IQ by definition fits the bell curve. A given normalized IQ test is only good for a given rang and time period and when incorrectly interpreted outside of that range will produce excesive people with high IQ's. Also a test that's accurate +/- 10 IQ points is going to bump more people from 150 to 160 than drop people from 160 to 150 simply because there are more people at 150 than 160. Not to mention the tendency for people to pick the highest score vs the average.
Yes, you can define a concept as a normal distribution centered at 100 with a standard deviation of 15. But if you claim that this concept can be measured with IQ tests, and then the observed distribution of IQ scores doesn't match the predicted curve - it's the curve that's wrong, not the tests.
The studies that the "fat tail" results were from all used the Stanford Binet L-M test, which has a ceiling of around 230.
Sit back and think about this for a second, if you define IQ in some other fashion you can't really have multiple IQ tests just the score on one specific test. As to having problems with a specific test, just because a specific laser rang finder has issues does not mean we need to redefine the inch.
PS: There is no way they had enough data to support an IQ rang up to 230. Do you bave any idea how many people you need to sample to have 50 people with an IQ between 205 and 220.
Do you have evidence for that? I've worked on cognitive ability testing as it related to workplace performance, and have never seen anything that deviated dramatically from a normal distribution, especially at the high end.
Google [iq fat tails] and there're a bunch of articles on it (and one comment I wrote here about 4 years ago). The original data source for most of the articles is Terman's 1921 study of high-IQ people; they've plotted out the observed frequency of Terman's data against a normal distribution and found that it deviated markedly after about 3-4 SD.
Some additional Googling seems to have found some other independent studies:
I'm curious what sort of population your work draws from. The results above showed that IQ follows a normal distribution until about 140; other papers I've read indicate that IQ correlates with life outcomes until an IQ of about 140, and then appears completely uncorrelated. If you're studying workplace performance, I wouldn't be surprised if a good fraction of high-IQ people simply aren't in the workplace. (See eg. Christopher Langan.)
The citation I was thinking of was from Daniel Goleman's Working with Emotional Intelligence, where one of the findings presented was that a moderately high IQ (usually in the 120-130 range) is often a prerequisite for entering a demanding profession like doctor, lawyer, or computer programmer, but continued success in the field depends more upon emotional skills like confidence, perseverance, resilience, social skills, and leadership.
With a bit of Googling, I've found some other support for this, including the Terman study:
"Our conclusion is that for subjects brought up under present-day educational regimes, excess in IQ above 140 or 150 adds little to one's achievement in the early adult years." - Louis Terman, 39th Yearbook of the National Society for the Study of Education Part I, pp. 83-84
> ut continued success in the field depends more upon emotional skills like confidence, perseverance, resilience, social skills, and leadership.
Eh. That simply sounds like the correlation weakens a bit, but is far from the claims people make like 'IQ is irrelevant'.
> "Our conclusion is that for subjects brought up under present-day educational regimes, excess in IQ above 140 or 150 adds little to one's achievement in the early adult years." - Louis Terman, 39th Yearbook of the National Society for the Study of Education Part I, pp. 83-84
Terman may have thought so, but with the full dataset this is clearly not so. Check out http://www.iza.org/conference_files/CoNoCoSk2011/gensowski_m... 'The Effects of Education, Personality, and IQ on Earnings of High-Ability Men', Gensowski et al 2011; IQ never stops mattering, even if personality factors start to matter more.
(Always funny how people can look at a study which goes something like 'X correlates .4 and Y correlates .3, but in the top 1% by X, the correlations are .3 and .4 respectively' and go 'X doesn't matter!' Says something about what they want to believe about X, I think.)
Just remember that the distribution of the test results depends on the tool itself as well as testing conditions. So I would say that such anomalies are rather the problem with the measurement than characteristic of the population.
PS: Often the limit is as low as 135.