Professional Documents
Culture Documents
4, 119-129 (1990)
ROBERT R. MCCRAE
Gerontology Research Center, National lnstitute on Aging,
National Institutes of Health, Baltimore, USA
Abstract
The jive-factor model of personality has repeatedly emerged from lexical studies of
natural languages. When adjective-based factor scales are correlated with otherpersona-
lity measures, the adequacy and comprehensiveness of thejive-factor model are demon-
strated at a broad level. However, English language adjectives do not necessarily capture
more subtle distinctions within the jive factors. In particular, of several facets of the
Openness factor, only Openness to Ideas and Values are well represented in single
terms. Openness to Fantasy, Aesthetics, Feelings, and Actions can be exprtwed in
phrases, sentences, and literary passages-as excerpts from Bunink 'Lika ' illustrate-
but not in single words. To maintain its relevance to personality psychology, the study
of personality language must continue to examine empirical links to other personality
systems and must move beyond the dictionary to analyses of natural language speech
and writing.
INTRODUCTION
Both the attraction that personality language studies have for some psychologists
and the distrust that they inspire in others are understandable. Aside from its intrinsic
interest to linguists and students of social perception, the attraction of this line
of research comes primarily from the promise of comprehensiveness that studies
of natural language hold. We have no definition of personality that would allow
us to identify a priori the range and scope of significant individual differences;
all we can do is note important traits and catalogue them, and then look for redun-
dancy and omissions in our lists. We might ask panels of experts to generate descrip-
tors, as Block (1961) did, or we might ask for spontaneous self-descriptions, as
John (1989) did. Natural language dictionaries offer a tempting alternative, because
lexicographers have already done most of the work of compiling items. If we are
All correspondence concerning this article should be addressed to: Robert R. McCrae, Gerontology
Research Center, National Institute on Aging, National Institute of Health, 4940 Eastern Avenue, Balti-
more, MD 21224, USA.
prepared to make the assumption that someone, somewhere in the course of centuries
of use by millions of speakers must surely have noted and codified every important
nuance of personality, then the dictionary provides an a posteriori definition of
personality. 1 will call this assumption, articulated by Norman (1963) and others,
the lexical hypothesis . Under this hypothesis, analyses of natural language dictionar-
ies promise to offer the basis for a comprehensive taxomony of personality traits
- surely an invaluable contribution to personality psychology.
There are two grounds for the distrust of the lexical hypothesis that many persona-
lity psychologists feel. The first is that there are many reasons for thinking that
natural language adjectives might not be an appropriate basis for scientific psy-
chology. Hofstee (1990) has listed several: the domain as defined by the dictionary
is ambiguous; individual terms are not necessarily translatable; natural language
is highly, perhaps excessively, evaluative;the internal structure of languages is unruly;
and terms are often inappropriate for first-person use. These obstacles are noted
by a proponent of personality language studies; opponents can point to several more.
Fundamentally, they are skeptical of the ability of laypersons to understand the
true basis of personality. Freudian psychologists, to take an extreme case, claim
that the true determinants of personality are unconscious and that the culture has
often conspired with its members to mask out the most central of human impulses.
Where are the natural language equivalents of Oedipal conflicts or anal fixations?
For the Freudian, the most important psychological characteristics are precisely
those that are unspoken. Psychologists from many different theoretical orientations
might disagree on the nature of the most important psychological constructs, but
they would share with Freudians the view that scientific theories often need to trans-
cend the implicit knowledge of the culture.
The second basis for distrust of the lexical hypothesis is quite different. It is that
there have been relatively few empirical efforts to relate lexical models to alternative
measures of personality. The five factors repeatedly found in natural languages have
generally been given names and interpretations that align them with familiar dimen-
sions such as Extraversion and Neuroticism, but until the 1980s there were virtually
no published attempts to confirm these interpretations empirically. I suspect this
occurred because students of language and personality generally take a rational
approach to personality assessment; if individuals say that they are sociable, talkative,
enthusiastic, and energetic , rationalists assume they must be extraverts.
Most other personality psychologists are not so easily persuaded. They have taken
seriously psychometricians’ criticisms of face validity and rational interpretation;
they have seen the chaos that can result when scales bearing the same label (e.g.
‘Type A’) are used interchangeably without evidence of empirical convergence. The
five-factor model was doomed to be a model of personality language, not personality,
until empirical links were made.
