You've neglected to name: a field, a sample group, and a definition of "research skills".<p>Among the general population, skill at taking physics tests probably correlates pretty well with skill at doing physics research -- both are dominated by people who know something about physics.<p>Once you're among the people in the physics Ph.D. program at an Ivy League university, the correlation is pretty weak.<p>Of course, to the extent that you measure "research skill" by things like "number of papers produced" or "amount of grant money attracted", anything that helps you get a more prestigious adviser at a prestigious university is correlated with "skill". Ergo, it doesn't hurt to be good at taking tests. But after you've passed your graduate qualifying exam, you can afford to stop worrying about tests, pretty much forever. Real life -- even <i>university</i> life -- is about other things.
No, they obviously do not. (Being bright is always helpful, of course, but they are pretty different skill sets, I think.) This is supported by the fact that marks/GRE scores aren't particularly important for admission to most top research PhD programs, in CS at least.