THE DESCRIPTION OF PERSONALITY.
I. FOUNDATIONS OF TRAIT
MEASUREMENT
BY RAYMOND B. CATTELL
Harvard University
I. THE PRESENT BARRIER IN DESCRIPTIVE PSYCHOLOGY
In spite of the impressive accumulation of data on personality
in relation both to environment and heredity, advances
in exactness of prediction or depth of theoretical understanding
have shown a distinct lull in the past few years. It is as
if the forces of research, while advancing by sheer pressureof
numbers on the flanks, have been held up in the centre by an
impassable bridgehead. The bridgehead controls transition
from the habit of defining personality in the vague terms of
popular speech to the practice of using exact and measurable
variables based on clear theoretical conceptions. The present
paper is a contribution to this transition.
Measurement of a kind—leading even to impressive
statistical treatments—admittedly exists. But a closer examination
shows the figures to be specious and lacking in
some or all of the characteristics required of true mensuration.
The proponents of measurement in this somewhat shadowy
world of personality qualities stake their defence on the well
known dictum of Thorndike (29) that "whatever exists,
exists in some quantity, and can therefore ultimately be
measured." Unfortunately the optimism engendered in some
psychometrists by this excellent statement blinds them to
certain, basic conditions of measurement, notably to the rule
of elementary algebra that added units must be of the same
kind. Therefore our first step must be to emphasize that
measurement can only follow upon the correct recognition
and definition of qualitative characters, i.e., upon advances
in descriptive psychology.
559
560 RAYMOND B. CATTELL
Because of the demoralizing disappointments which have
beset the prosecution of descriptive psychology—in attempting
to describe the stream of consciousness, in 'faculty'
psychology, in the Gestalt and the Geisteswissenschaftlich
approaches, and even in some aspects of factor analysis
(12, 18, 25)—it seems necessary to rally effort at this point
by reiterating the obvious truth that personality research
depends entirely on the soundness of personality description
and measurement. For, stripped of particularities, all research
methods consist of just this: measurement of personality
at a given moment, followed by lapse of time or application
of certain influences, and finally remeasurement. Consequently
the ability to deal with morphology is a prerequisite
in inquiring about growth and function. This is but another
way of saying that the genotype can be understood only
through studying the phenotype. Or again, looking at the
matter from the mathematical standard of degree of accuracy,
it may be said that the precision of predictions about the
growth and dynamic interactions of traits is limited by the
exactness with which personality can be described and
measured in cross section, i.e., statically.
This primary importance of description is not always
adequately realized or accepted. For the starkness of the
above methodological form is commonly hidden by complications.
And occasionallythe canon that description is primary
may seem completely flouted, as when one observes the great
progress achieved by psycho-analysis in understanding dynamics,
in spite of its having been singularly negligent of
description and measurement. But the contradiction is only
apparent, for in fact the technical level attained by psychiatry
in sheer description of psychotic and neurotic syndromes far
exceeded that of psychology in detecting and classifying individual
differences in normal subjects. Freud was able to
clarify the mechanisms of hysteria because he could recognize
when the hysterical syndrome became intensified or reduced;
whereas the psychologist found out little about the origins
of, say, suggestibility in children, because he was not able to
delimit the trait of suggestibility. The law that nosology precedes
etiology is not easily broken.
THE DESCRIPTION OF PERSONALITY. I 561
II. PRESENT FRONTIERS IN THE PROGRESS OF
TRAIT DEFINITION
The contribution of this paper to the problem of personality
description is, specifically, the propounding of a new
hypothesis concerning the nature of traits and the description,
in outline, of a methodology for the empirical determination
of trait unities. It thus prepares the ground for trait measurement.
The paper follows an earlier article (8) tentatively
clarifying certain personality concepts, and precedes, as a
theoretical introduction, two experimental researches (9, 10).
Reviewing the present position of trait theory, no one can
doubt that the most widely accepted hypothesis, especially in
educational, clinical, industrial and other branches of applied
psychology—in which rating scales flourish—is that personality
can be described in terms of discrete if not independent
traits. By adopting the creed that traits are single functional
entities, alike and comparable for different individuals, the
psychometrist is enabled to pursue, without any sense of sin,
the practice of converting merely qualitative into scientifically
quantitative description. Factor analysis also subscribes to
this hypothesis, but it can claim that its sins of assumption,
if they exist, are small ones; for its major independent factor
traits are built empirically on a foundation of many minor
traits, each of which, as a trait, is so narrow and specific that
no great assumption is made in presuming its unitariness.
This facile theoretical concession to the practical convenience
of applied psychology has not, however, remained
unchallenged. Psychologists of many different backgrounds
have protested that (i, 32) independent traits do not represent
the true structure of personality or have questioned the
alleged range and consistency of the traits employed (20, 27,
28). But the urgencies of practice seem to have brushed
these theoretical objections aside as so much hair-splitting.
Indeed, as often happens, the sheer volume of applied psychological
publication seems to have crowded more subtle
argument from the field of discourse, to the extent that a
casual reader might easily gain the impression that the critics,
562 RAYMOND B. CATTELL
rather than the proponents, of atomic traits, are intellectually
in hiding*
Those who are concerned to go beyond the use of traits
merely as counters, who wish to examine them more closely
and who insist that realism must have priority over convenience
or theoretical systematization, we may call, for this
discussion,' naturalists.' As representative of naturalism and
of the advances that can be made by shrewd observation and
insight, we may take the position of Allport. He begins by
recording the important phenomenological difference between
unique traits and common traits, saying (i, p. 297), "Strictly
speaking no two persons ever have precisely the same trait"
for, "What else could be expected in view of the unique
hereditary endowment, the differing developmental history,
and the never repeated external influences that determine
each personality?" On the other hand, he argues, heredity
and environment are sufficiently alike for the majority of
people in one culture to give substantially the same form to
the behavior of mature adults in many fields of quantitative
individual difference, e.g., in dominant-submissive behavior,
or radicalism-conservatism. In this way arise traits which,
by contrast to the above, may be called 'common traits.'
The view that all traits are essentially unique, but that
uniqueness approaches asymptotically the state of commonness,
is an indisputable conclusion alike of common sense and
clinical observation.1
By this view common traits would be
measurable, in terms of a common direction and common
units, but unique traits would not. Actually the mathematical
psychologist can claim that unique traits are measurable
in units unique to the individual, but this is rather a
Pyrrhic conquest for measurement.
Allport's second division of traits follows one made by
Stern (26) between driving traits (Richtungsdispositionen) and
instrumental traits (R-iistungsdispositionen), which traits, for
1
In agreeing with this view the present writer does not also subscribe to Allport's
criticism that factor analysis is incompatible with giving uniqueness to theindividual.
As Guilford (14) very lucidly argues, the uniqueness exists in the combination of
common factors. There are uniquenesses in Allport's sense, nevertheless, which
escape representation by common factors.
THE DESCRIPTION OF PERSONALITY. I 563
conformity with modern terminology, are called respectively
motivational and stylistic traits. The nature of the unity
underlying these traits differs, the former being teleonomic
(i.e., resident in common teleology, dynamic root, purpose or
goal of the parts) and the latter stylistic (i.e., resident in a
likeness of the parts with respect to the way in which an
end goal—of any kind—is striven for).
If space permitted, a preliminary review of the present
positions in regard to the nature of traits would illustrate
also the comparatively naive orthodox nomothetic view (3,
18), the view that there are not traits but only specific
habits, a position closely associated with stimulus-response
psychology (28, 32), the Gestalt view (4, 19), and the empirical
statistical view that a trait exists where the intercorrelations
of trait elements form a cluster of high values,
i.e., when there is an 'operational unity' (13, 15, 31). The
utility of these views will be discussed explicitly or by implication
below.
III. THE NATURE OF TRAIT UNITIES
At this point we propose to describe the kinds of unity or
integralness which, on a wide variety of psychological grounds,
can be argued to constitute the possible binding unities in the
elements of behavior. We shall discuss both the conceivable
kinds of trait unity and the kinds which are actually implicit
in some thousands 2
of trait terms in general use. Finally
we shall ask which of these unities are real and which fictitious
or misleading. In the ensuing section we shall try to determine
by what methods of empirical observation these various
types of true unity may be detected.
