You are on page 1of 11

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

net/publication/7834763

The NEO–PI–3: a more readable revised NEO personality inventory

Article in Journal of Personality Assessment · June 2005


DOI: 10.1207/s15327752jpa8403_05 · Source: PubMed

CITATIONS READS
502 15,298

3 authors, including:

Paul Costa Thomas A. Martin


Duke University Medical Center Susquehanna University
454 PUBLICATIONS 94,100 CITATIONS 28 PUBLICATIONS 2,445 CITATIONS

SEE PROFILE SEE PROFILE

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

GWAS of Personality View project

Lifespan and Intergenerational Effects of Childhood Malnutrition View project

All content following this page was uploaded by Paul Costa on 13 October 2015.

The user has requested enhancement of the downloaded file.


JOURNAL OF PERSONALITY ASSESSMENT, 84(3), 261–270
MCCRAE,
NEO–PI–3
COSTA, MARTIN

The NEO–PI–3: A More Readable Revised NEO


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

Thomas A. Martin
Department of Psychology
Susquehanna University

Use of the Revised NEO Personality Inventory (NEO–PI–R; Costa & McCrae, 1992) in adoles-
cent samples has shown that a few respondents have difficulty with a subset of items. We identi-
fied 30 items that were not understood by at least 2% of adolescent respondents and 18 addi-
tional items with low item-total correlations, and we wrote 2 trial replacement items for each.
We used self-report and observer rating data from 500 respondents aged 14 to 20 to select re-
placement items. The modified instrument retained the intended factor structure and showed
slightly better internal consistency, cross-observer agreement, and readability (Flesch–Kincaid
grade level = 5.3). The NEO–PI–3 appears to be useful in high school and college samples and
may have wider applicability to adults as well.

