After finding the Howard Wainer interview, I looked up the entire series of Profiles in Research published by the Journal of Educational and Behavioral Statistics. I don’t have much to say about most of these interviews: some of these people I’d never heard of, and I don’t really have much research overlap with the others. Probably I have the most overlap with R. D. Bock, who’s done a lot of work on multilevel modeling, but, for whatever reason, his stories didn’t grab my interest.
But I was curious about the interview with Arthur Jensen. I’ve never met him–he gave a talk at the Berkeley statistics department once when I was there, but for some reason I wasn’t able to attend the talk. But I’ve heard of him. As the interviewers (Daniel Robinson and Howard Wainer) state:
Dr. Jensen has authored over 435 articles, books, and book chapters and is perhaps best known for his controversial 123-page article that appeared in the Harvard Educational Review in 1969. In the article, Dr. Jensen concluded that the differences between Whites and Blacks on IQ tests were attributable to inherent intellectual differences between the two races. In 1980, his Bias in Mental Testing book concluded that intelligence tests were not biased against Blacks, resulting in even more controversy.
The interview had some interesting bits. First, something on education and individual differences:
The problems of schooling illustrate the first and second laws of individual differences. I call them laws because they are demonstrated without exception both in the psychological laboratory and in “real life.” Unfortunately, they happen to contradict the popular faith in education as the “great leveler.” The first law is that individual differences in learning and performance increase as task complexity increases. The second law is that individual differences in performance increase with continuing practice and experience, unless the particular task imposes an artificially low ceiling on proficiency.
One notable consequence of these laws is that successful attempts to raise performance by improving methods and amounts of instruction raises the overall mean of the treated group but at the same time widens the distribution of individual differences. The very same effect also applies to group differences. A benefit of raising the overall educational level of the whole population is that it moves a greater proportion of the population above the threshold levels of knowledge and skill required for gainful employment. The downside is the resulting increase in individual and group differences.
Then there are his ruminations on “g”:
The educated public today knows of Newton’s law of gravitation, Darwin’s natural selection, and Einstein’s equivalence of mass and energy. They should also know about Spearman’s g. Discovered in 1904, g is an essential concept for understanding variation in human abilities. . . . At the top of the factor hierarchy is g, the most general factor. Every cognitive ability that shows individual differences is loaded on the g factor. Tests differ in their g loadings, but their g loadings are not related to any particular knowledge or skills assessed by the various tests. So the possible indicators of g are of unlimited diversity. . . .
It is also important to understand what g is not. It is not a mixture or average of a number of diverse tests representing many different abilities. Rather, it is a distillate, representing the single factor that all different manifestations of cognition have in common. In fact, g is not really an ability at all. It does not reflect the tests’ contents per se, or any particular kind of performance. It defies description in psychological terms. [italics added] Actually, it reflects some properties of the brain that cause diverse forms of cognitive activity to be positively correlated, not only in psychometric tests but in all of life’s mental demands. IQ scores are an attempt to estimate g. But because IQ is just a vehicle for g, it inevitably reflects other broad factors as well, such as verbal, numerical, and spatial abilities, and the specific properties of the particular IQ test. Yet, g is the sine qua non of all IQ tests.
A bit over the top, no? I mean, I’m a political scientist and I think party id and ideology are important, and I even talk about our conceptual model in which each person has a position on a left-right scale and can get shifted by valence issues etc etc–but we know not to take that stuff too seriously!
The most interesting part of the interview, from a historical perspective, was Jensen’s discussion of the reaction to his papers. He tells a story in which he was giving a professional lecture in Chicago that was disrupted by 100 protesters who had infiltrated themselves into the crowd–and then he was rescued from the demonstration by a group of 10 police officers who had infiltrated themselves into the infiltrators.
I wonder what Jensen makes of the Flynn effect, in particular, the observation that most of the increases in IQ (relative to a fixed standard) occur in the lower quantiles of the distribution, with the upper quantiles showing little or no change. *Something* is acting like a great leveler…
Yes, Jensen's fixation on g would be nutty if it turns out to be only a statistical construct. But, the interesting question is whether by statistical techniques Jensen and co. have isolated a *biological* aspect of "quality of general brain function" — that is clearly what Jensen believes and he gives the evidence below (from the same interview).
For me, the reaction time results, which show a correlation between g and simple tests of recognition and reaction time, point to a possible neurological basis of what might otherwise be simply an abstract statistical quantity.
