Importance:
It Showed That Gifted Children Are Taller, Better-Looking, and
More Successful Than Average
The Terman-Study
results are quoted in nearly every discussion about what becomes
of gifted children.
Prior to the middle of the 19th century, youthful
prodigies were regarded "with a mixture of admiration, awe, and hopeful
expectation. His parents were envied, and the child was likely to be made the
protęgé of a prince or king."*
*
- (Genetic Studies of Genius, Vol. IV, "The Gifted Child
Grows Up", Terman and Odum, Stanford University Press, 1949, pg. 1)
Beginning about 1850, it became fashionable to perceive
highly intelligent children as misfits who "burned out",
or overtaxed their brains and died young. They were thought to be pale, asthenic,
bespectacled, and stoop-shouldered, and headed for adult stupidity or insanity.
"The myth became prevalent that many of the great geniuses were dunces in
childhood." The view developed that a rich and well-balanced maturity
depends upon the prolongation of infancy and the fullest experience with each
developmental stage. The bright child should protected from intellectual
stimulation, "and any tendency toward early cleverness should be positively
discouraged."
This must have been very comforting to the average citizen,
and to the intellectually challenged. That smart kid in class would pay for it
when she/he grew up.
The Terman Study showed that gifted children
grow up to become gifted adults, and that they tend to be
successful, well-adjusted, and long-lived.... healthier, wealthier, and wiser.
They're taller and better-looking, on average, and they're an invitation to
uncomfortable comparisons for the rest of us. It's hard not to feel competitive
with them, and at a competitive disadvantage.
Interestingly,
the Terman screening of the population revealed the fact that,
on the wings of the bell curve, out-lying ratio-IQs aren't distributed
in accordance with a Gaussian distribution. (Some of us are exploring the
hypothesis that the natural
logarithms of IQs are Gaussian distributed*.)
* - To obtain the logarithm of the ratio-IQ, divide the IQ by 100 to convert it to a decimal fraction. Then take its natural logarithm on any scientific calculator. Multiply the resulting fraction by 100 to convert it back into an IQ, and add 100 to it. The resulting number is the deviation "IQ", with a standard deviation of approximately 15 points. Note the interesting fact that there is nothing in this calculation that sets the standard deviation to be 15.
Background:
(Did They Make A Mistake By Letting Teachers Pre-Select the Candidates?) In
1921, Dr. Lewis M. Terman and Dr. Catherine Miles Cox were awarded
a grant of $34,300 (equivalent to, perhaps, $340,000 today) by
the Commonwealth Foundation to conduct a lifelong study of 1,444
gifted schoolchildren selected from a population estimated to
be at least 268,000 California schoolchildren. To this was added
another $8,000 in cash and $6,000 in services, plus additional
unbudgeted services, bringing the total to more than $50,000 (~$500,000
in today's dollars). This study is still underway among the "Termites"
who are still alive. These children were pre-selected by their teachers as having shown
signs of giftedness, and were then
further filtered by intelligence testing. Dr. Terman says (Genetic Studies of Genius - Vol.
I - Mental and physical Traits of a Thousand Gifted Children, Terman, et al, Oxford University Press,
London, 1925, pg. 20),
"The faults of subjective
ratings have been sufficiently exposed in numerous investigations
to show their great unreliability when used alone. Such ratings
are usually based too largely on the child's class work and are
almost certain to weigh too lightly the age factor. Age-grade
status is perhaps a better criterion, but its value is limited
(1) by the variable standards in different cities, or even in
different schools of the same city, and (2) by the difficulty
of equating various degrees of acceleration for children of the
different ages. The first-named objection is the most serious,
More0ver the age-grade status of a given child rests ultimately
upon a teacher's rating; that is, upon a judgment of fitness for
promotion. It is also likely to be affected by age of entering
school, regularity of attendance, adaptability to school requirements,
and various other factors having nothing to do with intelligence.
Achievement tests, although they are an objective method, were
not seriously considered as they would have been more costly than
intelligence tests and inferior to them as measures of native
ability."
