Furr-2008-Journal of Personality

A Framework for Profile Similarity: Integrating
Similarity, Normativeness, and Distinctiveness
R. Michael Furr
Wake Forest University
ABSTRACT Many questions in personality psychology lend them-

selves to the analysis of profile similarity. A profile approach to issues
such as personality judgment, personality similarity, behavioral consis-
tency, developmental stability, and person-environment fit is intuitively
appealing. However, it entails conceptual and statistical challenges arising
from the overlap among profile similarity and normativeness, which
presents potential confounds and potential opportunities. This article de-
scribes the normativeness problem, articulating the need to evaluate pro-
file similarity alongside normativeness and distinctiveness. It presents
conceptual and psychometric foundations of a framework differentiating
these elements for pairs of profiles. It derives two models from this frame-
work, and it discusses the application of their components to a variety
of research domains. Finally, it presents recommendations and implica-
tions regarding the use of these components and profile similarity
more generally. This approach can reveal and manage potential
confounds, and it can provide theoretical insights that might otherwise
be overlooked.
The analysis of profiles of psychological variables has an intuitive

appeal. From a profile-based approach, an individual is scored on a
set of variables obtained at one time, from one person, in one sit-
uation, or in response to one set of instructions. The individual’s
profile is then matched with a second profile of scores on the same set
of variables, obtained at a different time, from a different person, in
a different situation, or in response to a different set of instructions.
The similarity between two profiles can represent many psycholog-
ical phenomena such as personality similarity, personality judgment,
behavioral consistency, developmental stability, clinical diagnosis, or
person-environment fit.
Please address correspondence to Mike Furr, Department of Psychology, Wake
Forest University, Winston Salem, NC 27106; E-mail: [email protected].
Journal of Personality 76:5, October 2008

r 2008, Copyright the Authors
Journal compilation r 2008, Wiley Periodicals, Inc.
DOI: 10.1111/j.1467-6494.2008.00521.x
1268 Furr
Self and Informant Profiles

8
7 Self
Informant
6
5
Score
4
3
2
1
0
Neur. Ext Open. Agree. Consc.
Trait
Figure 1
Profiles of self trait ratings and Informant trait ratings.
Figure 1 illustrates two psychological profiles in the context of

self-other agreement of personality judgments. One profile repre-
sents a target person’s self-rated profile of personality trait scores,
and the other represents an acquaintance’s informant-rated profile of
the target’s personality. The similarity in the shape of the two profiles
reflects the degree to which the target and the informant agree upon
the traits that are more and less characteristic of the target’s per-
sonality. Profile similarity is usually defined in terms of the similarity
in the pattern of peaks and valleys common to the two profiles. For
example, both profiles in Figure 1 indicate that the target’s Extr-
aversion is greater than his or her Openness to Experience, and they
both indicate that the target’s Conscientiousness is greater than his
or her Agreeableness. Although the two profiles do not have iden-
tical shapes, the degree of similarity indicates that this target and
informant do have some agreement on the target’s personality
profile.
Despite its intuitive appeal, a profile-similarity approach raises
conceptual and statistical problems. These problems largely concern
the role of normativeness in creating apparently high levels of sim-
ilarity between profiles or in obscuring meaningful effects. Although
the normativeness problem in profile similarity has been recognized
in some domains of research (e.g., Cronbach, 1955), there have been
Profile Similarity 1269
few, if any, helpfully coherent and flexible frameworks for handling

the problem. Furthermore, there have been few, if any, frameworks
that are articulated for different domains of social and personality
psychology.
The current article presents a solution to the normativeness prob-
lem. It presents a conceptual and psychometric framework differen-
tiating normativeness and distinctiveness in the analysis of
profile similarity, illustrates the utility of the framework, places the
framework within the broader context of profile analysis, and
presents recommendations for the use of the framework across
many domains of social and personality psychology. The frame-
work affords researchers an opportunity to gauge and potentially
account for confounds in the analysis of profile similarity, and it
allows researchers to disentangle meaningful effects that might be
obscured.
This issue is important because profile similarity is used to exam-
ine many fundamental psychological phenomena. A profile ap-
proach has a long history in the study of personality judgment, in
which interjudge agreement is defined as the degree to which a
profile of trait scores as rated by one judge (e.g., the self) matches the
profile of trait scores as rated by another judge (e.g., an informant;
Baker & Block, 1957; Bernieri, Zuckerman, Koestner, & Rosenthal,
1994; Biesanz & West, 2000; Biesanz, West, & Millevoi, 2007;
Blackman & Funder, 1998; Chaplin & Panter, 1993; Colvin,
1993a, 1993b; Cronbach, 1955; Furr, Dougherty, Marsh, & Math-
ias, 2007; Letzring, Wells, & Funder, 2006; Starzyk, Holden,
Fabrigar, & MacDonald, 2006). A profile approach has also been
used extensively for phenomena such as husband-wife personality
similarity (Gaunt, 2006; Gonzaga, Campos, & Bradbury, 2007;
Luo & Klohnen, 2005), cross-situational behavioral consistency
(Champagne & Pervin, 1987; Furr & Funder, 2004; Krahé,
1990; Magnusson & Ekehammar, 1978; Ozer, 1986), personality
pathology (Carlson & Furr, 2007; Lynam & Widiger, 2001; McCrae
et al., 2001; Miller, Lynam, Widiger, & Leukefeld, 2001; Miller,
Pilkonis, & Morse, 2004; Miller, Reynolds, & Pilkonis, 2004;
Westen & Bradley, 2005; Westen, Shedler, & Bradley, 2006;
Widiger, Trull, Clarkin, Sanderson, & Costa, 2002), and the tempo-
ral stability of personality (Asendorpf & Van Aken, 1991; Biesanz,
West, & Kwok, 2003; Block, 1971; Caspi & Herbener, 1990; Ozer &
Gjerde, 1989).
1270 Furr
The Normativeness Problem
Profile normativeness creates conceptual and statistical challenges in

interpreting profile similarity. The ability to gauge profile norma-
tiveness alongside profile similarity provides a way of coping with
these challenges, and it represents an opportunity to explore new
ideas in domains such as self-other agreement, marital similarity,
behavioral consistency, and personality stability.
Profile normativeness is the degree to which a profile reflects an
average profile—the similarity between an individual’s profile of
scores and a group’s normative profile of scores. Figure 2 illustrates
this, with Figure 2a presenting three targets’ self-rated personality
profiles. Note the commonality across the three individual profiles;
two of the three profiles reflect a higher level of Extraversion than
Neuroticism, and all three profiles indicate that the targets see them-
selves as less Open than Extraverted and Agreeable. Figure 2b
presents the three targets’ self-rated profiles along with a profile re-
flecting these normative trends. This Normative self-rated profile
represents scores for each trait, as averaged across all targets’ self-
ratings. As Figure 2b illustrates, each target’s self-rated profile is, at
least, somewhat similar to the Normative self-rated profile, indicat-
ing that each target’s profile reflects some normativeness.
There are at least three likely facts about profile normativeness.
First, in analyses of trait profiles or behavioral profiles, the norma-
tive profile is likely to be ‘‘powerful’’ in that many individual profiles
will be somewhat, if not very, normative. That is, normativeness is
likely to characterize each individual’s profile to some degree, as
shown in Figure 2b. Second, two normative profiles are likely to be
highly similar to each other. For example, the normative self-rated
profile in Figure 2b is likely to be similar to a normative informant-
rated profile arising from the informants who rate the targets. For
example, Furr et al. (2007, Table 1) report normative trait profiles
from adolescents’ self-ratings and their mothers’ informant ratings;
the correlations between these profiles of facet scores were strong:
r 5 .83 and r 5 .94 for adolescents with and without Conduct Dis-
order, respectively. Third, a normative profile likely has psycholog-
ical meaning, particularly in terms of social desirability,
psychological adjustment, or adaptation to one’s environment. For
example, among the adolescents without Conduct Disorder exam-
ined by Furr et al., the normative profile of adolescent self-ratings
(a) Targets' Original Self–Rated Profiles

7
6
5
4
Score
3
2 ERIC
STAN
1 KYLE
0
Trait
(b) Targets' Original Self–Rated Profiles

7
6
5
4
Score
3
ERIC
2 STAN
KYLE
1 Normative
0
Trait
(c) Targets' Self–Rated Distinctive Profiles

2.5
2
1.5
1
0.5
Score
0
–0.5
–1
ERIC
–1.5 STAN
–2 KYLE
–2.5
Trait
Figure 2
Raw profiles, Normative profile, and Distinctive profiles.
1272 Furr
was highly correlated with independent ratings of trait social desir-

ability, r 5 .83 and r 5 .88 (note that these analyses were not re-
ported by Furr et al.). Together, these three likely facts imply that
normativeness is a powerful and meaningful component that can
affect profile similarity.
These likely facts have implications for the similarity between
profiles and for the psychological meaning of similarity. The first two
facts imply that any two profiles are likely to be similar, even without
an intrinsic connection between the two profiles. For example, one
target’s self-rated personality profile is likely to be somewhat, if not
very, similar to an informant-rated profile of another target. This is
likely because (a) each target’s self-rated profile is likely to be similar
to the normative self-rated profile, (b) each informant-rated profile is
likely to be similar to the normative informant-rated profile, and (c)
the normative self-rated profile is likely to be similar to the norma-
tive informant-rated profile. This implication, along with the likely
fact that normativeness has psychological meaning, implies that
profile similarity can represent social desirability, adjustment, or
adaptiveness. For example, high similarity between a target’s self-
rated profile and an informant-rated profile indicates high self-other
agreement, perhaps suggesting that the informant knows the target
well. But it may also, or alternatively, indicate that the target is well
adjusted because normativeness is likely to be associated with sim-
ilarity (as just described) and normativeness may reflect social
desirability or psychological well-being.
These implications create at least two forms of the normativeness
problem. First, they complicate the interpretation of average levels
of profile similarity. We might wish, for example, to interpret a
sample’s average level self-other agreement (i.e., averaged across all
target-informant pairs) as indicating the general depth of under-
standing between acquaintances. Unfortunately, this interpretation
is clouded by the fact that any given target’s self-rated profile of trait
scores will likely be similar to any given informant’s profile of trait
scores. Thus, we would likely obtain a positive degree of self-other
agreement even if acquaintances understood each other no better
than did strangers. Cronbach (1955) and others have recognized this
type of complication in the context of personality judgments, but it
also has important implications for other applications of profile
similarity. For example, any given wife’s profile will likely be similar
to any given husband’s profile, which clouds our ability to evaluate
the general level of personality similarity within married couples.

