Self-Organization
in Sentence Processing: Evidence Whitney Tabor Several recent parsing models (Kim, 2000; Stevenson, 1994, 1998; Vosse
& Kempen, 2000) assume that lexically anchored pieces of syntactic
trees are activated by the words in a sentence as they are encountered,
with each tree-fragment attempting to combine with other activated tree
fragments to form a bigger tree. These models share properties of self-organizing
systems, in that systemic behaviors arise through the interactions of
small, autonomously-acting but interconnected parts. In contrast to
current incremental, grammar-driven processing accounts, self-organizing
models predict that locally coherent fragments will be computed even
when they are ruled out by information the parser has (1) We did not believe the company would fire/hire truck drivers without consulting the union. In (2), the self-organization theory predicts that the locally coherent active reading of "the player tossed the frisbee" will interfere with the globally coherent (reduced relative) parse. (2) The coach chided the player (who was) tossed/thrown the frisbee. Indeed, Experiment 2 showed that reading times on "tossed" and the following words with "who was" absent were elevated compared to the other three cases, indicating that clausal structures can also produce distraction. Finally, Experiment 3 replicated the syntactic aspects of Experiment 2 and found evidence for early use of semantic information in the formation of local subparses. (3) The bandit worried about the prisoner/gold (which was) transported the whole way. Reading times on the verb "transported" and following words
were elevated when the manipulated noun was animate ("prisoner")
and "who was" was absent, compared to the other three cases,
indicating that the semantic coherence of the active sense of "prisoner
transported" created a local distraction. In addition to the empirical
support just noted, References Kim, A. E. (2000). The grammatical aspects of word recognition. PhD Dissertation, Department of Psychology, University of Pennsylvania. Stevenson, S. (1994). Competition and recency in a hybrid network model of syntactic disambiguation. Journal of Psycholinguistic Research, 23(4):295-322. Stevenson, S. (1998). Parsing as Incremental Restructuring. In Fodor, J. D. & Ferreira, F. (Eds.) Reanalysis in Sentence Processing, (pp. 327-364). Dordrecht: Kluwer Academic Publishers. Vosse, T. & Kempen, G. (2000). Syntactic structure assembly in human parsing: a computational model based on competitive inhibition and a lexicalist grammar. Cognition, 75: 105-143. |