You are on page 1of 12

Pergamon

Person. in&d. Diff: Vol. 21. No. 3. pp. 391402, 1996


Copyright 0 1996 Elsevier Science Ltd
Printed in Great Britain. All rights reserved
0191~8869/96 $15.00+0.00
SOl91-8869(96)00046-3

THE SIX-FACTOR MODEL OF PERSONALITY:


FACETS FROM THE BIG FIVE
Douglas N. Jackson, Michael C. Ashton and Jennifer L. Tomes
Departmentof Psychology. The University of Western Ontario, London, Ontario N6A X2.

Canada

(Received 4 September 1995)


Summary-One
hundred and forty-one respondents completed two scales designed to measure desirability
responding,
in addition
to the NE0 Personality
Inventory-Revised
(NEO-PI-R),
and the Six-Factor
Personality Questionnaire
(6FPQ), a new personality questionnaire
designed to appraise six broad factors
of personality
with 18 facet scales. Scores were subjected to principal components
factor analysis and
rotated to a seven-factor
solution targeting six content factors and a desirability factor. Coherent content
factors, corresponding
to four of the big five personality factors, emerged : Extraversion,
Agreeableness,
Independence
(low Neuroticism)
and Openness to Experience. Two additional content factors were identified as Methodicalness
and Industriousness,
which we interpreted as distinct aspects of Conscientiousness.
A seventh factor, Desirability,
also emerged, defined by the desirability marker scales and several NEOPI-R scales. It was concluded
that the six identified content factors could be well represented
by the
appropriate
6FPQ scales, as well as by certain scales from the NEO-PI-R.
Copyright
c 1996 Elsevier
Science Ltd.

THE NEO-PI-R

AND THE SIX-FACTOR

MODEL OF PERSONALITY

There has been a remarkably extensive literature reporting research supporting a five-factor model
of personality. Originally identified in ratings of others personality (Norman, 1963; Tupes &
Christal, 1961) using simple trait names or phrases, a number of investigators (e.g. Costa & McCrae.
1988; Digman, 1990) have maintained that the five-factor model applies to self-attributions in
personality questionnaires, but this issue remains controversial from a number of perspectives.
Block (1995) has argued that factors have emerged consistently because investigators chose variables
to elicit them, that the number of factors retained for interpretation is arbitrary in any particular
study, and that, while the number of factors has remained at five, their interpretation has been
inconsistent across different investigators. Paunonen (1993, August) has suggested that early studies
identifying the big five were based on peer judgments which are known to be less differentiated
than other methods of measurement, and that five is therefore a lower bound of the number of
general personality dimensions.
Our purpose in the present study is not to evaluate the appropriate number of factors in the
questionnaire domain. As Block (1995) pointed out, this has an arbitrary quality to it. Rather, our
intent is to evaluate the degree to which a well-known measure of the big five personality factors.
the NEO-PI-R, is represented in a dimensional space defined by six replicated factors derived from
the Personality Research Form (PRF; Jackson, 1984).
Costa and McCrae (1988) set out to classify the system of needs proposed by Murray (1938) in
terms of the five-factor model of personality. This was undertaken by performing a joint principal
components analysis of the NE0 Personality Inventory (NEO-PI; Costa & McCrae, 1985), which
was designed to measure the big five, and the PRF, which measures 20 variables of personality,
the large majority of which had their origin in Murrys (1938) classification. Actually. Costa and
McCrae (1988) reported that seven factors had eigenvalues greater than 1.O (p. 263), but chose
to examine a varimax-rotated five-factor solution. This was consistent with the purpose of their
study, which was not to define the number of factors underlying Murrays (1938) needs, but to
examine the correspondences between the PRF and NEO-PI dimensions within a common five
factor space (p. 261). In that five-factor space, all PRF scales showed salient loadings on at least
one factor, and these loadings generally made sense from a substantive viewpoint.
Jackson, Paunonen, Fraboni and Goffin (1996) undertook a set of three studies which provided
a different perspective on the factor structure of the PRF than that of Costa and McCrae (1988).
391

392

Douglas N. Jackson et al.

