The resonant dynamics of speech perception: interword integration and duration-dependent backward effects

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Speech Perception

Background:

• Listeners integrate temporally distributed phonemic information into coherent representations of syllables and words.
• Context-sensitive speech perception demonstrates how variations in timing and sound duration can alter word recognition.
• Previous research highlights perceptual shifts, such as 'gray chip' becoming 'great chip' with increased silence.

Purpose of the Study:

• To present the ARTWORD neural model for quantitatively simulating context-sensitive speech data.
• To explain the mechanism by which phonemic information is integrated into word percepts.
• To model the emergent properties of conscious speech perception.

Main Methods:

• Sequential storage of phonemic items in working memory.
• Bottom-up input to unitized list chunks that group variable-length item sequences.
• Competitive interaction between list chunks and feedback for activation boosting.

Main Results:

• The ARTWORD model quantitatively simulates context-sensitive speech perception data.
• The model demonstrates how temporal factors influence phonemic integration.
• Emergent conscious percepts are created through resonance and boosted activation.

Conclusions:

• The ARTWORD model provides a computational framework for understanding speech integration.
• Temporal dynamics and competitive processes are crucial for forming coherent speech percepts.
• The model successfully replicates human listeners' context-sensitive perception of speech sounds.