"Also, g is uniquely correated with variables outside the realm of psychometrics, particularly biological variables having behavioral correlates:
– The heritability (i.e., proportion of genetic variance) of various tests is directly related to the tests’ g loadings.
– Inbreeding depression of test scores is a purely genetic effect that lessens a quantitative trait. It results from the greater frequency of double-recessive alleles in the offspring of genetically related parents, such as cousins. The degree of inbreeding depression on various mental test scores is strongly related to the tests’ gloadings. The larger the gloading, the greater is the magnitude of inbreeding depression on the test scores.
– Anatomical and physiological brain variables are related to differences in tests’ gloadings: Brain size, brain glucose metabolic rate, the latency and amplitude of cortical evoked potentials, brain nerve conduction velocity, brain intracellular pH level, and certain biochemical neurotransmitters. Thus, g reflects biological components of intelligence more than any other psychometric factors."
Also, Jensen's general attitude towards science is admirable:
"To what extent my theoretical position is ultimately proven correct—or incorrect—will be determined by future scientific research. So whether people agree or disagree with my conclusions at any given time is much less important than my hope that they actually understand what I am saying. Criticism and further empirical research then can properly advance our knowledge."
Re: Flynn Effect, it is unclear whether the gains are actually gains in g, as opposed to some other factor. Flynn himself does NOT claim that they are.
"A bit over the top, no?"
As George Box once said, "All models are wrong, but some are useful." That goes for the g-factor too.
Are IQ tests useful? Yes they are, and the US armed forces has used them for over 60 years. The military routinely the gives Air Force Officers Qualifying test, (AFOQT) to all recruits, and make assignments based on the results. When they stop using the test, their washouts increase and costs go up. IQ scores are good predictors for many standardized tests including the SAT, MCAT, and the LSAT to name just a few. IQ scores are also good predictors of academic grades and socio-economic status.
We also know that "g" is highly heritable from studies on identical twins reared together and apart. Most likely g is related to some functional or structural aspect of the animal brain. Brain size (adjusted for body size) in humans has about a 0.4 correlation with IQ. At present g is a latent variable, but I'm sure we will one day relate it directly to some physical aspect of the brain. Let's remember that before 1900 most scientists didn't believe in atoms, and considered them a convenient fiction. Then Einstein with his paper on Brownian motion showed atoms really do exist. Only a few years ago IBM was able to image individual atoms.
So I'm not sure why you think Jensen is "over the top."
Steve, Zarkov:
I agree that this stuff is important, and it's unfortunate that this area of research has been so politicized. As a member of the editorial board of the Journal of Educational and Behavioral Statistics, I certainly have no objection to cognitive testing.
I thought Jensen was a bit over the top in that he seemed to be describing "g" in somewhat mystical terms. "g" is a statistical construct; it may very well be a helpful idea, but, no, I don't think it's real in the sense that atoms are real. This is not a criticism of Jensen's work; it's just the nature of psychology, a field in which many important concepts are, by their nature, impossible to define precisely. As I noted in my blog entry, this is true of political science as well.
<a>Cosma Shalizi's post from a couple of years ago had a lot to say about g. I am not qualified to evaluate his arguments, but I found them interesting. A taste:
… the case for g rests on a statistical technique, factor analysis, which works solely on correlations between tests. Factor analysis is handy for summarizing data, but can't tell us where the correlations came from; it always says that there is a general factor whenever there are only positive correlations. The appearance of g is a trivial reflection of that correlation structure. A clear example, known since 1916, shows that factor analysis can give the appearance of a general factor when there are actually many thousands of completely independent and equally strong causes at work. Heritability doesn't distinguish these alternatives either. Exploratory factor analysis being no good at discovering causal structure, it provides no support for the reality of g.
Andrew, I like your analogy. Rating people on some sort of liberal-conservative scale obviously goes a long way to predicting how they will stand on an issue, and is thus a useful construct even though it's not like there is a "true" value of it for each person. I may start using this analogy to intelligence instead of "athleticism," which is what I usually use.
A person's "intelligence" and "conservatism" or "athleticism" are not real in the same way that the person's mass or temperature is real; instead, each is a sort of index that agglomerates many parameters, often quite unrelated.
