Journal cj Educational Piyckoloiy

1963, Vo! 54, No. 1, 1-22

THEORY OF FLUID AND CRYSTALLIZED

INTELLIGENCE:
A CRITICAL EXPERIMENT1
RAYMOND B. CATTELL
University of Illinois

The theory of fluid and crystallired intelligence propounded 20 years ago

by the author has since accumulated support. However, the crucial is-
sue of whether 1 or 2 general factors subtend intellectual performances
has lacked an experiment adequately designed for accurate, determinate,
simple-structure rotation at the 2nd order. By factoring culturally em-
bedded with culture-fair intelligence measures on a background of pure
personality primaries (N = 277 7th and 8th grade boys and girls), it is
shown that 2 general factors indeed exist. A review, with some mathe-
matical formulations, is given of the theory's implications for the nature-
nurture ratio, brain injury, standard deviation of the IQ, growth curves,
the concept of a relational difficulty hierarchy, test standardization, and
the relative validities of traditional and culture-fair intelligence tests

PRESENT TRENDS IN INTELLIGENCE but widely surmised, and stated very

CONCEPTS clearly by Ferguson (1956), for example
(see also the review by Hunt, 1961),
Since Spearman's demonstration of that the formal structure of abilities is
the general ability factor (g) first put partly developmental^ determined by
intelligence testing in the framework of what might be called "generalized solu-
scientific research, the field has experi- tion instruments" (i.e., habits the acqui-
enced three major theoretical develop- sition of which becomes a key, opening
ments, which, however, have not hitherto the way to rapid advance in some gen-
been interrelated. eral area of cognitive problem solving).
These we shall briefly call "aids."
1. Thurstone's (1938) proof that any
comprehensive array of ability measures 3. The theory of "fluid and crystal-
will factor uniquely into a set of simple- lized general abilities," which states that
structure, oblique, primary abilities, with more refined analytical methods the
from which g can be obtained as a general ability factor now measured by
second-order factor, more uniquely de- intelligence tests will be found to be not
terminable than by the first-order one factor but two. In spite of marked
approach. cooperativeness of loading pattern (Cat-
tell, 1952), which makes them difficult
2. The notion, most frequently asso-
to separate, they have properties differ-
ciated with the name of Piaget (1947)
ing in vital ways for educational and
1
clinical prediction.
The research reported herein was per-
formed pursuant to a contract with the United This article puts forward entirely new
States Office of Education, Department of evidence specifically on the theory of
Health, Education, and Welfare. We are also fluid and crystallized g factors, but inte-
indebted to D. W. Dunnan, and the school
teachers of Springfield, Illinois, and to C. C.
grates all three of the above and has
Newman, and the school teachers of Paxton, implications for practical testing in
Illinois. terms of culture-fair intelligence tests.
l
 RAYMOND B. CATTELL

THEORY OF FLUID AND CRYSTALLIZED of the theory may be briefly stated as

GENERAL ABILITY follows:
Although the theory of fluid and crys- 1. That among the primary abilities
tallized general ability was first stated in the area semantically designated "in-
20 years ago (Cattell, 1941, 1943), telligence" in our culture (and which, as
and since developed more precisely Guilford points out, is not totally gen-
(Cattell, 1950, pp. 477-491; Cattell, eral to cognitive performance) there is
1957a, pp. 871-879; Cattell, 1957b), the not one "general ability" second-order
momentum of custom among intelligence factor, as in the Spearman-Thurstone
testers, and the absence of sufficiently resolution (Cattell, 1943), but more.
sustained research on the theory, have Two of these are highly cooperative
hitherto denied to practical test con- (Cattell, 1952) in the sense that they
struction and use the advantages in un- agree in loading positively most of the
derstanding that might be expected to general ability primaries and have
acrue from such dual measurement. The largely zero loadings outside the intelli-
main problem in research advance is gence field. Being cooperative they are
one of organization, since, as the fol- very difficult to separate, and most pre-
lowing summary of the theory will vious ability factorings have been satis-
show, coordination is necessary among fied to recognize the joint factor as one
specialists and practitioners in several factor. This conglomerate has been made
different fields. Parts of the following the basis of IQ measurement and has
theory have been stressed by certain been called "general intelligence." Tech-
investigators, e.g., in Ferguson's (1956) nical reasons for this confusion are dis-
use of "ability" and "learning set," cussed in the section "Implications for
Hebb's (1942) notion of A and B com- Experimental Checks on the Theory."
ponents in ability, Newland's2 (1962) 2. The two factors, which will be
use of process and product, and Hayes' called fluid (f) and crystallized (c)
(1962) notion of "experience-producing- general ability, gf and gc, and indexed8
drives" (though the last joins cognitive as U.I. (T) 1 and U.I. (T) 2, have
and dynamic components here kept sep- different properties, as follows:
arate). But the necessary experimental a. Crystallized ability loads more
conditions for constructive conclusions highly those cognitive performances in
are possible only if the total theory is which skilled judgment habits have
kept in focus, as is happily the case in,
for example, the recent work of Horn 8
In the Universal Index (VI.) system pro-
(unpublished). The interrelated aspects posed elsewhere (Cattell, 1957c; Cattell &
Warburton, 1963), to preserve factor identi-
2
In communicating to nonpsychologist teach- fication during changing explanatory devel-
ers, for example, the particular aspect of the opment, U.I. (T) 1 was assigned to general
fluid and crystallized ability theory expressed ability and UJ. (T) 2 to flexibility of closure,
in Newland's "process" and "product" proves in French's (1951) series. However, as pointed
to be readily conveyed by these terms. In any out elsewhere (Cattell, 1957a, p. 830), there is
more precise and comprehensive context the increasing evidence that flexibility of closure
terms mislead, however, for the capacity to is really the Personality Factor UJ. (T) 19,
learn is not only a function of the process outcropping in ability when ability tests only
(fluid ability) but often of the product, as are used. Accordingly it preserves the his-
well as of personality and motivation. Particu- torical importance of order in the index much
larly when aids have been acquired, the rate better if the U J. (T) 2 niche—flexibility going
of learning is a function not only of fluid to UJ. (T) 19—is now allocated to crystallized
ability but of the crystallized products of for- general ability, and flexibility removed to UJ.
mer application of process. 19.
 FLUID AND CRYSTALLIZED INTELLIGENCE

become crystallized (whence its name) d. The standard deviation of the

as the result of earlier learning appli- IQ calculated "classically" from
cation of some prior, more funda- mental ages, will be about 12-16
mental general ability to these fields. points for gc and 24-25 points for g{
Thurstone's Verbal and Numerical (Cattell, 1937, 1955, 1957b). The
primaries, or achievement in geogra- latter will be more constant and bio-
phy or history, would be examples of logically determined, whereas the
such products. Fluid general ability, former will hinge on culture habits.
on the other hand, shows more in tests For example, classroom systems which
requiring adaptation to new situations, slow the bright and press the dull will
where crystallized skills are of no par- reduce the possible absolute individ-
ticular advantage. ual difference in gc at any age level,
b. Before biological maturity (15- and therefore, gc standard deviation
20 years of age) individual differences of IQ, when annual increment is con-
in the difference of g, and gc will re- stant. Culture may similarly influence
flect mainly differences in cultural op- the sigma on gc in that a powerful
portunity and interest. Among adults education, with larger annual mean
these discrepancies will reflect also increments and, therefore, larger age
differences in age, since the gap be- performance sigma, will tend to re-
tween gt. and gf will tend to increase duce IQ sigma. Parenthetically, since
with experience and the time decay traditional intelligence tests are a
of g(. Although the speed of an intel- mixture of gc and gf, the standard
lectual performance is not a function deviation of the IQ on such tests will
of U.I. 1 and U.I. 2 alone, (for ex- be a function as follows:
ample, U.I. 22, Cortertia, is a power-
ful personality contributor; Cattell,
1957a), performances at speed, in any (p) . .