The NEO Personality Inventory (NEO–PI; Costa & McCrae, found that many of the same words were not understood by
1985) was designed for use by adults. Shortly after its publi- their respondents, and 15 of these 30 difficult items caused
cation, it became clear that college students could also use problems for Belgian children taking a Flemish version of
the instrument but required separate norms (Costa & Mc- the NEO–PI–R (De Fruyt, Mervielde, Hoekstra, & Rolland,
Crae, 1989). Recent studies (Baker & Victor, 2003; McCrae 2000).
et al., 2002) have shown that the Revised NEO–PI Even with these problematic items, Baker and Victor
(NEO–PI–R; Costa & McCrae, 1992) can be used in samples (2003) reported a clear replication of the adult five-factor
as young as 10 but that some younger respondents have diffi- structure, although some empirical weaknesses were noted.
culty understanding certain items. McCrae et al. adminis- Similarly, Allik, Laidra, Realo, & Pullmann (2004) reported
tered the NEO–PI–R to high school students with instruc- that the adult factor structure of the Estonian NEO
tions to leave blank any item they did not understand. For 210 Five-Factor Inventory was less clearly replicated in a
of the 240 items, at least 98% of respondents answered; the subsample of 12-year-olds than in older adolescents. These
remaining 30 items included terms such as fastidious, ad- limitations on factor replicability might be due to true devel-
here, and lackadaisical, which even some adults might find opmental changes in factor structure or to cognitive limita-
difficult. There is evidence that these 30 items are reliably tions in self-understanding that distort the accuracy of
problematic: Baker and Victor (2003), in a very diverse, pre- self-reports in younger adolescents. However, it is also pos-
dominantly Black U.S. sample with a mean age of 13.0 years, sible that younger children simply did not understand the vo-
cabulary in some of the items and that a simplified version of
the NEO–PI–R might produce a closer approximation to the
NEO–PI–R items in the Appendix are reproduced by special per- adult structure. Testing that hypothesis requires a modified
mission of the Publisher, Psychological Assessment Resources, version of the NEO–PI–R, and that more readable version
Inc., 16204 North Florida Avenue, Lutz, Florida 33549, from the
NEO Personality Inventory–Revised by Paul T. Costa, Jr., and Rob-
could make the instrument appropriate for a wider range of
ert R. McCrae, Copyright 1978, 1985, 1989, 1992 by PAR, Inc. Fur- respondents. In this article, we describe the development and
ther reproduction is prohibited without permission from PAR, Inc. preliminary validation of the NEO–PI–3.
262 MCCRAE, COSTA, MARTIN
Any change in an established test requires a balance be- states, although most were from Pennsylvania.
tween improving psychometrics and retaining known valid- Characteristics of the sample are given in Table 1. The sam-
ity. Because the NEO–PI–R has worked well in a variety of ple was predominantly White. High academic achievers
contexts, we chose a conservative approach in which the were overrepresented in this sample in which the majority
great majority of items were retained verbatim. However, the expected to obtain a postbaccalaureate degree. Most (474)
revision process also offered an opportunity to improve a listed English as their best language, although 82 respon-
small number of items that performed poorly for reasons dents also spoke other languages with their families.
other than readability. In the self-report form of the Potential participants were contacted by one of 53 under-
NEO–PI–R, 14 of the 30 facet scales had coefficient alphas graduate research assistants from two liberal arts colleges lo-
of .70 or less (Costa & McCrae, 1992). For these facets, we cated in the eastern and southern United States. Most
examined corrected item-total correlations in two samples: participants were acquaintances of the research assistants.
1,959 high school students (McCrae et al., 2002) and 1,492 Research assistants described the study to potential partici-
adults (see McCrae & Costa, 2004). A total of 24 items had pants and if they were under 18 years of age, to a parent as
item-total correlations less than .30 in both samples; 6 of well. If an individual expressed interest in participating (and,
these were among the 30 items already identified as problem- when necessary, the parent agreed), the research assistant
atic. Thus, 48 items, from 20 of the 30 facets, were selected submitted a referral with the identity, age, and contact infor-
for possible replacement in the NEO–PI–3. mation for the participant. Questionnaires were then mailed
One of the unusual features of the NEO–PI–R is the avail- to participants.
ability of two versions, Form S for self-reports and Form R To implement the cross-observer design, an attempt was
for observer ratings. Because these are parallel instruments, made to recruit sibling pairs in which both individuals were
it was necessary to change both, so data on both Form S and aged 14 to 20 and both agreed to rate and be rated by the
Form R were gathered. That design offers two advantages other. Complete data were obtained from 90 pairs. Individ-
over a simple self-report study. First, simultaneous item se- uals not paired with a sibling chose an anonymous target to
lection across the two forms is a kind of cross-validation, re- rate and were asked to specify their relationship to the target.
ducing the possibility of capitalizing on chance. Second, Most (226) rated friends, 79 rated siblings, and 12 rated a
because a subset of participants were sibling pairs who rated spouse or domestic partner.
themselves and each other, the design makes it possible to Approximately once per week, research assistants were
examine cross-observer validity evidence on the NEO–PI–R forwarded a list of all individuals they had referred within the
and NEO–PI–3 in an adolescent sample. past 10 weeks who had not yet returned their questionnaires.
Research assistants were encouraged to contact such partici-
METHOD pants to ask whether they had questions about any aspect of
the task, tell them that they could break their work into ses-
Participants and Procedure sions of whatever number and length were convenient to
them, and encourage them to return the questionnaire. This
Participants were 536 individuals ranging in age from 14 to approach worked well: 62.8% of all questionnaires mailed to
20 who constituted the first phase of a larger study spanning females were returned as were 48.5% of the questionnaires
the adolescent and adult life span. Age and sex were strati- sent to males. Participants and research assistants were each
fied, with overrepresentation of the late adolescent groups paid $20 for completed questionnaires.
because evidence suggests that age changes are most marked
then (Costa & McCrae, 2002). After tabulating the frequency Instruments
of missing items for preliminary analyses, protocols were
screened for validity using the criteria specified in the Re- The NEO–PI–R is a 240-item questionnaire that assesses 30
vised NEO Personality Inventory (NEO–PI–R) and NEO specific traits (or facets), 6 for each of the five basic personal-
Five-Factor Inventory (NEO-FFI) Professional Manual ity dimensions: Neuroticism (N), Extraversion (E), Open-
(Costa & McCrae, 1992) for validity checks, random re- ness to Experience (O), Agreeableness (A), and Conscien-
sponding, and missing items,1 and all further analyses were
conducted on the 500 respondents who had valid protocols
for both Form S and Form R.2 Participants resided in 26 .80 to .88 in the subsample of 36 invalid protocols; for Form R, they
ranged from .75 to .90. Correlations between the Form S domains
and the Subjective Well-Being scale remained significant for
1
The Manual specifies that a protocol is invalid if more than 40 Neuroticism (r = –.34, N = 35, p < .05) and Extraversion (r = .35, N =
items are missing. Because we included an additional 96 items in the 35, p < .05) as did the correlation between Conscientiousness and
test booklet, we increased that value proportionately to 56. self-reported grades (r = .42, N = 35, p < .05). These results suggest
2
In our experience, “invalid” protocols typically retain some va- that whereas researchers may wish to discard questionable proto-
lidity (cf. Carter et al., 2001). After recoding missing values as neu- cols, clinicians may still obtain some useful information from nomi-
tral, coefficient alphas for the NEO–PI–3 Form S scales ranged from nally invalid protocols, cautiously interpreted.
NEO–PI–3 263
TABLE 1 correlated (rs = .42 to .73, all ps < .001), so they were
Characteristics of the Sample summed to form a single Subjective Well-Being (SWB)
Characteristic n %
scale (α = .88). In adult samples, NEO–PI–R N has been
shown to be negatively related to subjective well-being,
Gender whereas E, A, and C are positively related (e.g., McCrae &
Male 242 48.4
Female 258 51.6 Costa, 1991). Well-being was included in this study as a cri-
Age terion variable to provide partial evidence of construct valid-
14 46 9.2 ity of the NEO–PI–3 in an adolescent sample.
15 48 9.6
16 46 9.2
17 64 12.8 Item Selection
18 100 20.0
19 98 19.6
20 98 19.6
The choice of replacement items was based on a number of
Ethnicity considerations. For each of the 20 facet scales that were to be
Hispanic 9 1.8 revised, we first created a subscale criterion score composed