To some extent, people seem to understand this: someone might be described as "socially liberal but economically conservative," or as having "excellent math skills but poor verbal skills," or being "fast and strong but uncoordinated." Seemingly, anyone who uses phrases like those, or merely understands them, realizes that conservatism, intelligence, and athleticism are not underlying traits from which individual parameters are derived, but are rather weighted averages (or something) of disparate traits that we combine for convenience. And yet, some of these same people argue that "intelligence" is a real thing, whereas few people would say that about athleticism. It's a bit of a puzzle.
Someone should mention —and I guess it will be me — that Alfred Binet, inventor of the Stanford-Binet intelligence test that established the field of modern intelligence testing, was well aware that "intelligence" has a somewhat fuzzy quality to it: when asked to define 'intelligence' he reputedly replied "Easy: 'intelligence' is what my test measures."
I lived through a couple of firefights over IQ, including whether a certain professor proposed to teach a seminar was racist because of views regarding g.
I had issue with the quality of the old data – as in, testing of immigrants showed European Jews below average and the tests themselves, if you looked at them, varied widely. I also couldn't understand how the discussion of g related to racial composition of groups when the racial make-up had great variety and was perhaps a social construct. For example, how does the relatively minor data that shows some mitochondrial link in Jewish males over time relate to the heritability of g? The connection seems tenuous at best.
jonathan, from what I recall the famous old study giving low IQ scores for Jewish immigrants involved people pre-selected for mental retardation. IQ tests were originally designed to find those who needed to be separated into special ed, and so the study people generally refer to was just about showing that the test gave the results it was supposed to. As far as I know, there wasn't any study showing Jews to actually have lower IQs than other ethnicities.
Ah, the general factor in intelligence.
I'm told that Spearman wrote somewhere:
"Every normal man, woman, and child is … a genius at something … It remains to discover at what … This must be a most difficult matter, owing to the very fact that it occurs in only a minute proportion of all possible abilities. It certainly cannot be detected by any of the testing procedures at present in current usage."
It's interesting that although g "explains" a chunk of the variance in IQ tests, people do research on trying to decompose what g is. Is it something to do with white matter tract integrity? Nerve conduction speed? Bits of prefrontal function? Lots of other things?
Do you know of anywhere else in stats where people grab some shared variance and then try to work out what it means?
Phil – I think you are confusing Binet with E. G. Boring (1923). (Besides Binet would have said it with a French accent.)
My main objection to Jensen is that he reaches such strong conclusions from a naive model that surely he knew to be incorrect – in the sense that the argument assumes environment and genetics are uncorrelated. They are massively correlated and interdependent. Even tiny differences in nature can lead to massive differences because they lead people to select the environments they experience and shape the environments (including other people) that shape them. Dickens & Flynn (2001) demonstrated how this could happen. Interestingly, most people missed the real point of that paper (in my opinion) – which was a demonstration that tiny innate differences coupled with small, cumulative environmental factors can produce massive differences between people.
"For example, how does the relatively minor data that shows some mitochondrial link in Jewish males over time relate to the heritability of g? The connection seems tenuous at best."
See the paper "Natural History of Ashkenazi Intelligence" by Gregory Cochran, Jason Hardy, Henry Harpending (on the web). The propose a theory and the means to test the theory. They say:
"There are several key observations that motivate our hypothesis. The first is that the
Ashkenazi Jews have the highest average IQ of any ethnic group, combined with an un-
usual cognitive profile, while no similar elevation of intelligence was observed among
Jews in classical times nor is one seen in Sephardic and Oriental Jews today." …
"We describe two main
clusters of Ashkenazi inherited disease, the sphingolipid cluster and the DNA repair
cluster, reviewing evidence that these modulate early central nervous system
development. A sample of Gaucher disease patients show a startling occupational
spectrum of high IQ jobs, and several other Ashkenazi disorders, idiopatdystonia and non-classical adrenal hyperplasia, are known to elevate IQ."
To test their theory, they propose we measure the IQ for the Ashkenazi Jews that carry the genes that cause certain nervous system diseases. Their paper is available on the web, just Google the title and authors.
The large heritability of IQ and other facts strongly suggest that g is not some kind of social construct, but something real and physical. Let's also remember that IQ simply induces a rank order. We don't say that a person with an IQ or 130 is 30% more intelligent than someone with an IQ of 100. We simply say that he will do better a cognitively intensive tasks. That's all.
Thom: the quote is generally attributed to Binet — do an online search for "intelligence is what my test measures" in quotes — but of course this proves nothing. You may well be right that somebody else said it.