(Knowing what we know
today about the low correlation between teachers' assessments
of giftedness--particularly, extreme giftedness--and children's
actual levels of giftedness, it would seem as though there would
have been room for judgment errors in the 1921 screening. On the other hand,
undoubtedly Dr. Terman did the best he could with the resources he had, and
produced a landmark study.)
Criteria
for Enrollment in the Study
The initial sifting method
consisted of a combination of teachers' ratings and the age-grade
status of the children. (In those days, apparently, the gifted
were commonly encouraged to skip grades, with whatever combination
of beneficial and harmful consequences that may have had.) Then
the nominees, ranging from 2% of the students in Los Angeles'
poorer schools, to 20% of the student population in communities
like, I suppose, Berkeley and Hollywood, and averaging 6% to 8%,
were given the National Intelligence Test as a screening tool.
Those who passed muster were then given an abbreviated version
of the 1916 Stanford-Binet IQ test. For children below 11, the
entrance criterion was an IQ of 140. For children above 11, ceiling
effects reduced the scores of the brighter students. Consequently,
the eligibility thresholds were reduced accordingly, from 139
at 11 to 11½ to 136 at 13 and to 132 at 14. Beyond 14,
the Terman Group Test (T. G. T.) was used for high school students.
Unfortunately, there don't seem to be any conversion tables in
Dr. Terman's book relating IQ's to T. G. T. scores. A raw score
of 195 is the entry level score for children 15 and older, and
presumably corresponds to an IQ of 140 and to a mental age of
21. However, the two highest scores made on this test are in the
215-219 interval, so it's difficult to assign an IQ to them.
Nominally, the study was to include the
top 1% of the scorers on the S-B test (IQ = 136 and above), but
the 1,528 schoolchildren that they enrolled in the study represent
only about 0.6% of 250,000. Furthermore, because there are more
IQs of 136+ than would be expected on the basis of a Gaussian
normal distribution, a 1921 S-B IQ of 136 would correspond to
the top 2% of the population. In reality, in the light of the
numbers of children they enrolled in each 10-point IQ range, I
suspect that they took some children in the IQ 136-150 range,
and missed a lot of others.
About 46 children in the
group were included because they demonstrated special talents.
Criticisms: (1) Too Many Children of Professionals; (2) Intervention by Dr.
Terman; (3) Labeling Anyone with an an IQ of 140+ a Genius One of the criticisms that has been leveled at
the study is that 30% of the children came from professional-technical
families rather than the 3% that characterized the average population.
Also, there was an under-representation of minorities in the study
cohort. Another criticism was that in their adult years, Dr. Terman
sometimes intervened on behalf of his "Termites", helping
them in various ways.
Another obvious criticism,
although I don't remember hearing much about it at the time, was
the fact that anyone with an IQ of 140+, occurring about to about
one in every 80 people, was designated a genius. More than 1,000,000
U. S. citizens satisfied this definition of genius in 1921. That's
a lot of promise to fulfill, and they didn't.
(It's easy to criticize with 20/20 hindsight, but at the time
this study began, the 1916 Stanford-Binet IQ test was only five years old, and
IQ testing was in its infancy in the United States.)
Many More Children
with Very-High IQs Than a Bell Curve Would Predict: The Terman Study unearthed
far
more high-IQ scores in its 1921-22 screening than would be predicted
by a Gaussian distribution(cf "Genius",
by Hans Eysenck, pg. 58). As Dr. Eysenck puts it,
"Leta Hollingworth
(1942) in her work with children
with IQs over 180 found far more than compatible with
a normal distribution, and so did Terman (1925) in his search
for children above 140 IQ."
