Parsing Complements:Comments on the Generality of the Principle of Minimal Attachment

Two experiments are described in this paper, which examine the processing of English sentences containing “complement” verbs, and which may be followed either by a nounphrase, as a direct object, or by a complement clause. It has been claimed by Frazier and Rayner (1982) that subjects are “garden-pathed” when reading reduced complements (lacking the overt complemetiser), and that this fact is strong evidence in support of the application of the principle of “Minimal Attachment” as a universal property of the human parser. Holmes, Kennedy, and Murray (1987) questioned this conclusion by providing evidence that full and reduced complements present equivalent problems, probably because of their greater structural complexity. Rayner and Frazier (1987) disputed this conclusion, ascribing it to an artefact resulting from the use of a self-paced reading task. The first experiment examines this controversy with a replication of the study by Holmes et al., measuring eye movements as the sentences are processed. The results replicate those of the original study and further show that the effects cannot be attributed to lexical preferences associated with the verb. The second experiment examines the format of the displayed text as one possible reason for the discrepancy between the two sets of results. When sentences are displayed across several lines, the line breaks are interpreted by subjects as signals for potential clause endings, “garden-pathing” the reader. However, misparsings induced in this way cannot be used to support the claim that readers are generally garden-pathed by the temporary ambiguity of reduced complement sentences.