Mistaking Covariance for Combination in Sensorimotor Adaptation: Regression Slopes Do Not Test Additivity

Area of Science:

• Neuroscience
• Motor Control
• Cognitive Science

Background:

• Sensorimotor adaptation involves implicit recalibration and explicit strategy.
• These processes are often assumed to be additive (A = I + E), underpinning measurement techniques and computational models.
• Recent studies questioned additivity based on regression slopes between implicit and explicit measures.

Purpose of the Study:

• To re-evaluate the interpretation of regression slopes in the context of sensorimotor adaptation additivity.
• To demonstrate that regression slopes reflect covariance, not the summation of learning processes within individuals.
• To establish that regression slopes do not serve as a valid benchmark for testing additivity.

Main Methods:

• Derivation of the expected regression slope under subtractive logic.
• Monte Carlo simulations to test benchmark rejection under various covariance structures.
• Analysis of existing regression slope data in light of shared-error models.

Main Results:

• Regression slopes indicate the covariance structure between learning processes across individuals, not their additive combination within individuals.
• The benchmark slope of β = -1 is only expected under specific uncorrelated conditions, which are often violated.
• Simulations show that even with enforced additivity, realistic covariance structures routinely reject this benchmark.
• Reported regression slopes are consistent with additive shared-error models.

Conclusions:

• Regression slopes are not a valid method for diagnosing the additivity of implicit and explicit sensorimotor learning.
• The additivity assumption requires direct investigation of motor output combination and formal model comparisons.
• Understanding the relationship between implicit and explicit learning requires distinguishing between inter-individual covariance and intra-individual summation.