The first 5 pages, shown as text so search engines and readers can find this document.
Page 1
Testing for Competence Rather
Than for "Intelligence"
DAVID C. McCLELLAND Harvard University1
The testing movement in the United States has
been a success, if one judges success by the usual
American criteria of size, influence, and profit-
ability. Intelligence and aptitude tests are used
nearly everywhere by schools, colleges, and em-
ployers. It is a sign of backwardness not to have
test scores in the school records of children. The
Educational Testing Service alone employs about
2,000 people, annually administers Scholastic Apti-
tude Tests to thousands of aspirants to college, and
makes enough money to support a large basic re-
search operation. Its tests have tremendous power
over the lives of young people by stamping some of
them "qualified" and others "less qualified" for col-
lege work. Until recent "exceptions" were made
(over the protest of some), the tests have served as
a very efficient device for screening out black,
Spanish-speaking, and other minority applicants to
colleges. Admissions officers have protested that
they take other qualities besides test achievements
into account in granting admission, but careful
studies by Wing and Wallach (1971) and others
have shown that this is true only to a very limited
degree.
Why should intelligence or aptitude tests have
all this power? What justifies the use of such tests
in selecting applicants for college entrance or jobs?
On what assumptions is the success of the move-
ment based? They deserve careful examination be-
fore we go on rather blindly promoting the use of
tests as instruments of power over the lives of many
Americans.
article contains the substance of remarks made at
a public lecture given at the Educational Testing Service,
Princeton, New Jersey, January 4, 1971.
Requests for reprints should be sent to David C. Mc-
Clelland, Department of Psychology and Social Relations,
Harvard University, William James Hall, Cambridge,
Massachusetts 02138.
The key issue is obviously the validity of so-
called intelligence tests. Their use could not be
justified unless they were valid, and it is my
conviction that the evidence for their validity is by
no means so overwhelming as most of us, rather un-
thinkingly, had come to think it was. In point of
fact, most of us just believed the results that the
testers gave us, without subjecting them to the
kind of fierce skepticism that greets, for example,
the latest attempt to show that ESP exists. My
objectives are to review skeptically the main lines of
evidence for the validity of intelligence and aptitude
tests and to draw some inferences from this review
as to new lines that testing might take in the future.
Let us grant at the outset that brain-damaged
or retarded people do less well on intelligence tests
than other people. Wechsler (19S8) initially used
this criterion to validate his instrument, although it
has an obvious weakness: brain-damaged people do
less well on almost any test so that it is hard to
argue that something unique called "lack of in-
telligence" is responsible for the deficiency in test
scores. The multimethod, multitrait criterion has
not been applied here.
Tests Predict Grades in School
The games people are required to play on apti-
tude tests are similar to the games teachers require
in the classroom. In fact, many of Binet's original
tests were taken from exercises that teachers used
in French schools. So it is scarcely surprising that
aptitude test scores are correlated highly with
grades in school. The whole Scholastic Aptitude
Testing movement rests its case largely on this
single undeniable fact. Defenders of intelligence
testing, like McNemar (1964), often seem to be
suggesting that this is the only kind of validity
necessary. McNemar remarked that "the manual
AMERICAN PSYCHOLOGIST January 1973 1
Page 2
of the Differential Aptitude Test of the Psychologi-
cal Corporation contains a staggering total of 4,096,
yes I counted 'em, validity coefficients." What
more could you ask for, ladies and gentlemen? It
was not until I looked at the manual myself (Mc-
Nemar certainly did not enlighten me) that I con-
firmed my suspicion that almost every one of those
"validity" coefficients involved predicting grades in
courses—in other words, performing on similar types
of tests.
