From a reader:
The question of whether or not IQ tests are equally valid for sociopaths is an interesting one. Essay tests typically measure not only subject material mastery, but also how closely the opinions expressed by a test taker match those of the test grader. Poorly written multiple-choice questions may follow simple patterns e.g. longest answer is always right. If someone administering a test knows the answers and gives non-verbal cues, then they may just be measuring a Clever Hans effect. And of course having a copy of the answer sheet before the test can reduce performance to an act of memorization.
Any of these systematic difficulties would drastically decrease the g-loading of a test. After going through all the ways that test questions can potentially be `gamed`, we must face the truism that a g-loaded question is g-loaded question. A given question may be solvable by more than one means, but if the ability to solve it by any and all of these means has a strong enough correlation with the ability to solve a diverse enough body of other seemingly unrelated problems involving complexity, then the ability to solve it is a mathematically valid demonstration of general intelligence per Spearman's factor analysis.
I've never heard a good argument against this, so I'm not interested in debating it.
On the other hand, I may be interested in debating subtler points about interplay of the general factor and specific factors amongst different groups of people with certain sets of DSM-IV diagnoses. For example it's generally accepted that high functioning autistics are better than the general population at performing some cognitive tasks, and worse than the general population at performing others. A significant proportion of autistics exhibit such large discrepancies on Raven's Advanced Progressive Matrixes vs. Wechsler tests that the discrepancies in scores actually far exceed what can be accounted for by the sum of these test's specific factors as normed on the general population. This is true even when comparing the Raven's scores against some of the Wechsler subtests considered to have the best g-loading.
Autism is not as well understood as some other DSM-IV diagnoses, but the effect involving IQ score discrepancies appears analogous to the way that ADHD can be accurately diagnosed from disparities between Wechsler series sub-test scores. There are non-IQ related cognitive skills tests which can effectively screen for sociopathy to the extent that test subjects are not aware of how the tests work. Additionally, there's some anecdotal evidence that sociopaths may generally fare better in chronometric IQ testing than in other forms of IQ testing.
There are some parallels between thought processes of autistics, sociopaths, and people with 3+ sigma general intelligence (1 or less out of every 1,000 for the general population, or IQ of 145+ with a standard deviation of 15). This mostly relates to being more rational/calculating as opposed to emotional/reactive. There are ways in which all three groups seem to act stupidly, but most of these don't really relate to lack of general intelligence. Some relate to different emotional needs, or emotion processing deficits in said neuroatypicals, and at least a few actually result from cognitive deficits in the aggregate population.
I know someone who's convinced that sociopathy occurs with a greater frequency among the highly intelligent. Personally I don't think true sociopathy occurs with much greater frequency, but I do think that similarities in dick-head behavior result from similar secondary causes. For example, I've noticed that extremely intelligent people:
* don't feel compelled to follow social norms for the sake of following social norms
* don't hold authority figures in high regard
* don't make decisions based on emotions, including empathy
* can be very adept at using self-manipulation while justifying unreasonable behavior
* tend to experience disdain to a heightened degree when they do experience it