Wednesday, September 25, 2013

Is Negation a Head?

A common idea (dating back at least to Chomsky 1957) is that do-support is triggered in English when adjacency between Tense and the verb is disrupted. The idea is that negation comes between Tense and V, blocking morphological merger of the two, and so do-support is required to support the tense affix:

(1)*Leprechauns not eat Lucky Charms.
(2) Leprechauns do not eat Lucky Charms.

An obvious problem for the adjacency account is that adverbs like often do not seem to disrupt the necessary adjacency:

(3) Leprechauns often eat Lucky Charms.
(4)*Leprechauns do often eat Lucky Charms.

In contrast, in other cases where adjacency is necessary, adverbs do count. For instance, verbs and objects must be adjacent in English, as must fronted auxiliaries and subjects:

(5)*Leprechauns eat often Lucky Charms.
(6)*Do often leprechauns eat Lucky Charms?

It is therefore doubtful that surface adjacency is really at issue in do-support. However, it has been proposed that the relevant adjacency is structural, and requires a head-complement relation (e.g., Embick and Noyer 2001). When negation is not present, Tense selects VP, and so their heads can merge morphologically. Adverbs, being adjoined to VP, do not disrupt this relation. Negation, on the other hand, is the head of a projection NegP that comes between T and VP, and that projection disrupts the necessary morphological merger, leading to do-support.

The question then is, why should we think that negation is a head that heads a NegP projection? In word order and other respects English negation acts very much like an adverb. I am aware of two arguments that it is a head. The first is that it can apparently license VP ellipsis, like an auxiliary verb. I have shown in other work that this argument does not go through (Bruening 2010). The second is what I want to address in this blog post, namely, that negation can move to C in subject-auxiliary inversion:

(7) Don't leprechauns eat Lucky Charms?

If subject-auxiliary inversion is head movement, and the auxiliary do got to C by moving first to T and then to C, it would seem to be necessary that the n't got picked up along the path of this head movement. That would only make sense if n't was a head Neg (between the starting position of do and T). Adverbs can never be carried along in head movement:

(8)*Do never leprechauns eat Lucky Charms?

There is another way of looking at negation that does not lead to this conclusion, however. Suppose that, universally, negation is actually a feature of clauses, or CPs. A negative clause has a feature [Neg] on the head C, and this feature is distributed (through selection or Agree) throughout the clause, to T and to V. Individual languages then decide how to spell this feature out. Passamaquoddy, for instance, marks it twice, once with a freestanding particle and again with a morpheme on the verb:

(9) Nihtol kete ma te '-kosiciy-a-wi-wa-l tan op wecessi-t.
that.Obv for.example Neg Emph 3-know-Dir-Neg-3P-Obv WH would IC.from.arrive-3Conj
`I mean, nobody knew where he could have come from.'

In Irish, negation appears to be a complementizer (e.g., Chung and McCloskey 1987):

(10) Ní ólann sé bainne ariamh.
Neg.Comp drink.Present he milk ever
`He doesn't ever drink milk.'

It therefore appears quite plausible that negation can be spelled out in various different positions within a clause (CP), most relevantly on C, and it can even be spelled out more than once. This makes sense if negation is a feature of the clause, which can be spelled out in various different ways. Coming back to English, we can describe the facts in the following way: the [Neg] feature needs to be spelled out, and in this language cannot be spelled out more than once. However, it can be spelled out in one of several different ways: it can be spelled out as an adverb adjoined to the complement of Tense (not); it can be spelled out as a clitic (or perhaps affix) to the head Tense (n't); or it can be spelled out as a clitic (or perhaps affix) on the head C (n't). The latter option can only be chosen if T-to-C movement applies; a precondition for cliticization/affixation to a head is that the head be occupied by an auxiliary. If the [Neg] feature is spelled out on C, it cannot be spelled out on T or as an adverb.

In this account, it is not necessary that negation be a head in order to account for its appearance in C. Negation appears in C not because it was picked up by head movement, but because the feature [Neg] originates on C and can be spelled out on C.

This analysis is cross-linguistically plausible, and the existence of a plausible alternative defuses the head-movement argument for treating negation as a head. The account also fits in with the general point of view that do-support is not about adjacency at all, either string adjacency or structural adjacency. (For a theory where do-support is not about adjacency, see Bruening 2010.)

References

Bruening, Benjamin (2010). Language-Particular Syntactic Rules and Constraints: English Locative Inversion and Do-Support. Language 86: 43-84.

Chomsky, Noam (1957). Syntactic Structures. The Hague: Mouton.

Chung, Sandra and James McCloskey (1987). Government, Barriers, and Small Clauses in Modern Irish. Linguistic Inquiry 18: 173-237.

Embick, David and Rolf Noyer (2001). Movement Operations after Syntax. Linguistic Inquiry 32: 555-595.