This is the accepted version of the journal article: Espinal, M. Teresa; Tubau, Susagna. «Interpreting argumental n-words as answers to negative wh-questions». Lingua, Vol. 177 (2016), p. 41-59. 19 pàg. DOI 10.1016/j.lingua.2015.12.013 This version is available at https://ddd.uab.cat/record/287758 under the terms of the license 1 Interpreting argumental n-words as answers to negative wh-questions ∗ 2 M.Teresa Espinal and Susagna Tubau (UAB). Accepted in Lingua, Dec. 2015 3 4 Abstract 5 This paper aims to provide an explanation of the lexical characterization and final 6 semantic interpretation associated with isolated argumental n-words in Question- 7 Answer pairs in Negative Concord languages, namely Catalan and Spanish. We 8 argue that there are two competing lexical variants of n-words in these languages: a 9 polarity variant and a negative existential quantifier variant. Accessibility to these 10 two lexical characterizations of n-words is correlated with one of the two possible 11 final interpretations of isolated argumental n-words when used as fragment answers 12 to negative wh-questions. Following a Structured Meaning approach to the 13 semantics of Question-Answer pairs, we present a new analysis of n-words as focus 14 constituents with respect to background wh-questions according to which a final 15 single negation reading can only be inferred from n-words conceived as indefinite 16 polarity items, whereas a Double Negation reading is inferred from negative 17 quantifiers. 18 19 Keywords: argumental n-words, indefinite polarity items, indefinite negative 20 quantifiers, Question-Answer pairs, Catalan, Spanish 21 22 1. Fragment answers, compositionality and ellipsis ∗ This research has been funded by two research grants awarded by the Spanish Ministerio de Ciencia e Innovación (FFI2011-23356, FFI2014-52015-P), and a grant awarded by the Generalitat de Catalunya to the Centre de Linguística Teòrica (2014SGR-1013). The first author also acknowledges an ICREA Academia award. We would like to thank O.Borik, M.Labelle, L.McNally and X.Villalba for their comments to previous versions of this manuscript, and V.Déprez for discussion on related issues. We are also grateful to the reviewers and editor of Lingua for their helpful comments and insightful questions. 1 23 In this paper we investigate why in Negative Concord (NC) languages such as 24 Catalan and Spanish isolated argumental n-words (e.g., Spanish nadie ‘nobody’, 25 Catalan res ‘nothing’) (Laka 1990) may be interpreted as conveying both single 26 negation and Double Negation (DN) when used as fragment answers to negative 27 wh-questions. To contextualise this puzzle, in Section 1.1 we briefly introduce the 28 behaviour of negative quantifiers (e.g., nobody, nothing) in this same context in 29 languages like Standard English and German, and show that they are expected to 30 yield only a DN interpretation. By contrast, in Section 1.2 we present the 31 empirically and theoretically challenging fact that Catalan and Spanish isolated n- 32 words can yield both a single negation reading and a DN reading, a property that is 33 not expected and, furthermore, is not explained under current syntactic and 34 semantic approaches to the distribution of n-words and the interpretation of NC 35 readings. 36 The rest of the article is organised as follows. In Section 2, a number of 37 theoretical assumptions about the lexical and syntactic status of argumental n- 38 words in Catalan and Spanish are introduced. In Section 3, we show that neither the 39 combination of a formal analysis of NC as syntactic Agree (Zeijlstra 2004 and ff.) 40 with an ellipsis analysis of fragment answers (Merchant 2001, 2004), nor a 41 semantic ellipsis account (Giannakidou 2000, 2006) can accommodate the two 42 potential interpretations that Catalan and Spanish n-words may have when used as 43 fragment answers to negative wh-questions. In Section 4, we offer a new analysis 44 within a Structured Meaning approach (von Stechow 1991; Krifka 2001, 2004, 45 2007, 2011) that allows us to derive both the single negation and the DN reading 46 that Catalan and Spanish isolated argumental n-words used as answers to negative 47 wh-questions may have. Finally, Section 5 concludes the paper. 2 48 49 1.1. Negative quantifiers in DN languages 50 In DN languages such as Standard English, German, Dutch and Afrikaans, when 51 negative indefinites such as nobody or nothing and their equivalents in the 52 aforementioned languages are used in isolation as answers to a negative wh- 53 question, a DN reading obtains, as shown in (1) and (2). Notice that the negative 54 questions in (1) and (2) are biased in the sense that they require non-neutral 55 contexts: a contrast set (of individuals who indeed did the homework, or of things 56 they read) needs to be established in order to make them felicitous questions (cf. 57 Han 1999, Romero and Han 2004, Asher and Reese 2005, Reese 2006). On the 58 other hand, these negative questions are felicitous if there is compelling contextual 59 evidence against p, in the sense that, in (1), not everybody did the homework (¬∀), 60 that is, that someone didn’t, or, in (2), they have not read everything, that is, 61 something was not read. 62 (1) Q: Who didn’t do the homework? (English) 63 A: Nobody. (= Nobody didn’t do the homework; 64 ! Everybody did the homework) 65 (2) Q: Was haben sie nicht gelesen? (German) 66 what have they not read 67 ‘What didn’t they read?’ 68 A: Nichts. 69 nothing 70 ‘Nothing’ (= They didn’t read nothing; 71 ! They read everything) 3 72 Negative indefinites such as nobody and nichts are commonly assumed to 73 contribute negation on their own in so-called DN languages, but this idea has been 74 implemented in various ways within the generative tradition.1 In the ’90s some 75 well-established proposals analysed these lexical items as negative quantifiers 76 (Zanuttini 1991; Haegeman and Zanuttini 1991, 1996; Haegeman 1995; Haegeman 77 and Lohndal 2010). In more recent analyses they are argued to be either non- 78 negative indefinites associated with a licensing abstract negative operator (Op¬) 79 that carries an interpretable negative feature [iNEG] (Penka 2011; Penka and 80 Zeiljstra 2005, 2010; Zeijlstra 2011), or inherently negative words bearing an 81 [iNEG] feature (Biberauer and Zeijlstra 2012). Discussing which position is 82 superior is beyond the scope of this paper. Rather, it will be assumed that negative 83 indefinites in DN languages always carry one instance of logical negation (¬). 84 Within the generative tradition the analysis of the answers in (1) and (2) has 85 been claimed to support an ellipsis account of fragment answers, according to 86 which part of the question is copied into the syntactic structure corresponding to the 87 answer. Since sentential negative markers in English and German (not/n’t and 88 nicht, respectively) are also assumed to introduce an instance of logical negation, 89 the copy operation depicted in the answers is predicted to result in DN (since the 90 negative quantifier takes wide scope over negation) and is hence expected to 91 receive a positive reading (¬∃¬ ⇒ ∀). This is indeed the case, as indicated by the 92 paraphrases included in parentheses in (1) and (2). 93 In the Minimalist Program (Chomsky 1995, 2001 and ff.), fragment answers 94 with a negative quantifier seem to be necessarily derived by means of ellipsis 95 (Merchant 2001, 2004). This analysis postulates that the negative indefinite moves 1 See Longobardi (2014, and previous work) for a criticism of the typological macroparametric distinction between DN and NC languages. 4 96 at narrow syntax to Spec, Foc(us) P(hrase) and that, finally, part of the structure is 97 PF-deleted. This movement-then-deletion approach is illustrated –although 98 somewhat simplified– in (3a, b), where the copied material and ellipsis are 99 indicated by the square brackets and the strikethrough, respectively. According to 100 this analysis of the answers in (1b) and (2b), represented in (3), a focused 101 constituent is assumed to have syntactically moved to the left periphery of the 102 clause, to a position above the c-command domain of the ellipsis-licensing head. 103 [E] refers to a formal feature that Foc0 carries in an elliptical structure. It is read at 104 the PF interface in such a way that the whole syntactic structure dominated by it is 105 elided. 106 (3) a. Q: Who didn’t do the homework? 107 A: [FocP nobodyi [E] [TP ti didn’t do the homework]] ‘Everybody.’ 108 b. Q: Was haben sie nicht gelesen? 109 A: [FocP nichtsi [E] [TP haben sie nicht gelesen ti]] ‘Everything.’ 110 Within such an ellipsis account, a pronounced fragment is an instance of 111 clausal ellipsis, and the meaning of nobody and nichts in combination with a copied 112 sentential negative marker contributes compositionally to a DN reading, as it 113 naturally follows from the lexical and syntactic properties of these elliptical 114 structures.2 At LF, the syntax of the answer is assumed to correspond strictly to the 115 syntax of the negative wh-question and the principle of Compositionality is 2 It should be noted, however, that this type of analysis can only account for those languages for which narrow syntactic movement to Focus can be postulated. We thank C. Poletto (p.c.) for this comment. See Weir (2014) for discussion of the hypothesis that fragments do move to Focus, but only at the PF component, since they are interpreted in situ. 5 116 believed to govern the interpretation of the answer. At PF, the material in the scope 117 of the feature [E] is PF-deleted and, hence, not phonologically realised.3 118 In so-called DN languages it is also possible to object to the negative 119 assumption (i.e., in (1) the presupposition Somebody didn’t do the homework) 120 activated by the negative wh-question by means of a different kind of ellipsis, the 121 target meaning being single negation rather than DN. To convey this meaning, 122 these languages resort not to clausal ellipsis, but rather either to full sentences, as in 123 (4Aa) and (5A), or, at most, to sentences with VP-ellipsis only, as in (4Ab).4 124 (4) Q: Who didn’t do the homework? (English) 125 A: a. Nobody did it. 126 b. Nobody did. 127 (5) Q: Wer hat die Hausaufgabe nicht gemacht? (German) 128 who has the homework not made 129 ‘Who didn’t do the homework?’ 130 A: Niemand hat sie gemacht. 131 nobody has she made 132 ‘Nobody did it.’ 133 The answers in (4A) and (5A) do not involve clausal ellipsis and therefore no 134 material is copied from the question onto the syntactic structure of the answer. The 3 This classical movement-then-deletion account of ellipsis is in line with theories in which PF “spells out” LF, that is, theories that introduce constraints on a particular type of correspondence between LF and PF representations: LF (broadly construed) is calculated first and determines PF (surface word order), and therefore scope at LF is matched by precedence at PF (Bobalijk 1995, 2002; Bobalijk and Wurmbrand 2012; a. o.). 4 We thank G. Kaiser, B. Gherke and K. Hartmann (p.c.) for sharing with us their intuitions on the German data. It seems that in this language prosodic stress on both the wh-word and the negative marker of the question favours a DN reading in the answer. This observation is interesting because it supports the claim made for many other languages that any explanation of DN must involve both prosody and syntax. See, among others, Corblin (1995, 1996) and Vinet (1998) for French; Corblin and Tovena (2003) for French and Italian; Zanuttini (1991, 1997), Godard and Marandin (2007) for Italian; Baltazani (2006) for Greek; Molnár (1998), Puskás (2006) for Hungarian; Huddlestone (2010) for Afrikaans; Tomioka (2010) for Japanese; Espinal and Prieto (2011) and Espinal at al. (2015) for Catalan. 