cognitive field, will load gt more than

gc, while tests without time limit (un- where p subscripts mean across people
less they are absolutely new and adap- at one age, and a subscripts mean
tive) , such as Thorndike called power across age (years), taken as the vari-
tests, will have relatively large load- ance of the means when 1 year sam-
ings of gc. ples are included across an agreed
c. Fluid and crystallized abilities number of yearly age groups. As
will differ on their age trend plots for <7f(p, and a f(6) can be assumed to be
the general population, gf reaching an biologically given, the IQ sigma on a
early maximum at 14—15 years, while traditional, culture-contaminated in-
g0 increases to 18, to 28, or beyond, telligence test will in general become
depending on the cultural learning smaller (over school years only)
period for the given subculture. where schools produce a large age
Thereafter, whereas gf will decline increment in scholarship and where
from about age 22 continuously to opportunity is not highly adapted to
old age, gc will show a later and lesser native ability differences, i.e., when
drop. In all these comments on gc, it rCf(P) is small.
must be understood that we are talk- e. For any same-age group the
ing about the general factor in a broad nature-nurture variance ratio will be
array of crystallized abilities, not any much higher for gt than gc on the
single crystallized ability. hypothesis that gf is directly physio-
 RAYMOND B. CATTEIX

logically determined whereas gc is a in gt. The effect of local brain dam-

product of environmentally varying, age will be always to lower gf in pro-
experientially determined investments portion to the magnitude of the dam-
of gf. At present the only available, age, but the effect on the g,. score will
and still scant evidence, reached with only be through the localized ability
the new multiple abstract variance changes, e.g., verbal aphasia (Lans-
analysis method (Cattell, Stice, & dell, 1962). Consequently, if the gc
Kristy, 19S7) for culture-fair tests level happens to be estimated from a
(which we shall show are largely set of primaries which do not involve
gt measures), does point to higher the particular area, as a representa-
inheritance* for gf. However, al- tive, in the standard array of sampled
though it is our hypothesis that gf areas in a traditional intelligence test,
is biologically and physiologically de- no significant fall in gc IQ will neces-
termined, as a function of total corti- sarily be detected following local
cal cell count, this does not mean that injury.
one would expect anything like com- h. The implication in Part a above,
plete hereditary determination. For en- that crystallized ability will be a gen-
vironment includes gestation period in- eral factor, i.e., covering a substantial
fluences and later physical trauma and total and variety of ability measures,
physiological change, all affecting gf. needs to be examined. For such a
f. The reversible, day-to-day or factor to arise it is necessary that
month-to-month fluctuations in level people (a) compete in, i.e., be inter-
will be different in gf and gC) the ested and active over, the same series
former varying only with general of experiential areas; and (b) experi-
physiological efficiency (Pribram, ence difference degrees of learning
1960) and the latter less with physi- opportunity (to acquire skills) simul-
taneously over all these areas. Evi-
ology and more with recent exercise
dence for the capacity of common in-
and interest.
struction and experience to generate
g. The effect of general brain dam- such group or general factors comes
age (e.g., arteriosclerotic) will be from the pioneer researches of Anas-
similar in both but more pronounced tasi (1936) and of Woodrow (1938);
and later work by Greene (19S3),
4
This conclusion requires some special tech- Fleishman (1954), and others. Only
nical discussion for which we have little space when the common variance of experi-
Previous studies, such as those of Blewett
(1954b) and Burks (1928), Freeman, Eysenck, ence is considerable would we expect
and others used twin methods in which the the magnitude (mean variance con-
environmental difference range (that between tribution over a standard sample of
twins) is smaller than in the total family of variables) of gc factor to exceed that
sibs, as in our studies. Since the actual ratios
we obtained with Culture Fair Tests are close of gf.
to the mean of those they obtained with tradi- Since the present g( levels of a set
tional tests, despite our environmental range of people are simple growth-plus-
being larger, the Culture Fair Tests can be accident functions of their gt levels
concluded to have lower environmental deter-
obtaining a few years earlier, and
mination. Our ratios were i ? M = ge- since their present gc levels are only
a somewhat more complex function of
netic contribution for twins, and J*-L5_= 90% gi levels over preceding years (see
90.0
for the genetic component in between-family Equation 3), one would theoretically
variance in gt (Cattell et al., 1957, p. 154). expect gf and gc to be substantially
 FLUID AND CSYSTAUIZED IKTZLUGENCE

correlated. The conditions favorable experimental research checks. However,

to most substantial correlation prevail there are further implications which can-
in subjects at school age with respect not be stated concentratedly in the nine
to g«. factor measured over school sub- propositions and which require freer dis-
jects. In adult life, though school sub- cussion, before the need for certain fea-r
jects will continue to correlate by tures in our own experimental design
"remnant correlations" (see below), a can be clear.
more valid measure of gc will be made A first major implication is that the
by weighting a much wider range of form of fluid ability—as a pattern of
performances, though any one is likely expressions and loadings—is due to an
to be less loaded than were school influence present and operative at the
subjects. The use of culture-fair tests, time of the experiment, whereas crystal-
to measure gt, becomes, under these Iked ability has a form determined by,
adult conditions especially, the more and representing, history. Further, one
reliable, valid, and intelligible intelli- is biologically, the other culturally deter-
gence measure. mined. Our earlier analogy for ge (Cat-
i. The rate of learning in any par- tell, 1943) of the shape of a dead coral
ticular area will be a function, as far formation, representing the properties
as cognitive components only are con- and accidental history of once living
cerned, not only of gf, but also of gc, coral organisms, and Newland's (1962)
and the level of the specific ability recent use of "process" and "product"
factor in that area. In so far as gc (or Guttman's "analysts" and "achieve-
and the specific are concerned, we are ment") for gf and gc, bring out this
thus saying that learning begets learn- meaning figuratively, in simple meta-
ing capacity. Even so, the effect of gc phors, but not quite as precisely or com-
as a general factor should prove to be pletely as the real scientific model needs
small relative to g{ in proportion as to do.
the learning is in a completely new In behavioral terms, the theory of gf
area. And in old areas the non-grpre- is that it is "a capacity to perceive rela-
dicted variance is more likely to de- tions and educe correlates" in Spear-
pend upon crystallized group factors man's (1932) original sense. Further
than the general factor, namely, those the theory postulates that relations, be-
associated with what we have called ginning with those between sensory ele-
aids. The demonstration of such ac- ments, can be arranged objectively in a
tion in learning increments, however, hierarchy5 of increasing abstraction and
involves complex and refined statisti- 5
The present writer's definition of intelli-
cal treatment, as discussed by Hum- gence proceeds to developments of Spearman's
phreys (19S9), Lord (1956), Tucker relational theory which should be distinguished
(19S9), Woodrow (1939), and others. from those for which Spearman would be
responsible. It assumes that the hierarchical
organization of relationships among relation-
IMPLICATIONS FOR EXPERIMENTAL ships will begin with sensory fundamentals.
CHECKS ON THE THEORY The special sense areas will therefore form
subsets, in the sense of mathematical set the-
The original theory of fluid and crys- ory, and in a psychological sense, in larger
tallized general ability has just been sets. From this one might be inclined to con-
clude that the most complex relations must
briefly recapitulated in the form of nine necessarily lie in the Cartesian space involving
propositions integrating quite diverse all six sensory sets. This may be, but at least
fields of psychological observation, and for the ordinary ranges of intelligence, suffi-
indicating a possible array of tangible cient complexity can be reached in the hier-
 RAYMOND B. CATTEIA

complexity, and that the level of gf of endeavor lead to the general conclusion,
an individual at a given time is denned however, that to place individuals in
by the position in the hierarchy at which their proper rank order on gf it is nec-
his insight fails, as shown by the onset essary that the test fundament mate-
of chance response behavior. It is a task rial should be either, (a) equally over-
for logicians (Braine, 1959), ecologists, learnt by all, or {b) equally new to
and machine-simulation specialists to all (and therefore "perceptual" in the
develop bases for this hierarchy; but our sense used in this particular field).
hypothesis would be that it will be found Anything in between will contaminate
to rest on sensory area fundaments, and the variance with experiential variance.
therefore, require for its genesis intact The sensory area, e.g., whether an
sensory neurological structure, as stated analogies test is visual or by sound,
in Footnote 5. Meanwhile psychologists seems unimportant. However, the ap-
can delineate a rough hierarchy by proach through a—equally overleamt—
trial-and-error item construction, or- is risky unless certain conditions are
dered by enlightened difficulty-level watched, and even though it ranks per-
analysis (extremity-vector analysis; sons correctly, it gives no correct account
Cattell, 1957, p. 360) experiments. of absolute level. Moreover, there is
Most attempts, operationally, to always the risk that individuals highly
measure intelligence, even when it has familiar with and trained in a field will
been conceived as some sort of constitu- know by rote the correct answers to rela-
tional, fluid-ability, relation-eduction ca- tional decisions into which they do not
pacity, have been made, however, have, and have never had, any insight.
through first measuring crystallized For example, many people can respond
ability (i.e., on fundaments and even that Einstein's famous formula is e —
relations upon which strong cultural me2, not e=wc 8 .
training has already occurred). The new With these introductions to the gen-
movement which began with studying eral theoretical position we are ready to
relation-education in perceptual tests bring it to precise formulation, and to
(Cattell, 1931, 1940; El Koussy, 1938; ask what the implications are for a
Fortes, 1932; Line, 1931; Penrose & crucial factor analytic experiment. At
Raven, 1936) apparently supports the this level of precision it can no longer be
theory, however, that quite complex in- assumed, at least in the ordinary cir-
telligence-demanding relations can be cumstances of experimental testing of an
successfully set up in entirely new intellectual performance, that perform-
(and therefore perceptual, not memory- ance level is going to be determined only
stored) material. Both experience and by fluid and crystallized general ability.
psychological reasoning in this area of The specification equation must put
these hypothesized two general abilities
archy within the relations arising from one
sensory area, to reach the limen of failure of
in the full context of specific abilities and
insight. This happens because there is a very temperamental and dynamic traits, as
large number of further subsets among which follows:
complex relations can hold, even within one
sensory area. The successful use of Culture PJ = SjiFj + s j2 F 2 +s Jn F N +
Fair Tests (involving largely spatial funda-
ments) but transcending merely spatial rela- s j tT T +s, I Ti+SjFj [2]
tions and ability, offers some proof of a
sufficiently high relational complexity being where Fx and F 2 are g( and gc, respec-
attainable upon a limited (two sense) sensory tively (as in the U.I. factor index);
foundation. the Subscripts T, I, and J represent
 FLUID AND CKYSTAIXIZED INTELLIGENCE

