Some analytical possibilities- Is “left branch” good enough? Will head movement break this?

Now that the proposal has been laid out, it is possible to more directly address a thorny question that has been lingering in the background. The generalization as it was presented early on was about the relative surface order of heads and modifiers. But of course there are a number of structures that can derive any particular order – a modifier might wind up to the right of a head, for example, even though it occupies a left branch. So perhaps it is simplistic to state the generalization as a simple matter of left branches versus right branches, as the proposed rule in (1) in Expressive predicate modification does. A related question, essentially a more specific form of it, is how this all interacts with head movement, which can change the relative surface order of modifier and head.

In fact, though, this kind of characterization may be an advantage. Among its important properties is that it may make it possible to derive the variation

in the form it takes from one language to another via head movement. Returning to Romance for a moment, the generalization there is not that postnominal adjectives must be restrictive, but that prenominal ones must be nonrestrictive. Taking into account the independent fact that Romance nouns move higher in their NP (Bernstein 1993; Cinque 1994; others), at least half this generalization follows. The reason both readings are available in surface postnominal positions is that there are both left and right branches that are spelled out right of the noun.

Moreover, there is actually more flexibility with respect to head movement in (1) than there may seem. It can actually restrict the availability of nonrestrictive interpretations on the basis of the position of the head in a more fine-grained way. More often than not, the position in which heads are interpreted doesn’t matter for the semantics, and usually it’s convenient to assume they are interpreted in their base position. But nothing requires this, and one could equally well suppose that (some) heads are interpreted in their surface position and semantically reconstruct, binding so-called “big” – that is, high-type – traces. In light of (1) and the binding properties of expressive modifiers, this has consequences. The trace of a head inside the scope of an expressive modifier, if bound from outside its scope, would bring about the (independently) ruled-out binding-across-dimensions configuration. Thus, a head can’t bind its trace in the expressive meaning from inside the ordinary meaning dimension. In turn, this and the left-branch requirement together create a system in which these nonrestrictive modifiers can only occur left of wherever a head is interpreted.

Given all this, some subtle predictions arise with respect to English as well. Assuming that verb movement in English is present but short (Johnson 1991) and that verbs are interpreted in their base positions, adverbs in English should admit nonrestrictive readings, even though right of the verb at the surface, if the verb has moved past them. This may be right:

In (1), the verb moves fromthe position indicated by the trace. Consequently, secretly should be able to get a nonrestrictive interpretation.