White 423 84.6 of all the items that were to be retained and correlated this
Black 14 2.8
Asian/Pacific Islander 38 7.6 score with the problematic item and its two alternatives in
Other 15 3.0 both Form S and Form R versions. In general, results from
Self-reported grades the two forms were similar. If one of the alternatives was cor-
All As 35 7.0
Mostly As 143 28.6 related at least as strongly with the criterion score as the orig-
Mostly As and Bs 204 40.8 inal item, it was selected. This strategy was questionable for
Mostly Bs 47 9.4 those facets in which several items were to be replaced be-
Mostly Bs and Cs 46 9.2
Mostly Cs 14 2.8 cause the few remaining items formed a less-than-optimal
Mostly Cs and Ds 6 1.2 criterion. We therefore also extracted a single general factor
Highest degree expected from all original and replacement items for each such facet;
High school 2 0.4
College 158 31.6
item selection was then based on factor loadings.
Master’s degree 229 45.8 Three of the replacement items that would have been cho-
Doctoral degree 102 20.4 sen by these methods of item selection were answered by less
Family income
than 98% of the sample and were thus deemed inappropriate
< $40,000 92 18.4
$40,000 to 80,000 175 35.0 for the NEO–PI–3. Some of the replacement items that would
> $80,000 158 31.6 have been selected for N5: Impulsiveness were more strongly
Note. ns do not sum to 500 because of missing data. related to C than to N, and were therefore discarded. In a few
instances, neither of the intended alternatives improved on the
original, but other alternative items written for the same facet
tiousness (C). Items are answered on a 5-point Likert scale did, and these were selected.
ranging from strongly disagree to strongly agree. Scales are
roughly balanced to control for the effects of acquiescence. RESULTS
Data on the reliability and validity of the instrument is sum-
marized in the Manual (Costa & McCrae, 1992). Parallel Preliminary Analyses
self-report (Form S) and observer rating (Form R) versions
have been validated. Respondents had been asked to circle any word or phrase
For each of the 48 items identified as potentially problem- they did not understand. A tally showed that the most diffi-
atic, two alternatives were written. The first was generally a cult words for this sample were fastidious (times circled in
restatement of the item in simpler words; the second, a new Form S = 57), panhandlers (39), permissiveness (33),
item theoretically relevant to the facet. The keying of the two shrewdness (28), lackadaisical (26), methodical (20), aes-
alternatives matched the original item so that scoring keys thetic (14), overindulge (12), and cynical (11). This list
for the NEO–PI–3 would match those of the NEO–PI–R. For shows substantial overlap with the problem words in Baker
this study, the 96 alternative items were interspersed among and Victor (2003).
the 240 original items. After discarding three Form S and seven Form R protocols
In addition to the NEO–PI–R, respondents completed sev- with more than 56 missing items, we counted the number of
eral brief measures of well-being: The Affect Balance Scale missing responses for each item and identified 18 items with
(Bradburn, 1969), a Satisfaction Index that asks about vari- less than 98% response rate; 13 of these overlapped with the 30
ous aspects of life (Costa & McCrae, 1984), the one-item De- identified by McCrae et al. (2002), and 3 were from the pool of
lighted–Terrible Scale (Andrews & Withey, 1976), and the alternative items. The similarity of results in Baker and Victor
5-item Satisfaction With Life Scale (Diener, Emmons, (2003), McCrae et al. (2002), and this study suggests that the
Larsen, & Griffin, 1985). These scales were all inter- most problematic items were accurately identified.
264 MCCRAE, COSTA, MARTIN
To gauge the representativeness of the sample of 500 valid shows a small drop in internal consistency, and both E4,
protocols, we scored the NEO–PI–R and compared mean Activity, and O4, Actions continue to show marginal internal
values with the published college student age norms. Both consistency despite the revision. Columns 5 and 6 show simi-
males and females had means of 50 ± 3 T-score points for all lar results for Form R.
domains and facets. Columns 4 and 7 show equivalence correlations between
the old and new versions of the scales that were revised for
Item Analyses Form S and Form R, respectively. The domain scores are vir-
tually unchanged, with all correlations .98 or higher as would
Acceptable replacements were found for 37 of the 48 items3 be expected given the extensive item overlap and data from a
slated for revision; these are shown in the Appendix. Five of single administration. In self-reports, the facet scale correla-
the 30 items with low readability in McCrae et al. (2002) were tions range from .83 for A6, Tender-Mindedness (in which ½
answered by more than 98% of this sample and had the items were changed) to .98 for O2, Aesthetics, and A5,
psychometric properties that could not be improved on with Modesty. Comparable results are seen for observer ratings.
the available alternatives. However, 25 of the 30 difficult items All these correlations support the view that the new scales
were replaced, which eliminated all of the frequently circled measure equivalent constructs.
words identified in the preliminary analysis with the exception Columns 8 and 9 of Table 3 report cross-observer correla-
of methodical. When Item 70, “I am not a very methodical per- tions, which compare favorably to most studies using adults
son,” was replaced by “I’m not a very orderly or methodical (cf. McCrae et al., 2004). Revisions made little difference at
person,” the item became intelligible to more respondents, and the domain level. At the facet level, notable improvements
the validity of the item improved considerably. Of the 18 addi- were found for N5, Impulsiveness, O6, Tender-Mindedness,
tional items identified as having low item–facet correlations in and C3, Dutifulness. Curiously, validity declined a bit for
previous studies, 12 were replaced, and 4 of the remaining A5, Modesty.
items worked well in this sample (corrected item–facet rs = .30 We examined the readability of NEO–PI–R and
to .38). Thus, only 2 consistently problematic items (48 and NEO–PI–3 facets using the Flesch–Kincaid reading grade
52) remain in the NEO–PI–3, and Item 52 is acceptable in level. For 17 of the 19 facets with item changes, readability
Form R (corrected item–facet r = .32). increased. Only E4, Activity, showed a small increase in
Table 2 reports data on the original and replacement items. reading difficulty from 3.2 years to 4.0 years. Across all 30
Columns 3 and 4 give Flesch–Kincaid grade level for the facets, the median grade level decreased slightly from 5.2 to
original and replacement item. For 21 of the 25 items with 5.0.
low readability, reading grade level was reduced with the
new item, and overall, the median grade for the 37 items de- Factor Structure
creased from 8.3 to 4.4. Columns 5 and 6 present corrected
item–facet correlations for the NEO–PI–R and NEO–PI–3 Principal component analysis was conducted on Form S and
for Form S; similar data for Form R are given in columns 7 Form R facets for both the NEO–PI–R and the NEO–PI–3. In
and 8. For Form S, correlations increased with 32 of the 37 each case, parallel analysis (Cota, Longman, Stewart,
replacements, and the median corrected item–facet correla- Holden, & Fekken, 1993) clearly indicated the presence of
tion increased substantially from .28 to .37. For Form R, cor- five factors. After varimax rotation, the intended structure
relations increased with 32 of the replacements, and the was found in each analysis, with congruence coefficients
median rose from .30 to .42. ranging from .91 to .98. McCrae, Zonderman, Costa, Bond,
and Paunonen (1996) recommended that factor replication be
Scale Analyses evaluated after rotation toward the normative adult Form S
structure (Costa & McCrae, 1992); results for Form S are re-
Table 3 reports analyses of the scales for both versions; they ported in Table 4. The structures of the original and revision
show that both work well in this adolescent sample and that are very similar, and both closely approximate the adult
the NEO–PI–3 scales appear to offer a slight improvement. structure, with congruence coefficients ranging from .94 to
Columns 2 and 3 of Table 3 report coefficient alphas for the .98. Across the revision, the most noteworthy changes are for
two versions of Form S. At the domain level, the two versions N5, Impulsiveness and E4, Activity, which show increased
have equivalent reliability (median α = .89); at the facet level, loading on their intended factor, and E5, Excitement Seeking
revisions lead to alphas that are the same or higher in 18 of and A6, Tender-Mindedness, which show decreased load-
the 19 cases, with notable improvements for A6, Ten- ings. For the latter two facets, the diminished factor loadings
der-Mindedness and C2, Dutifulness. Only O6, Values, in conjunction with higher internal consistency (see Table 3)
suggest that the scales contain more specific variance and
thus may be improved markers of their intended construct.
3
Items 19, 48, 52, 53, 87, 112, 201, 231, 233, 234, and 239 were Note that despite low internal consistency, E4, Activity and
not replaced. O4, Actions were clear definers of their intended factor.
NEO–PI–3 265
TABLE 2
Item Selection Statistics