In their first study, Jackson et al. (1996) used confirmatory factor analyses of two separate PRF
data sets (comprising 306 and 2141 Ss, respectively) to compare a five-factor model of the PRF
based on Costa and McCraes (1988) results with a six-factor model based on rational considerations
and the published results of factor analyses of other PRF data sets. In the six-factor model, the
Conscientiousness factor of Costa and McCrae (1988) was divided into two factors, with the PRF
Achievement, Endurance, and (negative) Play scales defining one of these factors and the Cognitive
Structure, (negative) Impulsivity, and Order scales defining another. The results for each of the
samples demonstrated that a six-factor model provided a significantly better fit to the data than did
the five-factor model.
The second study reported in Jackson et al. (1996) again used confirmatory factor analysis, this
time to compare the five- and six-factor models with respect to a new, briefer measure of the sixfactor model, whose development is described in the Method section, later. The results, computed
both for the 306-S sample described above and also for a new sample of 113 individuals, again
showed that the six-factor model provided a significantly better fit to the data than did the fivefactor model.
In the third study described by Jackson et al. (1996), evidence of convergent and discriminant
validity for this new measure, named the Six-Factor Personality Questionnaire (6FPQ; Jackson &
Paunonen, in press) was evaluated with a multitrait-multimethod
matrix using a peer-rating
criterion. All convergent validities were higher than any heteromethod-heterotrait
correlations. The
average convergent validity was 0.56, and the average absolute correlation with an irrelevant peer
rating was 0.11. These results provided encouragement for proceeding with further evaluations of
the 6FPQ.
Interestingly, the six-factor model reported by Jackson et al. (1996) seems to have emerged in
Costa and McCraes (1988) study. One of the two discarded factors reported by Costa and McCrae
was defined by the PRF scales for Achievement and Endurance (and also by Abasement), whereas
scales for Cognitive Structure, Impulsivity and Order loaded on a different factor.
Thus, the first purpose of this study is, in a sense, parallel to that of Costa and McCrae (1988).
Its aim is to locate the 30 facet scales of the NE0 Personality Inventory-Revised (NEO-PI-R; Costa
& McCrae, 1992) within the factor space of the six-factor model of personality, as proposed by
Jackson et al. (1996). Thereby, this study should also help to determine the substantive coherence
of the patterns of scales that define those factors, and to provide a further evaluation of the sixfactor model. As was true of Costa and McCraes (1988) analysis, the aim of this study is not to
discover the true number of factors in the personality domain, but to compare and evaluate the
relationship of one interpretation of a personality taxonomy with another.
It is also hoped that this study will provide some insight into the role of the social desirability
response style in the NEO-PI-R. In contrast to the approach taken in the construction of the PRF
and the 6FPQ, the social desirability of items was not a criterion for item selection in the
development of the NEO-PI (Costa & McCrae, 1988, p. 259). The rationale for not dealing with
desirability response bias was the belief that inherently desirable traits would be difficult to measure
if socially desirable items were removed.
The position that desirability responding should not be a consideration in personality scale
development is controversial. There is a potentially serious problem in confounding distinct constructs of content and desirability in a single measure: the presence of response style or common
method variance will sharply limit convergent and discriminant validity in comparison with sets of
measures that are more mutually independent. Put another way, one can have more confidence in
the construct validity and interpretability of scores that are not saturated with a construct-irrelevant
source of variance, especially when the construct-irrelevant source of variance is a general factor.
In the limiting case where all measures predominantly measure this general factor to the exclusion
of unique content variance, one would not expect evidence for convergent and discriminant validity
with respect to the same traits measured by distinct methods. The fact that it is not possible to
eliminate desirability variance completely in personality measures is not a sufficient reason for not
seeking to reduce its impact.
To support the claim that the social desirability of items cannot be separated from their content,
Costa and McCrae (1988) tested the following hypothesis: If the PRF Desirability scale is chiefly
a measure of a response set from which the other PRF scales were designed to be free, factor analysis

Six-factor model of personality

393

should show it loading on a second factor, perhaps along with NEO-PI variables. If, however,
responses to the Desirability scale are chiefly determined by item content, it should load on one or
more substantive factors defined by both PRF and NEO-PI variables. (pp. 259-260).
Unfortunately, the Costa and McCrae study only included one desirability scale, insufficient to
define a desirability factor unequivocally in factor analysis. The present study has included two
desirability scales, which, together with other personality scales drawn from each of the content
factors, provide a better opportunity to define a desirability factor that is conceptually and empirically distinct from any particular domain of content. Additional to this problem is the fact that the
variance attributable to a socially desirable response style tends to accumulate in the first unrotated
factor, but then to distribute itself among several factors upon rotation. Therefore, in order to
obtain a well-defined desirability factor in a rotated solution, it is often necessary to use targeted
rotations.
Accordingly, it was decided to rotate the factor matrix to an orthogonal seven-factor solution,
with six factors targeted for their expected 6FPQ and NEO-PI-R scales and a seventh targeted for
the desirability factor produced by targetings for the two desirability scales and for the one scale
within each NEO-PI-R and each 6FPQ factor which correlates most strongly with desirability.

METHOD
Respondents

Respondents were 144 undergraduate university students, 80 males and 64 females, who participated in this study as part of their introductory psychology course requirements. Responses for
three Ss with incomplete answer sheets were not included. Ages ranged from 17 to 25 yr; the median
age of all Ss was 19 yr.
Tests
Six-Factor Personality Questionnaire. The Six-Factor Personality Questionnaire (6FPQ; Jackson
& Paunonen, in press) was designed as a measure of the six-factor model found by Jackson et al.
(1996). The 6FPQ is a 108-item inventory that employs a five-point response format, and reports
scores for 18 facet scales and six factor scales, where each factor scale is the sum of three facet
scales. All but one of the 18 facets are measured by the PRF; the other facet was derived from the
Jackson Personality Inventory-Review (JPI-R; Jackson, 1994). The six items in each 6FPQ scale
were selected from the corresponding PRF or JPI-R trait scales. Items possessing optimal statistical
properties, including high item-total correlations, discrimination between scales, high item variance
and correlations with peer-ratings for the relevant facet, were chosen for inclusion in the 6FPQ,
with the constraint that three of the six items for each facet scale would be negatively-keyed.
The 6FPQ factor and facet scales are as follows (factor scales are italicized): E.ytraversion = Affiliation + Dominance + Exhibition; Agreeableness = Abasement + Even-tempered (low
Aggression) + Good-natured
(low
Defendence);
Methodicalness = Cognitive
Structure +
Deliberateness (low Impulsivity) + Order; Independence = Autonomy + Individualism (low Social
Recognition) + Self-reliance (low Succorance); Openness to Experience = Change + Understanding
+ Breadth of Interest; Industriousness = Achievement + Endurance + Seriousness (low Play).
The 6FPQ factors Extraversion, Agreeableness and Openness to Experience were expected to
correspond to the big five factors of the same names. Independence was expected to correspond
to low Neuroticism, and Methodicalness and Industriousness were considered likely to represent
different aspects of Conscientiousness.
NE0 Personality Inventory-Revised. The NE0 Personality Inventory-Revised (NEO-PI-R; Costa
& McCrae, 1992) was constructed to measure the big five personality factors, and several traits
within each factor. This 240-item questionnaire reports scores for five factors and for 30 facet scales.
Each factor represents the sum of six facet scales.
Desirability scales. Thirty-two Desirability items were chosen from the Desirability scales of forms
AA and BB of the PRF and converted to a five-point response format. Scores are reported for the
first 16 items (Desirability I) and for the last 16 (Desirability II); each half contained eight positivelykeyed and eight negatively-keyed items. These scales were intended to measure the social desirability