Empirical studies are important for another reason: they can point to convergences
and divergences that could not easily be predicted on rational grounds alone. Who
would have guessed that Norman’s Agreeableness factor would be strongly related
to the Jungian Thinking-Feeling function as assessed by the Myers-Briggs Type
Indicator (MBTI; McCrae and Costa, 1989a)?Isn’t it surprising that the factor often
labelled Intellect or Intelligence is quite modestly related to psychometric measures
of IQ (McCrae and Costa, 1985b)? Findings such as this underscore the importance
of studying the construct validity of adjective factors.
Openness to Experience 121
Until 1983, Paul Costa and I had been developing a three-factor model of personality,
measuring aspects of Neuroticism, Extraversion, and Openness to Experience. Gold-
berg’s 1981 and 1982 chapters excited our interest in linking our made1 to the five-
factor model, and he provided an instrument to measure the five factors as he con-
ceived them. Our studies were conducted on members of the Baltimore Longitudinal
Study of Aging (BLSA; Shock et al., 1984) and their spouses and peers. These
subjects are generally healthy, well-educated men and women who, in the past 30
years, have completed a wide variety of questionnaires and have been the targets
of a number of other assessments. Because of this, they formed an ideal sample
in which to examine the empirical correlates of the five-factor model.
Our first study (McCrae and Costa, 1985b) used self-reports on Goldberg’s 40
bipolar adjective scales, a supplementary set of 40 adjective scales we created, and
an early version of our personality questionnaire, the NEO Personality Inventory
(NEO-PI; Costa and McCrae, 1985). There were two major results from this study.
First, we were able to recover a version of the five-factor model in analyses of the
80 adjective scales, although we felt Goldberg’s Intellect factor was better interpreted
as Openness to Experience. Second, the Neuroticism, Extraversion, and Openness
factors were strongly related to our NEO-PI scales of the same names, suggesting
that questionnaire measures might be used to assess the five factors in more traditional
formats.
A follow-up study (McCrae and Costa, 1987) introduced new NEO-PI scales to
measure Agreeableness and Conscientiousness, and replicated the previous findings
on adjective factors using peer ratings instead of self-reports. We also showed cross-
observer and cross-instrument agreement, and in subsequent work we have used
the NEO-PI as our primary measure of the five factors.
Much of our work since has been devoted to an examination of the comprehensive-
ness of the five-factor model in accounting for dimensions of personality measured
by other instruments, including the MMPI, MBTI, Eysenck Personality Inventory,
Personality Research Form (PRF), Guilford-Zimmerman Temperament Survey,
Wiggins’ Interpersonal Adjective Scales, and Block’s California Q-Set (Costa and
McCrae, 1988a; Costa, Busch, Zonderman and McCrae, 1986; McCrae, 1989;
McCrae and Costa, 1985a).
Recently, John (1989) has offered definitions of the Big Five factors using items
from Gough and Heilbrun’s (1983) Adjective Check List (ACL). He and nine students
familiar with the literature on the five-factor model sorted each of the 300 ACL
items into one of the Big Five domains, or discarded it. Agreement of 90-100
per cent was found for the classification of about 40 per cent of the items; these
terms can be seen as common conceptions of the five factors, based on rational
interpretation of the lexical literature. The ACL has limitations both as a pool of
items and as an assessment technique, and John’s intent was primarily to describe
the factors in the familiar language of ACL adjectives. Here I examine these concep-
tions empirically.
The ACL was completed by 445 BLSA participants; of these, 410 had also taken
the NEO-PI, and smaller numbers had been rated by groups of peers or by spouses.
In its standard format, the ACL is seriously distorted by endorsement frequency
122 R. R. McCrae
because some subjectscheck many, some few items. Normally, scores are standardized
within groups differing by gender and total number of endorsements. For this analy-
sis, I created scores on John’s scales by summing the appropriate items; I then residua-
lized them by partialling out gender and the total number of items endorsed. Table
1 shows correlations between the residualized scores and NEO-PI self-reports and
ratings.
I have reordered John’s factors to conform to the NEO-PI convention and reflected
his Factor IV to align the two sets of measures. Convergent correlations in self-reports
range from 0.35 for Openness/Culture to 0.64 for Neuroticism versus Emotional
Stability. Somewhat smaller, but significant correlations are also seen when self-
reports on ACL scales are correlated with peer and spouse ratings. These correlations
are remarkably large, given that they are based on different observers using a different
instrument separated by intervals of from 1 to 4 years.
