A while back, I deduced that either shared environment produces Jensen Effects or that the race-genes-IQ debate should be over. And then I reasoned that since the debate is ongoing, shared environment probably produces Jensen Effects. To quote myself:
a. either the gap is causal cultural or causal biological.
if the gap is causal biological, it is either causal environmental-biological or causal genetic-biological.
the gap can not be causal environmental-biological (for reasons discussed before).
ergo, the gap must be either causal cultural or causal genetic-biological.
b. either the B/W gap is shared environmental or additive genetic (this follows from the logic of differential regression to the mean, discussed before)
c. either shared environmental influences can produce g(+) gaps or they can not.
the gap is decidedly g(+).
ergo, either the gap is not shared environmental or the gap is shared environmental + shared environmental influences can produce g(+) differences.
this reduces down to…
d. either the gap is additive genetic or the gap is shared environmental and shared environmental influences can produce g(+) gaps
e. either the debate over cause (wholly environmental versus some genetic) is over or it is not
it’s not over.
f. the Jensen effect on the MZA correlation either necessitates an anti-Jensen effect on the URT correlation or not.
if it does, since URT is a direct measure of the influence of shared environment, shared environmental influence must induce an anti-Jensen effect.
ergo, by d, the B/W would obviously have to be genetic.
h. the B/W gap is not obviously genetic. the inductive evidence of this is that the debate is still going on. And that I’m not compelled.
I believe that Jason Malloy and MH were the only ones who grasped my point. Whatever the case, I did and I decided to investigate. As expected, I found a meta-analytic Jensen Effect on Shared Environment. (This analysis is ongoing and the issue is not settled — but if I had to bet, based on n twins = 3000 K=10, I would say that the r (g x c^2) will come out to be non trivially positive and of a magnitude approaching r (g x h^2) — a conclusion that I don’t particularly appreciate). For example, here were the results from a large Austrian Twin Study based on the multidimensional aptitude battery n (tested twin pairs) ~ 1000.
MCV gives a r (g x c^2) of close to unity. This is even larger than the r (g x c^2) found in Samuelsson et al. (2007) based on achievement tests, n twin pairs = ~ 800. The upshot is that while James Flynn was completely wrong with regards to his suggestion that ethnic differences could be due to unshared environment (an idea on which I wasted an extraordinary amount of time) and with regards to his argument that the Flynn Effect was a Jensen Effect, Rushton, Jensen, Levin, Murray, Gottfredson, Woodely, Meisenberg, Hunt and many others seem to have been off in their arguments, suggestions, and imputations that a Jensen Effect implies biological or genetic etiology. Here, for example, is a typical example of the discussion on this topic:
If population group differences are greater on the more g- loaded and more heritable subtests, it implies they have agenetic origin (Jensen, 1973, 1998).
…A Jensen Effect for heritability has also been found, with the g loadings from various subtests correlating with the heritabilities of these same subtests (Jensen, 1998). A Jensen Effect for heritability provides biological evidence for a true genetic g, as opposed to the mere statistical reality of g .It makes problematic theories of intelligence that do not include a general factor as an underlying biological variable, but only explain the positive manifold, such as the model proposed by Dickens and Flynn (2001), and the mutualism model by van der Maas, Dolan, Grasman, Wicherts, Huizenga, and Raijmakers (2006)
(The rise and fall of the Flynn Effect as a reason to expect a narrowing of the Black–White IQ gap).
[T]he existence of the Jensen effect would appear to have significant ramifications. Firstly it effectively falsifies Thomsonite models of the development of intelligence (e.g. Bartholomew, Deary, & Lawn, 2009; Thomson, 1916; van der Maas et al., 2006), which argue that g arises chiefly from random sampling or mutualistic reinforcement amongst distinct ‘neural elements’ or bonds, rather than from the existence of a special quality or ‘mental energy’, as was first posited by Spearman (1927). This is because the Jensen effect reveals the existence of an apparent nexus amongst diverse biological variables and g , which suggests that the factor corresponds to something very fundamental to brain neurophysiology and genetics, rather than being a mere statistical regularity (Eysenck, 1987; Rushton & Jensen, 2010).
…It appears that dysgenic fertility is indeed a Jensen effect. This finding using a large and representative sample of the US population, along with subpopulations and a well validated measure of intelligence, therefore allows us to place dysgenic fertility into the genetic nexus of the Jensen effect. (A Jensen effect on dysgenic fertility: An analysis involving the National Longitudinal Survey of Youth.)