Douglas N. Jackson et al.

394

response style with content deliberately chosen to be heterogeneous. In previous research these
items have been shown to load on a factor distinct from PRF scales (Jackson & Lay, 1968; Jackson
& Morf, 1973; Morf & Jackson, 1972).
Method

of analysis

For each S, scores were calculated for the 18 6FPQ facet scales, the 30 NEO-PI-R facet scales,
and for the two desirability scales. Correlations between all scales were then calculated. This
correlation matrix is available from the authors upon request.
A principal components analysis was performed on the correlation matrix, with orthogonal
rotation to a targeted criterion (Schbnemann, 1966). Six of the targeted factors were designed to
correspond to the factors measured by the 6FPQ. Scales targeted for the Extraversion, Agreeableness
and Openness to Experience included all 6FPQ and NE0 trait scales keyed for each of these
respective factors. Scales targeted for the Independence/low Neuroticism factors included all 6FPQ
Independence and NE0 Neuroticism scales; the Neuroticism scales were given negative target
loadings. The targeted factor for Methodicalness comprised all 6FPQ Methodicalness facets and all
NE0 Conscientiousness facets; similarly, the targeted factor for Industriousness comprised all 6FPQ
Industriousness facets and all NE0 Conscientiousness facets.
The seventh factor was targeted using the two desirability scales and, in order better to define a
coherent factor, also using the one facet scale of each NEO-PI-R and 6FPQ factor which correlated
most strongly with the mean of the desirability scales. The only exception was that two facet scales
from NEO-PI-R Conscientiousness were targeted, so that targetings would correspond to those of
the 6FPQ, which treats Industriousness and Methodicalness as distinct aspects of Conscientiousness.
The facet scales targeted upon the Desirability factor were also targeted upon their respective
content factors.
Because targeted rotation has been criticized for the possibility that it may capitalize on chance
(Horn, 1967; Humphries, I&en, McGrath & Montanelli, 1969) we employed a technique proposed
by Paunonen (in press) to rotate our seven hypothesized factors to randomly-designated targets.
Within each factor the scales targeted to have high values were randomly permuted 1000 times.
These targeted values then served as the target for 1000 Procrustes rotations. Means and standard
deviations of congruence coefficients between the resulting factor loadings and the randomlydesignated targets were calculated, and then compared with congruence coefficients derived from
our analysis. This permitted a comparison of our Procrustes rotation based on our hypothesized
solution with the results from Procrustes rotations based on 1000 random targets. If the levels of
our hypothesized factor loadings fell within the range of the results from the random targets, then
it could be concluded that they were largely attributable to capitalization on chance.

RESULTS
The loadings obtained on the seven factors for all 50 scales are presented in Table 1. The seven
rotated factors can also be examined individually by considering all of the scales that display salient
loadings on each factor. The salient factor loadings for the seven rotated factors, and the 6FPQ and
NEO-PI-R scales that define them, are listed below. The minimum factor loading required for
salience was chosen as 0.45.
Factor I-E.xtraversion:
NEO-PI-R
Gregariousness
Exhibition
6FPQ
6FPQ
Affiliation
Dominance
6FPQ
NEO-PI-R
Assertiveness
Positive Emotions
NEO-PI-R
NEO-PI-R
Excitement Seeking
NEO-PI-R
Modesty

0.79
0.74
0.72
0.61
0.57
0.52
0.47
- -0.48

All of the 6FPQ scales targeted upon the Extraversion factor produced salient loadings, as did

Six-factor model of personality

395

Table I. Rotated matrix of factor loadings for NEO-PI-R, 6FPQ, and Desirability Scales

