Hmm, actually, your description of chess positions actually reminds me a lot of mistakes I’ve made when memorizing kanji: e.g. early on I’d only know one character with a given radical on the left side, say, so when I saw any kanji with that radical in that position, my instinct would be that it was the one that I knew. The last sentence in your comment rings very true, too. (But it’s presumably a worse problem with chess positions than it is for kanji!)

With chess I have to make myself take it slower, because a second really isn’t enough time to identify a complete chess position, and it’s easy to just say “Oh yeah, this is the position where there’s a pawn on c6, I remember the move is Qb6 here,” without really looking at the whole position, and of course there are plenty of positions that arise in my actual games that aren’t in my corpus at all, so it’s a real danger that I could, in a tournament, say “Right, pawn on c6, I’ll play Qb6,” without realizing that the position is slightly different from the one I have stored. Or conversely that I’ll get to that position and then be paralyzed because I can’t remember whether it’s really in my database or is subtly different.

Interesting to hear about the Esperanto reports – I’ve been wondering what will happen if/when I stop adding in new Japanese items to my system. I think you must also go through Esperanto items faster than I go through Japanese items? Or maybe not – I think the time/item is less when I’m going through 30 items than when going through 200.