Absolute distance from the equator and cranial capacity: 
 
r = +.62 (p < 10) 
 
cranial capacity = 2.5 cm3 × degrees latitude + 1257.3 cm3 
 
[Beals K. L., Smith C. & Dodd S. M. (1984). "Brain size, cranial morphology, climate, and time machines". Current Anthropology, 25, 301-318.] 
---- 
 
skin color to winter high temperature  
r = .85 (p < 0.001) 
 
IQ to skin color  
r = - .92 (p < 0.001) 
 
[Templer, D. I. and Arikawa, H. (2006). Temperature, skin color, per capita income, and IQ: An international perspective. Intelligence, 34, 121-139]

Sorry, it is based on an invention that has not been demonstrated as factual. 
 
The main alternative to it (that g is a quantitative aspect of brain physiology) hasn't been demonstrated as factual either. 
 
Your response is a non sequitur. The invention in question is the magic multiplier. Psychometric g is not an alternative. There is virtually no debate among scholars that intelligence consists of g and a few narrow abilities (which account for virtually no external validity). There is agreement that the variance in intelligence consists of both genetic and environmental factors. The disagreements are over magnitudes and mechanisms. 
 
I don't see why you jump to the conclusion, then, that Flynn's hypothesis is a "magical" invention. 
 
The Dickens-Flynn magic multipliers is an invention that has not been shown to exist. In fact, if it existed, adoption studies would have turned out differently. Massive intervention, including adoption does not alter adult intelligence. Flynn does not attempt to explain the things that show that his model is wrong (as expected). The fact is that all environmental influences combined account for 20% or less of the total variance in intelligence.  
 
Flynn does not have a credible explanation for environmental factors that account for the huge IQ difference between Ashkenazi Jews and other groups. Any reasonable model must account for every observed between-group difference. It needs to explain why Asians are smarter than Caucasians, but not smarter than Ashkenazi Jews and why 50/50 hybrids have IQs that are midway between the means of their two population groups, even when they share environments with adoptive families that do not share that mixture. 
 
Every finding to date shows that IQ is about 80% heritable in adults. The argument that there are environmental effects that hide and appear to be genetic simply does not stand close scrutiny and has never been demonstrated. The argument is no different than one that asserts the properties of an invisible god.

ben: 
 
Feel like backing up that claim? The Flynn-Dickens model is not based on "magic" but on science.  
 
Sorry, it is based on an invention that has not been demonstrated as factual. Even Flynn admits this. 
 
The idea that environment acts as a multiplier on genes has a great deal of explanatory power-- it explains the Flynn Effect as well as the twin studies. Flynn supports it quite well in his most recent book. What's your qualm with it, specifically? 
 
Can you cite sources that show any environmental factor that acts as a multiplier? As you know, Flynn has not cited any but you have accepted what he has imagined as a fact. Flynn has been quite clear in stating that he does not know what has caused the instability in test scores. In chatting with him, I found him to be honest, bright, and willing to admit when he is guessing. I asked him what portion of the IQ gains he has seen are g loaded. After repeating his thoughts about history as relating to testing, he admitted "I don't know." I asked the same question to John Raven and all he was willing to say is that people are getting better at solving abstract problems. As for g, he would not claim any gains.

I just looked at several papers that discuss the stability of intelligence.  
 
Results: 
 
The stability of general intelligence from early adulthood to middle-age 
Intelligence, In Press, Corrected Proof, Available online 8 February 2007 
Lars Larsen, Peter Hartmann and Helmuth Nyborg 
 
The differential previous stability coefficients were: 0.85 for g; 0.79 for arithmetic; and 0.82 for verbal ability. With respect to absolute previous termstabilitynext term of the specific tests, we found a significant increase in verbal score (mean scores; 107.16, 116.52), but no change in arithmetic score. Problems associated with different concepts of previous stability, level of analysis and potential practice effects were discussed. 
 
Overall, our findings provide support for the outcome of many other longitudinal studies, suggesting that general intelligence g shows high differential stability from early adulthood to middle-age. In fact, g measured in early adulthood predicts this very ability later in life with a precision that equals the reliability of the tests. 
 
### 
 
The Stability of Individual Differences in Mental Ability from Childhood to Old Age: Follow-up of the 1932 Scottish Mental Survey 
Intelligence, Volume 28, Issue 1, February 2000, Pages 49-55 
Ian J. Deary, Lawrence J. Whalley, Helen Lemmon, J. R. Crawford and John M. Starr 
 
Concurrent validity data are provided for the Moray House Test at age 11 (n=1,000) and age 77 years (n=97). The correlation between Moray House Test scores at age 11 and age 77 was 0.63, which adjusted to 0.73 when corrected for attenuation of ability range within the re-tested sample. This, the longest follow-up study of psychometric intelligence reported to date, shows that mental ability differences show substantial previous stability from childhood to late life. 
 
