ABSTRACT
What are the parameters governing the nature and position of clausal negation? I argue that variation in this domain is largely the reflex of the interaction of three independent factors:
• verb-movement parameters;
• the semantics of clausal negation;
• the basic condition that scope relations depend on asymmetric c-command.
The generalisation I would like to pursue here is the following:
(1) Clausal negators can appear in any position from which:
a. they c-command the binder of the event variable (Tense, at least in the cases we’ll look at here);
b. they can be licensed by Pol (Holmberg 2016).
(1) holds in virtue of (2) and (3):
(2) Postulate on the meaning of clausal negation (Partee 1973, Acquaviva 1995, Zeijlstra 2004, 2008, a.o.):
John didn’t leave means “there was no event of John’s leaving”: e[leave(j,e)
(3) The Tense Denotation Hypothesis:
Events are denoted through the Tense feature (i.e. through binding by a Tense- bearing head).
Licensing by Pol involves feature-inheritance: iNEG features are inherently associated with Pol, but may be donated (in the sense of Ouali 2008) to a clausal negator (Biberauer & Roberts 2011:38). Where this happens, we have an iNEG clausal negator, where it does not, we have a uNEG one (see Zeijlstra 2004, 2008, Biberauer & Zeijlstra 2012a,b on the implications of i/uNEG for negative concord). Interpretable NEG-features are interpreted as e and hence must have e in their scope, and hence must asymmetrically c-command the Tense-bearing head, by (3).
Hence we predict (i) that clausal negation cannot be higher than Pol, i.e. in the left periphery (unless it moves there with the finite verb, as in V2), and (ii) that clausal negation cannot precede a “high” subject in an SVO language. But where the subject is one notch lower, as in VSO Welsh and Irish, negation can precede it. Furthermore, the lowest possible position for negation will be vEAP, the vP which c-selects vEvP containing the event variable e, assuming the following structure for “IP”:
(4) [ M T A [VoiceP Voice [vEAP EA vEA [vEvP e vEv [v..P … IAs … [RootP .. “V” .. ]]]]]]
Since finite elements are always licensed higher than vEv (as they have to c-command it) and must either be in or have access to the TMA field, we can exclude positions inside the complement of the lower phase head (Voice, I assume). Hence, clausal negators must appear in the “structural window” boldfaced in (5):
(5) CP … [PolP Pol [PersP Pers [TMA …. [VoiceP Voice [vEAP [vEvP e ….
Further, I assume (6):
(6) Clausal negation must either c-command or combine with the finite head.
Finally, clausal negators are Negmin/max, i.e. they do not project/label and have no complements.
Most of the talk is devoted to showing how these assumptions pan out cross-linguistically. I will illustrate for “core” Romance (Standard Italian, etc.), Irish and Welsh, Haitian and Cape Verdean Creole, English, Finnish, French, Piedmontese, Afrikaans, West Flemish, and “core” V2 Germanic. Two very simple parameters emerge:
(7) i. is the negator a purely minimal category?
ii. if not, does it have a V-feature?
If (7i) is positive, then the negator is first-merged directly into the clausal spine and so has a V-feature by definition; if (7ii) is positive, the negator merges as a Specifier and then incorporates into the spine; if (7ii) is negative, the negator merges as a Specifier and cannot incorporate into the spine. The choice of external merge into the spine or internal merge into the spine reduces to a purely morphological one (free vs bound morpheme). Then the only syntactic option open to a clausal negator is (7ii), determining whether it interacts structurally with the clausal spine.