Tags
brain size, Dutch, Flynn Effect, gifted, height, IQ, James Flynn, Meng Hu, non-verbal IQ, nutrition, Raven progressive matrices, Richard Lynn, schooling, verbal-numerical IQ
As I explained in a previous post, a popular belief is that the Flynn Effect disproportionately impacts the lower parts of the bell curve. The reason this theory is so popular is because hypothesized causes of the Flynn Effect (increased schooling, better nutrition) are assumed to help primarily the most disadvantaged people, but have diminishing returns as you move up the scale. It’s also popular because it nicely explains why past generations had so many impressive scientists, thinkers and writers despite the average test score being so low. And for culturally biased tests like the SAT, the Flynn Effect is indeed much smaller at the high end. The percent of all American 17 years old (not just those who finish high school) capable of scoring at the highest levels on the Verbal SAT has not been increasing according to the book “The Bell Curve”, even though at the low end of the distribution, literary rates have been skyrocketing over the 20th century.
However when it comes to one of the best and most culture fair tests we have (the Raven Progressive Matrices), this theory absolutely collapses. In the previous post I talked about the dramatic Raven Flynn Effect in the brightest 10% of British men over the 20th century. But perhaps I wasn’t looking high enough. Perhaps the Flynn Effect disappears for the top 1%, or the the top 0.1%. Fortunately, the always resourceful James Flynn, long ago found the data. If you look at table 18 of a post by blogger Meng Hu, you can see Flynn’s data. What it shows is that in the Netherlands (where Raven scores jumped the equivalent of 21 IQ points in 30 years), getting even extremely high scores became a lot more common. By equating the changing frequency to the IQ distribution (mean 100, SD 15) we see that an IQ of 130 on the older norms equaled an IQ 110 on the newer norms. An IQ of 140 on the older norms equals an IQ of 120 on the newer norms, and an IQ of 150 on the older norms, equals an IQ of 130 on newer norms. So it seems the entire distribution increased by about 20 points from 1952 to 1982, including the extremely brilliant.
So what has caused these massive gains in such a short span of time? [UPDATE June 22/2014: I was informed in the comment section that the Flynn Effect gains among the gifted may have been theoretical rather than actually observed data, in which case, my above analysis is circular speculation]. James Flynn equated the older and younger generation on schooling and found that changes in education explained only about 1 IQ point of the massive rapid gains. Virtually nothing. And given the relatively high correlation between IQ and schooling (within generations), one can assume that people above IQ 150 were already finishing high school at high rates in 1952 so schooling seems an unlikely cause, especially for a test as seemingly culture fair as the Raven. So that brings us to Richard Lynn’s nutrition hypothesis which argues that 20th century nutrition has been increasing both intelligence and height. This theory is especially compelling when one consider that the unusually large Flynn Effect the Dutch enjoyed (7 points a decade!) is nicely paralleled by unusually rapid height gains over the same span of time.
Some interesting quotes from an article in USA Today:
The Dutch were not noted for their height until recently. It was only in the 1950s that they passed the Americans, who stood tallest for most of the last 200 years, said John Komlos, a leading expert on the subject who is professor of economic history at the University of Munich in Germany. He said the United States has now fallen behind Denmark.
So, like the Raven IQ gains, the height gains seem to have taken off in the 1950s
Many Dutch are much taller than average. So many, in fact, that four years ago the government adjusted building codes to raise the standards for door frames and ceilings. Doors must now be 7-feet, 6 1/2-inches high.
So like the Raven IQ gains, the height gains have been very strong at the highest levels.
In 1848, one man out of four was rejected by the Dutch military because he was shorter than 5-foot-2. Today, fewer than one in 1,000 is that short.
This is an astonishing statistic. It suggests that height among the Dutch has increased by 3.8 standard deviations since the 19th century. Does that imply that biological intelligence has increased by 57 IQ points?!! Obviously the nutrition theory is wrong to imply such absurd conclusions? Not so fast. Two points must be understood:
1) Primarily NON-VERBAL intelligence is affected by nutrition. Even the malnourished mind preserves virtually all its verbal-numerical intelligence. And this is not because these are crystallized skills acquired before the onset of malnutrition. Even when malnutrition occurs before birth, the ability to acquire verbal-numerical skills is virtually unscathed and these are the abilities that determine cultural achievements. Non-verbal disabilities are relatively invisible. So even if overall intelligence had increased by 3.8 SD, the increase would be extremely lopsided, with non-verbal IQ increasing much more, and verbal-numeral IQ hardly increasing at all. It’s unclear whether one should expect a 3.8 SD increase in overall IQ (verbal + non-verbal), or a 3.8 SD increase in just non-verbal IQ, in which case overall IQ would be much less increased. But either way, verbal-numerical ability (the building blocks of culture) would be virtually unchanged.