Any given unity is constituted by an integration of smaller
behavior elements, which we will call 'trait elements' and the
atomic nature of which, for the moment, may be assumed, by
reason of their being the result of repeated subdivision. The
words trait and syndrome are, therefore, used interchangeably
1
The list in question is that of Allport and Odbert (3) compiled from Webster's
dictionary, and the present comments arise from a detailed study of it in connection
with a further research (9).
564 RAYMOND B. CATTELL
here, both being larger aggregates of putative elementary
pieces of behavior.
The trait unity classification at which we arrive has six
categories, as follows.
I. Dynamic (teleological, teleonomic, motivational) unities.—
This form of unity is Allport's and Stern's directive unity,
in which the parts are integrated by the fact that they all lie
on the path to some one goal of the organism. They have a
common purpose. A very appreciable proportion of the more
important traits in trait lists 3
fall in this category, including
the definitely 'dispositional' (21) traits, e.g., timidity, amorousness,
assertiveness, acquisitiveness.
It is at first surprising that although language is well
supplied with trait terms corresponding to drives with a
relatively generalized goal object, such as needs, propensities
or ergs of any kind (8), it is practically mute with regard to
terms for the dynamic, teleological unities centering upon
special, immediate, local objects of the environment. These
latter unities are sentiments, attitudes and complexes acquired
in the individual life of the subject and may be covered by
the categorical label 'metanergs' (8). It may be that these
dynamic attachments of mixed ergic origin are generally too
narrow, transient or unimportant, either from the standpoint
of the single personality or of society, to earn a denomination.
Yet to decide that they can be thus dismissed because language
ignores them is too temptingly easy, and, to judge by
clinical experience, quite fallacious. It is most likely that
their great variety alone accounts for the failure of language
to crystallize terms. For where we find a sentiment unity,
or even a comparatively superficial attitude, that has some
claim to being a common trait there is usually a term for its
presence, e.g., patriotic, domestic, communist, fatherly.
This discovery that traits are to be found corresponding
'Murray's scheme (22) for rating individuals according to 'needs' is based on
just such an assumption of unity of behavioral intention or goal. The methodology
suggested later in this article should be capable of revealing the exact boundaries of
these important intentional unit systems and of deciding, for example, whether Murray's
fifty-one needs, or McDougall's sixteen propensities,or some other approach to
ergs most correctlydepicts personalitystructure.
THE DESCRIPTION OF PERSONALITY. I 565
to sentiments as well as to native ergs or propensities, reveals
that there are two senses in which the term 'dynamic trait'
is actually employed. Rating an individual for an ergic
trait, e.g., a need, requires an act of abstraction, which infers
the strength of a construct from analysis of a number of
actual sentiment manifestations. Certain sentiments may
be, among other things, manifestationsof, say, acquisitiveness
and the observer sums them from this aspect. The unity,
in short, is imposed by the observer—though along lines
indicated by biological and other considerations as to the
nature of this potential structure, the erg. It also resides in
the quality of satisfaction, as introspected by the subject.
On the other hand, the actual unities in the dynamic
structure of personality, the sentiments, formed by conditioning
etc., enjoying some degree of functional autonomy in
their attachment to particular objects, are derived from
diverse ergic roots. Without conditioning and all the environmentally
determined processes of sublimation and aiminhibition,
these latter structures would not exist. Their
existence makes it necessary to speak of two kinds of dynamic
trait unities: (i) Those unified about a real object. They
may be called metanergic dynamic traits; (2) those unified
about a perhaps never attained and frequently never conceptualized
biological goal, and which we may call ergic
dynamic trait unities.
In view of the uniquenessof sentiments and of their ergic
constitution, it might be contended that psychology, like
language, should concern itself little with metanergic dynamic
unities. But the clinician may reply that individual differences
in ergic endowment are less important than differences
in manner of cathexis or object attachment of these drives.
2. Social mould (environmental demand] unities.*—Many
trait terms point to a unity from the standpoint of society's
needs or as occasioned by the pressure of the environment
rather than from that of the individual's ergic goals. For
4
'Environmental mould' would be more correct, because more comprehensive.
But it is more cumbersome,and the great majority of such traits seem to be socially
rather than physically moulded.
566 RAYMOND B. CATTELL
example, society needs, in certain fields of the individual's
total activities, the quality of honesty. Educational pressure
is brought to bear creating a number of specific habits, each
of which is directed to the common social purpose of ensuring
public honesty. These behavior elements may be attached
to quite diverge dynamic systems within the individual, e.g.,
to affection, fear or self assertion, to patriotism or filial loyalty.
Such unities will arise when the press of the environment tends
to act with equal strength with respect to all the elements
demanded, as in a mould or die press. Thus with respect to
honesty, if the individual has lived in a good environment he
will have been strongly pressed for all kinds of honest behavior,
whereas if one kind of honest behavior has been neglected in
his upbringing it will be likely that all the rest have had an
equally faint imprint. Consequently, as between individuals,
the trait elements will vary together, and in one individual,
will be united by common origin and fate, notably common
rate of extinction, depending on common age of formation.
Traits which are largely social-environmental mould
unities are conscientiousness, courageousness, trustworthyness,
tactfulness, cynicalness, charm, superciliousness, humaneness,
selfishness, general inhibitedness (one form of
introversion?) religiosity, etc. It may be generalized that
they are largely traits for which there are opposite* in the dictionary,
whereas this is not true of dynamic traits, at least
of the ergic kind. Except where the opposites represent
reactions against society's pressure they are generally .merely
colorless refractory or inertia states, e.g., cowardly, untrustworthy,
lacking in charm.
Most social mould patterns appear to have a name, but
not all socially evaluated forms of behavior covered by a
name need issue in a real pattern in behavior, as the fifth
category (below) shows. Social mouldingand conscious social
recognition and labelling are not the same. The unity of the
trait elements is one of common variation, the conscioussocial
purpose being a common but not an invariable or basic
character of the unity. It is conceivable and likely that
many social and environmental mould traits exist for which
THE DESCRIPTION OF PERSONALITY. I 567
there is no name or social recognition, e.g., a set of acquired
habits may distinguish the good pedestrian from the bad
pedestrian, or the citizen experienced in undergoing bombing
from the inexperienced, With traits that develop through
moral and educational exhortation, however, it seems likely
that a better unity of the elements will arise when a trait
name exists and the social need which it meets is explicit,
conscious and educationally supported. Even then the unity
may nevertheless be of a low order, because the term is
differently understood by different educators and because
punishment and reward do not always act in the area perceived
only by the educator. Punctuality, for example, may
be conceived to apply to different patterns of situation in
different social groups (13), and even in the same social group
the individual may be most heavily punished for something
which never entered into the concept, e.g., not appearing
punctually for dental check-ups. On the whole, therefore,
the common possession (covariation) of elements in the unity
will not be great (13, 15). Incidentally we may note that in
so far as social mould traits form part of a single culture
pattern they will also be related (i.e., not statistically independent)
in the individual, e.g., if a culture demands of the
good citizen, among other things, politeness and courage, the
elements of each may show nearly as close an agreement with
those of the other as among themselves.
A few social mould unities will also be quite highly
developed dynamic unities of the second, acquired type
(sentiments), but, as indicated in the last paragraph, there
will be some social pressures which are the same for all
people yet which are never verbalized enough to be conscious
sentiments or sufficiently centered on a single object to be
unconscious sentiments. For some purposes the making of
a distinction between conscious and unconscious social mould
unities might be useful: the former would commonly have
some degree of dynamic organization as sentiments; the latter
would be collections of discrete attitudes and habits.
3. Constitutional, non-dynamic (temperamental) unities.—
There exist behavior -elements united in a common variation
568 RAYMOND B. CATTELL
and common fate because they spring from some unitary
constitutional endowment of the organism, which, however,
is not of a dynamic teleological nature.' The endowment
concerned is one of physical energy, nervous sensitivity,
fatigability, reaction-time speed, susceptibility to inhibition,
or some other unitary character not imposed by the environment.
