I had suggested the possibility of reading this out of whatever effects are left over from the conscientiousness indicator (achievement vs. intelligence), but I think that’s probably stretching the statistics a little too far for purposes of prediction. And without prediction, where’s the fun in theorizing? Better to come up with another indicator that is either a primary measurement, or an artifact of intelligence alone.
My first idea is to measure the rate at which a person can learn nonsense associations from flash cards. Maybe the flash cards have nonsense syllables on one side and nonsense ink blots on the other side. It’s based on the established fact that people are more likely to remember nonsense syllables that sound like real words that “make sense”. That is, the nonsense calls up something from semantic memory- maybe the nonsense “pne” reminds someone of “pen”, which is easier to remember. Similarly, we’re better able to remember the shape of an ink blot after we’ve interpreted it to look like something meaningful.
More practically speaking, a library of nonsense syllables and ink blots already exists: the Chinese alphabet. In short, my test is to hand someone a stack of Chinese alphabet flashcards and see how many they can memorize within a given time limit (where it is assumed that nonsense -> semantic free association is the best possible method), then compare that to their IQ centile. A higher number of flashcards indicates higher AH, with IQ held constant.
This leaves a couple of practical psychometric problems. 1) Some people do not realize free association can be used to learn information from flashcards much more quickly than mere rote and repetition. 2) You can get better at free association through practice. 3) An older person has more semantic associations to use, but younger people seem to be able to call them up more quickly. For theoretical purposes, all we’d have to do is assume that test-takers have practiced this method to the point of diminishing returns, but that presents a psychometric hurdle all its own. We’d have to measure this curve before even attempting to pull AH out of the data. It’s implausible that I could pull off such a research project, but maybe someone else could. (Maybe I’ll e-mail Cooijmans?)