Periodic fluctuations in reading times reflect multi-word-chunking

Area of Science:

• Cognitive psychology
• Neuroscience
• Computational linguistics

Background:

• Human memory requires recoding verbal information into limited-duration chunks.
• Neural oscillations (brain activity rhythms) may impose temporal limits on these chunks.
• Previous research predicted behavioral regularity in processing but lacked empirical evidence.

Purpose of the Study:

• To investigate if human sentence processing behavior exhibits periodicity.
• To determine if this periodicity aligns with cognitive chunking mechanisms.
• To explore the relationship between memory constraints, neural rhythms, and language comprehension.

Main Methods:

• Analysis of self-paced reading data from 181 participants.
• Computational formalization of multi-word chunks using dependency and part-of-speech information.
• Selection of final chunks based on normalized pointwise mutual information.
• Identification of periodic patterns in reading behavior at approximately 2 Hz.

Main Results:

• Behavioral periodicity was detected in sentence processing.
• This periodicity was found to be time-aligned with the identified multi-word chunks.
• The findings suggest that chunking minimizes memory demands during language comprehension.

Conclusions:

• This study provides the first evidence of periodic behavior in sentence processing.
• The results support the role of both memory limitations and endogenous neural rhythms in forming chunks during language comprehension.
• Chunking mechanisms appear to optimize information processing by minimizing memory load.