The second problem introduced by the normativeness implications
concerns the interpretation of any correlates of profile similarity. For
example, we might examine self-other agreement and private self-
consciousness, finding a positive correlation between the two. We
might interpret this as indicating that self-other agreement arises
when targets are attentive to their inner experiences and have good
self-insight. But self-other agreement, as a form of profile similarity,
may partially reflect psychological adjustment. Therefore, the pos-
itive correlation between self-other agreement and private self-con-
sciousness could be interpreted as indicating simply that people who
have high private self-consciousness are relatively well-adjusted, with
no clear implication for the nature of self-other agreement. This
creates ambiguity in interpreting ‘‘depth of mutual understanding of
a target’s personality’’ versus psychological adjustment as a correlate
of private self-consciousness.
Two Possible Analytic Strategies

There are at least two broad strategies for coping with the norma-
tiveness problem in profile-based analyses—a sample-level strategy
and a pair-level strategy. Although both strategies have been used to
account for normativeness, a pair-level strategy offers the potential
of greater flexibility and information.
Sample-Level Strategy (Random Dyads)
In a sample-level approach to normativeness, the average similarity

between pairs of profiles is compared to a normative level of sim-
ilarity across the entire sample. That is, similarity is computed be-
tween each profile and its true complementary profile (e.g., between
each target and his or her informant, or between each wife and her
husband) and similarity is computed between random pairs of pro-
files (e.g., between a given target and another target’s informant, or
between a wife from one couple and a husband from another cou-
ple). The average similarity between ‘‘true’’ pairs is then contrasted
with the average similarity between random pairs. In the context of
couples, Kenny et al. (2006, pp. 335–337) refer to this approach as
‘‘pseudo-couple analysis.’’
Letzring et al. (2006) present an excellent application of this
approach. In this profile-based examination of self-other agreement,
1274 Furr
participants interacted with each other in small groups, provided

self-reports of their personality, and then provided informant reports
of the other group members’ personalities. To examine the general
level of self-other agreement between two individuals who have in-
teracted with each other, Letzring and her colleagues computed a
similarity correlation between each target participant’s self-rated
profile of trait scores and an informant’s profile of trait ratings. A
strong positive correlation indicated agreement between the self’s
profile and the informant’s profile. Letzring et al. recognized the
complexity that normativeness introduces into this issue, stating that
‘‘the baseline correlation between two sets of ratings was not ex-
pected to be zero, as some items of the [trait questionnaire] are gen-
erally rated higher than others, regardless of the target’’ (p. 118).
Thus, a self-other agreement correlation might be positive, not be-
cause the informant knows anything distinctive about the target but
because the informant realizes that the typical person has higher
levels of some specific traits than of others. Addressing this problem,
Letzring et al. created ‘‘random pairs’’ by pairing a profile from one
target with a profile from an individual with whom the target had
not interacted, and they computed a profile-similarity correlation for
each random pair. After repeated pairings, the mean of the random
pair correlations was used as a baseline profile correlation reflecting
agreement between two unacquainted people. The average similarity
between the ‘‘real’’ acquainted self-other pairs was then contrasted
with the baseline profile correlation, and the difference was inter-
preted as the degree to which acquaintance affects self-other agree-
ment. In their experimental examination, Letzring et al. found that
‘‘real’’ self-other agreement was higher than baseline agreement
when more than minimal acquaintanceship was attained.
Similarly, Luo and Klohnen (2005) used a sample-level approach
in their examination of assortative mating (i.e., the hypothesis that
people select mates who are relatively similar to themselves). To ex-
plore the similarity between husbands and wives, Luo and Klohnen
computed a similarity correlation between each husband’s profile of
self-reported personality trait scores and his wife’s profile of self-re-
ported trait scores. Acknowledging the normativeness problem, Luo
and Klohnen stated that ‘‘individuals, on average, tend to be more
similar than dissimilar’’ (p. 311). That is, husbands and wives might
have similar profiles, but any two adults might have relatively similar
profiles. Addressing this facet of the normativeness problem, they
created ‘‘random couples’’ by pairing a profile from a husband in

one couple with a profile from a wife in a different couple, and they
computed a similarity correlation for each random couple. After
many random pairings, the mean correlation was interpreted as ‘‘the
average similarity between men and women’’ (p. 311). The average
similarity between the ‘‘real’’ couples was then contrasted with the
average similarity between randomly paired couples, and the differ-
ence was interpreted as the degree to which real couples are more
similar than are random pairs of men and women. In their examin-
ation of more than 20 domains of personality and attitudes, Luo
and Klohnen found that, on average, husbands and wives were in-
deed similar to each other, but they found that that the average real
couple showed significantly higher similarity than would be expected
based on the sample normativeness for only 5 domains (particularly
in attitude domains).
The sample-level strategy is an effective method of addressing one
form of the normativeness problem, allowing researchers to evaluate
the average level of similarity in a way that accounts for norma-
tiveness. Indeed, given the nature of the questions addressed by
Letzring et al. (2006) and given their experimental design, the
sample-level approach was reasonable and useful. Similarly, the
sample-level approach used by Luo and Klohnen (2005) was
sufficient to examine ‘‘general level of marital similarity.’’ Although
the sample-level strategy can be used to address one form of the
normativeness problem, a pair-level strategy can be used to address
both forms—the average level of similarity and the correlates of
similarity.
Pair-Level Strategy
In a pair-level strategy, the similarity between each pair of profiles is

decomposed into elements of normativeness or distinctiveness. For
example, in the study of self-other agreement, researchers might wish
to evaluate the normativeness of each self-rated profile or to index
the degree to which a target-informant pair agrees about the dis-
tinctive (i.e., nonnormative) qualities of the target’s personality.
Similarly, in the study of marital similarity, researchers might wish
to evaluate the normativeness of each spouse’s trait profile or to in-
dex the degree to which each couple shares distinctive personality
qualities.
1276 Furr
The framework presented in the next section is a pair-level strat-

egy decomposing overall profile similarity into components of dis-
tinctiveness and normativeness, and it is intended to accomplish four
goals. First, it provides a framework that is internally coherent, in
that its components are meaningful and their connections are psy-
chometrically well-defined. Second, it is applicable to a common
data structure in profile similarity—data in which each profile is
linked to only one other profile (e.g., each self-rating is linked to one
informant rating, each wife is linked to one husband). Third, the
framework is intended to be meaningfully applicable to any form of
profile similarity—self-other agreement, marital similarity, behav-
ioral consistency, personality stability, person-environment fit, and
so on. Fourth, it provides flexibility in the components that re-
searchers might examine, thus providing flexibility in the psycholog-
ical questions that can be addressed. Specifically, the framework to
be described allows researchers to examine several ways of concep-
tualizing and indexing profile normativeness, distinctiveness, and
similarity.
Conceptual Basis of the Framework

The current pair-level framework decomposes profile similarity into
components of normativeness and distinctiveness. This decomposi-
tion of profile similarity begins by decomposing each individual’s
profile into two specialized profiles.
Decomposition of Individual Profiles
An individual’s profile can be seen as arising from two profiles—a

normative profile representing a group’s average profile and a distinc-
tive profile representing the ways in which the individual is above or
below average on each trait. Figure 2c presents three distinctive pro-
files related to the profiles in Figure 2b. Eric’s distinctive self-reported
Neuroticism score is 2, reflecting the difference between Eric’s original
self-reported Neuroticism score (i.e., 5) and the normative self-
reported Neuroticism score (i.e., 3). Similarly, his distinctive Agree-
ableness score is 0, reflecting the fact that his self-reported Agreeable-
ness score (i.e., 5) matches the normative self-reported Agreeableness
score (i.e., 5). In sum, an individual’s distinctive profile includes
deviation scores reflecting the ways in which the individual deviates
from the normative profile. Consequently, an individual’s raw profile
Figure 3
Model 1 applied to self-other agreement in personality judgments.
of scores (e.g., Figure 2a) arises from a normative profile and from
his or her distinctive profile. Figures 3 and 4 present the three
basic profiles—profiles of raw/unadjusted traits scores, of normative
scores, and of distinctive scores.
Overall Similarity
In the current framework, Overall Similarity is the correlation be-

tween two raw, unadjusted profiles. This is illustrated in Figure 3, in
which the boxed terms represent various personality profiles (e.g., a
target’s self-reported personality profile, a targets’ distinctive pro-
file), and the solid lines and bold text represent the components of
self-other agreement, in terms of similarity, normativeness, and
distinctiveness. Overall similarity is the typical index of profile
similarity, often interpreted as agreement, consistency, stability, fit,
correspondence, depending on the relevant research domain. As
described above, Overall Similarity is potentially confounded with
normativeness, and the current framework decomposes Overall
Similarity into three types of components.
Distinctive Similarity
Distinctive Similarity reflects the similarity between the unique

aspects of the two profiles within a pair—the degree to which one
1278 Furr
Overall
Husband’s Similarity Wife’s
trait profile trait profile
Distinctive
Husband’s Wife’s
Similarity
Distinctive Distinctive
Husband-Wife trait profile trait profile Wife-Husband
Normativeness Normativeness
Generalized
Normative
Profile of the Similarity Profile of the
Normative Normative
Husband Wife
Figure 4
Elements of Model 2 applied to husband-wife similarity.
distinctive profile matches another distinctive profile. In self-other

agreement (see Figure 3), Distinctive Agreement is the similarity be-
tween the distinctive aspects of a target person’s self-perception and
the distinctive aspects of an informant’s perception. For example, a
target might describe himself in a way that reflects above-average
Extraversion and below-average Openness as compared to other
targets. If his informant also describes him in a way that reflects
above-average Extraversion and below-average Openness, then the
two profiles have high Distinctive Similarity. In marital similarity
(see Figure 4), Distinctive Similarity is the similarity between a wife
and husband’s distinctive profile—the degree to which the couple
shares unusually high and low trait levels.
Generalized Normative Similarity
Generalized Normative Similarity reflects the similarity between two

normative profiles. In self-other agreement, for example, this is the
similarity between the average self-rating and the average rating
provided by informants (i.e., the degree to which the average target’s
self-perception matches the average informant’s perception). As il-
lustrated in Figure 3, this component reflects Generalized Normative
Agreement. In marital similarity, this component reflects similarity
between the average husband and the average wife (see Figure 4).
Generalized Normative Similarity is constant across all pairs of pro-
files—that is, the similarity within each husband-wife pair (i.e., each
pair’s Overall Similarity) is partially explained by the similarity be-

tween the average husband and the average wife (i.e., the General-
ized Normative Similarity within the sample). As mentioned earlier,
two normative profiles are likely to be similar to each other (e.g., the
average wife will likely be similar to the average husband). There-
fore, Generalized Normative Similarity is likely to be large in many
applications of profile similarity.
Profile-Level Normativeness
The third type of component reflects the normativeness of each pro-

file or of each pair of profiles. These components can be conceptu-
alized in at least two ways, producing two meaningful models of
profile similarity.
In Model 1 (see Figure 3), there are three remaining components.
Two of these are Within-Profile Normativeness components reflect-
ing the normativeness of each profile. That is, they reflect the sim-
ilarity between each profile and the normative profile of its type. In
self-other agreement, one Within-Profile Normativeness component
reflects the degree to which the target’s self-perception is normative
—the similarity between a target’s self-rated trait profile and the av-
erage target’s self-rated trait profile. The other Within-Profile Nor-
mativeness component reflects the degree to which the informant’s
perception is normative—the similarity between an informant’s profile
of trait ratings and the average informant’s profile. Complementing
the two Within-Profile Normativeness components, Model 1 in-
cludes a Global Normativeness component. This component reflects
the normativeness of each pair of profiles—the degree to which the
average of a pair of profiles matches the average of all profiles. It is
the similarity between two new averaged profiles (see Figure 3). The
first new profile is a Pair Mean profile that includes scores averaged
across both profiles in a given pair. In self-other agreement, the Pair
Mean profile is the target’s personality profile averaged across self-
ratings and informant ratings. The second new average profile is a
Global Mean profile that includes scores averaged across all individ-
uals and all profiles. In self-other agreement, it is the profile of the
average person, averaged across all self-ratings and all informant
ratings. Based on the two averaged profiles, Global Normativeness
represents the similarity between a Pair Mean profile and the Global
Mean profile. In self-other agreement (see Figure 3), it reflects the
1280 Furr
degree to which a given target (as seen by self and informant)

matches the average target (as seen by all selves and all informants).
As presented in Figure 4, Model 2 provides a different conceptu-
alization of the pair-level normativeness components. In Model 2,
this includes two Cross-Profile Normativeness components reflecting
similarity between a profile and the average of the other type of
profile. Figure 4 illustrates this in the context of marital similarity.
Husband-wife normativeness represents the similarity between a
particular husband and the average wife, and wife-husband norma-
tiveness represents similarity between a particular wife and the av-
erage husband.
Table 1 summarizes the general conceptual framework as reflected
in both models. The models provide alternatives in the examination
of profile similarity from many domains of research. As detailed la-
ter, these alternatives allow researchers to use components most
closely fitting their theoretical interests, with the knowledge that they
are valid and coherent components of profile similarity.
Table 1 also summarizes the psychometric basis of these models.
The next section develops these psychometric foundations, demon-
strating that these models are internally coherent. That is, Overall
Similarity can be fully decomposed into the components presented in
Table 1.
Psychometric Foundations
In a profile approach to phenomena such as self-other agreement
and personality similarity, a person’s scores in one profile are
matched with scores in a second profile. For example, self-other
agreement requires an individual’s personality to be rated by two
sources—the target person’s self-ratings of personality comprise one
profile and the informant’s ratings comprise the second profile. The
following notation reflects the general case:
i 5 any one individual (i 5 1 to ni)
v 5 any one variable in a profile (v 5 1 to nv)
p 5 either of two profiles (p 5 1 to 2)
xivp 5 the score of individual i on variable v in profile p
For example, in the analysis of self-other agreement of personality