2) The reason height gains are relevant is that they parallel gains in brain size (and probably other properties of the brain too). But it’s unlikely that a 3.8 SD gain in height implies a 3.8 SD gain in brain size, since brain size probably has a much higher genetic floor than height does, which is why people of extremely short stature often appear to have big heads.
Lastly, the paradox of huge Raven gains (even at the high end) while no comparable high end gains on the Verbal SAT is resolved. Verbal-numerical ability is virtually untouched by nutrition, though the rise in education has improved crystallized forms of such skills at the low end (literacy rates) but not the high end because the gifted have always been educated. By contrast nutrition improves the ENTIRE distribution, hence even the gifted show huge Raven Flynn Effects. As for the Math SAT, that’s a mix of Verbal-numerical ability and NON-VERBAL problem solving, so it shows more high end gains than the Verbal SAT, but less than a pure non-verbal culture reduced test like the Raven.
Brain Size 
Thanks for pointing that out. You’re probably right. The huge gains at IQ 150 should have been a red flag given the relatively low ceiling of the Raven.I was trying to use actual data to prove the theory that Flynn Gains were uniform, but that data may have just been estimates based on the very theory I’m trying to prove. Circular analysis. I should add an update to that section of the post explaining the uncertainty.
i hope i’m not mischaracterizing anyone’s research in the following:
armstrong & woodley found a positive relationship between “g” & flynn effect on matrices type tasks – smarter people had more flynn effect on matrices type tasks, b/c smarter people are more adept at picking up pattern solving rules (e.g., if we teach a whole class of kids “here are 6 pattern rules that can help you solve matrices items” — smarter kids will benefit more from such instruction).
i think armstrong & woodley had it in their original article, but may have taken it out due to length concerns (or to elaborate upon it at a later point).
smart people benefit more from good instruction – smart people benefit more when taught general pattern solving rubrics (whether taught indirectly thru culture or directly).
Matrices gains without verbal gains may be due to such rule detection skills – &/or “g” going down, while matrix rule finding rules become more universal. &/or “the xbox/nintendo effect – we’re getting better at nonverbal/visual skills but losing vocabulary!” (the latter would play better in the media:)
btw, i’ve seen your excellent comments at other blogs & am pleased to find your blog!
btw, i’ve seen your excellent comments at other blogs & am pleased to find your blog!
Thank you so much! I’m pleased you’ve found my blog too.
I haven’t yet been able to access the Armstrong & Woodley paper, but from what I’ve read about the paper; it sounds like they implied that inductive reasoning can be taught which sounds impossible, because then intelligence itself could almost be taught..
But perhaps I’m oversimplifying or misunderstanding their argument, or perhaps their views have changed since writing it.
But a lot of people do believe intelligence can be taught. It’s sometimes said that the value of education is not so much what we learn, but that school teaches us how to think.
only the simple pattern rules of matrices type tasks can be taught or “gamed” – & are taught in most K-3 grades – do all the rows have the same things? do all the columns have the same things? do the outside boxes add up to the middle box?…
such teaching has lowered the value of matrices type tasks as a measure of intelligence – & caused it to be renormed much harder:)
intelligence can’t be taught – Jensen had enough evidence in 1969 to say it can’t – teaching matrices/pattern finding rules doesn’t teach intelligence – nothing like that ever generalizes to reality beyond that simple task.
the “miracle in milwaukee” & also raven feuerstein tried to convince us intelligence can be taught in the 1970s & 80’s with pie in the sky idiocy – teaching the test doesn’t generalize to anything else! (you can get a grant to work with inner city KG’ers for a year or 2 & show gains – but those go away quickly & are not real – you’re not going to get a grant to show how it works over 50 years, b/c the public doesn’t really want to know it doesn’t work!)