So what about grades? How valid are they as
predictors? Researchers have in fact had great
difficulty demonstrating that grades in school are
related to any other behaviors of importance—
other than doing well on aptitude tests. Yet the
general public—including many psychologists and
most college officials—simply has been unable to
believe or accept this fact. It seems so self-evident
to educators that those who do well in their classes
must go on to do better in life that they systemati-
cally have disregarded evidence to the contrary that
has been accumulating for some time. In the
early 1950s, a committee of the Social Science Re-
search Council of which I was chairman looked into
the matter and concluded that while grade level at-
tained seemed related to future measures of success
in life, performance within grade was related only
slightly. In other words, being a high school or
college graduate gave one a credential that opened
up certain higher level jobs, but the poorer students
in high school or college did as well in life as the
top students. As a college teacher, I found this
hard to believe until I made a simple check. I took
the top eight students in a class in the late 1940s
at Wesleyan University where I was teaching—all
straight A students—and contrasted what they
were doing in the early 1960s with what eight
really poor students were doing—all of whom were
getting barely passing averages in college (C— or
below). To my great surprise, I could not dis-
tinguish the two lists of men 15-18 years later.
There were lawyers, doctors, research scientists, and
college and high school teachers in both groups.
The only difference I noted was that those with
better grades got into better law or medical schools,
but even with this supposed advantage they did not
have notably more successful careers as compared
with the poorer students who had had to be satisfied
with "second-rate" law and medical schools at the
outset. Doubtless the C— students could not get
into even second-rate law and medical schools under
the stricter admissions testing standards of today.
Is that an advantage for society?
Such outcomes have been documented carefully
by many researchers (cf. Hoyt, 1965) both in
Britain (Hudson, 1960) and in the United States.
Berg (1970), in a book suggestively titled Educa-
tion and Jobs: The Great Training Robbery, has
summarized studies showing that neither amount
of education nor grades in school are related to voca-
tional success as a factory worker, bank teller, or
air traffic controller. Even for highly intellectual
jobs like scientific researcher, Taylor, Smith, and
Ghiselin (1963) have shown that superior on-the-
job performance is related in no way to better
grades in college. The average college grade for
the top third in research success was 2.73 (about
B — ) , and for the bottom third, 2.69 (also B-).
Such facts have been known for some time. They
make it abundantly clear that the testing movement
is in grave danger of perpetuating a mythological
meritocracy in which none of the measures of merit
bears significant demonstrable validity with respect
to any measures outside of the charmed circle.
Psychologists used to say as a kind of an "in" joke
that intelligence is what the intelligence tests mea-
sure. That seems to be uncomfortably near the
whole truth and nothing but the truth. But what's
funny about it, when the public took us more
seriously than we did ourselves and used the tests to
screen people out of opportunities for education
and high-status jobs? And why call excellence at
these test games intelligence?
Even further, why keep the best education for
those who are already doing well at the games?
This in effect is what the colleges are doing when
they select from their applicants those with the
highest Scholastic Aptitude Test scores. Isn't this
like saying that we will coach especially those who
already can play tennis well? One would think
that the purpose of education is precisely to im-
prove the performance of those who are not doing
very well. So when psychologists predict on the
basis of the Scholastic Aptitude Test who is most
likely to do well in college, they are suggesting
implicitly that these are the "best bets" to admit.
But in another sense, if the colleges were interested
in proving that they could educate people, high-
scoring students might be poor bets because they
would be less likely to show improvement in per-
formance. To be sure, the teachers want students
who will do well in their courses, but should society
AMERICAN PSYCHOLOGIST January 1973
Page 3
allow the teachers to determine who deserves to be
educated, particularly when the performance of in-
terest to teachers bears so little relation to any
other type of life performance?
Do Intelligence Tests Tap Abilities That
Are Responsible for Job Success?
Most psychologists think so; certainly the gen-
eral public thinks so (Cronbach, 1970, p. 300), but
the evidence is a whole lot less satisfactory than
one would think it ought to be to justify such
confidence.
