I didn't say much of anything about intelligence vs. individual age (longitudinal). All I said was regarding the Flynn effect, which is a cross-sectional measure.

In fact, fluid intelligence declines substantially with age (IIRC starting in the twenties), while crystallized intelligence rises (generally peaking around 60 with linguistic knowledge). I don't know the reliability of these trends across individuals (i.e. does everyone drop from 25, or do some start dropping at 20 and others not until 40?).

I also didn't take the test myself--I saw examples of test questions. I forget how it came up; I was in graduate school, and that was years ago. So I guess take the information with a grain of salt. It's also possible that I have mis-remembered the source of this question, so more grains of salt, I guess.

The point is that unless you're simply doing Raven's Progressive Matrices, there is often enough ambiguity so that rapid pattern-recognition and generalization is NOT the only cognitive skill needed for a high score. You also need, often, to have a sense of relative importance of differences. This, I think, provides some of the mechanism by which your lack of confidence in correctly calling 99.9th percentile could be justified.

Also, I'm pretty sure the part of the brain related to cognition is the whole brain. We find so many correlates of neural activity of one type in unexpected places that I have trouble envisioning how it could be otherwise. Maybe primary sensory cortex is mostly immune to this, but that's about it. But then you can train the precision of auditory maps in children and they tend to do better in school...so...I dunno. "Basically all of it" seems like the safe bet to me.