### 
 
School achievement strongly predicts midlife IQ  
Intelligence, Volume 35, Issue 6, November-December 2007, Pages 563-567 
Ruth Spinks, Stephan Arndt, Kristin Caspers, Rebecca Yucuis, L. William McKirgan, Christopher Pfalzgraf and Elijah Waterman 
 
The correlation between school achievement and WAIS-III FSIQ was r = 0.64, suggesting substantial life span previous stability in this relationship. Furthermore, elementary school achievement was substantially correlated with occupational status and household income at midlife. These findings suggest that standardized school achievement data is a useful measure of premorbid IQ. 
 
In summary the current report shows a moderate to high correlation between school achievement and IQ further substantiating the previous stability of IQ over time despite a wide variety of potential intervening factors. More significantly, grade school standardized school achievement test scores
More....

"If school mattered, IQ would increase in proportion to the years (and quality) of education received. That doesn't happen." 
 
But it does, doesn't it? N. Mackintosh, 'IQ and Human Intelligence', p.131: "One well-established educational correlate of IQ is the number of years of formal education recieved". The problem is to separate the chicken from the egg: do people with more/better education have higher IQs as a result, or do people have more/better education because they have higher IQs to start with? 
 
IQ and the number of years of education are strongly correlated and years of education can (if you understand the error involved) be used as a proxy for IQ. That does not mean that IQ is not stable, nor does it imply that IQ increases as people go through additional years of schooling. It simply means that more intelligent people are more inclined to pursue additional education and less intelligent ones are more inclined to do something else. This may be of interest to you: 
 
The prediction, from infancy, of adult IQ and achievement  
Intelligence, vol 35, issue 3, 2007 
Pages 225-231 
Joseph F. Fagan, Cynthia R. Holland and Karyn Wheeler 
 
IQ from 3 to 21 years appears to be quite stable and achievement at 21 years is predicted about as well by the IQ estimate at 3 years as it is by IQ at 21 years. 
 
Mackintosh, pp.131-139, quotes several lines of evidence suggesting that education does have some effect, the clearest being that children of the same age, who for institutional reasons have started school at different ages, differ significantly in IQ. Teasdale and Owen also report an educational effect in their studies of Danish army recruits. For older studies see Anastasi's 'Differential Psychology'. 
 
The only differences seen in IQ measurements that are attributable to education are gains in the non-g measurements that are attributable to specificity. So far, there have been no studies that have demonstrated that g increases as a result of any kind of training. The s gains are effectively errors in IQ measurement. If you coach someone to take the SAT, his scores will increase because he can apply learned approaches to some test items. His s loading will increase (for that test) and that means that the g loading of the test (for him) decreases. This score boost is not a boost in intelligence and is usually found to be temporary, such that a retest at a later date will show most of the boost to be gone. 
 
As to 'quality' of education, there were some relevant studies of the effects of the selective '11 plus' system in Britain. It was found that children who went to grammar schools (more academic education) had higher IQ at later ages than those who had the same tested IQ at age 11 but who for purely institutional reasons (e.g. shortage of places in their locality) went to the less academic 'secon
More....

Henri, 
 
vantage, see Flynn's response to question 3 which is pretty much the opposite of all you said. I know you probably disagree with him, but apparently he knows something about this subject. 
 
Yes, I disagree with him. He has stubbornly argued that the environment is the determinant of intelligence and has argued with weak and nonexistent evidence. Heritability can be measured by multiple and diverse approaches and all yield about the same result: 80% in adults. Flynn went to the extreme of inventing a magical relationship between the environment and genes and accompanied it with equations showing his idea to be possible, but it was entirely imaginary. [see Flynn and Dickens] 
 
I think it's plausible there are different types of intelligence and that emphasis on one or another can change those abilities or testing for them.  
 
Gardner and Sternberg also think there are different types of intelligence. Neither of them has demonstrated that such is the case. Like Flynn, they have used invention and assertion in place of science (factor analysis in particular). 
 
For example, there is no science instruction in Amish schools. Does that mean they will test poorly in raw cognitive abilities related to science even though their "overall average intelligence" seems to be the same? 
 
