Short-term balance training induces task-specific performance improvement,1–۳ and consequently balance should be seen more as a sum of task-specific skills than a general ability.2 Therefore, in regard to fall prevention, which is a major health and resource issue,4,5 patients should learn as many balance tasks as possible, and as close as possible to real-life fall contexts.2 The implementation of this recommendation requires optimization of balance task learning, especially in the setting of clinical rehabilitation where there is limited time to practice. The learning rate of a novel motor task can be improved with previous learning experience of relatively similar tasks.6 This phenomenon has been termed “learning to learn”.۷ One of the several possible mechanisms driving this effect is called structural learning, where invariants of several tasks sharing the same structure are extracted and facilitate the learning of a new task also sharing the same structure.8 The “learning to learn” effect has been demonstrated for cognitive9–۱۱ and visuomotor tasks,6,12 including a locomotion task relying on distorted visual feedback.13 To our knowledge, this “learning to learn” effect has never been specifically tested for complex full-body tasks such as balance tasks. This learning facilitation would be of interest to reduce the time to learn many different balance tasks. We hypothesized that prior practice of similar balance tasks improves the learning rate of a novel balance task. However, contrary to visuomotor or perceptual tasks, performance during balance tasks might not only depend on the optimization of the motor command or strategy, but also on the physical capacity to adequately perform the movement. In particular, some studies have reported an influence of lower limb strength and power on balance performance, although this finding is challenged by the results of other studies reporting no significant correlation.14,15 To shed light on this issue, we also measured lower limb power and assessed its influence on the learning rate.