Table 1. Correlationsof John’s (1989) ACL Big Five scales with NEO Personality Inventory
factors
ACL scale
NEO-PI factor -IV I V I1 I11
Self-reports
Neuroticism 6 4 -12* - 18* -24$ -231
Extraversion -08 5% 20$ 17$ 07
Openness 06 12* 35* -06 -17*
Agreeableness - 14t - 19$ 01 43$ 04
Conscientiousness - 17$ 03 16t 13* 50$
Mean peer ratings
Neuroticism 36$ 00 -02 -02 -11
Extraversion -06 a$ 14 18* -11
Openness 01 14 37$ 00 - 14
Agreeableness -11 -22t - 10 34.4 -09
Conscientiousness -03 09 00 00 33$
Spouse ratings
Neuroticism 33 -03 - 16 -22* -25*
Extraversion -09 42$ -01 15 02
Openness -11 -07 21* 01 -01
Agreeableness - 18 -07 05 49$ -02
Conscientiousness -09 02 -07 19 38t
Note: Ns = 410 for self-reports, 159 for peer ratings, and 100 for spouse ratings. Convergent correlations
are given in bold face; decimal points are omitted.
* p < 0.05; t p < 0.01; Sp < 0.001.
The lowest convergent correlations in Table 1 are between Openness and Factor
V. I believe that the problem lies with the Factor V scale: in this case, the lexical
tradition has failed to portray adequately one of the major dimensions of personality,
confounding intelligence, education and sophistication with a more basic personality
dimension that we call Openness to Experience.
One way to test this hypothesis is by factoring the items identified by John to
see if his Factor V is empirically recovered. Item factor analyses are problematic
with the ACL because of individual differencesin endorsement frequency and extreme
item splits, but it is possible to control these problems in part by factoring a partial
Openness to Experience 123
We believe that the fifth factor identified in lexical studies is best interpreted as
a variant of the more psychologically fundamental dimension of Openness to Ex-
perience. The NEO-PI measures six facets of Openness, in the areas of Fantasy,
Aesthetics, Feelings, Actions, Ideas and Values. We conceive of the open individual
as being interested in experience for its own sake, eager for variety, tolerant of
uncertainty, leading a richer, more complex, less conventional life. By contrast, the
closed person is seen as being impoverished in fantasy, insensitive to art and beauty,
restricted in affect, behaviourally rigid, bored by ideas, and ideologically dogmatic
(McCrae and Costa, 1985a, in press).
Our measure of Openness is correlated with Tellegen and Atkinson’s (1974)
Absorption; the MBTI Intuition scale; Zuckerman’s Sensation Seeking (especially
the Experience Seeking subscale);PRF needs for Change, Sentience and Understand-
ing; Holland’s Artistic and Investigative vocational interests; and Gough and Heil-
brun’s (1983) Creative Personality Scale. It has been related to such diverse criteria
as ego level, moral reasoning, divergent thinking scores and mid-career shifts.
The dimension of Openness thus appears to knit together a wide variety of traits
and topics of interest to personality psychologists-as any fundamental dimension
of personality should. This view of Openness stands in rather stark contrast to that
to which one would be led by a reading of the lexical literature. In most studies,
the closest approximation is some variant of Culture, Intellect or Intelligence-a
small factor with mixed loadings (e.g. Peabody, 1987).
There are, of course, many words in English that denote aspects of Openness.
But a conceptual analysis of adjective factors suggests that the usual Factor V only
points to the dimension of Openness-it does not register all the significant ways
in which individuals differ in regard to Openness. We are thus forced to adopt a
weak form of the lexical hypothesis and abandon the strong form which asserts
a rigorous parallelism between the structure of language and the structure of persona-
lity.
Some of the problems with the fifth adjective factor can be seen in the adjectives
themselves. Several adjective measures include wide interests or broad interests, and
breadth of interests is undoubtedly an attribute of open people. It is not, however,
an adjective, nor can I think of any single term that captures this basic trait. Prefers
124 R. R. McCrae
variety is another phrase we adopted (McCrae and Costa, 1985b) for want of a
single adjective. Many relevant terms, like closed-minded and down-to-earth, are
hyphenated, as if the language itself had recognized the deficiency of Openness terms
and begun to assemble new ones.
When examined by facet, it appears that English has many words that express
Opennessto Ideas, but fewer that capture other facets of Openness. There is a vocabul-
ary of political and social values, of course, but these terms have political and social
implications as well as personological-so much so that some have suggested that
they fall outside the domain of personality altogether.
