Thursday, 3 October 2013

The impossibility of precisely measuring high adult IQ

*

http://iqpersonalitygenius.blogspot.co.uk/2013/10/comparing-mental-age-ratio-with.html
*

16 comments:

  1. I agree with your thesis. There is another limit on measuring very high adult IQ: the subjects are very much cleverer than the people setting the tests. That problem must be much rarer with children.

    ReplyDelete
  2. @d - This does tend to induce test setters to go for speed of response to large numbers of not too difficult items, rather than intrinsic difficulty of content. Anne Roe candidly admitted this was the only way she could prepare a generic test of cognitive ability for her very high IQ sample in The Making of a Scientist.

    But you are correct that for this additional reason, IQ testing is much more coherent in children.

    Even a super-incredible high IQ of 190 in an *eight* year old - which is supposedly one-in-a-billion - corresponds to a mental age of slightly less than the average sixteen year old; and so is easily measurable without needing to make extrapolations or assumptions about distribution.

    ReplyDelete
  3. My own experience of clever people is that some are blessed with quick wits, some have great powers of reflection, and some seem to have both. Your Ms Roe would seem to have been probing only the first. But that may have more relevance to, say, a surgeon who has to make rapid decisions, than, say, to many scientists.

    ReplyDelete
  4. @d - Your general point is correct - but I don't think it applies to surgeons; or at least this comes pretty low down in the scale of what makes a good surgeon.

    The problems with relying on speed are that any deficiencies in vision, or muscles, or coordination become important - probably more important than intelligence differences.

    Also, you end up measuring specialist ability (eg in mathematics, or visuospatial skills) rather than 'g' or general intelligence.

    Probably, as I suggested in a previous post, physiological measurements of neural system efficiency, such as simple reaction time and other measures of that ilk - would be a better bet for measuring up to very high levels of g - such variables can also be measured right across the whole intelligence range, giving a measure of distribution (rather than working from an assumption of distribution).

    ReplyDelete
  5. "The problems with relying on speed are that any deficiencies in vision, or muscles, or coordination become important - probably more important than intelligence differences."

    Elsewhere you seem happy with the assumption that reaction speed is a good proxy for IQ.

    ReplyDelete
  6. @WmJas - Different point. You might recall when you did an IQ test there is a lot of scanning across the exam sheet to pick out different words and numbers and figures, picking out the place to mark your answer etc.

    Reaction times are just something like a light comes on and press a button a tiny distance - much less prone to this kind of problem.

    ReplyDelete
  7. @WmJas - It is not so much that RT is a 'proxy' for IQ - because IQ is a proxy as well - but that they are both measures of an as-yet ill defined property of the brain to do with speed and efficiency of processing.

    ReplyDelete
  8. Bruce, here is a question for you, regarding "reaction times" as a measure of intelligence, which is a concept that I think I understand although it still seems kind of strange.

    Just last night there were a few of us at a friend's house, and we were playing a card game that relies heavily on "reaction time" - you have to react in a certain way when certain cards are played. Do you think that in this sort of game, IQ will affect one's performance? Or am I still misunderstanding this concept?

    I happened to win the game, by the way, although I think that's because everyone except me was tipsy.

    ReplyDelete
  9. @SJ - Simple reaction time is only about a quarter to a third of a second - and the difference between fast and slow is on a range of a couple of hundred milliseconds.

    In simple RT there is no analysis, and perceptual and muscle factors are reduced to a minimum.

    Not many sports and recreations are affected by this kind of time differential between fast and slow RTs - and those which plausibly are would include foil fencing and ping pong - both dominated by higher IQ ethnicities.

    ReplyDelete
  10. In simple RT there is no analysis, and perceptual and muscle factors are reduced to a minimum.

    Oh, okay, this card game wouldn't count then. Anyway I think I get it now.

    ReplyDelete
  11. Why cant you recruit the high IQ people from Triple Nine Society or even Mensa? You wouldn't actually need to test 48 000 people...just find people who already have tested in the 0.1 percentile.

    ReplyDelete
  12. @Anon - the problem is knowing who really is in the 0.1 percentile - and for that you need a very large sample tat is representative of the population. There is no other way! The WORDSUM word-definition test on the GSS is the kind of sufficiently representative sample, but it has a very low IQ ceiling and only a few possible scores.

    ReplyDelete
  13. What you mean "who is really in the 0.1 percentile"? In Triple Nine Society you already have pool of people who have tested in the 0.1 percentile. Why could't you just recruit 15 of their members?

    ReplyDelete
  14. I think I just understood what you mean...you think people who have tested 0.1 percentile in various tests dont really have representative sample behind them?

    ReplyDelete
  15. @Same Anonymous - You need to understand sampling. But in a nutshell - maybe Tripe Nine Society claims or believes its members are in the top 0.1 % - but how could this claim actually be tested? It would need a very large sample, and the attempt would run into the problems I have outlined.

    ReplyDelete
  16. @Ooh - That's right. It is really difficult to get representative samples - in practice impossible. But even reasonably representative samples are unusual because it is very unusual to have a complete 'census' and equally unusual to be able to obtain a random sample of the complete population. This is just a fact!

    ReplyDelete