temperament, interest (dynamic, moti- the numerous other performances

vation) , and local area (specific) ability covered by F 2 ;
factors, respectively, and F N represents Tej is time specifically given only to
a factor of strength of immediate mem-
ory (for if, as in senility, the capacity to Ic represents the mean strength of
hold the material is deficient the ana- interest in learning at that earlier
lytical capacities cannot act upon it). time which, like T, is assumed
Actually these latter, in a fully ex- common to several pjc's—e.g.,
panded formula, would each be several those in school work;
factors, but that expansion is unneces- IeJ is interest specific to p j c only;
sary for our present statement of essen- Fie is the level of fluid general ability
tials. If, moreover, the performance (pj) over this earlier (not the pres-
is made a highly intellectual one, and ent) learning period;
there is good, uniform, experimental mo- F Me is a memory capacity factor, also
tivation, and sheer speed is not so im- at its earlier level at that time;
portant that it brings in temperament F Ae is an aid, i.e., a discovered "in-
factors, the analysis of the variance in strument for problem solving"
PJ can be restricted largely to the three mentioned earlier9 which oper-
cognitive factors. ates as a group factor over some
area less wide than the general
To understand better what would be
ability factors;
the theoretically expected outcome of a
Cj is a statement of the change, an
factor analysis of purely cognitive vari-
increment or deterioration by
ables (i.e., of "intellectual" performance
measured under such conditions) let us, 8
An aid is a term for the generalizable habit,
however, pry more closely into the na- discussed by Piaget (1947), Hunt (1961),
ture of F,. This is our specification Hayes (1962), Forgus (1955), Ferguson (1956),
and others which opens up faster learning in
equation notation for gc which, accord- a whole region, e.g., the infant's grasping the
ing to Propositions e and i, should itself possibilities of babble, the kindergartener's dis-
be resolvable historically into other fac- covery that doing to others as one would like
tors. In fact, any variable likely to be to have done to oneself is an avenue to social
goodwill, the junior high school student's reali-
highly loaded on gc might, by special zation that a triangle drawn on the board
factor analysis directed to splitting F 2 , means not that particular triangle but a class
be expected to yield factors consistent of triangles, or that letters in algebra can
with the Equation 3. This equation stand for numbers, and so on. Such a piece
of mental equipment stands in the class of
states that a crystallized ability perform- achievement-attainment measures, as the use
ance (pjC) is a function of previous time of FA is intended to remind us. In the last re-
(T) and interest (I) invested in the sort, it is itself a product of Fi, Fu, etc., and
application of fluid ability; memory; belongs to that subclass of attainments which
and specific, problem solving aids. have the special property of aiding further
attainment. As such it has an independence
of these other terms and a characteristic field
of operation, probably as a group factor like a
[sjiF,e + s JM F Me +s JA F Ae ] + Q [3 ] Thurstone Primary Ability. In view of the
length of time which Piaget and his followers
where: have been discussing such action it is surpris-
T e represents the time given earlier ing that there are still no multivariate experi-
(e) to applying one's abilities in mental studies demonstrating their existence,
and their nature and position among abilities
a whole field of intellectual per- generally; but they seem a reasonable specu-
formances—i.e., to pjc, Pki and lation.
 RAYMOND B. CATTELL

"forgetting" since the original they are gathered within the brackets
learning experience, due to time with T and I outside. (The TI, if one
and changing interests, retroac- wishes, is an integration of interest over
tive inhibitions, etc. (In the time.) The introduction of the memory
scholarship area one may think term (F M ) recognizes, as pointed out
of the effect upon a foreign earlier, that a subject's correct judg-
language area of neglect for a ment, even on a test item of a highly
decade.) intellectual kind, may depend on good-
ness of rote memory, operating over the
The s's are the usual situational in-
course of exposure to an intellectual edu-
dices showing by weights specific to the
cation. For example, a correct decision in
time, interest, ability, etc., how much
geometry may depend on remembering
intelligence, memory, etc., contribute to
the formula for the area of a circle, or
performance, p Jc . The w's are weights
solving a crossword puzzle on good rote
similarly given to number of repetitions
memory for Greek. As Cohen (1957)
(time of exercise) and strength of in-
shows, the variance in intelligence per-
terest, e.g., w^ is the rate at which in-
formances from this memory component
creased time expenditure brings increase
becomes large in older adults.
in p jc score.
Now our problem is to ask to what
Two statements in this formula need
kind of a factor structure the theoretical,
especial comment: (a) Whereas in a
psychological formulation in Equation 3
present performance the specification
would be expected to lead. Should we
equation for the factor analytic model
expect distinct factors for T and I? The
adds factors of ability and interest to
answer seems to be that although very
estimate the performance, we have here
specially planned and exceptionally exe-
multiplied ability, memory, etc., by inter-
cuted factor analyses might express a
est and by the time, recognizing thereby
product relation like T e I e Fi e as an addi-
a different quality in them from addi-
tion of three factors, the normal preci-
tive influences within the organism.
sion and choice of variables would yield
Neither the rather hollow science of
a single general factor loading most
learning theory (still helpless where
highly the variables which have simul-
human higher abilities are concerned)
taneously shared high interest, much
nor the degree of precision of factor
time and much involvement of fluid
analysis (Tucker, 19S9), permit deci-
ability. However, because they are mul-
sion yet as to whether an additive or
tiplied by the same TJ.e the intelligence,
product model is better. But the prod-
memory, and aid factors would tend to
uct formulation is consistent with our
be cooperative (Cattell, 1952) (i.e.,
formulations of motivation in the dy-
loading the same variables). And if the
namic calculus, elsewhere (Cattell,
variance in T J * were great, we should
1957a) and is contingently adopted for
expect a powerful second-order factor
these accumulative learning effects, (b)
among them defined by TeIe. Thus there
The T and I terms as indicated by their
are really two possible definitions of the
appearance outside the bracket, are not
crystallized general ability factor: (a)
given separate values for the gc, memory
As the first-order factor TJ e Fi e , repre-
factor and aid factors, though in an ex-
senting the past areas of investment of
panded formulation they might be given
fluid ability, and (b) as the second-
different weights. By the nature of the
order factor—strictly a general achieve-
learning situation intelligence, memory,
ment factor—corresponding to the in-
etc., would normally operate together so
 FLUID AND CRYSTALLIZED INTELLIGENCE

vestment of Tel e in fluid ability expres- telligence (Symbol gc) should be re-
sion, memory, and various aids and spe- stricted to the first of the two factors
cial primary abilities. mentioned three paragraphs above,
The situation is further complicated namely, to the Expression T e I e Fi e . The
by the fact that skills not only increase second-order factor covering T e I c [Fic+
but decay, and a common history of Fiie+FAe] is in any sense of the word a
change or decay after the pattern has general achievement factor, though it
been reached as in the first part of Equa- must be confessed that a substantial part
tion 3 could also lead to a common fac- of traditional intelligence test scores de-
tor structure. This, with some over- rive from this source. Both of these
condensation, we have represented by factors are at the mercy of culture, and
the term Cj in the latter part of Equa- even before one encounters the difficulty
tion 3. For example, if Cj is a time of cross-cultural comparisons thereon,
function and our sample is scattered one gets entangled in the shift which
from 30 to 60 years of age, a general occurs in them between the schooling
factor would tend to appear over all age and the adult activity age. The Pro-
school subjects due to their all being tean character is thus at best reducible
forgotten to the same degree, as a func- to four concepts, corresponding to four
tion of the time lapse. Such considera- crystallized general ability factor
tions have been grossly overlooked in patterns.
traditional adult intelligence tests, as 1. School age crystallized intelligence
well as the fact that a new general crys- (gcs)—common to gf expressions pow-
tallized ability factor, distinct from the ered by time and interest within the
scholastic one, is likely to arise through culturally set framework of the school
common learning by the application T e I c curriculum.
to whatever common areas of gf use 2. School age crystallized achieve-
exist for adults in our culture. Here ment (acg)—a second-order factor across
only culture-fair, but not traditional, intelligence, memory, special ability
intelligence tests seem to recognize that areas, etc., corresponding purely to the
in the majority of our population which effect of common time, interest, memory,
proceeds to "nonverbal" occupations and curriculum. It will differ markedly
(e.g., garage mechanics, farmers, etc.) a from gca in being predictive of achieve-
substantial decay of the attained crys- ments (e.g., athletics, foreign languages,
tallized ability levels in vocabulary and and rote areas of school learning) in
similar areas supervenes (Burt, 1955). which g t has never had any appreciable
Further, the normal adult intense invest- importance or role.
ment of skills in relatively narrow occu- 3. Adult activity crystallized intelli-
pations will tend to make the appear- gence (gca)—representing whatever
ance of a second "adult" crystallized skills develop strictly from fluid intelli-
ability factor, superseding the school gence being applied in whatever activity
curriculum crystallized general ability areas receive common time and interest
factor, a relatively poor substitute, of from adults, i.e., T a I a Fi a .
lower variance and poorer predictive
power. 4. Adult activity crystallized achieve-
ment (aca)—representing whatever gen-
In summary, one must face the com- eral factor is produced in adult achieve-
plication that the crystallized general ments through their common expendi-
ability factor is for various reasons Pro- tures of time and interest. Except in
tean. Strictly the term crystallized in- mixed samples from different cultures
10 RAYMOND B. CATTELL

