, ,

People who are unfamiliar with IQ research often feel it’s ridiculous to think that a simple IQ test, often measuring a small sample of mental abilities like pattern recognition or vocabulary, can possibly measure an entity as complex and multifaceted as intelligence. Well, the scientific rationale for using such one-dimensional tests to measure a trait as multidimensional as intelligence is that these test load high on g which stands for general intelligence. g is a hypothesized variable to explain why people who score high on one type of mental talent (i.e. rote memory) tend to score high on every other mental talent (i.e. verbal ability, spatial reasoning). So IQ tests don’t need to measure all mental talents (and impossible task); they just need to measure the few mental talents that load highest on g, because g predicts performance on every cognitive function ever studied by IQ researchers (and presumably every mental talent that exists).

But what if g has a maximum level and IQ tests continue to measure intelligence beyond it? In that case, at high levels, IQ tests will lose their theoretical justification and just be measuring an arbitrary sample of mental abilities that can’t possibly claim to be representative of all of intelligence. That doesn’t mean intelligence doesn’t exist beyond g, it just means that without g, IQ tests lose their scientific rigor.

How high does g go? In order to answer this, it helps to be able to measure it more directly. Eminent scientist and intelligence blogger Bruce Charlton believes (correctly in my opinion) that measures of reaction time (scored in milliseconds (ms)) are a more direct measure of g than paper and pencil IQ tests. Although g is not well understood at the biological level, it makes sense that the brain’s basic speed would influence virtually every cognitive function imaginable so reaction time likely plays a major role in whatever g is. Charlton recently wrote the following on this blog:

…I do not believe in the unity of g at very high levels – I think it breaks down into more specific abilities. More exactly, there is a maximum g – which would correspond roughly to a simple reaction time of 150 ms – and beyond that there are specific cognitive abilities added on. The IQ score will then vary according to the specific composition of the IQ test

Just as Charlton believes IQ scores don’t reflect g beyond a certain level, he also believes brain size does not reflect g beyond a certain level, once writing on this blog:

larger head/ brain size (if it is not pathological) usually goes with enhanced specialized non-g abilities – enlargement of some specific brain regions and cognitive functions rather than an increase in general processing efficiency which might underly g.

I’m not sure whether Charlton believes the g loading of brain size diminishes for the same reason that he believes IQ’s g loading diminishes or for some other reason(s), but either way, the implications of g maximizing commensurate with a maximum reaction speed is an incredibly significant insight. The only provisos I would add are 1) the putative 150 ms ceiling on reaction speed might be partly because of ceilings on sensory-motor speed and thus may not entirely reflect the maximum mental speed, and 2) g almost certainly consists of other brain properties beyond just speed and if these have higher ceilings, then individual differences in g can continue at levels far beyond where speed maxes out.

However because speed plays such a central role in g, even if g continues after speed maxes out, it will be greatly diminished.

So how fast is 150 ms? Scholar Arthur Jensen found that college students have a mean average reaction time of about 271 ms (Standard Deviation (SD) = 29 ms). If the SD seems small, it’s because they were given 20 trials and a lot of variability was cancelled out by taking each students average on all 20 trials. So if your reaction time was tested 20 times and you obtained an average reaction time of 150 ms, you would be 4.17 SD faster than the average college student. Assuming a Gaussian distribution, only about one in 75,000 college students should have an average reaction time this fast. Assuming the fastest reaction times are virtually exclusively found in people with at least some college education, and assuming about a third of Americans attend college, then an average reaction time this fast has a one in 225,000 level rarity in the general U.S. young adult population; equivalent to a reaction time IQ of 167 (sigma 15).

When I did the same analysis on choice reaction time using Jensen’s college stats, an average choice reaction time of 150 ms is equivalent to a reaction time IQ of 179. And this is using data on young adults. Because old people and children have much slower reaction times, their maximum reaction time IQ would be vastly higher.

So while I think Charlton is correct in asserting that g breaks down at very high levels, I believe those levels are so ridiculously high that this phenomenon is virtually never observed. IQ experts often speak of Spearman’s Law of Diminishing Returns (SLODR) which asserts that g explains less of the variability at high mental levels, but I think this phenomenon is hugely overrated and probably largely an artifact of low ceilings on many psychometric subtests. A lot of evidence contradicts SLODR. For example Jensen has reported that high IQ might actually be more genetic than low IQ and the Prometheus Society reported a potent 0.7 correlation between chronometric ability (Thinkfast scores) and SAT scores among extremely bright math students (though the sample size was small):

A study documented in the Prometheus MC Report, 1998/1999