You are on page 1of 14

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/232591551

Brief versions of the NEO PI-3

Article  in  Journal of Individual Differences · July 2007


DOI: 10.1027/1614-0001.28.3.116

CITATIONS READS

118 21,615

2 authors, including:

Paul Costa
Duke University Medical Center
433 PUBLICATIONS   65,781 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

SNPs and Personality View project

Cross-Cultural Studies of traits View project

All content following this page was uploaded by Paul Costa on 24 February 2015.

The user has requested enhancement of the downloaded file.


JournalR.R. McCrae
of Individual & P.T.
Published 2007,Co
Differences sta,
Hogrefe
2007;Jr.:Vol.
&Brief
HuberNEO-PI-3s
28(3):116–128
Publishers

Brief Versions of the NEO-PI-3


Robert R. McCrae and Paul T. Costa, Jr.
National Institute on Aging, National Institutes of Health,
Department of Health and Human Services, Baltimore, MD, USA

Abstract. Self-report (Form S) and observer rating (Form R) versions of two short forms of the NEO Personality Inventory-3 (NEO-
PI-3) were evaluated. The NEO Five-Factor Inventory-3 is a 60-item instrument that assesses the five factors. The NEO-PI-3 First Half
consists of the first 120 items of the NEO-PI-3, selected to optimally assess its 30 facet scales. Internal consistencies were systematically
lower for the brief scales than for the parent scales, but both brief instruments replicated the factor structure and showed cross-observer
validity. They appeared to work well in both adolescent and adult samples, and adequately in a middle school-age sample. Norms for
the NEO-FFI-3 are provided and it was shown that NEO-PI-3 norms can be used for NEO-PI-3 First Half scores if they are simply
doubled. Uses of the NEO-PI-3 First Half are discussed.

Keywords: personality assessment, short forms, NEO-PI-3, NEO-FFI

It is a statistical truism that, other things being equal, tests the NEO-FFI, confirming the research utility of brief
with more items are more reliable. There is evidence that measures of the FFM.
more reliable tests are more valid (although the correlation The NEO-FFI was based on items from the earliest
between reliability and validity is only about .33; see version of the NEO-PI, and did not offer the optimal short
Meyer et al., 2001, p. 131). It is also true that instruments form of the instrument. Some researchers have criticized
that measure an array of specific constructs are generally its psychometric properties, in particular its item factor
superior in the prediction of a variety of criteria to those structure (Becker, 2006; Egan, Deary, & Austin, 2000).
that assess only broad, global constructs (Paunonen & Ash- In 2004, McCrae and Costa, therefore, proposed a revi-
ton, 2001). Thus, personality assessment is likely to be op- sion of the NEO-FFI in which 14 items were replaced by
timized when a comprehensive set of specific traits is as- alternatives from the NEO-PI-R item pool. The Revised
sessed with multi-item scales. NEO-FFI (NEO-FFI-R) showed slightly improved psy-
However, there are also pragmatic reasons to use brief chometric properties and better readability compared to
measures, the most important of which is the limited time the NEO-FFI.
and cooperation of potential respondents. Survey re- To improve the psychometrics and readability of the
searchers, in particular, often wish to include brief mea- full NEO-PI-R, McCrae, Costa, and Martin (2005) devel-
sures of personality among a host of other variables of oped the NEO-PI-3, in which 37 of the NEO-PI-R items
interest. Investigators who target groups with low litera- were replaced. The NEO-PI-3 scales were essentially
cy (e.g., Trobst et al., 2000) may be more successful us- equivalent to those of the NEO-PI-R, but could be used
ing shorter and less taxing questionnaires. Fortunately, by adolescents as young as 12 (Costa, McCrae, & Martin,
there is evidence that brief scales can provide useful as- in press) as well as by adults (McCrae, Martin, & Costa,
sessments of personality constructs (Gosling, Rentfrow, 2005). One of the NEO-FFI-R items (“I’m hard-headed
& Swann, 2003; Herzberg & Brahler, 2006; Rammsted and tough-minded in my attitudes”) was dropped in the
& John, 2005). development of the NEO-PI-3, so the short version of the
The NEO Personality Inventory (NEO-PI[-R]; Costa NEO-PI-3–the NEO-FFI-3–consists of 59 NEO-FFI-R
& McCrae, 1985, 1992) is a measure of the dimensions items plus the replacement (“I have no sympathy for beg-
of the five-factor model (FFM; Digman, 1990), with six gars”) for the dropped item. In this article we report anal-
facet scales to assess specific aspects of each of the fac- yses of the reliability and item factor structure of the
tors, and with parallel forms for self-reports and observer NEO-FFI-3 and provide new norms for adolescents and
ratings. With 240 items, it assesses 30 traits, yet can be adults.
completed by most respondents in about 30 minutes. The NEO-FFI-3 does not provide specific facet infor-
Soon after its publication, however, it became clear that mation. Saucier (1998) proposed a scoring system for
many researchers required a briefer instrument, and in subcomponents of the NEO-FFI scales, but only 13 con-
1989, Costa and McCrae published a short version, the structs could be assessed, which generally did not corre-
NEO Five-Factor Inventory (NEO-FFI), that assesses the spond to the facets of the NEO-PI-R and, thus, cannot be
global factors with five 12-item scales. By December compared directly to the large literature on NEO-PI-R
2006, a PsycINFO search showed over 680 citations of facet scales. More desirable for research would be brief

Journal of Individual Differences 2007; Vol. 28(3):116–128 Published 2007, Hogrefe & Huber Publishers
DOI 10.1027/1614-0001.28.3.116
R.R. McCrae & P.T. Costa, Jr.: Brief NEO-PI-3s 117