trait judgments, x142 represents individual 1’s score on trait 4, as
rated by judge 2 (i.e., the informant).
Table 1
Profile Similarity Components Applied to Four Domains of Research
Models and Profile Similarity Components
Model 1
General conceptual framework
Overall Distinctive Generalized Within-Profile Within-Profile Global
Normative
Similarity 5 Similarity Similarity Normativeness 1 Normativeness 2 1Normativeness
Covariance terms
sxiv1 xiv2 5 sx0iv1 x0iv2 sxv1 xv2 sxv1 xiv1 sxv2 xiv2 14sxiv xv
Correlational terms s 0 s 0 r 0 0 s s r
x x x x xv1 xv2 xv1 xv2 sxv1 sxiv1 rxv1 xiv1 sxv2 sxiv2 rxv2 xiv2 þ4sxiv sxv rxiv xv
rxiv1 xiv2 ¼ iv1 iv2 iv1 iv2 sxiv1 sxiv2
Model 2
General conceptual framework
Overall Distinctive Generalized Cross-Profile Cross-profile
Normative
Similarity 5 Similarity Similarity 1 Normativeness 1 1 Normativeness 2
Covariance terms
sxiv1 xiv2 5 sx0iv1 x0iv2 sxv1 xv2 1 sxv1 xiv2 1 sxv2 xiv1
Correlational terms
sx0 sx0 rx0 x0 sxv1 sxv2 rxv1 xv2 þsxv1 sxiv2 rxv1 xiv2 þsxv2 sxiv1 rxv2 xiv1
iv1 iv2 iv1 iv2
rxiv1 xiv2 ¼ sxiv1 sxiv2
1282 Furr
Decomposition of Individual Profiles
As described earlier, an individual’s profile of scores can be seen as

arising from a normative profile and from a distinctive profile. More
technically, an individual’s observed score on a variable in a profile
can be seen as a composite of two elements—a normative element
and a distinctive element:
xivp ¼ xvp þ x0ivp Eq:1
where xvp is the mean score on the variable in that profile, averaged
over all individuals, and where x0ivp is the degree to which an indi-
vidual’s score on the variable in that profile deviates from the mean
score on the variable in that profile (i.e., x0ivp ¼ xivp xvp ). The
mean score reflects the group’s normative level of that variable in
that profile, and the deviation score reflects the degree to which the
individual is ‘‘distinctive’’ from the group norm on the variable in
the profile. An individual’s scores on a variable from two profiles
(e.g., a self-profile and an informant profile) are:
xiv1 ¼ xv1 þ x0 iv1 Eq:2a
xiv2 ¼ xv2 þ x0 iv2 Eq:2b

Table 2 presents a small example, with three individuals measured
on five variables (i.e., traits) by two raters—self and informant (i.e.,
Profile 1 and 2, respectively). Note, for example, that Kyle’s (indi-
vidual 1) score on Openness trait 3) in his self-rated profile (x131 5 1)
is the sum of the mean score on Openness across all self-ratings sit-
uation 1 ð x31 ¼ 2Þand his distinctiveness on Openness within the
self-ratings 1 ðx0131 ¼ 1). Thus, Kyle describes himself as distinc-
tively low on Openness.
Initial Decomposition of Profile Similarity
There are several ways to quantify profile similarity (e.g., Cattell

1949; Cohen, 1969; Cronbach & Gleser, 1953; Furr, 2007; McCrae,
1993), and the current framework focuses on the covariance and
correlation between profiles. A positive covariance or correlation
between profiles indicates that the relative order of variables in one
profile is similar to the relative order of variables in the other profile.
For example, a target individual’s extraversion, agreeableness, and
openness might be rated by himself and by an informant, with the
Table 2
Example Data for Computing Components of Similarity,
Normativeness, and Distinctiveness
Target 5 Kyle Distinctive Distinctive Kyle’s

Trait Self Inf. Self Inf. Mean
Rating Rating Rating Rating Profile
Neuroticism 1 1 2 0 1
Extraversion 3 3 1 0 3
Openness 1 2 1 1 1.5
Agreeableness 5 5 0 0 5
Conscientiousness 4 6 1 0 5
Mean 2.8 3.4 1 0.25 5
Std Dev 1.60 1.85 .71 .43 3.42
Target 5 Stan Distinctive Distinctive Stan’s

rating rating rating rating Profile
Extraversion 3 3 1 0 3
Agreeableness 5 6 0 1 5.5
Conscientiousness 5 5 0 1 5
Mean 3.6 4 0.25 0.75 3.8
Std Dev 1.20 1.79 .43 .83 1.29
Target 5 Eric Distinctive Distinctive Eric’s

rating rating rating rating Profile
Extraversion 6 3 2 0 4.5
Agreeableness 5 4 0 1 4.5
Conscientiousness 6 7 1 1 6.5
Mean 5 3.4 1.25 0.5 4.2
Std Dev 1.10 2.06 .83 .50 1.40
Norms Normative Normative Global

Trait Self-rating Inf. rating Mean
Profile
Neuroticism 3 1 2
Extraversion 4 3 3.5
Openness 2 3 2.5
Agreeableness 5 5 5
Conscientiousness 5 6 5.5
Mean 3.5 3 3.25
Std Dev 1.12 1.41 1.15
1284 Furr
target’s self-ratings indicating high extraversion, moderate agree-

ableness, and low openness. A positive covariance or correlation
between the self-rated profile and an informant-profile indicates that
the informant also sees the target as more extraverted than agree-
able, and more agreeable than open. Thus, the relative order of the
trait ratings in one profile (i.e., self-ratings) is similar to the relative
order of the trait ratings in the other profile (i.e., informant-ratings).
Because an individual’s observed score on each variable is concep-
tualized as a composite, the covariance between two profiles of scores
can be seen as the covariance between two composites. In this frame-
work, the covariance between two profiles initially can be decomposed
into four separate covariances (Nunnally & Bernstein, 1994, p. 171):
sxiv1 xiv2 ¼ sx0iv1 x0iv2 þ sxv1 xv2 þ sxv1 x0iv2 þ sxv2 x0iv1 Eq:3
The sxiv1 xiv2 component is Overall Similarity (i.e., the covariance

between individual i’s two profiles of observed scores), sx0iv1 x0iv2 is Dis-
tinctive Similarity (i.e., the covariance between the individual’s two
profiles of distinctiveness scores), and sxv1 xv2 is Generalized Nor-
mative Similarity (i.e., the covariance between the two group mean
profiles). Two additional components emerge from this initial de-
composition of overall similarity, although they are less likely to
have wide psychological meaning. Specifically, the sxv1 x0iv2 compo-
nent is the covariance between the individual’s distinctiveness scores
for Profile 2 and the group means for Profile 1, and sxv2 x0iv1 is
the covariance between the individual’s distinctiveness scores for
Profile 1 and the group means for Profile 2. See Appendix A for
algebraic details of this derivation, and for terms used in the two
substantive models.
This initial decomposition lays the psychometric foundation for
Model 1 and Model 2 (see Table 1), which are likely to have broad
psychological meaning and utility. Although a variety of models could
be derived, Model 1 and Model 2 are likely to be applicable across a
wide range of phenomena. The two models are equivalent in terms of
producing the same values for Overall Similarity, but they reflect
components with meaningfully different psychological implications.
Profile Similarity Model 1
As described earlier, Model 1 includes Distinctive Similarity, Gen-

eralized Normative Similarity, and three components of profile-level
normativeness (i.e., two Within-profile normativeness components

and a Global Normativeness component). On a covariance metric,
the initial decomposition can be re-parameterized (see Appendix B
for details) to reflect an individual’s Overall Similarity score as:
sxiv1 xiv2 ¼ sx0iv1 x0iv2 sxv1 xv2 sxv1 xiv1 sxv2 xiv2 þ 4sxiv xv Eq:4
where the first three components (i.e., sxiv1 xiv2 ; sx0iv1 x0iv2 , and sxv1 xv2 ) are
defined above—Overall Similarity, Distinctive Similarity, and Gen-
eralized Normative Similarity. Model 1 includes three additional
components. One Within-Profile Normativeness component reflects
the similarity between the individual’s scores in Profile 1 and the
group norms for Profile 1. Specifically, sxv1 xiv1 is covariance between
individual i’s scores in Profile 1 and the group mean scores for Profile
1. A second Within-Profile Normativeness component reflects the
similarity between the individual’s scores in Profile 2 and the group
norms for Profile 2. Specifically, sxv2 xiv2 is covariance between indi-
vidual i’s scores in Profile 2 and the group mean scores for Profile 2.
The final component of Model 1 is Global Pair-Level Normative-
ness, reflecting the degree to which the average of a pair of profiles
matches the average of all profiles. As discussed earlier (see Figure 1)
this component reflects the degree to which a given target (as seen by
self and informant) matches the average target (as seen by all selves
and all informants). Again, Global Normativeness is the match be-
tween two new profiles. The first is a ‘‘Pair Mean’’ profile—a set of
scores averaged across both profiles in a given pair, with each score
defined as:
xiv ¼ ðxiv1 þ xiv2 Þ=2 Eq:5
In self-other agreement, this reflects an individual’s profile of average
scores across a pair of judges (i.e., self and informant). The second
new profile is a ‘‘Global Mean’’ profile—a set of scores averaged
across all individuals and across both sets of profiles, with each score
defined as:
xv ¼ ðxv1 þ xv2 Þ=2 Eq:6
In self-other agreement, this reflects the average person as rated
across both kinds of judges. Based on these two profiles, Global
Pair-Level Normativeness is the similarity between a Pair Mean
profile and the Global Mean profile. Most generally, this indicates
the degree to which the individual on average is similar to the group
norm across all profiles. Table 2 presents the two averaged profiles
1286 Furr
for the example data. Table 3 presents the Model 1’s similarity co-
variances and correlations for the data in Table 1.
The derivation of the current framework emerges most easily in
terms of covariances, but most applications of profile similarity are
likely to use a correlations. Thus, Model 1 can be translated onto a
correlational metric:
rxiv1 xiv2 ¼
sx0iv1 sx0iv2 rx0iv1 x0iv2 sxv1 sxv2 rxv1 xv2 sxv1 sxiv1 rxv1 xiv1 sxv2 sxiv2 rxv2 xiv2 þ 4sxiv sxv rxiv xv
sxiv1 sxiv2
Eq:7
Table 1 summarizes Model 1 at the conceptual level, on a covari-

ance metric, and on a correlational metric.
Profile Similarity Model 2
As described earlier, Model 2 includes Distinctive Similarity, Gen-

eralized Normative Similarity, and two components of Profile-Level
Normativeness (i.e., two Cross-Profile Normativeness components).
On a covariance metric, the initial decomposition can be re-param-
eterized (see Appendix C for details) to reflect an individual’s Overall
Similarity score as:
sxiv1 xiv2 ¼ sx0iv1 x0iv2 sxv1 xv2 þ sxv1 xiv2 þ sxv2 xiv1 Eq:8
Again, sxiv1 xiv2 , sx0iv1 x0iv2 , and sxv1 xv2 represent Overall Similarity,
Distinctive Similarity, and Generalized Normative Similarity, re-
spectively. The first Cross-Profile Normativeness component is
sxv1 xiv2 —the covariance between individual i’s scores in Profile 2
and the group mean scores for Profile 1. As discussed earlier in terms
of marital similarity, this is the similarity between a particular hus-
band and the average wife. The second Cross-Profile Normativeness
component is sxv2 xiv1 —the covariance between individual i’s scores in
Profile 1 and the group mean scores for Profile 2. In terms of marital
similarity, this is the similarity between a particular wife and the
average husband.
Again, researchers working with profile similarity will generally
prefer a correlational similarity index (e.g., Luo & Klohnen, 2005).
Table 3
Profile Similarity Components for Three Models: Covariances and Correlations
Models/Profile Similarity Components
Both Models Model 1 Model 2
Overall Distinctive Gen. Norm. Self Other Global Other-self Self-Other

