since this item shows similar properties to those minimizers we are discussing here. See Table A3. Generalizers are elements denoting a maximally general type or class which contribute to a qualitative negation, 'by extending its scope to include everything in that maximal sortal domain' (Condoravdi and Kiparsky 2006, 1.2). 28 Younger speakers exhibit a more restricted use of PIs in non-veridical contexts and tend instead to use their positive counterpart (see Llop 2017, pp. 171-86). This tendency is in line with Martins (2000) hypothesis that the loss of polar versatility (i.e., PIs initially used in non-veridical contexts are progressively restricted to anti-veridical ones) is associated with diachronic change. 29 Here we will only be referring to cap as an NPI-which in Pallarese and Ribagorçan Catalan as well as in Aranese Occitan is used analogously to the element pas in Central Catalan (see Section 2), but whose characteristics are different from those of the PI/NCI indefinite pronoun cap 'any, no' discussed earlier. 30 The reader must bear in mind that the use of pas, ges and bric(a) as NPIs with enriched meanings is much more widespread than the use of brenca, molla and mica, which mainly correspond to scalar minimizers (see Llop 2017, 4.2.1.1). See Ledgeway (2017), who distinguishes between intensive and presuppositional emphatic negation to differentiate between the scalar and non-scalar (but pragmatically enriched) reading of such items. In Central Catalan this distinction can be exemplified by the respective use of gens (i) (intensive emphatic negation, i.e., scalar minimizer) and pas (ii) (presuppositional emphatic negation, i.e., NPI with enriched meaning). As pointed out by a reviewer, example (68) is an instance of a rhetorical question, which according to Paoli (2020) was instrumental in pas establishing itself as a negative operator. 32 Gómez (2011, pp. 299-300) describes a coincident process for the emergence of pas as the sentential negative marker in Roussillonese Catalan, but does not specify its chronology.