Séminaire de Recherche en Linguistique

Introduction: Negative concord (NC) languages are commonly classified in four distinct types, each assumed to exemplify a specific parametric choice: (Zeijlstra 2004, Biberauer & Zeijlstra 2012) 1) languages disallowing NC (English, Dutch), 2) languages generally enforcing Neg-doubling, i.e. strict NC (Czech, Romanian), 3) languages only enforcing Neg-doubling for post-verbal positions, i.e. non-strict NC (Italian, Spanish), and 4) languages disallowing Neg-doubling but allowing NC between NC-Items (NCI) (French). Here we investigate unexpected variability in otherwise solid strict NC languages like North Eastern Italian dialects and French Based Creoles that display obligatory Neg-doubling with post-verbal NCIs and variable situations with preverbal NCIs, with some requiring, others tolerating, and yet others banning Neg-doubling sometimes in the same language. Current models of NC are unable to capture this kind of variability as they rely on static and rigid classifications of NCI and negation types. We argue that a dynamic approach of what it means to be ‘negative’ is required to capture this unexpected observed variability while preserving previously reached generalizations.

The data: Venetian dialects manifest the following implicational scale with respect to obligatory doubling with preverbal NCI, a scale also valid for diachronic variation (Poletto 2017)

(1) A. negative coordination & negative adverbs >  B. focused direct objects >  C. subjects

          Doubling obligatory    ----------------------------------------------à      optional/impossible

That is, some dialects exhibit obligatory Neg-doubling with preverbal NCIs only when these are negative coordinations né and negative adverbs (never, not.even, not.yet), others when they are negative adverbs as well as focused objects, and yet others extend this obligatory doubling pattern also to subjects. But dialects that enforce doubling with subjects but not with focused objects or adverbs are not found. This same scale also applies to cases of optionality, so that, for instance in Zemignanese doubling with preverbal subjects is optional, yet obligatory with focused objects and adverbs. One might wonder about the foundations of such a scale, especially given the fact that the distribution of doubling appears quite different in French Based creoles (FBC), where the implicational scale appears rather to go in the opposite direction. In some FBC indeed, Déprez (2017) observed that preverbal NCI adverbs like ‘zamen’ are the most likely NCI type to occur without doubling, followed by NCIs meaning ‘nothing’, commonly argumental objects, and NCIs meaning nobody, commonly subjects. A similar scale with respect to doubling was also observed in the Quebecois dialect of Montreal by Burnett, Tremblay, Blondeau (2015).

(2)    Person > Nanye  > Zamen > (Mauritian Creole)

         No one  > nothing   > never

         Doubling obligatory----------------------------------------------àoptional/impossible

While the dimensions of the scales are not entirely parallel in both cases, more syntactic for the Venetian dialects, and more lexical for the French based creoles, the diametrical opposition in the Neg-doubling behavior of preverbal NCI adverbs is particularly striking.

Analysis: Our research questions thus are: a) what is the property triggering Neg-doubling with preverbal NCIs? b) What is the source of variability inside one implicational scale and across scales, i.e. why are Italian dialects and French creoles so different in appearance? c.) Why is comparable variability never observed in post-verbal position?

Contra Zeijlstra and others, we take core sentential negation markers to always be semantically negative and like other scope bearing elements to be interpretable in positions other than their surface ones. Obviating the need for an abstract negative operator, this commits us to the view that whenever sentential negation appears, it is interpreted. Extending an approach first developed in Déprez (2011,2017), we argue that the negative force of an NCI is dynamically computed as an interaction between its internal structure and the syntactic position it comes to occupy in a sentence. The core principle of this dynamic computation is given in (3):

(3) Negative features are only Visible [= semantically computable] at phase edge.

When not Visible, negative features are derivational time bombs (Preminger 2011) that must be defused/licensed, here by negation, to allow proper semantic interpretation.

To illustrate this dynamic interplay, consider first elements whose negative feature occurs inside their containing constituent, not at its edge and is hence ‘in-Visible’: i.e. NCI with a structure as in (4) 

(4) [_XP….[.. F_NEG]] 

NCIs like (4), with FNeg not located at the edge, and hence in-Visible, always require licensing, achievable under two conditions: either negation scopes directly over them or they reconstruct under its scope. The consequence for preverbal NCI of this type, is that obligatory Neg-doubling is an effect of reconstruction. On standard assumptions about reconstruction, only NCI occurring in A’ positions can reconstruct into their post-verbal position and thus have the same NC properties as post-verbal ones. In A positions, with reconstruction arguably unavailable, the interplay must be different. As is well known, languages can differ as to whether or not their preverbal subject position is an A or an A’-position and as to whether negation can scope over it or not. On our view, Neg-doubling with subjects like (4) is hence predicted to manifest variations along these lines.  Consider next elements whose negative feature occurs at the edge of their containing domain as in (5), possibly a phase, like DP, and are hence Visible:

(5). [_XP F_Neg  […. ]]

In this case, Fneg is Visible at the XP edge. With Chomsky’s (2001) Phase Impenetrability Condition (PIC), however, if NCI like (5) are merged inside a VP (objects or adverbs), their NegF will remain invisible at the sentential level and require licensing. When re-merged at the edge of a higher domain, say TP for subjects or CP for adverbs (simplifying here the complex structure these domains could have (Rizzi’s,Cinque’s)), the NegF of NCIs like (5) will be directly computable at phase edge, and hence no longer require negation; Lack of doubling is predicted as these NCIs are negative by themselves. This is the baseline of our proposed dynamic interaction between the inner structure of NCIs and their external positions for the Visibility of their NegF. Further predictions with respect to NC systems can be calculated from differences in NCI internal structure (i.e. possible internal movement of NegF under focus) and external positions. Taking Venetian adverbs to have a structure like (4) and to occur in A’ position predicts obligatory doubling as an effect of reconstruction. The opposite obtains for the preverbal creole adverbs: with an inner structure like (5), they manifest negative force and repel doubling. Subject NCIs occur in A or A’ positions and may have variable internal structures, which predicts variable Neg-doubling even within the same language, and also allows to model optionality. Able to account for the variability observed with preverbal NCIs, and the seemingly contradictory scales observed in our data, our dynamic model of negative interpretation further straightforwardly derives the empirical generalization that comparable variability is never observed in the post-verbal domain. As for fragment answers, given that they always occur at edges (Spec CP on Merchant's 2000 analysis) and generally under focus, XP internal Focus movement and their sentential edge position will ensure the Visisbility of their NegF, and their consequent negative interpretation.