II

111

Factors
IV

VI

VII

-0.11
0.16
0.03

0.08
0.01
0.20

-0.17
0.41
-0.17

0.31
-0.08
0.11

6FPQ Scales

0.72
0.61
0.74

0.04
-0.28
-0.13

Abasement
Even-Tempered
Good-Natured

-0.02
0.04
-0.08

0.47
0.75
0.62

-0.03
0.06
0.04

0.31
0.21
0.22

0.18
-0.08
0.08

0.02
0.11
-0.09

0.04
0.30
0.18

Cogmtive Structure
Deliberateness
Order

-0.20
-0.07
-007

0.05
0.25
-0.18

0.49
0.54
0.78

-038
-0.17
0.12

-0.14
-0.15
0.00

0.37
0.48
0.02

0.10
0.13
0.09

Autonomy
Indivtdualism
Self-Reliance

-0.38
-0.24
-0.28

-0.24
-0.06
-0.30

-0.20
0.07
-0.26

0.59
0.62
0.51

0.20
0.20
0.15

0.08
-0.21
0.21

0.06
0.28
0.16

Change
Understanding
Breadth of Interest

0.10
-0.01
0.11

0.09
0.02
0.20

0.32
-0.09
0.06

0.29
0.06
0.12

0.57
0.64
0.82

-0.14
0.40
0.10

-0.10
0.11
0.20

Achievement
Endurance
Seriousness

-0.05

0.07
0.00
0.10

0.40
0.15
0.27

-0.11
0.22
-0.10

0.13
0.18
--a.09

0.62
0.56
0.61

0.18
0.36
-0.20

-0.13
-0.07
-0.39
-0.34
0.22
-0.16

0.11
-0.68
0.04
0.09
-0.28
0.17

0.04
-0.07
-0.18
0.01
-0.30
-0.02

-0.79
-0.37
-0.66
-0.63
-0.44
-0.73

0.02
0.09
0.09
-0.17
0.17
0.05

0.09
-0.06
-0.07
0.06
-0.32
-0.20

-0.07
-0.40
-0.32
-0.34
-0.11
-0.36

0.45
0.79
0.57
0.43
0.47
0.52

0.26
0.08
-0.27
-0.31
-0.06
0.12

-0.09
0.02
-0.07
0.13
-0.05
0.01

-0.27
-0.09
0.15
0.00
0.18
-0.08

0.10
-0.05
0.15
0.12
0.05
0.13

-0.07
-021
0.22
0.16
-0.38
-0.21

0.58
0.03
0.20
0.29
-0.1:
0.61

Fantasy
Aesthettcs
Feelings
Actions
Ideas
Values

0.15
-0.08
0.10
0.25
0.04
0.13

-0.24
0.19
-0.26
0.03
0.10
0.22

-0.39
-0.03
-0.16
0.17
-0.18
-0.13

-0.26
-0.10
-0.45
0.20
0.20
0.24

0.32
0.76
0.35
0.67
0.60
0.39

-026
0.01
-0.16
-020
0.43
-0.10

0.26
0 24
0 3')
0.13
0.22
0.15

Trust
Stratghtforwardness
Altruism
Compliance
Modesty
Tender-Mindedness

0.09
-0.22
0.08
0.00
-0.48
0.05

0.64
0.56
0.48
0.83
0.42
0.41

-0.10
0.06
-0.17
-0.02
0.00
-0.17

0.08
-0.28
-0.22
-0.09
-0.13
-0.14

0.08
--a03
0.01
-0.09
0.05
0.15

-0.01
0.00
0.03
0.05
-022
-0 I7

0.40
0.4x
0 58
0.24
0 3.!
0.4i

Competence
Order
Dutifulness
Achievement Striving
Self-Discipline
Deliberateness

0.20
-0.14
-0.10
0.10
0.09
-0.14

-0.13
-0.28
0.01
-0.18
-0.13
0.28

0.24
0.75
0.43
0.34
0.53
0.50

0.20
0.13
0.02
-0.01
0.22
-0.01

-0.04
-0.03
-0.02
0.07
-0.01
-0.09

0.57
0.05
0.46
0.69
0.50
0.47

0.44
0.00
041
0.21
0.41
003

0.15
0.24

0.23
0.17

0.26
0.17

0.07
0.10

0.05
0.04

0.19
0.02

0.64
0.67

Affiliation
Dominance
Exhibition

-0.11
-0.35

-0.06
-0.09
-0.12

NEO-PI- R Scalus

Anxiety
Angry Hostility
Depression
Self-Consciousness
lmpulsivtty
Vulnerabtlity
Warmth
Gregariousness
Assertweness
Actiwty
Excnement Seeking
Posmve Emottons

Desirtrbrlir~~

Desirability I
Desirability II

Note. Targeted factors are as follows: I = Extraversion; II = Agreeableness; 111= Methodicalness; IV = Independence: V = Openness to
Experience; VI = Industriousness; VII = Desirability. Each scales loadings on its targeted factor(s) are in bold type.

four of the six targeted NEO-PI-R scales. The three 6FPQ scales all loaded above 0.60, and the
NEO-PI-R Gregariousness and Assertiveness scales achieved comparable results. Somewhat lower
loadings were produced for NEO-PI-R Positive Emotions and Excitement Seeking, while two other
NEO-PI-R Extraversion scales, Warmth and Activity, failed to yield salient loadings.
The NEO-PI-R Modesty scale, though targeted as a measure of Agreeableness, also yielded a

396

Douglas N. Jackson et al.

salient negative loading on the Extraversion factor. This is, however, consistent with the data listed
in the NEO-PI-R manual (Costa & McCrae, 1992) which reports a negative correlation between
Modesty and PRF Exhibition, and with the semantic opposition of the trait names themselves.
Factor 2-Agreeableness:
NEO-PI-R
Compliance
6FPQ
Even-tempered
NEO-PI-R
Trust
6FPQ
Good-natured
NEO-PI-R
Straightforwardness
NEO-PI-R
Altruism
6FPQ
Abasement
NEO-PI-R
Angry Hostility