Although the unitary character does not reside in the
environment the behavior elements are, of course, an interaction
between an heredity and an environment, and in some
cases the role of the environment is evident in provoking the
emergence of the pattern even though it plays little or no
part in deciding the nature of the pattern, e.g., in hereditary
psychotic unities, as when manic depressive disorder is
precipitated by environmental stress or epilepsy by a tumor.
In a true constitutional trait, environment, in so far as it
affects it, affects it as a whole. Prolonged excessive demands
on the part of environment might reduce the trait of 'energeticness,'
or disease (e.g., myxedema) might affect general
speed of response. On the other hand the ' general inhibition'
or 'introversion' produced by an unduly punishing environment
should be distinguishable from that general inhibition
and introversion due to constitutional sensitivity by the fact
that the former has a restricted pattern, depending on the
elements of the environmental mould which have been repressive.
Furthermore the constitutional trait will be more
common and similar for all individuals, in so far as human
genetic endowment is more uniform than are cultural patterns
and accidents of upbringing.
Contingently the following may be thought of as illustrating
constitutional unities: excitability, extraversion-introversion
(in one usage), intelligence, manual dexterity, general
inhibitableness, general emotionality, somatotonia, goodness
of retention, and some psychotic and neurotic syndromes.
4. Co-nascent (developmental, emergent) unities.—A basis
for classification which naturally occurs to one in thinking
over the principal grounds for the classification of living
objects and their characters is that of age and development.
A trait would then be considered unitary if its parts emerge
THE DESCRIPTION OF PERSONALITY. I 569
at the same time, ontogenetically or phylogenetically. Incidentally
the word co-nascent is employed because both
'developmental' and 'genetic' would be misleading. All
traits develop, and genetic is best used as synonymous with
'hereditary.' Actual trait names of this class are to be found,
though they are rare, somewhat anomalous and not particularly
important. We find 'infantile,' 'adolescent,' 'mature'
and 'senile.' 'Beastly' and 'monkey-like' perhaps illustrate
the phylogenetic usage.
Syndromes of. this type will be found to be in essence
either of a constitutional or of an environmental mould type
(almost invariably lacking dynamic unity), but possessing the
added peculiarity that their parts emerge together. This
historical 'accident' may be an aid in calling attention to the
unity, but it does not add anything to its structural nature.
Probably the greater proportion are unities of a constitutional
variety, i.e., syndromes determined by the constitution of the
organism, e.g., 'senility.' Some, however, e.g., adolescent, as
Mead's studies of adolescence in different cultures show, are
combined products of the environmental mould and of
environmental provocation of patterns intrinsically constitutional.
There is thus no reason for considering con-nascence
as an additional essential category.
5. Logical (stylistic,6
evaluative, semantic) unities.—This
category is suggested by Allport's stylistic trait (i), but
extends much more widely. He contrasted such traits with
dynamic traits, for the former describe the individual's
characteristic way of working toward some goal whereas the
latter are labelled by the goal itself. Examples are: forceful,
polite, dainty, gruff, eloquent, direct, debonnaire and cere-
monious.
6
In certain cases the stylistic type of logical trait may be or may have been part
of a dynamic unity. A trait such as politeness may have begun as an instrumental
habit in a dynamic trait. And, since functional autonomy is never complete,even in
the aged and rigid, the persistence of the style indicates some reinforcement and
reward, however obscure. Stylistic traits in Allport's sense in fact would bevariously
classified here. Some are dynamic,some a result of a pattern of early training (social
mould) and some may even be constitutional. When they are dynamic, however,
they are mainly brief instrumental habits and skills which can be variously employed
in the service of major traits. The majority of Allport's instances, however,seem to
be logically bounded,in the above sense.
570 RAYMOND B. CATTELL
These traits have no claim to functional unity. Primarily
the unity is an evaluative one, existing in the mind:
of the
observer. Now this is only a special case of those perceived
unities in which the classification of the elements together is
performed on some purely logical basis, divorced from any
intrinsic, functional unity. The logical bases may be very
varied. The elements may be classified together because
they have the same style, speed, social effect or purpose,
moral valuation, aesthetic character, etc., etc. If they have
the same social effect or purpose, e.g., evoking friendly
responses, performing services with respect to machinery,
they may secondarily, as a result of the social response to
them, acquire some functional unity, in which case they are
also true unities—social mould unities—but this is comparatively
rare. However, practically any basis of abstraction
in the mind of the observer is likely in some special situation
to have some practical use, so that every logical category is
potentially a functional category. This point is developed
below.
Examples of this wider logical unity are very common
among trait terms. Many of the earlier (and, unfortunately,
some of the later) 'special aptitude' categories in vocational
guidance are purely evaluative, e.g., clerical aptitude, social
intelligence, and have no demonstrated functional, psychological
unity in any of the above three senses. Personality
traits such as trustworthy, obnoxious, decent, formidable,
exhausting belong in this category.
An extreme variety of evaluative unity, in which even
logical coherence is lost, occurs with what may be called
semantic or etymological unity. Here the elements are bound
by nothing more than the social habit of referring to them by
the same name. This is the least excusable of the hypostatizations
which some psychological critics so strongly
deprecate. Pure instances are rare, but some degree, of arbitrary,
verbally-imposed unity may be suspected in such
traits as lady-like, chauvinistic, Falstaffian, Prussian, Babbittlike
and bourgeois. The origins of these counterfeits seem to
lie in a term having survived an historically real type, or in
THE DESCRIPTION OF PERSONALITY. I 571
etymological accidents, or in the far-fetched metaphorical
usage of a term accurate enough in its own context, e.g.,
'musty,' 'flabby,' 'asthenic,' 'acrid,' 'lionlike.' As Zola's
description of the Parisian woman or Kipling's description of
the empire builder is said to have helped create the type, so
these fictions may conceivably also provoke corresponding
real functional unities (social mould .type), in this way
resembling the true logical trait categories.
6. Miscellaneous forms of trait unity.—The kinds of unity
according to which trait elements could be grouped in unitary
traits are presumably unlimited. Classification could be
made, for example, according to such bizarre features as the
relative frequencyof use of various motor organs or according
to the alphabetical order of the stimuli which evoke the trait
elements. Our task has been, however, to discover the senses
in which the term trait is already implied by its dictionary
examples, and to ask which of these or other senses are
psychologically meaningful or desirable. The existing usages
which remain to be surveyed are few or unimportant. Chief
among them is the view that traits might be named by the
stimulus situations which evoke them or the kind of response
habit by which they are expressed.
Like most attempts at scientific description in terms of
stimulus and response—neglecting that third class of variable,
the state of the organism and its drives—this view seems to
have fitted reality too poorly to have produced verbal symbols
capable of standing the test of usage. One may track down
such apparent instances as 'alcoholic,' 'music-loving,' 'mechanically-minded,'
'sailorly' or 'fond of gardening,' in which
the stimulus situation labels the trait. But closer examination
shows one is being deceived by the ambiguity which
constantly dogs the term 'stimulus,' giving it at one time the
sense of an initiating situation and at another the sense of a
goal symbol. In so far as it is the latter, the labelling by a
stimulus is not different from that labelling by goal which is
characteristic of all dynamic traits. Indeed the above instances
are seen to be metanergic (sentiment) dynamic
unities.
572 RAYMOND B. CATTELL
These instances of relatively specific sentiment systems
invite a discussion of the view that there are no 'traits' but
only specific habits. As a representative of this view we may
take Weiss, who, criticizing the habitual hypostatization of
the nomothetic psychologists, says, regarding manifestations
of the alleged trait 'benevolence': "From the sensorimotor
standpoint these actions are all different, but because they
have the same biosocial effect they may be classified as
having . . . equivalence . . ." (32, p. 305). We argue, in
opposition to his insistence on the sensorimotor standpoint,
that it has been found of no use where traits are concerned.
The widest variety of sensorimotor connections may be
employed in the service of a single trait.
