Dec 22

Racism and group differences

At the end of my essay What good
is IQ?
, I suggested that taking IQ seriously might (among other
things) be an important step towards banishing racism. The behavioral
differences between two people who are far apart on the IQ scale are
far more significant than any we can associate with racial origin.
Stupidity isn’t a handicap only when solving logic problems; people
with low IQs tend to have poor impulse control because they’re not
good at thinking about the long-term consequences of their actions.

Somebody left a comment that, if what I was reporting about group
differences in average IQ is correct, the resulting behavior would be
indistinguishable from racism. In particular, American blacks (with
an average IQ of 85) would find themselves getting the shitty end of
the stick again, this time with allegedly scientific justification.

This is an ethically troubling point. It’s the main reason most
people who know the relevant statistical facts about IQ distribution
are either in elaborate denial or refusing to talk about what they know.
But is this concern really merited, or is it a form of tendermindedness
that does more harm than good?

Let’s start with a strict and careful definition: A racist is a
person who makes unjustified assumptions about the behavior or
character of individuals based on beliefs about group racial

I think racism, in this sense, is an unequivocally bad thing. I
think most decent human beings would agree with me. But if we’re
going to define racism as a bad thing, then it has to be a behavior
based on unjustified assumptions, because otherwise there
could be times when the fear of an accusation of racism could prevent
people from seeking or speaking the truth.

There are looser definitions abroad. Some people think it is
racist merely to believe there are significant differences
between racial groups. But that is an abuse of the term, because it
means that believing the objective truth, without any intent to use it
to prejudge individuals, can make you a racist.

It is, for example, a fact that black athletes tend to perform
better in hot weather, white ones in cool weather, and oriental asians
in cold weather. There is nothing mysterious about this; it has to do
with surface-area-to-volume ratios in the population’s typical
build. Tall, long-limbed people shed heat more rapidly than stocky and
short-limbed people. That’s an advantage in Africa, less of one in the
Caucasian homelands of Europe and Central Asia, and a disadvantage in
the north Asian homeland of oriental asians.

And that’s right, white men can’t jump; limb length matters there,
too. But whites can swim better than blacks, on average,
because their bones are less dense. I don’t have hard facts on
how asians fit that picture, but if you are making the same guess I am
(at the other extreme from blacks, that is better swimmers and worse
jumpers than white people) I would bet money we’re both correct. That
would be consistent with the pattern of many other observed racial

Sportswriter and ethicist Jon Entine has investigated the
statistics of racial differences in sports extensively. Blacks,
especially blacks of West African ancestry, dominate track-and-field
athletics thanks apparently to their more efficient lung structure and
abundance of fast-twitch muscle fiber. Whites, with proportionally
shorter legs and more powerful upper bodies, still rule in wrestling
and weightlifting. The bell curves overlap, but the means — and
the best performances at the high end of the curve — differ.

Even within these groups, there are racially-correlated
subdivisions. Within the runners, your top sprinters are likelier to
be black than your top long-distance runners. Blacks have more of an
advantage in burst exertion than they do in endurance. I don’t have
hard recent data on this as I do for the other factual claims I’m
making here, but it is my impression that whites cling to a thin lead
in sports that are long-haul endurance trials — marathons,
bicycle racing, triathlons, and the like.

It is not ‘racism’ to notice these things. Or, to put
it more precisely, if we define ‘racism’ to include
noticing these things, we broaden the word until we cannot justifiably
condemn ‘racism’ any more, because too much
‘racism’ is simply recognition of empirically verifiable
truths. It’s all there in the numbers.

Knowing about these racial-average differences in athletic
performance would not justify anyone in keeping a tall, long-limbed
white individual off the track team, or a stocky black person with
excellent upper-body strength off the wrestling team. But they do
make nonsense of the notion that every team should have a racial
composition mirroring the general population. If you care about
performance, your track team is going to be mostly black and your
wrestling team mostly white.

In fact, trying to achieve ‘equal‘ distribution is a
recipe for making disgruntled underperforming white runners and
basketball players, and digruntled underperforming black wrestlers and
swimmers. It’s no service to either group, you get neither efficiency
nor happiness out of that attempt.

Most people can follow the argument this far, but are frightened of
what happens when we apply the same kind of dispassionate analysis to
racial differences in various mental abilities. But the exact same
logic applies. Observing that blacks have an average IQ a standard
deviation below the average for whites is not in itself racist.
Jumping from that observation of group differences to denying an
individual black person a job because you think it means all black
people are stupid would be racist.

