I feel that the JCTI and other image tests are really measures of a specific skill of test-taking. I think I could move up several standard deviations if I learned what types of rules the more advanced questions were using. As it is, I'm just using 'there's a rotation theme here, and I'm going to assume that this texture repeats this pattern, and only one of the answers keeps the rotation theme and that texture in that pattern'. I suppose being able to eliminate 'wrong' answers and guess randomly is part of what the test is intended to measure, but I can see other patterns that eliminate the answer I chose. I'm picking the patterns based on an internalized guess of what I think the test maker is thinking when he is making the test.

For the record, I scored slightly higher than the Less Wrong median when I voted, so I don't think there's a large 'I don't get it, therefore it sucks' aspect to my opinion.