Thorndike and Hagen (19S9), for instance, ob-
tained 12,000 correlations between aptitude test
scores and various measures of later occupational
success on over 10,000 respondents and concluded
that the number of significant correlations did not
exceed what would be expected by chance. In
other words, the tests were invalid. Yet psycholo-
gists go on using them, trusting that the poor
validities must be due to restriction in range due
to the fact that occupations do not admit indi-
viduals with lower scores. But even here it is not
clear whether the characteristics required for entry
are, in fact, essential to success in the field. One
might suppose that finger dexterity is essential to
being a dentist, and require a minimum test score
for entry. Yet, it was found by Thorndike and
Hagen (1959) to be related negatively to income
as a dentist! Holland and Richards (1965) and
Elton and Shevel (1969) have shown that no con-
sistent relationships exist between scholastic apti-
tude scores in college students and their actual
accomplishments in social leadership, the arts, sci-
ence, music, writing, and speech and drama.
Yet what are we to make of Ghiselli's (1966, p.
121) conclusions, based on a review of 50 years of
research, that general intelligence tests correlate .42
with trainability and .23 with proficiency across all
types of jobs? Each of these correlations is based
on over 10,000 cases. It is small wonder that
psychologists believe intelligence tests are valid
predictors of job success. Unfortunately, it is im-
possible to evaluate Ghiselli's conclusion, as he does
not cite his sources and he does not state exactly
how job proficiency was measured for each of his
correlations. We can draw some conclusions from
his results, however, and we can make a good guess
that job proficiency often was measured by super-
visors' ratings or by such indirect indicators of
supervisors' opinions as turnover, promotion, salary
increases, and the like.
What is interesting to observe is that intelligence
test correlations with proficiency in higher status
jobs are regularly higher than with proficiency in
lower status jobs (Ghiselli, 1966, pp. 34, 78).
Consider the fact that intelligence test scores cor-
relate — .08 with proficiency as a canvasser or
solicitor and .45 with proficiency as a stock and
bond salesman. This should be a strong clue as to
what intelligence tests are getting at, but most ob-
servers have overlooked it or simply assumed that
it takes more general ability to be a stock and
bond salesman than a canvasser. But these two
jobs differ also in social status, in the language,
accent, clothing, manner, and connections by educa-
tion and family necessary for success in the job.
The basic problem with many job proficiency
measures for validating ability tests is that they
depend heavily on the credentials the man brings to
the job—the habits, values, accent, interests, etc.—
that mean he is acceptable to management and to
clients. Since we also know that social class back-
ground is related to getting higher ability test scores
(Nuttall & Fozard, 1970), as well as to having the
right personal credentials for success, the correla-
tion between intelligence test scores and job suc-
cess often may be an artifact, the product of their
joint association with class status. Employers may
have a right to select bond salesmen who have gone
to the right schools because they do better, but
psychologists do not have a right to argue that it
is their intelligence that makes them more proficient
in their jobs.
We know that correlation does not equal causa-
tion, but we keep forgetting it. Far too many psy-
chologists still report average-ability test scores for
high- and low-prestige occupations, inferring incor-
rectly that this evidence shows it takes more of this
type of brains to perform a high-level than a low-
level job. For instance, Jensen (1972) wrote
recently:
Can the I.Q. tell us anything of practical importance? Is it
related to our commonsense notions about mental ability
as we ordinarily think of it in connection with educational
and occupational performance? Yes, indeed, and there is
no doubt about it. ... The I.Q. obtained after 9 or 10
years of age also predicts final adult occupational status to
almost as high a degree as it predicts scholastic perform-
ance. . . . The average I.Q. of persons within a particular
occupation is closely related to that occupation's standing in
AMERICAN PSYCHOLOGIST January 1973
Page 4
terms of average income and the amount of prestige ac-
corded to it by the general public [p. 9].
He certainly leaves the impression that it is "mental
ability as we ordinarily think of it" that is re-
sponsible for this association between average IQ
scores and job prestige. But the association can
be interpreted as meaning, just as reasonably, that
it takes more -pull, more opportunity, to get the
vocabulary and other habits required by those in
power from incumbents of high-status positions.
