Chunk boundaries disrupt dependency processing in an AG: Reconciling incremental processing and discrete sampling.

Area of Science:

• Psycholinguistics
• Neuroscience
• Computational Linguistics

Background:

• Language comprehension relies on linking words, including those not adjacent.
• Speech is perceived in discrete chunks, posing challenges for processing non-adjacent dependencies (NADs).
• The neural mechanisms underlying NAD processing across chunk boundaries remain unclear.

Purpose of the Study:

• To investigate how the brain processes non-adjacent dependencies (NADs) in speech.
• To determine if NADs are processed differently when they cross chunk boundaries compared to within chunks.
• To reconcile incremental language processing theories with discrete speech sampling evidence.

Main Methods:

• An electroencephalography (EEG) study with 37 participants learning an artificial grammar (AG).
• Frequency-tagging approach used to analyze EEG responses to multi-syllable chunks and NADs.
• Comparison of electrophysiological responses to NADs within chunks versus across chunk boundaries.

Main Results:

• Participants successfully learned the artificial grammar chunks, confirmed by EEG spectral peaks.
• Electrophysiological responses to NADs crossing chunk boundaries were significantly smaller than within-chunk NADs.
• This indicates that processing NADs is less efficient when they span across perceived speech chunk boundaries.

Conclusions:

• The brain processes non-adjacent dependencies (NADs) less effectively when they cross perceived speech chunk boundaries.
• Findings suggest that discrete speech chunking may impose limitations on the processing of long-range linguistic dependencies.
• This research bridges the gap between incremental and discrete models of speech perception and has implications for language acquisition and syntax.