Individual Agreement Agreement Agreement Norm Norm Norm Norm Norm
Covariances
sxiv1 xiv2 sx0iv1 x0iv2 sxv1 xv2 sxv1 xiv1 sxv2 xiv2 sxiv xv sxv1 xiv2 sxv2 xiv1
Kyle 2.680 .000 1.520 1.760 3.160 2.280 1.880 2.320
Stan 1.000 .080 1.520 1.320 2.600 1.590 .800 1.640
Eric 1.200 .240 1.520 1.000 3.360 1.710 1.880 .600
Correlations
rxiv1 xiv2 rx0iv1 x0iv2 rxv1 xv2 rx0v1 xiv1 rxv2 xiv2 rxiv xv rxv1 xiv2 rxv2 xiv1
Kyle .903 .000 .748 .943 .977 .999 .869 .832
Stan .466 .196 .748 .943 .834 .899 .383 .784
Eric .532 .429 .748 .783 .936 .825 .783 .314
1288 Furr
On a correlational metric, Model 2 is:

sx0iv1 sx0iv2 rx0iv1 x0iv2 sxv1 sxv2 rxv1 xv2 þ sxv1 sxiv2 rxv1 xiv2 þ sxv2 sxiv1 rxv2 xiv1
rxiv1 xiv2 ¼
sxiv1 sxiv2
Eq:9
Table 1 summarizes Model 2 at the conceptual level, on a covari-
ance metric, and on a correlational metric.
Computations
The previous sections presented conceptual and psychometric foun-
dations of a framework differentiating similarity, normativeness,
and distinctiveness. The current section provides computational
foundations. Applied to the hypothetical data in Table 2, they pro-
duce the values in Table 3.1
The computational process can be summarized in three steps,
framed here in terms of self-other agreement. The first step is to
compute the normative profile for each of the two sets of profiles
(i.e., the xv1 values and the xv2 values for each variable, see the
Mean Self-Rating Profile and the Mean Informant-Rating Profile in
Table 2). One normative profile is the set of ‘‘self’’ mean scores for
each variable averaged across all targets’ self-ratings. The other nor-
mative profile is the ‘‘informant’’ set of mean scores for each vari-
able, averaged across all informants’ ratings. The second step is to
compute distinctive profiles, along with other mean profiles related
to Model 1 and/or Model 2. These profiles are computed as differ-
ences between profiles (i.e., a distinctive profile is computed as
x0ivp ¼ xivp xvp ) or as averages across profiles (e.g., Eric’s Mean
Profile). In the third step, correlations among all profiles are com-
puted for each pair (e.g., rxiv1 xiv2 , rx0iv1 x0iv2 ), producing the various com-
ponents seen Model 1 and 2.
As revealed by the correlations reported in Table 3, all three tar-
gets have relatively high self-other agreement. Kyle has the highest
Overall Agreement correlation, reflecting the fact that relatively high
scores in his self-rated profile of traits tend to be paired with rela-
tively high scores in his informant’s profile of ratings (and low scores
are paired with low scores). More specifically, Kyle describes himself
as more Agreeable and Conscientious than Neurotic, Extraverted,
1. A SAS program written to conduct these analyses is available online at: http://
www.wfu.wdu/ furrrm/profile1.sas
and Open, and Kyle’s informant describes him in a similar way.

Note that, for all three targets, Distinctive Agreement correlations
are lower than the Overall Agreement correlations. This pattern of
results indicates that, to some degree, Overall Agreement is driven by
normativeness. Nevertheless, two of the targets have positive Dis-
tinctive Agreement. For example, Eric’s level of Distinctive Agree-
ment indicates that the ways in which he described himself
distinctively (i.e., different from the normative target’s self-rating)
are in agreement with the ways in which he was described distinc-
tively by his informant. Note that all three targets have identical
values for Generalized Normative Agreement, which is the similarity
between the normative self-rating profile and the normative infor-
mant-rating profile.
Table 3 also shows the normativeness indices within the two
models. For Model 1, Kyle’s strong positive ‘‘Self-Normativeness’’
correlation reveals that he sees himself in a highly normative man-
ner. Similarly, Kyle’s high level of ‘‘Other Normativeness’’ indicates
that his informant sees him as typical, as compared to the way that
most informants saw their targets. Finally, the Global Normative-
ness correlations reflect the degree to which each participant’s mean
trait profile is similar to the group’s mean trait profile across both
raters (i.e., the Global Mean Profile in Table 2). For example, Kyle’s
very strong positive Global Normativeness correlation indicates that
he was seen by himself and his informant as nearly identical to
the average person. That is, his general trait profile is similar to the
average person’s general trait profile. For Model 2, there are
two Cross-Profile Normativeness correlations For example, Kyle’s
large ‘‘Other-self-normativeness’’ correlation indicates that his in-
formant described him has having a high level of traits that charac-
terized most self-ratings to a relatively high degree. That is, his
informant’s description of Kyle closely matches the typical target’s
self-rating.
Note that the current models produce components that are not
necessarily orthogonal across participants. However, by using cor-
relations as the indices of profile similarity (instead of covariances),
researchers avoid confounding individual differences in degree of
similarity with individual differences in degree of profile variability.
The covariances discussed earlier lay the psychometric foundation
for the framework, but researchers are most likely to examine the
components in terms of correlations.
1290 Furr
Applications of the Framework

Thus far, the framework’s meaning and its potential application
have been articulated within two domains of research: self-other
agreement and personality similarity. This section presents a broader
perspective on the framework’s meaning and utility by discussing (a)
insights gained through actual applications of the framework (or of
analysis that are conceptually consistent with the framework); (b)
applications of the framework in other common uses of profile sim-
ilarity (i.e., cross-situational behavioral consistency and temporal
stability of personality, see Table 4); and (c) the meaning of com-
ponents that were not discussed in the earlier examples (e.g., the
application of Model 2 components to self-other agreement). These
applications are intended to (a) illustrate that the current framework
is likely interpretable in any examination of profile similarity, (b)
provide a language for discussing the components within four very
common applications of profile similarity, and (c) provide sufficient
background so that the framework can be adapted easily to other
applications of profile similarity. As these applications illustrate, the
current framework can be used to evaluate the normativeness prob-
lem as a potential confound and to allow hypothesis-testing aimed
specifically for various components of normativeness and distinc-
tiveness.
Self-Other Agreement
There are at least two studies illustrating the useful application

of the current framework to self-other agreement. These studies
provide evidence for at least two important issues: (a) the utility
of examining Distinctive Similarity alongside Overall Similarity
and (b) the utility of examining one or more form of Norma-
tiveness to provide deeper insight into self-other agreement
processes.
One example is an examination of the hypothesis that self-other
agreement is lower among adolescents with personality pathology
than among adolescents without personality pathology (Furr et al.,
2007). Adolescents—some diagnosed with Conduct Disorder
(CD) and others without any disorder—rated themselves and were
rated by their mothers on a trait measure, and hypotheses
suggested lower agreement among adolescents with CD than among
Table 4
Profile Similarity Components Applied to Common Domains of Profile Research
Profile Similarity Components
Both Models Model 1 Model 2
General Terms
Overall Distinctive Generalized Within-profile Within-profile Global Cross-profile Cross-profile
Similarity Similarity Normative Normativeness 1 Normativeness 2 Normativeness Normativeness 1 Normativeness 2
Similarity
Self-other personality agreement (e.g., Furr et al., 2007)

Overall Distinctive Generalized Self Perceived Other Global Self-other Other-self
Personality Personality Normative Normativeness Perceived Personality Normativeness Normativeness
Agreement Agreement Agreement Normativeness Normativeness
Husband-wife similarity (e.g., Luo & Klohnen, (2005)
Overall Distinctive Generalized Husband Wife Global Couple Husband-wife Wife-husband
Personality Personality Normative Personality Personality Normativeness Similarity Similarity
Similarity Similarity Similarity Normativeness Normativeness
Personality stability (e.g., Ozer & Gjerde, 1989)

Overall Distinctive Generalized Time 1 Time 2 Global Developmental Developmental
Personality Personality Normative Personality Personality Personality Hyper-maturity Immaturity
Stability Stability Stability Normativeness Normativeness Normativeness
Cross-situational behavioral consistency (e.g., Furr & Funder, 2004)

Overall Distinctive Situational Situation 1 Situation 2 Global Cross-situational Cross-situational
Behavioral Behavioral Similarity Behavioral Behavioral Behavioral Behavioral Behavioral
Consistency Consistency Normativeness Normativeness Normativeness Normativeness 1 Normativeness 2
1292 Furr
control adolescents. For each adolescent, Overall Agreement,

Distinctive Agreement, Self-Perceived Normativeness, and Infor-
mant-Perceived Normativeness were calculated, reflecting elements
of Model 1.2
Results revealed3 the utility of components of Normativeness and
Distinctive Similarity. Results indicated that, as expected, Overall
Agreement was lower among adolescents with CD than among con-
trol adolescents. Results also revealed significantly lower levels of
normativeness within the CD group than within the control group,
raising the possibility that group difference in Overall Agreement
could emerge solely from group differences in normativeness. Evalu-
ating this facet of the normativeness problem, analyses revealed that,
despite differences in normativeness, the CD group had a significantly
lower level of Distinctive Agreement than the control group. Thus,
Furr et al. eliminated the possibility that the group difference in Over-
all Agreement was an artifact of group differences in normativeness.
A second example of the current framework’s application to self-
other agreement is an examination of the acquaintanceship effect—
Biesanz and his colleagues (2007) hypothesized that high levels of
acquaintance between targets and informants enhances self-other
agreement. They argued that previous research had produced am-
biguous results, because those examinations had not differentiated
similarity, normativeness, and distinctiveness. Biesanz et al. used
procedures consistent with Model 2, predicting different results for
three components.3 The first component was Overall Personality
Agreement, called raw self-acquaintance consensus by Biesanz et al.
The second was Other-self Normativeness, which is the correlation
2. To account for the large trait differences between the two groups, the
normativeness and distinctiveness scores were computed within each group.
That is, the normativeness and distinctiveness scores for a CD adolescent reflect
the way he or she compares to the average CD adolescent, and the normativeness
and distinctiveness scores for a control adolescent reflect the way he or she
compares to the average control adolescent.
3. Although their approach was very consistent with the conceptual and psycho-
metric logic of the current frameworks, Beisanz et al. (2007) quantified distinctive
agreement is a way that differs slightly from the framework articulated in this
article Specifically, whereas the current framework mean deviates each response
from the normative profile, they standardized each response in relation to the
normative profile. That is, they mean deviated each response from the normative
profile and then adjusted for normative standard deviations.
between an acquaintance’s profile of trait ratings and the average

profile of self-ratings. Biesanz and his colleagues conceptualized this
as stereotype accuracy. The third component was very closely akin to
Distinctive Personality Agreement, called differential accuracy. The
researchers predicted that increasing acquaintance would produce
no stable effect on Overall Agreement, it would decrease Other-self
Normativeness, and it would increase Distinctive Agreement.
Results generally confirmed these predictions across two studies.
Specifically, Biesanz et al. (2007) found a nonsignificant association
between acquaintance and Overall Agreement, a significant negative
association between acquaintance and Other-self Normativeness, and
a significant positive association between acquaintance and Distinctive
Agreement. By differentiating similarity, normativeness, and distinc-
tiveness, Biesanz and his colleagues resolved ambiguities from previ-
ous studies. Specifically, they found that increasing acquaintanceship
enhances our ability to understand what makes an individual different
and unique. Consequently, as our level of acquaintance increases, our
impressions rely ‘‘less on how people are in general’’ (p. 132).
Marital Similarity
To my knowledge, previous examinations of marital similarity have

not included strategies akin to the current framework. However, as
described earlier (see Figure 2), Model 2 could be usefully applied to
marital similarity (i.e., Distinctive Personality Agreement, General-
ized Personality Similarity, Husband-Wife Normativeness, and
Wife-Husband Normativeness, etc). Such an application could pro-
duce insights into the antecedents and consequences of similarity
between husbands and wives.
As presented in Table 4, two components of Model 1 also
might be usefully applied in this domain. Specifically, the two
Within-Profile Normativeness components would provide informa-
tion regarding the normativeness of each spouse’s personality, in
the context of his or her own sex. Husband Personality Normative-
ness reflects the degree to which a particular husband is similar to
the average husband, and Wife Personality Normativeness reflects
the degree to which a particular wife is similar to the average wife.
More technically, these components are the correlations between an
individual’s trait profile and the trait profile of the average husband
(for husbands) or wife (for wives).
1294 Furr
Personality Stability Across Time
Developmental researchers have examined the stability of personal-