6 135 structure corresponding to these answers excludes the possibility of obtaining a DN 136 reading, while allowing the answer to express single negation, as conveyed by the 137 logical negation of the negative quantifier. It seems, therefore, that so-called DN 138 languages tend to avoid clausal ellipsis when the intended interpretation of the 139 negative indefinite does not match the one that results from the syntax of fragment 140 answers, which is necessarily DN if the wh-question is negative. 141 Contrary to what happens in Standard English and German, in NC languages 142 (e.g., Catalan, Spanish) answering a negative wh-question with an isolated n-word 143 mostly results in NC, but the sentences do not show VP-ellipsis. In the next section 144 we will show that the possibility of obtaining a single negation reading, which 145 corresponds to the default interpretation for a population of native speakers, calls 146 into question not only the analysis that must be attributed to n-words, but also the 147 analysis of argumental n-words as fragment answers. The reason for this claim is 148 that if isolated n-words were considered negative quantifiers in all natural 149 languages, they should only license a DN reading under a standard clausal ellipsis 150 account. 151 152 1.2. N-words as answers to negative wh-questions in Catalan and Spanish 153 Argumental n-words such as Spanish nadie (‘nobody’) and nada (‘nothing’) and 154 Catalan ningú (‘nobody’) and res (‘nothing’) are most commonly interpreted as 155 conveying single negation. This is illustrated in (6) and (7) for Spanish.5 156 (6) Q: ¿Quién no llevaba gafas? 6 (Spanish) 5 In French, by contrast, the most natural interpretation for personne and rien in similar contexts is DN, which suggests that these expressions are not the same type of n-word we find in Spanish and Catalan. We thank M. Labelle (p.c.) for pointing out this contrast to us. 6 Interestingly, as pointed out to us by a native speaker, if special prominence is placed on the sentential negative marker in the question, as indicated by the capitals in ¿Quién NO llevaba gafas? (lit. who not wore glasses ‘Who was not wearing glasses?’), a DN reading for the reply Nadie is 7 157 who not wore glasses 158 ‘Who wasn’t wearing glasses?’ 159 A: Nadie. 160 nobody (= Nobody was wearing glasses) 161 (7) Q: ¿Qué no han leído los estudiantes? (Spanish) 162 what not have read the students 163 ‘What didn’t the students read?’ 164 A: Nada. 165 nothing (= The students didn’t read anything) 166 In relation to these kind of data composed of a negative wh-question and an n- 167 word isolated answer, we refer the reader to a number of perception experimental 168 studies run with native speakers of Catalan and Spanish (Espinal and Prieto 2011, 169 Prieto et al. 2013, Espinal et al. 2015) that aimed to foster on our knowledge of the 170 interaction between syntax and prosody. In these studies participants had to indicate 171 whether they interpreted an n-word in the answer to a negative wh-question as 172 expressing either a single negation reading (i.e., ‘nobody’ / ‘nothing’) or a DN 173 reading (i.e., ‘everybody’ / ‘everything’). Specifically, the result we would like to 174 focus on in this paper, as it is both empirically and theoretically challenging, is that, 175 when the intonation contour of the isolated answer was unmarked (i.e., had a fall 176 boundary tone, also described as L+H*L% in Cat_ToBI and Sp_ToBI), participants 177 associated the argumental n-word with a single negation reading only 57.5% of the 178 time in the case of Catalan speakers, and 66% of the time in the case of Spanish 179 speakers, not 100% of the processed items, as we would have expected in NC favoured. We believe that DN is inferred in this case as the output of the following PF-LF interaction: stress on the sentential negative marker in the question gives the instruction to copy the negative clause in the question onto the answer, thus resulting in DN. 8 180 languages in which n-words are indefinite expressions. 7 Furthermore, the 181 proportion of DN responses in the interpretation of Catalan ningú/res and Spanish 182 nadie/nada is surprising, as single negation is the only possible interpretation that is 183 in accordance with the description of n-words in traditional/prescriptive grammars 184 and in some descriptive/theoretical studies for Catalan (cf. Fabra 1956; Solà 1973; 185 Vallduví 1994; Espinal 2000, 2002) and for Spanish (cf. Bosque 1980, Sánchez 186 1999, RAE 2009). Hence, the aim of the present article is to provide an analysis of 187 isolated argumental n-words in these languages that accounts for the fact that, in the 188 absence of a marked prosodic contour (that is, in the absence of any linguistic 189 trigger of a denial interpretation), a compositionally-driven DN interpretation is not 190 discarded in either of these NC languages. 191 192 2. Theoretical assumptions 7 See Espinal et al. (2015) for a full discussion of the experimental conditions and statistical significance of the mean perceived DN of the results we refer to. Participants had to rate two target n-words (ningú/res in Catalan and nadie/nada in Spanish) produced in a Q-A discourse context by two pairs of subjects as meaning either ‘nobody/nothing’ or ‘everybody/everything’. The hypotheses tested in this paper were that n-words associated with two distinct intonation contours, either a rise- fall pitch contour consisting of a rising pitch accent followed by a final fall boundary tone (L+H* L%) or a rising pitch accent associated with the stressed syllable followed by a low-rising boundary tone (L+H* L!H%), both in isolation and in preverbal position, had different interpretations and that the latter intonation was responsible for triggering an increase in DN interpretations. Examples of the sort of data that Catalan and Spanish speakers were presented with are given in (i) and (ii), respectively. (i) Q: Qui no ha menjat postres? who not has eaten dessert ‘Who has not eaten dessert?’ A: Ningú. nobody (ii) Q. A fecha de hoy ¿qué no nos han mandado? at date of today what not us have sent ‘As of today, what haven’t they sent us?’ A. Nada. nothing See also Espinal and Prieto (2011), Prieto et al. (2013) and Espinal et al. (2015) for arguments in support of the correlation between a marked prosody (and gesture) and a marked interpretation such as denial, and Horn (1989) and Geurts (1998), for general discussion on metalinguistic negation and mechanisms of denial. 9 193 In the upcoming sections, a number of theoretical assumptions about the nature of 194 n-words are introduced. Likewise, we consider the syntactic and semantic status of 195 isolated n-words when they occur as answers to wh-questions. 196 A new analysis for interpreting isolated argumental n-words in the Romance 197 languages under study in this paper is based on the assumption that these items 198 come in two lexical variants. One variant of these items consists of indefinite 199 expressions characterized semantically with a polarity feature, a variant we will 200 refer to as n-words1. A competing variant (Kroch 2000) for n-words is variably 201 available, and is characterized semantically as a negative existential quantifier; we 202 will refer to these items as n-words2.8 203 This lexical characterization is related to the fact that, for a population of 204 Catalan and Spanish native speakers, n-words are basically assumed to be PIs that 205 nevertheless can be associated with an uninterpretable negative formal feature in 206 the syntax to guarantee a single negation or NC reading (by means of an Agree- 207 chain with an interpretable matching feature), no matter whether they occur in 208 preverbal or postverbal position. However, for another population of native 209 speakers of Catalan and Spanish, n-words can also be increasingly characterized as 210 indefinite negative quantifiers, that is, as inherently negative words that do not 211 participate in NC structures, which may license a DN reading. This is argued in 212 Section 2.1. Later on, in Section 4 we will relate these two variants to either the 8 See Herburger (2001) for an antecedent of the hypothesis that n-words in Spanish are lexically ambiguous between polarity items and inherent negative expressions (referred to as negative elements). She assumes that while polarity n-words are non-negative existential quantifiers that need to be licensed in downward entailing contexts, n-words in answers to wh-questions are negative elements that are inherently negative and create downward entailing contexts. Herburger’s account predicts that, when used as answers to negative wh-questions, isolated n-words in Spanish will be interpreted as conveying a DN reading. See Déprez et al. (2015) for an experimental investigation of Catalan n-words that provides empirical independent support for the ambiguous status of these lexical items. 10 213 single negation or the DN interpretation associated with isolated argumental n- 214 words when used as answers to negative wh-questions. 215 Furthermore, n-words that occur as answers to questions are assumed to sit in a 216 syntactic Focus position, regardless of whether they license a NC or a DN 217 interpretation. This is argued for in Section 2.2 by evaluating the behaviour of 218 preverbal n-words –which isolated n-words resemble most– with respect to their 219 distribution, as well as left periphery and information structure considerations. 220 221 2.1. N-words as indefinite polarity items and as indefinite negative quantifiers 222 As has been extensively discussed in the literature, n-words in NC languages may 223 display an ambivalent syntactic behaviour: they have been claimed to behave like 224 negative quantifiers when occurring preverbally and in isolation, but as polarity 225 items when occurring postverbally (Laka 1990; Ladusaw 1992; Déprez 1997 and 226 ff.; Giannakidou 1998, 1999; Herburger 2001; Zeijlstra 2004; a. o.). That is, n- 227 words in so-called NC languages can be in the scope of the sentential negative 228 marker without yielding DN. 229 Following the microparametric approach developed in Labelle and Espinal 230 (2014), our lexical characterization of n-words in Catalan and Spanish relies on a 231 semantic feature, Chierchia’s (2006: 559) [+σ], and a syntactic feature, Zeijlstra’s 232 (2004 and ff.) [uNEG]. While the [+σ] feature stands for a strong scalar feature, 233 taken to be responsible for PIs activating a process of domain widening (i.e., PIs are 234 scalar elements that activate alternatives within smaller domains; hence, they are 235 felicitous in downward entailing contexts, see Ladusaw 1980), the [uNEG] feature 236 guarantees that n-words establish a syntactic dependency relation with negation. 237 Consider the example in (8). 11 238 (8) Los estudiantes no han leído nada. (Spanish) 239 the students not have read nothing 240 ‘The students didn’t read anything.’ 241 In (8) nada is a PI, inherently characterized as [+σ], which is a requirement for 242 its interpretation as a (negative) PI. As such, it basically means some, with the 243 addition that it forces us to consider not only the current (pragmatic) context, but 244 the largest contextually relevant domain (e.g., by saying The students didn’t read 245 anything, the speaker includes not only a contextually salient set corresponding to, 246 for example, textbooks, but a larger set including journal articles, book chapters, 247 etc.), thus it is domain-widening. Furthermore, nada (like English any) activates 248 alternatives within smaller domains, which means that it introduces the implicature 249 that a statement containing any is the pragmatically strongest statement possible in 250 context (e.g., if The students didn’t read anything, a fortiori, The students didn’t 251 read a textbook). It is the feature [+σ], where σ stands for strong, that is assumed to 252 be responsible for the activation of alternatives of scalar items.9 253 We assume that n-words, more precisely n-words1, are lexical roots 254 semantically characterized by an abstract [+σ] feature. As such, n-words1 will 255 require a covert exhaustifier σ operator in a c-commanding position to be properly 256 licensed (Chierchia 2006; cf. Giannakidou 1998). Furthermore, in (8), nada is a PI 257 licensed by the negative marker, and the two items (the negative marker and the n- 258 word) are in a syntactic relationship mediated by Agree that guarantees a NC 259 reading. In order to formalize the distinction between being polar sensitive to some 260 exhaustifying operator on the one hand, and participating in an Agree relationship 9 See Longobardi (2014) for an alternative operational rule that assigns [+ANY] to a lexical head when “that phrase is interpreted as an existentially bound variable in the immediate scope of a distinct negative (non-veridical) operator and nowhere else” (Longobardi 2014: 226). 