TEST FOR GENIUS
2/27/98 Percentile rankings for 26 correct and above were missing a 9. Thanks to David Omtzigt for spotting it.
9/20/98 Added a footnote: all problems have now been solved and a new norming study will be undertaken in 2000.

                      PAUL  COOIJMANS

                          PB   44
                     5737 ZG  LIESHOUT
                         NEDERLAND

                      NORMING  REPORT
               REVISION 1998 - FOURTH NORMING

     The Test For Genius now consists of 43 items, 3 of which are
already completed.  Compared to REVISION 1996, two items were dis-
carded.  As of October 1997, the obtained scores on the 40 problems of
REVISION 1998 were ("G" is number correct, "f" is frequency of that
score):

 G:  f:         2.5 1          7   2          12   1         21   1
                3   6          8   2          13.5 1         22   1
 0   3          4   6          9   1          16   1         25   2
 1   6          5   1          10  2          17   2         25.5 1
 2   7          6   1          11  2          18   1         27.5 1

     Testees reported over 150 prior scores on other tests.  For
norming purposes, only scores from tests showing a moderate to very
high correlation with raw scores on the TFG were used.  Not used were
scores from tests that showed a less than moderate correlation, or for
which too few scores were reported to make assessment of correlation
possible.  The prior scores used came from:

Test:                                   # scores: cor. with TFG:

The Final Test                          22        .96
Test for Genius (long version)          33        .91
Mega/Titan/Ultra Test (Hoeflin)         15        .83
Raven's Advanced Progressive Matrices   10        .64

     The first two are tests created by Paul Cooijmans.  Info on them
will be published later or supplied on request.  The prior scores were
converted to an IQ scale with 16 points per standard deviation if
needed, and thus formed 80 "raw score/prior IQ" pairs.  Raw scores and
IQ's were both put in columns in numerical order, the columns facing
each other one-to-one, and for each raw score the median of the IQ's
facing it was used to norm that raw score.  This method was applied to
norm raw scores 1 through 27.5; a score of 0 was normed as "lower
than...", and scores above 27.5 were normed by extrapolation.
Missing raw scores were normed by interpolation.  This gave the
following norm table:

G:    IQ:  %ile:         11  154  99.96         26  178  99.99994
                         12  155  99.97         27  181  99.99998
                         13  156  99.98         28  185  99.999995
                         14  158  99.986        29  187  99.9999976
0    <130                15  160  99.992        30  189  99.9999986
1     130  96.9          16  162  99.994        31  191  99.9999992
2     132  97.7          17  165  99.998        32  193  99.9999996
3     138  99.1          18  167  99.9987       33  195  99.9999998
4     139  99.2          19  169  99.9994       34  196  99.9999999
5     140  99.3          20  170  99.9995       35  198  99.99999994
6     141  99.4          21  171  99.9996       36  200  99.99999997
7     144  99.7          22  172  99.9997       37  202  99.99999999
8     146  99.79         23  173  99.99976      38  204  99.999999994
9     148  99.87         24  175  99.99985      39  206  99.999999998
10    152  99.94         25  176  99.9999       40 >207  99.999999999

     Percentiles were computed according to the "normal distribution'.
Compared to the third norming, the only significant change is that
accuracy has improved for scores below IQ 140; this is a result of the
removal of two items.

     This norming may be considered preliminary; it is expected that
by the year 2000 enough data will be available for a fifth norming,
based on over 100 answer sheets.  With regard to the prior scores
used, one should realize that this test was conceived and put into use
long before its creator got in contact with the American high-IQ
community.  Many testees came from the Netherlands and other countries
where IQ testing is not as big as it is in the USA, and therefore were
not able to report usable previous scores.  Norming this test is
somewhat more complicated than norming a test that is basically spread
in the USA, where one has the luxury that most testees have prior
scores on college and graduate school entrance test, that are widely
used there.  The author hopes that his current activities in various
IQ societies will generate more test submissions and usable previous
scores during the years to come.

   1 Two items in the TFG are yet unsolved.  More information on test
items is not supplied in relation to this norming.  Item analysis will
be performed when sufficient numbers of correct answers have come in
for the harder items; this may take a few years.  Note that it has
never before been attempted to make and study the statistical behav-
iour of problems of this hardness and quality.

     The Test For Genius is accepted for admission to the Top One
Percent Society (99th %ile), the One-in-A-THousand Society (99.9th
%ile). The Glia Society (IQ 150) and the Giga Society (IQ 196).  It is
recommended as an entrance test for the Prometheus (IQ 164) and Mega
(IQ 176) Societies.  It is one of the few instruments capable of
measuring intelligence in the 140-176 IQ range and unnervingly ac-
curate and reliable in doing so, hardly language-biased and the only
test potentially able to discriminate above the "mega" level.


1 All items have been solved now, and a new norming study will be
undertaken in 2000.  -- Paul Cooijmans in a Sept 2, 1998 letter.



                Copyright: Paul Cooijmans 1997, 1998
      Permission granted to Darryl Miyaguchi to publish online

Return to the Uncommonly Difficult I.Q. Tests page.