ity profiles across time—to what degree is a person’s trait profile at
one developmental period similar to his or her profile at another
developmental period? To my knowledge, profile-based examina-
tions of temporal stability have not included strategies akin to the
current framework; the framework, however, offers potential in-
sights in this area (see Table 4). Personality stability can be defined as
the degree to which an individual’s personality profile remains stable
across time. Thus, Overall Personality Stability is the similarity be-
tween an individual’s trait profile at one time and his or her trait
profile at a later time (e.g., Ozer & Gjerde, 1989).
Overall Personality Stability might have multiple sources, intro-
ducing ambiguity into its interpretation. An individual might obtain
high Overall Stability in at least two ways. First, high Overall Sta-
bility could reflect a tendency to retain highly idiosyncratic person-
ality profile in the face of changing age norms. That is, each phase of
psychological development may have a normative personality pro-
file, but an individual might be nonnormative in important ways. To
the degree that those nonnormative or distinctive qualities are stable
across time, the individual will exhibit high Overall Stability. Second,
high Overall Stability could reflect a tendency to be consistently
normative. In such cases, high Overall Stability emerges if the nor-
mative profiles of the time points are similar to each other. That is, if
two developmental periods have similar norms and if an individual is
normative within each age period, then he or she will exhibit high
Overall Stability. Moreover, normative profiles are likely to reflect
social desirability or psychological well-being. If so, an Overall Sta-
bility that is driven by normativeness reflects both stability and de-
sirability. This presents a potential confound in analyses of Overall
Stability.
In personality stability, the components of Model 1 are highly
informative. Distinctive Personality Stability is the degree to which
an individual retains a no-normative profile across time, and it arises
from the individual’s two distinctive trait profiles. These profiles re-
flect the distinctive levels of each trait measured at the two time
points (e.g., scores representing the degree to which the individual
was distinctively extraverted at both times). Thus, Distinctive Per-
sonality Stability is the degree to which an individual’s distinctive
trait profile at one time is correlated with his or her distinctive trait
profile at the later time. A second component is Generalized Nor-
mative Stability, which reflects the degree to which the two age pe-
riods have similar normative trait profiles. The third and fourth
components are Within-time Normativeness components reflecting
the normativeness of an individual’s personality within each age pe-
riod (i.e., the degree to which an individual is like the average person
in each age period). Each of these is a correlation between an indi-
vidual’s trait profile at one time and the normative profile for that
time point. The fifth component is Global Personality Normative-
ness, reflecting the normativeness of an individual’s general trait
profile. More specifically, it is the correlation between the individ-
ual’s Mean Trait Profile (i.e., his or her trait profile, averaged across
both age periods) and the Global Trait Profile (i.e., the trait profile of
the average person averaged across both time points).
The two Cross-Profile Normativeness components also reveal in-
teresting concepts in personality stability. One Cross-Profile Nor-
mativeness component can be interpreted as Developmental Hyper-
maturity, reflecting the degree to which an individual’s personality at
a relatively young age resembles the normative personality of an
older age period. The other Cross-Profile Normativeness component
might be interpreted as Developmental Immaturity, reflecting the
degree to which an individual’s personality at a relatively old age
resembles the normative personality of a younger age period.
Cross-Situational Behavioral Consistency
From a profile-based approach, cross-situational behavioral consis-

tency can be defined as the degree to which an individual’s profile of
behaviors in one situation is consistent with his or her profile of be-
haviors in a second situation (e.g., Furr & Funder, 2004). That is, it
is the degree to which the most salient behaviors that a person ex-
hibits in one situation are also the most salient behaviors that he or
she exhibits in another situation. The current framework can be ap-
plied very usefully to this domain (see Table 4), and Overall Con-
sistency is the correlation between an individual’s behavioral profile
as measured in one situation and his or her behavioral profile in
another situation.
Overall Consistency might have multiple sources, creating ambi-
guity in its interpretation. An individual might obtain high Overall
1296 Furr
Consistency in at least two ways. First, high Overall Consistency

could emerge from a highly personality-driven behavioral style. That
is, an individual’s behavior might be determined strongly by his or
her distinctive personality characteristics, rather than situational
cues. To the degree that those personality characteristics are stable
across situations, the individual will exhibit high behavioral Overall
Consistency. Second, an individual’s behavior might be determined
strongly by situational forces, rather than distinctive personality
characteristics. In such cases, high Overall Consistency emerges if the
two situations are highly similar to each other. That is, if two sit-
uations have highly similar norms or cues, and if an individual be-
haves solely in response to those cues, then he or she will exhibit high
Overall Consistency. Although few, if any, behaviors are either to-
tally immune to situational cues or totally determined by situational
cues, a given individual might tend toward one possibility or the
other. The current framework affords insight into such potentially
important tendencies.
In behavioral consistency, the components of Model 1 are highly
informative. Distinctive Consistency reflects the degree to which an
individual behaves in a consistently non-normative manner, and it
arises from the individual’s two distinctive profiles of behavior.
These profiles reflect the distinctive levels of each behavior mea-
sured in the situations (e.g., scores representing the degree to which
the individual was distinctively talkative in Situation 1). Thus, Dis-
tinctive Consistency is the degree to which an individual’s distinctive
profile in one situation is correlated with his or her distinctive profile
in the second situation. High Distinctive Consistency emerges only
when an individual consistently acts in contrast to situational cues. A
second component is Situational Similarity, reflecting the degree to
which situations have similar behavioral norms. It arises from
normative behavioral profiles in each of two situations—the behav-
ioral profile of the average person in each situation. The third
and fourth components are Within-Situation Normativeness com-
ponents reflecting the normativeness of an individual’s behavior in
each situation (i.e., the degree to which an individual acts like
the average person in each situation). Each is the correlation be-
tween an individual’s behavioral profile in a situation and the
normative profile in that situation. The fifth component is Global
Behavioral Normativeness, reflecting the normativeness of an
individual’s general behavioral style. It is the correlation between
the individual’s Mean Behavioral Profile (i.e., his or her behavioral

profile, averaged across situations) and the Global behavioral profile
(i.e., the behavioral profile of the average person in the average
situation).
Although Distinctive Consistency, Situational Similarity, and the
other components of Model 1 are informative in any analysis be-
havioral consistency, Model 2’s Cross-Profile Normativeness
components may be meaningful only when the situations are
highly complementary. For example, a researcher might observe
participants in an interaction with a close friend and in an interac-
tion with a stranger. One Cross-Profile Normativeness would
reflect the degree to which an individual’s behavior with a friend
resembles the typical person’s behavior with a stranger (i.e., pre-
sumably revealing an unusual level of psychosocial distance, or ill at
ease with a friend). The other Cross-Profile Normativeness would
reflect the degree to which the individual’s behavior with a stranger
resembles the typical person’s behavior with a friend (i.e., presum-
ably revealing an unusual level of familiarity, ease, or closeness to a
stranger).
Further illustrating the utility of the framework, components of
the framework have been applied to the examination of the link be-
tween behavioral consistency and self-monitoring. Researchers
have hypothesized that low self-monitors are more behaviorally con-
sistent than high self-monitors (Gangestad & Snyder, 2000; Snyder
& Ickes, 1985), but previous research provides mixed support for
this prediction (Cheek, 1982; Snyder & Monson, 1975; Wymer &
Penner, 1985). Furr and Funder (2005) adopted a profile-similarity
approach to examine this prediction. In this research, participants
completed the self-monitoring scale, and they engaged in dyadic
interactions in two situations. Behavior was measured in each
situation, and an Overall Behavioral Consistency correlation was
computed for each participant, representing the degree to which
he or she exhibited a consistent profile of behavior across the
pair of situations. Analyses of Overall Consistency revealed that be-
havioral consistency was correlated with self-monitoring, but
this effect was rather small, only marginally significant, and in
the opposite direction of predictions (r 5 .17, po.10). These results
appear to contradict predictions again: if there is any difference,
high self-monitors seem somewhat more consistent than low
self-monitors.
1298 Furr
Although Furr and Funder’s (2005) analysis of Overall Consis-

tency seems to contradict the self-monitoring hypothesis, a consid-
eration of Distinctive Behavioral Consistency offers new insight into
the phenomenon. Specifically, high self-monitors might manifest a
relatively normative behavioral style—managing their behavior to
conform to situational cues and norms. In contrast, low self-mon-
itors might manifest a more distinctive style—basing their behavior
on their distinctive personality characteristics more than on situa-
tional norms. Since Overall Behavioral Consistency is affected by
normativeness, the normativeness of high self-monitors could be
‘‘competing’’ with the distinctiveness of low self-monitors. This pro-
duces results that reflect neither clearly. To examine this possibility,
components from Model 1 were computed for each participant, and
correlations between these components and self-monitoring were
examined (Furr & Funder, 2005). As expected, self-monitoring was
positively correlated with all indices of normativeness, suggesting
that high self-monitors tend to be behave in relatively normative
ways. Conversely, self-monitoring was negatively correlated with
Distinctive Consistency (r 5 .18, po.05), as expected. Although
this correlation was not large, it was statistically significant and its
direction was consistent with the original predictions regarding self-
monitoring and consistency.
Thus, the pair-level decomposition of behavioral consistency re-
veals importantly different facets of behavioral consistency, behav-
ioral normativeness, and their links to self-monitoring. Considering
several components together provided deeper insight into the psy-
chological meaning of behavioral consistency, in this case producing
results consistent with theory.
Implications and Recommendations

The framework presented in this article allows researchers to exam-
ine profile similarity in a way that copes with the normativeness
problem. A primary purpose of the current article is to present the
conceptual and psychometric foundations of the framework and to
articulate its wide applicability and meaning. Several implications
and recommendations emerge from these applications.
When Is the Framework Applicable?
The primary implication is that the framework is likely to be
applicable to any analysis of the similarity between two profiles of
traits or behaviors. As illustrated in Table 4, the framework provides

insights into a variety of domains in which profile similarity is
examined.
The normativeness problem must be considered when
examining profile similarity, and the current framework offers
two opportunities in this regard. First, it allows researchers to gauge
and potentially eliminate normativeness as a potential confound.
For example, a high level of self-other agreement could arise
from (a) an informant’s genuine understanding of a target’s
personality, (b) coincidental normativeness (i.e., any informant’s
rating is likely to be similar to any target’s self-rating), or (c) a
psychological characteristic that is associated with normative-
ness and thereby associated with agreement (e.g., social desirability
or psychological adjustment). An analysis in which each
target/informant pair’s Overall Agreement is decomposed into
normative and distinctive components affords the opportunity
to understand more fully the processes driving the correlation
and to evaluate the potential confounding effects of norma-
tiveness.
A second opportunity afforded by the current framework is that it
can reveal findings that would otherwise remain hidden. Normative-
ness should not be seen simply as a nuisance variable or potential
confound; instead, normativeness and distinctiveness are
potentially important psychological phenomena in their own right,
and the current framework can help disentangle their effects (e.g.,
Biesanz et al., 2007). For example, researchers might find that
couples in which the spouses are both normative might have
different qualities than couples in which the spouses are nonnorma-
tive. Furthermore, these qualities might differ from the qualities
of couples in which the spouses share distinctive qualities.
Thus, the differentiation of similarity, normativeness, and distinc-
tiveness can expand the scope of psychological questions to be
addressed.
Whether the framework is interpreted in terms of gauging poten-
tial confounds (e.g., social desirability inherent in the normative
profiles) or in terms of gaining insights into important new facets of
similarity, distinctiveness and normativeness (e.g., developmental
hyper-maturity), researchers are likely to benefit from the informa-
tion provided by the current framework within in any application of
profile similarity.
1300 Furr
Which Components Should Be Examined?
The current framework provides a psychometrically coherent per-

