Re: Rushton's postings

Wed, 26 Oct 1994 10:52:14 CDT

Steve Gangestad here. I'm no expert on theories of intelligence and
intelligence testing. Moreover, I've not yet read "The Bell Curve." I do,
however, have some familiarity with aspects of the literature that Phil
Rushton has mentioned as evidence for heritable racial differences in IQ:
specifically, the g-loading criterion, speed of processing effects, and
cranial capacity differences. My comments largely have to do with the
last of these. Apologies for (and warning of) the length of this post.

(Before getting into cranial differences, a few brief remarks on the first
two of these issues. First, tests that load high on g tend to best
differentiate American blacks and whites, be most heritable, and be
supposedly "culture-fair." But they're also tests that show large increases
in performance in Western countries ober the past 30 years [about .5 sd]
due to environmental factors that have yet to be identified [surely not
SES or gross quality of education; see Flynn, 1987, Psych. Bulletin] --
indeed, increases greater than those observed for the supposedly less
culture-fair, less heritable tests of verbal intelligence. Thus,
performance on these tests is clearly malleable in response to conditions
we don't yet understand. Second, although interest in g as some innate
"biological" property was revived in the 1980s due to demonstrated
relations between IQ and simple speed of information processing [including
peripheral nerve conduction speed], that literature is filled with
failures to replicate and alternative explanations [e.g., Vernon, who
originally demonstrated IQ's relation to peripheral nerve conduction
speed, has recently published himself the third failure to replicate
the finding; reported relations of IQ to latency to early components
of visual or auditory evoked potentials have similarly been inconsistent;
results from research on the Hicks paradigm, which purportedly showed
that high IQ folks process additional bits of information in a simple
discrimination task faster than others, have now been shown to merely
demonstrate that high IQ folks learn faster on the task than others --
an interesting effect perhaps, but my no means evidence that IQ relates
to some simple property of nerve conduction; other sorts of reaction
time effects similarly don't compellingly demonstrate relations with
basic "speed of processing"; see the journal Intelligence for this
literature]. That g reflects variation in some innate and simple
property of the nervous system related to processing speed is a matter
of considerable debate.)

Now, for comments on cranial capacity (CC): As Phil notes in his
writings, across-population differences in cranial capacity have been
acknowledged and demonstrated by others. Beals et al. (1984, Current
Anthropology) summarized the relevant data by means of a computer-
generated clinal map of CC, one that shows "high" points and "low"
points. What Phil has not been so clear on, at least in his posts to
the list, is that the variation is not neatly captured by race.
Within races, a great deal of variation exists. Thus, for instance,
certain Asian populations have CCs less than that of some European or
African populations. Moreover, CC is not merely a function of the recency
of the emergence of populations as Homo S. migrated out of Africa. As
Beals et al. note, New World populations all emerged in relatively recent
times, yet demonstrate the full range of CC observed worldwide. In light
of these facts, it seems odd that Phil continues to suggest that race is
the joint of nature that organizes these data: It ain't.

Like Phil, Beals et al. argue that variation in CC resulted from selection
pressures associated wioth climatic conditions: Bigger CCc are observed
in cold,dry climates. Unlike Phil, they suggested that this effect
doesn't reflect increased pressures for cognitive skills, but rather
reflects pressures related to thermoregulation. Their argument against
these differences having much to do with cognitive differences was based
on the premise that observed variations in human cranial capacity don't
relate to variations in intellectual performance. Phil claims that
this premise is wrong, that differences in CC do translate into differences
in IQ. He furthermore claims that the correlation within populations may
be about .3-.4. In light of this association, Phil suggests that the
variations in CC across populations also translate into IQ differences --
something that could be expected if variations in brain size "cause"
differences in IQ.

Of course, it need not be the case that between-population differences
in CC translate into IQ differences in the same way that within-
population differences do. The within-population correlation does
not demonstrate some necessary causal implact of brain size on IQ.
The correlation may arise from the operation of some third variable
(e.g., nutritional status, illness). (On a related note, meaningful
variations in cranial volumes [either within- or between-populations]
obviously need not be due to direct genetic effects. Jensen and
Johnson (1994, Intelligence) report a sibling intraclass correlation
of head circumference [a proxy measure of CC] of about .4, indicating
a narrow heritability of at most .8 [and less if common environmental
variance > 0]. Similarly, some variations across populations may well
be due to diffrerences in environmental factors, such as pathogen
prevalence.) We simply don't know the relevance of the within-population
correlations to an understanding of between-population variations in CC
vis a vis differential intelligence.

