Journal of Personality Assessment

ISSN: 0022-3891 (Print) 1532-7752 (Online) Journal homepage: https://www.tandfonline.com/loi/hjpa20

Measuring Identity From an Eriksonian

Perspective: Two Sides of the Same Coin?

Seth J. Schwartz , Byron L. Zamboanga , Wei Wang & Janine V. Olthuis

To cite this article: Seth J. Schwartz , Byron L. Zamboanga , Wei Wang & Janine V. Olthuis (2009)
Measuring Identity From an Eriksonian Perspective: Two Sides of the Same Coin?, Journal of
Personality Assessment, 91:2, 143-154, DOI: 10.1080/00223890802634266

To link to this article: https://doi.org/10.1080/00223890802634266

Published online: 10 Feb 2009.

Submit your article to this journal

Article views: 901

View related articles

Citing articles: 61 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=hjpa20
Journal of Personality Assessment, 91(2), 143–154, 2009
Copyright
C Taylor & Francis Group, LLC
ISSN: 0022-3891 print / 1532-7752 online
DOI: 10.1080/00223890802634266

Measuring Identity From an Eriksonian Perspective: Two Sides

of the Same Coin?
SETH J. SCHWARTZ,1 BYRON L. ZAMBOANGA,2 WEI WANG,3 AND JANINE V. OLTHUIS4

1
Center for Family Studies, Department of Epidemiology and Public Health, Leonard M. Miller School of Medicine, University of Miami
2
Department of Psychology, Smith College
3
Department of Epidemiology and Biostatistics, University of South Florida
4
Department of Psychology, Dalhousie University, Halifax, Canada

In this article, we report the results of 3 studies evaluating the psychometric properties of scores generated using the Erikson Psychosocial Stage
Inventory (EPSI; Rosenthal, Gurney, & Moore, 1981) with emerging adults. In Study 1, a hybrid bifactor solution, consisting of an overall identity
factor as well as of “method effects” factors for identity synthesis and identity confusion, provided a better fit to the data than did either one or
two-factor solutions. This bifactor solution was largely invariant across gender and across Whites, Blacks, and Hispanics. In Study 2, the overall
identity, identity synthesis, and identity confusion scores were shown to possess convergent validity with another Eriksonian measure and with
measures of identity status. In Study 3, the EPSI subscale scores were shown to possess construct validity vis-à-vis self-esteem, purpose in life,
internal locus of control, ego strength, anxiety, and depression. We discuss implications for the measurement of identity.

