[pewdiepotpie]

it sounds like there is an experiment to be designed on this basis. One thing that comes to mind, is my experience with IQ estimation tests, is the question or concern the tester had regarding any past testing, there was a very different heuristic for people who have had even one previous test. I have been tested on 6 occasions each estimate at a high percentile, we would need to find someone in apx. 10th percentile and follow thier score distribution as a function of test frequency, with some non trivial considerations for experimental controll of confounds.