this would be a very tenuous factor of there is much to suggest that product
little practical predictive use. relations are often revealed by coopera-
Only in the case of Concept 1 would tive factor patterns. Consequently, since
we expect the correlation of fluid and cooperative factors are hard to rotate
crystallized intelligence factors to be correctly, it would seem best at this
really substantial, and the term intelli- stage not to attempt such finesse but to
gence is best eschewed completely for accept variables sufficient to reveal crys-
Concepts 2 and 4. Also the terms "school tallized intelligence as a "factor in
practised" and "non-school-practised" being," if it so exists. When such deeper
abilities are misleading for gc and gf, analysis is eventually undertaken it
for the first would describe both Con- would be well to keep in mind that I e
cepts 1 and 2 and the second could be might split again into a number of per-
either Concepts 3 or 4, or gf. Any com- sonality and dynamic factors (e.g., super
bination from decay influences is omitted ego strength, emotional stability) affect-
from this reduction to four general fac- ing the investment of fluid intelligence in
tor concepts. crystallized intelligence skills. Essen-
tially this has recently been powerfully
DESIGN OF THE EXPERIMENT IN argued by Hayes (1962). Decay factors
RELATION TO EXPECTATIONS will also need consideration.7
Testing the theory of fluid and crys- 2. Provisions must be made for defi-
tallized intelligence requires not one ex- nition by rotation of factors possibly
periment but a comprehensive organiza- general to every cognitive test employed.
tion of experiments. Yet one which One reason why the present theory,
needs to be performed before all others, rather than the Spearman-Thurstone-
as crucial to further work, is a factor Thomson theory of a single general in-
analysis to check whether one factor or telligence factor, has not hitherto re-
two runs through most intellectual per- ceived due and serious consideration is
formances. that most workers in the cognitive field
Technically, two questions need care have included only cognitive tests and
in planning such a factor analytic experi- thus rendered quite impossible any reve-
ment: lation of a distinctly rotatable second
1. Are we going to aim at abstracting 7
A summary of the main influences now
TeIeFjo as a single factor or at splitting posited in any performance in the area se-
it? Theoretically it is not beyond the mantically designated Intelligence might be
power of fine factor analysis to split a convenient at this final step: (a) a fluid
presently operating whole into the com- general intelligence factor; (6) a crystallized
ponents which historically produced it. general intelligence factor; (c) one or more
crystallized auxiliaries, corresponding to time
A product relation would be approxi- and application acting with sheer capacities
mated by additive factors as the plas- to memorize at the former among time (e.g.,
mode studied by Bargmann (1955), knowledge of vocabulary as distinct from nice
Cattell and Dickman (1962), Cattell judgment of meaning among familiar basic
English words); (d) factors corresponding to
and Sullivan (1962), and Thurstone any general patterns of change or decay since
(1947) sufficiently demonstrate. But in learning; (e) existing personality and dynamic
the present case T e , I e , and F l e would (motivation) factors helping application to an
be cooperative, because by the necessary intellectual test.
choice of tests for intelligence they would Here b, c, and d have structure in virtue of
be variables also on which common T e history, while a and to some extent e are struc-
tures due to present organismic functional
and Io have been expended. Besides unities.
 FLUID AND CRYSTALLIZED INTELLIGENCE 11

general factor. The procedures of Burt well-known factors, not merely variables,
(1955) and Vernon (1950) for example, so that they would fan out into a maxi-
neither permit unique determination of mum number of dimensions to give sta-
the alleged general intelligence entity nor bility of rotation of the ability factors
separation of a possible second general with respect to many reference points,
factor. and (c) themselves to have a known
The technical failure in factor analysis second-order structure, the emergence of
here, by which investigators have failed which in the present blind rotation would
to see the wood for the trees, is only give a check on its general correctness.
a special case of ignoring a general These requirements9 we felt to be best
principle in scientific method—namely, met by taking Personality Factors A
the figure-ground principle, according to through Q4 (having the same meaning as
which, if we wish to define X adequately on the 16 PF Test) on the High School
we must introduce data which is not X. Personality Questionnaire (HSPQ; Cat-
In this case to rotate with any approach tell & Beloff, 1959).
to precision a factor expected to be gen- The nine ability variables consisted
eral to all abilities one must include a of five Thurstone Primary Abilities—
wide framework of nonability factors Verbal, Spatial, Reasoning, Number, and
(not merely variables) as hyperplane Fluency—and four subtests from the
stuff (Cattell, 1952). Indeed, in spite Institute of Personality and Ability Test-
of the immense importance of the intelli- ing (IPAT) Culture Fair Intelligence
gence concept to education, the second- Test, Scale 2a—namely, Perceptual
order general ability factors appear never Series, Perceptual Classification, Ma-
to have been so rotated, in any research trices, and Topology. If our hypothesis is
discoverable by the writer, to a compre- correct all the Thurstone primaries, with
hensively determined simple structure the exception of Spatial and possibly
position.8 Fluency preformances should load on a
In the present study we planned to crystallized ability factor (as also should
reach the second-order analysis with be the case with each specific aid fac-
only nine ability measures—a bare but tor). 10 On the other hand, the Culture
sufficient minimum for testing the main
cognitive hypotheses—but with 13 non- °This procedure of entering the ability re-
ability measures. The latter were chosen search with more personality than ability
(a) to outnumber abilities, {b) to be measures, may seem quixotic to conservative
researchers, but in addition to the reasons
8
In many cases investigators have shown a given above it is actually strategically neces-
curious inconsistency in believing that first- sary in terms of demonstrating that second-
order factors should be rotated but second order ability factors are in fact distinct from
orders should be left where they fall! (The personality factors. This can by no means be
Schmidt-Leiman determination of loadings of taken for granted, since some "primary abili-
variables on second orders, for example, as- ties," e.g., flexibility of closure, now appear
sumes this.) In other cases, the explanation (Cattell, 1957a) as outcroppings of personality
seems to be the usual exhaustion phenomenon, factors "in disguise" in the cognitive realm.
10
in which, by the time the investigator gets to A small note on a large issue is necessary
the second order of a long factor analysis, here in connection with the notion raised above
neither time nor energy remains to press on to that an accumulation of separate aid develop-
a determinate second-order resolution. Mainly, ments might constitute a general factor. The
however, it seems that like early climbers on magnitude of the aid factor, F*1 is a joint func-
Everest, they have not dragged along at the tion of Fi and some environmental "accident"
early stages of the expedition the equipment A1, though whether it should be written as
which foresight might have indicated to be (a) Fi = Fx X A, or, (6) FA = Fi •+• A is un-
essential at the last stage of ascent. certain. In either case A would be operating as
12 RAYMOND B. CATTEIX

Fair subtests, being purely perceptual by the wider realm of personality meas-
and lacking even pictorial reference to ures) into such performances as Verbal,
the not-present should load only a fluid Spatial, etc., primaries.
general ability factor. However, since The choice of sample was made on the
even the Verbal item decisions involve assumption that Horn's (unpublished)
some immediate adaptive performance, research would use adults. For a com-
the fluid factor should extend to some plete answer on the above hypothesis
degree (in the rotation position denned requires a check on the expectation that
the crystallized intelligence factor will
a group factor covering from two or three to
a large number of performances.
behave very differently (as go, and gca)
In the early stages of the operation of an aid, with children and adults. There is some
as suggested above, it should be detectable in advantage in beginning with, say, 12-
either case as a group factor. In later stages, year-old children, because we may ex-
it would be possible for an accumulation of pect less complication from any decay
overlapping group factors to be resolved as a
general factor, indeed, they would constitute term ( Q in Equation 3). Also one might
part of the crystallized general ability. That a expect the gCB factor, based on so uni-
set of overlapping small group factors could form an area of earlier application, to
alternatively appear as a general factor has be larger than with adults (about the
puzzled factor analysts since the early debates
of Thomson and Spearman on the interpreta- same variance as the gr) and thus easier
tion of g as a "sampling of bonds" versus a to locate in a first study.
"single power." This factor analytic prob-
lem can be kept clear only by invoking the
principle of organismic setting. DATA AND ITS ANALYSIS
This principle states that although each of The sample consisted of 124 eighth
a set of separate covariational entities may ap-
pear as a separate group factor, yet, if they
grade boys and girls from Paxton, Illi-
all have some quality in common, so that they nois, and 154 from the seventh and
can appear as part of a larger organization, eighth grades of a junior high school in
then, by simple structure resolution, their sum Springfield, Illinois, a total of 277 (i.e.,
will also appear as a single factor, in the con-
text of several other, qualitatively different,
one dropped) being carried forward to
such accumulations. Thus although each leaf in the analysis. The Thurstone Primaries
a heap of leaves is a separate entity and (if its (except for Fluency) were given in both
variation, e.g., in perception from various forms, and HSPQ was similarly given
angles, were considered, a separate factor) yet
the heap also has its dimensions, as an entity,
in separate A and B scores for each
among other heaps. Similarly for cells in an primary personality factor; but the Cul-
organism or people in a team. What one gets ture Fair Intelligence Test used only
factor analytically depends upon the organis- one form (which must be remembered in
mic level at which measurements and rotations
are made.
appraising later saturations). Testing
took about 3 hours and was under one
Even before one gets a shift of organismic
level, a mere change of scale, from microscopic, examiner.
or, in terms of psychological variables, from The first order analysis required
low to high density of representation of vari- product-moment correlations11 for 44
ables, can convert the resolution from a mul-
11
tiple group factor outcome to a general factor The correlation matrix, centroid, lambda
outcome. Pending further experiment our hy- transformation, and final simple structure
pothesis is that the variance produced by matrix have been deposited with the Ameri-
cummulative action of distinct aids will, except can Documentation Institute. Order Document
for what is left in the truncated group factors No. 7381 from ADI Auxiliary Publications
(Cattell, 1962) become part of the crystallized Project, Photoduplication Service, Library of
general ability factor, i.e., Fi and F* in Equa- Congress; Washington 25, D. C , remitting in
tion 3 will both go into its variance. This advance $1.25 for microfilm or $1.25 for photo-
states an important difference from Hunt's copies. Make checks payable to: Chief, Photo-
(1961) theory. duplication Service, Library of Congress.
 FLUID AND CRYSTALLIZED IKTTT.UGE.NCE 13