Occasionally one encounters instrumental sensorimotor
habit systems, e.g., auto-driving, footballing, verbal eloquence,
chain smoking, or others of a more individual and unverbalizable
kind, such as have been called 'actones,' skills and,
sometimes, attitudes, which have reached such a degree of
development that they are important enough to include in
the category of metanergic dynamic traits. They are, however,
so frequently attached as servants to some larger sentiment
or complex integrate, or available for interchangeable
service in the interests of a variety of sentiments and purposes,
that they can scarcely rank as dynamic traits in themselves
and belong rather in the realm of motor skills.6
The last conceivable criterion of trait unity which we have
to consider is by no means the least, and has been left till last
because it requires extensive discussion, running through the
rest of this article. It is embodied in the view that the only
real unity of trait elements lies in their correlating positively
together in a 'cluster' or mathematical 'factor.' This empirical
view has been uncompromisingly stated by Thurstone,
in an article criticizing "The traditional methods of dealing
with these complexities (trait structures)" because they "have
'The restricted trait spoken of here seems to be identifiable with Hull's (16)
'habit family hierarchy' denned as "a group of two or more habit sequences all of
which may be initiated by a particular stimulus and terminated by a particular
reaction."
THE DESCRIPTION OF PERSONALITY. I 573
been speculative, bibliographical or merely literary in character"
(30, p. 2).
The view that traits are only to be discovered by empirical
studies of covariance is strongly maintained and developed
in all the following discussion. But it is equally strongly
disputed that the definition of a trait merely as a mathematical
factor or, still more inadequately, as a simple cluster
of correlating elements, is defensible.
Our solution to the problem of traits is thus to recognize
that several—basically three—distinct types of trait unity
exist. Now in the past much discussion of the nature of
traits and the methods whereby they might be investigated
has been dogged by the misleading and often unconscious
assumption that they are all of one nature, or that only one
kind of unity matters. A consideration of something more
tangible, say agricultural implements, which may be classified
according to color, weight, cost, place of manufacture, agricultural
function etc., each grouping having some utility and
some groupings being more universally functional than others,
will suffice to remind one that organization is relative to a
purpose.
For most clinical work dynamic traits and constitutional
traits seem to be of primary importance. In education there
is relatively more attention to social mould traits. In
personnel selection and vocational guidance, as already
pointed out, the logical, evaluative trait has enjoyed great
popularity. Such traits, e.g., clerical aptitude, consist of
functionally unrelated elements of behavior which happen to
have in common certain social or physical effects. But even
the logical unity may be false, for the trait in question is
often philosophically hard to define. Does honesty, for
example, include aesthetic and intellectual honesty? Each
psychometrist draws his own 'logical' boundaries for the
trait he is testing, turning applied psychology into a Tower of
Babel.
Even if this difficulty were overcome, through agreement
to define traits by fiat, the logical type of trait unity would
remain an undesirable and unreal one. In this matter the
574 RAYMOND B. CATTELL
pure psychologist is at the moment being more practical than
the practical psychologist who favors such traits. For although
a test of, say, sociability, leadership or clerical aptitude,
made up of a farrago of unrelated behavior elements
falling in anyone of these fields, may tell us about the individual's
behavior, from the evaluative standpoint of society
or vocation (as the psychologist sees it), at that particular
moment in that particular field, only measures of the real
underlying unities of a dynamic, constitutional or social
mould type can enable us to generalize about the person's
performance in modified circumstances or to predict performance
in years to come.
From this discussion it follows that there are only three
kinds of unity which are truly intrinsic and functional (in
respect to one or more senses of functional), namely, (i)
dynamic trait unities, (2) constitutional trait unities, (3)
social mould trait unities. For some social purposes it is
useful also to consider logical trait unities.
IV. BASIC METHODS IN DETERMINING TRAIT UNITIES:
THE CASE OF DYNAMIC TRAITS
Accepting the viewpoint that the task of personality
research is to investigate not one, but several kinds of trait
unity, by the combination of examples of which the personality
can be described, we may now ask how these unities are
to be experimentally discovered. That is to say, by what
methods may the boundaries of dynamic, constitutional and
social mould traits be explored. (The boundaries of logical
trait unities can, of course, be arbitrarily fixed.)
Let us admit at the beginning that the methods by which
psychology—behavioristic psychology—establishes causal or
functional connections are in essence no different from those
of other sciences. Introspection, as involved for example in
psychoanalysis and a great variety of current methods
employing naive verbal report, may give guidance, but
ultimately it has to be stiffened by objective proof. The
methodology of psychology, in so far as it is concerned with
establishing causal connections or functional unities, has to
THE DESCRIPTION OF PERSONALITY. I 575
proceed by the universal method of discovering covariation
and inevitable sequence,among observed events. This fact is
frequently overlooked but seldom contested: it is seriously
contested only by the view that 'equivalence' of behavior
elements can be established by intuition and observation of
similarity of form, a view which will be discussed later.
The objective methods of establishing trait unities which
have so far received approval are the calculation of correlations,
the application of factor analysis or the use of analysis
of variance. All these are methods which look for common
variations in trait variables in the field of individual differences.
The notion has been implicit, in such correlation
studies, that the method is capable of detecting any kind of
trait unity, and it has been suggested that it be applied to
discovering dynamic unities and deciding, for example,
whether the massive drives posited by Freud, or the smaller
propensity units defined by McDougall, or the still finer
differentiations represented by Murray's needs, give the most
accurate account of the ergic structure of personality.
This suggestion seems to overlook the important principle
that different kinds of trait unity will yield different kinds of
correlation pattern. Dynamic traits, indeed, stand in a
special position in this respect, and we may well approach
the general problem of deducing trait structure from correlation
data by devoting this first section to the special case of
dynamic traits.
The basic generalization we wish to stress here is that,
among individuals possessed of equal endowment in a particular
drive, different manifestations will vary inversely and be
negatively rather than positively correlated. The situation
may be explored more fully by means of Fig. I, in which we
take a minimum population of two persons, possessing differing
endowments in the basic erg (in this case sex drive) and differing
amounts of investment in different manifestations. For
clarity of discussion it is necessary, further, to introduce the
notion of 'levels' of expression, representing stages inontogenetic
development and degrees of arborization in the dynamic
structure. Thus in this case there is first a break between
576 RAYMOND B. CATTELL
AREAS OF
B c o E F e
BEHAVIOR
A B C 0 E F
3RD LEVEL
ATTITUDES
2ND LEVEL *"
SENTIMENTS
1ST LEVEL
ERGS, DISPOSITIONS
FIRST INDIVIDUAL
FIG. i.
SECOND INDIVIDUAL
adult and infantile sexuality and then various levels of
sublimation or complexity of expression of adult sexuality.
The strength of the investments is indicated by numbers 7
which, at all levels, add up to the same quantity—that of the
individual's total endowment, as indicated at the root.
It is obvious that in this, or any larger population, the
7
The use of quantities here provokes discussion again of the meaning of common
and uniquetraits, for it may well be argued that, especially in dynamic traits, the finer
ramifications are unique. Our whole purpose is to measure common traits, but we
must admit that the small trait elements out of which, by correlation, the larger common
traits may be constructed, are themselves unique and measurable only in logical
(metric) not population (normative) units. There are theoretically two ways of
handling the measurement of these trait elements. We may make our small logical
categories so fine and numerous that even the most bizarre form of behavior will find
a category in which its presence or absence can bte recorded. Or we may retain categories
of reasonable size and record qualitative as well as quantitative variations in
each, expressing the qualitative variations as directions of vector quantities. The
latter seems impracticable, for we have no means of making a combined factor analysis
of directions and quantities.
This consideration reminds one that all trait forms eventually depend on measurements
in logical trait categories. All traits are relations between organism and
environment. They do not reside only in the organism. Since the cultural environment,
and to a lesser extent the heredity of organisms, slowly change, the functional
unity of a given common trait is not eternal. The common traits of an ancient
Egyptian might not be measurable in terms of common traits established today.
But the logical trait categories of the trait elements are (or can be) permanent. They
are the dust from which the organic, functional unities are built and to which they
return.
THE DESCRIPTION OF PERSONALITY. I 577
manifestations a + b on the one hand and the manifestations
c + d on the other, will tend to show a negative correlation
because they are alternate and complementary manifestations
of the individual's total endowment. They will also tend
to show a positive correlation because they spring from the
same common endowment. If a factor analysis is made of
the correlations of the A, B, C, etc., variables we should
expect that the factor pattern would yield (i) a general
factor, corresponding to the total ergic endowment (n), (2) a
bi-polar factor, positive in the c + d derivatives and negative
in the a + b derivatives, (3) superimposed bi-polar factors
subtending a smaller number of variables, actually two or
three in these instances. The relation of the magnitudes of
the variance due respectively to the positive general factor
and the bipolar factors will depend on the relative importance
of the (presumably hereditary) differences in total endowment
in this erg (drive) and the (presumably environmentallydetermined)
individualdifferences in cathexis, or internal and
external inhibitions.
