Insensitivity of the human sentence-processing system to hierarchical structure

Area of Science:

• Cognitive science
• Computational linguistics
• Psycholinguistics

Background:

• Hierarchical phrase structures are widely accepted for describing human language.
• The precise role of these structures in cognitive processing remains under investigation.

Purpose of the Study:

• To investigate the role of hierarchical structure in sentence processing.
• To compare the predictive power of hierarchical versus sequential language models on reading times.

Main Methods:

• Implemented probabilistic language models, varying in their reliance on hierarchical or sequential structure.
• Estimated syntactic category occurrence probabilities for sentences with available reading-time data.
• Correlated model probability estimates with empirical reading-time data.

Main Results:

• Hierarchical structure models did not explain reading time variance beyond sequential models.
• Sequential structure models accounted for a significant portion of reading time variance.
• The contribution of hierarchical structure to predicting reading times was negligible.

Conclusions:

• Sentence hierarchical structure does not appear to significantly influence the generation of expectations for upcoming words.
• Sequential information in language processing plays a more dominant role than hierarchical structure in this context.
• Findings challenge the assumption of hierarchical structure's direct impact on real-time sentence comprehension.