variables (9 Thurstone abilities, 4 Cul- rotation, but with at least 20 variables to

ture Fair subtests, and 28 HSPQ per- define the hyperplane, this should be
sonality variables), i.e., in all 41 plus 3 a lesser risk and offer a better design.
random variables for hyperplane testing. Centroid factoring of the first-order
Since they consisted mainly of paired 22-factor matrix yielded, by the same
markers for factors, it was not surprising pair of criteria as before, 8 factors. Their
that the Tucker (Thurstone, 1938) and simple structure rotation is shown in
Guttman (Cattell & Sullivan, 1962) Table 1, together with Cp of their cor-
tests (separately applied) pointed to relations (see Footnote 11). The simple
20-24 factors to be extracted. We settled structure here is no longer quite as clean
on 22 as the limit for determinate, con- as at the first order. If the oddity of
verged, iterated communalities and iter- Factor 7 is omitted (with which we are
ated four times to reasonable stability. not much concerned) the count is, how-
The rotation for simple structure was ever, still 59%, which is high among
carried out with extreme care, since as reported studies and at any rate the
will now become apparent, all conclu- essential maximum or plateau attainable
sions in the second order hinge upon the among present possible rotations.
angle values settled upon in the first. The structure of Table 1 reveals un-
Simple structure reached, after four mistakeably the distinctness of the fluid
overall rotations, the very high value of general ability pattern, which is Factor 1
82% of all variables lying within the (loading the four subtests of the Culture
±.10 hyperplane. At this point in the Fair, the Thurstone Spatial, and the
blind rotations an inspection showed that HSPQ Ego Strength, C, factor), and the
the usually accepted structure for both crystallized general ability factor, Fac-
the Thurstone Primary Abilities and tor 2, which loads Verbal ability, Rea-
HSPQ primary personality factors had soning, and Number most highly. That
emerged, independently, with excellent these two ability factors, roughly orthog-
clarity. The two prescribed markers, onal to the remaining six, are essenti-
and virtually nothing else, loaded each ally in the right position is attested
named factor. Although only one marker further by the correctness of the second-
had been inserted for Fluency and each order factors emerging among these lat-
Culture Fair subtest, these five picked up ter personality factors. Factor 3 is the
enough on other variables, e.g., classifi- typical U.I. Q. I. (Cattell, 1957a) sec-
cation had loading on HSPQ Factor B, ond-order extraversion factor (Carrigan,
for each to appear as a nonspecific fac- 1960) covering the primaries A, D( —),
tor, though they are in the usual wider F, H, Q 2 ( - ) . Factor 4 is the general
sense, specifics. Anxiety versus Adjustment factor—
The C R matrix for this carefully ro- C( - ) , 0 , and Q,( - ) , (with Q3 slightly
tated study was now inverted to give anomalous); and Factor 5 continues
the correlations among the 22 factors these previously recognized second-order
(C F =DC E " 1 D) and this was again fac- factors to include that indexed as U.I. Q.
tored. The advantage of proceeding to I l l (Cattell, 1957a). Although these
the second order from the C B instead are to be remarked here only as confir-
of from correlations of actual battery mation of the "setting" for the two main
measures of the factors is that we elimi- ability factors, it is possible that Factor
nate distortions from the unreliabilities 6—"super-ego-good-upbringing" or Con-
and imperfect validities of actual scales. trol—has also some interaction with
In exchange we pick up whatever un- crystallized ability.
certainties remain in the simple structure Let us for the moment postpone ex-
14 RAYMOND B. CATTELL

TABLE 1
SECOND-ORDER FACTORS AND THEIR INTERCORRELATIONS

Part A: Factor loadings (in reference-vector correlation form)

F, F,
F.
Fi
F, F, F7 F.
First order factor Exvia- UJ.
variable g< ge Invia Anxiety Q.in Control ? ?

Thurstone primaries
Verbal .15 .46 -.17 — .03 .04 .07 -.OS -12
Spatial .32 .14 -.02 -.03 -.03 .OS -27 .04
Reasoning .08 .50 .08 .02 -.02 .02 38 -.14
Number .OS .59 -.04 .OS -.07 -.05 -.10 .03
Fluency .07 .09 -.52 -.18 .12 .26 .13 -.07
IPAT Culture Fair subtests
Series .35 .43 .05 -.12 .06 .10 .24 .04
Classification .63 -.02 -.05 -.OS .00 .14 .16 .06
Matrices .SO .10 -.23 .04 -.01 -.08 -25 .42
Topology .51 .09 .16 .12 .03 -.03 -.14 -.00
IPAT HSPQ
A Cyclothymia -.04 .52 .31 -.23 .05 .11 .20 .03
C Ego Strength .21 -.07 -.04 -.52 -.08 -.09 -.49 .12
D Excitability -.04 -.44 -.44 -.09 .09 .04 .14 -.27
E Dominance -.15 -.01 -.OS -.00 .61 -.04 -.02 -.17
F Surgency -.05 .09 .39 -.07 .06 -.40 -.20 -.09
G Super Ego Strength -.14 .08 -.01 -.10 -.10 .43 .41 .OS
H Parmia .21 -.04 .55 -.17 .00 .13 .03 -.03
I Premsia -.09 -.29 -.05 .26 -.54 -.OS .43 -.01
J Coasthenia -.10 -.04 .01 39 — .12 .45 .01 .16
Q Timidity .16 -.01 .01 .73 .09 -.04 .10 .17
Q. Self-Sufficient -.06 .05 -37 -.05 -.43 -.04 -.04 .05
Q» Self-Sentiment Control .05 -.02 .06 .00 31 35 .56 .11
Q. High Ergic Tension -.04 .37 .03 .45 -.00 -37 — .14 .05
% in ± 1 0 hyperplane .51 .60 .60 .51 .73 .55 32 .60

Part B : Factor correlations

1. Fluid general ability 1.00 .47 .29 35 .09 .21 -.01 -.15
2. Crystallized general ability .47 1.00 .17 .15 -.03 .28 .03 -20
3. Extroversion versus In via .29 .17 1.00 .17 -.01 .10 — .24 -.OS
4. Anxiety versus Adjustment .35 .15 .17 1.00 .24 .06 .14 -.12
5. UJ. Q. HI .09 -.03 -.01 24 1.00 -.46 .06 .26
6. Control .21 .28 .10 .06 -.46 1.00 .07 -.70
7. ? -.01 .03 -.24 .14 .06 .07 1.00 — .11
8. ? -.15 -20 -.05 -.12 .26 -.70 -.11 1.00

perimental examination of our assump- theory, on the grounds that both must
tions about the inherent modes of ex- be highly related to an entity which
pression of these two factors as shown existed a few years earlier, and which we
by loadings and pause to examine their have symbolized (Equation 3) by Fie
mutual relation as shown in Tables 1 (or gfe)—the fluid ability level earlier.
and 2. The substantial positive corre- The present crystallized ability level—
lation, + .47, shown in Table 1, Part B, F
2 (or gc)—is a function of F t e and the
is what would be expected from our time and interest applied at that time.
 FLUID AND CRYSTALLIZED INTELLIGENCE 15