It would seem, therefore, that with suitable criteria to
permit a realistic rotation of axes, factor analysis could detect
and delimit common dynamic unities. We could not, of
course, measuredirectly the strengths at the deeper, non-overt
levels of dynamic integration indicated in Fig. I, but these
could be deduced factor patterns, discovered as a plumber
could deduce the volume of water running through street
conduits merely from observing the faucets in use in many
houses. The picture would also be complicated by the fact
that the type of dynamic connection indicated only lightly in
the diagram, namely a confluence of drives, through which
any piece of high level behavior springs from several basic
drives, would in most human behavior be far more frequent.
The mathematical picture would therefore be expected to
be more complicated, involving many more factors, than that
which we are at present accustomed to find in factor analysis
of more constitutional kinds of traits. But the method, with
suitable improvements, seems to be rightly oriented for
discovering the major sentiments which form common traits.
578 RAYMOND B. CATTELL
An indication of its correctness and practicality is offered by
one of the few soundly based analyses of orectic traits yet
available, namely Burt's study (6) of recorded (not rated)
emotional behavior in children. There emerged precisely the
type of structure here argued on theoretical grounds: a
general factor of total emotionality and various superposed
bi-polar factors corresponding to alternative expressions.
Even without elaboration of mathematical procedures the
method could be successfully applied to animal motivation
studies, in which the highly complicated sentiment and habit
superstructure of human beings would only be faintly repre-
sented.
Dynamic traits, however, in contrast to the two remaining
types of trait unity, can be investigated by other means than
the factor analysis of individual differences in behavior. For
there exist a number of approaches which we may call temporal
sequence methods,or intra-individual studies.
Bete.
Ef'etc.
FIG. 2.
Temporal sequence study looks for covariations within
the individual instead of variations between individuals, as
in factor analysis. It is, in other words, an intra-individual
rather than an inter-individual method, but this is only one
aspect of the difference. There are really two problems and
two methods here, and if we wish to use existing terms which
approximately indicate the difference we have in mind we
should call them the problems of equivalence and of subsidiation
or purposive sequence. Figure 2 puts these problems
schematically. A', A", and A'" are forms of behavior
THE DESCRIPTION OF PERSONALITY. I 579
springing from the same ergic root and directed to the same
biological goal. They are, in Allport's terms, equivalents.
A', B', C', etc. are successive pieces of behavior on the path
to the goal by way of the A' route. They are items in a
purposive sequence. The investigation of the structure of a
dynamic trait (in this case, of course, an individual, unique
trait, though it may be common if others also possess it)
requires elucidation of both kinds of connection: equivalence
and purposive sequence. These correspond to bipolar and
other factor structures in the factor analysis approach.
Now it is sometimes claimed that the first kind of connection—equivalence—can
be established by inspection of
the quality of the behavior, namely of the symbolic and other
resemblances which enable the insightful psychologist to
discover by intuition the equivalence of purpose, as when he
explains to you that a man kicking a chair really wants to
punch his opponent's head. Similarly some psychologists
suppose that purposive sequence or subsidiation can be
discovered without study of covariation, for in clinical practice
and everyday life it is usual simply to ask a man for what
more remote purpose he acquires a certain piece of behavior,
or we make an intuitive judgment or we ask him to introspect
by way of free association.
It is to be noted, however, that even in the hurried
conditions of routine work we do not entirely trust these
short cuts. We do not believe a conversion hysteric's
explanation of why he cannot walk, or our friend's rationalizations;
nor can one tell from inspection of style whether a
man who is thumping a piano does so to release aggressiveness,
lust or fear. In fact dynamic, teleological links have to be
discovered, in the end, like any others, by observing how
events vary together.8
This bleak truth is less escapable
when one considers studies of animal motivation.
8
It is contended by Allport, on the other hand, that the unity of a trait is such
that from understanding it one can predict the individual's behavior in a field with
regard to which there have been no previous observations made. This is a prediction
from the formal, essential nature of the trait. In the view of the present writer this
might be done, but only when the formal or intentional nature of the trait has been
established by factor analysis. Thus if we find that all the highly 'g' saturated
performances involve the eduction of complex relations we may risk the prediction
580 RAYMOND B. CATTELL
Temporal sequence studies aim at detecting these covariations
and they do so essentially in two ways, corresponding
to the two kinds of problem. First we may observe temporal
variations in two or more pieces of behavior and see if a
sequence of observations on one correlates with a sequenceof
observations on the other. Thus if we plot an individual's
sociability and his interest in the opposite sex and find the
daily variations are negatively correlated we may conclude
that these two kinds of behavior are equivalents, like A' and
A" in Fig. 2. An all or nothing variation, as in the observation
of the repeated disappearance of one symptom with the
coincident emergence of another, constitutes a special case of
such correlation. This method of temporal covariation permits
experiment, in addition to passive observation, for one can,
for example, remove one form of behavior, by deprivation or
inhibition, and record changes in the other.
To interpret the structure and general nature of the
dynamic connection in such a case, from the magnitude of the
correlation, is, however, no simple undertaking. For example,
a complete absence of significant correlation in temporal
covariation studies does not prove absence of 'equivalence'
in the two pieces of behavior concerned. Decline in
A' behavior may result in an increase of A'" behavior rather
than a change in the particular variable, A", one has chosen
to pair with the first. In fact we meet again here the problem
we have already encountered in the cross-sectional, factor
analysis approach,- but here its manifestations lie in the
variations in a single individual. For here too there will be
a general factor among the increments, due to the tendency
for equivalent elements of a purposive behavior trait to vary
together as well as inversely. The common variation will
arise here, not from differences in hereditary endowment but,
for example, from endocrine changes or swings of appetite,
that a person with the trait of high intelligence will do well in a situation obviously
characterized by the need for complex relation eduction, even though we have no
previous experience of the correlation of this test situation with the tests by which
the trait 'g' was identified. However,in so far as our conceptualization of the essence
of the trait might be faulty or limited, such prediction would always involve some
guess-work.
THE DESCRIPTION OF PERSONALITY. I 581
producing covariation in all manifestations of, for example,
sexual behavior. On the other hand the bipolar factors will
result, as in the previous case, from circumstances of internal
and external inhibition, and will produce inverse common
variation of equivalent (i.e., alternative) manifestations of
the disposition. It seems desirable, therefore, to pursue
longitudinal, temporal covariation studies by observing variation
in quite a number of manifestations at the same time,
rather than in a pair only.
The second form of temporal sequence study, which deals
with the unravelling of purposive sequence or subsidiation
(22, n, 23), we may call the method of temporal inwriance.
It is concerned with finding out what invariably follows what.
It is therefore identical with the search for any kind of causal
connection, except that in teleological, final causation one is
interested in the sequence in both directions.9
The aim of temporal invariance study is to find out for
what more remote purpose a given piece of behavior is being
carried out, i.e., to discover the A' to B' to C' linkages in
Fig. 2. One piece of behavior (a trait element) serves and is
connected with another if the two form an invariable sequence.
The more remote goal is the behavior or satisfaction, among
many following pieces, which alone invariably follows.
Psychoanalytic free association isa mathematically unchecked,
introspective and loose application of this method. As in the
preceding method, deliberate experiment, by manipulating
dynamic successions and by stimulating and depriving, can
supplement passive observation. Generally this particular
longitudinal approach requires little or no mathematical elaboration,
the task being one only of recording frequencies of behavior
elements preceding or following a given piece of
behavior.
Dynamic unities, we have argued, are of two kinds: ergic
unities, e.g., needs and dispositions, in which all behavior
8
In efficient causality one looks for the invariable predecessor; in final, teleological
'causality' for the invariable successor. The present writer considers that all teleological
causality in psychology is ultimately a manifestation of efficient causality, but
that it may be advantageous to establish the latter by first establishing the former,
which is simpler.