spective that includes a variety of components related to similarity,
normativeness, and distinctiveness (see Table 1); however, the com-
ponents are more important in practice than are the mathematical
associations among them. The purpose of the psychometric models
outlined in this article is to articulate precisely the formal derivation
of overall similarity from a coherent set of components and to dem-
onstrate that the components comprise a meaningful framework in-
tegrating similarity, normativeness, and distinctiveness. In practice,
the components can be used with little concern for the formal psy-
chometric connections among them.
One’s choice of components should fit one’s specific research
questions. As illustrated earlier, all components are meaningful in
most applications of profile similarity. However, some components
might be most fundamental to particular questions or hypotheses.
For example, in their examination of self-other agreement, Furr
et al. (2007) examined Overall Agreement, Distinctive Agreement,
and two indexes of Within-Profile Normativeness. Thus, they fo-
cused on some, but not all, elements of Model 1. Similarly, Biesanz
et al. (2007) examined Overall Agreement, Distinctive Agreement,
and one Cross-Profile Agreement index—focusing on some, but not
all, elements of Model 2. Finally, Starzyk et al. (2006) focused solely
on an index akin to Distinctive Agreement. Although such choices
reflect the specific research questions being addressed, they may also
reflect the lack of a coherent framework for profile similarity. The
availability of a coherent and generally applicable framework may
enhance the range of components and questions that researchers
consider examining in their analysis.
In addition to their differential relevance to specific psychological
questions, some components may have less meaning when applied to
particular forms of data. At least two such cases may exist. First,
Global Normativeness may be less meaningful when each pair of
profiles characterizes two people instead of one. For example,
Global Normativeness may not be highly meaningful in analyses
of marital similarity. In such cases (i.e., cases in which one profile
describes one individual and the second describes another), the Pair
Mean profile would describe the ‘‘average person in the couple.’’
This component is likely to be more meaningful when each profile
within a pair describes a single person. For example, in the case of

self-other personality agreement, both profiles describe a target in-
dividual. Therefore, the Pair Mean profile reflects the target
person as averaged across two raters, and Global Normativeness
reflects the degree to which the target person’s average profile fits the
average person in the entire sample. The Cross-Profile Normative-
ness components are a second case in which some components might
have less utility. Specifically, in cross-situational behavioral consis-
tency, these components might be interesting only when the two
situations are mutually complementary. Recall that, in cross-situa-
tional behavioral consistency, the Cross-profile normativeness com-
ponents reflect the correlation between one’s behavior in a situation
and the normative behavioral profile in a second situation. If the two
situations are not complementary to each other, then the Cross-
Profile Normativeness components from Model 2 are probably less
interesting than the Within-Profile Normativeness components from
Model 1.
Profile distinguishability also has implications for the application
of the current models. The current models are highly applicable for
profiles that are clearly distinguishable from each other. For exam-
ple, the personality-rating profiles in Table 1 can be distinguished in
terms of the rater who provided them (i.e., self or informant). Like-
wise, in marital similarity, researchers distinguish husbands’ profiles
from wives’ profiles. But not all profile-oriented research involves
distinguishable profiles. For example, analysis of roommate
similarity requires two profiles—one for roommate 1 and one for
roommate 2— but which roommate should be designated as 1 or 2?
When there is no clear way to distinguish one profile from the
other, the profiles are nondistinguishable. Furthermore, when
working with nondistinguishable profiles, the procedures outlined
in this article could produce results that depend partially on the
way in which profiles are designated. Models to eliminate such
ambiguity could be developed for ‘‘nondistinguishable’’ analytic
situations. Until then, researchers could use the current frameworks
using a multistage analytic strategy. That is, researchers could ran-
domly assign each member of the dyad to profile 1 or profile 2,
run analyses based on the current models, reassign members to the
profiles, rerun the analysis, and so on. This process could be con-
ducted repeatedly, and the results could be aggregated across the
repetitions.
1302 Furr
No Index of Pure Similarity
As implied by the current models, there is no index of ‘‘pure’’ sim-

ilarity. That is, no component reflects profile similarity as completely
independent of normativeness and distinctiveness. The components
represent blends of similarity, distinctiveness, and normativeness,
and they can be used to gauge the presence and potential meaning of
normativeness and distinctiveness. Because no single component
represents pure’’similarity, the components are most informative
when examined together.
Relatedly, when examining Distinctive Similarity, researchers
should recognize that it is an important but conservative form of
profile similarity. That is, Distinctive Similarity reflects similarity
only in ways in which an individual is distinctive from the group
norm. For example, in self-other agreement, Distinctive Agreement
reflects the degree to which a target and a judge agree on the ways in
which the target is nonnormative. Therefore, when computing Dis-
tinctive Agreement scores, no ‘‘credit’’ is given for accurate evalu-
ations of the ways in which a target is truly normative.
Consequently, a relatively normative target is likely to receive a
lower Distinctive Agreement correlation than is a relatively nonnor-
mative target. Given the conservative nature of Distinctive Similar-
ity, logic suggests that it often should be examined alongside other
indices of similarity and normativeness, particularly Overall Simi-
larity. Thus, Distinctive Similarity is not the index of pure similarity;
rather, it is an important component of the complete picture of sim-
ilarity. A value of the current framework is that it provides coherent
conceptual and psychometric models that tie Distinctive Similarity
to Overall Similarity and to meaningful facets of Normativeness.
Researchers using this framework can have confidence that the com-
ponents are conceptually and statistically cohesive, and they can
understand precisely how the components complement each other.
Scaling and Fisher’s Transformation
Two additional technical recommendations are relevant to correla-

tional profile analysis. First, researchers examining profile similarity
should evaluate Cohen’s (1969) concerns about scaling. Specifically,
the direction in which variables are scored can affect a correlational
index of profile similarity. Cohen suggests that one solution to this
problem is to reflect each variable about the scales’ midpoint, to

append these values onto the normally scored profiles, and to com-
pute the correlation between expanded profiles. The current models
are valid for this index. In two of the applications described above,
analyses were conducted on profiles as originally scored and on re-
flected profiles. In neither case did the analyses produce different
conclusions. Nevertheless, researchers should examine the possibility
that reflections lead to different conclusions. Second, researchers
should use Fisher’s transformation when working with profile sim-
ilarity correlations, particularly when conducting significance tests of
mean correlations. In the applications described above, the similar-
ity, normativeness, and distinctiveness correlations were computed,
and then transformed using Fisher’s r to z transformation.
Integration With and Differentiation From Other Approaches to

Profile Similarity
Psychologists have grappled with the challenges of profile similarity
since at least the 1940s (e.g., Cattell, 1949; du Mas, 1946). A full
articulation of all relevant issues is beyond the scope of this article,
but the current models can be placed in the context of two relatively
well-known perspectives.
Cronbach’s Decomposition of Personality Judgments
Cronbach (1955) examined components of profile similarity in per-

sonality judgments. His highly influential perspective concerns de-
signs in which a single judge rates multiple targets, who also provide
self-ratings. In his framework (Equation 3a, p. 192), a judge’s general
accuracy arises from four elements: Elevation, Differential Eleva-
tion, Stereotype Accuracy, and Differential Accuracy. Elevation is
the difference between a judge’s mean rating of targets (averaged
across all items and multiple targets) and the mean rating provided
by those multiple targets (averaged across all items and all targets).
Differential Elevation is the degree to which a judge accurately de-
tects differences among the targets’ average ratings, averaged across
items. Stereotype Accuracy is the accuracy with which a judge rates
‘‘the generalized other’’ (p. 179). That is, it is the similarity between
a judge’s profile of mean ratings (averaged across targets) and
the target’s profile of mean self-ratings (averaged across targets).
Finally, Differential Accuracy is a judge’s ‘‘ability to predict
1304 Furr
differences between [targets] on any item.’’ As originally defined,

Elevation, Differential Elevation, and Stereotype Accuracy are not
applicable at the level of a pair of profiles. That is, they are not
quantified for each target-judge pair.
With a significant reconceptualization, Cronbach’s (1955) Stereotype
Accuracy and Differential Accuracy can be integrated with the models
presented in the current article. Cronbach’s data structure can be re-
conceptualized as including multiple unique judge/target pairs. That is,
rather than conceptualizing the two sets of profiles as being provided by
a single judge rating multiple targets with each target providing a self-
rating, they could be seen as arising from multiple judges, each of whom
rates a single different target (e.g., Table 2). This reconceptualization
creates an association between Cronbach’s Stereotype Accuracy and
the Generalized Normative Agreement component in the models above
(i.e., Generalized Mean Similarity, more broadly). Specifically, Cron-
bach (Equation 2, p. 179) decomposes Stereotype Accuracy into vari-
ances and a correlation. If the data are reconceptualized as described
above, then this correlation is the Generalized Normative Agreement
correlation in the current models (Equation 9). Similarly, Cronbach’s
Differential Accuracy component is closely related to the ‘‘Distinctive
Agreement’’ component in the models above (i.e., Distinctive Similarity
more broadly). Cronbach (p. 192) describes two possible decomposi-
tions of Differential Accuracy. Again, if the data are reconceptualized
as described above, then one of these possible decompositions (labeled
DA2oj in Cronbach’s notation) includes a correlation that is equivalent to
the Distinctive Agreement correlation in the models above.4
Although the current framework can be integrated with Cronbach
(1955), it differs in at least two important ways. First, Cronbach’s
framework is articulated in two highly specific contexts: (a) the sub-
stantive context of personality judgment and (b) the methodological
context of a single judge providing ratings of multiple targets, with
each target providing self-ratings. This specificity served Cronbach’s
original goals, but it limits his framework as a general perspective on
profile similarity. That is, researchers must recast the conceptual is-
sues from personality judgment to their domain of study, and they
must reconceptualize (if possible) the data structure in a way that fits
their methods. In contrast, the current models are articulated in two
4. A detailed discussion of the integrations and differentiations is available online

at: http://www.wfu.wdu/ furrrm/integrations.doc
general ways: (a) in terms that are applicable to a wide range of sub-
stantive contexts (see Table 4) and (b) with regard to a data structure
very common in profile similarity research. Specifically, many, if not
most, examinations of profile similarity involve the pairing of each pro-
file with a single other profile (e.g., when each husband is paired with his
wife, when each individual’s behavior in one situation is paired with his
or her behavior in a second situation, or when each individual’s person-
ality profile at one age is paired with his or her profile at another age).
More importantly, the second departure from Cronbach’s frame-
work is that the current models provide more information. That is,
some of the components in the current models are not reflected in
Cronbach’s framework, even when it is reconceptualized to fit the
relevant data structure Specifically, the Within-Profile Normative-
ness, the Cross-Profile Normativeness components, and the Global
Normativeness components are unique to the current models. As il-
lustrated earlier, these components are likely to be highly meaningful
in many examinations of profile similarity.
Kenny’s Decomposition of Dyadic Data
Along with Cronbach (1955), Kenny and his colleagues (Kenny,