versions of each of the 30 NEO-PI-3 facet scales. A com-


puterized adaptive test (CAT) analysis of NEO-PI-R data
Method
(Reise & Henson, 2000) suggested that such a project
might be feasible. Reise and Henson concluded that, us- Participants
ing CAT methods, most of the information from the
The present study reanalyzes data from three samples that
eight-item facet scales could be obtained from adminis-
completed the NEO-PI-3. All participants were recruited
tration of the four best items. In this article we will iden-
from the community by psychology students, and all com-
tify four-item NEO-PI-3 facet scales through convention-
pleted questionnaires at home and mailed them back to the
al regression analyses and evaluate their psychometric
investigators. In the first study, the NEO-PI-R and 96 can-
properties in samples of adults and adolescents. By pre-
didate replacement items were administered to 536 individ-
senting the 120 selected items at the beginning of the
uals aged 14 to 20, who described themselves and a sibling
NEO-PI-3 questionnaire, the same test booklet can con-
or an anonymous target of the same age group (McCrae,
veniently be used for either the full or the brief version
Costa, et al., 2005). In the subset of 500 individuals with
of the NEO-PI-3. Because of this item placement, we will
valid protocols (the Adolescent sample), 51.6% were fe-
refer to the brief version as the NEO-PI-3 First Half
male; 84.6% were White, 7.6% were Asian/Pacific Island-
(NEO-PI-3FH). This strategy was used by Morey (1991)
er, and 7.8% were other ethnicity; and most described
for the Personality Assessment Inventory (PAI), and a re-
themselves as A or B students. Of these, 180 described a
cent evaluation of the PAI short form scales suggested
sibling who also described them. Cross-observer validity
that they could be appropriately used in some circum-
coefficients can be obtained from these pairs.
stances (Frazier, Naugle, & Haggerty, 2006).
The second study administered the same items to 652
It might be argued that even briefer versions could be
adults aged 21 to 91 who also described themselves and an
created: Facets could be roughly estimated by two items or
adult partner or anonymous target (McCrae, Martin, et al.,
even one (Mullins-Sweatt, Jamerson, Samuel, Olson, &
2005). In the subset of 635 respondents with valid data (the
Widiger, 2006). However, even when such short measures
Adult sample), 56.1% were women; 92.6% were White,
show reasonable convergent validity, they contain specific
1.6% Asian/Pacific Islander, and 5.8% other ethnicity; and
variance in addition to the common, valid variance and er-
75.6% had more than a high school education. Mutual rat-
ror; in very large samples (e.g., from the Internet) such spe-
ings of spouses, friends, and other targets were obtained
cific variance can show significant correlations with other
from a subset of 532 individuals.
variables. In single-item scales, content is always con-
founded with acquiescence. When single-item scales are In the third study, the 240 items selected for the NEO-
used it is impossible to determine whether observed asso- PI-3 in the first study were administered to 449 boys and
ciations result from the intended construct or idiosyncratic girls aged 12 or 13 (Costa et al., in press). Valid self-reports
properties of the item. Averaging across four items not only were obtained from a subset of 202 respondents; valid rat-
reduces random error; it also tends to neutralize the system- ings of anonymous targets of the same age and gender were
atic bias of each individual item. If keying is balanced, the obtained from a different subset of 222 respondents. Most
effects of acquiescence are largely controlled. of the respondents (the Middle School sample; 52.4% fe-
Normally the introduction of new scales requires pub- male) were White (94.5%); in the 6th, 7th, or 8th grade;
lication of norms. It would be easy to generate means and and described themselves as A or B students.
standard deviations for NEO-PI-3FH scales in the same
samples on which the NEO-PI-3 is normed but use of
those norms would be cumbersome. The NEO-PI-3 al- Instrument
ready has norms for different ages, genders, and forms;
duplication of all these norms–and profile sheets–for a The NEO-PI-3 is a 240-item questionnaire measuring the
First Half version would be costly and might lead to er- FFM and 30 specific traits that define the five factors or
rors in using the correct norms. It might be argued that a domains: Neuroticism (N), Extraversion (E), Openness to
research instrument like the NEO-PI-3FH does not re- Experience (O), Agreeableness (A), and Conscientiousness
quire norms, because group analyses can be conducted (C). Items use a 5-point Likert-scale format, with responses
on raw scores, using age or gender as covariates if appro- from strongly disagree to strongly agree. Keying is roughly
priate. However, even in research contexts it is useful to balanced to minimize the effects of acquiescent respond-
have a rough idea of how new samples compare to the ing. Form S contains items in the first person; Form R con-
general population. One possible solution is provided by sists of the same items phrased in the third person with
the simple expedient of doubling the First Half scores and separate versions for male and female targets.
interpreting them using existing NEO-PI-3 norms. That Scales of the NEO-PI-3 are essentially equivalent to
would be reasonable if, and only if, the means and stand- those of the NEO-PI-R (McCrae, Martin, et al., 2005), and
ard deviations of the doubled First Half scores are com- evidence of the reliability and validity of the latter is pre-
parable to those of the full scores. We will examine that sented in the Manual (Costa & McCrae, 1992).
possibility. The NEO-FFI-3 is a 60-item questionnaire assessing the

Published 2007, Hogrefe & Huber Publishers Journal of Individual Differences 2007; Vol. 28(3):116–128
118 R.R. McCrae & P.T. Costa, Jr.: Brief NEO-PI-3s

Table 1. Coefficient alphas, equivalence coefficients, and factor/scale correlations for NEO-FFI-3 scales
Form S Domain Form R Domain
Sample N E O A C N E O A C
Adolescent (N = 500)
Coefficient α .82 .80 .78 .72 .83 .81 .79 .78 .82 .88
Equivalence coefficient .91 .92 .93 .89 .92 .91 .92 .91 .92 .94
Factor/Scale correlation .96 .96 .95 .94 .97 .96 .96 .96 .94 .97
Cross-observer correlation .39 .57 .52 .45 .47 .34 .56 .49 .41 .50
Adult (N = 635)
Coefficient α .86 .79 .78 .79 .82 .86 .80 .77 .84 .88
Equivalence coefficient .93 .90 .91 .91 .90 .93 .91 .90 .93 .93
Factor/Scale correlation .96 .95 .97 .95 .94 .95 .96 .94 .95 .96
Cross-observer correlation .52 .60 .50 .55 .49 .54 .60 .55 .52 .49
Middle School (Ns = 202, 222)
Coefficient α .79 .76 .71 .76 .87 .75 .76 .66 .78 .87
Equivalence coefficient .92 .90 .88 .92 .94 .88 .90 .87 .91 .95
Factor/Scale correlation .94 .91 .95 .83 .94 .81 .94 .89 .91 .90
Note. Cross-observer correlations are between NEO-FFI-3 scales and corresponding NEO-PI-3 domains of the other Form, N values = 180
adolescents, 532 adults. All correlations are significant at p < .001.

five domains only. It is a modification of the NEO-FFI-R .83, .81, and .81 for Adolescent, Adult, and Middle School
(McCrae & Costa, 2004) by the substitution of one new samples, respectively.
NEO-PI-3 item. The NEO-FFI-R itself was developed to We conducted item factor analyses separately for each
improve the readability and item factor structure of the sample and form. In each, we extracted five factors, per-
NEO-FFI (Costa & McCrae, 1992). Items for the NEO- formed a varimax rotation, and generated factor scores. The
FFI-3 are taken from 28 of the 30 facet scales (all except factor structure was well replicated in both Adolescent and
N5: Impulsiveness and C6: Deliberation). There are both Adult samples. For example, in the Adult Form S analysis,
self-report and observer rating versions of the NEO-FFI-3. all items loaded above .30 on their intended factor, and 50
of them (83.3%) loaded above .40. Only two items had
loadings of .40 or higher on an unintended factor. The rep-
licability of the factor structure can be summarized by cor-
Results relating the obtained factor scores with the a priori domain
scales. These data are reported in Table 1, and show excel-
Validity and Normative Information for the lent correspondence (all r values ≥ .94) in both Adolescent
NEO-FFI-3 and Adult samples. Somewhat lower correlations were
found in the Middle School sample. In part, this may be a
Table 1 summarizes analyses of the reliability and factor result of the relatively small sample sizes (n values = 202,
structure of the NEO-FFI-3. Internal consistency across the 222), which can lead to arbitrary variations in the factor
two forms ranged from .72 to .88 in the Adolescent and rotation. An alternative is orthogonal validimax rotation
Adult samples, with a median of .82. This value is compa- (McCrae & Costa, 1989), in which the item factors are
rable to the median value (.80) reported in the manual for maximally aligned with the scale scores. Using this proce-
the NEO-FFI in an adult sample. In the Middle School sam- dure in the Middle School sample, factor/scale correlations
ple somewhat lower values are seen, with a median of .76. for N, E, O, A, and C were .94, .94, .95, .92, and .93, re-
Equivalence coefficients are calculated as the correlation spectively, for self-report data, and .85, .94, .88, .91, and
between the full 48-item NEO-PI-3 domain scale and the .90, respectively, for observer rating data.
corresponding 12-item NEO-FFI-3 scale. Most of these We also examined the item factor structure of the NEO-
values are above .90 in the Adolescent and Adult samples, FFI-3 separately for males and females in both Adolescent
and near .90 in the Middle School sample. These values are and Adult samples, using standard varimax rotation. For
inflated to some extent by shared method variance, because Form S, factor/scale correlations ranged from .91 to .97
both scales were computed from the same data. A more (Mdn = .95). For Form R, they ranged from .85 to .97 (Mdn
conservative procedure would be to correlate NEO-FFI-3 = .95). Thus, the NEO-FFI-3 structure is clearly generaliz-
scales with the sum of the 36 remaining items from each able across genders.
domain. Across the two forms, these corrected part/total In the adolescent and adult samples we had self-re-
correlations ranged from .72 to .89, with median values of ports and observer ratings of the same targets, allowing

Journal of Individual Differences 2007; Vol. 28(3):116–128 Published 2007, Hogrefe & Huber Publishers
R.R. McCrae & P.T. Costa, Jr.: Brief NEO-PI-3s 119

an assessment of cross-observer validity. Heteromethod regression can capitalize on chance differences among
correlations of NEO-FFI-3 scales with corresponding items. We, therefore, conducted two cross-validation stud-
NEO-PI-3 domain scores from the other Form are given ies, using Form R data from the Adolescent sample (N =
in Table 1, and show that the 12-item scales are strongly 500) and Form S data from the Adult sample (N = 635).
related to independent assessments of personality. A The same procedure was used in both new samples as in
comparison with cross-observer correlations for the full the original sample. Ten reverse-keyed items were neces-
domain scales (McCrae, Costa, et al., 2005; McCrae, sarily included in all three item selections; of the 110 addi-
Martin, et al., 2005) shows that the NEO-FFI-3 scales tional items to be selected from 230 candidates, choices
explain about 90% as much variance as do the NEO-PI-3 from the Adolescent Form R data matched those in the der-
domain scales. ivation sample in 72 cases (χ² = 26.3, p < .001), and choices
Norms for the NEO-FFI-3 are given in the Appendix, from the Adult Form S data matched those in the derivation
using the same age groupings that are employed for the sample in 76 cases (χ² = 38.2, p < .001). These data suggest
NEO-PI-3 (McCrae, Martin, et al., 2005). To assess the rep- that the initial selection succeeded in identifying short
resentativeness of these norms, we compared them to mean scales that should be generalizable across methods of mea-
values in the independent samples used to develop the surement and age groups.
NEO-FFI-R (McCrae & Costa, 2004). Scored with adoles- In an initial assessment of these First Half scales in the
cent norms, the high school students in the development Middle School sample, we computed equivalence coeffi-
sample had T scores ranging from 47 to 52. A wider range cients between the full and half scales. In one case, O6:
of variation was seen in the adult development sample, Values, the correlation was lower than .80. When selection
where T scores (using older adult norms) ranged from 44 procedures were applied to the items of O6: Values in the
to 56. Those adults were well-educated members of the Middle School Form S data (N = 202), three of the initially
Baltimore Longitudinal Study of Aging (Shock et al., selected items were again chosen, but the fourth item dif-
1984), and it is not surprising that they scored lower in N fered. That item, “I believe that we should look to our re-
and higher in O than the present sample. Although none of ligious authorities for decisions on moral issues,” had also
these samples can be considered truly representative of the been selected in the two cross-validation analyses previ-
general population, experience with the NEO-PI-R and ously described, and it was substituted for the initial selec-
NEO-FFI, which have similar normative groups, suggests tion in the final version of the NEO-PI-3FH. To make
that they should be adequate. scores comparable to those of the NEO-PI-3, the sum of
the four items in each brief facet scale is doubled. NEO-PI-
3FH domain scores are computed as the sum of the relevant
Analyses of the NEO-PI-3 First Half six doubled facet scales.

Item Selection
Reliability, Equivalence, and Factor Structure
Item selection was performed on Form S data from the
Adolescent sample (N = 500). Form S data were given Table 2 reports the internal consistency of the First Half
priority because self-reports are more commonly used domain and facet scores in the three samples for both
than observer ratings. Selection was made in the adoles- forms. These values are, of course, lower than those nor-
cent sample on the assumption that items that were more mally seen for full-length NEO-PI-3 scales, but they are
difficult to read or understand would be more easily iden- acceptably high for the domains and most of the facet
tified in this sample and selected out. To reduce the ef- scales. Reliabilities for the Middle School sample are con-
fects of acquiescence, two positively keyed and two neg- sistently lower than those found in older samples, which is
atively keyed items were chosen for each facet. Each also true for the full-length scales. Among 12- and 13-year-
eight-item facet scale was predicted in a stepwise regres- olds, O4: Actions and O6: Values have αs that most re-
sion by the positively-keyed items. The two best items searchers would consider unacceptably low; certainly
were then entered as the first block of a second regres- scores on those facets should be interpreted with caution.
sion; for the second block, the remaining reverse-keyed However, in evaluating these data it must be recalled that
items were included in a stepwise fashion, and the first multiple regression tends to select maximally independent
two selected were chosen. For five scales with only two predictors of the criterion, so internal consistency is often
reverse-keyed items, those were used in the first block of minimized, whereas validity is maximized.
the regression, and the two best positively-keyed items Table 3 reports equivalence correlations between the
were selected from the second block. For three of the brief and full versions of the domain and facet scales. These
scales, E4: Activity, O4: Actions, and O6: Values, adjust- values are uniformly high for the domain scales, and all
ed R2s were less than .80 (.76, .72, and .75, respectively). exceed .80 for the facet scales. These values are an upper
Adjusted R2s for the remaining scales ranged from .80 to bound to equivalence because they are calculated on the
.91 (Mdn = .84). same data and share any systematic error. For the domains,
Individual items are relatively unreliable, and multiple corrected part-total correlations (in which the brief versions

Published 2007, Hogrefe & Huber Publishers Journal of Individual Differences 2007; Vol. 28(3):116–128
120 R.R. McCrae & P.T. Costa, Jr.: Brief NEO-PI-3s

Table 2. Coefficient α for NEO-PI-3FH Scales


Form S Form R
NEO-PI-3FH Scale Adolescent Adult MS Adolescent Adult MS
N: Neuroticism .85 .88 .82 .86 .88 .83
E: Extraversion .82 .80 .79 .82 .82 .79
O: Openness .82 .81 .72 .79 .79 .71
A: Agreeableness .78 .82 .74 .84 .87 .81
C: Conscientiousness .86 .86 .87 .89 .88 .89
Mdn .82 .82 .79 .84 .87 .81
N1: Anxiety .67 .73 .61 .69 .72 .57
N2: Angry Hostility .67 .70 .70 .73 .81 .71
N3: Depression .70 .78 .56 .70 .73 .57
N4: Self-Consciousness .58 .62 .40 .66 .62 .42
N5: Impulsiveness .48 .55 .24 .54 .57 .38
N6: Vulnerability .61 .67 .51 .55 .67 .60
E1: Warmth .59 .60 .59 .57 .67 .57
E2: Gregariousness .68 .69 .70 .66 .75 .64
E3: Assertiveness .66 .66 .68 .63 .69 .63
E4: Activity .43 .49 .34 .43 .57 .42
E5: Excitement Seeking .48 .46 .30 .56 .52 .36
E6: Positive Emotions .60 .67 .59 .60 .63 .54
O1: Fantasy .63 .62 .60 .59 .62 .61
O2: Aesthetics .72 .75 .64 .68 .75 .66
O3: Feelings .63 .61 .41 .61 .64 .54
O4: Actions .26 .33 .11 .38 .40 .11
O5: Ideas .64 .65 .54 .70 .72 .53
O6: Values .39 .49 .20 .33 .42 –.07
A1: Trust .65 .69 .29 .69 .75 .48
A2: Straightforwardness .59 .67 .56 .67 .74 .54
A3: Altruism .57 .65 .61 .64 .67 .64
A4: Compliance .58 .58 .55 .63 .70 .57
A5: Modesty .59 .57 .48 .63 .66 .58
A6: Tender-Mindedness .52 .55 .38 .64 .68 .51
C1: Competence .50 .58 .57 .63 .65 .66
C2: Order .76 .76 .80 .81 .82 .76
C3: Dutifulness .47 .51 .43 .60 .55 .50
C4: Achievement Striving .65 .65 .57 .68 .65 .62
C5: Self-Discipline .64 .69 .67 .71 .75 .70
C6: Deliberation .65 .63 .54 .69 .66 .54
Mdn .61 .64 .56 .64 .67 .57
Note. MS = Middle School sample.

were correlated with the sum of the second-half items) PI-R structure (McCrae, Zonderman, Costa, Bond, &
ranged from .76 to .95, with median values of .85, .86, and Paunonen, 1996); we then evaluated variable, factor, and
.84 for Adolescent, Adult, and Middle-School samples, re- total congruence coefficients. As summarized in Table 4,
spectively. the brief scales retained the structure of the parent instru-
A first test of the construct validity of the First Half ment. All but one of the factor congruences exceeded .90,
scales examines the internal convergent and discriminant and all were higher than the .85 criterion of factor repli-
validity of the facet scales via factor analysis. In each cability (Haven & ten Berge, 1977). Of 180 variable con-
sample and form, we extracted five factors and rotated gruence coefficients, all but seven were higher than 95%
them to maximum similarity to the normative adult NEO- of rotations from random data (McCrae et al., 1996). Of

Journal of Individual Differences 2007; Vol. 28(3):116–128 Published 2007, Hogrefe & Huber Publishers
R.R. McCrae & P.T. Costa, Jr.: Brief NEO-PI-3s 121

Table 3. Equivalence coefficients for NEO-PI-3FH vs. NEO-PI-3 scales


Form S Form R
NEO-PI-3 Scale Adolescent Adult MS Adolescent Adult MS
N: Neuroticism .97 .97 .96 .97 .97 .95
E: Extraversion .97 .97 .96 .97 .97 .94
O: Openness .97 .97 .96 .96 .97 .94
A: Agreeableness .96 .97 .94 .97 .97 .96
C: Conscientiousness .97 .97 .96 .98 .97 .97
Mdn .97 .97 .96 .97 .97 .95
N1: Anxiety .92 .94 .91 .92 .94 .90
N2: Angry Hostility .91 .90 .90 .92 .94 .92
N3: Depression .91 .93 .90 .91 .93 .89
N4: Self-Consciousness .91 .91 .88 .92 .89 .87
N5: Impulsiveness .91 .89 .85 .89 .91 .85
N6: Vulnerability .92 .92 .90 .91 .94 .91
E1: Warmth .93 .94 .92 .92 .95 .90
E2: Gregariousness .92 .92 .90 .92 .93 .89
E3: Assertiveness .92 .93 .93 .92 .93 .91
E4: Activity .87 .90 .85 .85 .91 .89
E5: Excitement Seeking .89 .89 .87 .89 .89 .84
E6: Positive Emotions .91 .92 .90 .91 .92 .92
O1: Fantasy .92 .91 .91 .91 .92 .89
O2: Aesthetics .95 .95 .93 .95 .96 .92
O3: Feelings .92 .91 .91 .89 .92 .89
O4: Actions .85 .86 .83 .87 .87 .80
O5: Ideas .94 .94 .90 .94 .94 .90
O6: Values .86 .90 .84 .86 .88 .82
A1: Trust .92 .94 .84 .93 .94 .85
A2: Straightforwardness .89 .92 .91 .93 .94 .91
A3: Altruism .92 .92 .90 .92 .93 .90
A4: Compliance .91 .91 .92 .92 .92 .89
A5: Modesty .92 .91 .88 .92 .92 .91
A6: Tender-Mindedness .89 .89 .89 .91 .91 .91
C1: Competence .91 .91 .89 .91 .91 .92
C2: Order .94 .93 .92 .95 .95 .93
C3: Dutifulness .89 .91 .87 .91 .91 .92
C4: Achievement Striving .91 .90 .87 .92 .91 .91
C5: Self-Discipline .94 .92 .93 .94 .94 .93
C6: Deliberation .91 .91 .87 .91 .92 .88
Mdn .91 .91 .90 .92 .92 .90
Note. MS = Middle School sample.

Published 2007, Hogrefe & Huber Publishers Journal of Individual Differences 2007; Vol. 28(3):116–128
122 R.R. McCrae & P.T. Costa, Jr.: Brief NEO-PI-3s

Table 4. Congruence coefficients from NEO-PI-3 First Half factor analyses


Factor Variable
Sample N E O A C Total Congruences ≥ .86
Form S (Self-Reports)
Adolescent .98 .96 .93 .96 .97 .96 30
Adult .97 .98 .97 .98 .96 .97 29
Middle school .97 .94 .90 .94 .96 .95 28
Form R (Observer Ratings)
Adolescent .97 .96 .92 .94 .96 .95 30
Adult .97 .98 .93 .97 .97 .96 29
Middle school .94 .95 .87 .90 .94 .92 27
Note. Variable congruence coefficients ≥ .86 are higher than 95% of rotations from random data. Variable congruences below .86 were N1 (.85)
and O6 (.75) in Middle School Form S; O4 (.75), O6 (.72) and A6 (.79) in Middle School Form R; O6 (.83) in Adult Form S; and O6 (.79) in
Adult Form R.

Figure 1. NEO-PI-3 First Half Form S domain and facet scores plotted against combined-sex adult or adolescent NEO-PI-3
norms (McCrae, Martin, & Costa, 2005). The five domain scores are given on the left; toward the right the facet scales
are grouped by factor.

particular interest are the factor loadings of the two vari- on the O factor ranging from .27 to .59, with a median of
ables with the most problematic internal consistency, O4: .41. In general, the lowest factor loadings for these two
Actions and O6: Values. Across the six factor analyses facets (.27 to .34) were found in the Middle School sam-
summarized in Table 3, these variables showed loadings ple.

Journal of Individual Differences 2007; Vol. 28(3):116–128 Published 2007, Hogrefe & Huber Publishers
R.R. McCrae & P.T. Costa, Jr.: Brief NEO-PI-3s 123

Table 5. Cross-observer validity and comparative validity of NEO-PI-3 First Half scales
Correlation Relative Variancea
Adolescent Adult Adolescent Adult
b c b c
NEO-PI-3FH Scale S R S R S R S R
N: Neuroticism .42 .36 .56 .55 1.19 .89 1.04 .99
E: Extraversion .58 .59 .61 .63 .95 .97 .89 .94
O: Openness .57 .54 .57 .56 1.09 .99 1.00 .97
A: Agreeableness .43 .45 .56 .55 .86 .91 .93 .92
C: Conscientiousness .51 .52 .51 .52 .92 .96 .95 .98
Mdn .51 .52 .56 .55 .95 .96 .95 .97
N1: Anxiety .44 .38 .48 .49 1.24 .90 .91 .95
N2: Angry Hostility .33 .32 .55 .49 .91 .85 1.06 .84
N3: Depression .38 .31 .51 .50 1.10 .72 1.01 .96
N4: Self-Consciousness .41 .40 .39 .39 1.10 1.01 .82 .84
N5: Impulsiveness .21 .19 .38 .40 1.02 .79 .82 .91
N6: Vulnerability .27 .32 .45 .45 .85 1.12 .88 .88
E1: Warmth .41 .45 .53 .55 .89 1.06 .84 .88
E2: Gregariousness .44 .45 .55 .57 .75 .78 .81 .86
E3: Assertiveness .43 .44 .53 .55 .98 1.01 .90 .97
E4: Activity .45 .44 .44 .49 .78 .73 .69 .86
E5: Excitement Seeking .43 .54 .63 .65 .63 .97 .85 .90
E6: Positive Emotions .48 .46 .39 .42 .96 .88 .75 .89
O1: Fantasy .36 .37 .37 .39 1.07 1.14 .86 .93
O2: Aesthetics .57 .56 .54 .55 .96 .90 .92 .95
O3: Feelings .46 .36 .39 .40 1.00 .59 .80 .85
O4: Actions .33 .24 .43 .39 1.30 .68 1.07 .86
O5: Ideas .45 .45 .55 .54 .90 .89 .88 .83
O6: Values .36 .40 .46 .47 .69 .86 .82 .87
A1: Trust .22 .21 .40 .43 .71 .68 .81 .95
A2: Straightforwardness .22 .20 .34 .35 1.00 .79 .79 .84
A3: Altruism .31 .33 .47 .44 .83 .95 .81 .71
A4: Compliance .47 .46 .55 .50 .84 .79 .97 .81
A5: Modesty .32 .39 .43 .45 .68 1.04 .81 .88
A6: Tender-Mindedness .24 .26 .42 .46 1.09 1.27 .76 .94
C1: Competence .32 .31 .39 .38 .87 .82 .84 .80
C2: Order .48 .47 .62 .61 1.08 1.06 .92 .89
C3: Dutifulness .44 .41 .42 .41 .93 .80 .89 .87
C4: Achievement Striving .38 .40 .41 .44 .76 .86 .73 .85
C5: Self-Discipline .39 .42 .42 .43 .85 .98 .83 .87
C6: Deliberation .43 .48 .42 .41 .70 .87 .92 .87
Mdn .40 .40 .44 .45 .91 .88 .84 .87
Note. N = 180 for Adolescent sample, 532 for Adult sample. aRelative variance = (variance in criterion accounted for by First Half scale)/(variance
in criterion accounted for by full scale; see McCrae, Costa, & Martin, 2005, McCrae, Martin, & Costa, 2005). Calculations based on correlations
to three decimal places. bCorrelation of First Half Form S scale with Form R scale. cCorrelation of First Half Form R scale with Form S scale.

Published 2007, Hogrefe & Huber Publishers Journal of Individual Differences 2007; Vol. 28(3):116–128
124 R.R. McCrae & P.T. Costa, Jr.: Brief NEO-PI-3s

Cross-Observer Correlations norms to interpret observer-rated personality scores of or by


Middle School-aged boys and girls.
Standard deviations of NEO-PI-3FH scales are consistent-
Table 5 reports cross-observer validity coefficients for the
ly larger than those of the full scales. For example, in the
NEO-PI-3FH scales. The first four data columns report the
Adult Form S data, standard deviations are from 5% to 26%
correlation of each First Half scale with the full scale of the
larger than the corresponding NEO-PI-3 scale. This is a pure-
alternative form. For example, the first value tabled, .42, is
ly statistical phenomenon, the result of the fact that the two
the correlation of the 24-item self-reported First Half N
“halves” of the doubled brief scale are perfectly correlated
scale with the 48-item observer-rated N scale in the Ado-
and, thus, provide a slightly inflated estimate of the variance
lescent sample. All correlations are statistically significant,
of a true eight-item scale. The effect of this artifact is that
and they are comparable in magnitude to most cross-ob-
when First Half scales are interpreted with full scale norms,
server correlations (e.g., McCrae et al., 2004).
deviation from the mean will be somewhat exaggerated, es-
A more telling comparison, however, is with cross-ob- pecially for very high and very low scores. For example, an
server correlations in the same samples using the full NEO- individual with the maximum possible score, 32, on E1:
PI-3 scales. Those values have been reported previously Warmth will have an estimated T-score of 72 using NEO-PI-3
(McCrae, Costa, et al., 2005; McCrae, Martin, et al, 2005); norms; the true T-score, based on NEO-PI-3FH norms, is 70.
in the fifth through eighth data columns we report the For the purposes for which the NEO-PI-3FH is intended, such
squared ratio of the First Half validities divided by the full minor distortions do not appear to be a problem.1
NEO-PI-3 validities. These values reflect the proportion of
the variance in the criterion that the brief scale predicts
relative to the full scale. The median values show that about
95% of the cross-observer validity is retained by the brief
domain scales, and 85%-90% of the cross-observer validity Discussion
is retained by the brief facet scales. (Values above 1.0 sug-
gest that, in this sample, the First Half scale is a better pre- This article has examined two brief forms of a new version
dictor than the corresponding full scale.) of the NEO-PI-R, the NEO-PI-3. How adequate was their
development? Smith, McCarthy, and Anderson (2000)
identified a number of “sins” commonly committed in de-
veloping short forms of tests, and the eight relevant sins
Means and Standard Deviations have been avoided or addressed here:
1. The original instruments are well-validated;
Figure 1 shows NEO-PI-3FH Form S domain and facet scales 2. content coverage has been systematically retained in the
plotted using norms from the full NEO-PI-3. Scores for the NEO-PI-3FH, where each facet scale is represented by
complete Adult sample were standardized as T-scores (M = four items;
50, SD = 10) using combined-sex means and standard devi- 3. internal consistency of domains scores is generally
ations for the Adult (21+) NEO-PI-3; scores for the Adoles- above the .70 criterion Smith et al. recommend, although
cent and Middle School samples were standardized using facet scales have lower αs, and users are cautioned about
combined-sex Adolescent (14–20) NEO-PI-3 norms. (Recall scales with notably lower reliability;
that the NEO-PI-3 norms are derived from the Adult and 4. equivalence with the full scales has been reported, al-
Adolescent samples examined here; for these groups, this though it is based on the same data, so users are cau-
analysis is a comparison of scales, not of samples.) As the tioned that this is an upper-bound estimate;
figure makes clear, means for the full version (where all T 5. factor structure has been retained;
values = 50) and doubled brief version are very similar, espe- 6. in the development of the NEO-FFI-3, items were select-
cially for adults and older adolescents, suggesting that NEO- ed in part (although not systematically) to represent the
PI-3 norms can meaningfully be used to interpret NEO-PI- full range of facets on each factor (as Smith et al. noted);
3FH scores both for groups and for individuals. Middle 7. cross-observer validity was demonstrated in independ-
School-aged children tend to score lower than older adoles- ent samples; and
cents on O, and that age difference is preserved in the NEO- 8. information is provided from which users can evaluate
PI-3FH. A similar plot was examined for Form R NEO-PI- the loss of validity in comparison to the savings in time.
3FH scales; again, very similar values were found for adults Test users share responsibility with test developers for
and older adolescents. Middle School-aged targets scored evaluating alternative versions of a test; this article pro-
substantially lower on N and higher on A and C than later vides the data needed to do that.
adolescent targets, which may reflect a rating bias in younger
raters (Costa et al., in press). These differences would need to The NEO-FFI-3 is a relatively minor revision of the widely
be borne in mind when using Adolescent Form R NEO-PI-3 used NEO-FFI. It has slightly better psychometric proper-

1 NEO-PI-3 First Half norms for the standard age and gender groups are available from the first author.

Journal of Individual Differences 2007; Vol. 28(3):116–128 Published 2007, Hogrefe & Huber Publishers
R.R. McCrae & P.T. Costa, Jr.: Brief NEO-PI-3s 125

ties than the original scale and eliminates a few items that during the second half of the test can still be scored on
presented difficulties in comprehension to younger respon- responses to the first half.
dents (McCrae & Costa, 2004). Data in the present article The NEO-PI-3FH is not intended for clinical assessment
continue to support its use when only global information in any age group. Clinicians make decisions that can have
about the five broad factors is needed. The instrument a profound impact on their patients’ lives and well-being,
works very well in adults and in older adolescents, and al- and they should base those decisions on the best possible
most as well in Middle School-aged respondents. and most reliable information. That said, it is sometimes
The second instrument is completely new. In the NEO- the case that the best possible information is whatever the
PI-3FH, the four best items from each of the 30 facets of patient is willing to give, and a depressed or demoralized
the NEO-PI-3 are used to estimate full scale facet scores. patient may leave questionnaires unfinished. The NEO-PI-
This instrument would allow researchers to cut in half the 3FH scoring provides a way of salvaging some useful in-
time and the number of items they need to assess the more formation from an incomplete clinical assessment.
specific traits of the FFM, allowing them to conduct re-
search on more difficult populations, or to collect data on
other variables.2 There are costs associated with this Authors’ Note
choice: Internal consistency is lower, predictive validity is
reduced by about 10%, and normative information is less Robert R. McCrae and Paul T. Costa, Jr., both receive
accurate than for the full scales. Researchers need to bal- royalties from the NEO-PI-R. This research was support-
ance these considerations carefully in choosing an instru- ed in part by the Intramural Research Program of the
ment and designing a study. For example, relative validity NIH, National Institute on Aging. The NEO-PI-R profile
of different scales ought to be part of a power analysis: Use form from the Revised NEO Personality Inventory by
of a brief scale might lead to a 10% reduction in a margin- Paul T. Costa, Jr. and Robert R. McCrae is reproduced by
ally-significant correlation, making it nonsignificant – and special permission of the publisher, Psychological As-
the research unpublishable. Instead of viewing it as a sessment Resources, Inc., 16204 North Florida Avenue,
“quick-and-dirty” alternative to the NEO-PI-3, our hope is Lutz, FL 33549, USA. Copyright © 1978, 1985, 1989,
that the NEO-PI-3FH will be seen as an upgrade to the 1991, 1992 by Psychological Assessment Resources, Inc.
NEO-FFI-3, giving detailed facet information as well as (PAR). Further reproduction is prohibited without per-
more reliable estimates of the five factors by adding a mere mission of PAR.
60 items.
Survey researchers are unlikely to include a 120-item
questionnaire in their interviews, but they, and other re-
searchers, might have an interest in a subset of traits. Given,
say, just 12 items to assess personality, it might make sense References
to pick three facets and administer the First Half items,
which could be licensed by the publisher. Becker, G. (2006). NEO-FFI scores in college men and women:
One impetus to the development of a brief NEO-PI-3 A view from McDonald’s unified treatment of test theory.
was concern that younger adolescents would be unwilling Journal of Research in Personality, 40, 911–941.
or unable to complete a 240-item questionnaire, although Costa, P.T., Jr., & McCrae, R.R. (1985). The NEO Personality
they might respond to one half that long. It is not yet clear Inventory manual. Odessa, FL: Psychological Assessment Re-
whether the NEO-PI-3FH ought to be recommended for sources.
this age group. Our experience showed that many 12- and Costa, P.T., Jr., & McCrae, R.R. (1989). The NEO-PI/NEO-FFI
13-year-olds can and will complete the full inventory (Cos- manual supplement. Odessa, FL: Psychological Assessment
Resources.
ta, McCrae, et al., 2005), and also showed that it is precisely
Costa, P.T., Jr., & McCrae, R.R. (1992). Revised NEO Personality
this age group that most needs the longer version, to com-
Inventory (NEO-PI-R) and NEO Five-Factor Inventory (NEO-
pensate for the generally lower psychometric quality of the
FFI) professional manual. Odessa, FL: Psychological Assess-
data they produce. However, the Middle School sample ex- ment Resources.
amined here consisted chiefly of academically superior stu- Costa, P.T., Jr., McCrae, R.R., & Martin, T.A. (in press). Incipient
dents, and the briefer First Half might be required in a more adult personality: The NEO-PI-3 in middle school-aged chil-
representative population. We suggest that researchers in- dren. British Journal of Developmental Psychology.
terested in assessing personality in 12- and 13-year-olds Digman, J.M. (1990). Personality structure: Emergence of the
begin by administering the full NEO-PI-3, and turn to the five-factor model. Annual Review of Psychology, 41, 417–440.
brief version only if there are substantial problems with Egan, V., Deary, I., & Austin, E. (2000). The NEO-FFI: Emerging
participant cooperation. Note that any student who quits British norms and an item-level analysis suggest N, A, and C

2 The NEO-PI-3FH is not designed as a cheaper alternative to the full NEO-PI-3; there are no plans to publish it at a discounted rate. To do
so would be to offer a financial incentive to use a less-than-optimal instrument, and that does not seem to be in the interest of advancing
personality psychology.

Published 2007, Hogrefe & Huber Publishers Journal of Individual Differences 2007; Vol. 28(3):116–128
126 R.R. McCrae & P.T. Costa, Jr.: Brief NEO-PI-3s

are more reliable than O and E. Personality and Individual Mullins-Sweatt, S.N., Jamerson, J.E., Samuel, D.B., Olson, D.R.,
Differences, 29, 907–920. & Widiger, T.A. (2006). Psychometric properties of an abbre-
Frazier, T.W., Naugle, R.I., & Haggerty, K.A. (2006). Psychomet- viated instrument of the five-factor model. Assessment, 13,
ric adequacy and comparability of the short and full forms of 119–137.
the Personality Assessment Inventory. Psychological Assess- Paunonen, S.V., & Ashton, M.C. (2001). Big-Five factors and fac-
ment, 18, 324–333. ets and the prediction of behavior. Journal of Personality and
Gosling, S.D., Rentfrow, P.J., & Swann, W.B., Jr. (2003). A very Social Psychology, 81, 524–539.
brief measure of the Big-Five personality domains. Journal of Rammstedt, B., & John, O.P. (2005). Kurzversion des Big Five
Research in Personality, 37, 504–528. Inventory (BFI-K): Entwicklung und Validierung eines öko-
Haven, S., & ten Berge, J.M.F. (1977). Tucker’s coefficient of con- nomischen Inventars zur Erfassung der fünf Faktoren der Per-
gruence as a measure of factorial invariance: An empirical study sönlichkeit [Short version of the Big Five Inventory (BFI-K):
(Heymans Bulletin No. 290 EX). University of Groningen. Development and validation of an economic inventory for as-
Herzberg, P.Y., & Brahler, E. (2006). Assessing the Big-Five per- sessment of the five factors of personality]. Diagnostica, 51,
sonality domains via short forms: A cautionary note and a pro- 195–206.
posal. European Journal of Psychological Assessment, 22, Reise, S.P., & Henson, J.M. (2000). Computerization and adaptive
139–148. administration of the NEO-PI-R. Assessment, 7, 347–364.
McCrae, R.R., & Costa, P.T., Jr. (1989). Rotation to maximize the Saucier, G. (1998). Replicable item-cluster subcomponents in the
construct validity of factors in the NEO Personality Inventory. NEO Five-Factor Inventory. Journal of Personality Assess-
Multivariate Behavioral Research, 24, 107–124. ment, 70, 263–276.
McCrae, R.R., & Costa, P.T., Jr. (2004). A contemplated revision Shock, N.W., Greulich, R.C., Andres, R., Arenberg, D., Costa,
of the NEO Five-Factor Inventory. Personality and Individual P.T., Jr., Lakatta, E.G., et al. (1984). Normal human aging: The
Differences, 36, 587–596. Baltimore Longitudinal Study of Aging (NIH Publication
McCrae, R.R., Costa P.T., Jr., & Martin, T.A. (2005). The NEO- No. 84–2450). Bethesda, MD: National Institutes of Health.
PI-3: A more readable Revised NEO Personality Inventory. Smith, G.T., McCarthy, D.M., & Anderson, K.G. (2000). On the
Journal of Personality Assessment, 84, 261–270. sins of short-form development. Psychological Assessment,
McCrae, R.R., Costa, P.T., Jr., Martin, T.A., Oryol, V.E., Ruka- 12, 102–111.
vishnikov, A.A., Senin, I.G., et al. (2004). Consensual valida- Trobst, K.K., Wiggins, J.S., Costa, P.T., Jr., Herbst, J.H., McCrae,
tion of personality traits across cultures. Journal of Research R.R., & Masters, H.L., III. (2000). Personality psychology and
in Personality, 38, 179–201. problem behaviors: HIV risk and the five-factor model. Jour-
McCrae, R.R., Martin, T.A., & Costa, P.T., Jr. (2005). Age trends nal of Personality, 68, 1233–1252.
and age norms for the NEO Personality Inventory-3 in adoles-
cents and adults. Assessment, 12, 363–373.
McCrae, R.R., Zonderman, A.B., Costa, P.T., Jr., Bond, M.H., &
Paunonen, S.V. (1996). Evaluating replicability of factors in Robert R. McCrae
the Revised NEO Personality Inventory: Confirmatory factor
Box #03
analysis versus Procrustes rotation. Journal of Personality and
Gerontology Research Center
Social Psychology, 70, 552–566.
5600 Nathan Shock Drive
Meyer, G.J., Finn, S.E., Eyde, L.D., Kay, G.G., Moreland, K.L.,
Baltimore, MD 21224-6825
Dies, R.R., et al. (2001). Psychological testing and psycholog-
USA
ical assessment. American Psychologist, 56, 128–165.
E-mail mccraej@grc.nia.nih.gov
Morey, L. (1991). Personality Assessment Inventory: Profession-
al manual. Odessa, FL: Psychological Assessment Resources.

Journal of Individual Differences 2007; Vol. 28(3):116–128 Published 2007, Hogrefe & Huber Publishers
R.R. McCrae & P.T. Costa, Jr.: Brief NEO-PI-3s 127

Appendix
Means and Standard Deviations for NEO-FFI-3 Scales in Different Age Groups
Men Women Combined
NEO-FFI-3 Scale Mean SD Mean SD Mean SD
Adolescents Aged 14–20
Form S (242, 258, 500)
N: Neuroticism 23.3 7.3 24.9 7.0 24.1 7.2
E: Extraversion 29.7 6.3 32.4 6.2 31.1 6.4
O: Openness to Experience 29.1 6.8 31.9 6.1 30.6 6.6
A: Agreeableness 28.2 5.8 30.3 5.6 29.3 5.8
C: Conscientiousness 27.9 6.8 29.5 7.1 28.7 7.0
Form R (211, 254, 465)
N: Neuroticism 22.8 6.3 25.5 7.2 24.3 6.9
E: Extraversion 28.3 5.6 30.6 6.9 29.6 6.4
O: Openness to Experience 25.5 6.2 27.3 6.3 26.5 6.3
A: Agreeableness 26.6 6.8 27.4 7.2 27.0 7.1
C: Conscientiousness 25.6 8.0 27.5 8.3 26.6 8.2
Younger Adults, 21–30
Form S (97, 121, 218)
N: Neuroticism 19.7 7.3 23.5 8.0 21.8 7.9
E: Extraversion 28.9 5.2 30.3 6.1 29.6 5.7
O: Openness to Experience 28.0 7.4 30.9 5.9 29.6 6.7
A: Agreeableness 27.9 5.7 31.8 6.5 30.1 6.4
C: Conscientiousness 30.9 6.4 32.0 7.1 31.6 6.8
Form R (112, 122, 234)
N: Neuroticism 19.8 7.1 23.6 7.1 21.8 7.3
E: Extraversion 29.3 6.2 29.7 5.5 29.5 5.8
O: Openness to Experience 25.8 5.9 27.4 5.8 26.7 5.9
A: Agreeableness 29.0 6.5 30.2 6.9 29.7 6.8
C: Conscientiousness 30.3 7.8 31.8 7.4 31.1 7.6
Older Adults, 31+
Form S (182, 235, 417)
N: Neuroticism 18.8 6.9 21.5 7.8 20.3 7.6
E: Extraversion 26.3 6.3 28.4 6.2 27.5 6.3
O: Openness to Experience 26.9 5.6 28.5 6.1 27.8 6.0
A: Agreeableness 31.1 5.4 34.7 5.1 33.1 5.5
C: Conscientiousness 32.9 5.7 33.2 6.2 33.0 6.0
Form R (189, 226, 415)
N: Neuroticism 18.2 7.7 21.9 7.2 20.2 7.7
E: Extraversion 25.4 7.4 28.3 5.9 26.9 6.8
O: Openness to Experience 24.1 7.0 25.9 5.6 25.1 6.3
A: Agreeableness 30.6 7.6 32.0 6.3 31.4 7.0
C: Conscientiousness 33.6 7.3 33.5 6.9 33.5 7.1

Published 2007, Hogrefe & Huber Publishers Journal of Individual Differences 2007; Vol. 28(3):116–128
128 R.R. McCrae & P.T. Costa, Jr.: Brief NEO-PI-3s

Men Women Combined


NEO-FFI-3 Scale Mean SD Mean SD Mean SD
Adults, 21+
Form S (279, 356, 635)
N: Neuroticism 19.1 7.1 22.2 7.9 20.8 7.7
E: Extraversion 27.2 6.1 29.0 6.2 28.2 6.2
O: Openness to Experience 27.3 6.3 29.3 6.2 28.4 6.3
A: Agreeableness 30.0 5.7 33.7 5.7 32.1 6.0
C: Conscientiousness 32.2 6.0 32.8 6.5 32.5 6.3
Form R (301, 348, 649)
N: Neuroticism 18.8 7.5 22.5 7.2 20.8 7.6
E: Extraversion 26.8 7.2 28.7 5.8 27.9 6.6
O: Openness to Experience 24.7 6.6 26.4 5.7 25.6 6.2
A: Agreeableness 30.0 7.3 31.4 6.6 30.8 6.9
C: Conscientiousness 32.4 7.6 32.9 7.1 32.6 7.3
All Ages, 14+
Form S (521, 614, 1,135)
N: Neuroticism 21.0 7.5 23.3 7.7 22.3 7.7
E: Extraversion 28.4 6.3 30.4 6.4 29.5 6.5
O: Openness to Experience 28.1 6.6 30.4 6.3 29.4 6.5
A: Agreeableness 29.2 5.8 32.3 5.9 30.8 6.1
C: Conscientiousness 30.2 6.7 31.4 7.0 30.8 6.9
Form R (512, 602, 1,114)
N: Neuroticism 20.4 7.3 23.8 7.3 22.2 7.5
E: Extraversion 27.5 6.6 29.5 6.4 28.6 6.6
O: Openness to Experience 25.0 6.5 26.8 6.0 26.0 6.3
A: Agreeableness 28.6 7.3 29.7 7.1 29.2 7.2
C: Conscientiousness 29.6 8.5 30.6 8.1 30.1 8.3
Childred, Aged 12–13
Form S (101, 101, 202)
N: Neuroticism 23.4 6.7 24.8 7.4 24.1 7.1
E: Extraversion 31.6 5.7 35.3 6.0 33.5 6.1
O: Openness to Experience 25.8 5.8 28.5 5.9 27.1 6.0
A: Agreeableness 28.4 6.4 32.0 5.6 30.2 6.3
C: Conscientiousness 27.3 7.9 29.0 7.4 28.1 7.7
Form R (102, 120, 222)
N: Neuroticism 18.7 6.3 20.4 6.6 19.6 6.5
E: Extraversion 32.0 5.6 33.5 6.6 32.8 6.2
O: Openness to Experience 23.1 5.1 25.5 5.8 24.4 5.6
A: Agreeableness 29.0 6.1 31.0 7.4 30.0 6.9
C: Conscientiousness 28.7 7.3 28.9 8.4 28.8 7.9
Note. N values for men, women, and combined sex are given in parentheses.

Journal of Individual Differences 2007; Vol. 28(3):116–128 Published 2007, Hogrefe & Huber Publishers

View publication stats

You might also like