12 261 on the other, we assume (following Espinal and Tubau in press) that a polar n-word 262 may acquire, in the course of the syntactic derivation, a formal [uNEG] feature that 263 makes it syntactically dependent on a matching interpretable feature. In this view, 264 n-words1 are the output of merging indefinite polarity roots defined [+σ] with a 265 formal syntactic [uNEG] feature at syntax, as shown in (9).10 We depart from 266 Zeijlstra’s (2004 and ff.) lexical endowment of an inherent [uNEG] feature in order 267 to claim explicitly that only the NC reading, but not semantic polarity, is 268 syntactically-driven. 269 (9) n-word1[+ ] [uNEG] σ 270 271 [uNEG] √ [+σ] 272 This merge operation applies under the assumption that [uNEG] heads select 273 items specified [+σ], but notice that not all polar roots need to be syntactically 274 specified [uNEG]. For instance, French qui/quoi que ce soit (lit. who/whatever it 275 may be) ‘anybody/anything’ and Romanian cine știe ce (lit. who knows what) 276 ‘anything’ occur in non-negative polar contexts (and are hence specified as [+σ]), 277 but cannot occur in negative contexts, thus supporting the claim that they do not 278 merge with a [uNEG] feature (Espinal and Tubau in press). 279 Polarity items defined [+σ] will require a σ operator adjoined to a conditional 280 or an interrogative operator for their interpretation. Alternatively, a lexical item 281 encoding semantic negation, such as verbs that express fear and doubt, prepositions 282 that express the concepts of ‘before’ and ‘until’, and other contexts licensing 283 expletive negation, can also host the licensing σ operator. 10 For the idea that word formation is a syntactic process (i.e., it follows the same syntactic rules that are used to build up clauses) see the Distributed Morphology model (Halle and Marantz 1993, Embick and Noyer 2007, a. o.). 13 284 The relationship between [+σ] and the Op σ is a semantic dependency (cf. 285 Giannakidou 1998). However, when a root defined [+σ] merges with [uNEG] a 286 syntactic requirement comes into play. Given that [uNEG] is an uninterpretable 287 syntactic feature, it must be checked (i.e., deleted) before the derivation is sent to 288 the interfaces. The feature [uNEG] needs to establish a relation of Agree with a 289 matching syntactic [iNEG] feature to be checked. Such a feature is provided either 290 by the overt negative operator [iNEG] (i.e., the negative marker no), which signals 291 the scope of negation (Zeijlstra 2004: 271), or by a covert negative operator, which 292 is inserted when the n-word occupies a position that outscopes NegP (Zeijlstra 293 2004: 259).11 294 Accordingly, following Zeijlstra (2004, 2012), Haegeman and Lohndal (2010), 295 and Biberauer and Zeijlstra (2012), we assume that Catalan and Spanish n-words 296 participate in an (Inverse) Agree relation with a negative operator that carries an 297 [iNEG] feature.12 Following this approach, the Spanish example in (8) can be 298 schematically analysed as in (10). 299 (10) Los estudiantes no[iNEG] han leído nada[+ ][uNEG]. σ 11 A reviewer asks why a sentence such as *Juan ha comido nada (lit. Juan has eaten nothing) is not grammatical in Spanish. That is, what prevents an abstract negative operator from rescuing such a sentence? We follow Zeijlstra (2004: 271) in assuming that the overt negative operator marks the scope of negation. Hence, either it is clear that negation is sentential by means of the presence of preverbal n-words, or the negative marker must be overt. Likewise, another reviewer asks why postverbal n-words cannot occur without an overt licenser if a negative quantifier variant (n-word2) is available in Spanish and Catalan. Our answer relies on the fact that this still emergent negative quantifier variant is inherently specified as [uFoc] and, as such, it can only be appropriately licensed after movement to a left peripheral Focus position: NADA ha comido Juan (lit. nothing has eaten Juan). See also Déprez et al. (2015). 12 Unlike Chomsky’s (2000, 2001) formulation of Agree, where a Probe c-commands a Goal, the constituent that is assumed to carry an uninterpretable feature, Inverse/Reverse Agree is formalized as in (i) in Zeijlstra (2012: 514), where the Goal c-commands the Probe. (i) Inverse/Reverse Agree: α can agree with β iff: a. α carries at least one uninterpretable feature and β carries a matching interpretable feature. b. β c-commands α. c. β is the closest goal to α. 14 300 Let us now consider preverbal n-words, as in (11a). If, under the present 301 assumption, n-words are lexical roots characterized as PIs (i.e., defined by a [+σ] 302 feature), when they occur in preverbal position an abstract negative operator, Op¬, 303 specified as [iNEG], is postulated to license the n-word specified as [uNEG]. This 304 Last Resort operation is required in order to guarantee their negative interpretation 305 as n-words, as shown in (11b) (Zeijlstra 2004). 306 (11) a. Nadie llevaba gafas (Spanish) 307 nobody wore glasses 308 ‘Nobody was wearing glasses.’ 309 b. [Opσ [Op¬[iNEG] [nadie[+σ][uNEG] llevaba gafas ]]] 310 Summing up, our assumptions regarding n-words are as follows. For a 311 population of native speakers, n-words are lexically specified as indefinite PIs, 312 semantically characterized by a strong scalar feature [+σ] that forces their 313 interpretation in a domain-widening context. Conceived as polar roots, these items 314 may merge in the course of the derivation with an abstract syntactic [uNEG] 315 feature. Once this merge has occurred, it imposes on the resulting n-words1 the 316 requirement that they participate in an Agree chain with an [iNEG] operator, thus 317 resulting in a single negation or a concordant reading. Whether they occur 318 postverbally, preverbally or in isolation, such n-words1 entail semantically a 319 numeral zero meaning (we will come back to this claim in Section 4).13 13 See Déprez (1997) for the original analysis of French rien and personne as negative expressions whose quantificational force is identified with the numeral zero: “the assumption that the numeral that n-word incorporates means zero provides a straightforward explanation as to why answers with n-words are always interpreted negatively” (p. 123). This means that for this author negating a zero numeral produces the cancelling effect of DN (Déprez 2000: 269). (i) Je n’ai pas vu personne. (French) I NEG.have not seen zero.person ‘I did not see nobody.’ (⇒ I saw at least one person) Although we will not make any strong claim about French n-words in this paper, we suspect that their semantic characterization cannot be exactly the same as the one postulated for the two 15 320 However, a second population of native speakers of Catalan and Spanish seem 321 to have in their grammar a competing lexical variant for n-words, which we will 322 call henceforth n-words2. This emergent variant (in the sense of not showing the 323 basic polar meaning of n-words in these NC languages) is lexically characterized as 324 being inherently negative, and its meaning corresponds to a negative quantifier, 325 defined semantically as ¬∃ (similar to negative quantifiers as defined in first-order 326 logic). This variant, encoding an inherent semantic negation, does not require any 327 sort of syntactic checking to be negative and therefore does not participate in NC 328 structures.14 Furthermore, in line with the work of Zanuttini (1991) and Déprez 329 (1997), an n-word2 is decomposable into a negative component and a 330 quantificational one, and when it combines with an external negative operator it 331 necessarily conveys a DN reading. 332 In Sections 3 and 4 we will return to the importance of the ambiguity of 333 argumental n-words that we postulate in order to explain their interpretation when 334 they occur as isolated answers to negative wh-questions. For the time being we will 335 consider in the next section their syntactic distribution in comparison to preverbal Romance languages studied in this article. This is supported by the fact that the corresponding Catalan and Spanish examples in (ii) and (iii) do not license a DN meaning. (ii) No he visto a nadie. (Spanish) not have seen DOM anybody ‘I did not see anybody.’ (iii) No he vist pas a ningú. (Catalan) not have seen not DOM anybody ‘I did not see anybody.’ See also footnote 5 above. 14 The emergence of this n-word2 variant raises the question of what prevents a negative quantifier from merging with [uNEG]. The answer follows from an assumption we have specified above, namely that a [uNEG] feature selects for polar roots specified [+σ]. If negative quantifiers merged with [uNEG], the prediction would be that they should occur in NC constructions, but this does not seem to be the case, since when two negative quantifiers co-occur in a clause, DN is predicted to arise. One possible way out of this puzzle is to assume, in keeping with Déprez’s (1992) and de Swart and Sag’s (2002) approach to n-words, that an operation of resumption is responsible for the NC reading of sequences of negative quantifiers (whereas an operation of iteration accounts for DN readings). But see Déprez et al. (2015) for arguments against the validity of a resumption analysis in Catalan. 16 336 n-words and their interpretation when we take into account left peripheral and 337 Topic-Focus information structure considerations (cf. Chafe 1976, Vallduví 1993, 338 Krifka 2007). 339 340 2.2. Preverbal and isolated n-words as Focus 341 In this section we address the syntactic and informational status of isolated 342 argumental n-words used as answers to negative wh-questions, regardless of 343 whether they are finally interpreted as conveying a single negation or given a DN 344 interpretation. In similarity with preverbal n-words, which occupy a left-peripheral 345 position in a cartographic approach to clause structure (Rizzi 1997, Cinque and 346 Rizzi 2008, a. o.), isolated n-words are also candidates to occupy a syntactic 347 position at the left periphery of the clause. First, if n-words in fragment answers are 348 assumed to be part of larger structures that have undergone clausal ellipsis (see 349 Section 1), it is not surprising that they behave similarly to preverbal n-words. 350 Second, given that isolated n-words are the answer to wh-constituent questions, 351 from the two available candidate positions −Topic or Focus− that have been 352 postulated for them following the Split-CP hypothesis, they are expected to sit in 353 Focus.15 Third, as happens with preverbal n-words, which generally fail the tests of 354 Topichood (Vallduví 1993, Espinal 2007, for Catalan; RAE 2009, for Spanish) and, 355 hence, can never be prototypical Topics, isolated n-words cannot be Topics either. 356 As shown in the Catalan examples in (12) and (13), preverbal n-words and isolated 357 n-words can be preceded by Topic constituents but not followed by them (Vallduví 358 1993; Espinal 2007: 62 and 63). The same seems to apply to Spanish. 