Prospects for top-down derivation

Abstract

This article explores a model of grammar involving top-down derivations, where each step ("splitmerge") yields an asymmetric pair of elements relevant to the expression of order, information, and grammatical features. These derivations are inevitably layered, in the sense that the output of a previous derivation may appear as an atom in the numeration for the next derivation. It is suggested that opacity effects follow from the layering of derivations, not from conditions on movement. While the main questions surrounding the model contemplated here involve the 'when' and 'what' of merge, this article focuses on the more preliminary question of the 'how' of merge.

The complexity of narrow syntax: Minimalism, representational economy, and simplest Merge

“…the sequence of operations Merge turning a set of elements ('Numeration') into a hierarchical structure. In current minimalism, as mentioned above in section 3.1, such a derivation is taken to be punctuated, consisting of various phases (Chomsky 2001) or derivation layers (Zwart 2009a), each feeding the interface components independently of other phases/layers that are part of the same derivation. As argued in more detail below, this punctuated nature of the derivation calls for a reconsideration of what it is that makes a derivation recursive, thereby bearing on Chomsky's original argument for natural language grammars as being of the context-free type.…”

“…This conception of Merge covers both parts of the disjunction in (10a), the rewrite rule yielding a terminal at the first step of the procedure, and a pair consisting of a terminal and a nonterminal after that. Similarly, in the top-down procedure proposed in Zwart (2009a), the final step in the procedure will involve just a single terminal, leaving only the empty set as what remains of the Numeration. Since the Numeration, in this system, is a nonterminal, each step yields a pair of a terminal and a nonterminal, except the last, which yields just a terminal (or a terminal and an empty set).…”

“…We refer to Zwart (2009a;2011a) for a characterization of layered derivations and the idiosyncratic properties (including locality) that they yield. As stated above, accepting the possibility of layered derivations, terminals can be phrases, and the differences between Merge (in its simplest conception) and the finite-state rule in (10a) disappear.…”

Easy to (re)analyse

“…I argue in this article that íow^/¡-constructions receive a simple and natural analysis within hnguistic minimalism. A crucial ingredient of minimalism, in the view adopted here, is the construction of a network of (suh)derivations (derivation layering), with interface effects showing up each time an element passes from one derivation layer to the next (Zwart 2009(Zwart , 2011a. I argue that the reanalysis of the complex string easy to please as a single adjective can be understood as such an effect of derivation layering.…”

Precede-and-command revisited revisited

“…Many have argued that in order to achieve a maximally simple derivation D of any linear string, we must allow D to involve subderivations, which Chomsky (2001) calls phases. But subderivations can be organized in many different ways (Uriagereka 1999, Zwart 2009, de Vries 2012, and it is not a priori clear that a division in syntactic categories, as in 3, is necessary. 12 An alternative to phases, argued in Zwart 2009, is readily available, since (as is universally agreed) any element α merged in the context of derivation D1 may be the output of a separate (prior) derivation D2; if so, every derivation is potentially layered, that is, a network of subderivations.…”