Careful studies that try to separate the credential
factor from the ability factor in job success have
been very few in number.
Ghiselli (1966) simply did not deal with the
problem of what the criteria of job proficiency may
mean for validating the tests. For example, he re-
ported a correlation of .27 between intelligence test
scores and proficiency as a policeman or a detective
(p. 83), with no attention given to the very im-
portant issues involved in how a policeman's per-
formance is to be evaluated. Will supervisors' rat-
ings do? If so, it discriminates against black
policemen (Baehr, Furcon, & Froemel, 1968) be-
cause white supervisors regard them as inferior.
And what about the public? Shouldn't their opin-
ion as to how they are served by the police be part
of the criterion? The most recent careful review
(Kent & Eisenberg, 1972) of the evidence relating
ability test scores to police performance concluded
that there is no stable, significant relationship.
Here is concrete evidence that one must view with
considerable skepticism the assumed relation of in-
telligence test scores to success on the job.
One other illustration may serve to warn the
unwary about accepting uncritically simple state-
ments about the role of ability, as measured by
intelligence tests, in life outcomes. It is stated
widely that intelligence promotes general adjust-
ment and results in lower neuroticism. For ex-
ample, Anderson (1960) reported a significant cor-
relation between intelligence test scores obtained
from boys in 1950, age 14-17, and follow-up ratings
of general adjustment made five years later. Can
we assume that intelligence promotes better ad-
justment to life as has been often claimed? It
sounds reasonable until we reflect that the "intel-
ligence" test is a test of ability to do well in school
(to take academic type tests), that many of Ander-
son's sample were still in school or getting started
on careers, and that those who are not doing well
in school or getting a good first job because of it
are likely to be considered poorly adjusted by them-
selves and others. Here the test has become part
of the criterion and has introduced the correlation
artificially. In case this sounds like special reason-
ing, consider the fact, not commented on particu-
larly by Anderson, that the same correlation be-
tween "intelligence" test scores and adjustment in
girls was an insignificant .06. Are we to con-
clude that intelligence does not promote adjustment
in girls? It would seem more reasonable to argue
that the particular ability tested, here associated
with scholastic success, is more important to success
(and hence adjustment) for boys than for girls.
But this is a far cry from the careless inference that
intelligence tests tap a general ability to adapt suc-
cessfully to life's problems because high-IQ children
(read "men") have better mental health (Jensen,
1972).
To make the point even more vividly, suppose
you are a ghetto resident in the Roxbury section of
Boston. To qualify for being a policeman you have
to take a three-hour-long general intelligence test
in which you must know the meaning of words like
"quell," "pyromaniac," and "lexicon." If you do
not know enough of those words or cannot play
analogy games with them, you do not qualify and
must be satisfied with some such job as being a
janitor for which an "intelligence" test is not re-
quired yet by the Massachusetts Civil Service Com-
mission. You, not unreasonably, feel angry, upset,
and unsuccessful. Because you do not know those
words, you are considered to have low intelligence,
and since you consequently have to take a low-
status job and are unhappy, you contribute to the
celebrated correlations of low intelligence with low
occupational status and poor adjustment. Psycholo-
gists should be ashamed of themselves for promot-
ing a view of general intelligence that has en-
couraged such a testing program, particularly when
there is no solid evidence that significantly relates
performance on this type of intelligence test with
performance as a policeman.