359 (12) a. [Topic A mi], [Focus ningú] no em deu res. 15 See also Holmberg (2013) for an account of answer particles (‘yes’ and ‘no’) as sitting in Focus. 17 360 to me nobody not me owe nothing 361 ‘Nobody owes me anything.’ 362 b. *[Focus Ningú] [Topic a mi] no em deu res. 363 nobody to me not me owe nothing 364 (13) Q. Qui no et deu res? 365 who not you owes nothing 366 ‘Who doesn’t owe you anything?’ 367 A. a. [Topic A mi], [Focus ningú]. 368 to me nobody 369 ‘To me, nobody’ (= Nobody owes me anything) 370 b. *[Focus Ningú], [Topic a mi] 371 Thus, it is plausible to assume that preverbal and isolated n-words sit in Focus.16 372 Besides these syntactic arguments, from a semantic point of view isolated n- 373 words, used as answers to negative wh-questions, have a meaning that contrasts 374 with the necessary specific interpretation characteristic of Topic constituents 375 (Cohen and Erteschik-Shir 2002). 17 This interpretation is compatible with the 376 indefinite meaning of n-words, be it the indefinite PI variant (n-word1) or the 377 indefinite negative quantifier variant (n-word2). When isolated n-words in answers 378 are indefinite PIs, they must be interpreted in relation to a wh-domain constituted 16 Some arguments in support of the claim that preverbal n-words fail the tests of Topichood, taken from Vallduví (1993) and Espinal (2007), are the following: when a contrastive context is built into the wh-question, the n-word is diagnosed as a Contrastive Focus rather than as a Topic; n-words cannot appear to the left of fronted wh-questions as Topics do; unlike left-detached complements, n- words do not allow a coindexed clitic attached to the verb; unlike left-detachment, which allows several fronted constituents, preverbal argumental n-words are usually reduced to one single constituent (with the exception of specific n-word combinations with adjuncts; e.g., mai ‘never’); unlike subjects from embedded clauses, n-words cannot be left-adjoined to the matrix sentential node; and unlike preverbal subjects, n-words are not interpreted as informational links to a previous discourse. 17 According to Kiss’s (1998) distinction between identificational focus and information focus, isolated n-words should be said to express exhaustive identification and, therefore, to function as identificational focus, which is identified as the exhaustive subset of the set of contextually or situationally given elements, for which the predicate phrase actually holds. See also Szabolcsi (1981), who regards exhaustive listing as the predominant semantic characteristic of Focus. 18 379 by a set of individual entities. As a result, the variable n-word is under the scope of 380 the negative logical operator occurring in the background question, thus entailing 381 that no individual of the set of relevant entity-type denoting expressions can be 382 interpreted as an argument of the predicate provided by the background question. 383 When isolated n-words in answers are indefinite negative quantifiers, by contrast, 384 they must be interpreted in relation to a wh-domain formed by a set of generalized 385 quantifier expressions, and the output interpretation is therefore that the negative 386 generalized quantifier applies as a function to the predicate provided by the 387 discourse question.18 388 To sum up, in this section we have presented the two main theoretical 389 assumptions that guide our analysis of argumental n-words used as answers to 390 negative wh-questions. First, we have analysed n-words as lexically ambiguous 391 between PIs, which can participate in a syntactic relation of NC (n-words1), and 392 indefinite negative quantifiers (n-words2), which cannot. Second, we have argued 393 that preverbal and isolated argumental n-words must be considered Focus 394 constituents, and that interpretively they function as Focus. This is why in answers 395 to constituent wh-questions, our analysis of isolated argumental n-words (which are 396 in Focus) leads to a Background-Focus structure along the lines which will be 397 developed in Section 4. 398 399 3. Isolated argumental n-words analysed in terms of ellipsis 18 We will come back to this issue, namely the two possible ways of conceiving the wh-domain (as either a set of entities of type 〈e〉 or a set of generalized quantifiers of type 〈〈e,t〉t〉), in Section 4. For the time being, we would like to point out that positive universal quantifiers such as Spanish todo el mundo lit. all the world ‘everybody’ or Catalan tothom ‘everybody’ would not be felicitous as answers to negative wh-questions; they are only so to positive ones. The inability of ordinary universals to scope over negation has also been observed for Greek (Veloudis 1982, Giannakidou 2000) and Hungarian (Szabolcsi 1981), but is less clear in English (Beghelli and Stowell 1997). 19 400 Before presenting our own analysis, in this section we evaluate the extent to which 401 it is possible to explain the interpretation(s) that n-words may have as answers to 402 negative wh-questions in NC languages in terms of ellipsis. In Section 3.1 an 403 ellipsis account à la Merchant (2001, 2004) is combined with an analysis of NC à 404 la Zeijlstra (2004 and ff.). In Section 3.2 we evaluate the predictions of a semantic 405 ellipsis account. It will be shown that, as they stand, neither of these approaches 406 allows us to account for the single negation reading of n-words in Catalan and 407 Spanish answers to negative wh-questions. 408 409 3.1. A syntactic ellipsis account 410 In this section we will show that, combining Zeijlstra’s (2004, and ff.) analysis of 411 NC in Romance −conceived as syntactic Agree− with Merchant’s (2001, 2004) 412 clausal ellipsis account of fragment answers, DN is predicted to be the only 413 possible interpretation associated with Catalan and Spanish n-words used as 414 answers to negative wh-questions. This follows from the presence of two [iNEG] 415 features in the syntactic structure: one in the covert Op¬[iNEG] that licenses the 416 isolated n-word moved to a syntactic Focus position, and another one that is copied 417 from the negative wh-question, as shown in (14) and (15).19 These structures would 418 be postulated for the Spanish fragment answers in (6A) and (7A), respectively, 419 repeated here for convenience as (16A) and (17A).20, 21 19 As already pointed out by Biberauer and Zeijlstra (2012: 352) for Italian, under a syntactically- oriented approach Spanish sequences of the sort Nada … no, and Nadie no..., are predicted to be ungrammatical because no[iNEG] does not c-command the preverbal n-word [uNEG]. Such a configuration triggers the presence of a covert c-commanding Op¬[iNEG], DN being the only possible interpretation that can be obtained. 20 One could assume that object and subject n-words should be interpreted differently, since object n-words could in principle have their [uNEG] feature checked prior to Focus movement. The prediction would be that object n-words should preferably license a single negation reading, whereas subject n-words should license a positive one. However, in Espinal et al. (2015) no subject-object difference has been observed in Catalan, and only a slight difference in Spanish. This difference, 20 420 (14) [Op¬[iNEG] [FocP nadie [uNEG] [E] [TP ti no[iNEG] llevaba gafas]]] 421 (15) [Op¬[iNEG] [FocP nadai [uNEG] [E] [TP los estudiantes no[iNEG] han leído ti]]] 422 (16) Q: ¿Quién no llevaba gafas? 423 who not wore glasses 424 ‘Who wasn’t wearing glasses?’ 425 A: Nadie. 426 nobody (= Nobody was wearing glasses) 427 (17) Q: ¿Qué no han leído los estudiantes? 428 what not have read the students 429 ‘What didn’t the students read?’ 430 A: Nada. 431 nothing (= The students didn’t read anything) 432 Recall that the analysis of fragment answers as complete sentences that have 433 undergone ellipsis (Merchant 2001, 2004; Merchant et al. 2013) proceeds in two 434 steps: first, the fragment is A'-moved from its first Merge position to a functional 435 category in the left periphery, which results in the need for a last resort Op¬[iNEG] 436 that c-commands a [uNEG] feature of the n-word, so that Agree can delete it; 437 second, PF-deletion of the rest of the clause applies.22 438 For the present purposes what is crucial is that following this combined 439 syntactic analysis of argumental isolated n-words as fragment answers, the single nonetheless, would support that objects are less often interpreted as conveying DN when compared to subjects. We thank M. Labelle (p.c.) for pointing this out to us. 21 In the syntactic representations of this section we omit, for simplicity, the semantic binding of the [+σ] feature. 22 The arguments gathered in Merchant (2004) to sustain the movement part of his analysis of fragments are many and varied: preposition stranding, the distribution of pronominals in various languages, islands, complementizer deletion, c-selectional effects in raising and control infinitivals, restrictions in predicate answers in English, the distribution of polarity items in English, Greek and Irish, generic objects in Turkish and caseless fragments in Korean and Japanese. The evidence for ellipsis is also vast: connectivity effects involving case-matching in a variety of languages, Greek anaphoric dependencies, binding and the distribution of scope and bound pronouns in English. 21 440 negation interpretation illustrated in (16A) and (17A) cannot be explained 441 straightforwardly, unless an extra assumption (i.e., that the negated part of the 442 negative question may but does not have to license the elided part) is introduced in 443 the discussion (Zeijlstra 2004).23 444 Among NC languages, the situation in Catalan poses an additional challenge 445 for an account that combines syntactic Agree and ellipsis. As illustrated in (18), for 446 a population of native speakers of Catalan the sentential negative marker optionally 447 co-occurs with preverbal n-words. 448 (18) Ningú (no) porta ulleres. (Catalan) 449 nobody not wears glasses 450 ‘Nobody is wearing glasses.’ 451 The optionality of no has been a serious puzzle for all theories of NC that have 452 attempted to account for the Catalan data. Both van der Wouden and Zwarts (1993) 453 and, more recently, Zeijlstra (2004) have postulated the existence of two dialects of 454 Catalan in an attempt to account for such optionality. The latter, in particular, 455 assumes that this language manifests in “one variety that is a Strict NC variation 456 (Catalan I), and one variety that exhibits Non-Strict NC behaviour (Catalan II)” 457 (Zeijlstra 2004: 133).24 This same author further assumes that there is a crucial 458 difference between the two varieties of Catalan with respect to the featural content 23 A reader might suggest that if it is assumed that the n-word reconstructs to its base-generated position under the scope of negation instead of remaining in Spec,FocP, the NC reading of the fragment answer can be straightforwardly explained. However, as has been shown in Section 2.2, n- words as answers to negative wh-questions must be considered instances of (identificational) focus. In addition, if reconstruction of n-words in Spec,FocP were possible, we would expect the reconstructed answer (e.g. No llevaba gafas nadie, lit. not wore glasses nobody, ‘Nobody was wearing glasses’) to be acceptable as an answer to a wh-question such as (16Q). However, in such a context it is not appropriate to use a postverbal n-word with a falling boundary tone. Sentences where the n-word occurs in postverbal position (e.g., Los estudiantes no han leído nada, lit. the students not have read nothing, ‘The students didn’t read anything’) are also inappropriate as answers to wh-questions about the object argument. 