Does comprehension (sometimes) go wrong for noncanonical sentences?

Area of Science:

• Psycholinguistics
• Cognitive Science
• Computational Linguistics

Background:

• The human parsing mechanism (HPM) is debated regarding its use of algorithmic versus non-algorithmic strategies.
• Misinterpretation effects in noncanonical sentences are often cited as evidence for non-algorithmic processing.
• It remains unclear if these effects stem from genuine parsing errors or comprehension task limitations.

Purpose of the Study:

• To investigate whether the human parsing mechanism (HPM) exclusively uses algorithmic processing.
• To determine if observed misinterpretation effects in sentence processing are due to parsing errors or task-specific difficulties.
• To provide evidence for or against models of the HPM that incorporate non-algorithmic strategies.

Main Methods:

• Two experiments were conducted using canonical and noncanonical sentences.
• Participants completed two tasks: plausibility judgment and agent/patient naming.
• The order of these tasks was manipulated across experiments to assess task-specific influences.

Main Results:

• Participants consistently made errors in agent/patient naming, irrespective of sentence plausibility judgments.
• Performance on plausibility judgments did not correlate with errors in the naming task.
• Task order did not significantly alter the pattern of errors in agent/patient identification.

Conclusions:

• Misinterpretation effects in sentence processing do not reflect errors in the HPM's core parsing mechanisms.
• The observed errors are likely attributable to the demands of specific comprehension tasks, not non-algorithmic parsing.
• Results strongly support models positing that the HPM relies solely on algorithmic processing for deriving sentence meaning.