Let’s pick neurosurgery as an example. Here is a profession where
IQ matters in an obvious and powerful way. If you’re screening people
for a job as a neurosurgeon, it would nevertheless be wrong to use the
standard-deviation difference in average IQ as a reason to exclude an
individual black candidate, or black candidates as a class. This
would not be justified by the facts; it would be stupid and
immoral. Excluding the black neurosurgeon-candidate who is
sufficiently bright would be a disservice to a society that needs all
the brains and talent it can get in jobs like that, regardless of skin

On the other hand, anyone who expects the racial composition of the
entire population of neurosurgeons to be ‘balanced’ in
terms of the population at large is living in a delusion. The most
efficient and fair outcome would be for that population to be balanced
in terms of the distribution of IQ — at each level of IQ the
racial mix mirrors the frequency of that IQ
within different groups. Since that minimum IQ for
competency in neurosurgery is closer to the population means for
whites and asians than the mean for blacks, we can expect the
fair-outcome population of neurosurgeons to be predominantly white and

If you try to social-engineer a different outcome, you’ll simply
create a cohort of black neurosurgeons who aren’t really bright enough
for their jobs. This, too, would be a disservice to society (not to
mention the individual patients they might harm, and the competent
black neurosurgeons that would be discredited by association). It’s
an error far more serious than trying to social-engineer too many
black wrestlers or swimmers into existence. And yet, in pursuit of a
so-called equality, we make this sort of error over and over again,
injuring all involved and creating resentments for racists to feed

Nov 17

What good is IQ?

A reader asks:

To clarify, while I believe natural selection explains a lot I have
caveats about IQ as a tool for testing intelligence. If you can’t
measure the coast of France with a single number how can you do it
with human intelligence?

Easily. Human intelligence is a great deal less complex than the
coast of France. :-)

It’s fashionable nowadays to believe that intelligence is some
complicated multifactor thing that can’t be captured in one number.
However, one of the best-established facts in psychometry (the science
of measuring mind) is that it is quite difficult to write a test of
mental ability that is not at least 50% correlated with all other such
tests. Or, to put it another way, no matter how you design ten tests for
mental ability, at least about half the variance in the scores for any one
of them statistically appears to be due to a “general intelligence”
that shows up on the other nine tests as well.

Psychometricians call this general intelligence measure “g”. It
turns out to predict important real-world success measures quite well
— not just performance in school but income and job success as
well. The fundamental weakness in multiple-factor theories of intelligence
is that measures of intelligence other than g appear to predict
very little about real-world outcomes. So you can call a lot of other
things “intelligence” if you want to make people feel warm and fuzzy,
but doing so simply isn’t very useful in the real world.

Some multifactor theorists, for example, like to describe accurate
proprioception (an acute sense of body position and balance) as a kind
of intelligence. Let’s say we call this “p”. The trouble with this
is that there are very few situations in which a combination of high p
and low g is actually useful — people need to be able to balance
checkbooks more often than they need to walk high wires. Furthermore,
g is easier to substitute for p than the other way around; a person
with high g but low p can think up a way to not have to walk a high
wire far better than a person with low g but high p can think up a way
not to have to balance a checkbook. So g is in a strict functional
sense more powerful than p. Similar arguments apply to most of the
other kinds of specialized non-g ‘intelligence’ that have been

Once you know about g, you can rank mental-capability tests by
how well their score correlates with g. IQ is valuable because a
well-composed IQ test measures g quite effectively. For purposes
of non-technical discussion, g and IQ can be considered the same, and
pychometricians now accept that an IQ test which does not closely track
g is defective.

A lot of ink has been spent by people who aren’t psychometricians
on insisting that g is a meaningless statistical artifact. The most
famous polemic on this topic was Stephen Jay Gould’s 1981 book
The Mismeasure of Man, a book which was muddled,
, and in some respects rather dishonest. Gould was a
believing Marxist; his detestation of g was part of what he perceived
as a vitally important left-versus right kulturkampf. It is
very unfortunate that he was such a persuasive writer.