0.83
0.75
0.64
0.62
0.56
0.48
0.47
- .0.68

As was the case for Extraversion, all three of the 6FPQ scales, and four of the six NEO-PI-R
scales, that were targeted for the Agreeableness factor yielded salient loadings. Certain scales, such
as NEO-PI-R Compliance and 6FPQ Even-tempered, defined this factor particularly well, while
others produced somewhat smaller loadings. Two NEO-PI-R Agreeableness scales, Modesty and
Tender-Mindedness, did not achieve salient loadings.
One of the higher loadings for this factor was for NEO-PI-R Angry Hostility, a scale that was
keyed on Neuroticism by Costa and McCrae, and that was therefore targeted (negatively) on the
Independence factor. This result raises questions about the distinctiveness of the neurotic hostility
and antagonistic hostility constructs that have been hypothesized as relevant to risk of coronary
heart disease (Costa & McCrae, 1992).
Factor 3-Methodicalness:
Order
6FPQ
NEO-PI-R
Order
Deliberateness
6FPQ
NEO-PI-R
Self-Discipline
NEO-PI-R
Deliberation
Cognitive Structure
6FPQ

0.78
0.75
0.54
0.53
0.50
0.49

The three 6FPQ scales targeted for Methodicalness all yielded salient loadings, as did three of
the six NEO-PI-R Conscientiousness scales, which were targeted for both Methodicalness and
Industriousness. The very high loadings for the Order scales of both instruments suggest that
orderliness and neatness are central to the Methodicalness factor. The other three NEO-PI-R scales
with salient Methodicalness loadings load almost as highly on Industriousness (Factor 6) as on
Methodicalness.
Factor &Independence:
Individualism
6FPQ
Autonomy
6FPQ
Self-Reliance
6FPQ
NEO-PI-R
Feelings
NEO-PI-R
Self-Consciousness
NEO-PI-R
Depression
NEO-PI-R
Vulnerability
NEO-PI-R
Anxiety

0.62
0.59
0.51
- 0.45
-0.63
-0.66
-0.73
-0.79

All three of the 6FPQ scales targeted for this factor produced salient loadings, but the high
negative loadings for four NEO-PI-R Neuroticism scales suggest that susceptibility to negative
emotions is the essence of low Independence. Two other NEO-PI-R Neuroticism scales, Angry
Hostility and Impulsiveness, did not produce salient loadings.
The salient loading for NEO-PI-R Feelings, its highest among the seven factors, suggests that it
is a better measure of Neuroticism or low Independence than of Openness to Experience, its targeted
factor.

397

Six-factor model of personality


Factor S--Openness to Experience:
Breadth of Interest
6FPQ
NEO-PI-R
Aesthetics
Actions
NEO-PI-R
Understanding
6FPQ
NEO-PI-R
Ideas
6FPQ
Change

0.82
0.76
0.67
0.64
0.60
0.57

The three 6FPQ scales targeted for the Openness to Experience factor yielded high salient loadings,
as did three of the six NEO-PI-R scales targeted upon this factor. Three other NEO-PI-R Openness
to Experience scales-Fantasy,
Feelings and Values-did not produce salient loadings on this factor.
Factor &-Industriousness:
NEO-PI-R
Achievement Striving
6FPQ
Achievement
Seriousness
6FPQ
NEO-PI-R
Competence
6FPQ
Endurance
NEO-PI-R
Self-Discipline
Deliberateness
6FPQ
NEO-PI-R
Deliberation
NEO-PI-R
Dutifulness

0.69
0.62
0.61
0.57
0.56
0.50
0.48
0.47
0.46

The three 6FPQ scales targeted upon the Industriousness factor all produce salient loadings, as
do five of the six NEO-PI-R Conscientiousness scales, which were targeted for both Methodicalness
and Industriousness. Two of the NEO-PI-R scales, Achievement Striving and Competence, produced quite substantial loadings, while the other three, Self-Discipline, Deliberation and Dutifulness,
produced somewhat lower loadings on both Methodicalness and Industriousness.
One scale from the 6FPQ Methodicalness factor, Deliberateness, shows a salient secondary
loading on the Industriousness factor, indicating that this scale involves Industriousness in addition
to Methodicalness.
Factor 7-Desirability:
Desirability II
Desirability I
NEO-PI-R Positive Emotions
NEO-PI-R Warmth
NEO-PI-R Altruism
NEO-PI-R Straightforwardness
NEO-PI-R Tender-Mindedness

0.67
0.64
0.61
0.58
0.58
0.48
0.45

In addition to the two Desirability scales, five NEO-PI-R scales yielded salient loadings on the
Desirability factor. Three of these scales loadings approach those of the Desirability scales themselves, suggesting that they are substantially saturated with desirability variance. None of the 6FPQ
scales were salient upon this factor.
Mean factor loadings for targeted NEO-PI-R and 6FPQ scales

The mean loadings obtained by the targeted scales of each questionnaire on the six content factors
are displayed below.

6FPQ

E.rtrmersion
Agreeableness
Methodicalness
Independence
Openness
to E.rperienre
Industriousness
AWage

Scules

0.69
0.62
0.60
0.57
0.68
0.60
0.63

NE@

PI-R

0.54
0.56
0.47
0.61
0.62
0.46
0.54

S~olr.s

398

Douglas N. Jackson et al.

