Testing Bayesian models of human coincidence timing.

Area of Science:

• Neuroscience
• Human Sensorimotor Control
• Bayesian Inference

Background:

• Accurate timing estimation is crucial for sensorimotor tasks like hitting a ball.
• Previous studies often overlooked target timing variability.
• Bayesian theory suggests optimal estimation integrates sensory data with prior knowledge of variations.

Purpose of the Study:

• To investigate if Bayesian integration is used in coincidence-timing sensorimotor tasks.
• To test if the human sensorimotor system adapts to manipulated trial-by-trial target timing variability.

Main Methods:

• Manipulated the trial-by-trial variability (prior distribution) of target timing in a sensorimotor task.
• Observed human subjects' timing behavior over several hundred learning trials.
• Compared behavioral adjustments when prior distributions changed from wide to narrow and vice versa.

Main Results:

• Subjects developed systematic timing behavior aligned with the prior distribution's width, matching Bayesian model predictions.
• Behavioral adaptation occurred when switching between narrow and wide prior distributions.
• Adaptation was slower when transitioning from a wide to a narrow prior distribution.

Conclusions:

• Bayesian integration is a fundamental mechanism across all aspects of human sensorimotor control, including timing tasks.
• The human sensorimotor system demonstrates flexibility and adaptability in Bayesian learning.
• Findings support the role of Bayesian inference in optimizing sensorimotor performance under uncertainty.