Unfortunately for Gould, g is no statistical phantom. Recently g
and IQ have been shown to correlate with measurable physiological
variables such as the level of trace zinc in your hair and performance
on various sorts of reaction-time tests. There are hints in the
recent literature that g may be largely a measure of the default level
of a particular neurotransmitter associated with states of mental
alertness and speed of thought; it appears that calling people of
subnormal intelligence “slow” may not be just a metaphor!

IQ is one of several large science-related issues on which
political bias in the dominant media culture has lead it to present as
fact a distorted or even reversed version of the actual science. In
1994, after Murray and Herrnstein’s The Bell Curve got a
thoroughly undeserved trashing, fifty leading psychometricians and
psychologists co-signed a summary of mainstream
science on intelligence
. It makes eye-opening reading.

The reasons many popular and journalistic accounts continue to
insist that IQ testing is at best meaningless and at worst a sinister
plot are twofold. First, this belief flatters half of the population.
“My IQ may be below average, but that doesn’t matter because IQ is
meaningless and I have high emotional intelligence!” is,
understandably, a favorite evasion maneuver among dimwits. But that
isn’t the worst of it. The real dynamite is not in
individual differences but rather that the distribution of IQ (and
hence of g) varies considerably across groups in ways that are
politically explosive.

Men vs. women is the least of it. With other variables controlled,
men and women in a population have the same mean IQ, but the
dispersion differs. The female bell curve is slightly narrower, so
women have fewer idiots and fewer geniuses among them. Where this
gets touchy is that it may do a better job than cultural sexism of
explaining why most of the highest achievers in most fields are male
rather than female. Equal opportunity does not guarantee equal
results, and lot of feminist theory goes out the window.

But male/female differences are insignificant compared to the real
hot potato: differences in the mean IQ of racial and ethnic groups.
These differences are real and they are large enough to have severe
impact in the real world. In previous blog entries I’ve mentioned the
one-standard-deviation advantage of Ashkenazic Jews over gentile
whites; that’s roughly fifteen points of IQ. Pacific-rim Asians
(Chinese, Japanese, Koreans etc.) are also brighter on average by a
comparable margin. So, oddly enough, are ethnic Scots — though
not their close kin the Irish. Go figure…

And the part that, if you are a decent human being and not a racist
bigot, you have been dreading: American blacks average a standard
deviation lower in IQ than American whites at about 85. And
it gets worse: the average IQ of African blacks is lower
still, not far above what is considered the threshold of mental
retardation in the U.S. And yes, it’s genetic; g seems to be about
85% heritable, and recent studies of effects like regression towards
the mean suggest strongly that most of the heritability is DNA rather
than nurturance effects.

For anyone who believe that racial equality is an important goal,
this is absolutely horrible news. Which is why a lot of
well-intentioned people refuse to look at these facts, and will
attempt to shout down anyone who speaks them in public. There have
been several occasions on which leading psychometricians have had
their books canceled or withdrawn by publishers who found the actual
scientific evidence about IQ so appalling that they refused to print

Unfortunately, denial of the facts doesn’t make them go away. Far from
being meaningless, IQ may be the single most important statistic about
human beings, in the precise sense that differences in g probably drive
individual and social outcomes more than any other single measurable
attribute of human beings.

Mean IQ differences do not justify making assumptions about any individual.
There are African black geniuses and Ashkenazic Jewish morons; humanity and
ethics demand that we meet each individual human being as an individual,
without prejudice. At the same time, group differences have a significance
too great to ignore. In the U.S., blacks are 12% of the population but
commit 50% of violent crimes; can anyone honestly think this is
unconnected to the fact that they average 15 points of IQ lower than the
general population? That stupid people are more violent is a fact
independent of skin color.

And that is actually a valuable hint about how to get beyond
racism. A black man with an IQ of 85 and a white man with an IQ of 85
are about equally likely to have the character traits of poor impulse
control and violent behavior associated with criminality — and
both are far more likely to have them than a white or black man with
an IQ of 110. If we could stop being afraid of IQ and face up to it,
that would give us an objective standard that would banish racism per
se. IQ matters so much more than skin color that if we started paying
serious attention to the former, we might be able to stop paying
attention to the latter.

