IQ tests often feature puzzles where the subject must identify and extrapolate a pattern in a sequence of symbols. One might hypothesize that someone who is good at doing that would also be good at finding patterns in random symbols, or otherwise finding connections between unrelated things. This may lead to a tendency to believe that things are connected when they aren't. Is there any research demonstrating a relationship between the ability to see patterns and the tendency to find patterns when they "aren't there?"

Clearly, someone who can't see any patterns would never commit type-II error, but on the other hand someone who can see lots of patterns may have well-calibrated instincts from experience with other patterns that appeared to them but later turned out to not be there. It's not obvious to me how those competing factors would balance out as you went from low-g to high-g.

  • $\begingroup$ If I understand correctly, you interpret correctly solving an item from the reasoning part of an IQ test as avoiding a type-I error, that is, correctly rejecting a null hypothesis of no pattern being present. You argue that people who are good at this should also tend to make more type-II errors, because they tend to more easily reject the null hypothesis of no pattern, even when it's true. However, as in statistical testing, it is entirely possible for someone to make both fewer type-I and type-II errors than another person, as these two error rates are not always negatively correlated. $\endgroup$ – fujiu Sep 1 '19 at 13:19
  • $\begingroup$ @fujiu Finding one person or even many people who make fewer of both types of errors doesn't mean that the two variables can't be negatively correlated, only that they can't be perfectly negatively correlated. $\endgroup$ – Retracted Sep 2 '19 at 5:10
  • 1
    $\begingroup$ You're absolutely right. What I meant was probably not clear enough. It's absolutely possible that such a negative correlation might exist, I just wanted to point out that no correlation or even a positive correlation would also be possible. Unfortunately, I don't know any research addressing this question. $\endgroup$ – fujiu Sep 2 '19 at 8:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.