The present fluid ability level—F2 (or TABLE 2

gf)—is an organic continuation of Fie, THIRD-ORDER FACTORS AND INTERCORRELATIONS
falling short of perfect correlation with
it only because different times have Third-order primary factors
Second-order factors
elapsed for different people and the in- I. 2. 3. 4.
terlude has treated them more or less
kindly in terms of brain damage, etc. Fluid intelligence
Crystallized intelligence
.69 .02
.63 —.04
—.07
.07
.00
.32
If next we ask what this correlation Extraversion versus Invia
(U I. Q. I) .18 .03 .38 .23
implies in terms of a third-order analysis, Anxiety versus Adjustment
(U.I. Q. II) -.01 .01 00 .79
different theories will give different an- U.I. Q III .09 —.51 —.07 .32
Control .01 .99 —.03 —.05
swers. Our theory would posit that Fj FT .06 .03 —.74 .04
and F 2 (g<; and g() should come together Fs .02 —.69 08 —.06
% in it 10 byperplane
in a single higher order factor, because
historically a single influence, Fu, is re- Primary factor correlations at third order
sponsible for them both. It has, so to 1. 1.00 .38 .14 .25
speak, fathered Fj out of time (as "the 2. .38 1.00 .00 .10
3. .14 .00 1.00 —.16
child is father of the man") by simple 4. .25 .10 —.16 1.00
maturation, and F 2 out of experience by
interest and learning investment.
obtained after five overall rotations, with
Whether the third-order F ]e should load
results shown in Table 2.
higher on Fj or F 2 depends on many
If our hypotheses is correct it would
things. High similarity of experiences
not be right to conclude from Table 2
(including time length) and interest
that "fluid and crystallized general
should raise the loading on F 2 , whereas ability form a single third-order factor,"
heterogeneity of the group in this respect but rather that a single influence, which
should lower it. Heterogeneity of the is fluid ability as it stood during the
group in age and time lapse since the formative period of crystallized ability,
formative period of gc should lower the is causative to the present levels of both.
loading in F L In a group of school That this "formative fluid ability" influ-
children such as the present, in the ence is in fact Fi in Table 2 could be
midst of their growth period, the Fie checked by preserved records of the
loading in Fj should be quite high and subjects' earlier fluid ability scores. It
that in F 2 a little short thereof. In adults is noteworthy in Table 2 that this fluid
or children more heterogeneous for age, ability factor has no significant loading
the loadings should both be lower. on any personality factor, but stands as
The third-order analysis designed to a measure of pure ability. On the other
test these hypotheses again had, by rea- hand, it has slight, significant correla-
son of the initial planning, enough hyper- tions as a factor with control and anxiety
plane stuff to determine the rotation of of a kind we have come to expect from
this still broader general ability factor. the linking of intelligence with other
Further, the rotation was again made factors through the natural selection of
with pure factor measures, not error- social status (Cattell, 1945; Fisher,
contaminated actual battery scores, for 1958). The factor contingently dubbed
simple structure was attained in the ref- Control here has, on the other hand,
erence vector matrix underlying C F significant relations to crystallized
shown in Table 1. In this C F four factors ability in Table 1 and must be consid-
were indicated by the Tucker test and a ered one of the personality factors con-
satisfactory 63% in the hyperplane was tributing, partly by aiding school memo-
16 RAYMOND B. CATTEIX

rizing, to crystallized ability, via FM, and crystallized general abilities, as well
etc., in Equation 3 above. as by such aids (usually primary, group
factor patterns) as are relevant to the
NATURES OF FLUID AND CRYSTALLIZED
particular field (Equation 3). Conse-
INTELLIGENCE quently, crystallized ability must be con-
With concepts denned and checked at sidered to be begotten by crystallized
the experimental level we can now pro- ability as well as by fluid ability. This
ceed to broader formulations. The state- may seem to contradict the concept
ments so far are: above that crystallized ability is a
1. That intelligence test and school consequence and function of fluid ability
performances must be considered, more levels, but it is not so if we consider (a)
comprehensively than hitherto, to be de- that fluid ability is so to speak put out
termined, at the moment, by personality at compound interest, and recognize
and motivation factors as well as by two also that the rate of return may be dif-
distinct second-order general factors, ferent over different periods of learning,
(fluid and crystallized intelligence) as depending on interest, etc.; and (b)
symbolized in Equation 2. Equation 2 that crystallized ability is not only a
is primarily an expression for perform- function of fluid ability but also of per-
ance in a timed test, but with trivial sonality factors, etc. Thus personality
adjustments, is also an expression for factor "deposits" occur in what at the
rate of learning. time of learning is considered an ability
2. As a result of this mechanism of —crystallized general ability—and also
learning operating uniformly across in aids. If space permitted, these rela-
school subjects, and other areas, for time tions of earlier personality and motiva-
intervals differing for different people, tion factors, as operating in Equation 2
third-order general factors of school to crystallized abilities as in Equation
achievement (for children) and adult 3, could be worked out in a fourth or
activity achievement (for adults) are "historical depth" equation.
created. (There is no space to discuss As to the descriptive similarities and
their relation here, but they should be differences of gt and ge, both are very
slightly positively correlated.) The term general to complex, intellectual, rela-
crystallized intelligence, however, is se- tion-perceiving performances, and even
mantically better reserved for the in regard to the variables chosen to dis-
second-order factor in such performances criminate them one notices, in Figure 1,
as derive their positive correlations only that each has some tendency to load the
from the action of fluid ability, not from variables set to mark the other. However,
memory and various special abilities the slight loading of crystallized intelli-
operating in the school (or adult) set- gence in say Pluency is not as great as
tings. Crystallized intelligence and that of fluid ability in Verbal or Series.
achievement are thus not identical gen- (Note with oblique axes the lines of
eral factors, the former being narrowed projection are drawn parallel to the axes
in manifestation to the handling of com- in Figure 1.) Parenthetically, it will be
plex relations, whereas the latter could noted that if the experiment had been
cover whatever the breadth of the cul- undertaken without personality fac-
tural schooling curriculum embraces. tors, etc., as hyperplane substance, the
rotator would almost certainly have
A special logical complexity arises
fallen into the false resolution of making
here, however, in that rate of learning in
these factors more highly correlated
complex fields is assisted both by fluid
 FLUID AND CRYSTALLIZED INTELLIGENCE 17

CKVSTAUJZED t
greater freedom of the second-order
fluid ability from personality loadings,
which supports our hypothesis, is not
incompatible with its having correla-
tions therewith at the third order, point-
ing to genetic and social status ties.12
A word is necessary here to forestall
a misunderstanding which will almost
certainly arise in the attempt of con-
servatives to avoid any major restructur-
ing of ideas. This will be the view that
gf is nothing more than the k or "prac-
tical ability" factor, located by such
factorizations as those of Vernon (1950)
FIG. 1. Culture-fair subtests and Thurstone and Cohen (1957). There are three good
Primaries in relation to %t and g e . (Rotation reasons for rejecting this. The first,
assisted by 16 PF factors in hyperplane.)
comparatively trivial, is that k has al-
(especially in the absence of D, I, A, ways shown decided associations with
and Q2) by going through the ability masculinity and mechanical experience
variables only, e.g., through Verbal and whereas in our mixed boy and girl sam-
Topology. ple gc shows statistically no difference
and is actually a shade higher for the
Figure 1 also confirms the whole argu-
girls. Secondly, and more systemati-
ment of Equation 2 to the effect that
cally, the whole mode of factor resolution
crystallized intelligence, being partly the
in the Burt-Holzinger-Vernon bipolar
product of motivational and personality
system is quite different from multiple-
history, will have more significant asso-
factor simple-structure resolution and
ciation with personality factors. This
never yields a one-to-one matching. The
association is not only statistically sig-
former method puts verbal ability on one
nificant but makes psychological sense,
side and practical-mechanical on the
for crystallized ability correlates sub-
other. In the present writer's opinion
stantially positively with Cyclothymia
this is in any case a fallacious form of
(Factor A in HSPQ), and negatively
factor resolution, for since the first gen-
with Excitability (D), and Premsia (I)
eral factor is indeterminate, the whole
which other researches show to be
hierarchy is fictitious. But, fallacious or
similarly related to general school
not, it is, except by some accidental
achievement. Similarly in the motiva-
tion realm (Connor, 1961), docility (low 12
In looking for indications of some sep-
self-assertion) and super ego strength arate action of the interest-personality factors
have the highest correlation with total posited in Equations 1, 2, etc., one may note
school achievement. Their correlations in Table 1 that Factors 6 and 7 admit of a
single factor being rotated from them which
with the Verbal (V) factor, the Rea- loads the two Super Ego factors, G and Q»;
soning (R) factor, and the Culture Fair Desurgency, F( — ) ; and positive performance
are, respectively, .20, .48, and .02; and on Reasoning, Fluency, Series, and Classi-
.19, .41, and - . 0 2 , (see Footnote 10) fication. It makes good psychological sense that
such cognitive performances should be aided
the Culture Fair (fluid ability) measures by such personality factors, but it suggests
consistently being more free of associa- they are acting in the present test situation
tion with personality factors affecting (as in Equation 1) rather than through help in
achievement. As stated above, this past performance as in Equation 3.
18 RAYMOND B. CATTELL

caricature, too different for gt ever to overlearnt by all persons to be tested, it

