[mastre_]

The last part makes sense, it's similar to a game where they introduce more and more unfamiliar symbols to you as you progress to higher levels, the more unknowns you have to deal with the more you slow down. Eventually, if/when you learn them well you get most of the speed back. I guess most people learning 2nd languages don't learn them well enough to know the 2nd language really well; technically, if you know both languages equally well, the speed penalty/overhead will be minimal, similar to expanding one's vocabulary in the same language (as long as you don't exceed some threshold of mental capacity).

[sho_hn]

I'm not sure that's it. I need to preface this by saying that I'm working from really dim memories here, but from what I recall they were using a definition of 'bilingual' that people who learn a second language after a certain age don't fit, and the tested bilinguals should feel at home in both languages.

Rather, it seems that having more than one language active in the brain concurrently (which requires them to be similarly well developed and in regular use) requires the brain to do more active selection work. I'm not sure this actually makes sense or is just my reading of it, but basically it can't jump to one thing immediately, but has to eliminate the alternatives first. This gives the bilingual brain an edge in tasks that benefit from being well-trained to do that step (and seems to help stave off dementia, where inhibitory control goes out the window, which is a necessity for this selection process to happen), but slows it down in other tasks that would benefit from not having to do the extra work.

In other words, if you're asked to sort or bin a list of English words, it seems to help if your brain doesn't have to wade through a soup made of multiple languages. But if you're asked to do something that exercises similar muscles as wading through this kind of soup (maybe "process strained analogies"? - I kid), it gives you a different sort of edge.