One potential problem with the above comparison involves the NEO-PI-R Conscientiousness
scales, whose targetings for both Methodicalness and Industriousness may have reduced each scales
loadings on one scale or the other. If loadings on both Methodicalness and Industriousness factors
are averaged, both for the NEO-PI-R Conscientiousness scales and the 6FPQ Methodicalness and
Industriousness scales, the result is an average loading on a Conscientiousness factor of 0.45 for the
6FPQ scales and 0.47 for the NEO-PI-R scales. These results indicate that both the 6FPQ and
NEO-PI-R are effective measures of both the five- and six-factor personality models.
In considering the above comparisons, it should be recognized that factor loadings are a function
of scale reliabilities, which are in turn influenced by scale length and the overall number of items in
a questionnaire. Given that the 6FPQ contains a total of only 108 items as compared with the NEOPI-Rs 240, the data indicate that the 6FPQ achieves at least comparable levels of factor separation
with fewer than half of the items contained in the NEO-PI-R. This may be a consequence of several
NEO-PI-R scales that load on inappropriate content factors or on the Desirability factor. It may
also be a result of NEO-PI-R scales that measure constructs not closely related to those measured
by the 6FPQ.
Desirability saturation of NEO-PI-R and 6FPQ factors

The overall desirability saturation for each of the five NEO-PI-R factors and six 6FPQ factors
was also determined. This was accomplished by averaging the absolute factor loadings on Desirability yielded by the scales within each factor. Thus, for each 6FPQ factor, the Desirability loadings
for all three component scales were averaged, while for each NEO-PI-R factor, the Desirability
loadings of all six component scales were averaged. These mean loadings are presented below:

Agrerablrnes.s
Ertrocersion
Independence/low Nruroticim
0pmne.n to Erperirnw
Consrienti0urn~s.s
Indu.stri0u.snes.s
Methodmhe.w
AW-i3ge

NEO-PI-R

6FPQ

0.41
0.31
0.27
0.23
0.26

0.17
0.16
0.17
0.14

0.30

0.24
0.11
0.17

Thus, four of the five NEO-PI-R factors are more highly saturated with desirability than any
6FPQ factor, and the other NEO-PI-R factor is more highly saturated with desirability than all but
one of the 6FPQ factors. These results clearly reflect the differing philosophies of the tests constructors: Jackson tried to minimize social desirability when selecting PRF and JPI items, while
Costa and McCrae consciously chose not to do so during NEO-PI-R scale construction.
Stability of results: salience and congruence

It should be noted that the sample size, number of variables and number of factors in this study
were appropriate to ensure that the results would be stable across samples. According to Guadagnoli
and Velicers (1988) criterion for producing a stable solution in factor or component analysis, a
sample size of 141 requires, for each factor, an average loading of 0.49 for each salient variable on
that factor. Our salience threshold of 0.45 was sufficient to meet this requirement; the lowest mean
salient loading for any factor was 0.55, for Industriousness.
Similarly, the solution provided by the targeted rotation provided evidence that our reported
rotated factor analytic results are more than chance results. Table 2 presents a comparison of the
congruence coefficients between our factor loadings and the target matrix employed in the rotation
with the corresponding mean congruence coefficients obtained from rotating our factor matrix to
1000 randomly-derived targets. For all seven factors the congruence coefficients exceed by more
than 2.5 standard deviations those obtained from random rotation. Clearly, we can reject the
hypothesis that our results are traceable to rotating to a random target.

399

Six-factor model of personality


Table 2. Congruence

coefficients obtained for targeted


randomly-rotated
factors

Factor
Extraversion
Agreeableness
Methodicalness
Independence
Openness
Industriousness
Desirability

Targeted
rotation
_~
0.81**
0.78
o.so***
0.81***
0.89***
0.72***
0.64***

Random
rotation
(mean)
0.40
0 38
0.39
0.32
0.38
0.40
0.39

factors and for

Random
rotation
(SD)
0.10
0.10
0.09
0.12
0 II
O.OY
0.09

No/e. ***p
< 0.001
Congruence coefficients were computed between a binary target matrix
and rotated principal component loadings The tirst column refers to
congruence coefficients based on hypothesized loadings. the second
and third columns report respectively the mean and standard dewation of congruence coefficients based on 1000 random permutations
of the hypothesized target loadings for each factor.

DISCUSSION

Coherence of the factors

The seven factors obtained in this analysis were both comprehensible and plausible. All of the six
content-based factors were clearly defined by the targeted 6FPQ scales and by several NEO-PI-R
scales. On average, scales from the two instruments load about equally highly on the six factors.
The Extraversion, Agreeableness and Openness to Experience factors clearly are very similar
across the two instruments. The close similarity of Independence and low Neuroticism is also
confirmed. All scales keyed on 6FPQ Methodicalness and Industriousness factors load on the
expected factors, and all scales from NEO-PI-R Conscientiousness load on at least one of these.
The Desirability scales are at the heart of a seventh factor, with strong contributions from many
NEO-PI-R scales whose very names imply social desirability: Positive Emotions. Warmth and
Altruism.
NEO-PI-R scales corresponding to 6FPQ factors

The results of this study indicate that Jackson et al.s (1996) six-factor model of personality can
be adequately represented by NEO-PI-R scales, in addition to those of the 6FPQ. Indeed, formulas
can be constructed that will define each of the 6FPQ personality factors in terms of NEO-PI-R
scales, and these are listed below (scales loading approximately equally on two or more factors have
been excluded, to avoid inflating correlations between these scale combinations):
Estraversion = Gregariousness + Assertiveness; Agreeableness = Compliance + Trust - Angry Hostility;
Methodicalness = Order; Independence = -Vulnerability-Anxiety
- Depression - Self-Consciousness; Openness to Experience = Aesthetics + Actions + Ideas; and Industriousness = Achievement
Striving + Competence.
Desirability responding in the NEO-PI-R