Flesch–Kincaid Grade Item–Total, Form S Item–Total, Form R

Facet Item NEO–PI–R NEO–PI–3 NEO–PI–R NEO–PI–3 NEO–PI–R NEO–PI–3

N1 121a,b 10.3 3.6 .22 .46 .24 .55


N4 16a 5.6 4.9 .51 .60 .49 .67
N4 136a 8.3 2.6 .49 .51 .42 .55
N4 226 4.8 5.2 .27 .59 .20 .51
N5 21a,b 9.3 2.1 .33 .26 .34 .27
N5 81 12.0 4.4 .30 .39 .18 .37
E3 42a 3.2 2.6 .19 .33 .33 .40
E4 17 2.3 1.0 .07 .15 .00 .06
E4 47a,b 3.7 12.0 .32 .38 .28 .40
E5 202 5.2 0.7 .15 .26 .15 .42
E6 57a 9.6 7.6 .40 .48 .44 .52
E6 147a 12.0 6.3 .31 .52 .36 .55
O2 8a,b 11.5 5.6 .52 .67 .42 .58
O4 78 3.0 5.2 .17 .21 .15 .30
O4 168 8.7 3.7 .18 .25 .12 .29
O6 118a 9.4 4.4 .32 .31 .24 .23
O6 148a 12.0 7.5 .49 .35 .34 .33
O6 238a,b 11.2 9.6 .24 .31 .15 .20
A1 4a,b 5.4 3.6 .36 .41 .44 .49
A2 219a,b 4.9 3.6 .30 .32 .30 .32
A4 229 9.5 9.9 .31 .37 .39 .43
A5 114a 0.5 0.7 .21 .42 .40 .59
A6 29a 8.4 5.9 .19 .39 .35 .42
A6 59 9.0 5.6 –.03 .43 .07 .52
A6 119a,b 6.4 4.4 .37 .39 .34 .47
A6 149a 12.0 8.0 .24 .32 .31 .30
C1 5a,b 2.2 0.7 .27 .30 .48 .51
C1 35 8.8 9.5 .29 .28 .25 .40
C1 125a 2.3 0.5 .42 .47 .36 .61
C1 185a,b 7.6 0.7 .38 .37 .41 .40
C2 10 7.7 2.3 .13 .52 .22 .63
C2 70a,b 7.3 11.1 .21 .54 .30 .58
C2 160a,b 9.6 2.2 .19 .27 .11 .36
C3 105 5.6 3.7 .12 .29 .25 .34
C3 165a 8.4 10.3 .21 .32 .27 .36
C3 225 3.0 3.3 .31 .36 .47 .44
C4 20a,b 12.0 6.6 .28 .41 .07 .46
Mdn 8.3 4.4 .28 .37 .30 .42

Note. For item–total analyses, N = 500. See Table 4 for facet labels. NEO–PI–R = Revised NEO Personality Inventory; N = Neuroticism; E = Extraversion; O =
Openness; A = Agreeableness; C = Conscientiousness.
aItem left missing by more than 2% of the sample in McCrae et al. (2002). bItem left missing by more than 2% of the present sample.

Results for Form R are very similar, with congruence co- ps < .05). Consistent with earlier studies of adolescents (Mc-
efficients ranging from .92 to .98 after targeted rotation. Sim- Crae et al., 2002; Robins, Fraley, Roberts, & Trzesniewski,
ilar changes in loadings are found for N5, Impulsiveness, E4, 2001), there were small cross-sectional increases in O and C
Activity, E5, Excitement Seeking, and A6, Tender-Minded- (rs = .11 and .12, respectively; ps < .05) but no age associa-
ness in the Form R revision. tions for the other factors. Self-reported grades were related
only to C (r = .33, p < .001), which has been noted as the main
Correlates predictor of grades in other populations (e.g.,
Chamorro-Premuzic & Furnham, 2003; Lievens, Coetsier,
To provide preliminary evidence of validity, we correlated De Fruyt, & De Maeseneer, 2002). All these correlates sug-
NEO–PI–3 scales with SWB, gender, age, and self-reported gest that NEO–PI–3 scales have convergent and discriminant
grades. As hypothesized, SWB was significantly related to validity when used in an adolescent population.
N, E, A, and C (rs = –.53, .30, .17, and .39, respectively; all ps
< .001) but not to O (r = –.07, ns). Consistent with research Analyses in Subsamples
on gender differences (Costa, Terracciano, & McCrae,
2001), women scored higher than men on all five domains, The NEO–PI–R has been used extensively in college-age
although the effects were small in magnitude (rs = .10 to .24, samples, and the NEO–PI–3 is a relatively minor modifica-
266 MCCRAE, COSTA, MARTIN
TABLE 3
Internal Consistency, Cross-Version Correlations, and Cross-Observer Correlations for NEO–PI–R
and NEO–PI–3 Scales

