culture g
The central point of my prior post on the subject of race and intelligence was that the proposition that differences in performance on IQ tests between various racial and ethnic groups can be materially explained by differences in the genetics of the groups is unproven. A substantive complaint of those who make what I call in the post the “econometric” argument (as opposed to the “physical mechanism” argument) for this proposition is that the hereditarians keep knocking down hypothesized environmental explanation after environmental explanation for difference in racial group IQ test performance, and that therefore the remaining “space” for a non-falsified environment-only theory has become so narrow and twisted that any environment-only theory must be extremely ad hoc and counter-intuitive. In contrast, they argue that the elegant genetic theory keeps passing falsification tests. The hereditarians claim that the environmental theorists seem like they will just keep throwing up any objection to avoid drawing natural conclusions from the data, and that therefore the environmental theorists have moved beyond legitimate scientific debate to pure denialism.
On one hand, it seems to me that there is something to this argument. To the extent that the hereditarians have, in the face of massive social and professional disdain, disproven the crudest alternative theories, they have done a service to science and society. On the other hand I think they have drastically over-stated their case, both rhetorically and substantively. I think it is quite possible to present an environmental theory that conforms to current finding without being either ad hoc or counter-intuitive. Here goes:
As a starting point illustration, consider the following type of natural experiment that is well-known in the research literature, but less frequently noted in the public debate: practice tests increase measured IQ – a lot. Taking one practice IQ test raises average IQ on major IQ tests by about 7 points (~0.5 STDEV). Think about that: I can close half or more of several widely-discussed IQ gaps between groups just by letting the lower-scoring group take one practice test. Now, there are huge limitations to this: scores don’t keep improving as a subject takes more practice tests; it doesn’t appear to create lasting gains; and, importantly, it doesn’t improve g, which is to say, that it doesn’t create the ability to answer a more generalized set of questions across a variety of tests. In the limit case, if I gave someone an answer sheet to an IQ test and then had them copy these answers on the test form, I could create an arbitrarily high IQ score without increasing g at all. That is, I would presumably produce no increase in the subject’s ability to answer as-yet-unseen relevant questions.
But now consider the example of the reverse digit span question, which has been an important component of many IQ tests, and is highly predictive of overall IQ. In this question the subject is presented orally with a digit sequence (e.g., ‘76138’) and asked to repeat it backwards. Additional sequences of increasing length are presented until the subject can no longer successfully complete the task. Try it – it’s harder than it seems like it might be, even with short sequences. Now, when next presented with a sequence, silently repeat the number twice to yourself, and then visually imagine it written in a bright color (also try the alternative trick of imaging dialing the number on a phone keypad, which works better for some people). Now, as you get to longer sequences continue this procedure, but also “chunk” the sequence into sub-sequences of three or four digits each. Practice this repeatedly. If you’re like me, this will increase your reverse digit span. Congratulations, your IQ just went up – feel smarter?
A valid complaint about this is that we’ve just gamed the question and haven’t really increased your g. To be fair, though, I’ve just improved your ability to answer a whole class of questions, rather than just a single question – we’re clearly operating a higher level of abstraction than just rote memorization of a given list of answers.
But now think about moving to a yet higher level of abstraction: the ability to invent such tricks in the face of as-yet unseen question classes. In the case of reverse digit span, it would be the capability to invent some kind of (likely unarticulated) processes like the ones that I have described on-the-fly when presented with this question, never having considered this question until that moment. Surely at this point we are describing something in the neighborhood of g. But ask yourself this: do you think that a person would be more likely to invent such processes on-the-fly after spending 17 years immersed in a culture that demanded some combination of rote memorization of often unintelligible sequences, frequent use of numbers and constant testing than someone of identical genetic endowment who had not? Does this description remind you of any specific cultures?