The highest IQ score found
during the screening was 200, a score that would be expected to occur only once
among every 5,000,000,000 test-takers, and 29.6 points above
the highest expected 1-in-250,000 score of 171.4. The chance of finding such a score among 268,000
schoolchildren would thus have been 1 in 20,000, or to say it
another way, it suggests that there are 20,000 times as many
people with an IQ
of 200 as would have been expected
on the basis of a Gaussian distribution. There were 26
children who achieved scores of 180 or above (cf Ellen
Winner's "Gifted
Children: Myths and Realities"). There should be
only one such score in every 3,500,000
individuals. Consequently, the frequency
with which IQ's of 180+ occurred in the 1921-22 screening was
about 365 times the number that would have been predicted
by a Gaussian distribution. Seventy-seven children were identified
with IQ's that fell above 170 (cf "Genius",
pg. 60), whereas only 1.5 would have been predicted, an "enrichment
ratio" of about 49
to 1. There were about 12
times as many Termites with IQ's of 160 as would have
been expected, and about 3.4 times as many with IQ's of 150 as there
should have been.
Dr. Terman pointed out the deviation from a Gaussian
distribution in a paper he published in 1925.
These Deviations
from a Bell Curve Seem to Have Been a Closely Guarded Secret in
the 1940's and the 1950's
What is shocking to me about
this is that when I began reading about IQ testing in the mid-40's,
there was no mention of the fact that IQ distributions deviated
blatantly from a Gaussian normal distribution. The only hint of
it that I remember was Leta Hollingworth's mentioning the fact
that she was finding 2 or 3 children per million in New York City
with IQ's above 180 rather than the one per million that would
be expected. (Actually, a Gaussian distribution calls for 1 IQ
of 180 or above only in every 3.5 million people. The actual
frequency of occurrence turns out to be, perhaps, 45 per million.)
Nor did I hear anything about it as an undergraduate psychology
major at Ohio State in the early 50's..
Table 1, below, shows the
way the Terman numbers may be mapped into equivalent deviation
IQ's.
|
|
||||
|
IQ |
Frequency of Occurrence |
Frequency of Occurrence |
Deviation IQ |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
159.6 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Terman Hyperabundance
of the Hyperbright Found in Other Studies As Well (Leading to
a Switch to "Deviation IQs"?)
These results agree well
with other, later studies of larger populations. A landmark University
of London Master's thesis by Geoffrey Thomas Sare in 1951 generated
the following table (Table 2):
|
|
||||
|
IQ |
Frequency of Occurrence |
Frequency of Occurrence |
Deviation IQ |
Observed to Expected |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
The Termites' deviation-IQs
actually reflect slightly higher-than-expected IQs, consistent
with an enriched population. However, the Termites' score distribution
closely mirrors what the Sare study would predict.
A
Log-Normal Distribution Fits the Terman and Geoffrey Sare Data
Very Well.
Although the scores themselves
deviate from the tails of a bell curve, the natural logarithm
of the Gaussian error function is normally distributed.
John Scoville has observed that a log-normal
distribution fits Vernon Sare's data very well, as shown
in Table 3 below. The log-normal distribution equates the deviation
IQ to the logarithm of the ratio-IQ.
Deviation-IQ = ln(ratio IQ/100) = 106.667 ln(ratio IQ/100) + 100,
And conversely,
Ratio-IQ = 100 exp((deviation-IQ - 100)/106.667)
|
|
||||
|
IQ |
Frequency of Occurrence |
Frequency of Occurrence |
Deviation IQ |
Expected to Observed |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
It has been discovered that
although Gaussian
distributions fit observed IQ distributions well within one or
two standard deviations of the mean value of 100 (between IQ levels
of 68 and 132), they rapidly become inaccurate for IQ's lying
outside this range. Hans Eysenck
discusses this on page 58 in his book "Genius", observing
that Cyril Burt proposed a Pearson Type IV distribution to fit
the observed data. A Pearson Type IV curve predicts that an IQ
of 200 will occur with a frequency of 6.2 per million, and one
IQ of 160 per 1,100 test subjects. The 6.2 per million agrees
well with the Terman data, but not as well with the Vernon Sare
results (which predict 2 per million). On the other hand, one
IQ of 160 per 1,100 test subjects agrees well with the Sare table
above.
Caveat:
I also need to flag the
fact that Darryl Miyaguchi is not sure about the reality of the
Geoffrey Thomas Sare paper that several of us have quoted. Darryl
is going to try to make sure that such a Master's thesis exists.
