Domain Minimization in English Verb-Particle Constructions

The placement of the particle before or after an object in the English verb-particle construction is influenced by a variety of factors. We argue that many of them can be subsumed under a single simple principle, motivated by considerations of processing efficiency: to the extent that the domains of syntactic and semantic dependencies can be minimized, processing is facilitated. We use a more precise formulation of this idea to make several predictions about the distribution of particles based on the size of the object NP and the semantic dependencies among the verb, the particle, and the object. Corpus studies confirm the predictions, providing evidence for the principle of domain minimization.*

1. Introduction

English verb-particle phrases display a basic syntactic alternation: verb and particle can occur either joined, as in 1a, or split, as in 1b.

(1)

a. She _vp [looked up _np [the number]]

b. She _vp [looked _np [the number] up]¹

Numerous factors have been proposed to account for the distribution of joined vs. split orders in English performance. Most of these factors fall into one of three groups: syntactic factors like the length of the NP (e.g. Chen 1986, Hawkins 1994), semantic factors like the degree of idiomaticity (e.g. Fraser 1976, Chen 1986), and discourse factors like givenness (e.g. Chen 1986) or focus (e.g. Dehé 2002). These studies often focus on just one factor at the expense of others. If they consider multiple factors, it is unclear how they interact. What is also lacking in this research literature is a satisfactory explanation for the distribution.

In his recent corpus study of particle placement, Gries (2003) demonstrated the need for a multifactorial approach. He initially considers all twenty-five factors that have ever been proposed and eliminates those that prove to be statistically insignificant. He then shows how the remaining factors can be subsumed under the processing hypothesis; that is, they all relate to differences in processing cost for different orderings. In order to do this, however, he has to appeal to a variety of theoretical frameworks and to processing principles that range from the proposed processing cost associated with identifying and activating referents in discourse (e.g. Givón 1992, Lambrecht 1994) to that associated with various constituent orders (e.g. Hawkins 1994).

We take a different approach here. We integrate several of the factors that have traditionally been proposed to affect ordering preferences into a single and simple principle of processing efficiency and complexity. We argue that the ordering preferences that can be observed for verb-particle phrases in performance data are strongly linked to the overall size of the processing domains for the various syntactic and semantic relations that are involved in these constructions. Applying the theoretical framework presented in Hawkins 2001, 2003, and 2004, we focus on two types of basic relations: combinatorial and dependency relations.

A relation of combination is defined in 2.

(2) Combination

Two categories A and B are in a relation of combination iff they occur within the same syntactic mother phrase or maximal projection (phrasal combination), or if they occur within the same lexical cooccurrence frame (lexical combination).

The phrasal combinations of a language are determined by its phrase-structure rules: a verb combines with a direct object NP within the same VP, the subject combines with the VP within S (or IP), and so on. Subjects, objects, and other complements are also in a relation of lexical combination with the verb and are listed in its lexical entry; adjunct phrases are not so listed.

The definition for dependency given in Hawkins 2001, 2003, and 2004 is summarized in 3.

(3) Dependency

Two categories A and B are in a relation of dependency iff the parsing of B requires access to A for the assignment of syntactic or semantic properties to B with respect to which B is zero-specified or ambiguously or polysemously specified.

For example, the subject NP in The key opened the door to the wine cellar is zero-specified with regard to its theta-role and the parser needs to...