If there are cultural environments that provide experiences that, at a high level of abstraction, develop a set of mental skills that can be translated on-the-fly to create “tricks” that enable someone to answer even highly g-loaded previously-unseen classes of IQ test questions, we can easily see how such difficult-to-quantify cultural influences could increase measured g. But just as we might infer that there may be some common innate characteristic g at the individual level that is not obvious upon inspection but appears as an artifact of the subject’s demonstrated level of ability on lots of superficially unrelated test questions, why could we not equivalently hypothesize that differences between IQ scores of various cultural groups could easily be (in part) the product of such cultural factors that demonstrate measured impact across seemingly-disparate test questions: what we could call “culture g.”
I find this hypothesis to be neither ad hoc nor counter-intuitive – in fact, it’s not even especially new, as it lines up well with folk wisdom that predates the modern era of IQ testing. It is not falsified, as far as I can see, by any major research findings. None of this is to say that it is true, or even more probable than some genetic or environmental alternative; merely that one can construct an environmental theory for specific group differences that is neither falsified nor implausible on its face.
Jim —
The crucial issue here is that IQ is not a ratio scale measurement. That is, it is not measured in meters or seconds. This is at the root of the controversy over comparisons.
No one disputes that people in industrialized countries have gotten taller in the last century, as height is measured in meters and is a ratio scale quantity.
That is, we know that 1 meter is 1 meter, whether it was measured in the US or Africa, whether it was measured in 1907 or 2007, whether it was measured by country X or person Y. But IQ is not a ratio scale measurement and does not display the same time/space/group invariance.
If instead we had a ratio scale measurement which was a proxy for IQ — say in cubic centimeters or milliseconds or chemical molarities — then we’d be in much better shape.
One thing you should be aware of is that there are several such neurophysiological predictors of g. These include:
1) MRI measured brain volumes
http://today.uci.edu/news/release_detail.asp?key=1187
Human intelligence determined by volume and location of gray matter tissue in brain
2) Reaction times
http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6SYP-3TGVR4P-B&_user=145269&_coverDate=09%2F30%2F1998&_rdoc=1&_fmt=&_orig=search&_sort=d&view=c&_acct=C000012078&_version=1&_urlVersion=0&_userid=145269&md5=78131c1f41a069dbfb9b3d61e9a9d0b7
The relationship of speed-of-information-processing (SIP), as derived from reaction times (RTs) on experimental tasks, and intelligence has been extensively studied.
3) Glucose metabolic rates
http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6WNP-4CTN475-2&_user=145269&_coverDate=09%2F30%2F2004&_rdoc=1&_fmt=&_orig=search&_sort=d&view=c&_acct=C000012078&_version=1&_urlVersion=0&_userid=145269&md5=9be3aed44e684a6a785579b5dc621b34
Having more gray matter in an area available for processing may also account for the inverse correlations reported in several studies between brain activation and good performance on g-loaded tasks (Haier et al., 1988 and Parks et al., 1988), provided that more gray matter results in less energy use when that area is employed (efficiently) for specific cognitive tasks. Although inverse correlations have been reported between brain size and cerebral glucose metabolic rate (Haier et al., 1995; Hatazawa et al., 1987 and Yoshii et al., 1988), additional study is needed of regional correlations between cerebral structure and function.
4) Neuronal conduction velocities
http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6T0D-4G94J35-1&_user=145269&_coverDate=12%2F31%2F2005&_rdoc=1&_fmt=&_orig=search&_sort=d&view=c&_acct=C000012078&_version=1&_urlVersion=0&_userid=145269&md5=190ce8dc7dffd8e9e2922a5f02cac39c
Brain nerve conduction velocity is a valid and useful construct for studying human cognitive abilities
——-
No one has yet done the study to the best of my knowledge, but it is a simple matter conceptually to measure a number of these covariates (there are others) on the same individuals along with IQ. One could then regress IQ on these covariates to obtain intelligence as a function of indisputable ratio scale measurements.
One would want to repeat this study in different locales, at different times, and with different groups to see whether the regression equation was time/space/group invariant.
It would be particularly interesting to see if groups that scored differently on IQ tests actually had neurophysiological differences — differences in brain morphology, differences in the velocity with which their axons transmitted signals, differences in the refresh rates of their synaptic clefts, etcetera.
