Race-Related IQ

STAN MULAIK pscccsm at prism.gatech.edu
Wed Dec 7 01:11:00 EST 1994

Chris Schadt <gash at u.washington.edu> writes:

>On 4 Dec 1994, STAN MULAIK wrote:

>> ......  On the other hand, H&M in The Bell Curve,
>> say that simply assuming that the mean differences are environmental in
>> origin does not pin down specifically what sorts of environmental differences
>> produce the differences in mean I.Q..  Many of the easy candidates for
>> explanation just don't hold up when examined empirically.  This does not
>> mean that the results ultimately will be that heredity plays the causal
>> role.  It means we haven't isolated for various groups why some are
>> higher on the average and others lower on the average.  But keep in mind
>> that unlike in experiments, means do not reflect something common to
>> every member of a group.  Just because Asians have on average higher
>> I.Q.'s than Caucasian Americans does not mean every Asian has some
>> component that places him/her above every American.  Interpreting mean
>> differences in natural populations is a very difficult thing to do
>> properly.
>This Paragraph represents the fundamental limitations (and dangers) of 
>hastily, without a further and greater analysis of the problem, applying 
>research such as I understand the book the Bell Curve to contain, to  
>societal issues. It doesn't take to much of a statistical background to 
>know that in the averaging of any two samples that the chance of obtaining 
>the same statistical mean approaches zero, without adding in environmental 
>differences or inherant flaws in the sample or metodology, such as the 
>conventional I.Q. test may represent, as has been eluded to in many 
>previous postings.  


I think Herrnstein and Murray created more problems for themselves than
they deserved by discussing the mean differences on IQ among various
racial groups.  Even they admitted that, granting that they would like
to regard the differences as being both environmental and hereditary,
they couldhn't specify how much of each was involved.  In fact they
sort of end the chapter saying that they wanted to keep the question
open and felt that those who try to declare it as totally environmental
are doing so without providing the needed empirical support.  It is
sort of an assumption some people *want* to make because for them to
think otherwise is the unthinkable.  For me, it is an issue that
goes nowhere because at this point we don't have the knowledge to
resolve it.  If all Herrnstein and Murray were trying to accomplish is
to get us to leave the question open, then they did so in such a way
as to get a lot of folks unnecessarily nervous.

But shifting the discussion a bit, the real interesting stuff in their
book involves regression relationships between IQ and other measures
and the interesting fact that various things seem predictable to more
than a little degree from IQ.  Also, in predicting academic performance
or some job-performance variables from I.Q., regression slopes between
different ethnic groups seem to be much the same, but the
intercepts for blacks, say, are lower than for whites, so that if
you use the white prediction equation, you overpredict the black
performance, which is bias in blacks' favor, rather than the usually
expected other way around.  (This is not new, but maybe not generally
Much of The Bell Curve is not about ethnic group differences, and I
think that is getting overlooked.  Actually the media is giving a
very distorted view of what the book is about.  NBC's attempt to
link the book to the Pioneer Foundation grossly misled people as to
what the book is about.  With the exception of Arthur Jensen, the
book relies very little in citations on some of the characters
NBC brought out into public view as being involved with the
Pioneer Foundation.  Some I've never heard of and I take a professional
interest in the area of individual differences.  Jensen does happen
to be a careful experimental psychologist and he has conducted studies
to test various interpretations of black-white differences on
I.Q. loaded tests and found many of them unsupported.  Herrnstein
and Murray cite such studies, and I think fairly.  These studies do
not prove heredity is the basis for the differences, but rather that
usual environmentalist explanations don't account for the differences.
A negative does not prove a positive.  There may be other environmental
explanations not considered.  But at least many of the cliche
explanations environmentalists offer don't work.  Anyway, The Bell
Curve does not depend on the work of some of these egregious
characters that NBC tried to link to the book.  It is based mainly
on doing regression, partial correlation analyses of data collected
not by them but by others, particularly the National Longitudinal
Survey of Labor Market Experience of Youth (NLSY)  involving 
originally 12,686 youths in a sample followed by the National
Opinion Research Council under supervision of the Center for
Human Resources Research at Ohio State University. [I don't know
if the Pioneer Foundation provided the funding for the study, but
for the present I doubt it, because it's main thrust was not to
prove anything about white racial superiority, and further, it was
joined in 1980 by the Department of Defense who used the sample
to update norms for the Armed Forces Qualification Test.]  

Let me say that I don't agree with everything Herrnstein and Murray
say in their book.  For example, as editor of Multivariate
Behavioral Research I arranged for the publication of a paper by
Louis Guttman that was highly critical of Arthur Jensen's use
of the fact that the mean differences between blacks and whites
were greatest on the first principal component of certain cognitive
ability tests.  Jensen wanted to equate I.Q. with first principal
component.  Guttman's argument was that if you compare any two
groups on variables on which they differ in their means, they
will differ the most in their first principal component scores, and
this is a mathematical and not an empirical relationship, for it
would apply, no matter what the subject matter of the variables.

I had several distinguished experts in the field of individual
differences and multivariate statistics comment on Guttman's paper.
I felt the interpretations were varied.  Herrnstein and Murray
report in a footnote that these experts decisively refuted Guttman.
I didn't reach that conclusion when I read those papers.  What
I saw were a variety of opinions, some critical of Guttman.  But
no one, in my opinion, decisively refuted his main point.

I.Q. cannot be defined by equating it to the first principal
component of any set of cognitive variables.  There has to be
some independent way of defining it that permits you to identify
it as what is heavily saturated say in a first-principal component.
And I have already indicated how that could be done using some
of Guttman's analyses of what goes on in intelligence test

Stan Mulaik
Georgia Institute of Technology, Atlanta Georgia, 30332
uucp:	  ...!{decvax,hplabs,ncar,purdue,rutgers}!gatech!prism!pscccsm
Internet: pscccsm at prism.gatech.edu

More information about the Bioforum mailing list