Form S Form R

Coefficient α Coefficient α Cross-Observer r

Scale NEO–PI–R NEO–PI–3 rRa3 NEO–PI–R NEO–PI–3 rR3 NEO–PI–R NEO–PI–3

Domains
N 90 91 99 90 91 99 37*** 38***
E 89 89 99 89 90 99 60*** 60***
O 89 89 99 87 88 99 56*** 55***
A 87 87 98 91 91 99 48*** 47***
C 91 92 98 94 95 99 52*** 53***
Mdn 89 89 99 90 91 99 52 53
Facets
N1 73 77 97 75 79 98 41*** 40***
N2b 71 78 35***
N3b 77 76 36***
N4 69 75 92 68 76 92 35*** 39***
N5 66 66 92 66 68 91 16* 21**
N6b 72 71 30***
E1b 76 77 44***
E2b 75 76 51***
E3 73 75 97 73 74 97 44*** 44***
E4 52 57 93 51 57 93 47*** 51***
E5 62 64 94 68 73 95 50*** 54***
E6 73 77 96 75 78 96 48*** 49***
O1b 77 76 34***
O2 79 81 98 79 81 98 57*** 59***
O3b 76 72 46***
O4 44 48 89 46 56 89 27*** 29***
O5b 79 84 47***
O6 65 63 90 56 57 88 39*** 44***
A1 77 77 97 80 81 98 26*** 26***
A2 66 67 97 74 74 97 23** 22**
A3b 72 80 34***
A4 66 67 97 73 73 97 52*** 51***
A5 73 75 98 79 81 98 42*** 38***
A6 50 66 83 59 71 85 17* 23**
C1 68 70 88 73 78 91 34*** 35***
C2 67 79 92 73 83 93 45*** 46***
C3 62 69 89 76 78 93 39*** 46***
C4 76 78 97 76 81 96 45*** 43***
C5b 77 83 43***
C6b 76 79 52***
Mdn 73 75 97 75 78 98 42 43

Note. N = 500, except for cross-observer correlations (n = 180). See Table 4 for facet labels. Decimal points are omitted. Median for facets is across all 30 scales.
NEO–PI–R = Revised NEO Personality Inventory; N = Neuroticism; E = Extraversion; O = Openness; A = Agreeableness; C = Conscientiousness.
aEquivalence coefficient, NEO–PI–R with NEO–PI–3 scale. bFacet unchanged.

*p < .05. **p < .01. ***p < .001.

tion, so results so far are not surprising. There are, however, ally the same. Allik et al. (2004), in an Estonian sample,
only a handful of studies using the NEO–PI–R in high school found that 14-year-olds had personality structures very simi-
age samples. We therefore repeated the major analyses, re- lar to adults and that the structure was “practically indistin-
stricting the sample to the 204 respondents aged 14 to 17. Re- guishable from adult personality by age 16” (p. 445). This
sults are reported in Table 5; values for NEO–PI–R and appears to be the case for American adolescents as well, from
NEO–PI–3 domains are similar, with a slight overall advan- both self and observer perspectives.
tage for the revised version. Compared to the total group, the Finally, academically successful adolescents were
internal consistency is slightly higher, whereas cross-ob- overrepresented in this sample, although the NEO–PI–3 was
server agreement, factor structure,4 and correlates are virtu- designed chiefly for individuals with limited vocabularies.
To examine the performance of the instrument in a group
4
The factor structure among these younger adolescents was virtu- more closely resembling those on whom it may be used, we
ally identical to the structure of the full adolescent sample, with fac- analyzed NEO–PI–3 responses of 113 individuals who re-
tor congruencies for Forms S and R all .98 or .99. ported getting grades in the B to D range (see Table 1). Inter-
NEO–PI–3 267
TABLE 4
Factor Structure for Form S NEO–PI–R and NEO–PI–3 Facet Scales After Targeted Rotation

Procrustes-Rotated Principal Component

N E O A C VCa

Scale and Facet R 3 R 3 R 3 R 3 R 3 R 3

N1: Anxiety .80 .81 .05 .03 .02 –.01 .09 .11 .07 .05 .97b .97b
N2: Angry Hostilityc .63 .64 –.04 –.02 –.06 –.07 –.51 –.51 –.04 –.03 .99b .99b
N3: Depressionc .80 .79 –.14 –.14 .01 .02 .02 .01 –.16 –.15 .99b .99b
N4: Self-Consciousness .75 .74 –.25 –.31 –.11 –.10 .17 .18 –.07 –.12 .98b .97b
N5: Impulsiveness .44 .52 .32 .33 .14 .10 –.25 –.22 –.37 –.38 .98b .99b
N6: Vulnerabilityc .77 .77 –.04 –.04 –.10 –.10 .05 .07 –.32 –.30 .99b .98b
E1: Warmthc –.11 –.13 .76 .75 .23 .25 .29 .28 .17 .18 .99b .98b
E2: Gregariousnessc –.10 –.12 .80 .79 –.05 –.03 –.07 –.07 –.03 –.04 .96b .97b
E3: Assertiveness –.37 –.38 .47 .46 .17 .19 –.34 –.39 .31 .30 .99b .99b
E4: Activity –.02 .02 .58 .64 .05 .05 –.21 –.23 .38 .37 .98b .98b
E5: Excitement Seeking .01 –.01 .67 .60 .13 .19 –.23 –.25 –.06 –.12 .97b .97b
E6: Positive Emotions –.09 –.10 .62 .65 .37 .30 .21 .24 .06 .10 .95b .96b
O1: Fantasyc .18 .18 .35 .34 .53 .53 .00 .01 –.26 –.25 .95b .95b
O2: Aesthetics .20 .17 .09 .10 .74 .73 .16 .18 .03 .03 .99b .99b
O3: Feelingsc .34 .34 .43 .42 .57 .57 .12 .11 .15 .17 .98b .98b
O4: Actions –.18 –.20 .19 .25 .50 .51 .08 .05 –.15 –.11 .98b .99b
O5: Ideasc –.06 –.07 .03 .01 .79 .80 .02 .01 .16 .16 .98b .99b
O6: Values –.01 –.02 .10 .07 .64 .67 .19 .23 .03 –.00 .84 .83
A1: Trust –.25 –.25 .32 .29 .05 .07 .61 .61 .06 .05 .97b .98b
A2: Straightforwardness –.06 –.03 .06 .04 –.08 –.08 .70 .70 .19 .19 .96b .96b
A3: Altruismc –.10 –.11 .43 .42 .21 .23 .57 .56 .21 .23 .94b .93d
A4: Compliance –.17 –.16 –.08 –.14 –.04 –.08 .79 .75 .02 –.03 .99b .99b
A5: Modesty .25 .24 –.16 –.18 .12 .12 .48 .50 –.02 –.02 .86d .87d
A6: Tender-Mindedness .13 .19 .23 .32 .26 .30 .58 .51 –.06 .00 .97b .93d
C1: Competence –.34 –.35 .18 .16 .13 .10 –.02 .04 .71 .74 .99b .99b
C2: Order .06 .07 –.08 –.08 –.10 –.10 –.04 –.03 .69 .65 .96b .96b
C3: Dutifulness –.12 –.10 .02 .02 .06 .05 .23 .27 .74 .78 .99b .98b
C4: Achievement Striving –.05 –.08 .20 .24 .06 .12 –.07 –.04 .81 .79 .99b .99b
C5: Self-Disciplineb –.28 –.28 .09 .07 –.09 –.08 .06 .06 .77 .79 .99b .99b
C6: Deliberationd –.07 –.09 –.23 –.25 –.12 –.12 .35 .35 .68 .66 .95b .96b
Congruencee .98b .98b .97b .97b .94b .94b .96b .96b .98b .98b .97b .97b

