Modeling Interactions between Speech Production and Perception: Speech Error Detection at Semantic and Phonological...

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Speech Production and Perception

Background:

• Speech production and comprehension are interconnected, relying on an inner speech loop for error monitoring.
• Spontaneous speech errors are infrequent, necessitating experimental paradigms like picture naming and halt tasks to study them.

Purpose of the Study:

• To investigate the neural mechanisms underlying self-detection of speech errors.
• To develop a biologically inspired neural model of the inner speech loop using the Neural Engineering Framework (NEF).

Main Methods:

• Simulated semantic and phonological speech errors in the first experiment.
• Modeled a picture naming and halt task with varying target-distractor word similarity in the second experiment.
• Utilized a spiking neural network comprising approximately 500,000 neurons.

Main Results:

• The inner speech loop model successfully detected simulated speech errors.
• The model accurately replicated human behavioral data from the picture naming and halt task.
• Phonologically similar distractors resulted in lower target word halting rates compared to semantically similar or dissimilar distractors.

Conclusions:

• The proposed neural architecture for the inner speech loop effectively models key interactions between speech production and perception.
• The findings highlight the critical role of phonological and semantic processing in self-monitoring speech.
• The study provides insights into the neural basis of speech error detection and correction mechanisms.