be identified with k. Finally, however, is of interest to discover what these
gf is not particularly loaded on mechani- areas may be. A systematic attack on
cal aptitude variables (the evidence is this was begun some years ago in the
not given here, but in Horn, unpub- work of Cattell (1940), Line (1931),
lished), or in spatial ability measures, or Fortes (1931), and others and pointed
one-sidedly in any of those performances to simple spatial perception, shade and
which have been included in this vague pitch perception, and probably tactile
and impractical practical ability artifact. experience and awareness of the body,
On the last issue—that of giving us a as being sufficiently overlearnt in most
far more complete descriptive differenti- cultures, but this exploration was never
ation of gf and gc in the extent and continued. Meanwhile, from a practical
nature of their loading fields—there is intelligence test construction standpoint
admittedly a great deal still to be found it is interesting to note that the Classi-
out. Considering the volumes of ability fication, Topology and somewhat less,
correlating that have gone on over the the Matrices tests are good measures of
past 20 years it has been, to say the fluid ability (at any rate as used in the
least, theoretically uninspired. Espe- Culture Fair Test) and Number, Rea-
cially the hypothesis should be tested soning and Verbal primaries of crystal-
that gc's loading pattern will alter with lized ability. Incidentally, the evidence
age, subculture, and culture whereas g('s of cultural learning in the Series test
will not. Already such studies as that is a bit surprising.
of Sisk (1940) point to the effect on the While commenting on odd loadings in
form of the crystallized intelligence pat- Table 1 one may note also the loading
tern of such very local influences as an which a factor, Number 8, outside the
"interest in study" over particular col- general ability factors, has on the Ma-
lege courses shared by members of a trices test. The present writer has else-
group. However, one unjustified as- where argued, on the theoretical likeli-
sumption needs to be scotched which hood of such a loading, against the le-
has been, among teachers, as obstructive gitimacy of using Matrices alone, for
in holding up the practical introduction measuring fluid general intelligence, in-
of culture-fair tests, as the k illusion has stead of a complete Culture Fair battery
been in buttressing outworn theory. This averaging several subtests. This shows
is the "faith validity" judgment that be- that indeed appreciable specifics or
cause such tests employ shapes they extra factors are as likely to occur in
must be testing spatial ability. As Fig- perceptual as in traditional subtests. It
ure 1 shows, the loading of the fluid is possible they will not prove to be cog-
general intelligence factor is not as great nitive specifics, but, as the present analy-
in Spatial as in Matrices, Topology, and sis suggests, less than half of the variance
Series tests, and though such tests de- in Matrices appears to lie in general in-
pend on perception there is as yet no telligence, and the rest is sheer unknown
evidence that they load unduly Thurs- bias in the test measurement if Matrices
tone's perceptual factors. alone is used as an intelligence test.
If, as stated at the outset, the theo- As to the present accuracy of estimate
retical aim of culture-fair intelligence of Fi and F 2 the loadings of the subtests
subtest construction is to require com- in the Culture Fair battery18 on the fluid
plex relation eduction in material that 18
The search for further good subtests might
is either completely new to or completely now be guided by the emerging theory that
 FLUID AMD CRYSTALLIZED INTELLIGENCE 19

ability factor are, in general, as good as, or sources of high performance, to per-
or a little better than, those of the mit prediction of more remote perfojrm-
Thurstone primaries on the crystallized ances in time and place, and to permit us
ability factor (mean=.40 as against to apply a psychological science via
.36). functional analytical concepts. The
writer had a student who never went to
NEEDED RESEARCH school till 17, who was soon getting
straight A's as a junior in college, he
At the present moment researches on did splendidly on a culture-fair test at
these theories are as vital in the applied 17 but would have been excluded from
as in the basic field. For if the main college on his traditional intelligence
concepts are correct, present routine test- test or achievement performance. The
ing practices in school, and still more in discrepancy of gc and g t is, however,
vocational selection of adults, are wast- mostly small in the school years, so long
ing a lot of talent. The traditional in- as we have homogeneity of school, cul-
telligence test, because it confounds fluid tural subgroup, and social status. It is
with crystallized intelligence and even in the adult field, e.g., in selecting for an
with the third-order general attainment intelligence-demanding learning oppor-
factor, gives a speciously good correla- tunity among persons of different ages
tion, at least in schools, with the crite- and distances from school activities, in
rion. A test in June predicts academic different regions, that failure to measure
achievement in June (or even in the gt and gc separately could lead to serious
following September) for the simple misdirection. How fair is the Miller
reason that it already surreptitiously Analogies to engineering students com-
contains in itself most of the criterion it peting for graduate school positions
is claiming to predict. In fact, if short- against English majors, compared with
term correlation is all that is required it a culture-fair test? It is well known
would be sensible to abolish intelligence that the norms of extensively standard-
tests and predict from today's school ized traditional intelligence tests are
achievement to tomorrow's. invalidated within 2 or 3 years of their
But the idea of introducing intelli- production by changing (usually in-
gence tests has surely much broader ob- creased!) school and cultural education
jectives. It aims to give psychological levels. Is this intergenerational "creep"
understanding of causes of backwardness of norms on crystallized intelligence
tests avoided by culture-fair tests as
fluid ability shows itself most highly, relative
to crystallized ability, in complex relation and preliminary data (Cattell, 1957b) in-
correlate eduction performances where crys- dicates? These are the problems which
tallized ability is eliminated either by (o) all applied research should be facing.
subjects being entirely untrained in the per-
formance, or (6) aU subjects being com- In terms of basic research a realm
pletely overlearnt in the fundaments (but not of fascinating concepts and possibilities
the relations, else intelligence would disap- is opened up. Is it possible that gf
pear) ! Possibly Culture Fair Subtest Number minus gc might be the best predictor of
4, Topology, is an example of the former and
Spatial ability of the latter, since the latter is organic brain damage? Can gt and gc
a common requirement of bodily movement also be separated by intraindividual, P
for people in all cultures. This definition of technique factoring, demonstrating the
one property of the fluid general ability needs functional unity of each in terms of sus-
to be experimentally pursued in relation to the
other hypothesized properties above, now ceptibility to day-to-day changing con-
that the factor can be isolated. ditions and stimuli? What clarity might
 RAYMOND B. CATTELL

be brought to the vexed question of pre- groups at 30 and 50 years of age; with
school ability organization by these con- adults heterogeneous as to age; and with
cepts? (There are important leads here mixed "normals" and known brain dam-
in the work of Burt, 1955; Hebb, 1942; aged cases. It needs to be tried with
Hofstaetter, 1954; McNemar, 1940; larger ranges of variables, identically
Reitan, 1958.) Will nature-nurture ra- used with two or three different cultural
tios calculated on gf measures prove to groups, to see whether the third-order
be altogether more highly genetically achievement general factor can be sepa-
determined than gc measures? Is the gf rated both from the second-order crys-
IQ more constant? tallized intelligence and fluid intelligence
Is it possible that what Thurstone factors, and especially we need age plots
called primary ability factors are psy- of the differing course of these two fac-
chologically the same as what we have tors, each with and without speed limits
called aids above? Probably both will in testing.
consistently appear as first-order factors,
relative to gc and gf as second order, and, SUMMARY
although some may well be genetic in The theory of fluid and crystallized
origin, the theory is worth exploring general intelligence factors has been ex-
that many first orders will prove to be
amined in regard to an array of implica-
growths from a specific breakthrough
tions. A distinction has been drawn be-
(in the form of hitting upon a habit of
tween a general achievement factor and
analysis which becomes instrumental to
the crystallized intelligence factor. Both,
a whole sequence of skill). In view of
the elaborate writing of Piaget, and still as distinct from fluid intelligence, will
more the speculations of his followers, so change their pattern (a) with age,
regarding structures of this kind (see and (b) with culture as to constitute
the recent summary by Hunt, 1961) it is relatively poor predictors of other per-
high time that more precise correlational formances in groups of different ages
methods should attempt to demonstrate and heterogeneous subcultures.
these patterns and relate them to the An experiment factoring 44 hypothe-
three types of formation here discussed.14 ses-relevant variables measured on 277
seventh and eighth grade boys and girls,
But perhaps good research strategy with sufficient noncognitive variables to
indicates that the first necessity today is permit effective rotation of any general
the checking of the existence of the two cognitive factors which might appear,
general factors in groups of differing demonstrated the existence of two gen-
ages, homogeneities, and cultural set- eral ability factors. One fits the crystal-
tings, in relation to the obvious differ- lized ability factor measured in tradi-
ences which would be expected from the tional intelligence tests and the other a
theory. For example, this experiment fluid general ability measured in culture-
needs repeating with age-homogeneous fair intelligence tests. Evidence is offered
14 that the latter is neither a spatial ability
In so far as the discovery of each aid (in-
strument for further problem solving) must factor nor the so-called "practical
be partly a function of fluid ability level, a ability" (k) factor, but basic general
part of the variance of all aid developments intelligence. These two general abilities
will be absorbed in the general factor of crys- appear in a single third-order factor hy-
tallized ability. The F/s of Equation 3 would
represent, therefore, only the narrow group pothesized to express the "formative
factor variance left over after this absorption fluid ability" partly responsible for the
in the general factor. present level of both of them.
 FLUID AND CRYSTALLIZED INTELLIGENCE 21

