Contextual Cues and Transition Statistics Drive Expression of Competing Motor Memories

Learning multiple motor skills without interference and expressing the correct one in a changing environment is a fundamental challenge. Contextual cues are known to help separate these memories, but how they interact during retrieval is not well understood. We investigated how the stability, recency, and transitional statistics of learning environments influence this process. Across six visuomotor adaptation experiments, participants learned opposing rotations (Tasks A and B) tagged with distinct contextual cues under different schedules (blocked or interleaved) and were tested in stable or dynamic environments. We found that while contextual cues can successfully separate memories, expression is systematically biased by learned transition statistics: towards more stable memories after imbalanced training, and towards more recent memories when stabilities are matched. Critically, when the stable statistics of training mismatched the volatile statistics of testing, cue-based retrieval collapsed, and behavior was dominated by these stability or recency biases. Conversely, learning in a high-entropy, interleaved environment enabled precise, cue-appropriate expression regardless of the testing schedule. These results demonstrate that memory retrieval is not cue-driven but arises from an arbitration process between cues and transition priors. Our findings reveal that memory retrieval involves weighting sensory information against latent priors derived from the history of context transitions. This work provides a unifying theoretical framework for understanding adaptive memory expression, positing that the brain leverages the learned statistical structure of the environment to infer which memory to recall, thereby balancing cue-driven selection with the stability and predictability of past experience. This principle offers a unifying explanation for interference, spontaneous recovery, and the benefits of variable practice, providing a more holistic model of adaptive motor behavior.