Types of Parametric Change
Ian Roberts
Downing College, University of Cambridge
 

1. Clark & Roberts (1993), Roberts (1993, 1996, 1997, 1999) and Roberts & Roussou (2000, 2003) all propose that the central mechanism of syntactic change is the loss of movement dependencies. Arguably, this fact follows from the conservative nature of the language acquisition device, which prefers relatively simple representations over relatively more complex ones. If movement creates relatively complex representations (essentially by introducing extra features as triggers), then we see why it is dispreferred and therefore may be diachronically lost if not robustly triggered for a given generation of first-language acquirers.

2. In current syntactic theory, the target position of all movement is a functional category. What is moved may be either lexical or functional. For illustration, let us take all movement to be of the form "L to F". It then follows that, where L-to-F movement is lost, the formerly moved material must suffer one of three fates in the new grammar: (i) it is merged in F, (ii) it is merged in L and remains there, (iii) it disappears completely. Case (i) is grammaticalisation, case (ii) is word-order change and case (iii) is loss. We can contrast cases (i) and (ii) in the obvious way as "upward" and "downward" changes. Then we can observe the following differences:

Upward changes apply just to a morphologically-defined subclass of L, recategorising it as F; these changes are local, unproductive (but sensitive to morphological subregularities), and associated with semantic and phonological "reduction".

Downward changes apply to all L, giving rise to word-order change; these changes are fully productive, and involve no semantic or phonological change to L-roots.

Examples of upward changes are the reanalysis of English modals in the 16th century, the development of Romance future/conditional forms from Latin habere (both V becoming T), the development of negative words from minimisers (N becomes D/Q), the development of articles from demonstratives (A becomes D), etc. Examples of downward changes are the loss of V-to-T, VSO > SVO, OV > VO (cf. Roberts (1997)) and (a case where the moved element is also functional) the loss of V2.

3. A very important aspect of this dichotomy is the fact that upward changes involve changes to interface properties (phonological and semantic reduction), while downward changes do not. I will suggest, following the ideas sketched in Roberts & Roussou (2003, Chapter 5), that this is due to the inherent properties of functional categories. Such categories are defective at the interfaces and hence reanalysis as a functional category involves phonological and semantic reduction. Since downward movement does not require categorial reanalysis as a functional category, no such reduction is observed. I will also suggest that this very defectivity of functional categories underlies the preference for relatively simpler representations, understood as a ban on "extra" features, i.e. as a preference for minimal content. The central concept in this account of types of syntactic change is thus the general idea that functional categories are atomic, in that they preferentially lack structure in syntax, and obligatorily lack it at the interfaces.
 

*****

Remarks on Head Movement
Ian Roberts
Downing College, University of Cambridge
 

Chomsky (2001) argues against the inclusion of head-movement in "narrow syntax" on a number of grounds: (a) head-movement does not have the formal properties of other operations, (b) it never affects interpretation, (c) it is unclear which features trigger it, (d) it is subject to special locality constraints, and (e) some head-movement operations are sensitive to PF notions such as second position.

I will propose a version of head-movement which does have the problem in (a), and which at the same time solves the problem in (c). I will argue that the claims in (b), (d) and (e) are false. The result is that a kind of head movement, not exactly the same formal operation as construed hitherto, has a place in narrow syntax.