On Wed, 19 Dec 2012 13:19:36 -0800,
Jim Clark wrote:
Hi
I know the following will be viewed dubiously (by Mike P at least),
Boy, how predictable I've become. ;-)
coming from
a Canadian and one with University of Western Ontario (aka Western
University
Canada, another story) connections, but here goes anyway.
You should just quit here and tap out. ;-)
The short summary states: "No one component, or IQ, explained everything.
Furthermore, the scientists used a brain scanning technique known as
functional
magnetic resonance imaging (fMRI), to show that these differences in
cognitive
ability map onto distinct circuits in the brain."
But of course no respectable g person says that g is everything.
No, if memory serves, "they" say that there is a general intellectual
"factor"
that underlies all other intellectual "factors". I emphasize "factor" to
keep
the focus on the reification of intelligence into statistical entities,
namely factors,
which are simply solutions to a set of equations and does not provide info
about processes. In other words, factors serve as the mathematical basis
for observed correlation and may actually provide little insight into the
processes
that generated the correlations (which I think is the main point of the
"Neuron"
paper). To demonstrate how to beat a dead horse, I refer the interested
reader to the classic paper by Armstrong "Derivation of Theory by Means of
Factor Analysis or Tom Swift and His Electric Factor Analysis Machine".
A copy is available here:
http://repository.upenn.edu/cgi/viewcontent.cgi?article=1015&context=marketing_papers
Rather, the
standard model today (I thought) was a hierarchical one with general and
specific abilities. Did the researchers (not IQ researchers) indeed find
rs=0
among their three components? That would, I think, be surprising given the
extensive literature on IQ tests.
When one is trying to explain the pattern among correlations in a
correlation
matrix, one can continue to factor analyze subsequent matrices (i.e.,
correlations
among factors) in a hierarchical manner until one has achieved a rank=1,
that is,
the correlation matrices have been reduced to a single number. One might
want to call this number "g" or "general intelligence" or "pool-pah" or
whatever.
Let me explain this in the context of another but similar situation.
Unfortunately,
I'll have to use the work of another Canadian research, Bob Altemeyer to do
so --
I apologize. ;-) Altemeyer developed his Right Wing Authoritarianism (RWA)
scale on the basis of Adorno et al's research in "The Authoritarian
Personality".
While Adorno et al thought that there were nine components to authoritarian
personality, Altemeyer decided to select three of them as the basis for the
RWA
scale: authoritarian submission (AS), conventionalism (CO), and
authoritarian
aggression (AA).
Over the course of a few decades, Altemeyer refined his RWA scale to such
a degree that one can readily obtain large Cronbach's alpha for it even with
small
sample sizes; I use the RWA in one of the studies I have students perform in
experimental psych lab have consistently gotten alphas greater than .90 over
a dozen semesters. Altemeyer points out that when he factor analyzes the
RWA,
he typically get a single factor (sometime two) which helps support his
position
that authoritarianism is a unidimensional construct.
However, is it really unidimensional? Altemeyer constructed the RWA items
by basing them on aspects of the three components he took from Adorno et al.
He intentionally mixed up aspects of AS, CO, and AA in the items, so perhaps
it should not be surprising that alphas are so large or that a single factor
is obtained.
But Cronbach's alpha is not a measure of unidimensionality, rather it is
dependent
upon the pattern of correlations among items and these correlations could be
based on:
(1) A single factor "Authoritarianism"
or
(2) Three correlated factors of "Auth Submission", "Conventionalism", and
"Auth
Aggression"
If authoritarianism is dependent upon the "profile" of values on the three
factors,
one will see that there will be different types or groups of authoritarians.
Of the
three factors, authoritarian aggression, that is, committing violence in the
service
of someone in authority, is the most dangerous characteristic -- this helps
to explain
why one would actively hurt others. But what if one were low on
authoritarian
aggression but high on authoritarian submission and conventionalism? Some
religious groups, for example, the Quakers, are pacifistic but would be high
on
submission and conventionalism -- how many kill crazy Quaker mass murderers
have there been?
But if one gave Quakers the RWA, their score may be very high, perhaps
similar
to that of neo-nazis or other hate groups. The solution might be to
reconstruct
the RWA into three subscales, one for each of the factors, and then
interpret the
profile that a person produces. Friedrich Funke in a 2005 article in the
journal
"Political Psychology" developed a three factor authoritarianism scale and
finds
that a structural equation model with three correlated latent variables (one
for each
factor) does a good job describing the responses (it's a little more
complicated
but the main point is correct; NOTE: the items were in German and he used
German
participants).
So, we're left with the following dilemma
(1) Should one describe the degree of a person's authoritarianism in terms
of a
single score which assume a unidimensional scale
or
(2) Should one describe the degree of a person's authoritarianism in terms
of three
scores reflecting the degree of influence of each latent variable
To return to the original point of this post, which is more correct:
(1) Should one describe the degree of a person's intelligence in terms of a
single score which assume a unidimensional scale
or
(2) Should one describe the degree of a person's intelligence in terms of a
profile
of scores reflecting the degree of influence of each latent variable
representing
different cognitive abilities?
It seems to me that a single score descriptions is likely to be misleading
and
theoretically unproductive.
Perhaps we will hear from Phil Rushton and Tony Vernon, also at Western,
and (I
believe in both cases) advocates for some contribution from general
intelligence.
I don't know. Rushton might too busy measuring penises and correlating
penis length
with intelligence. Vernon might be too busy measuring head size and
correlating it
with intelligence. The things and body parts these "researchers" focus on.
-Mike Palij
New York University
[email protected]
"Mike Palij" <[email protected]> 19-Dec-12 2:31 PM >>>
A research article in the journal "Neuron" argues that it is incorrect to
think of intelligence as being a represented by a single underlying factor,
usually referred to as "g". Instead, it is asserted that "g" is an artifact
of
interacting distinct brain areas engaged in different types of processing.
A press release describing the article is accessible here:
http://www.uwo.ca/its/brain/iqmyth/
A version of the published article is available on this webpage (lower
right side; look for "Fractionating human intelligence (.pdf)" or click
here:
http://www.uwo.ca/its/brain/iqmyth/Hampshire%20Owen%20IQ%20Neuron.pdf
Though I admit to being partial to this kind of thinking I'd like to point
out two caveats:
(1) Looking at the neuroimaging component and relating it psychometric
performance worries me because such analysis is quite complex and
really requires replication,
and
(2) the researchers are Canadians, eh, which should worry everyone. ;-)
Just trying to get back to normal.
---
You are currently subscribed to tips as: [email protected].
To unsubscribe click here:
http://fsulist.frostburg.edu/u?id=13090.68da6e6e5325aa33287ff385b70df5d5&n=T&l=tips&o=22450
or send a blank email to
leave-22450-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu
