Viviane Déprez led the third seminar with a presentation entitled Negative Concord and the Syntax of N-words. Her approach focuses on the internal micro-parametric syntax of N-expressions as the source for variation in concord relations rather than on the nature sentential negation. Concord readings can stem on this view from at least two distinct semantic operations: resumptive quantification between negative quantifiers (Déprez 1997, following May 1979; de Swart and Sag 2002), or variable binding by a negative operator (Ladusaw, Giannakidou), with N-expressions interpreted as indefinite variables. A central factor determining the choice of the semantic operation is the internal structure of N-expressions. Only N-expressions that are quantificational and occupy a high position in the DP allow resumptive quantification concord. This is standard French concord in sentences like Personne n’a rien vu, with French N-words distinguished from NPIs by the fact that they allow modification by a-adverbs (almost/absolutely), license concord under locality conditions, have negative readings in many polarity contexts, obligatory double negation readings when used with sentential negation and a possible one when in pairs. Variable binding occurs with structural dependency of N-expressions on negative operators, which arises when the N-expression features a null determiner-like internal category. N-words in Martinique and Haitian Creole resemble bare nouns with a null determiner (Déprez 2005); they contrast with French ones in requiring sentential negation, except in certain NPI contexts, never allowing double negation readings with it or among each-other, but allowing concord at long distance. The distinct internal structures of N-words are revealed through modification. Regular determiners and adjectival modification, whether pre-nominal or post-nominal or with autre, is prohibited with contemporary French N-words rien or personne. These further disallow cliticisation with en (*J’en ai vu personne cf *J’en ai vu quelqu’un vs J’en ai vu quelques-uns) and are devoid of characteristic nominal features such as gender or number. Following recent works on DP structure (Zamparelli 1995 among others), these properties strongly suggest that such N-words occupy a high position in their functional D-structure and are quantificational and pro-nominal rather than noun-like in nature. In contrast, Creole anyen and pèsonn allow regular post-nominal modification and are plural, displaying in this regards relevant noun-like properties. Akin to the alternating positions of quantifiers with strong and weak readings, the position of N-words is taken to be able to alternate within their functional structure to account for variability in their concord properties, in micro-parametric fashion. This allows an explanation of the peculiar behaviour of N-expressions that behave differently within a particular (variety of a) language. An example provided by Déprez is Martinique Creole janm which contrary to anyen or pèsonn, can be used without pa, and allow double negation readings in some cases. Alternating structures appears to be prominent in Quebecois French where N-words seem to behave in some cases quite similarly to the standard French ones, while in other cases, as in J’ai pas vu personne when they concord with negation pas, they display NPI like characteristics: long distance dependency, positive interpretation in NPI environment, no double negative reading. Syntactic differences to European French would include the possible use of pas personne in subject position, the possibility of floating personne in cases like Ils sont pas venus personne. The rather notable variation between speakers – one informant (Auger) reports co-occurrence of pas with personne but not with rien – provides additional support to a micro-parametric approach. A case of possible micro-parametric difference across varieties of French was discussed: N-words behave like NPIs in the context of the preposition sans where as Déprez notes, noun phrases with null determiners are still allowed, furthering the correlation between bare noun structures and NPI behaviour.
Déprez draws support for a micro-parametric analysis based on the internal structure of N-expressions from her analysis of the historical evolution of French N-words in the Frantext corpus data base. The evolution of aucun involves the loss of feature variablitity, changes in possible determination and position (sans raison aucune) along with an increasing negative interpretation. Interpretation, features and modification (particularly by autre) provide potential criteria for changes of status. Typical nominal features and pre-determiners are lost for rien by the first half of C16 according to Frantext data. As discovered by Déprez, patterns of modification for rien and personne particularly by autre went from prenominal up to C16, to a direct post-nominal modification (rien autre) in C17-18 and from C19 to complex post-nominal modification introduced by de (rien d'autre). She proposes to model these changes as a gradual move of the N-word up to the edge of their D-projection and argues that it was not until the N-words had reached the upper edge of their internal functional projection that they became negative in nature. This would mean that interpretable features of N-expressions becoming visible to the semantic module when reaching the syntactic edge of the projection. On this view, the emergence of the negative meaning of N-words is a case of Visibility at the Edge (Déprez 2006 for other such cases based on the grammaticalization of creole determiners), subject to Chomsky’s Phase Impenetrability Condition.
The discussion was led by Maj-Britt Mosegaard Hansen. She raised the question of which variety of European French was being described. Cases of polarity uses were evidenced by written data from Frantext and spoken Elicop data. The general observation is that even when fully negative, N-words can be used in strong NPI contexts, such as comparatives; this is the case for Portuguese N-words that are otherwise not normally found in other polarity environments (Anne Breitbarth). A related issue is the use of pas with N-words, and it was discussed whether the ban on this in European French is a normative prescription or a grammatical requirement. This raises the question of the potential disjunction between grammatical representations and sociolinguistic ones, and child language data might help to adjudicate the point (Eric Haeberli). The question of criteria was discussed, and it was agreed that the a-adverbs test was not fully reliable. Locality probably is, despite the fact that it is infringed in denials (Je n'ai demandé qu'ils arrêtent personne, voyons!) and in Quebecois as well as Catalan. The claim made following varying informant judgments that Quebecois does not allow double negation is at odds with attested examples. Whether speaker judgements are the best data to work from was considered.
The date of the historical change was envisaged. Richard Ingham noted that if one uses a wider corpus than Schnedecker and Prevost's 2002 paper, it can be seen that nul starts to give ground to aucun in non-affirmative clauses as early as the C13-14, when null determiners were still normal in French. So the association sought by Déprez and Martineau between the modifier stage of aucun and null determiners looks plausible, in fact more so than if thought to have been a development of the C16-17, when null determiners were beginning to be lost. Pierre Larrivée raised the issue of N-word néant, which is negative very early on despite being a noun without a determiner. This goes to the heart of the hypothesis that the change of category to pronoun and raising to D was a necessary component of N-words.