582 RAYMOND B. CATTELL
elements directed by intention to one biological goal are
abstracted (this appears as a general factor in factor analysis);
and metanergic unities, e.g., sentiments, in which drives from
different ergic roots fuse in a cathexis upon a single real
object. Temporal sequence study, in either of the above
sub-methods, seems to be the only means, in extra-individual
study, by which either type of unity can be discovered. The
method suggested by Baldwin (5) to explore the sentiments
of a single individualby recording the frequency of coincidence
of diverse emotional reactions to the same objects, is a. variety
of sequence study, since the identification of the drives
involved in the sentiment fusions depends on sequence study.
Among other suggested alternatives to the basic methods
here described as exclusive is the method of validating a trait
entity by 'prediction.' One writer argues that "successful
prediction from life history would establish traits and nothing
in this process demands factor analysis, covariance or even
quantity." But unless the prediction proceeds successfully
from one observed element in the trait to an entirely different
element (in which case the proof rests on exactly the same
grounds as correlation through simultaneous covariation,
above), the prediction provides evidence of nothing more
than consistency, reliability or absence of function fluctuation
in the trait element concerned. Function fluctuation is, of
course, an issue distinct from, and simpler than, that now
being discussed.
In exploring dynamic trait unities, then, we are confined
definitely to the study of covariance, in inter-individual and
intra-individual circumstances. Only the latter circumstance
is capable of leading to knowledge of truly unique traits.
The collation of both kinds of results, however, permitting
sequence study to aid the choice of factors in factor analysis
data, seems necessary at present for the successful exploration
of common traits. For unless unforeseen methodological
improvements make factor analysis more self-sufficient and
definitive, and allow us to handle the problem of unknown
and mixed levels of dynamic trait manifestation, the sequence
study must remain not just an auxiliary but a necessary
preliminary.
THE DESCRIPTION OF PERSONALITY. I 583
V. BASIC METHODS IN DETERMINING TRAIT UNITIES:
THE SITUATION FOR TRAITS IN GENERAL
Traits other than dynamic traits cannot be investigated
by sequence studies, for constitutional and social mould traits
do not involve, in the trait unity, striving for a goal. The
covariation of elements which constitutes the unity of these
traits has to be discovered by that inspection of correlation
coefficients which has been called factor and cluster analysis.
The straightforward application of factor analysis, however,
may be insufficient to bring out these trait unities.
For it is obvious, in the first place, that the factors, clusters
or correlation patterns arrived at depend, in their nature and
number, upon the sources of variability in the particular
population used. In most factor analysis it has been customary
to make the population homogeneous for those
aspects of personality with which one is not particularly
concerned, e.g., age, sex, education, natio-racial sample,
cultural background, etc., in which case one considers the
pattern established only for persons with that background.
For the shape of clusters, the factor saturations of variables
and even the very emergence of a factor will depend on this
preliminary arrangement. The point is illustrated by such a
common observation as that stature or reading speed will
correlate highly with the general ability factor in an ageextended
child population, but little or not at all in an adult
group.
Now it is argued here that a deliberate manipulation of
the population sample and its circumstances, in order to
contrast the results of different circumstances, may be
necessary to discover trait unities of various kinds, and that
factor analysis has to some extent failed in clarifying the
field of personality because it has confined itself to analysis
of variations as they exist in a typical mixed population at a
given moment. This statement applies to Q-technique as
well as R-technique, for though the former has certain
advantages in exploring personality aspects otherwise difficult
to approach, it yields, as Burt has shown (7), the same factors
as R-technique on similar populations.
584 RAYMOND B. CATTELL
An important new source of variations for exploring trait
unities lies in the increments obtained in a population with
age, with training or cultural influences with physiological
influences and with experiences affecting personality. If one
employs this method as a check on the ordinary factor
analysis, the emergence of the same factors would indicate
that one is dealing with real functional entities instead of mere
mathematical conveniences. Thus if, for example, the 'c'
factor of surgency (extraversion core) is found by static
factor analysis to saturate especially such traits as sociability,
quickness of apprehension, originality, informality and cheerfulness,
we should expect the same factor to emerge and
saturate, proportionately, the same traits in an analysis of
the increments of these traits occurring in a given group of
individuals with increasing age. That is to say, not only
should the more sociable individual be more quick, but, if
surgency is to be considered a real, psychological, functional
entity, we should find that the individual grows more quick
as he grows more sociable.
Other studies which do not depend on existing individual
differences in a group would be found in measurements of
fluctuation of traits (not steady increments with age or other
influences but short term variations), for we should expect
the elements of a true functional unity also to fluctuate
together.10
Again analyses could be made of the correlations
of the differences of twins reared apart, a procedure which
might be expected to eliminate the correlation clusters due to
constitutional traits leaving only those due to dynamic and
social mould traits. Incremental, fluctuation and hereditary
relation factor analyses might be grouped apart, under the
general label of differential factor analyses to distinguish them
from those static factor analyses, on homogeneous or nonhomogeneous
populations, in which the existing differences
in a group of individuals at a given moment are taken as the
basis of correlation. Our contention is that personality study
requires the extension of the present static factor analysis
10
This is the method, though without factor analysis, already implied by, for
example, the study of cyclothyme traits by Johnson (17).
THE DESCRIPTION OF PERSONALITY. I 585
researches by the inclusion of differential factor analyses, and
that the only factors which can be regarded as corresponding
to real trait unities are those which emerge from both
approaches.
Even with this principle in mind, however, the research
worker has still the task of extracting, from any given set of
correlations, clusters and factors, the psychologically meaningful
unitary traits which may lie behind them. How this
task may be accomplished is best seen by imagining unitary
traits of the types set out above and deducing the structure
of variables which would follow from them. By this approach
from within we may see how the researcher will need to
proceed in the reverse process of trying to arrive at the traits
from the correlations.
In Fig. 3 the eight lines beneath the letters represent a set
of trait elements—fragments from the total surfaceof personality—chosen
for rating and inter-correlation in a typical
sample of the population. Let us suppose that there were
SOCIAL AND PHYSICAL PRESSURES
AND DEMANDS
HONESTY PATTERN COURAGE PATTERN
ENERGY ESCAPE
OF CONSTITUTION ERG
CONSTITUTIONAL PATTERNS
IN THE ORGANISM
FIG. 3. Definition of trait elements.
A'. Respecting property of friends. B"'. Avoiding exaggeration in relating
A". Respecting property of strangers. stories.
B'. Being truthful to friends. C. Facing physical dangers.
B". Being truthful to strangers. D'. Keeping one's head in emergencies.
D". Magnitude of vegetative n. a. response-in
fear.
586 RAYMOND B. CATTELL
originally, in the mind of the investigators, four traits—
A, B, C and D—each constituting some logical, traditional,
stylistic or semantic unity, and that they were broken down
into narrower, behaviorally-defined traits—A', A", B', B",
B"', etc.—as a precautionary step, to avoid the trap of false
presuppositions from verbal unities.
Trait elements which vary together, in magnitude of and
direction of variance, will be drawn, by convention, at the
same level on this diagram. Now the forces which act upon
these behavior elements and which cause them to vary and
covary are of two kinds. On the one hand there are patterns
of external stimulation and inhibition, corresponding to the
social and environmental mould or die as already described.
By reason of the slight or deep impress of any particular die,
i.e., of any group of environmental forces which, for social or
other reasons happen to operate together, from a single focus,
the individual will have a high or low development of the set
of trait elements constituting that pattern, i.e., they will
covary within the population. On the other hand there are
patterns arisingfrom dynamic and non-dynamic constitutional
unities in the organism and these also will tend to produce
covariation in the elements which belong to them. The
problem is to see what covariations of the elements will result
from these super-posed patterns of covariance.
Emergence of covariance patterns (illustrated by imaginary
example}.—It is obvious, as illustrated by the example, that
neither the internal nor the external patterns of covariance
will succeed in impressing themselves on the superficial,
observed variability of the trait elements. The highest
correlations, the manifest correlation clusters, will occur where
a group of trait elements share both common constitutional
elements and common environmental mould influences, as in
C and D', or A' and B'. Trait A", which is part of the same
honesty trait and trait B" which is part of the same dynamic
need (with regard to A' and B', respectively) do not show
more than moderate correlations with other items in the same
trait. The most negligible correlations are those among
elements in logical, stylistic, semantic unities (e.g., B' with
B").