intelligence is .75 to .80 heredity — & the part that’s not genetic: we have no idea what influences that. we only know how to make intelligence a bit lower than it’s preset to be (parasites, starvation, head injuries, etc.)
school doesn’t make anyone smarter. the general public might like to believe it, but they’re wrong.
schools teaching “critical thinking” skills is idiocy – that’s school taking credit for the normal maturation of smart brains. other brains cannot be taught so well. good instruction separates out the distribution even more – the brightest gain the most from it, the dullest gain very little. the world needs to grow up & understand this 🙂
only the simple pattern rules of matrices type tasks can be taught or “gamed” – & are taught in most K-3 grades – do all the rows have the same things? do all the columns have the same things? do the outside boxes add up to the middle box?…
such teaching has lowered the value of matrices type tasks as a measure of intelligence – & caused it to be renormed much harder:)
This is a very good point I hadn’t considered. All of our modern practice reading charts and tables with columns and rows could certainly be having a narrow transfer effect on Raven scores. However the most likely place where we practice such skills is in school, and Flynn found that controlling for schooling across generations only reduced the huge Dutch Raven 21 point Flynn Effect by about 1 point.
Now you imply the effect might be most pronounced in kids (K through grade 3) and the Dutch Flynn Effect was in adults, all of whom probably had at least primary school education in both cohorts so such early grade skills might be less relevant at this age. But it is in adults where the biggest Raven Flynn Effects have been documented. Although perhaps even equating for schooling, younger generations have had more experience interpreting columns and rows? However when you consider that there are other non-verbal tests that don’t require such skills which show equally large Flynn Effects, such explanations become superfluous, especially when you also consider that the full Flynn Effect is apparent in toddlers, perhaps even infants. It is for these reasons that Richard Lynn’s nutrition theory is so compelling.
intelligence can’t be taught – Jensen had enough evidence in 1969 to say it can’t – teaching matrices/pattern finding rules doesn’t teach intelligence – nothing like that ever generalizes to reality beyond that simple task.
the “miracle in milwaukee” & also raven feuerstein tried to convince us intelligence can be taught in the 1970s & 80′s with pie in the sky idiocy – teaching the test doesn’t generalize to anything else! (you can get a grant to work with inner city KG’ers for a year or 2 & show gains – but those go away quickly & are not real – you’re not going to get a grant to show how it works over 50 years, b/c the public doesn’t really want to know it doesn’t work!)
I agree, however the fact that gains are short-term should not be used to dismiss them. As “IC” argued on another thread, intelligence might be like a muscle that gets stronger with exercise but atrophies with lack of use. See the comment section:
https://brainsize.wordpress.com/2014/06/10/more-thoughts-on-the-flynn-effect/
I personally rejected that analogy, but whatever the correct analogy might be, the real problem with the “miracle in Milwaukee” was not its fleetingness, but its lack of transfer beyond the IQ test itself, as you point out..
intelligence is .75 to .80 heredity — & the part that’s not genetic: we have no idea what influences that. we only know how to make intelligence a bit lower than it’s preset to be (parasites, starvation, head injuries, etc.)
Yes, as Jensen said, the preponderance of evidence suggests that g is a physiological variable, uninfluenced by cultural or psychological experience. Only genetics and the BIOLOGICAL environment (i.e. prenatal health, alcohol, drugs, head injuries, nutrition ect) seem to explain the (American) variation in IQ (or at least g) by later adulthood.
This is an old thread, but I found it, so …
“And for culturally biased tests like the SAT, ”
Not true. If it were, it would be very easy to show which questions were guilty. Results on the SAT vary by race, but the rank order of difficulty does not. If a question were culturally biased against one group, the difficulty rank of that question would increase for that group. All questions are pre-tested – they are asked and scored, but not used in the student’s final score. Any questions that fail the rank order test are pulled. So we know that African American results by question are worse than for European Americans, the rank order difficulty of the different questions does not. The End.
The cultural bias thing was looked at very closely by those in the field and sorted out decades ago. Get with the millenium.
If the criterion the test is trying to predict is itself culturally biased, then predictive validity does not prove culture fairness. Also a test can be culturally biased between generation while fair to all races of the same generation. Further, the black-white SAT gap has narrowed in recent decades while the IQ gap has remained constant.