24 NC is referred to as Strict if the sentential negative marker always co-occurs with the n-words in all contexts. Conversely, NC is Non-Strict if the sentential negative marker co-occurs with postverbal n-words, but needs to be absent when n-words occur preverbally (Giannakidou 1998). 22 459 of the sentential negative marker: while in Catalan I (i.e., the variety where a 460 preverbal n-word is followed by no), the sentential negative marker is assumed to 461 bear a [uNEG] feature, in Catalan II (i.e., the variety where a preverbal n-word is 462 not followed by no) it is assumed to carry an [iNEG] feature. As illustrated in (19), 463 a Catalan I sentence containing a preverbal n-word and the sentential negative 464 marker would accordingly be analysed with a covert Op¬, defined as [iNEG], 465 simultaneously licensing the preverbal n-word and the sentential negative marker, 466 both characterised as [uNEG].25 467 (19) Op¬[iNEG] ningú [uNEG] no [uNEG] porta ulleres (Catalan I) 468 In Catalan II, by contrast, as shown in (20), there is no overt sentential negative 469 marker co-occurring with preverbal n-words and, hence, a covert Op¬[iNEG] licenses 470 the n-word.26 471 (20) Op¬[iNEG] ningú [uNEG] porta ulleres (Catalan II) 472 By itself, however, Zeijlstra’s analysis does not predict the possible DN 473 interpretation that a population of native speakers of Catalan may attribute to 25 In contrast to Zeijlstra’s proposal, we assume that in all varieties of Catalan single negation corresponds to an overt sentential negative marker that always carries an [iNEG] feature. Regarding the population of speakers of this language that optionally use no with preverbal n-words (a variety that does not correspond to a Strict NC language), we postulate that the speakers’ lexicon has, in addition to a no1 variant characterized [iNEG], a polarity-like no2 (lit. not) marker, characterized [+σ], that we postulate both for NC structures of the sort preverbal n-words + no + V and in prototypical pleonastic or expletive negation contexts (cf. Jespersen 1917, a. o.). The fact that no2 needs an antiveridical operator as a licenser explains why it cannot appear in non-veridical contexts such as those introduced by conditional or interrogative operators. See Longobardi (2014) for the hypothesis that Catalan no is ambiguous, a topic which we will not develop any further because it is beyond the scope of the present article. See Déprez et al. (2015) for empirical support for the lexical ambiguity of n-words, and the possible ambiguity of no in Catalan. 26 Following Zeijlstra (2004), given that the sentential negative marker is defined as [uNEG] in Catalan I, an Op¬ with a licensing [iNEG] feature is required even when there is no n-word present in the derivation. This is illustrated in (ia). In Catalan II, by contrast, the sentential negative marker carries an [iNEG] feature, as illustrated in (ib). (i) En Joan no porta ulleres the John not wears glasses ‘John isn’t wearing glasses.’ a. En Joan Op¬[iNEG] no[uNEG] porta ulleres (Catalan I) b. En Joan no[iNEG] porta ulleres (Catalan II) In Zeijlstra’s (2004) account, Catalan I is assumed to align with Strict NC languages like Romanian, while Catalan II is assumed to align with Non-Strict NC languages like Spanish. 23 474 isolated n-words in answers to negative wh-questions (see Section 1.2). See 475 example (21). 476 (21) Q: Qui no porta ulleres? 477 who not wears glasses 478 ‘Who is not wearing glasses?’ 479 A: Ningú. 480 nobody (= Nobody is wearing glasses / Everybody is wearing glasses) 481 However, in combination with Merchant’s (2001, 2004) deletion under ellipsis 482 a DN interpretation is the only reading that is predicted to arise in Catalan. 483 Following such a syntactic analysis, consider the structure in (22) for isolated 484 n-words in Catalan I (Zeijlstra’s Strict NC variety). The Op¬[iNEG] that licenses the 485 n-word in Focus (inherently specified with a [uNEG] feature; cf. our n-words1 486 variant) and the Op¬[iNEG] that licenses the sentential negative marker copied from 487 the negative wh-question would cancel each other out, thus yielding a DN reading, 488 contrary to the only possible interpretation that one population of speakers associate 489 with this sentence. 490 (22) [Op¬[iNEG] [FocP ningúi [uNEG] [E] [TP ti Op¬[iNEG] no [uNEG] porta ulleres]]] 491 In Catalan II (Zeijlstra’s Non-Strict NC variety), two instances of [iNEG] (one in 492 the Op¬ and one in the copied sentential negative marker) would also co-occur. As 493 shown in (23), a DN reading would again be the only possible interpretation for the 494 isolated n-word. 495 (23) [Op¬[iNEG] [FocP Ningúi [uNEG] [E] [TP ti no [iNEG] porta ulleres]]] 496 In conclusion, a purely syntactically-oriented account that combines an 497 analysis of n-words merging with a [uNEG] feature at narrow syntax with an 498 analysis of isolated n-words as subject to syntactic clausal ellipsis cannot account 24 499 for the single negation interpretation of Catalan and Spanish isolated n-words, 500 unless for these languages it is claimed that VP-ellipsis may apply too (as we 501 discussed for English and German (4) and (5)). The problem is that we do not have 502 any sort of evidence from Catalan or Spanish to claim that isolated n-words are 503 sometimes the output of clausal ellipsis (with negation in the TP domain, which 504 would guarantee a DN reading) and other times the output of VP-ellipsis (which 505 would guarantee a single negation reading). Furthermore, a syntactic ellipsis 506 account would not be able to explain the fact that a single negation interpretation 507 for isolated argumental n-words is generally preferred by native speakers in both 508 Catalan and Spanish.27 509 Similarly, if an ellipsis account is combined with the assumption that n-words 510 are inherently negative quantifiers (our n-words2 variant), the result is not any 511 better, for we continue to obtain only a DN reading. In that case, in contrast to the 512 structures in (22) and (23) above, having an n-word semantically defined as ¬∃ 513 would constitute a first source of negation, whereas the negative operator (either 514 covert or overt, specified as [iNEG]) associated with the negative question would 27 An additional problem for this syntactically-oriented approach appears when Romanian data are considered. In Romanian, a Romance Strict NC language, the n-word in (iA) can only receive a single negation interpretation. However, when combining Zeijlstra (2004) and Merchant (2001, 2004), we would predict the structure seen in (ii). (i) Q: Ce nu au citit studenţii? (Romanian) what not have read the students ‘What didn’t the students read?’ A: Nimic. nothing (= The students didn’t read anything) (ii) [Op¬[iNEG] [FocP nimici [uNEG] [E] [TP Op¬[iNEG] [TP studenţii nu [uNEG] au citit]]]] Following the models under discussion, this structure contains two Op¬ specified as [iNEG], one that licenses nimic and another one that licenses nu. These operators are predicted to cancel each other out, thus yielding a positive interpretation for (iA). Contrary to this theoretical prediction, our informants agree on the claim that a DN reading cannot possibly be obtained for isolated n-words even if a special marked intonation is provided. Incidentally, one of our informants (E. Ciutescu, p.c.) reports that n-words in Romanian can only be interpreted as conveying DN when they are part of a full structure with two n-words and the sentential negative marker (n-word + nu + V + n-word), and some special stress is added to the first n-word, thus confirming Falaus’ (2007) claims for this language. 25 515 constitute a second source of negation. Both negations would cancel each other out, 516 thus resulting in DN. 517 In the next section we consider a semantic ellipsis account for n-words acting 518 as fragment answers to negative wh-questions. We show that this account suffers 519 from similar shortcomings. 520 521 3.2. A semantic ellipsis account 522 Apart from the limitations that have been raised in the previous section, the analysis 523 of fragment answers as complete sentences that have undergone ellipsis (Merchant 524 2001, 2004; Merchant et al. 2013) has some additional problems. First, establishing 525 what constitutes the antecedent of a fragment is not uncontroversial, since it has 526 been regarded as a question either of syntactic LF isomorphism (cf. Watanabe 527 2004) or of semantic propositional isomorphism (cf. Giannakidou 2000, 2006). 528 The first solution will have the same problems as a narrow syntactic analysis 529 has both for n-words of the type n-word1 (i.e., n-words merging a [uNEG] feature) 530 and for n-words of the type n-word2 (i.e., negative existential quantifiers). More 531 specifically, although applying syntactic LF isomorphism to Catalan and Spanish 532 may correctly predict the meaning of isolated n-word answers produced with an 533 unmarked intonation contour, the grammatical motivation for copying more or less 534 syntactic material from the question into the answer looks like a stipulation. Thus, 535 under such an ellipsis account, the DN and single negation readings should be 536 argued to follow from two different LFs in association with n-words of the type n- 537 word1; that is, the DN reading would result from (24Aa), which contains the 538 sentential negative marker present in the antecedent question, and the single 26 539 negation reading would follow from (24Ab), which does not contain a sentential 540 negative marker in the discourse context. 541 (24) Q: ¿Quién [NegP no [iNEG] [TP ha comido postre]]? (Spanish) 542 who not has eaten dessert 543 ‘Who didn’t eat dessert?’ 544 A: a. [Op¬[iNEG] [FocP nadie [uNEG] [NegP no [iNEG] ha comido postre]]] 545 nobody not has eaten dessert 546 b. [Op¬[iNEG] [FocP nadie [uNEG] [TP ha comido postre]]] 547 nobody has eaten dessert 548 Assuming that n-words are negative quantifiers (i.e., of the type n-word2) 549 would require the same operation for the right readings to emerge: if the sentential 550 negative marker is present in the antecedent question, the negative quantifier would 551 convey a DN reading; if the sentential negative marker is absent, the negative 552 quantifier would lead to a single negation reading. However, what remains 553 uncertain in this type of analysis is how speakers determine exactly which material 554 from the antecedent question gets copied onto the answer (either NegP or TP). 555 Let us now consider whether the two meanings of the fragment answers in (24) 556 can be composed by applying a semantic propositional isomorphism. 557 Defendants of this approach, such as Giannakidou (2000, 2006), assume in line 558 with Karttunen (1977) that questions denote the set of their true answers. 559 Giannakidou’s (2006) idea is that the “negative meaning in elliptical fragments then 560 arises not as an inherent contribution of the n-words, but rather as the result of their 561 being associated with negation at the level at which ellipsis is resolved” (p. 363). In 562 other words, following Merchant (2001), Giannakidou claims that the elliptical 563 proposition is licensed semantically if it can be inferred from the antecedent 27 564 question. Hence, an answer such as nadie to a negative wh-question such as (24Q) 565 would have to be interpreted as being the elliptical counterpart of one of the 566 answers in the wh-set in (25), namely the last one in the set. 567 (25) Wh-set of answers: {Juan didn’t eat dessert, Pedro didn’t eat dessert, 568 someone didn’t eat dessert, everybody didn’t eat dessert, nobody didn’t eat 569 dessert} 570 At this point we must consider what sort of negation is expressed in negative 571 wh-questions of the sort exemplified in (24Q). Following Ladd (1981), Romero and 572 Han (2004), and Reese (2006), we claim that these type of negative wh-questions 573 can only express inner / inside negation, as supported by the morphosyntactic 574 distribution of the negative PI tampoco ‘either’ (vs. the positive PI también ‘too’). 