Benjamin Swets & Fernanda Ferreira
Michigan State University
Various theories of language production make different assumptions about how much syntactic information is buffered before and during articulation. While some evidence supports theories maintaining that syntactic planning occurs in small, phrasal chunks (Schriefers, Teurel & Meinhausen, 1998; Smith & Wheeldon, 2001), other evidence shows that syntactic planning takes place over larger domains (Ferreira & Swets, 2002; Meyer, 1996). It has also been argued recently that syntactic planning is not a costly process once articulation has begun (Smith & Wheeldon, 2001). However, most of the above studies have investigated the planning of relatively simple structures. The goal of the present study was to test these different assumptions of incremental syntactic planning with a task that would involve the production of complex structures --- English relative clause sentences in which resumptive pronouns are uttered behind wh-island boundaries (see Example 1, below).
The experiment utilized a task which required English speakers to describe, with full sentences, a picture array that included partial descriptions. We compared initiation times to begin articulation and utterance duration measures of utterances like Example 1, below, with two other syntactic conditions. In one condition, the surface form of the target sentence was similar (Example 2, below). A condition with a cross-clausal dependency between a wh-element and its trace that did not have a wh-island boundary was also included (Example 3, below). The hypothesis that syntactic planning operates in sub-clausal or phrasal increments would predict that, because each sentence begins with the same clause, no effects of initiation time should be observed.
In fact, the target sentences did not reveal initiation time effects between conditions. However, effects of the late-arising syntactic complexity were observed on speech durations from regions early on in the sentences. Specifically, when a sentence with a wh-island resumptive pronoun was uttered, more time was spent uttering both the head noun ("donkey", below) and the subsequent relative pronoun ("that") (both of which are separated by at least a clause from the source of the wh-island complexity), compared with the two control conditions. These findings are not consistent with either the phrasally incremental hypothesis or the hypothesis that syntactic planning is cost-free once articulation has begun. The findings are more consistent with theories of production that assume less incremental syntactic planning (Ferreira, 2000).
Example target sentences
(1) This is a donkey that I don't know where it lives.
(2) This is a donkey that doesn't know where it lives.
(3) This is a donkey that I didn't say lives in Brazil.
References
Ferreira, F. (2000). Syntax in language production: An approach using tree-adjoining grammars. In L. Wheeldon (Ed.), Aspects of Language Production. Cambridge, MA: MIT Press.
Ferreira, F., & Swets, B. (2002). How incremental is language production? Evidence from the production of utterances requiring the computation of arithmetic sums. Journal of Memory & Language, 46, 57-84.
Meyer, A.S. (1996). Lexical access in phrase and sentence production: Results from picture-word interference experiments. Journal of Memory & Language, 35, 477-496.
Schriefers, H., Teruel, E., & Meinhausen, R. M. (1998). Producing simple sentences: Results from picture-word interference experiments. Journal of Memory & Language, 39, 609-632.
Smith, M. & Wheeldon, L. (2001). Syntactic priming in spoken sentence production: An online study. Cognition, 78, 123-164.