Identity has been the subject of theoretical and empirical anal- ized by identity synthesis would be expected to have explored
ysis ever since Erikson’s (1950) seminal book Childhood and and committed to identity alternatives and would be placed
Society. As conceptualized by Erikson (1950), personal1 identity into the identity achieved status. Conversely, someone char-
represents the amount of self-knowledge, synthesis, and consis- acterized by identity confusion would be expected to be un-
tency that a person possesses over time and across situations. committed and to have undergone little, if any, systematic
For example, in Western cultural contexts, it is important to “be identity exploration. Such a person would be placed into the
the same person” at home, at work or school, and with friends identity diffused status. Marcia (1966) also included two inter-
(Cross, Gore, & Morris, 2003; Suh, 2002). The extent to which mediate statuses—foreclosure (commitments without prior ex-
the various aspects of one’s identity fit together—known as iden- ploration) and moratorium (active identity exploration without
tity synthesis (Erikson, 1968)—is assumed to strongly predict commitments).
psychosocial functioning in areas such as self-worth, depres- The vast majority of neo-Eriksonian identity research has re-
sion, and relationships. On the other hand, identity confusion lied on the identity status model (Schwartz, 2001, 2005), with
represents a sense of feeling “mixed up,” that is, being unable to comparatively few studies having attempted to operationalize
enact and maintain lasting commitments to life alternatives and Erikson’s (1950, 1968) identity dimension directly. This may
lacking a clear sense of purpose and direction. Erikson (1950, be because Erikson discussed identity in figurative and clinical
1968) had proceeded from the assumption that identity synthe- terms, used a number of case examples, and spoke in highly
sis and identity confusion represent opposing ends of a single abstract language. As a result, it may be difficult to extract clear
continuum. In other words, increases in identity synthesis are as- operationalizations from his work (Côté, 1993). As noted ear-
sumed to be accompanied by equivalent and opposing decreases lier, Erikson’s theory and the identity status model overlap in
in identity confusion and vice versa. terms of diffusion and achievement (Côté & Schwartz, 2002),
A number of empirical operationalizations of Erikson’s work whereas the moratorium and foreclosed statuses do not map
have been introduced, the most enduring of which has been the neatly onto Erikson’s synthesis-confusion dimension. Although
identity status model (Marcia, 1966). Marcia (1966) extracted moratorium has been conceptualized as the “route” to achieve-
from Erikson’s (1950, 1968) work the dimensions of explo- ment, it has been found to be more similar to diffusion in terms
ration and commitment. Exploration refers to sorting through of identity confusion and psychological distress (Kidwell, Dun-
potential identity alternatives, and commitment refers to de- ham, Bacho, Pastorino, & Portes, 1995). Foreclosure appears to
ciding on and adhering to a specific set of goals, values, and index something entirely separate from Erikson’s model, as it
beliefs. From Marcia’s (1966) viewpoint, someone character- is not clear whether commitments enacted without exploration
represent identity synthesis, identity confusion, or neither. As
a result, it may be important to operationalize and measure
identity directly from an Eriksonian perspective as well as us-
Received August 28, 2007; Revised July 22, 2008. ing measures of identity status. Such work would allow for
Address correspondence to Seth J. Schwartz, Center for Family Studies,
Department of Psychiatry and Behavioral Sciences, Leonard M. Miller School an empirical comparison of Eriksonian and identity status ap-
of Medicine, University of Miami, 1425 NW 10th Avenue, 2nd Floor, Miami, proaches to identity as well as for attempts to extract higher
FL 33136; Email: [email protected] order identity consolidation and confusion constructs that draw
1 Although multiple dimensions of identity have been discussed, for purposes on both Erikson’s and Marcia’s (1966) work (e.g., Schwartz,
of clarity and brevity, the term identity is used here to refer to personal identity. 2007).
143
144 SCHWARTZ, ZAMBOANGA, WANG, OLTHUIS