Our results show evidence of a desirability factor that is distinct from content factors, with
substantial loadings for several NEO-PI-R scales. Our interpretation of this finding is that desirability responding represents a confound that detracts from the construct interpretation of scales
loading this factor. A portion of the variance on these scales is subject to an alternative interpretation
from that implied by their scale names. Although it is true that desirability responding might
demonstrate empirical validities in terms of certain criterion measures (Jackson & Messick, 1958),
the presence of such a confound always interferes with the logical or construct validity of a
measure (Cronbach, 1950).
Possible threats to validity

Our results indicate that it is possible to identify six content factors that are common to the 6FPQ
and the NEO. It would be useful, however, to consider threats to the validity and generalizability

400

Douglas N. Jackson et al.

of our findings in the light of the present and previously published evidence. We list and discuss six
possible threats to validity.
(1) Our results are based on a single sample with a relatively large number of variables in relation
to sample size. Guadagnoli and Velicers (1988) important work on the stability of principal component factor loadings demonstrated that the subject-to-variable ratio was irrelevant in determining
factorial stability. These authors examined the similarity between sample and population component
patterns across seven levels of sample size, four levels of the number of variables, three levels of the
number of components, and three levels of the saturations of loadings. They devised an overall
statistic, g2, to reflect the average squared difference between comparable loadings of corresponding
sample and population component patterns. Based on their extensive Monte Carlo analyses, functions were calculated relating sample size, loading saturation and the level of g2. Results of their
analyses indicated the importance of moderate to high levels of saturation and of absolute sample
size, and the unimportance of long-standing rules of thumb, such as the subject-to-variable ratio.
As noted in the Results section above, the sample size and levels of saturation in our study meet
the standard set by Guadagnoli and Velicer for obtaining stable results. Nevertheless, we recommend
that our analyses be replicated on other samples.
(2) The use of targeted or Procrustes rotations capitalizes on chance. Moreover, within the context
of targeted rotations, other methods of factor extraction or rotation should be considered. We have
already demonstrated through the use of a Monte Carlo procedure that the results that we report
are quite distinct from what would be expected on the basis of chance. It is true that just as with
exploratory analyses, targeted rotation using a certain number of factors does invite the criticism
of possible arbitrariness in the choice of a model. But in this case we had the benefit of previous
hypothesis testing statistical structural equation modelling results on which to base our targeted
facets that comprise the 6FPQ.
Our choice of an orthogonal rotation was based on our view that the six factors, while not
perfectly uncorrelated, should be distinct enough to produce viable orthogonal solutions. Although
we chose not to examine oblique solutions, we did decide to examine the results of extraction
methods other than principal components analysis. For example, we performed a principal factor
analysis on the correlation matrix, using as diagonal values the variables squared multiple correlations with all other variables. The congruence coefficients between these rotated principal axis
factors and the rotated principal components averaged 0.999.
(3) Industriousness and Methodicalness represent the same factor. Our primary intent in this paper
was not to evaluate the relative merits of the five- and six-factor models of personality. Rather, we
sought to evaluate the degree to which it was possible to interpret results from a prominent device
for assessing the big five personality factors in terms of a six-factor solution. But in a previous
paper (Jackson et al., 1996) employing structural equation modelling explicitly designed to evaluate
five vs six factors for the domain represented by the Personality Research Form, the six-factor
solution was found to be the preferred model. Separate correlated factors of Methodicalness and
Industriousness consistently yielded a significantly better-fitting model than did a single Conscientious factor in three distinct samples. Our present results should thus not be evaluated in
isolation, but in the context of previous results supporting a six-factor interpretation.
(4) The six-factor model does not emerge in analyses of other instruments. Important evidence for
the existence of a basic personality factor is provided when the factor emerges in analyses of
different instruments. For the Methodicalness and Industriousness factors, such evidence is perhaps
forthcoming from the results of research by Zuckerman, Kuhlman and colleagues. In two factoranalytic studies of wide batteries of personality test scales (Zuckerman, Kuhlman & Camac, 1988;
Zuckerman, Kuhlman, Thornquist & Kiers, 1991), a stable five-factor solution emerged, despite the
lack of any scales measuring intellectual or artistic interests. Three of the factors that were found
correspond to the big five factors of Extraversion, Neuroticism and Agreeableness. The highestloading scale on another factor, which was called Psychoticism-Impulsive
Uninhibited Sensation
Seeking by Zuckerman et al., was PRF Cognitive Structure, suggesting that this factor would
correspond to a broad version of 6FPQ Methodicalness. The second-highest-loading scale on the
remaining factor, which was named Activity by Zuckerman et al., was JPI Energy Level. The
correlations of this scale with PRF Achievement and Endurance (Jackson, 1994, p. 64) suggest that
this factor may be strongly related to the 6FPQ Industriousness factor. Presumably, if scales

Six-factor model of personality

401

measuring intellectual or artistic interests had been included in the Zuckerman et al. studies, a sixfactor solution could have emerged.
(5) The desirability factor can be interpreted in terms of content. It is important to emphasize our
interpretation of the seventh factor obtained in this study as one reflecting style rather than content.
Although the NEO-PI-R scales with high loadings on this factor-Warmth,
Altruism and Positive
Emotions-might
suggest an interpretation of friendliness or happiness for this factor, it is important
to note that this factor was most strongly defined by the two desirability scales themselves, whose
items span a range of content not limited to that represented by the scales above. Moreover, three
of the NEO-PI-R Conscientiousness scales-Competence,
Dutifulness and Self-Discipline--show
loadings of above 0.40 on this seventh factor, complicating any unitary content-based interpretation
of the factor.
(6) Our sample of university students was homogeneous in terms ojage. Despite the similarity in
age of the participants in our sample, there is a large body of evidence indicating that standard
deviations of personality scale scores for the general population are no larger than those for
university students. Accordingly, we consider our choice of a sample not to have compromised our
conclusions.