THE DESCRIPTION OF PERSONALITY. I 587
Our conclusion therefore is that factors, which represent
sources of covariance, rather than high clusters, correspond to
traits in the above senses. This conclusion, incidentally,
does not deny the fact that there are great difficulties in the
way of choosing, from a great variety of factor sets arrived
at mathematically, the one set which alone represents the
psychological reality. That one should follow this difficult
and uncertain path of discovering factors, instead of accepting
the view that an empirically obvious correlation clusteru
constitutes a trait, requires some justification. The justification
is (i) That meaningful traits, in the sense of dynamic,
constitutional and social mould unities, correspond to factors
rather than clusters. (2) That factors may be expected to
recurr, in identifiable form, in a variety of populations,
circumstances and modes of factor analysis (e.g., in differential
as well as static correlation analyses). A constitutional
hyper-thyroid pattern, for example, may appear identifiably
in different cultural groups. A cluster, on the other hand,
may be a local artefact, corresponding to a chance overlap of
social mould and constitutional trait factors. (3) That the
factor traits have the value of being more widely useful in
calculation and prediction, for they correspond to the influences
that are functionally distinct in the growth of
personality. This is true even of social mould traits, which
have no functional unity in the individual, being only a
collection of correlating habits, but which represent the
influence of a single force in the environment. Even if this
force may never operate again,18
in that the connection of the
elements is purely historical, the factor constitutes a con-
u
It should be made clear, perhaps, that the term 'cluster' here is used in its
simple, obvious sense, not in the special technical sense employed by Tryon in his
technique,of 'cluster analysis' (31). The latter seems to be a novel way of arriving
at factors of a generalcharacter.
Moreover, it is necessary to remember that some clusters will be due to all the
component traits' having a high saturation with one factor. To say that clusters
are factor overlaps assumes that other things (factor saturations) are approximately
equal.11
This observation reminds one that the social mould patterns may be expected
to be far more complicatedthan the constitutional ones. This occursnot only because
of the great variety of social mould influences and their changes with place and time,
but also because, as the individual grows older, they operate differently or vanish
entirely. His nature is, in respect to these patterns, a geological deposit.
588 RAYMOND B. CATTELL
venient measuring rod, for it has once operated with a similar
pattern on all individuals, producing similar features.
The above discussion illustrates a certain danger in the
term 'operational unity,' which has sometimes been applied
to simple, empirical clusters of correlation coefficients, on the
grounds that the trait elements involved certainly do, operationally,
'go together.' Strictly, this term should be applied
just as much to factors. For what shall be called an operational
unity depends on the nature of the operations one has
in mind, and if they are widely conceived, to include all
varieties of psychological circumstance and situation, the
factor is the truer1S
operational unity.
To avoid undue complication of presentation the problem
was not raised, in relation to the diagram above, that the
ergic or constitutional roots of a given social mould trait
pattern might be different for different people. The mathematical
analysis takes care of this, presenting a composite,
average picture. It seems extremely likely (indeed certain if
one adopts a holistic psychology) that each constitutional
trait shows itself to some extent in every piece of behavior.
That isto say it will appear as a generalfactor. Its saturation
of any trait element will thus be a function of the average
extent to which, in the population, the particular ergic root
or constitutional tendency enters into the formation of the
trait.
Finally we have to note that all the above methodology
says nothing about the permanence of traits within the
individual. Except in factor analysis one is taking a picture
with a flashlight, discovering configurations that exist
at a given moment. (In most rating, as opposed to test,
studies, it would be a fairly long 'moment.') Thus one
might catch, in addition to what are usually called traits,
the patterns of states of maladjustment or even moods,
needs-in-action, and physiological transient states. These
patterns could be sorted out from the more permanent ones
by examining consistency coefficients over various intervals.
M
This does not deny that some clusters may correspond to traits. The highest
observable cluster may, even so, be an overlap of two such 'real' clusters.
THE DESCRIPTION OF PERSONALITY. I 589
What degree of permanence is required to constitute a trait
rather than a state is a relatively arbitrary issue.
VI. THE UTILITIES OF TRAITS, ATTITUDES, SENTIMENTS
AND INTERESTS, AS CONCEPTS
The present discussion on methodology cannot be seen in
the general perspective of personality research as a whole
without a brief clarification of the relation of traits to sentiments,
attitudes and interests. The present writer has
pointed out elsewhere (8) the confusion and loss brought
about by incoordinated terminology in this field, and has
suggested that the trend of discourse has been to crystallize
'attitude' to mean a finer ramificationof a sentiment (related
to it as a twig to a bough), so that a collection of attitudes
grow out of a sentiment, in its interactions with environment.
The sentiment, being deeper, expresses itself more often in
feelings and actions; the attitude in opinions. This relationship,
of course, dismisses the use of attitude to describe a
merely momentary mental stance, state or experience of
derived emotion, and retains the term strictly in the senseof
a neuro-psychic disposition, i,e., a mental structure, or, if
structural connotations are not desired, in the sense of a
potential and recurrent pattern of behavior.
It has, further, been suggested in the above systematization
that 'interest' best applies to the quantitative—as
contrasted to qualitative or directional—vector aspect of
metanergs (attitudes, sentiments) and ergs (primary drives).
An interest, however, may loosely be used to define also, in
terms of its goal object (direction or quality), any erg or
metanerg of which it is the quantitative aspect. Thus we
properly speak of amount of patriotic interest, but we loosely
speak of patriotic interests instead of sentiments.
What, then, is the relation of the term trait to these
terms? That depends upon the trait. Only dynamic and
social mould traits have any relation to sentiments, attitudes
and interest. Indeed, the second, metanergic form of dynamic
trait is nothing but a sentiment, being an organization
of drives about a real object. Social mould traits, on the
590 RAYMOND B. CATTELL
other hand, are collections, often dynamically unintegrated,
of sentiments, attitudes and habits.
Actual language usage, as already indicated, gives heavier
emphasis to the use of 'trait' with respect to constitutional
and ergic (dispositional type) dynamic traits. Such usage
implies that sentiments (metanergic dynamic traits) are more
specific and local than traits, as ^attitudes are more specific
than sentiments. The justification for this habit of language
lies in such observations as that of a recent research (2) to
the effect that individuals can undergo a complete cultural
transplantation, changing attitude and even sentiment attachments,
without manifesting any appreciable transformation
of what is generally spoken of as personality (traits).
But, as we have seen, all traits, even common traits, slowly
change.
Great as the need is for some general term to refer to
species of personality manifestations, the practice of restricting
'trait' to the deeper, more important, stable and
abstracted variables of personality has some justification,
in usage and meaning. For normally we infer a sentiment
from a collection of observed attitudes and we similarly
abstract an individual's trait rating from observations of the
character of his sentiments. Thus, for example, we judge
that a man has an assertive disposition because we are
presented with an accumulation of sentiments and attitudes
loaded with assertiveness. He may be very assertive in
some and submissive in others. We strike an average.
The abstraction which we perform in rating a trait is also
indicated by the fact that the object of an attitude, and to a
lesser extent of a sentiment or complex, can always be designated,
whereas the trait is an attitude to life generally.
(Where ergic dynamic traits are concerned it is the attitude
of stressing a particular biological goal.) This generalization
of the trait, in contrast to attitude, is only another way of
expressing the mathematical statement that it is a 'factor'
derived from a collection of particular responses intercorrelating;
or of saying that this conception of trait is of
common traits, whereas sentiments and attitudes are more
frequently unique 'traits.'
THE DESCRIPTION OF PERSONALITY. I 591
Nevertheless, it is not possible simply to equate the
antithesis 'common-vs.-unique' with 'constitutional and ergicvs.-metanergic,'
or with 'important, major trait-vs.-unimportant,
superficial trait.' For a very narrow, superficial
attitude, and one not even remotely related to constitution,
may yet be a thoroughly common trait. Again, a constitutional
trait (as produced by biological variation or mutation)
may be highly specific and unique. Similarly a unique trait
may be highly important, at least in relation to the individual
personality. In short, we have to admit three or four14
distinct characteristics of traits, even though they correlate
so that in general the more constitutional is also the more
common and the more important. In conclusion, therefore,
this semantic problem seems most reasonably solved by
applying the term trait generically to all manifestations of
personality. Where the dictionary or a factor analysis
indicates 'common traits' we are dealing with the more
substantial manifestations, whereas unique traits are more
likely to be found among such slighter manifestations as
attitudes and some sentiments.