A non trivial positive r (g x c^2) makes the above reasonings untenable. What is suggested by this is that there is a strong anti-Jensen Effect on unshared environment, rather than a strong Jensen Effect on genes. In terms of path diagrams, while the effect of unshared environment is smaller on g than non g, the effect of both genes and shared environment is larger on g than non g. If you were just to pit genes versus shared environment — something I have yet to do — you might find no difference in terms of the correlation with g-loadings. Following Jensen & Rushton’s reasoning this would imply a shared environmental origin to g. Or, better, one might just say a non-shared environmental non-origin.
So I think that this is a problem for Jensen inferences. And this leads me to the question: Can MCV be salvaged?
Recently, I proposed to Meng Hu a multivariate approach to MCV. In instances where you have a vector of difference, a vector of g-loading, and a vector of cause e.g., heritability, biologicality, shared environmentality, etc. simply run a regression. You would want to look for mediation. Oddly, no one seems to have tried this.
To given an example, Kan (2011) found a Jensen Effect on heritability and on cultural loadings. This led him to devise a complex Cov(GE)-g model. To super simplify: people seek out, in accordance with their genetic dispositions, environments which increase g — hence: r g-loadings-“culture loading” and r g-loadings-h^2. With a path running, presumably: genes–>cultural–>g. But I noted that cultural loaded tests are more cognitively complex than cultural less loaded tests and so likely to be more g-loaded. In counter, Kees-Jan argued that, regardless, culturally loaded tests should be less heritable, because they are more culturally loaded. Presumably, he and his Co. make the same argument in “Kan, Wicherts, Dolan, and van der Maas. On the Nature and Nurture of Intelligence and Specific Cognitive Abilities: The More Heritable, the More Culture-Dependent? Psychological Science”. But this was just an argument from the “heritability paradox”. On this paradox, Jensen, 2006 pg. 133) noted:
This is the name given to the frequent and seemingly surprising finding that h2 increases as a function of task complexity and also as a function of the degree to which tasks call for prior learned knowledge and skills. The notion that this is paradoxical results from the expectation of some theorists that the more elemental or basic components of performance in the causal hierarchy of cognition, going from stimulus (the problem) to response (the answer), are far less removed from the underlying brain mechanisms than are the final outcomes of complex problem solving. The tests of greater complexity are considered to be more differentially influenced by environmental conditions, such as prior cultural–educational opportunities for having acquired the specific knowledge or skills called for by the test. Therefore, it is expected that individual differences in tests that are more dependent on past-acquired knowledge and skills, such as typical tests of IQ and scholastic achievement, should reflect genetic factors to a lesser degree than RT on relatively simple ECTs, which are assumed to depend almost exclusively on the most basic or elemental cognitive processes. Presumably, individual differences in RT on such relatively simple ECTs scarcely involve differences in environmental factors influencing specific cultural or scholastic knowledge and skills, and therefore ECTs should have higher h2 than the more experientially loaded psychometric tests. It turns out, however, that the empirical finding described as the “heritability paradox” is not really paradoxical at all. It is actually an example of a well-known effect in psychometrics — the aggregation of causal effects, whereby the sum or average of a number of correlated factors has greater reliability and generality than the average of the reliability coefficients of each of the aggregate’s separate components. The factor common to a number of somewhat different but correlated ECTs, therefore, should be expected to have a higher phenotype–genotype correlation (and thus higher h2) than its separate elements.
“Cultural loaded” — really: “informationally loaded” — tests can index a broader range of functioning which gives a more reliable index of psychometric g and by way of psychometric g of heritable g. “Cultural loadedness” will only reduce heritability if the background information is heterogeneous with respect to the sample — in this case between twin pairs. This trivially obvious point seemed to have missed my interlocutor.
Whatever the case, this is speculation either way. Culture-loadings correlate with heritability and with g. And heritability correlates with g. Does g mediate the cultural-heritability correlation as I would predict or does culture mediate the heritability-g correlation? For example, here was some data from Kan (2011):
And here were the bivariate correlations:
Well, to make a shorter explanation even shorter, here were the multivariate results:
Controlling for test battery, the association between heritability and cultural loading was fully mediated by g, but the association between heritability and g was only partially mediated by cultural loading. So, based on this analysis, it would look as if my position was more correct. But, of course, one would want to conduct several Multivariate Method of Correlated Vector (MMCV) analyses before deciding anything.
My point here is simply to (re?)-introduce a less simplistic version of MCV, a version which might be of use since, if I am right about C^2 and g, the Jensen Effect network is more complex than previously thought.
Now, what can be explored? Well, for starters, Kan (2011) chapter 4 presents g-loadings, cultural loadings, and B/W difference data. My prediction: Same as above — the association between cultural loading and b/w d is mediated in full by g. Also, Spitz (1988) presents data on g-loadings, mental retardation scores, and heritability estimates. I guess the immediate issue of interest is whether various r (vector1 x h^2) is mediated by g. For MR, that can be checked with the Spitz results.
(Let me know what you get MH.)