High-IQ Societies and the Tests
They Accept for Admission Purposes

Kevin Langdon

Part 4

Table 5. Number and Mean Score on Various Tests by Society Membership,
Based on Previous Scores Reported by 3580 Recent LAIT Testees

Excerpted from "Reply to Paul Maxim on the Relative Performance of Mensa and ISPE
Members on Various Measures of Intellectual Ability," Noesis #122, August 1996

    Note: Results were not reported for N less than three. There was not enough data to include other tests or societies.

Table 6. Number and Mean Score on Polymath Systems Tests, by Society Membership

1. Cincinnatus Society, founded by Grady Ward, a breakaway group from TNS, active from 1987 to 1989; now defunct.
2. Members of Four Sigma were selected by means of the LAIT, so their LAIT scores cannot be compared directly with those of members of other societies. For tests other than the LAIT, only those who were already Four Sigma members are counted in these statistics.
3. Langdon Adult Intelligence Test (1976), Four Sigma Qualifying Test (by Kevin Langdon, Peter Shearer, William Kolb, George Koch, Ronald K. Hoeflin, and Ellen Tadin, 1985), Polymath Intellectual Ability Scale (1987), Langdon Intellectual Gradient High-range Test (1992) , Langdon Short Form Intelligence Test (1993), The Mobius Test (by Cyril Edwards and Kevin Langdon, preliminary version [by Cyril Edwards], 1978; Polymath Systems version, 1996). The tests are by Kevin Langdon, except as noted. The LIGHT and the LSFIT are revised forms of the LAIT. The number of data points for the FSQT, the LIGHT, and The Mobius Test is very small; data from these tests can be expected to be less reliable than that for the other three.
4. At that time, there were two separate societies, known as the Mega Society and the Titan Society in the late 1980's (the Titan Society was also known, at various times, as the Mega Society, the One-in-a-Million Society, the Noetic Society and the Hoeflin Research Group). The societies merged in 1991. The mean score for three Mega members was 164 and the mean for three Titan members was 157.

Discussion of the Above

The surprisingly strong showing of Mensa and Intertel members in the comparisons in both tables is probably due to self-selection; those who took the LAIT and my other tests were hardly a random sample. Their relatively high scores on the Cattell Verbal and the CTMM are, of course, due to their selection by means of these tests.

It is interesting to note that, of all the tests in Table 5, only the Mega Test and the Stanford-Binet show a strong difference between Four Sigma and the lower-cutoff societies. This would tend to indicate that the Prometheus Society has made a mistake in adding a number of standard tests to its list of scores accepted for admission. These tests do not seem to be capable of discriminating at the four-sigma level, as indicated in Table 4 (for the best-known of them), but the Stanford-Binet might be suitable for this purpose, provided that it is not taken too early (early childhood scores do not correlate well with adult scores).

There is abundant evidence that there are more high scores than would be predicted by a model of the distribution of levels of intelligence based on the normal curve on the Stanford-Binet and other childhood tests. John Scoville has proposed, in "Statistical Distribution of Childhood IQ Scores" <http://sac.uky.edu/~jcscov0/ratioiq.htm>, that childhood scores are distributed log-normally. A chart contained on that page places the four-sigma level at IQ 183; a score this high cannot be earned by a child older than 12. The ages when Stanford-Binet scores might be useful for identifying people at the four-sigma level are about 8-12.

Here's a fast conversion method, based on Scoville's chart, from childhood scores to adult IQ's, accurate to within about one point up to childhood IQ 204 = adult IQ 176 (sigma = 16), the one-in-a-million level: 115 and below, no adjustment; 116-155, adjusted IQ score = 115 + .8 * [excess of IQ over 115]; above 155, adjusted IQ = 147 + .6 * [excess over 155].

The Four Sigma members who took the Stanford-Binet listed in Table 5 (who were, of course, selected by means of Polymath Systems tests) scored an average of 167.8. Given the correction of this Stanford-Binet score by means of the formula above (to approximately 155), the correlation of .68 with the Stanford-Binet that I obtained in my most recent norming study of the LAIT, and the likelihood that this correlation would be substantially higher for a sample drawn from an unselected sample, this is very close to the amount of regression to the mean that is to be expected. Their mean score on the Mega (34.1, approximately 161) was somewhat higher than expected. The amount of regression in Table 6 is also reasonable, given that there seems to be a significant amount of self-selection within the eligibility pool of each of the societies (the average Mensan is above the 99th percentile, the average TNS member is above the 99.95th percentile, etc.).

In contrast with Table 5, Table 6 shows clear discrimination through at least the four-sigma level, for the tests with adequate-sized samples. Clearly, high-range, highly g loaded tests are the best selection instruments at very high levels.

[Part 3]  [<]

[High-IQ Societies]

[Consciousness and Mind]

[Polymath Systems Home Page]