Back to reality: differences in learning strategy in a simplified virtual and a real throwing task.

Area of Science:

• Motor Neuroscience
• Rehabilitation Technology
• Human Movement Science

Background:

• Virtual environments (VEs) offer controlled sensorimotor conditions for motor neuroscience and rehabilitation.
• Previous studies often assess performance within VEs but not its real-world validity.
• The transferability of skills learned in VEs to real-world scenarios remains a critical question.

Purpose of the Study:

• To compare human throwing performance in a precisely matched real-world environment and a virtual environment.
• To investigate whether virtual environment performance accurately represents real-world motor behavior.
• To analyze the learning process and movement variability in both environments.

Main Methods:

• Comparison of throwing accuracy and precision between a real-world task and a simplified virtual reality task.
• Precise matching of task parameters between the virtual and real environments.
• Decomposition of movement variability into deterministic and stochastic components using the tolerance-noise-covariation method.

Main Results:

• Throwing accuracy and precision were significantly worse in the virtual environment initially, despite simplified movements.
• Real-world task performance improved faster, reaching similar success rates and error levels after 3 practice days.
• Movement variability analysis revealed distinct learning stages: higher tolerance and exploration in the virtual environment, and more covariation and noise (fine-tuning) in the real environment.

Conclusions:

• Simplified virtual environments may require more familiarization and exploration than initially assumed.
• The findings suggest caution in using VEs for research and rehabilitation due to potential challenges in skill transfer.
• Real-world tasks, with fewer constraints, may allow for more adaptive strategies like modifying the solution manifold for improved learning.