In the meantime, I need to raise a flag concerning this "data"
pending Darryl's investigation, since Im relying upon it for confirmation
of the log-normal model that John Scoville and Dr. Robert Dick
have devised.
Of course, the Terman data
fits the log-normal model well, as shown by comparing Table I
with Table III.
Distributions
of Other Physical Parameters, Such as Height, Are Also Non Gaussian It's worth noting
that Gaussian distributions for other physical parameters such
as height also predict frequencies of occurrence fairly accurately
for values close to the norm, but become increasingly erroneous
on the wings of the curve. For example, if the average male height
is taken to be 5' 8', with a standard deviation of 4 inches, heights
greater than 7' 8" and less than 3' 8" should occur
in fewer than one in a billion cases. In practice, their frequencies
of occurrence are much higher than that.
Conclusions: "Termites" Were
Taller, Better-Educated, More Successful Than Average
The most significant result
of the Terman Study was the revelation that the Termites were,
on the whole, taller, better educated, more successful, and better
adjusted than average. For example, 70% graduated from college,
with 67% of the men and 60% of the women continuing on to graduate
school (in the 30's)..Among the men, 86% were to be found in the
professions, the semi-professions, and higher business. No gifted
men were found on the lowest occupational level, compared to 13%
of the general population.
But—No
Geniuses Were Found
At the same time, it was
also the case that no geniuses emerged from the group. The two
members of the screened population who became Nobel Laureates--Drs.
William Shockley, for the invention of the transistor, and Dr.
Luis Alvarez, for the liquid hydrogen bubble chamber--failed to
make the cut admitting them to the program. However, a couple
of points might need to be considered. In Dr. Shockley's case,
three men shared the Nobel Prize: Dr. Shockley, Dr. John Bardeen,
and Dr. Walter Brittain. Dr. Bardeen was a theoretician who was
the first physicist to receive a second Nobel Prize in physics
(for superconductivity). (As a child, could Dr. Alvarez have spoken
English as a second language, and have faced special problems
on an English-based IQ test.?) Also, the students had to first
be selected by their teachers, and later-studies revealed that
teachers have an imperfect track record in identifying the gifted.
But the fact remains that
the there were two Nobel-Prize winners in physics who were among
those screened by the Terman Study and they were rejected as having
IQs too low to be included in the program.
But—How
Many Geniuses Among 268,000 People Anyway?
In
assessing the accomplishments of the Termites, it's important
to realize that 250,000 people corresponds to the population of
a city the size of Grand Rapids, Michigan, or Ft. Lauderdale,
Florida. Another way to say it is that 250,000 represent 0.1%
of the current U. S. population. How many geniuses will Grand
Rapids produce? Ft. Lauderdale? How many would 0.1.% of the geniuses
in the United States comprise. The Terman Study produced 70 individuals
who, by or before the age of 40, had qualified for American
Men of Science (before World War II). Three had been elected
to the national Academy of Sciences. ten more appeared in Directory
of American Scholars, and 31 made it into Who's Who in
America. Thirteen of these were faculty members, eight were
top-ranking executives, and three were diplomats. Of the rest,
one was an internationally known scientist who was the provost
of a leading university, another was a distinguished literary
scholar who was the vice-chancellor at one of the country's largest
universities. A third, with a doctorate in theology, was the president
of a small denominational college. It also included a famous oceanographer
who headed a major institute of oceanography, the dean of a leading
medical school, and an internationally known physiologist who
directed an internationally known laboratory. Collectively, they
had published 2000+ scientific and technical papers, 60 books
and monographs, 230 patents, 33 novels, 375 short stories, and
265 articles. And all of these accomplishments had been registered
by or before the age of 40.
Age At Which
the Termites Learned to Read
"Terman found back
in the 1920's that the age at which reading began was one of the
few traits that distinguished the more gifted from the not-as-gifted."
In the 1921 study, among
children a or above IQ 170, only 43% began to read before they
were 5, and only 13% of these learned to read before the age of
4. The median reading age was a little above 5. 57% learned to
read when they were 5 or older..
