Subtypes have no empirical basis at all. I think Gulenko's two papers on cognitive styles (which also reads like a torture manual) and activity groups are two parts of a really interesting hypothesis besides being good fodder for science fiction. I think it would be awesome to test some claims of socionics empirically and Gulenko does provide empirical claims, though all the Soviet and tabloid stuff and people declaring themselves ILE geniuses are scaring too many scientists away. Such is the life of a fringe science (I mean fringe science as a neutral term even though it can be somewhat stigmatized despite fringe theories like plate tectonics sometimes being proven right, so if anyone has a better term please tell me.)