Divisive attenuation based on noisy sensorimotor predictions accounts for excess variability in self-touch

Area of Science:

• Neuroscience
• Computational Neuroscience
• Human Motor Control

Background:

• Tactile sensation is attenuated when self-generated compared to externally applied forces.
• Previous force-matching tasks show healthy participants overestimate self-generated forces.
• This exaggeration is absent when force is controlled indirectly via a motor.

Purpose of the Study:

• To provide the first detailed computational model of sensory attenuation in force-matching tasks.
• To elucidate the roles of motor predictions and noise in force exaggeration.
• To quantify sensory attenuation by comparing direct and indirect self-generated forces.

Main Methods:

• Developed a computational model incorporating sensory signal attenuation based on motor predictions.
• Analyzed existing and new force-matching data from nearly 500 participants.
• Quantitatively fitted the model to data to assess noise contributions and attenuation mechanisms.

Main Results:

• The model successfully accounts for force exaggeration and highlights contributions from memory and prediction noise.
• Attenuated sensations (self-generated forces) exhibit greater trial-to-trial variability than unattenuated ones.
• Evidence supports a divisive, rather than subtractive, mechanism for sensory attenuation based on predicted reafference.

Conclusions:

• Sensory attenuation reduces the perceived intensity of self-generated forces.
• This attenuation incorporates noise from motor prediction, leading to increased variability.
• The findings support a predictive coding framework for sensory processing of self-generated actions.