You'd have to have some sort of test that correlates highly with general intelligence—that is, people who score highly on the test also score highly on many different intelligence-based tasks. To create a really good test, you'd have to have a really good definition of intelligence, and not everyone agrees on a single definition.

In short, I can't describe a specific set of hypothetical evidence because doing so requires having a sturdy definition of intelligence, which I don't have.