Note. N = 500. Components are rotated toward the adult normative structure (Costa & McCrae, 1992). Loadings greater than .40 in absolute magnitude are
underlined. NEO–PI–R = Revised NEO Personality Inventory; N = Neuroticism; E = Extraversion; O = Openness; A = Agreeableness; C = Conscientiousness. R
= NEO–PI–R; 3 = NEO–PI–3.
aVariable congruence coefficient; total congruence coefficient in the last row. bCongruence higher than that of 99% of rotations from random data. cScale

unchanged in revision. dCongruence higher than that of 95% of rotations from random data. eCongruence with adult normative NEO–PI–R structure.

nal consistencies for the five domains in this subsample were found for N, E, and C (rs = –.42, .22, and .33, respectively; all
only slightly lower than those seen in the full sample, ranging ps < .05). Females scored higher than males on E, O, and A.
from .85 to .89 for Form S and from .84 to .93 for Form R. Al- Neither age nor self-reported grades were related to person-
though factor analyses normally require a minimum of 200 ality; the former because the expected effects are too small to
participants, analyses in this subsample nevertheless closely be seen in this smaller sample and the latter perhaps because
approximated the adult structure, with congruence coeffi- of restriction of grade range from B to D. In general, how-
cients of .88 to .96 for Form S and .87 to .93 for Form R.5 All ever, it appears that the NEO–PI–3 functions well in academ-
of these can be considered replications using Haven and ten ically average adolescents.
Berge’s (1977) rule of thumb. Only 44 of these respondents
had sibling raters, but the cross-observer correlations were DISCUSSION
significant for N (r = .38, p < .05), E (r = .55, p < .001), O (r =
.49, p < .001), and C (r = .46, p < .01) and showed a trend for Schinka and Borum (1994) reported that the NEO–PI–R had
A (r = .29, p < .10). Finally, we correlated domain scores an overall Flesch–Kincaid grade level of 5.7 but noted that
with SWB, gender, age, and grades. Even in this much some items and scales had higher reading levels and con-
smaller subsample, significant correlations with SWB were cluded that “a reasonable degree of confidence that the sub-
ject has the capability of comprehending all inventory items
is probably only warranted if the subject has completed at
5
Congruences with the factor structure in the full adolescent sam- least 10 years of formal education with average grades or
ple ranged from .95 to .99 for Form S and from .94 to .98 for Form R. better” (p. 98). That conclusion might seem to sharply limit
268 MCCRAE, COSTA, MARTIN
TABLE 5
Internal Consistency, Cross-Observer Agreement, Factor Congruences, and Correlations With SWB
for NEO–PI–R and NEO–PI–3 Domain Scales in Respondents Aged 14 to 17

NEO–PI–R/NEO–PI–3 Domain

Variable N E O A C

Coefficient alpha
Form S .90/.91 .89/.90 .89/.89 .88/.88 .90/.92
Form R .91/.92 .89/.90 .86/.87 .91/.90 .93/.94
Cross-observer correlation
Form S/Form R .35***/.38*** .58***/.57*** .62***/.61*** .53***/.52*** .53***/.54***
Factor congruence coefficienta
Form S .97/.98 .96/.96 .95/.96 .96/.95 .97/.97
Form R .97/.97 .96/.97 .93/.90 .94/.94 .95/.95
Form S correlation
SWB –.47***/–.50*** .24***/.26*** –.09/–.08 .29***/.28*** .47***/.52***
Genderb .11/.12 .26***/.25*** .20**/.20** .19**/.20** .03/.05
Grades –.09/–.09 .07/.06 .08/.08 .15*/.13 .31***/.33***

Note. n = 204 for coefficient alpha and congruence coefficients; n = 111 for cross-observer correlations; ns = 201 to 204 for Form S correlations. SWB =
Subjective Well-Being Scale; NEO–PI–R = Revised NEO Personality Inventory; N = Neuroticism; E = Extraversion; O = Openness; A = Agreeableness; C =
Conscientionsness; Form S = self-reports; Form R = observer ratings.
aCongruence coefficients with adult normative NEO–PI–R structure after Procrustes rotation; all coefficients higher than those of 99% of rotations from random

data. bMale = 0, female = 1.


*p < .05. **p < .01. ***p < .001.

the portion of the adolescent population for whom the quent research shows that these results generalize to other age
NEO–PI–R would be appropriate. In fact, however, there is groups, it might be reasonable to consider the NEO–PI–3 as a
no requirement that test respondents understand each and ev- replacement for the NEO–PI–R for respondents of all ages. In
ery item in an inventory. One of the reasons to have the meantime, either the NEO–PI–R or the NEO–PI–3 can ap-
multi-item scales is to allow an occasional missing or misun- propriately be used for adolescents age 14 and up.
derstood item without seriously compromising the validity
of the assessment. That is presumably why the NEO–PI–R
has already proven useful in adolescent populations (Baker ACKNOWLEDGMENTS
& Victor, 2003; De Fruyt et al., 2000; McCrae et al., 2002).
Nevertheless, it is surely the case that one would want to We thank the undergraduates who worked on this project. We
minimize the number of misunderstood items, and the especially thank Erin Weller for helping to develop the mate-
NEO–PI–3 is a step in that direction. It has an overall reading rials, procedures, and files on which data collection relied.
grade level of 5.3 and has eliminated most of the items that We thank Erin Weller, Kathy Wulderk, Amanda Jones,
adolescents aged 14 to 20 find difficult. For those with lim- Rebecca Fritschie, Danielle Quigg, and Rachael Gebely for
ited reading skills, these changes should increase the accu- managing the considerable work of preparing and mailing
racy of assessments as well as making the test-taking questionnaires, communicating with research assistants,
experience less frustrating. Psychometrically, the NEO–PI–3 scanning data, and resolving problems as they arose. We
shows modest improvements over the generally good perfor- greatly appreciated their warmth, straightforwardness, com-
mance of the NEO–PI–R. petence, and orderliness. Thanks, too, to each of the 53 un-
An examination of the items in the Appendix shows that dergraduates who recruited participants and reminded them
the new items are shorter and more up to date but also more of our interest in receiving their packets.
colloquial. Translators of the NEO–PI–R have frequently re-
marked on the informal and idiomatic language of the instru-
ment and have been encouraged to maintain the sense rather REFERENCES
than the literal wording of the items. The NEO–PI–R has
generally worked well in translation (e.g., Costa, McCrae, & Allik, J., Laidra, K., Realo, A., & Pullmann, H. (2004). Personality develop-
Jónsson, 2002), and the NEO–PI–3 presumably would, too. ment from 12 to 18 years of age: Changes in mean levels and structures of
Although a more readable NEO–PI–R was initially con- traits. European Journal of Personality, 18, 445–462.
ceived as an adaptation for adolescent respondents, the Andrews, F. M., & Withey, S. B. (1976). Social indicators of well-being:
Americans’ perceptions of life quality. New York: Plenum.
NEO–PI–3 may have wider applicability. It would be reason- Baker, S. R., & Victor, J. B. (2003, August). Adolescent self-reports of per-
able to examine its use in younger respondents (cf. Markey, sonality and temperament: NEO–PI–R and TPQ. Paper presented at the
Markey, Tinsley, & Ericksen, 2002) and in adults. If subse- 11th European Conference on Developmental Psychology, Milan, Italy.
NEO–PI–3 269
Bradburn, N. M. (1969). The structure of psychological well-being. Chi- Procrustes rotation. Journal of Personality and Social Psychology, 70,
cago: Aldine. 552–566.
Carter, J. A., Herbst, J. H., Stoller, K. B., King, V. L., Kidorf, M. S., Costa, P. Robins, R. W., Fraley, R. C., Roberts, B. W., & Trzesniewski, K. H. (2001).
T., Jr., et al. (2001). Short-term stability of NEO–PI–R personality trait A longitudinal study of personality change in young adulthood. Journal of
scores in opioid-dependent outpatients. Psychology of Addictive Behav- Personality, 69, 617–640.
iors, 15, 255–260. Schinka, J. A., & Borum, R. (1994). Readability of normal personality in-
Chamorro-Premuzic, T., & Furnham, A. (2003). Personality predicts aca- ventories. Journal of Personality Assessment, 62, 95–101.
demic performance: Evidence from two longitudinal university samples.
Journal of Research in Personality, 37, 319–338.
APPENDIX
Costa, P. T., Jr., & McCrae, R. R. (1984). Personality as a lifelong determi-
nant of well-being. In C. Malatesta & C. Izard (Eds.), Affective processes Deleted Items and Their Replacements
in adult development and aging (pp. 141–157). Beverly Hills, CA: Sage. (in Boldface) for the NEO–PI–3
Costa, P. T., Jr., & McCrae, R. R. (1985). The NEO Personality Inventory
manual. Odessa, FL: Psychological Assessment Resources. N1:
Costa, P. T., Jr., & McCrae, R. R. (1989). The NEO–PI/NEO–FFI manual
supplement. Odessa, FL: Psychological Assessment Resources. 121. I’m seldom apprehensive about the future.
Costa, P. T., Jr., & McCrae, R. R. (1992). Revised NEO Personality Inventory 121. I seldom feel nervous.
(NEO–PI–R) and NEO Five-Factor Inventory (NEO–FFI) professional N4:
manual. Odessa, FL: Psychological Assessment Resources. 16. In dealing with other people, I always dread making a social blunder.
Costa, P. T., Jr., & McCrae, R. R. (2002). Looking backward: Changes in the 16. When I’m around people, I worry that I’ll make a fool of my-
mean levels of personality traits from 80 to 12. In D. Cervone & W. self.
Mischel (Eds.), Advances in personality science (pp. 219–237). New 136. I often feel inferior to others.
York: Guilford. 136. I often feel that I am not as good as others.
Costa, P. T., Jr., McCrae, R. R., & Jónsson, F. H. (2002). Validity and utility 226. When people I know do foolish things, I get embarrassed for them.
of the Revised NEO Personality Inventory: Examples from Europe. In B. 226. I feel awkward around people.
De Raad & M. Perugini (Eds.), Big Five assessment (pp. 61–77).
Göttingen, Germany: Hogrefe & Huber. N5:
Costa, P. T., Jr., Terracciano, A., & McCrae, R. R. (2001). Gender differ- 21. I rarely overindulge in anything.