One of the most important consequences of such a study would be that it would resolve the Flynn effect. Given an equation for IQ which depended entirely on ratio scale variables, one could simply see whether the underlying neurophysiological ratio scale measurements have been changing (as they have been with height).
One could also see which environmental interventions actually changed these neurophysiological variables.
Even in the absence of such a study, though, it is important to note that at least some of the variables known to predict IQ differ between populations — MRI measured brain volumes among them.
— john hahvahd · Nov 29, 05:00 PM · #
I teach theater full-time in a small selective liberal arts college. I am also a professional actor. Memorizing texts for performance (and teaching others how to memorize them) has been an important part of my life for thirty-five years. As anyone who deals with memorization issues knows, how something is memorized affects how completely it is memorized and how long it’s retained. Think for example how many songs you probably know the lyrics to along with thier melodies even though you may not have heard or sung them in years or, in my case, decades.
The example you gave on the digit test points out a very simple fact. It is among other things a test of short term memory. The “tricks” you used to memorize longer and longer strings are nothing more than devices designed to improve memory which can be taught or simply invented on the spot. And they help a lot. I learned to visualize phone numbers to remember them because taking pen and paper into bars seemed, well…obvious. I learned a long time ago to repeat someone’s name aloud as soon we’ve been introduced and use it two-three times in the subsequent conversation. These things are tricks in a sense but you’d be surprised by how much they impress people (someone you want to date for example whose phone number you don’t write down but remember!) with the level of attention you seem to have paid to them.
My point is this. IQ tests which purport to measure “G” actually measure a number of things that we call by other names but are not, in lay discussions of the subject at least, regarded as part of intelligence. In this case it’s memory. In other cases it measures the tolerance for handling the pressure of taking tests in the first place. In other cases
it’s a snapshot of skill sets or subject area knowledge (vocabulary for instance) or perhaps something as simple as eyesight or the presence or absence of dyslexia. Proponents of genetic differences in IQ are (deeply) fond of telling us how that they control for variables in analyzing IQ scores across ethnic groups in order to isolate the inheritable bits. Perhaps they should consider designing IQ tests that control for variables that may affect the scores but cannot themselves be called intelligence in any meaningful way. Either that or admit that what they call intelligence is itself too vague a concept to have the heavy meanings they ascribe to it.
— davido · Nov 29, 09:23 PM · #
Note that this is a clearer, better-developed version of Flynn’s dogs and rabbits are both mammals line. I think it applies best to tests such as the Raven, which obviously has nothing to do with dogs or rabbits, but still requires subjects to construct formal classifications.
The training effect certainly gets nowhere near enough attention. Children in modern Western societies are exposed to a tremendous number of IQ-like tests and cognitive strategies, which I’m sure have done an enormous amount to complicate psychometry. If the Flynn Effect shows anything, it shows that cross-cultural IQ studies, such as those of Lynn, are almost useless. There is no conceivable distinction between cross-cultural and cross-temporal studies, and the latter are clearly producing a false signal. QED.
Nonetheless, there remains a large base of studies with much smaller cross-cultural variation. The Minnesota Transracial Adoption study being of course the gold standard, especially because it was conducted not by Pioneer fundees but by researchers who clearly hoped for and expected the opposite result.
And you still have not answered the most fundamental criticism which I and others have raised, which is that you have not one but two competing hypotheses: either (a) there are significant variations in cognitive capacity among modern human populations, or (b) there aren’t.
As John Derbyshire points out, some rather important decisions rest on the question of (a) versus (b).
When you accept the framework that we must assume (b) until (a) is “proven” – a word that is really better reserved for mathematical deduction – you have already made a philosophical error. If decisions must be made, the same standards of evidence should be applied to both hypotheses, and we should look at which is more compelling.
Examining the case for either hypothesis in isolation is a good way to confuse oneself.
— Mencius · Nov 29, 09:33 PM · #
Another interesting study would seek to compute ratio-variance in personality types among different ethnic groups. Insofar as race can be tied to geography, it stands to reason that different environments would demand different combinations of logic-sets (personality-types).