Kashy, & Cook, 2006; Kenny & Winquist, 2001) provide an example
of the decomposition of profile similarity into components reflecting
some form of distinctiveness or normativeness. In a discussion of the
measurement of interpersonal sensitivity, Kenny and Winquist de-
scribe the profile analysis applicable when there is one perceiver per
target and one target per perceiver (i.e., a ‘‘1P1T’’ design). This cor-
responds exactly with the common design for many analysis of pro-
file similarity (e.g., one husband per wife and one wife per husband),
which is the basis of the current models. In their discussion of a
profile analysis of these data (pp. 275–277), Kenny and Winquist
describe a stereotype accuracy correlation analogous to the Gener-
alized Normative Agreement in the current models (see the discus-
sion of Cronbach’s stereotype accuracy, above). In addition, they
‘‘remove the effects of stereotype accuracy’’ (p. 276) by computing a
correlation that parallels the Distinctive Agreement correlation in
the current models. Although the models described in the current
article share some overlap with Kenny and Winquist’s (2001) anal-
ysis, they differ in an important way. Specifically, the current models go
beyond the Kenny and Winquist approach by articulating a wider set of
1306 Furr
components, by providing multiple ways of decomposing similarity, by

placing the components within coherent, fully integrated models, and
by extending the logic to multiple domains of psychological research.
Summary
The analysis of profile similarity is an intuitively appealing procedure
used in many areas of psychology. Researchers have long recognized
that, despite its intuitive appeal, profile similarity has conceptual and
statistical problems. Despite this recognition, there has been little con-
sensus regarding a conceptually and psychometrically coherent way of
handling such problems. Of specific interest for the current article is the
normativeness problem and its effects on profile similarity correlations.
The framework presented in the current article decomposes the ‘‘shape’’
facet of profile similarity into components reflecting blends of similar-
ity, normativeness, and distinctiveness. The primary purpose of the
current articlewas to detail the conceptual and psychometric founda-
tions of the framework and to place it in the context of profile similarity
more generally. In addition, the article presented four applications of
the framework, demonstrating its utility and importance.
Acknowledgments
I thank Will Fleeson, David Funder, Dan Ozer, and Dustin Wood for
their very helpful comments on earlier drafts of this paper. This research
was partially supported by a grant from NIMH (ROI-MH-63908).
REFERENCES
Asendorpf, J. B., & Van Aken, M. A. (1991). Correlates of the temporal consis-
tency of personality patterns in childhood. Journal of Personality, 59, 689–703.
Baker, B. O., & Block, J. (1957). Accuracy of interpersonal predictions as a func-
tion of judge and object characteristics. Journal of Abnormal and Social Psy-
chology, 54, 37–43.
Bernieri, F. J., Zuckerman, M., Koestner, R., & Rosenthal, R. (1994). Measuring
person perception accuracy: Another look at self-other agreement. Personality
and Social Psychology Bulletin, 20, 367–378.
Biesanz, J. C., & West, S. G. (2000). Personality coherence: Moderating self-other
profile agreement and profile consensus. Journal of Personality and Social Psy-
chology, 79, 425–437.
Biesanz, J. C., West, S. G., & Kwok, O.-M. (2003). Personality over time: Meth-
odological approaches to the study of short-terms and long-terms development
and change. Journal of Personality, 71, 905–941.
Biesanz, J. C., West, S. G., & Millevoi, A. (2007). What do you learn about
someone over time? The relationship between length of acquaintance and con-
sensus and self-other agreement in judgments of personality. Journal of Per-
sonality and Social Psychology, 92, 119–135.
Blackman, M. C., & Funder, D. C. (1998). The effect of information on consensus
and accuracy in personality judgment. Journal of Experimental Social Psy-
chology, 34, 164–181.
Block, J. (1971). Lives through time. Berkeley, CA: Bancroft Books.
Carlson, E., & Furr, R. M. (2007). Evaluating a trait profile approach to personality
pathology. Poster presented at the Association for Research in Personality pre-
conference of the 8th Annual Meeting of the Society for Personality and Social
Psychology, Memphis, TN.
Caspi, A., & Herbener, E. S. (1990). Continuity and change: Assortative marriage
and the consistency of personality in adulthood. Journal of Personality and
Social Psychology, 58, 250–258.
Cattell, R. B. (1949). rp and other coefficients of pattern similarity. Psychometrika,
14, 279–298.
Champagne, B. M., & Pervin, L. A. (1987). The relation of perceived situation
similarity to perceived behavior similarity: Implications for social learning
theory. European Journal of Personality, 1, 79–91.
Chaplin, W. F., & Panter, A. T. (1993). Shared meaning and the convergence among
observers’ personality descriptions. Journal of Personality, 61, 553–585.
Cheek, J. M. (1982). Aggregation, moderator variables, and the validity of per-
sonality tests: A peer-rating study. Journal of Personality and Social Psychol-
ogy, 43, 1254–1269.
Cohen, J. (1969). rc: A profile similarity coefficient invariant over variable reflec-
tion. Psychological Bulletin, 71, 281–284.
Colvin, C. R. (1993a). ‘‘Judgable’’ people: Personality, behavior, and competing
explanations. Psychological Bulletin, 52, 177–193.
Colvin, C. R. (1993b). Childhood antecedents of young-adult judgability. Journal
of Personality, 61, 611–635.
Cronbach, L. J. (1955). Processes affecting scores on ‘‘understanding of others’’
and ‘‘assumed similarity.’’ Psychological Bulletin, 52, 177–193.
Cronbach, L. J., & Gleser, G. C. (1953). Assessing similarity between profiles.
Psychological Bulletin, 50, 456–473.
du Mas, F. M. (1946). A quick method for analyzing the similarity of profiles.
Journal of Clinical Psychology, 2, 80–83.
Furr, R M. (2007). The intraclass correlation as an index of profile similarity.
Manuscript in preparation, Wake Forest University.
Furr, R. M., Dougherty, D. M., Marsh, D. M., & Mathias, C. W. (2007). Per-
sonality judgment and personality pathology: Self-other agreement in adoles-
cents with conduct disorder. Journal of Personality, 75, 629–662.
Furr, R. M., & Funder, D. C. (2004). Situational similarity and behavioral con-
sistency: Subjective, objective, variable-centered, and person-centered ap-
proaches. Journal of Research in Personality, 38, 421–447.
Furr, R. M., & Funder, D. C. (2005). Individual differences in multi-measurement
constructs. Unpublished manuscript, Wake Forest University.
1308 Furr
Gangestad, S. W., & Snyder, M. (2000). Self-monitoring: Appraisal and reap-

praisal. Psychological Bulletin, 126, 530–555.
Gaunt, R. (2006). Couple similarity and marital satisfaction: Are similar spouses
happier? Journal of Personality, 74, 1401–1420.
Gonzaga, G. C., Campos, B., & Bradbury, T. (2007). Similarity, convergence, and
relationship satisfaction in dating and married couples. Journal of Personality
and Social Psychology, 93, 34–48.
Kenny, D. A., Kashy, D. A., & Cook, W. L. (2006). Dyadic data analysis. New
York: Guilford Press.
Kenny, D. A., & Winquist, L. (2001). The measurement of interpersonal sensi-
tivity: Considerations of design, components, and unit of analysis. In J. Hall &
F. Bernieri (Eds) Interpersonal sensitivity: Theory, measurement, and applica-
tions (pp. 265–302). Englewood Cliffs, NJ: Erlbaum.
Krahé, B. (1990). Situation cognition and coherence in personality: An individual-
centred approach. Cambridge: Cambridge University Press.
Letzring, T. D., Wells, S. M., & Funder, D. C. (2006). Quantity and quality of
available information affect the realistic accuracy of personality judgment.
Journal of Personality and Social Psychology, 91, 111–123.
Luo, S., & Klohnen, E. C. (2005). Assortative mating and martial quality in
newlyweds: A couple-centered approach. Journal of Personality and Social
Psychology, 88, 304–326.
Lynam, D. R., & Widiger, T. (2001). Using the five factor model to represent the
personality disorders: An expert consensus approach. Journal of Abnormal
Psychology, 110, 401–412.
Magnusson, D., & Ekehammar, B. (1978). Similar situations—similar behaviors?
A study of the intraindividual congruence between situation perception and
situation reactions. Journal of Research in Personality, 12, 41–48.
McCrae, R. R. (1993). Agreement of personality profiles across observers. Multi-
variate Behavioral Research, 28, 2540.
McCrae, R. R., Yang, J., Costa, P. T., Dai, X., Yao, S., & Cai, T., et al. (2001).
Personality profiles and the prediction of categorical personality disorders.
Journal of Personality, 69, 155–174.
Miller, J. D., Lynam, D. R., Widiger, T. A., & Leukefeld, C. (2001). Personality dis-
orders as extreme variants of common personality dimensions: Can the five-factor
model adequately represent psychopathy? Journal of Personality, 69, 253–276.
Miller, J. D., Pilkonis, P. A., & Morse, J. Q. (2004). Five-factor model prototypes
for personality disorders: The utility of self-reports and observer ratings. As-
sessment, 11, 127–138.
Miller, J. D., Reynolds, S. K., & Pilkonis, P. A. (2004). The validity of the five-
factor model prototypes for personality disorders in two clinical samples. Psy-
chological Assessment, 16, 310–322.
Nunnally, J. C., & Bernstein, I. (1994). Psychometric theory (3rd ed.). New York:
McGraw Hill.
Ozer, D. J. (1986). Consistency in personality: A methodological framework. Berlin,
Germany: Springer.
Ozer, D. J., & Gjerde, P. F. (1989). Patterns of personality consistency and change
from childhood through adolescence. Journal of Personality, 57, 483–507.
Snyder, M., & Ickes, W. (1985). Personality and social behavior. In G. Lindzey &
E. Aronson (Eds.), Handbook of social psychology (Vol. 2, pp. 883–948). Read-
ing, MA: Addison-Wesley.
Snyder, M., & Monson, T. C. (1975). Persons, situations, and the control of social
behavior. Journal of Personality and Social Psychology, 32, 637–644.
Starzyk, K. B., Holden, R. R., Fabrigar, L. R., & MacDonald, T. K. (2006). The
Personal Acquaintance Measure: A tool for appraising one’s acquaintance
with any person. Journal of Personality and Social Psychology, 90, 833–847.
Westen, D., & Bradley, R. G. (2005). Prototype diagnosis of personality. In S.
Strack (Ed.), Handbook of personology and psychopathology (pp. 238–256).
New York: Wiley.
Westen, D., Shedler, J., & Bradley, R. (2006). A prototype approach to person-
ality diagnosis. American Journal of Psychiatry, 163, 846–856.
Widiger, T. A., Trull, T. J., Clarkin, J. F., Sanderson, C., & Costa, P. T. (2002). A
description of the DSM-IV personality disorders with the five factor model of
personality. In P. T. Costa & T. A. Widiger (Eds), Personality Disorders and
the Five Factor Model of Personality (2nd ed., pp. 89–99). Washington, DC:
American Psychological Association.
Wymer, W. E., & Penner, L. A. (1985). Moderator variables and different types of
predictability: Do you have a match? Journal of Personality and Social Psy-
chology, 49, 1002–1015.
Appendix A
Details of the Initial Decomposition
This Appendix presents algebraic details of the components and
accuracy of the initial decomposition (Equation 3).
Defining the components of the initial decomposition

This section presents the definitional and computational formulas
for the covariance components of the initial decomposition:
X
1 nv
Overall Similarity sxiv1 xiv2 ¼ ðxiv1 xi1 Þðxiv2 xi2 Þ
nv v¼1
X
1 nv
¼ xiv1 xiv2 xi1 xi2
nv v¼1
1310 Furr
Generalized Normative Similarity

X
1 nv
sxv1 xv2 ¼ ðxv1 x1 Þðxv2 x2 Þ
nv v¼1
X
1 nv
¼ xv1 xv2 x1 x2
nv v¼1
The remaining covariances include the ‘‘distinctiveness’’ scores,

and they rely on the mean of an individual’s profile of distinctiveness
scores. An individual’s distinctiveness on a variable ðx0ivp Þ and his or
her mean distinctiveness across variables (x0ip ) are:
x0ivp ¼ xivp xvp
1X nv
1X nv
1X nv
x0ip ¼ xvp ¼
xivp xivp xvp ¼ xip xp
nv v¼1 nv v¼1 nv v¼1
Distinctive Similarity
X
1 nv 0
sx0iv1 x0iv2 ¼ xiv1 x0i1 x0iv2 x0i2
nv v¼1
X
1 nv
sx0iv1 x0iv2 ¼ ðxiv1 xv1 xi1 þ x1 Þðxiv2 xv2 xi2 þ x2 Þ
nv v¼1
" # " #
1 X nv
1 X nv
sx0iv1 x0iv2 ¼ xiv1 xiv2 xi1 xi2 xiv1 xv2 xi1 x2
nv v¼1 nv v¼1
" # " #
1 X nv
1 X nv
xiv2 xv1 xi2 x1 þ xv1 xv2 x1 x2
nv v¼1 nv v¼1
Mean/Distinctive Similarity 1
X
1 nv 0
sxiv1 xv2 ¼
0 xiv1 x0i1 ðxv2 x2 Þsx0iv1 xv2
nv v¼1
X
1 nv
¼ ðxiv1 xv1 xi1 þ x1 Þðxv2 x2 Þ
nv v¼1
" # " #
1 X nv
1 X nv
sx0iv1 xv2 ¼ xiv1 xv2 xi1 x2 xv1 xv2 x1 x2
nv v¼1 nv v¼1
Mean/Distinctive Similarity 2
X
1 nv 0
sx0iv2 xv1 ¼ xiv2 x0i2 ðxv1 x1 Þ
nv v¼1
X
1 nv
sx0iv2 xv1 ¼ ðxiv2 xv2 xi2 þ x2 Þðxv1 x1 Þ
nv v¼1
" # " #
1 X nv
1 X nv
sx0iv2 xv1 ¼ xiv2 xv1 xi2 x1 xv1 xv2 x1 x2
nv v¼1 nv v¼1
Algebraic Proof of the Initial Decomposition
The following demonstrates the accuracy of the initial decomposi-

tion (Equation 3 in the text):
sxiv1 xiv2 ¼ sx0iv1 x0iv2 þ sxv1 xv2 þ sxv1 x0iv2 þ sxv2 x0iv1
Generalized Mean Mean