Nonetheless, let's take Phil's argument on its own terms and see where
it leads us. Does it suggest that there truly do exist meaningful,
heritable differences in the IQs of blacks and whites? To carry out
this exercise, let's first lay out the assumptions. First, we'll
assume that the association between CC and IQ is entirely due to genetic
variation in CC (or brain size) affecting IQ. (Thus, we'll assume
that the covariation is not at all environmentally mediated -- almost
certainly wrong but, again, I want to take a strong form of the argument
Phil offers on its own terms.) Next, we'll assume a correlation
between height- and weight- corrected CC and IQ of .3 (although at
least one study has shown a larger correlation, multiple studies have
shown smaller ones; based on a sample of 14,000 kids, Jensen and Johnson
[1994, Intelligence] estimated a value of about .3). This correlation,
together with estimates of the standard deviations of CC and IQ, allow
us to estimate the "effect" of CC on IQ -- that is, the boost in IQ
we get from a given increase in CC (i.e., the unstandardized regression
weight). Finally, we'll assume that the difference in CC across
African-Americans and Caucasian-Americans translates into differences in
IQ in precisely the same way that within-race variations do. Again, we
can reasonable question this assumption, but we're working here with
Phil's argument. The question is, how much difference between the
average IQ of blacks and whites do we expect to be accounted for by
genetical variations in CC, in light of these assumptions?

The data we can use comes from Phil's own research. Using external
cranial measurements of over 6000 US military personnel, he applied
formulas derived by Lee and Pearson (1901) to estimate CC (Rushton,
1991, Intelligence). African-Americans were estimated to have
significantly smaller CC than Caucasian-Americans. Estimates of CC
had a weighted standard deviation within groups of male and female
blacks and whites of about 93. After controlling for height, weight,
sex, and rank (enlisted vs officers), whites had estimated CCs 21 cubic
cm greater than that of blacks -- .22 within-group standard deviations.
Given an "effect" of .3 sd on IQ for every 1 sd change in CC, we estimate
(under the assumptions listed above) that the "race" difference in CC
translates into a .067 sd difference in IQ. This effect amounts to 1 IQ
point. (No, I'm not talking about 1 sd here; I'm talking about 1 point,
as in the difference between IQs of 100 and 101.)

A replication based on very recent analyses is possible. Jensen and
Johnson (1994) analyzed head circumference data on 14,000 4- and 7-year-
old children in the National Collaborative Perinatal Project. Using
the same logic applied to Phil's data, we can estimate that 7 year-old
blacks and whites differ by about .15 sd on the proxy measure of CC --
which translates into a .045 sd difference in IQ or .67 points.

Now, as Jensen and Johnson (1994) point out, the proxy measures of CC
will underestimate systematic relations of CC with other variables
(such as race) because they're fallible. Head circumference correlates
about .5 with CC and, hence, we can adjust the estimated standardized
effect of race on CC by multiuplying by 2. Phil's proxy should be a
somewhat better estimate of CC than mere head circumference (Jensen
& Johnson, 1994) and thus adjust effects in his study by a factor of
less than 2. For both studies, we estimate (under assumptions laid
out above) a difference of about 1.5 IQ points between the races as
a function of CC differences.

Clearly, 1.5 IQ points is not a very bid difference. In a population
of equal proportions of blacks and whites and that sort of difference,
"race" would account for one-quarter of one percent of the the variance
in IQ.

Now, I am not, of course, suggesting that there does in fact exist a
1.5 IQ point difference (small though it may be) due to racial
differences in heritable CC. Given the likely existence and direction
of this model's assumption departures (e.g., the fact that, however
slightly, nutritional status and infection probably affect CC and in
ways that increase racial differences), we should expect the actual
effect (if it exists at all) to be even smaller. Rather, I am
suggesting that under a set of assumptions charitable to Phil's
favored interpretation of this data, we still can expect only a 1.5
IQ point difference between the races due to cranial volume differences --
practically nothing. If meaningful racial differences in evolved
intelligence exist, they ain't gonna be explained in terms of
differences in brain size.

Bottom line: Please, Phil, unless you can come up with a model that
shows otherwise (and I mean a specific model with realistically and
empirically estimated parameters that allows derivation of precise
expected IQ differences), don't tell us that African-Americans score
substantively lower on IQ tests *because* they have smaller brains.

(One final note: Here, for sake of carrying out what I understand
Phil's argument to be, I haven't questioned the meaningfulness of
global IQ scores and g. Of course, many others have done so -- but
that's another issue.)