Zarkov: I think any reasonable person would agree that some people are more intelligent than others, so intelligence is not just a "social construct." But that doesn't mean "g" is a single, real, underlying factor. And its heritability doesn't provide any evidence that it is. Consider the fact that some people are more athletic than others, and athleticism is heritable. When we say "athleticism" we mean some combination of speed, proprioception, coordination, strength, physical trainability, and many other factors. Some people are really good at all of these and some people are really bad at all of these, but many people are good at some and bad at others. If we make up a bunch of tests to test these various traits, then do some sort of weighted sum of the test results, we'll have an index that may be useful for predicting who has more potential to be a baseball player or a ballet dancer, and that is heritable, but nobody (I think) would think we have measured some single underlying quantity. I think the situation with IQ is similar.
A type of problem on some IQ tests shows drawings of several shapes rotated in various ways, and asks which of them depicts a reference shape viewed from a different angle. People with high IQ tend to do better at this than people with low IQ. And pigeons can be trained to do better at this than college students. So here's a cognitive challenge at which pigeons outperform people. People are clearly more intelligent than pigeons — not just a social construct! — but that doesn't mean a univariate intelligence measure is "real."
Andy,
Don't we do this every time we estimate an error variance? I do.
'g' is an induction.
Scot
"Dickens & Flynn (2001) demonstrated how this could happen. Interestingly, most people missed the real point of that paper (in my opinion) – which was a demonstration that tiny innate differences coupled with small, cumulative environmental factors can produce massive differences between people."
Bob Williams has responded to this:
There is a concept, referred to as emergenesis by David Lykken. It basically says that people find their genetic abilities and arrange their lives in order to hone their natural abilities. To some extent, this is only common sense. The problem arises when people extrapolate that intelligence is caused by self-selecting behavior, such as study. When a person studies something, he learns the material and may learn it to the point that he can automatize some information processing tasks. This is an ability, not intelligence.
Intelligence is not the product of education. Intelligence is the ability which is reflected in psychometric g. If education (or the act of valuing education) were a cause of intelligence, we would have longitudinal studies in which those individuals who pursued much education would have ever increasing IQs and those who dropped out of school would not show IQ gains. This is not what such studies have shown. The reality is that intelligence differences found at about the age people enter school are the same as found later in life. IQ stabilizes long before most people complete their formal educations.
The notion that macro environmental conditions can affect intelligence is largely limited to injury or extraordinary circumstances. Sandra Scarr: "Within the range of 'humane environments,' variations in family socioeconomic characteristics and in child-rearing practices have little or no effect on IQ measured in adolescence." Jensen: "There is simply no good evidence that social environmental factors have a large effect on IQ, particularly in adolescence and beyond, except in cases of extreme environmental deprivation."
Note that the heritability of IQ, in fact, increases from early childhood to later maturity. A predominantly environmental theory of IQ variation would have to predict the opposite.
There is an additional problem, when one argues that macro environmental conditions influence intelligence. It is that there are a number of physical correlates to intelligence which are not likely to be altered by such human practices as study, thinking, praise, etc. We know that smart brains are physically larger; they operate with lower glucose uptake; they have higher nerve conduction velocities; they respond more quickly to such chronometric measures as inspection time; they respond with less skew to chronometric measurements; they have more complex electroencephalography traces (from average evoked potential measurements); and they display a variety of other chemical and physical properties which differ from less intelligent brains. While these attributes are statistical and are not individually deterministic, the weight of such physical evidence strongly supports the notion that what is being observed as a variation in intelligence is to a large extent physiological. Both chronometric and electroencephalography measurements can be done in such a way as to measure intelligence with a correlation to IQ tests which is as good as the correlations from one standard IQ test to another.
"A bit over the top, no?"
Not really.
g is a very difficult concept to grasp — Jensen has been studying it since the mid-1960s – but it's also very useful. I'd recommend Jensen's magnum opus from 1998, The g Factor.
As for "athleticism," that concept has changed over my lifetime to become something very analogous to g. Over the 40+ years I've been following sports, the term "athleticism" has come to mean something much more like g than it did back in the 1960s. Today, when coaches talk about "athleticism" they mean something general but specific that isn't just a weighted average of abilities. Just as, say, all theoretical physicists but only some Presidents have high g, all NFL cornerbacks but only some NFL quarterbacks have high athleticism.