By contrast, In Leta Hollingsworth's
study of "Children Above 180 IQ", 83% were reading before
age 5, 67% were reading before the age of 4, 58% were reading
before 3,, and 8% were reading before the age of 2. Their average
reading age was 3.67.
A similar study
by Van Tassel and Baska found that 80% were reading before 5,
and 55% were reading before 4.
In Mirica Gross'
1980's study of extremely gifted children in Australia, among
the four children who scored at the 200 level on the 1973 edition
of the Stanford Binet, Adrian was reading before 2, Christopher
was reading between 2 and 3, Ian was reading before 2, and Hadley
was reading at 18 months. Christopher and Ian scored exactly 200,
while Adrian, at 6, scored 233.
Did the Terman
Study Miss More Potentially-Eligible Children Than It Enrolled?
You'll remember that I mentioned
at the outset that the Terman Study only enrolled about 0.6% of
the 250,000 schoolchildren that it screened. At the same time,
the test claimed to have screened the top 1%. But that would have
corresponded to a ratio IQ of 137 or a deviation IQ of 133.5 or
about the top 2% of all the schoolchildren. So what happened to
the other 1.3%? As you'll see below, there is reason to believe
that the study missed more children who would later become eligible
adults than it picked up.
Follow-Up:
The Termites Adult IQs
In 1940, when the average
age of the Termites was 29, they were given a follow-up IQ test
called the Concept Mastery Test (CMT). Table 4, below, shows the
number of people who took the CMT in each IQ range, together with
their average IQ scores on the CMT. What we're seeking
here is a comparison between the "Termites" childhood
IQ scores and their adult IQ scores.
Column 1 lists the childhood ratio-IQ ranges.
Column 2 lists the estimated deviation IQ ranges
equivalent to Column 1's ratio IQ ranges.
Column 3 lists the estimated average ratio-IQ
within each range in Column 1.
Column 4 lists the estimated average deviation-IQ
within each range in Column 2.
Column 5 lists the number of people who took
the adult (CMT) IQ test in 1940 whose childhood IQs had fallen
within the ranges in Column 1.
Column 6 lists the average adult CTM scores
for each childhood IQ range.
Column 7 lists the average declines in IQ, from
child to adult, if the CMT measured ratio IQs.
Column 8 lists the average declines in IQ, from
child to adult, if the CMT measured deviation IQs.
|
|
|||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
* - Guesstimate
At first
blush, since deviation iQs weren't introduced until the early
1960's, one would suppose that the CMT could only be measuring
ratio IQs. However, Terman, et al, must have been well aware of
the over-representation of high scores among his "Termites".
It's possible that they normed the CMT on the basis of percentiles,
and assigned IQ scores based upon percentile rankings, thereby,
in effect, generating deviation IQ scores.
The data in Columns 1, 5, and 6
are derived from a wonderful discussion of the severely gifted
written by Grady Towers, entitled, "The
Outsiders". In it, Grady says,
|
"To
obtain such insight requires close observation by a gifted observer.
Fortunately, those insights are available to us in Leta S. Hollingworth's
book, Children above 180 IQ. Hollingworth not only observed her
subjects as children, she also continued to maintain some contact
with them after they had reached maturity. So although her book
is ostensibly about children, it is in fact laced throughout
by her observations on exceptionally gifted adults as well.
"The average childhood IQ score for those with childhood IQs above 170 was 177.7 for men, and 177.6 for women. That's quite close to the 180 cutoff used by Leta Hollingworth in selecting her subjects. Note that Terman's subjects who scored above 170 IQ as children averaged 155.8 on the CMT-T at age 41, a score quite close to the 155.16 made by the average Triple Nine member. Such a close match makes it reasonable to generalize Hollingworth's findings to members of both the Triple Nine Society and the Prometheus Society. |
A
Rift Within the Lute: Grady Towers vs. Hans Eysenck
There's some
confusion written into this last paragraph. Grady compares the
childhood ratio-IQ average of 177.7
for the children who scored 170
or above as children
with their adult average CMT IQ of 155.8,
treating the CMT score as a ratio-IQ. He then compares
this 155.8 IQ score with the 155.16
average IQ of Triple Nine Members. But the entry-level IQ required for admission
into the Triple Nine Society is a deviation IQ of 150, corresponding to a ratio IQ of
160. Obviously, the 155.16
average IQ of the triple Niners (corresponding to an average ratio
IQ of about 168) has to be a deviation IQ. To
further muddy the waters, Dr. Hans Eysenck says of the adult Termites,
"But it is with respect to intellectual ability
that the group shines most brilliantly; on tests like the Concept
Mastery Test specially devised for high-fliers, the majority scored
close to the top percentile; if anything, they showed a gain in
IQ over the period covered."
