Linguistic inputs must be syntactically parsable to fully engage the language network

Area of Science:

• Neuroscience
• Psycholinguistics
• Cognitive Science

Background:

• Human language comprehension is robust to errors, questioning the role of syntactic parsing.
• Some theories propose shallow, semantics-based processing over deep syntactic analysis.
• Alternative view: error correction mechanisms reconstruct ill-formed inputs after symbolic parsing.

Purpose of the Study:

• To test whether syntactic parsing is essential for language comprehension.
• To differentiate between syntactic reconstruction and shallow semantic processing.
• To investigate brain responses to syntactically challenging vs. semantically implausible sentences.

Main Methods:

• Used fMRI to measure brain activity during language comprehension.
• Developed a behavioral paradigm to assess syntactic reconstruction.
• Presented participants with backward word order sentences and semantically implausible sentences.

Main Results:

• Backward word order sentences impaired grammatical structure recovery.
• These backward sentences elicited low responses in language areas (fMRI).
• Semantically implausible but grammatical sentences engaged language areas similarly to plausible sentences.

Conclusions:

• Syntactic parsing is necessary and sufficient for full engagement of the language network.
• Findings support a generalized reliance on syntactic parsing in human language comprehension.
• Robustness to errors may stem from error correction, not absence of parsing.