I was really missing this from the article. I'd describe it as "play" to get to know the system you are trying to learn. This knowledge than helps with the specific task.
Given the study I assume a different mechanism coming in to play also:
If you try to play a musical piece on an instrument teachers will advise you to practice as slow as needed, not to make a single mistake, that could end up being learned.
It's like trying to correct yourself on word you mangled up while speaking, you'll be so fixated you immediately make the mistake again and again.
This memory effect might have been countered by the variations in the study.
I struggled with exactly that, and failed on that instrument. Never did quite get the same thing out of it twice, without a lot of focus on that thing.
Later, with little formal practice, I would play with a different one. Ended up with some modest skill that remains today.
Instead of learning to correct toward a given nominal, we get a sense of what nominal is, given an input, essentially.