Self-, other-, and joint monitoring using forward models

Area of Science:

• Psycholinguistics
• Cognitive Science
• Computational Linguistics

Background:

• Traditional self-monitoring models in language psychology rely on comprehension.
• An alternative framework by Pickering and Garrod (2013) posits predictive modeling in speech production.

Purpose of the Study:

• To elaborate on Pickering and Garrod's (2013) forward-model-based account of language self-monitoring.
• To extend this framework to speech comprehension and dialogue monitoring.

Main Methods:

• Constructing forward models of upcoming utterances for comparison during production.
• Computing inverse models from production errors to refine speech commands.
• Simulating perceived utterances via covert imitation and forward models for comprehension.
• Utilizing discrepancies between simulated and heard speech to update production command representations.

Main Results:

• Speakers predict and compare their utterances, using error signals to adjust production.
• Listeners monitor speech by simulating utterances and comparing them to auditory input.
• Discrepancies inform inverse model computation for refining or revising production commands.
• The model accounts for monitoring in dialogue, including feedback and alignment.

Conclusions:

• Self-monitoring in language involves predictive forward models for both production and comprehension.
• Error signals derived from model-vs-utterance discrepancies are crucial for adaptation.
• This framework offers a unified view of monitoring across production, comprehension, and dialogue.