The general question of the relationships between various
types of trait unity cannot be left without some brief, and
necessarily rather speculative discussion concerning the issue
of the relative 'importance' of common-vs.-unique traits and
of the three varieties of common traits. A language count
would, of course, vote that common traits are far more
important than unique traits, and the views of many psychologists
imply the same judgment. It is probably true
that the behavior of most individuals in a mass-educating
culture can be fairly completely predicted in terms of common
traits; but under more special circumstances the unique trait
can be the major factor in personality prediction. Again the
dictionary gives, among common traits, a seeming predominance
of dynamic terms, supporting the clinic in its emphasis
on the first of the three varieties of unity.
"Another characteristic difference, implied by the above discussions, is that
common traits can be measured in units derived from the dispersion of a population
(which, for lack of a better term, we will call 'normative' units) whereas unique
traits, having to be defined logically, can only be measured in'metric' units. Common
traits can be measured in both metric and normative units.
592 RAYMOND B. CATTELL
The relative importance and utility of the three varieties
of common traits could theoretically be expressed precisely
in terms of the proportion of the total variance (in a number
of representative fields of behavior) contributed by traits of
each type. From a glance at available factor analyses one
might hazard the guess that constitutional traits, e.g., surgency,
intelligence, general emotionality, take an appreciable
slice of the variance. Further, there is some clear evidence,
e.g., in the lower 'g' saturation of intelligence tests with
adults, and in the higher inter-correlations of forms of honest
behavior in older and mentally older subjects (15), that social
mould traits become more definite and account for more of
the variance as individuals become longer exposed to the
culture. Certain social mould traits, notably the 'w,' character
integration factor, obviously contribute a major amount
to the variance in very important fields; but in general one
may suspect that the utility of social mould traits is reduced
by the fact that they are specific to a culture—to some
extent even to the provinces of a culture—so that their
predictive value is apt to be local or temporary.
VII.
1. All traits are really unique, but in a population with
common racial and cultural backgrounds a majority are so
nearly common that they can be treated as common traits,
measurable on common axes.
2. It is contended, with Gordon Allport (i), that "it is
more important to discover intelligible traits than independent
ones," i.e., mathematically independent ones, for the former
have functional existence in the personality and society, and
can be more widely used in prediction. Such intelligible
unities seem to be of three kinds—dynamic, constitutional and
social mould. Co-nascent and logical trait unities also have
utility in special circumstances, but their present too facile
and frequent use in education and guidance seems mistaken.
3. These three kinds of traits manifest themselves as
mathematical factors (not necessarily,or even probably, of an
independent kind) in the factor analysis of trait element
THE DESCRIPTION OF PERSONALITY. I 593
inter-correlations. To discover them, however, it will be
necessary to collate a variety of static factor analyses with a
well chosen variety of differential factor analyses, thereby
evolving criteria for the rotation of axes distinct from the
• unpsychological methods—such as 'simple structure'—now
employed.
4. Dynamic traits alone may be supplementarily investigated,
both as unique and as common traits, by temporal
sequence studies which are longitudinal, intra-individual
methods.
5. Clusters (of highly positively inter-correlating trait
elements) are unlikely to be traits. Dynamic traits, one may
deduce, are likely to manifest themselves as general factors
with superimposed bi-polar factors. Constitutional traits
will appear as simple general factors, probably with a more
even saturation of behavior elements than is found for
dynamic general factors. Social mould traits are likely to
appear as much restricted group factors. Such considerations
contribute towards, but do not provide, a unique determination
of trait unities by factor analysis. The possibility of a
truly unique solution to a factor analysis, yielding the psychologically
real trait unities in personality, is discussed in a
later article (10).
6. All traits, being relations between a changingorganism
and a changing environment, are only temporary patterns.
The common traits, however, are likely to be at least as
stable as a culture pattern.
7. Common traits can be measured in either metric or
normative (population relative) units; but unique traits,
having to be denned by logical dimensions, can only be
expressed in metric units.
REFERENCES
i. ALLPOB.T, G. W. Personality, a psychological interpretation. New York: Henry
Holt, 1937.
2. , BRUNER, J. S., & JANDORF, E, M. Personality under social catastrophe:
ninety life histories of the Nazi revolution. Character W Pers., 1941, 10, l-zz.
3. , & ODBERT, H. S. Trait names: a psycho-lexical study. Psychol. Monogr,,
19367 4T, No. 211, pp. vii, 171.
4. AKGYAI, A, Foundationsfor a scienceof personality. New York: Commonwealth
Fund, 1941.
594 RAYMOND B. CATTELL
5. BALDWIN, A, L. The statistical analysis of the structure of a single personality.
Psychol. Butt., 1940, 37, 518-519.
6. BURT, C. The analysis oftemperament. Brit.J. med, Psychol., 1938,17,158-188.
7. . Thefactors of the mind. London: University of London Press, 1940.
8. CATTELL, R. B. Sentiment or attitude? The core of a terminological problem
in personality research. Character y Pen., 1940, g, 6-17.
9. . -The descriptionofpersonality. II. Basic traits and the personality sphere.
/, abnorm. & soc. Psychol., 1943, 38, 4, 414-434.
10. . The description of personality. III. Principles and findings in a factor
analysis of the personality sphere. (Publication to be announced.)
11. CHRISTENSON, J. A. The analysis of personality; a general study. /. gen,
Psychol., 1941, 25, 41-57-
12. DILTHEY, W. Weltanschauung, Philosophie und Religion in Darstellungen. Leipzig:
Barth, 1911.
13. DUDYCHA, G. D. An objective study of punctuality in relation to personality
and achievement. Arch. Psychol., 1936, No. 204.
14. GUILFORD, J. P. Unitary traits of personality and factor theory. Amer. J.
Psychol., 1936, 48, 673-680.
15. HARTSHORNE, H., MAY, M. A., & SHUTTLEWORTH, F. K. Studies in the organization
of character. New York: Macmillan, 1930.
16. HULL, C. L. The concept of the habit-family hierarchy and maze learning.
PSYCHOL. REV., 1934, 41, Part I, 33-52; Part II, 134-152.
17. JOHNSON, W. B. Euphoric and depressed moods in normal subjects, Part II.
Character & Pers., 1938, 6, 188-202.
18. KLAGES, L. The science of character. London: Allen & Unwin, 1929.
19. LEWIN, K. d dynamic theory of personality. New York: McGraw-Hill, 1935.
20. LORGE, I, Personality traits by fiat. /. educ. Psychol., 1935, 26, 273-278.
21. McDouoALL, W. Outline of psychology. New York: Scribner, 1923.
Z2, MURRAY, H. A., et al. Explorations in personality. New York: Oxford Univ.
Press, 1938.
23. SEWARD, G. A. The 'validation' of drives. PSYCHOL. REV., 1942, 49, 88-95.
24. SPEARMAN, C. The abilitiesof man. London: MacMillan, 1932.
25. SPRANGER, E. Types of men: the psychology and ethics of personality. Halle:
Niemeyer, 1938. ,
26. STERN, W. Die menschliche Personlichkeit. Leipzig: Barth, 1923.
27. SYMONDS, P. M. Diagnosingpersonality and conduct. New York: Century, 1931.
28. THORNDIKE, E. L. Educational psychology. New York: Columbia Univ. Press,
1910.
29. . Mental andsocial measurements. NewYork: Columbia Univ. Press, 1913.
30. THURSTONE, L. L. The vectors of mind. PSYCHOL. REV., 1934, 41, 1-32.
31. TRYON, R. C. Cluster analysis. Ann Arbor: Edwards, 1939.
32. WEISS, A. P. The bio-social standpoint in psychology. In Psychologies of 1930
(C. Murchison, ed.). Worcester: Clark Univ. Press, 1930.