The Role oj Power in Controlling Life-
Outcome Criteria
Psychologists have been, until recently, incredibly
naive about the role of powerful interests in con-
trolling the criteria against which psychologists
have validated their tests. Terman felt that his
AMERICAN PSYCHOLOGIST January 1973
Page 5
studies had proved conclusively that "giftedness,"
as he measured it with psychological tests, was a
key factor in life success. By and large, psycholo-
gists have agreed with him. Kohlberg, LaCrosse,
and Ricks (1970), for instance, in a recent sum-
mary statement concluded that Terman and Oden's
(1947) study "indicated the gifted were more suc-
cessful occupationally, maritally, and socially than
the average group, and were lower in 'morally
deviant' forms of psychopathology (e.g., alcoholism,
homosexuality)." Jensen (1972) agreed:
One of the most convincing demonstrations that I.Q. is
related to "real life" indicators of ability was provided in
a classic study by Terman and his associates at Stanford
University. . . . Terman found that for the most part
these high-I.Q. children in later adulthood markedly ex-
celled the general population on every indicator of achieve-
ment that was examined: a higher level of education com-
pleted; more scholastic honors and awards; higher occupa-
tional status; higher income; production of more articles,
books, patents and other signs of creativity; more entries
in Who's Who; a lower mortality rate; better physical and
mental health; and a lower divorce rate. . . . Findings such
as these establish beyond a doubt that I.Q. tests measure
characteristics that are obviously of considerable importance
in our present technological society. To say that the kind
of ability measured by intelligence tests is irrelevant or
unimportant would be tantamount to repudiating civiliza-
tion as we know it [p. 9],
I do not want to repudiate civilization as we know
it, or even to dismiss intelligence tests as irrelevant
or unimportant, but I do want to state, as em-
phatically as possible, that Terman's studies do not
demonstrate unequivocally that it is the kind of
ability measured by the intelligence tests that is
responsible for (i.e., causes) the greater success of
the high-IQ children. Terman's studies may show
only that the rich and powerful have more oppor-
tunities, and therefore do better in life. And if that
is even possibly true, it is socially irresponsible to
state that psychologists have established "beyond a
doubt" that the kind of ability measured by intel-
ligence tests is essential for high-level performance
in our society. For, by current methodological
standards, Terman's studies (and others like them)
were naive. No attempt was made to equate for
opportunity to be successful occupationally and
socially. His gifted people clearly came from
superior socioeconomic backgrounds to those he
compared them with (at one point all men in Cali-
fornia, including day laborers). He had no un-
equivocal evidence that it was "giftedness" (as re-
flected in his test scores) that was responsible for
TABLE 1
Numbers of Students in Various IQ and SES Cate-
gories (Sixth Grade) and Percentage Subsequently
Going to College
1Q
High
Low
Socioeconomic status
High
51
33
% to college
71
18
Low
57
96
% to college
23
5
Note, x> = 11.99, t < .01, estimated tetrachoric r = .35, SES X IQ.
(Table adapted from Havighurst et al., 1962. Copyrighted by Wiley,
1962.)
the superior performance of his group. It would be
as legitimate (though also not proven) to conclude
that sons of the rich, powerful, and educated were
apt to be more successful occupationally, maritally,
and socially because they had more material ad-
vantages. To make the point in another way, con-
sider the data in Table 1, which are fairly repre-
sentative of findings in this area. They were ob-
tained by Havighurst, Bowman, Liddle, Matthews,
and Pierce (1962) from a typical town in Middle
America. One observes the usual strong relation-
ship between social class and IQ and between IQ
and college-going—which leads on to occupational
success. The traditional interpretation of such
findings is that more stupid children come from the
lower classes because their parents are also stupid
which explains why they are lower class. A higher
proportion of children with high IQ go to college
because they are more intelligent and more suited
to college study. This is as it should be because
IQ predicts academic success. The fact that more
intelligent people going to college come more often
from the upper class follows naturally because the
upper classes contain more intelligent people. So
the traditional argument has gone for years. It
seemed all very simple and obvious to Terman and
his followers.
However, a closer look at Table 1 suggests an-
other interpretation that is equally plausible, though
not more required by the data than the one just
given. Compare the percentages going to college
in the "deviant" boxes—high socioeconomic status
and low IQ versus high IQ and low socioeconomic
status. It appears to be no more likely for the
bright children (high IQ) from the lower classes to
go to college (despite their high aptitude for it)
than for the "stupid" children from the upper
classes. Why is this? An obvious possibility is
AMERICAN PSYCHOLOGIST January 1973