On the intelligence question, variance in intelligence distribution among groups is almost certainly insignificant. Since all human societies ultimately sprang from a genetic bottle-neck in the not too distant past, and since differences in enviro-social complexity — surely the determiner of group intelligence distributions — are not nearly as pronounced as we imagine they are*, observed intelligence differences are most likely noise.
The implication, it seems to me, is that currently-poorly-performing ethnic groups hold intelligence potential in escrow, awaiting a set of cultural conditions precedent.
* algorithmic complexity: the amount of irreducible information coded into a society.
— JA · Nov 29, 11:46 PM · #
This is a common assumption, but it is clearly unjustified. We know that significant human evolution has taken place in the last 50,000 years – in fact, in the last 10,000. See lactose tolerance alleles, for example.
— Mencius · Nov 30, 02:50 AM · #
“The central point of my prior post on the subject of race and intelligence was that the proposition that differences in performance on IQ tests between various racial and ethnic groups can be materially explained by differences in the genetics of the groups is unproven.”
Well, how about the General Theory of Relativity? Is that “unproven,” too?
From a Popperian perspective, of course Einstein’s theory is unproven. That’s why we call it a “theory” as opposed to a “law,” like Newton’s “Law of Gravity” which it superseded.
It’s “unproven,” and Einstein’s theory may someday be superseded by a better theory, just as it superseded Newton’s theory. But, clearly, Einstein’s theory is a tremendous step forward in human understanding, just as Newton’s was before it.
Clearly, our current understanding of human intelligence is far less well established than our current understanding of gravity. But, it’s progressing rapidly, and likely to make a major leap forward in our lifetimes as the DNA data comes pouring in.
What non-scientific thinkers can do is help prepare ourselves for what’s very likely to emerge from the genome studies? At present, the conventional wisdom
But, the important thing to keep in mind is that what this discussion has been about is whether James Watson should have been fired for saying something that is totally contradicted by the science.
As this posting of yours points out, the great weight of scientific evidence is — contrary to 99% of what was written in the Main Stream Media — on Watson’s side. But, you say, there’s still a possibility that Watson’s view may turn out to be wrong. Okay, sure, no question about it — there’s still a possibility he’ll be proven wrong.
In the meantime, however, can he have his job back? Can he get an apology from all the ignoramuses and would-be Thought Policemen?
— Steve Sailer · Nov 30, 04:58 AM · #
Jim, I realize I’m probably being a little scientifically naive, but isn’t your theory testable?
If I understand you correctly, you are hypothesizing, at least for argument, that there is some cultural difference between the upbringing of the various genetic pools that have tested differently for IQ, and that this currently unknown cultural difference operates similarly to test preparation.
That seems like a fairly easily testable hypothesis, regardless of the mechanism. Twin adoption studies would be cleanest, but I would also be interested in cross-racial adoptions. If star-bellied sneeches have the highest g, spiral are in the middle, and square-bellied sneeches are the lowest, what happens to the g values for square-bellied babies adopted and raised by star-bellied and spiral-bellied parents, and vice versa?
Is it your position that that sort of cross-cultural comparisons (1) hasn’t been or (2) couldn’t be reliably performed? Thanks!
— J Mann · Dec 3, 07:53 PM · #
J:
Yes, I believe the hypothesis is theoretically falsifiable. Empirically, it has been very difficult to conduct such experiments in a free society with sufficient controls to achieve resolution. If you go to my first post on this (linked in this post), you can read the section on the Minnesota Transracial Adoption Study and click through on the links under “opposing interpretations” to see the confounding issues that are under debate in this study (which is widely considered to be the gold standard of such studies in terms of rigor).
Does this mean that it is impractical to ever really test the hypothesis at finite cost? – I don’t know. As I idnicated in the first post, my expectation is that resolution of the general issue will likely require analysis of the physical mechanism of hypothesized genetic transmission.
— Jim Manzi · Dec 3, 11:59 PM · #
Thanks, Jim, and mea culpa for not reading your prior post before commenting. That’s very helpful.
— J Mann · Dec 4, 05:17 PM · #