UPDATE: An excellent summary of science relating to g
is here

Nov 14

Selecting for intelligence

Mike Smith relays an interesting possible explanation for the observed
statistical fact that American and European Jews have a mean IQ a
standard deviation higher than Caucasian gentiles:

During the period from ancient times to modern times, there was a
constant phenomenon of Jews converting to Christianity (there were
many social pressures to do so). In a nutshell, the idea is that the
lower-IQ Jews were statistically more likely to convert, as it freed
them from having to learn to read Torah. During the Middle Ages, it
was not worth the effort for most people to become literate; the
payback was not worth it. Books were rare and expensive, and learning
to read was no guarantee of getting ahead in life. Of course, people
like to do what they’re especially good at, and the higher-IQ’s among
the Jews did not find learning to read to be such a burden. As such,
they were statistically less likely to convert (and statistically more
likely to become fathers of many children in a culture that valued
intelligence.) It is worth noting that in ancient times, Jews were not
stereotyped as especially intelligent; that stereotype arose in the
Middle Ages.

This is a special case of one of my favorite Damned Ideas, originally
developed by John W. Campbell in the 1960s from some speculations
by a forgotten French anthropologist. Campbell proposed that the
manhood initiation rituals found in many primitive tribes are a
selective machine designed to permit adulthood and reproduction only
to those who can demonstrate verbal fluency and the ability to override
instinctive fears on verbal command.

Campbell suggests that all living humans are descended from groups
of hominids that, having evolved full-human mental capability in some
of their members, found the overhead of supporting the dullards too
high. So they began selecting for traits correlated with intelligence
through initiation rituals timed for just as their offspring were
achieving reproductive capacity; losers got driven out, or possibly
killed and eaten.

Campbell pointed out that the common elements of tribal initiations
are (a) scarring or cicatricing of the skin, opening the way for
lethal infections, (b) alteration or mutilation of the genitals,
threatening the ability to reproduce, and (b) alteration of the mouth
and teeth, threatening the ability to eat. These seem particularly
well optimized for inducing maximum instinctive fear in the subject
while actually being relatively safe under controlled and relatively
hygenic conditions. The core test of initiation is this: can the
subject conquer fear and submit to the initiation on the basis
of learned (verbal, in preliterate societies) command?

Campbell noticed the first order effect was to shift the mean of
the IQ bell curve upwards over generations. The second-order effect,
which if he noticed he didn’t talk about, was to start an arms race in
initiation rituals; competing bands experimented with different
selective filters (not consciously but through random variation).
Setting the bar too low or too high would create a bad tradeoff
between IQ selectivity and maintaining raw reproductive capacity. So
we’re descended from the hominids who found the right tradeoff to push
their mean IQ up as rapidly as possible and outcompeted the groups
that chose less well.

It doesn’t seem to have occurred to Campbell or his sources, but
this theory explains why initiation rituals for girls are a rare and
usually post-literate phenomenon. Male reproductive capacity is
cheap; a healthy young man can impregnate several young women a day,
and healthy young men are instinct-wired to do exactly that whenever
they can get away with it. Female reproductive capacity, on
the other hand, is scarce and precious. So it makes sense to select
the boys ruthlessly and give the girls a pass. Of course if you push
this too far you don’t get enough hunters and fighters, but the right
tradeoff pretty clearly is not 1-to-1.

(This would also explain why humans are designed for mild polygyny,
1 to 3 sexual partners per male. You can spot this by looking at
where human beings are on various physical characteristics that
correlate with degree of polygyny in other primates — disparity in
average size between males and females, for example, is strongly
correlated with it.)

What Campbell did notice is that this theory of selection
by initiation would neatly explain one of the mysteries of human
paleoanthropology — how human beings got so smart so fast. The
differences between H. Erectus and H. Sapiens are not large in
absolute genetic terms (they can’t be, we share over 94% of our genome
with chimps) but they’re hard to credit given normal rates of
morphological change in mammals and only two million years to work
in. Something must have been putting hominids under
abnormally strong selective pressure — and Campbell’s idea
is that we did it to ourselves!

Now, I’m not sure I believe Jews bootstrapping themselves up a
whole standard deviation in less than 2000 years, but if you apply
a similar idea to a longer timeframe it begins to look pretty
reasonable. (And Campbell did suggest that the Jewish practice of
infant circumcision had originally been a manhood rite.)

Within my lifetime, I expect we’re going to have the ability to do
germ-line enhancement of human intelligence. I strongly suspect that that
will set off another arms race — because cultures that suppress
that technology will be once again doomed against cultures that do. And
this time, we’re smart enough to know that in advance…