Journal of Experimental Psychology: Learning, Memoiy, and Cognition 1998, Vol. 24, No. 4, 940-961 Copyright 1998 by the American Psychological Association, Inc. 0278-7393/98/S3.00 Plausibility and Recovery From Garden Paths: An Eye-Tracking Study Martin J. Pickering University of Glasgow Matthew J. Traxler Florida State University and University of Glasgow Three eye-tracking experiments investigated plausibility effects on recovery from misanalysis in sentence comprehension. On the initially favored analysis, a noun phrase served as the object of the preceding verb. On the ultimately correct analysis, it served as the subject of a main clause in Experiments 1 and 3 and of a complement clause in Experiment 2. If the object analysis was implausible, disruption occurred during processing of the noun phrase. If it was plausible, disruption occurred after disambiguation. In Experiment 3, discourse context affected plausibility of the initial analysis and subsequent reanalysis. The authors argue that readers performed substantial semantic processing on the initial analysis and committed strongly when it was plausible. Experiment 3 showed that these effects were not due to selectional restrictions or word co-occurrences and that the interpretation of the target sentence was not computed in isolation. Research in sentence processing has focused overwhelm- ingly on the resolution of syntactic ambiguity by considering syntactic misanalysis and reanalysis in "garden-path" sen- tences (Bever, 1970; Frazier, 1979; Frazier & Rayner, 1982). (In this article, syntactic misanalysis and reanalysis may refer to adopting and abandoning a single analysis or to favoring and subsequently disfavoring an analysis in com- parison to others.) Much recent research has been concerned with the role of semantics in initial syntactic analysis (e.g., Altmann, Garnham, & Dennis, 1992; Altmann & Steedman, 1988; Ferreira & Clifton, 1986; Mitchell, Corley, & Gar- nham, 1992; Rayner, Carlson, & Frazier, 1983; Taraban & McClelland, 1988; Trueswell, Tanenhaus, & Garnsey, 1994). In this article, we address a related question: How do semantic factors affect the process of recovery from misan- alysis when there is no doubt that initial misanalysis has Martin J. Pickering, Human Communication Research Centre, Department of Psychology, University of Glasgow, Glasgow, United Kingdom; Matthew J. Traxler, Department of Psychology, Florida State University, and Human Communication Research Centre, Department of Psychology, University of Glasgow, Glasgow, United Kingdom. The order of authorship is arbitrary. This research was supported by Economic and Social Research Council Grant R000234542 and a British Academy postdoctoral fellowship awarded to Martin J. Pickering. We gratefully acknowledge Melody Terras for her assistance in data collection and analysis. We are particularly indebted to Keith Edwards for extremely capable technical assis- tance. We also acknowledge the members of the Human Communi- cation Research Centre Sentence Processing Group at the Universi- ties of Edinburgh and Glasgow for their advice and encouragement. We thank Simon Garrod, Simon Liversedge, Keith Rayner, and Tom Urbach for insightful commentary, theoretical input, and technical assistance. Correspondence concerning this article should be addressed to Martin J. Pickering, Human Communication Research Centre, Department of Psychology, University of Glasgow, Florentine House, 53 Hillhead Street, Glasgow G12 8QF United Kingdom. Electronic mail may be sent to martin@psy.gla.ac.uk. taken place? We also address the question of what kinds of information affect the computation of the plausibility of an analysis: Is the relevant assessment of plausibility based solely on sentence-internal factors, or is it affected by discourse context? Incremental understanding rapidly affects processing. For example, resolution of a noun phrase with respect to discourse context rapidly affects syntactic analysis (Altmann et al., 1992; Altmann, Garnham, & Henstra, 1994; Altmann & Steedman, 1988; Britt, Perfetti, Garrod, & Rayner, 1992; Britt, 1994; Spivey-Knowlton, Trueswell, & Tanenhaus, 1993; Trueswell & Tanenhaus, 1991). Semantic context also rapidly affects processing of spoken and written language (e.g., Boland, Tanenhaus, Garnsey, & Carlson, 1995; Gar- rod, Freudenthal, & Boyle, 1994; Marslen-Wilson, 1973, 1975; Marslen-Wilson, Tyler, & Koster, 1994; Swinney, 1979; Traxler & Pickering, 1996; Trueswell et al., 1994; Tyler & Marslen-Wilson, 1977). Hence, it might be expected that semantics would rapidly affect syntactic reanalysis. The fact that people interpret sentences incrementally allows us to make some general predictions about semantic influence on syntactic processing. We propose that people experience greater difficulty during the initial processing of syntactically ambiguous fragments when the analysis that they adopt (or foreground) is semantically implausible than when it is semantically plausible. Further, we propose that people experience greater difficulty during processing of syntactically disambiguating information (after misanalysis) when the initial analysis had a plausible interpretation than when it had an implausible interpretation. If people con- struct an analysis for a sentence fragment that has a plausible interpretation, they should strongly commit to that analysis, integrating its interpretation with general knowledge. They should then find reanalysis comparatively difficult. In con- trast, if people construct an analysis for a sentence fragment that has an implausible interpretation, they should less strongly commit to that analysis. Either they reanalyze immediately or they retain the analysis but integrate it less strongly with general knowledge. However, it is also 940