BIBLIOGRAPHY
ANASTASI, A. The influence of specific experi- structure and measurement. New York:
ence upon mental organization. Genet. Psy- Harcourt, Brace & World, 1957. (a)
chol. Monogr., 1936, 18, 24S-3SS. CATTELL, R. B. The IPAT Culture Fair In-
BARGMAKN, R. A demonstration study of the telligence Scales. Champaign, 111.: Institute
effectiveness of factor analytical models. for Personality and Ability Testing, 1957
Frankfurt: Hochschule fuer Internationale (b)
Paedagogische Forschung, 1955. CATTELL, R. B. A universal index for psycho-
BLEWETT, D . Twin studies on the inheritance logical factors. Psychologia, 19S7, 1, 74-
of intelligence and general neuroticism. Un- 85. (c)
published doctoral dissertation, University CATTELL, R. B. The basis of recognition and
of London Library, 19S4. (a) interpretation of factors. Educ. psychol.,
BLEWETT, D. B. An experimental study of the Measmt., 1962, in press.
inheritance of intelligence. / . menl. Sci., CATTELL, R. B., & BELOFF, H. The High School
1954, 100, 54-69. (b) Personality Questionnaire. Champaign, 111.-
BRAINE, M. D . S. The ontogeny of certain Institute for Personality and Ability Testing,
logical operations: Piaget's formulation ex- 1959.
amined by nonverbal methods. Psychol. CATTELL, R. B., & DICKMAN, K. A dynamic
Monogr., 1959, 73(5, Whole No. 475). mode} of physical influences demonstrating
BURKS, B. S. The relative influence of nature the necessity of oblique simple structure.
and nurture upon mental development. Psychol. Bull., 1962, 59, 389-400.
Yearbook Nat. Soc. Stud. Educ, 1928, 27, CATTELL, R. B., STICE, G. F., & KRISTY, N. F.
219-236. A first approximation to nature-nurture ra-
BURT, C. L. The evidence for the concept of tios for eleven primary personality factors in
intelligence. Brit. J. educ. Psychol., 1955, 25, objective tests. / . abnorm. soc. Psychol.,
158-177. 1957, 54, 143-159.
CARRIGAN, P. M. Extraversion-introversion as CATTELL, R. B., & SULLIVAN, W. The scien-
a dimension of personality: A re-appraisal. tific nature of factors: A demonstration by
Psychol. Bull., 1960, 57, 329-360. cups of coffee. Behav. Sci., 1962, 7, 184-193.
CARROLL, J. B. The factorial representation of CATTELL, R. B., & WARBURTON, F. W. Objec-
mental ability and academic achievement. tive personality and motivation tests: A
Educ. psychol. Ateasmt., 1943, 4, 307-332. theoretical introduction and practical com-
CATTELL, R. B. The Cattell Group Intelligence pendium. Urbana: Univer. Illinois Press,
Tests: Scales 0,1, 2 and 3. London: Harrap, 1963.
1931. COHEN, J. The factorial structure of the WAIS
CATTELL, R. B. The fight for our national in- between early adulthood and old age. J.
telligence. London: P. S. King, 1937. consult. Psychol., 1957, 21, 283-290.
CATTELL, R. B. A culture-free intelligence test: CONNOR, D . The prediction of educational
Part I. J. educ. Psychol., 1940, 31, 161-179. achievement from objective motivation fac-
CATTELL, R. B. Some theoretical issues in adult tor measurement. Paper read at Midwestern
intelligence testing. Psychol. Bull., 1941, Psychological Association Annual Meeting,
38, 592. Chicago, May 1961.
CATTELL, R. B. The measurement of adult EL KOUSSY, A. H. A note on the Greys Anal-
intelligence. Psychol. Bull., 1943, 3, 153-193. ogy Test. Brit. J. educ. Psychol, 1938, 4,
CATTELL, R. B. The cultural functions of so- 294.
cial stratification: I. Regarding the genetic FERGUSON, G. A. On transfer and the abilities
basis of society. / . soc. Psychol., 1945, 21, of man. Canad. J. Psychol, 1956, 10, 121-
3-23. 131.
CATTELL, R. B. Personality: A systematic the- FISHES, R. A. The genetical theory of natural
oretical and factual study. New York: selection. (2nd rev. ed.) New York: Dover,
McGraw-Hill, 1950. 1958.
CATTELL, R. B. Factor analysis. New York: FLEISHMAN, E. A., & HEMPEL, W. E. Changes
Harper, 1952. in factor structure of a complex psycho-
CATTELL, R. B. A note on Dr. Sloan's evidence motor task as a function of practice. Psy-
regarding the value of culture free intelli- chometrika, 1954, 19, 239-252.
gence tests. Amer. J. ment. Defic, 1955, 59, FORGUS, R. H. Early visual and motor experi-
504-506. ence as determiners of complex maze learn-
CATTELL, R. B. Personality and motivation ing ability under complex and rich and re-
22 RAYMOND B. CATTELL

duced stimulation. J. comp. physiol. Psy- LORD, F. M. Further problems in measurement

ckol., 195S, 48, 215-220. of growth. Educ. psychol. Measmt., 1958,
FORTES, M. A study of cognitive error. Brit. 18, 437-451.
J. educ. Psychol., 1932, 2, 297-318. MCNEMAB, Q. A critical examination of the
FRENCH, J. W. The description of aptitude and University of Iowa studies of environmental
achievement tests in terms of rotated factors. influences upon the IQ. Psychol. Bull., 1940,
Psychological Monograph No. 5, 1951, Uni- 37, 63-92.
versity of Chicago Press. NEWLAND, T. E. The assessment of excep-
GREENE, E. B. An analysis of random and tional children. In W. M. Cruickshank (Ed.),
systematic changes with practice. Psycho- Psychology of exceptional children and
metrika, 1953, 8, 37-52. youth. New York: Prentice-Hall, 1962.
GUIUORD, J. P., CHRISTENSEN, P. R., BOND, Ch. 2.
M. A., & SUTTON, M. A. A factor analysis PENROSE, L. F., & RAVEN, J. G. A new series
study of human interests. Psychol. Monogr., of perceptual test: Preliminary communica-
1954, 68(4, Whole No. 375). tion. Brit. J. med. Psychol., 1936, 16, 28-30.
HARIOW, H. F. Motivation as a factor in the PIAGET, J. The psychology of intelligence.
acquisition of new responses. In, Current London: Routledge & Regan Paul, 1947.
theory and research in motivation: A sym- PRIBRAM, K. H. A review of theory in physi-
posium. Lincoln: Univer. Nebraska Press, ological psychology. Annu. Rev. Psychol.,
1953. Pp. 24-49. I960, 11, 1-40.
HAYES, K. J. Genes, drives and intellect. Psy- REITEN, R. M. Qualitative versus quantitative
chol. Rep., 1962, 10, 299-342. mental changes following brain damage. J.
HEBB, D. O. The effect of early and late brain Psychol., 1958, 46, 339-346.
injury upon test scores, and the nature of SISK, H. L. A multiple factor analysis of men-
normal adult intelligence. Proc. Amer. Phil. tal abilities in the freshman engineering cur-
Soc, 1942, 85, 275-292. riculum. / . Psychol., 1940, 9, 165-177.
HOFSTAETTER, P. R. The changing composi- SPEARMAN, C. The abilities of man. London:
tion of intelligence: A study in T-technique. Macmillan, 1932.
/ . genet. Psychol., 1954, 85, 1S9-164. SWENEY, A. B., & CATTEIX, R. B. Dynamic
HUMPHREYS, L. G. Discussion of Dr. Fergu- factors in twelve year old children as re-
son's paper. In P. H. Dubois, W. H. Man- vealed in measures of integrated motivation.
ning, & C. F. Spies (Eds), Factor analysis / . din. Psychol., 1961, 17, 360-369.
and related techniques in the study of THURSTONE, L. L. Primary mental abilities.
learning. Technical Report No. 7,1959, Con- Chicago: Univer. Chicago Press, 1938.
tract No. Nonr-816(02), Office of Naval THURSTONE, L. L. Multiple factor analysis.
Research. Pp. 183-187. Chicago: Univer. Chicago Press, 1947.
HUMPHREYS, L. G. The organization of hu- TUCKER, L. R. Determination of generalized
man abilities. Amer. Psychologist, 1962, 17, learning curves by factor analysis. In P. H.
475-483. Dubois, W. H. Manning, & C. F. Spies
HUNT, J. McV. Intelligence and experience. (Eds.), Factor analysis and related tech-
New York: Ronald, 1961. niques in the study of learning. Technical
LANSDEXX, H. Laterality of verbal intelligence Report No. 7, 1959, Contract No. Nonr-
in the brain. Science, 1962, 135, 922-923. 816(02), Office of Naval Research.
LASHXEY, K. S. Cerebral organization and VERNON, P. E. The structure of human abili-
behavior. In K. S. Lashley (Ed.), The ties. New York: Wiley, 1950.
neuropsychology of Lashley. New York: WOODROW, H. The relation between abilities
McGraw-Hill, 1960. and improvement with practice. J. educ.
LINE, W. The growth of visual perception in Psychol., 1938, 29, 215-230.
children. Brit. J. Psychol. monogr. Suppl., WOODROW, H. Factors in improvement with
1931, No. IS. practice. / . Psychol., 1939, 7, 55-70.
LORD, F. M. The measurement of growth.
Educ. psychol. iieasmt., 1956, 16, 421-437. (Received April 20, 1962)