Underspecification of syntactic ambiguities: evidence from self-paced reading

Area of Science:

• Psycholinguistics
• Cognitive Psychology
• Computational Linguistics

Background:

• Syntactic ambiguity in sentences can lead to faster reading times compared to unambiguous sentences.
• Existing parsing models explain this via structural race or reanalysis.
• An alternative hypothesis suggests readers underspecify interpretations of ambiguous attachments to save time.

Purpose of the Study:

• To investigate whether readers strategically underspecify interpretations of ambiguous attachments.
• To examine if question types influence reading and parsing strategies.
• To test the "good-enough" and construal models of sentence processing.

Main Methods:

• A self-paced reading study was conducted.
• Participants read sentences containing relative clauses with two possible attachment sites.
• Question type (superficial vs. specific interpretation) was manipulated between participants.

Main Results:

• An "ambiguity advantage" in reading times was found, but only when superficial comprehension questions were expected.
• When specific interpretation questions were anticipated, reading times in disambiguated regions increased, reducing the ambiguity advantage.
• Despite more careful inspection, question-answering times suggested underspecified representations for ambiguous attachments.

Conclusions:

• Reading time advantages for ambiguous sentences are influenced by expected task demands.
• Readers may employ "good-enough" processing, strategically underspecifying interpretations.
• Findings support construal and "good-enough" models of sentence parsing.