575 Therefore, we hold that these negative wh-questions are not ambiguous. 576 (26) ¿Quién no ha comido postre tampoco /*también? (Spanish) 577 who not has eaten dessert either too 578 ‘Who didn’t eat dessert either /*too?’ 579 Moreover, questions containing inner negation are negatively biased and require a 580 non-neutral context, in the sense that the question is about the proposition that 581 Someone didn’t eat dessert (see also the discussion around examples (1) and (2) 582 above), which means that this type of question is felicitous when there is 583 compelling contextual evidence against p (p being in this particular case Everybody 584 ate dessert) (cf. Büring and Gunlogson 2000). 585 Giannakidou (2006) postulates that n-words can be either universal quantifiers 586 or existential ones, with the idea that the universal negation is equivalent to the 587 existential version. That is, ∀x [PERSON(x) → ¬eat(dessert,x)] is equivalent to ¬∃x 588 [PERSON(x) & eat(dessert,x)] for Nadie ha comido postre. However, in (24) the 28 589 prediction would be that the isolated argumental n-word Nadie stands for Nadie no 590 ha comido postre lit. nobody not has eaten dessert, which contains the inherent 591 negation driven by the n-word (¬∃, our n-word2) plus the inner negation conveyed 592 by the negative question. That is, Nadie would stand for the following logical 593 representation: ¬∃x [PERSON(x) & ¬eat(dessert,x)]. Accordingly, Nadie as an 594 answer to ¿Quién no ha comido postre? could only be interpreted as conveying a 595 DN reading.28, 29 596 597 3.3. Summary 598 To sum up, in this section we have reviewed three different analyses for fragment 599 answers, and we have shown that only one of them (the one based on LF 600 isomorphism) can account for the two possible readings that Spanish (and Catalan) 601 speakers attribute to isolated argumental n-words as answers to negative wh- 602 questions, but only under the dubious assumption that there is a grammatical 603 criterion for deciding how much material from the question gets copied. The other 604 two approaches (the syntactic ellipsis account and the semantic propositional 605 isomorphism) can only predict the DN reading, which means that a not insignificant 606 amount of data still remains to be explained.30 28 As discussed in Reese (2006: 336), a similar problem arises from the semantic approach of Romero and Han (2004) inside negative polar questions, where the proposition in the scope of the VERUM operator is ¬ϕ. If a no answer negates the embedded proposition, then it should convey ϕ rather than ¬ϕ. 29 Giannakidou’s (2000: 501) analysis for a Strict NC language like Greek makes a distinction between ordinary universal quantifiers (e.g., kathe ‘every’), which cannot take scope over negation, and more specific universal quantifiers (e.g., KANENAS ‘nobody’) that are lexically specified as universals that take scope over negation. The universal-over-negation reading of universal emphatics is achieved by Quantifier Raising at LF. However, the analysis of emphatic quantifiers in Greek cannot be extended to n-words in Catalan and Spanish; this is an additional argument for exploring a new account for the interpretation of n-words in Catalan and Spanish. 30 See also de Swart (2010: 30-34) for an overview of the arguments in favour of and against an ellipsis account of fragment answers. 29 607 This discussion leads us to consider that an alternative analysis of isolated 608 argumental n-words as answers to negative wh-questions needs to be developed. 609 This new analysis should account not only for the marked DN reading that one 610 population of Catalan and Spanish speakers associate with n-words, but, crucially, 611 also for the unmarked single negation reading that another population of Catalan 612 and Spanish speakers associate with isolated n-words. 613 We have additionally argued that, since the negative wh-questions under study 614 contain an inner negation, we are dealing not with cases of denial but rather with 615 NC vs. DN contrasts at the level of Question-Answer pairs. 616 617 4. Towards a new analysis for interpreting argumental n-words 618 Recall from Section 1.2 that the main goal of this paper is to account for the 619 ‘surprising’ single negation reading that Catalan and Spanish isolated argumental n- 620 words may express when they are used as answers to negative wh-questions. In 621 Section 2 we argued that for one population of speakers of these languages n-words 622 are mainly indefinite PIs, characterized inherently with a [+σ] semantic formal 623 feature. For these speakers n-words become syntactically negative only after merge 624 with a formal [uNEG] feature, and an Agree relation is established with a negative 625 operator characterized with a formal [iNEG] feature. We pursue the idea that polar 626 n-words that are in an Agree-chain with a negative operator specified [iNEG] entail 627 semantically a numeral zero meaning. In this section we will show how this 628 inference is obtained. 629 Recall also from Section 2.1 that a second population of speakers has access to 630 n-words characterized as indefinite negative existential quantifiers. This variant, 631 which corresponds to a generalized quantifier defined semantically as 30 632 λP¬∃x[P(x)], can neither merge with [uNEG] nor participate in NC (see footnote 633 13). 634 We assume from Section 2.2 that the isolated argumental n-word answer 635 corresponds to the wh-part of the constituent question and is identified with a 636 syntactic Focus position, independently of whether the argumental n-word is an 637 indefinite PI or an indefinite negative quantifier. 638 In this section we explore an analysis of isolated n-words as focus answers to 639 background negative questions along the lines of the semantic ellipsis account that 640 follows from the Structured Meaning approach developed by von Stechow (1991) 641 and Krifka (2001, 2004, 2007, 2011). We distinguish between an analysis of 642 isolated n-words conceived as indefinite PIs and an analysis of isolated n-words as 643 existential negative quantifiers. We postulate a congruency criterion on Question- 644 Answer pairs as defined in Krifka (2004). By combining the lexical ambiguity of n- 645 words with the congruency criterion, we explain that the single negation reading of 646 isolated n-words correlates with the polarity variant, whereas the DN reading 647 correlates with the negative quantifier variant. 648 The idea in the Structured Meaning approach is that “a wh-question sets the 649 background for an answer, which, in turn, determines the focus of the answer” (von 650 Stechow 1991: 38). Accordingly, the meaning of a fragment answer, like the 651 meaning of a full sentence, is organized into a background (B) part and a focus part 652 (i.e., a 〈B,F〉 information structure). Isolated n-words used as fragment answers 653 must be considered focus with respect to background questions, and, as focus 654 constituents, they determine a structured meaning. Furthermore, the isolated n- 655 words, as focus, are an element of the domain of the question. That is, with respect 656 to the wh-constituent question, the n-word answer indicates the existence of a set of 31 657 alternatives of the denotation (Krifka 2007): the alternative denotations must be 658 comparable to the denotation of the expression in focus, and must be of the same 659 type (entities 〈e〉, or generalized quantifiers 〈〈e,t〉t〉), and of the same ontological 660 sort (persons, things, places, etc.).31 661 This Structured Meaning approach to Question-Answer pairs, applied to the 662 meaning of isolated answers to constituent questions, predicts: (i) that the n-word 663 that appears in the Answer indicates the existence of a wh-set of potential answers 664 and triggers a set of alternatives for interpretation; and (ii) that the n-word Answer 665 must be congruent with respect to the Question. 666 Regarding the first of these predictions, it should be noted that since native 667 speakers of Spanish (and Catalan) take n-words as being either indefinite PIs or 668 indefinite negative quantifiers, and n-words are always the focus of the information 669 structure, two different wh-domains may be relevant when interpreting a Question- 670 Answer pair such as the one given in (24), repeated in (27) for convenience, namely 671 (28a) or (28b). In (28a) the wh-domain is constituted by a set of entities of type 〈e〉: 672 j stands for the denotation of Juan, m stands for the denotation of María, and – 673 among others– x[+ ] stands for the denotation of the indefinite expression nadie1, a σ 674 variable inherently specified with the semantic [+σ] feature. By contrast, in (28b) 675 the wh-domain is formed by a set of generalized quantifiers of type 〈〈e,t〉t〉: 676 λR[R(j)] denotes the set of sets of which Juan is a member, λQ[Q(m)] denotes the 31 Krifka (2007) opposes the term expression focus (ia) to denotation focus (ib), defined in the following terms: (i) A property F of an expression α is a Focus property iff F signals a. that alternatives of (parts of) the expression α or b. alternatives of the denotation of (parts of) α are relevant for the interpretation of α. 32 677 set of sets of which María is a member, and –among others– λP¬∃x[P(x)] is the 678 semantic interpretation of the negative existential quantifier nadie2.32 679 (27) Q: ¿Quién no ha comido postre? (Spanish) 680 who not has eaten dessert 681 ‘Who didn’t eat dessert?’ 682 A: Nadie. 683 anybody // nobody 684 (28) a. Wh-domain1: M = {j, m, ..., x[+ ]} σ 685 b. Wh-domain2: M = {λR[R(j)], λQ[Q(m)], …, λP¬∃x[P(x)]} 686 Regarding the second prediction, in a Structured Meaning approach to the 687 meaning of Answers, the latter are conceived as congruent or incongruent with 688 respect to the Question. But what does it mean to be a ‘congruent answer’? And, 689 more specifically, what is a congruent answer to a negative wh-question? 690 According to Krifka (2008) “the obvious congruence criterion in this representation 691 is that the question meaning should correspond to the background of the answer, in 692 the sense that the question meaning differs from the background of the answer only 693 insofar as it might have more restricted domains. (…) In addition, the focus must be 694 an element of the domain of the question” (Krifka 2008: 149).33 N-words do not 695 escape this congruence criterion. However, in addition, n-words satisfy the negative 696 bias of negative wh-questions in the following terms. 32 A reviewer is concerned about the accessibility of these two domains by native speakers that have the two variants in their lexicon. We think that access to either one or the other of these two indefinite variants is free, and that the final choice made determines whether a single negation or a DN is compositionally composed. 33 These two conditions are represented as follows (Krifka 2008: 149): (i) A question – answer pair Q - A with meanings !Q" and !A" = is congruent if and only if: a. !Q"⊆ B b. F ∈ DOM(!Q") 33 697 Negative wh-questions are interpreted as requests regarding a negative 698 proposition and restrict future moves. Along the lines of Cohen and Krifka (2011) 699 and Krifka (to appear), negative wh-questions are speech acts that can be seen as 700 functions from input commitments to output commitments. Input commitments 701 correspond to the set of commitment states that constitute the commitment space 702 with respect to which the negative wh-question must be interpreted, and they 703 include the commitment state that there is compelling contextual evidence against 704 p. Going back to the relevant question in (27), its meaning can be defined as in 705 (29).34 706 (29) 〈…, C〉 + REQUESTS1,S2 (ASS(¬p)) 707 where ¬p = [¬ATE(DESSERT)(WHO)] 708 and ∃c ∈ C |c: ∃x [¬ATE(DESSERT)(x)] 709 Output commitments relate to the expected negative reply of such a negative 710 question.35 711 With respect to the meaning of the question as expressed in (29), depending on 712 whether the wh-domain is (28a) or (28b), the final interpretation associated with the 713 isolated n-word used as a fragment answer is going to be single negation or DN. 714 Furthermore, in both cases the Answer is going to be congruent relative to the 715 Question, in accordance with Krifka’s criterion. 