My colleagues and I am working on a detailed account of working memory in parsing. Human parsing often handles complex structures easily without taxing working memory, but sometimes fails dramatically. Consider the following examples:
(1) The bird chased the mouse that scared the cat that saw the dog that ate the pumkin that grew in the garden.
(2) The salmon that the man that the dog chased smoked fell.
Right-branching structures, such as example (1) can be easily understood despite multiple levels of embedding (four embedded clauses in this example). By contrast, center-embedded structures such as (2) lead to severe difficulty at quite shallow levels (in this case, two embedded clauses).
The relevant empirical data are challenging, because the contrasts between the easy and difficult structures are often quite subtle and defy simple characterizations (particularly when considering cross-linguistic data). For example, (3) below contains double center-embedded clauses, but does not lead to the same kind of difficulty as (2):
- (3) That the food that John ordered tasted good pleased him.
The theory I proposed arose initially out of a concern for efficient parsing within the Soar production system architecture. Later work showed that the structure of the model could be motivated by a general psychological principle, similarity-based interference, which is known to play a significant role in other kinds of non-linguistic short-term memory. The resulting theory is a detailed account of syntactic working memory that makes contact with a general principle of human memory (similarity-based interference). I'm currently working on a new model of syntactic interference in the ACT-R architecture,with a focus on serial odrer representation.
Weare testing the model cross-linguistically in Japanese, English, and Hindi.
Here are some papers and presentations that discuss working memory in parsing:
- Lewis, R. L. & Nakayama, M. (to appear). Syntactic and positional similarity effects in the processing of Japanese embeddings. In Nakayama, M. (Ed.) Sentence Processing in East Asian Languages. Stanford: CSLI Publications. [PDF]
- Lewis, R. L. & Nakayama, M. (2001). "The Representation of Serial Order Information in Sentence Processing: Insights from Work on Short Term Memory and Implications for Processing Complexity. " Fourteenth Annual CUNY Sentence Processing Conference, University of Pennsylvania, Philadelphia, March 2001. [PDF]
- "Working Memory for Syntactic Processing: Serial Order Information and Cue-based Retrieval." Department of Psychology, Michigan State University, February 2001. [PDF]
- "Attachment Without Competition: A Race-based Model of Ambiguity Resolution in a Limited Working Memory". CUNY Sentence Processing Conference, March 1999. [postscript]
- Lewis, R.L. (1999). Accounting for the fine structure of syntactic working memory: Similarity-based interference as a unifying principle. Behavioral and Brain Sciences 22:105-106 [PDF]
- Young, R.M. and Lewis, R.L. (1998). The Soar Cognitive Architecture and Human Working Memory. In Miyake, A. and Shah, P. (Eds), Models of Working Memory: Mechanisms of Active Maintenance and Executive Control. New York: Cambridge University Press. [PDF].
- Lewis, R. L. (1996). Interference in short-term memory: The magical number two (or three) in sentence processing. The Journal of Psycholinguistic Research 25:93-115. [postscript; hardcopy reprints also available; please email a request and include your address.]
- Lewis, R. L. (1996). A theory of grammatical but unacceptable embeddings. Manuscript. [postscript]
- 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.
[Home | Top of page]