ences in personality traits across cultures: Robust and surprising findings. 21. It doesn’t bother me too much if I can’t get what I want.
Journal of Personality and Social Psychology, 81, 322–331. 81. I have little difficulty resisting temptation.
Cota, A. A., Longman, R., Stewart, R., Holden, R. R., & Fekken, G. C. 81. I’m always in control of myself.
(1993). Comparing different methods for implementing parallel analysis:
E3:
A practical index of accuracy. Educational and Psychological Measure-
42. I sometimes fail to assert myself as much as I should.
ment, 53, 865–876.
42. Sometimes I don’t stand up for my rights like I should.
De Fruyt, F., Mervielde, I., Hoekstra, H. A., & Rolland, J.-P. (2000). As-
sessing adolescents’ personality with the NEO–PI–R. Assessment, 7, E4:
329–345. 17. I have a leisurely style in work and play.
Diener, E., Emmons, R. A., Larsen, R. J., & Griffin, S. (1985). The Satisfac- 17. I have a laid-back style in work and play.
tion With Life Scale. Journal of Personality Assessment, 49, 71–75. 47. When I do things, I do them vigorously.
Haven, S., & ten Berge, J. M. F. (1977). Tucker’s coefficient of congruence 47. I act forcefully and energetically.
as a measure of factorial invariance: An empirical study (Heymans Bulle-
tin 290 EX). University of Groningen, Groningen, The Netherlands. E5:
Lievens, F., Coetsier, P., De Fruyt, F., & De Maeseneer, J. (2002). Medical 202. I’m attracted to bright colors and flashy styles.
students’ personality characteristics and academic performance: A 202. I like loud music.
Five-Factor Model perspective. Medical Education, 36, 1050–1056. E6:
Markey, P. M., Markey, C. N., Tinsley, B. J., & Ericksen, A. J. (2002). A pre- 57. I have sometimes experienced intense joy or ecstasy.
liminary validation of preadolescents’ self-reports using the Five-Factor 57. I have felt overpowering joy.
Model of personality. Journal of Research in Personality, 36, 173–181. 147. I don’t consider myself especially “light-hearted.”
McCrae, R. R., & Costa, P. T., Jr. (1991). Adding Liebe und Arbeit: The full 147. I’m not happy-go-lucky.
Five-Factor Model and well-being. Personality and Social Psychology
Bulletin, 17, 227–232. O2:
McCrae, R. R., & Costa, P. T., Jr. (2004). A contemplated revision of the 8. Aesthetic and artistic concerns aren’t very important to me.
NEO Five-Factor Inventory. Personality and Individual Differences, 36, 8. I’m not really interested in the arts.
587–596.
O4:
McCrae, R. R., Costa, P. T., Jr., Martin, T. A., Oryol, V. E.,
78. Once I find the right way to do something, I stick to it.
Rukavishnikov, A. A., Senin, I. G., et al. (2004). Consensual validation
78. I like the old-fashioned methods I’m used to.
of personality traits across cultures. Journal of Research in Personal-
168. Sometimes I make changes around the house just to try something
ity, 38, 179–201.
different.
McCrae, R. R., Costa, P. T., Jr., Terracciano, A., Parker, W. D., Mills, C. J.,
168. I believe variety is the spice of life.
De Fruyt, F., et al. (2002). Personality trait development from 12 to 18:
Longitudinal, cross-sectional, and cross-cultural analyses. Journal of Per- O6:
sonality and Social Psychology, 83, 1456–1468. 118. I believe that the different ideas of right and wrong that people in
McCrae, R. R., Zonderman, A. B., Costa, P. T., Jr., Bond, M. H., & other societies have may be valid for them.
Paunonen, S. V. (1996). Evaluating replicability of factors in the Re- 118. Our ideas of right and wrong may not be right for everyone in
vised NEO Personality Inventory: Confirmatory factor analysis versus the world.
270 MCCRAE, COSTA, MARTIN
148. I believe that loyalty to one’s ideals and principles is more impor- 35. I don’t take civic duties like voting very seriously.
tant than “open-mindedness.” 35. I sometimes act thoughtlessly.
148. I believe that it’s better to stick to your own principles than to 125. I pride myself on my sound judgment.
be open-minded. 125. I have good judgment.
238. I believe that the “new morality” of permissiveness is no morality 185. I’m a very competent person.
at all. 185. I have many skills.
238. People should honor traditional values, not question them.
C2:
A1: 10. I would rather keep my options open than plan everything in advance.
4. I tend to be cynical and skeptical of others’ intentions. 10. I don’t mind a little clutter in my room.
4. Often, people aren’t as nice as they seem to be. 70. I am not a very methodical person.
70. I’m not a very orderly or methodical person.
A2: 160. I tend to be somewhat fastidious or exacting.
219. I pride myself on my shrewdness in handling people. 160. I’m picky about how jobs should be done.
219. I’m pretty slick when it comes to dealing with people.
C3:
A4: 105. Sometimes I cheat when I play solitaire.
229. I often get into arguments with my family and co-workers. 105. I ignore a lot of silly little rules.
229. I sometimes get into arguments. 165. I adhere strictly to my ethical principles.
A5: 165. I follow my ethical principles strictly.
114. I try to be humble. 225. I’d really have to be sick before I’d miss a day of work.
114. I’m not a show-off. 225. I try to go to work or school even when I’m not feeling well.

A6: C4:
29. Political leaders need to be more aware of the human side of their 20. I am easy-going and lackadaisical.
policies. 20. I’m not very ambitious.
29. When making laws and social policies, we need to think about
who might be hurt.
59. I’m hard-headed and tough-minded in my attitudes. Robert R. McCrae
59. I don’t worry much about the homeless. Box 03
119. I have no sympathy for panhandlers. Gerontology Research Center
119. I have no sympathy for beggars. 5600 Nathan Shock Drive
149. Human need should always take priority over economic consid-
Baltimore, MD 21224–6825
erations.
149. Human need is more important than economics. Email: mccraej@grc.nia.nih.gov
C1:
5. I’m known for my prudence and common sense. Received June 18, 2004
5. I’m known for my common sense. Revised August 25, 2004

View publication stats

You might also like