Abstract
Sensorimotor adaptation is supported by at least two parallel learning systems: an explicit strategy which is intentionally controlled, and an involuntary implicit learning system. These two systems are generally studied in laboratory settings using visuomotor rotations, studies that have shown subconscious learning systems to be driven in part by sensory prediction error (i.e., the mismatch between the realized and expected outcome of an action). Here we used a ball rolling task to explore whether sensory prediction errors can drive implicit motor adaptation outside of the standard and highly constrained laboratory environment. After application of a visual shift, participants rapidly adapted their rolling angles to reduce the error between the ball and target. We removed all visual feedback and told participants to aim their throw directly toward the primary target, revealing an unintentional 5.19° implicit adjustment to reach angles that decayed over time. To determine whether this implicit adaptation was driven by sensory prediction error, we tested participants in a second paradigm similar to Mazzoni and Krakauer (2006) in which participants were given an aiming target that would ‘solve’ the visual shift. As expected, the aiming group showed no difference in implicit learning but developed a larger explicit strategy than the control group. Remarkably, after rapidly reducing ball rolling error to zero, rolling angles in the aiming group gradually drifted over time, resulting in an overcompensation of 3.28° beyond the target (and towards the aiming target). This involuntary drift in rolling angle, which resulted in worsening task performance, is a hallmark of implicit learning driven by sensory prediction error. These results show for the first time that implicit processes driven by sensory prediction error studied in the laboratory actively contribute to motor adaptation in naturalistic skill-based tasks with more complex body motion.
Research summary Current theories posit that implicit, i.e., subconscious, motor learning is driven in part by sensory prediction error (SPE): a mismatch between a movement’s outcome versus the expected result1–7. The notion that errors are a critical implicit learning substrate is central to both descriptive and computational motor control models1,8–11. The visuomotor rotation (VMR) paradigm has been applied extensively to examine error-based adaptation, in which participants use a robotic manipulandum or tablet to move a cursor towards a virtual target (Fig 1, Inset). A visual rotation is applied to the cursor’s path, inducing an SPE. While this constrained environment allows precise error manipulation2,6, skill learning in natural settings is more complex, involving multiple unconstrained control variables and internal models that predict the motion of external physical objects. Do SPE learning mechanisms uncovered in laboratory settings extend to these more complicated skill-based behaviors?
Competing Interest Statement
The authors have declared no competing interest.