Overall Distinctive
¼ þ Normative þ Distinctive þ Distinctive
Similarity Similarity
Similarity Similarity 1 Similarity 2
Overall similarity
8 n n 9
>
> 1
Pv 1
Pv >
>
>
< nv xiv1 xiv2 xi1 xi2 nv xiv1 xv2 xi1 x2 >
=
v¼1 v¼1
¼ n n
>
> Pv Pv >
>
: n1v xiv2 xv1 xi2 x1 þ n1v xv1 xv2 x1 x2 >>
;
v¼1 v¼1
( )
1 X nv
þ xv1 xv2 x1 x2
nv v¼1
(" # " #)
1 X nv
1 X nv
þ xiv1 xv2 xi1 x2 xv1 xv2 x1 x2
nv v¼1 nv v¼1
(" # " #)
1 X nv
1 X nv
þ xiv2 xv1 xi2 x1 xv1 xv2 x1 x2
nv v¼1 nv v¼1
Terms cancel each other out, reducing to:

X
1 nv
Overall similarity ¼ xiv1 xiv2 xi1 xi2
nv v¼1
1312 Furr
Appendix B
Decomposition of Model 1
This Appendix presents algebraic details of the accuracy of Model 1
(Equations 4 and 7).
Defining the Components of Model 1

for the covariance components of Model 1. The Overall Similarity,
Generalized Normative Similarity, and Distinctive Similarity com-
ponents are as defined in Appendix A. Model 1 includes two Within-
Profile Normativeness components and a Global Normativeness
component:
Within-Profile Normativeness 1
X
1 nv
sxiv1 xv1 ¼ ðxiv1 xi1 Þðxv1 x1 Þ
nv v¼1
X
1 nv
¼ xiv1 xv1 xi1 x1
nv v¼1
Within-Profile Normativeness 2
X
1 nv
nv v¼1
X
1 nv
¼ xiv2 xv2 xi2 x2
nv v¼1
The Global Normativeness component reflects the similarity be-
tween the Pair Mean profile and the Global Mean Profile. Recall
(Eq. 5 in the text) that a score on a variable in the Pair Mean profile
is xiv , making the mean of the profile xi . In addition, a score on a
variable in the Global Mean profile is xv , making the mean of this
profile x .
Global Normativeness
X
1 nv
sxiv xv ¼ ðxiv xi Þðxv x Þ
nv v¼1
X
1 nv
¼ xiv xv xi x
nv v¼1
The elements in this Equation can be broken down further,

which will allow convergence with the other components in
Model 1:
xiv1 þ xiv2 xv1 þ xv2 xi1 þ xi2
xiv ¼ xv ¼ xi ¼
2 2 2
x1 þ x2
x ¼
2
Replacing terms,
Global Normativeness
nv
1X xiv1 þ xiv2 xv1 þ xv2 xi1 þ xi2 x1 þ x2
¼
nv v¼1 2 2 2 2
! !
1 1X nv
1 1X nv
¼ xiv1 xv1 xi1 x1 þ xiv2 xv2 xi2 x2
4 nv v¼1 4 nv v¼1
! !
1 1X nv
1 1X nv
þ xiv1 xv2 xi1 x2 þ xiv2 xv1 xi2 x1
4 nv v¼1 4 nv v¼1
Note that the elements comprising Global Normativeness are

exactly the two ‘‘Within-Profile Normativeness’’ scores, defined
above, and the two ‘‘Cross-Profile Normativeness’’ scores defined
in Model 1.
Algebraic Proof of Model 1
The following demonstrates the accuracy of Model 1 (Equation 4 in

the text)
sxiv1 xiv2 ¼ sx0iv1 x0iv2 sxv1 xv2 sxv1 xiv1 sxv2 xiv2 þ 4sxiv xv
Generalized Within- Within-

Overall Distinctive 4 Global
¼ Normative Profile Profile þ
Similarity Similarity Normativeness
Similarity Normativeness 1 Normativeness 2
Over similarity
8 n n 9
>
> 1
Pv 1
Pv >
>
>
< nv x x
iv1 iv2
x x
i1 i2 nv x
x
iv1 v2
x x
i1 2 >
=
v¼1 v¼1
¼ n n
>
> Pv Pv >
> 1
: nv xiv2 xv1 xi2 x1 þ n1v xv1 xv2 x1 x2 >>
;
v¼1 v¼1
( )
1 X nv
xv1 xv2 x1 x2
nv v¼1
1314 Furr
( ) ( )
1 X nv
1 X nv
xiv1 xv1 xi1 x1 xiv2 xv2 xi2 x2
nv v¼1 nv v¼1
8 nv n 9
> P Pv
> 1
< 4 nv
1 1
xiv1 xv1 xi1 x1 þ 4 nv 1
xiv2 xv2 xi2 x2 > >
=
þ4 v¼1 v¼1
>
> Pv
n Pv
n
>
: þ 14 n1v xiv1 xv2 xi1 x2 þ 14 n1v xiv2 xv1 xi2 x1 >
;
v¼1 v¼1
And once again, terms cancel each other out, reducing to:
X
1 nv
nv v¼1
Finally, Equation 4 can be transformed to Equation 7,

because each covariance is the product of a corresponding
correlation and the standard deviations of the two variables, pro-
ducing.
rxiv1 xiv2 ¼
sx0iv1 sx0iv2 rx0iv1 x0iv2 sxv1 sxv2 rxv1 xv2 sxv1 sxiv1 rxv1 xiv1 sxv2 sxiv2 rxv2 xiv2 þ 4sxiv sxv rxiv xv
sxiv1 sxiv2
Appendix C
Details of Model 2
This Appendix presents algebraic details of the accuracy of Model 2
(Equations 8 and 9). To enhance clarity, this proof focuses primarily
on the Model as defined on the covariance metric.
Defining the Components of Model 2

for the covariance components of Model 2. Again, Overall Similar-
ity, Generalized Normative Similarity, and Distinctive Similarity
are as defined in Appendix A. Model 2 includes two Cross-Profile
Normativeness components:
Cross-Profile Normativeness 1
X
1 nv
nv v¼1
X
1 nv
¼ xiv1 xv2 xi1 x2
nv v¼1
Cross-Profile Normativeness 2
X
1 nv
nv v¼1
X
1 nv
¼ xiv2 xv1 xi2 x1
nv v¼1
Algebraic Proof of Model 2
The following demonstrates the accuracy of Model 2 (Equation 8 in

the text)
sxiv1 xiv2 ¼ sx0iv1 x0iv2 sxv1 xv2 þ sxv1 xiv2 þ sxv2 xiv1
Generalized
Overall Distinctive Cross-Profile Cross-Profile
¼ Normative þ þ
Similarity Similarity Normativeness 1 Normativeness 2
Similarity
Overall similarity
8 n n 9
>
> 1
Pv 1
Pv >
>
>
< nv xiv1 xiv2 xi1 xi2 nv xiv1 xv2 xi1 x2 >
=
v¼1 v¼1
¼ n n
>
> Pv Pv >
>
: n1v xiv2 xv1 xi2 x1 þ n1v xv1 xv2 x1 x2 >>
;
v¼1 v¼1
( )
1 X nv
xv1 xv2 x1 x2
nv v¼1
( ) ( )
1 X nv
1 X nv
þ xiv1 xv2 xi1 x2 þ xiv2 xv1 xi2 x1
nv v¼1 nv v¼1
Again, terms cancel each other out, reducing to:

X
1 nv
nv v¼1
1316 Furr
Equation 8 can be transformed to Equation 9, because each co-

variance is the product of a corresponding correlation and the stan-
dard deviations of the two variables, producing:
rxiv1 xiv2 ¼
sx0iv1 sx0iv2 rx0iv1 x0iv2 sxv1 sxv2 rxv1 xv2 þ sxv1 sxiv2 rxv1 xiv2 þ sxv2 sxiv1 rxv2 xiv1
sxiv1 sxiv2

Furr-2008-Journal of Personality

Uploaded by

Copyright:

Available Formats

Furr-2008-Journal of Personality

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Furr-2008-Journal of Personality

Uploaded by

Copyright:

Available Formats

A Framework for Profile Similarity: Integrating

Similarity, Normativeness, and Distinctiveness

ABSTRACT Many questions in personality psychology lend them-

The analysis of proﬁles of psychological variables has an intuitive

Journal of Personality 76:5, October 2008

Self and Informant Profiles

Figure 1 illustrates two psychological proﬁles in the context of

few, if any, helpfully coherent and ﬂexible frameworks for handling

The Normativeness Problem

Proﬁle normativeness creates conceptual and statistical challenges in

(a) Targets' Original Self–Rated Profiles

(b) Targets' Original Self–Rated Profiles

(c) Targets' Self–Rated Distinctive Profiles

was highly correlated with independent ratings of trait social desir-

the general level of personality similarity within married couples.

Two Possible Analytic Strategies

Sample-Level Strategy (Random Dyads)

In a sample-level approach to normativeness, the average similarity

participants interacted with each other in small groups, provided

created ‘‘random couples’’ by pairing a proﬁle from a husband in

In a pair-level strategy, the similarity between each pair of proﬁles is

The framework presented in the next section is a pair-level strat-

Conceptual Basis of the Framework

Decomposition of Individual Proﬁles

An individual’s proﬁle can be seen as arising from two proﬁles—a

In the current framework, Overall Similarity is the correlation be-

Distinctive Similarity reﬂects the similarity between the unique

distinctive proﬁle matches another distinctive proﬁle. In self-other

Generalized Normative Similarity

Generalized Normative Similarity reﬂects the similarity between two

pair’s Overall Similarity) is partially explained by the similarity be-

The third type of component reﬂects the normativeness of each pro-

degree to which a given target (as seen by self and informant)

For example, in the analysis of self-other agreement of personality

Models and Proﬁle Similarity Components

Decomposition of Individual Proﬁles

As described earlier, an individual’s proﬁle of scores can be seen as

xiv2 ¼ xv2 þ x0 iv2 Eq:2b

Initial Decomposition of Proﬁle Similarity

There are several ways to quantify proﬁle similarity (e.g., Cattell

Target 5 Kyle Distinctive Distinctive Kyle’s

Target 5 Stan Distinctive Distinctive Stan’s

Target 5 Eric Distinctive Distinctive Eric’s

Norms Normative Normative Global

target’s self-ratings indicating high extraversion, moderate agree-

The sxiv1 xiv2 component is Overall Similarity (i.e., the covariance

Proﬁle Similarity Model 1

As described earlier, Model 1 includes Distinctive Similarity, Gen-

normativeness (i.e., two Within-proﬁle normativeness components

Table 1 summarizes Model 1 at the conceptual level, on a covari-

Proﬁle Similarity Model 2

As described earlier, Model 2 includes Distinctive Similarity, Gen-

Models/Proﬁle Similarity Components

Both Models Model 1 Model 2

Overall Distinctive Gen. Norm. Self Other Global Other-self Self-Other

On a correlational metric, Model 2 is:

and Open, and Kyle’s informant describes him in a similar way.

Applications of the Framework

There are at least two studies illustrating the useful application

xiv2 ¼ xv2 þ x0 iv2 Eq:2b