A number of years after introducing the identity status model, Erikson’s (1950, 1968) identity stage, a two-factor solution for
Marcia (2002) began to explore the possibility of operationaliz- the EPSI (with Synthesis as the first factor and Confusion as the
ing some of Erikson’s (1950, 1968) other stages (e.g., intimacy, second factor) should provide a significantly better fit to the data
generativity, integrity) for empirical research. As part of this than a one-factor solution with Synthesis and Confusion cast as
effort, Marcia (2002) recast Erikson’s identity stage as identity polar opposites. Moreover, if identity synthesis and confusion
synthesis with identity confusion (whereas Erikson had used the are largely opposing but still possess some unshared variability
term versus). This recasting carried the assumption that some vis-à-vis one another, a hybrid solution—representing a mixture
degree of both synthesis and confusion would be required for of one- and two-factor models (cf. Quilty, Oakman, & Risko,
healthy identity development. Although a high degree of coher- 2006)—may fit the data better than either the one- or two-factor
ence in one’s sense of self is clearly adaptive, the possibility of models. Further, to the extent to which Erikson’s theory and
continuing identity development is contingent on some “empty Marcia’s (1966) identity status theory converge (see Waterman,
space” within one’s self-knowledge—that is, some space that re- 1999, for a further discussion), significant correlations should
mains to be filled in (cf. Luyckx, Goossens, & Soenens, 2006). emerge between scores on the EPSI and on the identity status
The assumption that identity synthesis and confusion can co- subscales. Provided that they are strong, these correlations may
exist to some extent implies that they are not perfect opposites be indicative of higher order identity consolidation and confu-
and that a one-dimensional understanding of identity may not sion constructs that may subsume both Eriksonian and identity
be accurate. A two-dimensional model of identity, with synthe- status indexes (cf. Schwartz, 2007). Measuring such constructs
sis and confusion as separate factors, requires a certain degree requires a reasonably valid and reliable factor structure for the
of independence between these two dimensions. It is not en- EPSI.
tirely clear what the factor structure of identity would look like Another important issue in the measurement of identity con-
from an Eriksonian perspective—and as a result, this remains structs is the appropriateness of identity measures for diverse
an empirical question. populations. Until recently, the vast majority of identity re-
The structure of identity is also important with regard to its re- search had been conducted with largely White samples (Sneed,
lationship to indexes of psychosocial functioning. The value of Schwartz, & Cross, 2006). However, as the population of the
identity as a social-psychological construct lies in its relevance United States, and of other Western countries, becomes more
for healthy and maladaptive functioning (Schwartz, 2005). That ethnically and culturally diverse, it has become imperative for
is, a coherent sense of identity is valuable only inasmuch as identity research to include participants from various ethnic
it is associated with adaptive mental health outcomes such as groups. Although some identity status studies have been con-
self-esteem, resilience, and life purpose. Similarly, identity con- ducted with ethnically diverse samples (Hall & Brassard, 2008;
fusion is to be avoided only inasmuch as it is associated with Streitmatter, 1988), this has not yet been done with Eriksonian
maladaptive mental health outcomes such as anxiety and de- identity measures.
pression. Relevance to mental health and behavioral outcomes Provided that EPSI scores correlate with exploration, com-
was one of the primary emphases within Erikson’s (1950, 1968) mitment, and identity status in theoretically consistent ways and
writings, and it must be the benchmark by which identity is provided that these correlations are consistent across gender and
evaluated empirically. ethnicity, the EPSI can be considered to possess adequate crite-
rion validity. Moreover, to the extent to which scores on the EPSI
MEASUREMENT OF IDENTITY correlate with measures of psychosocial functioning (e.g., self-
Researchers have developed a number of measurement instru- esteem, anxiety, depression) in theoretically consistent ways
ments to index personal identity development. Perhaps the most across gender and ethnicity, the measure can be assumed to
popular of these are the Extended Objective Measure of Ego possess adequate construct validity.
Identity Status (EOM–EIS–II; Bennion & Adams, 1986) and the In this article, we report the results of three studies using com-
Ego Identity Process Questionnaire (EIPQ; Balistreri, Busch- pletely separate and independent samples. In all three studies,
Rossnagel, & Geisinger, 1995), both of which are grounded we used ethnically diverse samples—which, as stated earlier,
in Marcia’s (1966) identity status model. Eriksonian measures represents an important need (and advance) in identity research.
have been much less widely used; perhaps the most commonly In Study 1, we examined the factor structure of the EPSI, com-
used is the Erikson Psychosocial Stage Inventory (EPSI; Rosen- paring a one-factor solution, a two-factor solution, and a hybrid
thal, Gurney, & Moore, 1981). The EPSI contains subscales “bifactor” solution that posits one primary factor with two lower
assessing several of Erikson’s (1950, 1968) stages including order “method” factors included to account for the presence of
identity. The Identity subscale was designed to yield a single positively and negatively worded items. Given that identity syn-
set of scores, with the assumption that these scores would re- thesis and identity confusion are posited as largely opposing,
flect identity synthesis and the absence of identity confusion. with some degree of independence between them, the bifactor
However, whereas in-depth psychometric evaluations have been solution was hypothesized to provide the best fit to the data.
conducted on the EIPQ (Schwartz & Montgomery, 2002) and The bifactor solution provides both an overall identity score and
the EOM–EIS–II (Schwartz, Adamson, Ferrer-Wreder, Dillon, separate scores for identity synthesis and identity confusion (cf.
& Berman, 2006) to support the hypothesized factor structures Reise, Morizot, & Hays, 2007). A similar procedure has been
of scores generated by these instruments, this has not yet been used to address method effects resulting from negatively worded
examined with the EPSI. items in measures of self-esteem (Quilty et al., 2006). There is
Although the structure of scores from the EPSI may appear to evidence that a bifactor solution provides a better fit to the data
be solely a methodological issue, it carries important theoretical than does a second-order solution where the larger second-order
implications as well. To the extent to which “identity synthesis factor relates to the items only through the first-order method
with identity confusion” is the most accurate interpretation of factors (Chen, West, & Sousa, 2006).
FACTOR STRUCTURE OF IDENTITY 145

