I'll add my own $1 here.
Robert Sternberg is a man who's done much in the area of psychometric (i.e the measuring of mental abilities). He has constructed several tests - when he found out his previous ones were biased, he made new ones, yet he doesn't seem too happy about it. In Skeptic's magazine, he was interviewd about tthe book The Bell Curve, from which Mikod has gotten most of his info.
Worth noticing is that that book was written for the media - not for sicentists. It tries to portray that scientists are in agreement of its conclusions, but the truth is different. I'll quote from the interview.
Sternberg: I think that there is definitely some heritability of intelligence in the White population. Almost every psychologist believes there is some
heritability of IQ and I agree. But the public may not understand just what that means. If you accept the use of the heritability statistic, about .5 is probably
right.
Skeptic: Can you explain wherein the general public's conception or the media's description of what is meant by heritability is wrong?
Sternberg: The commonplace understanding of heritability often doesn't realize that heritability is calculated within a range of environments, at a given time,
for a given population. So heritability is not the same in every population. In fact there is wide variation in populations over time and space. It is not a fixed
statistic. The value you obtain (for heritability) depends on the population, where it is, and when it is. But the major misunderstanding relates to the role of
the environment and to the role of teachability. With respect to teachability, even if heritability is fairly high, it does not mean that we cannot modify
intelligence.
Skeptic: They do review studies that deal with race differences.
Sternberg: Yes, but there is evidence that they do not review at all. There is nothing in the book that suggests that race differences are genetic. They even say
that. But what they do say is that is what we would infer given the data, even though probably somewhere else, they would have one sentence to the effect that
there is one study. And they don't cite a number of studies that suggest that race differences are not genetic.
Ah yes, agreement amongst sicentists. I'd call it an intellectually dishonest work.
Skeptic: Which studies don't Herrnstein and Murray cite?
Sternberg: Well, one study that they cite and distort the results of is the Scarr-Weinberg study. What Scarr-Weinberg and several people have done is look at blood groups (associated with Whiteness or Blackness), or skin color, and looked at the correlation with IQ. The typical correlation is about .15, suggesting
that you are accounting for about 1-2% of the variance. And even that less than 2% could be due to the way darker versus lighter people are treated. So when you look at the studies that have been done, they counter-indicate the conclusion that Herrnstein and Murray draw.
Ouch.
Skeptic: You say that Herrnstein and Murray build their whole argument on the (often wrong) interpretations of statistics. Can you be more specific?
Sternberg: One example is taking studies that show that within group heritabilities have nothing to do with between group heritabilities and then insinuating
that they do. Another example is the issue of causation and correlation. They know, and anyone who takes statistics knows, you can't draw any real causal
conclusions from correlational data. Lots of things correlate with lots of things, IQ being one of them. To draw causal inferences from correlational data,
which is what all their data are, is statistically incorrect. Another thing that many may not realize is that virtually all their data are based on one study, the
National Longitudinal Study of Youth (NLSY), which was not a study that was particularly representative of the United States population.
Skeptic: In what sense?
Sternberg: The mean was low. I think the mean IQ in that group was around 94 and the standard deviation was not 15 or 16. It was not a typical US
population. Another thing they do, in comparing correlations, is that they don't take into account the reliability and precision of the measures being used. For
example, almost every measure we use is a proxy for something else. If you ask yourself, "How good is 'number of years of schooling' for measuring how
much education a person has?" it's not a very good measure. Two people could each have 16 years of schooling, but if you compare somebody who was a
straight A student in a really good college with someone who was a D student in a really poor college, the number of years of education will be the same but
their educational attainments will be vastly different. So as a measure of how much schooling you have really had, years of education is extremely imprecise
and it's not going to look very good in correlational analyses. In contrast, IQ is a pretty good measure of that narrow construct, compared to the other types of measurement that we have. That will make IQ look more powerful than the other measures because the other measures are such crude proxies for the
constructs that they are trying to measure.
Yay.
