Learning is difficult. Why? Your brain must change somehow to endow you with new knowledge and skills. We don’t yet know how exactly new knowledge is stored in the brain, let alone why this process takes time. We approach this question from the perspective of neural population activity. I will present a particular behavior (brain-computer interface control) that enables us to query the state of a  population of task-relevant neurons prior to learning, and then challenge that population to change in ways that promote new abilities. We find that learning is slow when it requires the formation of new patterns of population activity, but learning can be fast when pre-existing neural activity patterns are sufficient for the learned behavior. This means that we might eventually be able to examine the activity of a population of neurons and predict what will be easy or difficult to learn.