716 How are the two meanings composed? When the wh-domain is of the type in 717 (28a) the interpretation of the focus answer in (27A) has the structured property in 718 (30), with a B(ackground) part that comes from the meaning of the question and a 34 C is a commitment space, a set of commitment states; c is a specific commitment state corresponding to the presupposition of the negative wh-question; REQUEST corresponds to a request question between S1, the speaker, and S2, the hearer; this REQUEST operation applies to an assertion ASS of a negative proposition. 35 It would not be appropriate as a future move to the question in (27) to reply, for example, todos ‘everyone’. This positive universal quantifier in the reply would pragmatically clash with the negative question. See note 18 above. 34 719 F(ocus) part that is a member of the domain of the question. In this logical formula 720 the B part is the functor and the F part is the argument. 721 (30) 〈λx[¬ATE(DESSERT)(x)],〈x[+ ]〉〉 σ 722 Notice that the B part contains a logical negative operator, and the F contains 723 the lexical meaning associated with n-words1 when conceived as PIs: they 724 contribute a variable that in the case of nadie1 has additional properties: being 725 countable and of the person ontological sort, which must be taken as additional 726 felicity conditions on the interpretation of the variable. Since this variable is under 727 the scope of the only negative logical operator occurring in the B part, an inference 728 that follows is that quantity(x)=0. This inference is not part of the lexical meaning 729 of the variable x, since it is deduced only when the variable it introduces is under 730 the scope of the negative logical operator occurring in the B part.36 In that case the 731 negative propositional antecedent corresponding to the B question combines at the 732 level of logical representation with the lexical endowment of an n-word1, yielding a 733 single negation reading. 734 We therefore explain the possible single negation reading of the Spanish 735 isolated n-word in (27) (like the previous ones in (6) and (7)) by inferring the 736 negative meaning of n-words1, when used as focused answers, from their being 737 under the scope of a negative logical operator coming from the background 738 discourse. In this way the single negation interpretation for isolated argumental n- 36 The claim that this entailment is not part of the lexical endowment of the n-word is due to the fact that n-words of the type PI (n-word1), mainly in Catalan, can also occur in other polarity contexts (interrogatives, conditionals) where the inferred quantity(x) might be 0, 1 or more than 1. (i) a. Has vist res? have.2SG seen anything ‘Did you see anything?’ b. Si veus res, avisa’m. if see.2SG anything tell.me ‘If you see anything, let me know.’ 35 739 words as fragment answers to negative wh-questions is semantically predicted to be 740 possible, although it is not the output of syntactic NC. 741 Suppose, by contrast, that the wh-domain with respect to which the answer to 742 the negative wh-question in (27) is computed is that given in (28b). In that case, the 743 logical representation corresponding to the 〈B,F〉 information structure would 744 require a more elaborate form in which the F part (the generalized quantifier) is the 745 functor, and the B part (the structured property corresponding to the wh-question) is 746 the argument. In (31) we use script ℘ as a symbol for such higher-order 747 generalized quantifiers that take the B part as their argument. 748 (31) λ℘.℘{λx[¬ATE(DESSERT)(x)]} 749 By applying the contents of the negative existential quantifier to this formula, the 750 semantic derivation is as shown in (32). 751 (32) λ℘.℘{λx[¬ATE(DESSERT)(x)]} (λP¬∃y[P(y)]) 752 = λP¬∃y[P(y)]{λx[¬ATE(DESSERT)(x)]} 753 = ¬∃y[λx[¬ATE(DESSERT)(x)](y)] 754 = ¬∃y[¬ATE(DESSERT)(y)] 755 In this case the B part contains a logical negative operator, and the F contains a 756 logical negative operator too, which is overtly expressed in the case of English 757 nobody: ¬body but covertly expressed in the case of Spanish (and Catalan) n- 758 words2, when conceived as negative existential quantifiers such as nadie2: 759 λP¬∃x[P(x). The output interpretation of such an Answer, given the computation 760 in (32), conveys compositionally a DN reading, which is inferred only by some 761 segment of the population of the languages we here describe (see Section 1.2 762 above). 36 763 To sum up, in this section we have shown that argumental n-words used as 764 fragment answers to negative wh-questions impose special restrictions on Question- 765 Answer pairs, depending on whether their lexical semantics is that corresponding to 766 indefinite polarity variables x[+ ] of type 〈e〉, or to indefinite negative existential σ 767 quantifiers λP¬∃x[P(x)] of type 〈〈e,t〉t〉. We have argued that isolated argumental 768 n-words that convey a single negation reading correspond to n-words1, whereas 769 those that convey a DN reading correspond to n-words2. That is, for those speakers 770 that treat n-words as indefinite PIs (the n-word1 variant), isolated argumental n- 771 words used as fragment answers to negative wh-questions are interpreted as 772 conveying a single negation reading, since n-words contribute a variable in F that 773 cannot cancel the logical negative operator in B, but can entail a numeral zero 774 interpretation. Conversely, for those speakers that treat n-words as negative 775 quantifiers, isolated argumental n-words (in the competing n-word2 variant) are 776 interpreted as conveying a DN reading when they occur as answers to negative wh- 777 questions. That is, when used as fragment answers to negative wh-questions, only 778 n-words2 contribute a negative operator in F that cancels the logical negative 779 operator in B. 780 781 5. Conclusions 782 In this article our point of departure was the empirical fact that in Catalan and 783 Spanish two possible interpretations may be attributed to isolated argumental n- 784 words when used with a falling boundary tone (i.e., the unmarked one) as fragment 785 answers to negative wh-questions (Espinal et al. 2015). 786 We have argued that n-words come in two types: the PI variant, which may 787 acquire a formal [uNEG] feature when it occurs in negative syntactic contexts, 37 788 yielding a NC structure, and the negative quantifier variant, which does not 789 participate in NC structures. We have semantically characterized n-words1 as 790 variables carrying a polarity-sensitive formal feature (i.e., x[+ ]), and n-words2 as σ 791 negative generalized quantifiers (i.e., λP¬∃x[P(x)]). 792 We have argued that a population of speakers have access to the n-word1 793 variant; this variant is licensed in polarity contexts, and in NC configurations 794 mediated by syntactic Agree. Another population of speakers have access to the n- 795 word1 variant but also to a competing n-word2 variant. This variant does not 796 participate in NC structures, but it guarantees a possible DN reading when n-words 797 combine with a second negative logical operator, triggered by either a sentential 798 negative marker or an additional negative quantifier. 799 We have also argued that isolated n-words are in Focus when they are used as 800 fragment answers, no matter whether they correspond to indefinite variables or 801 indefinite negative quantifiers, and no matter whether the final interpretation is 802 single negation or DN. 803 We have discussed the shortcomings of both syntactic and semantic ellipsis 804 accounts when it comes to explaining the possibility of a single negation 805 interpretation for isolated n-words when used as fragment answers to negative wh- 806 questions in Catalan and Spanish. Negative wh-questions express inner negation, 807 and convey a backgrounded attribute against p. With respect to this non-neutral 808 biased context, an n-word answer is congruent since it is an element of the domain 809 of the wh-question. The n-word answer indicates the existence of a set of 810 alternatives of the denotation, either a set of entities of type 〈e〉 or a set of 811 generalized quantifiers of type 〈〈e,t〉t〉. 38 812 Finally, following a Structured Meaning approach to the semantics of 813 Question-Answer pairs, we have presented a new account of the two possible 814 readings that native speakers of Catalan and Spanish assign to isolated argumental 815 n-words used as answers to negative wh-questions. We have correlated the two 816 lexical characterizations of n-words with the two possible final interpretations that 817 can be inferred, namely the single negation reading (which combines the logical 818 negative operator in the wh-question with the n-word1 variant), and the 819 compositional DN reading (which follows from the cancellation of a logical 820 negative operator in the wh-question by the negative quantifier variant of an n- 821 word2). Furthermore, we have shown that the possible single negation reading 822 assigned to isolated n-words1 is not the output of syntactic NC but rather an 823 inference driven when these items are interpreted against background negative wh- 824 questions that are requests regarding a negative proposition. 825 826 References 827 Asher, N., Reese, B., 2005. Negative bias in polar questions. In: Maier, E., Bary, 828 C., Huintik, J. (Eds.), Proceedings of SuB9, pp.30-43. 829 Baltazani, M., 2006. Intonation and pragmatic interpretation of negation in Greek. 830 Journal of Pragmatics 38(10), 1658-1676. 831 Beghelli, F., Stowell, T., 1997. Distributivity and negation: the syntax of each and 832 every. In: Szabolcsi, A. (Ed.), Ways of Scope Taking. Kluwer, Dordrecht, pp. 833 71-107. 834 Biberauer, T., Zeijlstra, H., 2012. Negative concord in Afrikaans: filling a 835 typological gap. Journal of Semantics 29, 345-371. 39 836 Bobalijk, J. D., 1995. Morphosyntax: The syntax of verbal inflection. Ph.D. thesis. 837 Massachusetts Institute of Technology. 838 Bobalijk, J. D., 2002. A-chains at the PF-interface: Copies and ‘covert’ movement. 839 Natural Language and Linguistics Theory 20(2), 197-267. 840 Bobalijk, J. D., Wurmbrand, S., 2012. Word order and scope: Transparent 841 interfaces and the ¾ signature. Linguistic Inquiry 43(3), 371-421. 842 Bosque, I., 1980. Sobre la Negación. Cátedra, Madrid. 843 Büring, D., Gunlogson, C., 2000. Aren’t positive and negative polar questions the 844 same?. Ms. UCSC/UCLA. Available at http://www. semanticsarchive.net. 845 Chafe, W. L., 1976. Givenness, contrastiveness, definiteness, subjects, topics and 846 point of view. In: Li, C. N. (Ed.), Subject and Topic. Academic Press, New 847 York, pp. 27-55. 848 Chierchia, G., 2006. Broaden your views: Implicatures of domain widening and the 849 ‘logicality’ of language. Linguistic Inquiry 37(4), 535-590. 850 Chomsky, N., 1995. The Minimalist Program. The MIT Press, Cambridge, MA. 851 Chomsky, N., 2000. Minimalist inquiries: the framework. In: Martin, R., Michaels, 852 D., Uriagereka, J. (Eds.), Step by step: Essays on Minimalist Syntax in Honor of 853 Howard Lasnik. The MIT Press, Cambridge, MA, pp. 89-155. 854 Chomsky, N., 2001. Derivation by phase. In: Kenstowicz, M. (Ed.), Ken Hale: a 855 Life in Language. The MIT Press, Cambridge, MA, pp. 1-53. 856 Cinque, G., Rizzi, L., 2008. The cartography of syntactic structures. In: Moscati, V. 857 (Ed.), CISCL Working Papers on Language and Cognition, 2, 43-59. 