If you test someone on a subject that he has not been taught, you are administering an invalid test. You already know that don't you? If you give a Frenchman an IQ test in Chinese, the test is not valid, unless the Frenchman is a fluent (at native language level) Chinese speaker. But you can give Amish, French, and Chinese the same Raven's test and the results will have similar predictive validity.

So not only do the Amish have a 19th century lifestyle, their children are taught in a 19th century manner. I think this is closer to America in 1900 than anything you'll find in Brazil. And they score around the norm on standardized tests which might indicate "average intelligence".  
 
This is what you would expect. Intelligence is a reflection of biological factors, not of educational attainment. People tend to get educations that are roughly in line with their level of intelligence, but whether they do or don't, their IQs will remain relatively stable from about the age that they start school. If school mattered, IQ would increase in proportion to the years (and quality) of education received. That doesn't happen.

David, 
... suggested that the effect of increasing family size was to reduce mean IQ by between 2 and 4 points per extra child. Presumably reducing family size would increase IQ by a similar amount. But reducing average family size from, say, 4 to 2, would hardly account for more than 8 points of the Flynn Effect total of over 20.  
 
A good while back, parity was assumed to account for a decrease in IQ on a within family basis. Then a paper came out saying that the effect was entirely due to low IQ people having larger families (statistically) and smart people having small families. The family size thing is true. Researchers bought the explanation, until this year. A new study, with large N, was reported from Norway, showing that the first idea was correct. It showed that the largest IQ decrease was for the second born child. This opened the door (again) to arguments that some of the Flynn effect is due to a steady decrease in family size.

I think your point is valid, but I have not seen any papers that have reported changes in test g loading. I expect to be talking to the most knowledgable people (Lynn, Rushton, Deary, Flynn, etc.) in this area in just a few days and will see what they have to say. 
 
One factor is that the gains have been greatest among low IQ people. This may reflect a decrease in the number of people who are adversely affected by serious cognitive disorders. At the low end, the variance in IQ is believed to be overwhelmingly due to higher or lower g. The question is whether increased scores change the total test loading between the three factors: g, s, and e. (s = specificity and e = random error) The sum of the squares of these three factors must equal 1.0. That means that, if s loading increases g will decrease, assuming e is constant.

Matt wrote: "One thing I think I've come around to agreeing with Flynn on is that focus on g might not be the best path to fruitful research. My impression from what I've read up on is that most of the variance in g can be explained by a combination of working memory and transmission efficiency, which seem like pretty tractable problems to work on. g needs to be decomposed eventually, and now seems like a fine time for it." 
 
It is true that both working memory and speed related measurements correlate strongly with g, as does brain volume, especially at the cognitive centers. If by "decomposed" you mean that we should focus on the underlying biology, I agree with you. If you mean to simply understand the fine structure, that is already available from the factor analysis. 
 
"He himself has pointed to the fact that the scores are highest on the Raven's and that the questions on the WISC that correlate most highly with Raven's scores tend to show the biggest gains as a way of refuting Rushton. Which suggests that while crystallized g hasn't changed much, fluid g has." 
 
I disagree. No test, even the Raven's is free from s loading. There is plenty of specificity in the Raven's and WISC to allow for raw score increases. Keep in mind that Gc is about 80% g loaded. As you may know, Bouchard and Johnson have argued rather strongly that Gc is not a separate g (as in the Cattell-Horn model that identifies Gc and Gf, but does not extract g).

David wrote: 
 
There is respectable evidence for the proposition that IQ tests are more highly g-loaded at lower IQ levels as compared with higher IQ levels.  
 
This has been shown by factor analysis of top and bottom halves and by correlating the scores between two different IQ tests for high and low range testees. See: Kane, H. & Brand, C. R. (2001). 'The Structure of Intelligence in groups of varying cognitive ability: a test of Carroll's three-stratum theory.'  
 
This could be interpreted as showing that there are more ways of being intelligent than of being stupid. Or it could be that as you move up the scale of difficulty in an IQ test (e.g. Wechsler or Raven's) the questions are actually testing a more diverse range of abilities. Or both. 
 
It shows that group factors (expressed as residuals) are increasingly important as intelligence increases. Over most of the populatiuon, group factors (narrow abilities) account for virtually none of the external validity of the test. 
 
 
What I would like to know is whether this has actually been observed.  
I have not seen any papers that show a secular decline in g loading. One paper showed the expected FA gain over 20 years, but no change in inspection time measurements. The study was very carefully done with the same equipment, same school, and same researchers. If there was a g loading on the FA gain, the IT should have decreased. It did not. 
 
So far, there is little evidence of a gain in g, but there is one paper that will be presented next week that argues for a small g increase.