-----Genius, pg. 61
Dr. Eysenck is saying
that the Termites' IQs may have risen a little since childhood.
Obviously, there is a rift within the lute here that demands resolution
before continuing with this analysis.
In 1940, deviation IQs didn't
exist*. However, Dr. Terman and his associates would have been
well aware of the disparities betweeen Gaussian predictions and
the data they had amassed in their 1921 IQ assessment. I suppose
it's possible that they normed the CMT on the basis of frequency
data rather than mental ages, thereby anticipating what in the
1960's would be called deviation IQs. If this is the case, then
the Termites adult CMT scores, cited in the sixth column of Table
4 above, would have been only a little less than their equivalent
childhood deviation IQ scores. There would have been regression
to the mean, but it wouldn't have been startling. On the other
hand, if the CMT IQs were ratio IQs, then the regression to the
mean was sizable, as Grady Towers describes above.
This may help explain why the Termites didn't necessarily set the world on fire. They were smart, but not that smart. The 26 adults who tested at or above 180 as children (with a corresponding average deviation IQ of, perhaps, 163) had dropped to an average score of, at a guess, 161 on the CMT, a drop of 2 points. However, if their average adult score of 161 on the CMT is a ratio-IQ score, they will have dropped to a deviation IQ of about 151—a decline of about 12 points of deviation IQ. They would have registered, perhaps, 20 to 25 points of IQ above the average Termite. That's still a significant difference, but considering the handicaps in interfacing with a society designed (from their perspective) by mental pygmies, it may not be surprising that there was no difference in worldly success. We know today that there's a "sweet spot" with deviation IQs ranging from about 120 to about 150 (ratio IQ of 160) in which IQs are sufficiently high to give their owners an edge, and yet not so high that they're at risk of being out of joint with the world.
This result has two implications.
First, if, on average, the
Termites' IQs dropped significantly below the levels that would
be expected in a population of 250,000—e. g.,
to the levels that would be expected in a population fo 100,000—then
this would suggest that some of "late blooming" children
who were missed in the 1921 screening later surpassed, in adulthood, some of
the Termites. There should have been enough of these adults to
account for the expected number of adult gifted in a population
of 250,000. Presumably, the missing "late bloomers"
would have fallen near the bottom of the range, since the numbers
of Termites in the higher IQ ranges in childhood is approximately
what would be expected based upon Vernon Sare's data. (In fact,
if anything, there are more Termites in these ranges than would
be predicted by the Vernon Sare study, due, perhaps, to the high
fraction of children from professional families included in the
study.)
Second, with the adults
who tested as children at or above a ratio IQ of 180,
we're looking at a corresponding adult deviation IQ threshold
of about 159, and an estimated average adult deviation iQ of
161. Then we're comparing these individuals with childhood
ratio IQs above 180 with the average Termite, possessing an average
adult IQ of, perhaps, 135. That's 26 points of IQ, but it's not an enormous
difference.
* - The 1957 Form S version of the California Test of Mental Maturity (CTMM), and my early-60's copy of the Slossen Intelligence Test (SIT) are both ratio tests, and are predicated upon the mental age construct. The maximum mental age on the SIT is 27, and the corresponding ceiling on the CTMM is 32. Interestingly enough, the background material for the SIT assigns a rarity of 1 in 10,000 to an adult ratio IQ of 160, although as we've seen above, it must have been known from the time of the initial Terman screening of the 250,000 California schoolchildren in 1921 that the frequencies of occurrence of these ratio scores were far larger than is predicted by a Gaussian normal distribution.