For coordinate structures a structural parallelism effect has been observed during reading, i.e., the second conjunct is read faster when it is structurally similar to the first one (see Frazier et al., 1984; Frazier, Munn & Clifton, 2000).
In (1a) the noun phrase [a short poem] is processed faster than in (1b), since in (1a) the two conjoined noun phrases are structurally identical, which does not hold for the sentence in (1b). However, the parallel-structure effect does not occur in every case a phrase with a particular shape is repeated. The parallelism effect seems to be specific to coordinate structures (see Frazier et al., 2000).
Current models of sentence processing (e.g., Frazier & Clifton, 1996; Schneider, 1999) are not able to account for the parallel-structure effect. In a parallel coordinate structure they propose to build up the same structure again for the second conjunct thereby providing no means to explain its more rapid processing.
We propose the iteration model, a parsing model for the processing of coordinate structures, based on an iteration mechanism which exploits the particularities of these constructions, namely their syntactic redundancies. This way the iteration model is able to account for the parallel-structure effect. In addition, the proposed model can also account for the garden-path effects in coordinate structures (Blodgett & Boland, 1998; Frazier, 1987) which reflect the strategies of the human sentence processor when a conjunction is encountered.
In contrast to conventional two-dimensional trees which are normally used by parsing models for the structural representation of sentences, the data structures resulting from our iteration mechanism are three-dimensional trees. We argue this to support a syntactic analysis which allows for three-dimensional syntactic structure as proposed, e.g., by Moltmann (1992) and Wilder (1999).
An advantage of the iteration mechanism is that it can be described independent of other syntactic processes and comes with a clear interface between the sequential syntactic processes and the iteration mechanism needed for coordination. Therefore, existing parsing models can be extended easily by the proposed iteration mechanism. This is especially interesting for computational models, since coordinate structures tend to be one of the largest error sources for the computational processing of natural language (see, e.g., Argamon, Dagan & Krymolowski, 1999; Collins, 1999).
References
Argamon, S., Dagan, I., & Krymolowski, Y. (1999). A memory-based approach to learning shallow natural language patterns. Journal of Experimental and Theoretical AI, 11, 369-390.
Blodgett, A., & Boland, J. E. (1998). Understanding BOTH In Coordinate Structures: the Osu Subject Pool Both Surprised Allison and Julie! Poster session at the 11th CUNY Conference on Human Sentence Processing, New Brunswick, NJ.
Collins, M. (1999). Head-Driven Statistical Models for Natural Language Parsing. Doctoral dissertation, University of Pennsylvania.
Frazier, L. (1987). Syntactic Processing: Evidence from Dutch. Natural Language and Linguistic Theory, 5, 519-559.
Frazier, L., & Clifton, C. (1996). Construal. Cambridge, MA: MIT Press.
Frazier, L., Munn, A., & Clifton, C. (2000). Processing Coordinate Structures. Journal of Psycholinguistic Research, 29 (4), 343-370.
Frazier, L., Taft, L., Roeper, T., Clifton, C., & Ehrlich, K. (1984). Parallel Structure: A source of facilitation in sentence comprehension. Memory & Cognition, 12 (5), 421-430.
Moltmann, F. (1992). Coordination and Comparatives. Doctoral dissertation, MIT, Cambridge, Mass.
Phillips, C. (2003). Linear Order and Constituency. Linguistic Inquiry, 34 (1), 37-90, in press.
Schneider, D. A. (1999). Parsing and Incrementality. Doctoral dissertation, University of Delaware.
Wilder, C. (1999). Right Node Raising and the LCA. In Proceedings of WCCFL.