Openness to Fantasy, Aesthetics and Feelings are peculiarly underrepresented in
English adjectives. Artistic can mean susceptible to aesthetic experiences, but more
commonly it means artistically talented. We have other terms like originaland creative
to signify artistic talent, but no single words to express the receptivity to aesthetic
impressions. In the case of Openness to Feelings, the problem is somewhat different.
There are of course many adjectives to express emotionality, both positive and nega-
tive, but positive and negative affects are primarily signs of Extraversion and Neuroti-
cism. Openness to Feelings is seen in highly varied and differentiated emotional
reactions, in affective responsiveness to the environment and to one’s own internal
feelings (Rogers, 1961). The word sensitive can take this meaning, but frequently
it means touchy or defensive-aspects of Neuroticism.
within myself for something that was on the point of becoming defined,
of taking shape . . .. My thoughts became confused, my mind again tor-
menting me with its spontaneity, its desultoriness, thronged again with
the most diverse feelings, thoughts and fancies. (p. 190).
Complexity, ambivalence, even inconsistency are frequently seen in the narrator’s
emotional reactions. In a boarding house in which he stayed he was surprised to
find hanging over the sofa a portrait of the deceased owner in his coffin, and he
was struck by the tone of the place and the mundane activities of the other boarders:
And in everything-the banal song, the rhythmical homely tapping, the
old, cheap print, and even in the dead man whose life seemed to continue
in the senselessly happy world of the boarding house, there was a sadness,
delicious and bitter. . . (p. 158).
What adjective expresses the ability to perceive senseless happiness as a bitter and
delicious sadness?!
Alexei Arsenyev’s role as an aspiring artist sometimes conflicted with his sense
of beauty:
From the mail I would pick out the latest of periodicals . . .. A new Chek-
hov story! The sight of that name alone affected me so that I could not
even read the opening lines-the ecstasy I could foresee blinded me with
such poignant envy! (p. 182).
We say that open people have an appreciation for art, a particular susceptibility
to aesthetic experiences. Sometimes the result is highly emotional:
Music was one of my most complex and tormenting delights. I adored
her when she played something beautiful. My heart ached with a raptur-
ously self-sacrificing tenderness for her. I wanted to live on and on. (pp.
166-167).
Sometimes there is greater detachment, a disinterested absorption in the experience:
There’s no country in the world more beautiful than the Ukraine [he
tells Lika]. And the main thing is it no longer has any history, it ended
its history forever, long ago. There’s only the past, the songs and legends
of the old days-a sort of timelessness. This fascinates me more than
anything else.
You use the word “fascinate” very often, don’t you? [she remarks].
But life should be fascinating.. . (p. 217).
CONCLUSIONS
within the Big Five will be revealed by further analyses of adjectives. I have argued
that Openness to Experience is inadequately represented in English language trait
terms. I cannot claim that the six facets of Openness measured by the NEO-PI
exhaustively assess the domain. But I think the NEO-PI facets of Openness to Fan-
tasy, Aesthetics, Feelings, Actions, Ideas and Values come closer to depicting the
full domain than any list of adjectives would.
I would make two suggestionsfor further research by those interested in personality
and language. First, I think that more attention must be paid to empirical studies
linking adjective factors to psychological constructs drawn from other traditions.
Gough (1987), for example, has used adjective descriptors to help explain the psycho-
logical meaning of his scales; the converse process can also be illuminating to students
of personality language.
Second, I think that the future may lie in the analysis of phrases, sentences and
paragraphs rather than single terms. The vast range of this material makes systematic
analyses difficult or impossible, but such studies can be used to seek new aspects
of personality and thus extend the provisional taxonomy offered by lexical studies.
Several approaches might be used. Asch’s (1946) classic attempt to understand impres-
sions based on a configuration of traits has recently been updated by Hampson’s
(1990) analysis of inconsistent pairs of traits. McGuire’s (1984) examination of the
spontaneous self-concept, De Boeck and Mattheus’ (1989) analyses of everyday
speech, and McAdams (1988) studies of life narratives may provide some leads.
It would be possible, and perhaps profitable, to abstract personality descriptions
from the writings of Shakespeare, Stendhal or Conrad to see if they suggest new
personality traits. Those of us who respect the power and subtlety of natural language
must be willing to follow it beyond the confines of the dictionary into its natural
habitats in speech and literature.
REFERENCES
Asch, S. E. (1946). ‘Forming impressions of personality’, Journal of Abnormal and Social
Psychology, 41: 258-290.
