My colleagues and I are developing a detailed process model of how human parsing recovers automatically from misinterpretations of structurally ambiguous material. For example, consider the following sentences:
(1) Mary forgot her husband yesterday.
(2) Mary forgot her husband went to the store yesterday.
(3) Although Mary forgot her husband wasn't angry.
How does the parser decide in real-time whether "her husband" is the direct object of "forgot" (as in (1)), or the subject of a new clause, as in (2) and (3)? Most psycholinguistic theorizing has focused on what guides the initial interpretation of local ambiguities, but has not explained what happens when the chosen interpretation proves to be inconsistent with later parts of the linguistic input. Even though the information needed to resolve ambiguities is often not available at the decision point, many structural ambiguities are not problematic for human parsing (as example (2) illustrates). Limited repair parsing provides a novel reformulation of the parsing search space that explains this fact, while at the same time accounting for the classic breakdowns on difficult garden path sentences, such as example (3). The limited repair model provides an account of the contrasts between dozens of easy and difficult ambiguities across several typologically distinct languages.
Here are some papers that discuss reanlaysis:
Van Dyke, J. & Lewis, R. L. (2002, submitted). Distinguishing effects of structure and decay on attachment and repair: A retrieval interference theory of recovery from misanalyzed ambiguities. [PDF]
Lewis, R. L. (1999). Specifying architectures for language processing: Process, control, and memory in parsing and interpretation. In Crocker, M., Pickering, M., and Clifton, C. Jr., editors, Architectures and Mechanisms for Language Processing. Cambridge University Press. [postscript]
- Lewis, R.L. (1998). Leaping off the garden path: Reanalysis and limited repair parsing. In Fodor, J.D. and Ferreira, F. (eds), Reanalysis in Sentence Processing. Boston: Kluwer Academic. [PDF]
- Lewis, R. L. (1993). An Architecturally-based theory of Human Sentence Comprehension. PhD thesis, Carnegie Mellon University. [PDF] Also available as Computer Science Tech Report CMU-CS-93-226; send requests to reports@cs.cmu.edu.