There has been quite a bit of discussion recently about Prometheus'
admission standards. It will come as no surprise to anyone that I have
some ideas about this. Instead of responding individually to 50 or 100
messages, I'll try to summarize what seem to me to be the key areas.
*The Ferguson Formula*
The psychometrician George Ferguson described a method for estimating
"true" I.Q. given two or more different test scores. If the tests are
imperfectly correlated, two high scores indicate a somewhat higher
"true" score. The Mega Society's constitution [Bylaws]provides that the
Ferguson formula *may* be used for admissions to Mega, but no member
has actually been admitted on this basis.
This makes some sense for Mega, as the problem of discriminating at
the 99.9999th percentile has, in my opinion, not yet been solved. (I
think that Mega actually discriminates at around the 99.9997th
percentile.) But I don't think that it's a big problem that Prometheus
is "too exclusive," so my inclination with regard to fooling around
with the Ferguson formula is that our admissions policy ain't broke.
But if we should find ourselves, at some point, without acceptable
high-range tests, a combination of the Terman Concept Mastery and
the Raven Advanced Progressive Matrices should provide us with a
reasonably good measure with a little bit of top to spare at the
four-sigma level.
*Chronometric Testing*
It's very interesting that certain physiological measurements have
turned out to correlate significantly with psychometric I.Q., but
there are three problems with this approach at the present time:
1. According to Dr. Arthur Jensen, one of the leading figures in
chronometric testing, the best scientific consensus figure for
the correlation between chronometric tests and standard I.Q. tests
is only around .4, though some studies have yielded higher figures. [2]
One interesting question is how well these tests correlate with
the tests Prometheus uses for admission, but it appears that the
state of the art in chronometric testing is not yet sufficient for
our purposes.
2. The LAIT, Mega, and Titan tests are designed specifically to
discriminate at the high end of the I.Q. scale. Calibration of
chronometric tests at very high levels is needed. Hedley [St. John-
Wilson], as someone with a strong interest in both chronometric testing
and the high end, what can you tell us about this? [3]
3. According to messages from several participants on this list,
there is a significant practice effect with ThinkFast. [4] Of course,
taking the LAIT or the Mega Test is time-consuming, too, but it's
*interesting*; without the enjoyment of engaging one's abstract
reasoning powers in challenging tasks and the small epiphanies
one experiences in finding the solutions it may be difficult to
get very-high-I.Q. people to put in the time required.
*Elo* [5]
Elo, yourself. Although it's an interesting idea to set up a rating
system analogous to that for chess, it's unnecessarily complicated.
Occam's razor should put this one to rest.
And, having groaned under the task of devising, publishing,
distributing, norming, and scoring tests for many years, I can tell
you that administering anything like what's been suggested would
take an enormous amount of somebody's time and energy.
*Publication of Answers to Problems on the Mega Test*
This is a serious problem. I think that Ron will have to retire this
test soon, and the Titan Test may also need to be retired in the not-
too-distant future. The problem for Ron is compounded by the fact
that he has used items from the Mega and Titan tests in constructing
his Hoeflin Power Test. But very few people have taken the Power Test.
Ron would do better to discard all the compromised items he knows of
(those who have material on this should send it to Ron), devise some
new problems if necessary, and publish a new version of the test.
It seems likely that the lifetime of high-range tests will be shorter
in the future as the Internet continues to grow.
New high-quality tests are needed. I have a test ready to go as soon
as legalities permit. Alan Aax' *Eight Item Test* is an interesting
experiment, but it's too short and it hasn't been normed. There are
several problems with Paul Cooijmans' tests, which I won't go into
here, but he may eventually be able to produce something usable.
*"Practicing Psychology without a Licence"*
The California Board of Psychology maintains that administering I.Q.
tests by mail for a fee constitutes the practice of psychology without
a license. A statute that was pending about a year ago in New York
state would have made it illegal, with or without a fee. (Does anyone
know the status of the proposed statute?)
In my opinion, the actions of state governments to prohibit testing by
persons who are not licensed psychologists violate the constitutional
right of high-I.Q. societies to select their own members. I intend to
mount a legal challenge to the California Board of Psychology. I would
be most grateful for whatever assistance others can offer for this
action in defense of our right to make use of the means necessary for
selecting members of the societies.
Kevin Langdon
Notes
[1] The Prometheus Society's ``fire'' listserv is a very active e-mail list with many interesting messages every day. A number of nonmembers of Prometheus participate on the list. For information, write to [Alfred Simpson, Membership Officer, 209-25 Villa Rd., Etobicoke, ON M8W 1M4, Canada or see http://www.prometheussociety.org/].
[2] Although I obtained the .4 figure directly from Dr. Jensen, there's more to it than that. The following passage is from Chapter 8 of The g Factor, ``Information Processing and g'':
It is sufficient here to note that the combined RTs [reaction times] from a number of ECTs [elementary cognitive tasks] and IQ or other highly g-loaded measure approach the correlations typically found between various psychometric power tests, ranging up to correlations of about .70. A review of several studies in which RTs (and RTSD [standard deviation of RT]) from four or five different ECTs were combined shows multiple correlations (R) ranging from .431 to .745, with an average R of .61 for RT, .60 for RTSD (i.e., intraindividual variability in RT), and .67 for RT + RTSD.
Dr. Jensen went on to point out that, with an appropriate correction for the restricted range of the college-student samples on which these statistics are based, these figures would be larger by at least .10. This is impressive, approaching the correlations of the best psychometric tests with one another.
It may well be that we will be able to make use of ``chronometric'' tests in the future, but none has yet been demonstrated to discriminate accurately in the range of interest to us.
[3] In a reply to my message on the ``fire'' list, dated July 9, Hedley St. John-Wilson wrote:
Some tests are more g loaded than others. For example, complex reaction times are more g loaded than simple reaction times. There is more information on the field of Chronometrics by Jensen and Cognitive Diagnostics at the following addresses:
[4] ThinkFast is a computer program which consists of timed drills which are claimed to provide a highly g loaded measure of ability. ThinkFast is described on the ``BrainTainment Center'' Web site (http://www.brain.com), where it is offered for sale. However, I am not convinced that the claims made for the program are justified.
[5] Arpad Elo, a Hungarian-American mathematician, devised the system of ratings which bears his name. The Elo system is used by FIDE, the world chess federation, and by many national chess organizations. Bill McGaugh, a member of Prometheus, has outlined details of a new intelligence scale based on Elo on his Web page (http://www.pe.net/~bmcgaugh/eloiq.htm).