Reading focus
Syllabus: How Does Practice Build Skill?
Identify what varied across practice and what structure remained stable. Compare the implications of learning that structure with Logan's account of retrieving instances, without treating experiments in different tasks as a direct contest.
A question to consider. What new task would distinguish learning a useful family of transformations from merely becoming faster on practiced examples?
Why this reading
Experience can prepare a learner for a task they have not encountered in exactly that form. These experiments investigate whether variation teaches relationships among changes, narrowing the adjustments needed when a new task shares that structure.
Before you read
A visuomotor transformation changes the relation between a hand movement and its displayed result. A rotation changes its direction by an angle; a family of rotations varies that angle while retaining the same kind of transformation. Structural learning concerns such common relationships, rather than memorizing one setting or learning the average movement.
Read for
- Compare varied training within one transformation family with naïve participants and with variation across a broader family. Identify what each control contributes.
- Separate faster adaptation, reduced interference, and the direction of exploration as evidence for structure. Follow the distinction between initial movement and later corrections.
- Ask what each comparison excludes: familiarity with a particular angle, a general increase in adaptability, or learning relationships that apply to a restricted family of tasks.
Consider afterwards
- Training on many rotations improves adaptation to a new rotation. What additional test would distinguish learning the rotation family from simply becoming better at handling any visual disturbance?
Editorial context
The evidence concerns structured laboratory movement transformations. Everyday activities in the introduction are motivating analogies, not tested outcomes. The experiments constrain accounts of learning from variation without directly measuring the neural representation proposed to explain it. Nor does random variation guarantee transfer: what varies, what remains stable, and what the subsequent task requires are central to the inference.