858 Cohen, A., Erteschik-Shir, N., 2002. Topic, focus, and the interpretation of bare 859 plurals. Natural Language Semantics 10, 125-165. 860 Cohen, A., Krifka, M., 2011. Superlative quantifiers as modifiers of meta- speech 40 861 acts. The Baltic International Yearbook of Cognition, Logic and Communi- 862 cation 6, 1-56. 863 Corblin, F., 1995. Compositionality and complexity in multiple negation. Logic Jnl 864 IGPL 3, 449-471. 865 Corblin, F., 1996. Multiple negation processing in natural languages. Theoria 17, 866 214-259. 867 Corblin, F., Tovena, L., 2003. L’expression de la négation dans les langues 868 romanes. In: Godard, D. (Ed.), Les Langues Romanes. Problèmes de la phrase 869 simple. CNRS, Paris, pp. 281-343. 870 Déprez, V., 1992. Raising constructions in Haitian Creole. Natural Language and 871 Lingusitic Theory 10, 191-231. 872 Déprez, V., 1997. Two types of Negative Concord. Probus 9, 103-143. 873 Déprez, V., 2000. Parallel (a)symmetries and the structure of negative expressions. 874 Natural Language and Linguistic Theory 18(2), 253-342. 875 Déprez, V., Tubau, S., Cheylus, A., Espinal, M.T., 2015. Double Negation in a 876 Negative Concord language: An experimental investigation. Lingua 163, 75- 877 107. 878 Embick, D., Noyer, R., 2007. Distributed Morphology and the syntax-morphology 879 interface. In: Ramchand, G., Reiss, Ch. (Eds.), The Oxford Handbook of 880 Linguistic Interfaces. Oxford University Press, Oxford, pp. 289-324. 881 Espinal, M. T., 2000. On the semantic status of n-words in Catalan and Spanish. 882 Lingua 110, 557-580. 883 Espinal, M. T., 2002. La negació. In: Solà, J., Lloret, M. R., Mascaró, J., Pérez 884 Saldanya, M. (Eds.), Gramàtica del Català Contemporani. Empúries / Grup 62, 885 Barcelona, pp. 2687–2757. 41 886 Espinal, M. T., 2007. Licensing expletive negation and negative concord in Catalan 887 and Spanish. In : Floricic, F. (Ed.), La Négation Dans les Langues Romanes. 888 John Benjamins Linguisticae Investigationes Supplementa 26, Amsterdam, pp. 889 49-74. 890 Espinal, M. T., Prieto, P., 2011. Intonational encoding of double negation in 891 Catalan. Journal of Pragmatics 43, 2392–2410. 892 Espinal, M. T., Tubau, S., Borràs-Comes, J., Prieto, P., 2015. Double negation in 893 Catalan and Spanish. Interaction between syntax and prosody. In: Larrivée, P., 894 Lee, C. (Eds.), Negation and Polarity: Cognitive and Experimental Perspectives. 895 Springer, Berlin, pp. 145-176. 896 Espinal, M. T., Tubau, S., In press. Meaning of words, meaning of sentences. 897 Building the meaning of n-words. In: Fischer, S., Gabriel, C. (Eds.), 898 Grammatical Interfaces in Romance Linguistics. Mouton de Gruyter, Berlin. 899 Fabra, P., 1956. Gramàtica Catalana. Teide, Barcelona. 900 Falaus, A. M., 2007. Le paradoxe de la double négation dans une langue à 901 concordance négative stricte. In: Floricick, F. (Ed.), La Négation Dans les 902 Langues Romanes. John Benjamins, Linguisticae Investigationes Supplementa 903 26, Amsterdam, pp. 75-97. 904 Giannakidou, A., 1998. Polarity Sensitivity as (Non)Veridical Dependency. John 905 Benjamins, Amsterdam. 906 Giannakidou, A., 2000. Negative ...Concord?. Natural Language and Linguistic 907 Theory 18, 457-523. 908 Giannakidou, A., 2006. N-words and Negative Concord. In: Everaert, M., van 909 Riemsdijk, H. (Eds.), The Blackwell Companion to Syntax. Blackwell, Oxford, 910 pp. 327-391. 42 911 Geurts, B., 1998. The Mechanisms of denial. Language 74, 274-307. 912 Godard, D., Marandin, J.M., 2007. Aspects pragmatiques de la négation renforcée 913 en italien. In: Floricic, F. (Ed.), La Négation Dans les Langues Romanes. John 914 Benjamins Linguisticae Investigationes Supplementa 26, Amsterdam, pp. 137- 915 160. 916 Haegeman, L., 1995. The Syntax of Negation. Cambridge University Press, 917 Cambridge. 918 Haegeman, L., Zanuttini, R., 1991. Negative heads and the Neg-Criterion. The 919 Linguistic Review 8, 233–252. 920 Haegeman, L., Zanuttini, R., 1996. Negative Concord in West Flemish. In: Belletti, 921 A., Rizzi, L. (Eds.), Parameters and Functional Heads. Essays in Comparative 922 Syntax. Oxford University Press, Oxford, pp. 117-179. 923 Haegeman, L., Lohndal, T., 2010. Negative Concord and multiple Agree: a case 924 study of West Flemish. Linguistic Inquiry 41(2), 181-211. 925 Halle, M., Marantz, A., 1993. Distributed Morphology and the pieces of inflection. 926 In: Hale, K., Kayser, S. J. (Eds.), The View From Building 20: Essays in 927 Linguistics in Honor of Sylvain Bromberger. The MIT Press, Cambridge, MA, 928 pp. 11-176. 929 Han, C., 1999. The structure and interpretation of imperatives: Mood and Force in 930 Universal Grammar. Ph.D. thesis. University of Pennsylvania. 931 Herburger, E., 2001. The Negative Concord puzzle revisited. Natural Language 932 Semantics 9, 289-333. 933 Holmberg, A., 2013. The syntax of answers to polar questions in English and 934 Swedish. Lingua 128, 31-50. 43 935 Horn, L. R., 1989. A Natural History of Negation. University of Chicago Press, 936 Chicago. 937 Huddlestone, K. M., 2010. Negative indefinites in Afrikaans. Ph.D. Thesis, Utrecht 938 University. 939 Jespersen, O. 1917., Negation in English and Other Languages, reprinted in 940 Selected Writings of Otto Jespersen (1962). George Allen and Unwin, London, 941 pp. 3-351. 942 Kartunnen, L. 1977., Syntax and semantics of questions. Linguistics and 943 Philosophy 1, 3-44. 944 Kiss, K. É., 1998. Identificational focus versus information focus. Language 74(2), 945 245-73. 946 Krifka, M., 2001. For a structured meaning account of questions and answers. In: 947 Féry, C., Sternefeld, W. (Eds.), Audiatur vox sapientia. A Festschrift for Arnim 948 von Stechow. Akademie Verlag, Berlin, pp. 287-319. 949 Krifka, M., 2007. Basic notions of Information Structure. In: Féry, C., Fanselow, 950 G., Krifka, M. (Eds.), Interdisciplinary Studies on Information Structure. 951 Universitäts verlag Potsdam, Potsdam, pp. 13-55. 952 Krifka, M., 2008. The semantics of questions and the focusation of answers. In: 953 Lee, C., Gordon, M., Büring, D. (Eds.), Topic and Focus: a Cross-linguistic 954 Perspective. Kluwer, Dordrecht, pp. 139-151. 955 Krifka, M., 2011. Questions. In: von Heusinger, K., Maienborn, C., Portner, P. 956 (Eds.), Semantics. An International Handbook of Natural Language Meaning, 957 vol. 2. Mouton de Gruyter, Berlin, pp. 1742-1758. 44 958 Krifka, M., To appear. Negated polarity questions as denegations of assertions. In: 959 Kiefer, F., Lee, C. (Eds.), Contrastiveness and Scalar Implicatures. Springer, 960 Dordrecht. 961 Kroch, A., 2000. Syntactic change. In: Baltin, M., Collins, C. (Eds.), The 962 Handbook of Contemporary Syntactic Theory. Blackwell, Oxford, pp. 699-729. 963 Labelle, M., Espinal, M. T., 2014. Diachronic changes in negative expressions: the 964 case of French. Lingua 145, 194-225. 965 Ladd, R. D., 1981. A first look at the semantics and pragmatics of negative 966 questions and tag questions. Proceedings of Chicago Linguistic Society 17, 164- 967 171. 968 Ladusaw, W.A., 1992. Expressing Negation. In: Barker, C., Dowty, D. (Eds.), 969 SALT II. Ithaca, New York, Cornell Linguistic Circle. 970 Laka, I., 1990. Negation in Syntax: On the Nature of Functional Categories and 971 Projections. Ph.D. Thesis. Massachusetts Institute of Technology. 972 Longobardi, G., 2014. Theory and experiment in parametric minimalism. The case 973 of Romance negation. In: Pensalfini, R., Turpin, M., Guillemin, D. (Eds.), 974 Language Description Informed by Theory. John Benjamins, Amsterdam, pp. 975 217-261. 976 Merchant, J., 2001. The Syntax of Silence: Sluicing, Islands, and the Theory of 977 Ellipsis. Oxford University Press, Oxford. 978 Merchant, J., 2004. Fragments and ellipsis. Linguistics and Philosophy 27, 661- 979 738. 980 Merchant, J., Frazier, L., Weskott, T., Clifton Jr., C., 2013. Fragment answers to 981 questions: a case of inaudible syntax. In: Goldstein, L. (Ed.), Brevity. Oxford 982 University Press, Oxford, pp. 21-35. 45 983 Molnár, V., 1998. Topic in focus. On the syntax, phonology, semantics and 984 pragmàtics of the so-called ‘contrastive topic’ in Hungarian and German. Acta 985 Linguistica Hungarica 45(1-2), 89-166. 986 Penka, D., 2011. Negative Indefinites. Oxford University Press, Oxford. 987 Penka, D., Zeijlstra, H., 2005. Negative indefinites in Dutch and German. Ms. 988 University of Tuebingen. 989 Penka, D., Zeijlstra, H., 2010. Negation and polarity: an introduction. Natural 990 Language and Linguistics Theory 28(4), 771-786. 991 Prieto, P., Borràs-Comes, J., Tubau, S., Espinal, M. T., 2013. Prosody and gesture 992 constrain the interpretation of double negation. Lingua 131, 136-150. 993 Puskás, G., 2006. Double negation and information structure: somewhere between 994 Topic and Focus. In: Molnár, V., Winkler, S. (Eds.), The Architecture of Focus. 995 Mouton de Gruyter, Berlin, pp. 291-317. 996 Real Academia Española (RAE). 2009. Nueva Gramática de la Lengua Española. 997 Espasa Calpe, Madrid. 998 Reese, B. J., 2006. The meaning and use of negative polar interrogatives. In: 999 Bonami, O., Cabredo-Hofherr, P. (Eds.), Empirical Issues in Syntax and 1000 Semantics 6, 331-354. 1001 Rizzi, L., 1997. The Fine Structure of the Left Periphery. In Haegeman, L. (Ed.), 1002 Elements of Grammar. Kluwer, Amsterdam, pp. 281-337. 1003 Romero, M., Han, C., 2004. On negative yes/no questions. Linguistics and 1004 Philosophy 27(5), 609-658. 1005 Sánchez, C., 1999. La negación. In Bosque, I., Demonte, V. (Eds.), Gramática 1006 Descriptiva de la Lengua Española. Espasa, Madrid, pp. 2561-2630. 46 1007 Solà, J., 1973. La negació. Estudis de Sintaxi Catalana, vol. 2. Edicions 62, 1008 Barcelona, pp. 87−118. 1009 von Stechow, A., 1991. Focusing and backgrounding operators. In: Abraham, W. 1010 (Ed.), Discourse Particles. John Benjamins, Amsterdam, pp. 37-84. 1011 de Swart, H., 2010. Expression and Interpretation of Negation. An OT typology. 1012 Springer, Dordrecht. 1013 Szabolcsi, A., 1981. The semantics of topic-focus articulation. In: Groendendijk, J., 1014 Janssen, T., Stokhof, M. (Eds.), Formal Methods in the Study of Language. 1015 Mathematisch Centrum, Amsterdam, pp. 513-541. 1016 Tomioka, S., 2010. Contrastive topics operate on speech acts. In: Zimmermann, M., 1017 Féry, C. (Eds.), Information Structure: Theoretical, Typological and 1018 Experimental Perspectives. Oxford University Press, Oxford, pp. 115-138. 1019 Vallduví, E., 1993. Catalan as VOS: evidence from information packaging. In: 1020 Ashby, W. J., Mithun, M., Perissinotto, G., Raposo, E. (Eds.), Linguistic 1021 Perspectives on the Romance Languages. John Benjamins, Amsterdam, pp. 1022 335−350. 1023 Vallduví, E., 1994. Polarity items, n-words, and minimizers in Catalan and Spanish. 1024 Probus 6, 263-294. 1025 Veloudis, I., 1982. Negation in Modern Greek. PhD Dissertation. University of 1026 Reading. 1027 Vinet, M-T., 1998. Contrastive focus, n-words and variation. Revue Canadienne de 1028 linguistique 43(1), 121-141. 1029 Watanabe, A., 2004. The genesis of negative concord: syntax and morphology of 1030 negative doubling. Linguistic Inquiry 35, 559-612. 47 1031 van der Wouden, T., Zwarts, F., 1993. A semantic analysis of negative concord. In: 1032 Lahiri, U., Wyner, A. (Eds.), Semantics and Linguistic Theory (SALT) III. 1033 Cornell University, Ithaca, NY, pp. 202-219. 1034 Zanuttini, R., 1991. Syntactic properties of sentential negation: A comparative 1035 study of Romance languages. Ph.D. Thesis. University of Pennsylvania. 1036 Zanuttini, R., 1997. Negation and verb movement. In: Haegeman, L. (Ed.), The 1037 New Comparative Syntax. Longman, New York, pp. 214-245. 1038 Zeijlstra, H., 2004. Sentential negation and Negative Concord. Ph.D. Thesis. 1039 University of Amsterdam. 1040 Zeijlstra, H., 2011. On the syntactically complex status of negative indefinites. 1041 Journal of Comparative Linguistics 14, 111-138. 1042 Zeijlstra, H., 2012. There is only one way to agree. The Linguistic Review 29, 491- 1043 539. 48