Block, J. (1961). The Q-sort Method in Personality Assessment and Psychiatric Research,
Charles C. Thomas, Springfield, IL.
Bunin, I. (1963). ‘Lika’. ‘The Gentleman from Sun Francisco’ and Other Stories (0.Shartse,
Trans.), pp. 139-249, Washington Square, New York. (Original work written in 1933.)
Costa, P. T., Jr. and McCrae, R. R. (1985). The NEO Personality Inventory Manual, Psycho-
logical Assessment Resources, Odessa, FL.
Costa, P. T., Jr. and McCrae, R. R. (1988a). ‘From catalog to classification: Murray’s needs
and the five-factor model’, Journal of Personality and Social Psychology, 55: 258-265.
Costa, P. T., Jr., Busch, C. M., Zonderman, A. B. and McCrae, R. R. (1986). ‘Correlations
of MMPI factor scales with measures of the five-factor model of personality’, Journal
of Personality Assessment, 50: 640-650.
De Boeck, P. and Mattheus, R. (1989). ‘Categories of everyday third-person talk’, Paper
presented at the Invited Workshop on Personality Language, University of Groningen,
The Netherlands.
Goldberg, L. R. (1981). ‘Language and individual differences: the search for universals in
personality lexicons’. In: Wheeler, L. (Ed.), Review of Personality and Social Psychology,
Vol. 2, pp. 141-165, Sage, Beverly Hills, CA.
Goldberg, L. R. (1982). ‘From Ace to Zombie: some explorations in the language of persona-
lity’. In: Spielberger, C. D. and Butcher, J. N. (Eds), Advances in Personality Assessment,
Vol. 1, pp. 203-234, Erlbaum, Hillsdale, NJ.
128 R. R. McCrae
RESUMI?
Le modtle de la personnalitk a cinq facteurs est apparu a maintes reprises dans des Ctudes
lexicales de langues naturelles. Quand les Cchelles de facteur basCes sur des adjectifs sont
corrClCes avec d’autres mesures de la personnalitk, l’adtquation et l’ampleur du modkle i
Openness to Experience 129
cinque facteurs y sont largement demontrees. Dans la langue anglaise toutefois, les adjectifs
ne traduisent pas necessairement les difftrences subtiles a l’interieur des cinq facteurs. Ceci
vaut en particulier pour les differentes facettes du facteur Franchise. Seules Franchise- au
niveau des IdCes et des Valeurs peuvent Ctre bien representees par des termes simples. Etre
franc pour les Fantasmes, I’EsthCtique, les Sentiments et les Actions peut Ctre traduit par
des expressions, des phrases et des passages litteraires-comme il en ressort des citations
de ‘Lika’ de Bunin-t non par des termes simples. L’ttude du language de la personnalite,
afin de conserver sa pertinance pour la psychologie de la personnalitt, doit continuer d’en
rechercher les liens empiriques avec les autres systkmes de la personnalite. Le langage nature1
parle et ecrit doit par la-mCme Ctre &die en complement des analyses du dictionnaire.
ZUSAMMENFASSUNG
Das Funf-Faktoren Model1 der Personlichkeit hat sich wiederholt aus lexikalischen Studien
naturlicher Sprachen ergeben. Wenn auf Adjektiven beruhende Faktorenskalen mit anderen
PersonlichkeitsmaRen korreliert werden, zeigt sich die Angemessenheit und das Umfassende
des Fiinf-Faktoren Modells auf einer breiten Ebene. Jedoch erfassen englischsprachige Adjek-
tive nicht unbedingt subtilere Unterscheidungen innerhalb der funf Faktoren. Insbesondere
sind von den verschiedenen Facetten des Offenheits-Faktors nur die Ideen: und die Werte-
Subfacette gut durch einzelne Begriffe reprasentiert. Offenheit fur Phantasie, Asthetik, Gefiihle
und Handlungen konnen durch Phrasen, Satze und literarische Textstellen, nicht jedoch durch
einzelne Worte ausgedruckt werden, wie Exzerpte aus Bunin’s ‘Lika’ illustrieren. Um ihre
Relevanz fur die Personlichkeitspsychologie zu erhalten, muR die Studie der Personlichkeits-
sprache fortfahren, empirische Beziehungen zu anderen Personlichkeitssystemen zu unter-
suchen. Sie muR vom Lexikon fortschreiten zur Analyse des natiirlichen Sprechens und
Schreibens.