CONCLUSIONS
The results of this study suggest the following conclusions.
Six coherent content factors are well-defined by both 6FPQ and NEO-PI-R scales. The salient
loadings on these factors clearly correspond to those posited by the six-factor model.
The distinction between the Methodicalness and Industriousness factors is supported by the data.
Several NEO-PI-R Conscientiousness scales, and two 6FPQ scales, load on both factors, but several
others show substantial loadings on only one.
The scales of the NEO-PI-R are more heavily saturated with the social desirability response style
than are the scales of the 6FPQ. Several NEO-PI-R scales load more highly on the desirability
factor than on their targeted content factors.
Five or six major personality factors can be measured effectively by the NEO-PI-R or the 6FPQ.
The 6FPQ does so with considerably fewer scales and items.
Acknowledgemenrs--This
research was supported
in part by a research grant from the Social Sciences and Humanities
Research Council of Canada. The authors thank Ilana Barak and Steven Mitchell for their help in data collection.

REFERENCES
Block,J.

(1995). A contrarian

view of the five-factor

approach

to personality

description,

Psychological

Bullerin I/7,

187

215.
R. R. (1985). NE0 Personaiii.v Inventory Prqfessional Manual. Odessa, FL: Psychological
Assessment Resources, Inc.
Costa, P. J., Jr. & McCrae, R. R. (1988). From catalog to classification:
Murrays needs and five-factor model. Journul (!f
Personalit), and Social Ps_vcholog,v,55, 258-265.
Costa. P. J., Jr. & McCrae. R. R. (1992). Revised NE0 Personulitv Inoentor_vProfessional Manual. Odessa, FL: Psychological
Assessment Resources. Inc.
Cronbach,
L. J. (1950). Further evidence on response sets and test design. Educationtd and Psvchologicul Mecrsurrmerrt. /It.
3-31.
Digman, J. M. (1990). Personality structure: Emergence of the five-factor model. Annual Reoiew ofPsvchologv. 41, 417440.
Guadagnoli,
E. & Velicer. W. L. (1988). Relation of sample size to stability of component
patterns, Psvcholo,yicu/ Bulketin.
103, 265-275.
Horn, J. L. (1967). On subjectivity in factor analysis. Educational and Pq~chological Measurement, 27, 81 l-820.
Humphries.
L. G., Ilgen, D.. McGrath,
D. & Montanelli,
R. (1969). Capitalization
on chance in rotation
of factors.
Educational and Psychological Measurement. 29, 259-271.
Jackson, D. N. (1994). Jackson Personality Inventory-Rerised Munual. Port Huron, Ml: Sigma Assessment Systems.
Jackson, D. N. (1984). Personality Research Form Manual. Port Huron, MI: Sigma Assessment Systems.
Jackson, D. N. & Lay, C. H. (I 968). Homogeneous
dimensions of personality scale content. Multiuariate Behavioral Research,
3, 321-337.
Jackson, D. N. & Messick, S. (1958). Content and style in personality assessment. Psvchological Bulletin, 55, 2433252.
Jackson, D. N. & Morf. M. (1973). An empirical evaluation of factor reliability. Muttiwriate Behaaioral Research. X. 439459.
Jackson, D. N. & Paunonen,
S. V. (in press). Six Factor Personality Questionnaire. Port Huron, MI: Sigma Assessment
Systems.

Costa, P. J., Jr. & McCrae,

402

Douglas

N. Jackson

et al.

Jackson, D. N., Paunonen. S. V., Fraboni, M. & Goffin, R. D. (1996). A five-factor versus six-factor model of personality.
Personalily and Individual Differences, 20, 3346.

Morf, M. & Jackson, D. N. (1972). An analysis of two response styles: True responding and item endorsement, Educational
and Psychological Measurement. 32, 329-353.

Murray, H. A. (1938). Explorations in personalify. New York: Oxford University Press.


Norman, W. T. (1963). Toward an adequate taxonomy of personality attributes: Replicated factor structure in peer
nomination personality ratings. Journal of Abnormal and Social Psychology, 66, 574-583.
Paunonen, S. V. (1993, August). Sense, nonsense, and rhe Big Five factors of personality. Paper presented at the meeting of
Division Five of the American Psychological Association, Toronto, Ontario, Canada.
Paunonen, S. V. (in press). On chance and factor congruence following orthogonal Procrustes rotation. Educafional and
Psychological Measurement.

Schonemann, P. T. (1966). The generalized solution of the orthogonal procrustes problem. Psychometrics, 31, 1-16.
Tupes, E. C. & Christal, R. E. (1961). Recurrenf personalifyfacfors based on trait rafings. USAF ASD Technical Report No.
61-97. Lacklund Air Force Base, TX: U.S. Air Force.
Zuckerman, M., Kuhlman, D. M. & Camac, C. (1988). What lies beyond E and N? Factor analyses of scales believed to
measure basic and dimensions of personality. Journal of Personality and Social Psychology, 54,96-107.
Zuckennan, M., Kuhlman, D. M., Thornquist, M. & Kiers, H. (1991). Five (or three) robust questionnaire scale factors of
personality without culture. Personality and Individual Differences, 12, 929-941.

You might also like