In Study 2, we examined the convergent validity of the EPSI Measure: EPSI Identity Subscale
subscales in relation to two other Eriksonian measures and to Adolescent identity was measured using the 12-item identity
measures of exploration, commitment, and identity status. We subscale from the EPSI (Rosenthal et al., 1981), which mea-
hypothesized that the EPSI would converge well with other sures the extent to which participants have a clear sense of who
Eriksonian measures, that identity synthesis would relate most they are and what they believe in. Six items are worded in a
closely to the achieved status, and that identity confusion would “positive” direction (toward identity synthesis), and 6 items are
relate most closely to the diffused status. Provided that the bifac- worded in a “negative” direction (toward identity confusion).
tor solution were to be retained, we would hypothesize that the Possible scores on the EPSI identity scale (heretofore refer-
second-order “total identity” factor would represent Erikson’s enced simply as the EPSI) range from 12 to 60. Sample items
(1950, 1968) overall identity dimension and would relate posi- from this measure include “I know what kind of person I am”
tively to the achieved status and negatively to the diffused status. (identity synthesis) and “I feel mixed up” (identity confusion).
In Study 3, we examined the relationships of the empirically The response scale used for the EPSI ranges from 1 (Strongly
derived EPSI subscales (extracted in Study 1) to measures of Disagree) to 5 (Strongly Agree).
adaptive and maladaptive psychosocial functioning. For Study The EPSI has been used successfully with middle school
3, we selected measures that would appropriately index adaptive (Schwartz, Pantin, Prado, Sullivan, & Szapocznik, 2005), high
(self-esteem, purpose in life, internal locus of control, and ego school (Reis & Youniss, 2004), and college (Markstrom &
strength) and maladaptive (anxiety and depression) psychoso- Kalmanir, 2001) students. As reported by Rosenthal et al.
cial functioning in emerging adulthood. Well-being and distress (1981), scores on the identity scale were positively related to
represent important positive and negative aspects of adjustment the identity subscale from the Personal Maturity Inventory (r =
(Keyes, 2005). Emerging adulthood, which bridges adolescence .56; Greenberger & Sorensen, 1974). Moreover, in a longitudi-
and full adulthood, is critical as a time in which individuals nal study following adolescents from their sophomore to senior
must “find their own way” in many Western cultural contexts year of high school, Reis and Youniss (2004) found that the most
(Arnett, 2000). Self-esteem, purpose in life, internal locus of progressive changes in EPSI scores (i.e., increases in identity
control, and ego strength represent both a sense of well-being coherence and decreases in identity confusion) across adoles-
and a sense of self-direction that is essential for “getting ahead” cence were associated with positive relationships (e.g., adaptive
in the United States and other Western societies (Côté, 2002). communication, high support, low avoidance, low conflict) with
Anxiety and depression, on the other hand, represent aspects parents and peers.
of maladaptive functioning that can undermine progress toward
gainful employment, committed partnership, parenthood, and
other adult responsibilities (Schulenberg & Zarrett, 2006). We Results
hypothesized that both the overall identity score and identity We first tested the one-factor, two-factor, and bifactor models
synthesis would relate positively to adaptive psychosocial func- against one another. In all three solutions, the metric for each
tioning and negatively to maladaptive psychosocial functioning. latent variable was set by constraining the variance of each
We hypothesized that the converse would be true of identity con- latent variable to 1. Additionally, in the bifactor model, the
fusion. We conducted all analyses using Mplus 5.0 (Muthén & correlations between the overall factor and the method factors
Muthén, 2007) using maximum likelihood estimation. are set to zero (Chen et al., 2006). This is done so that the
method effects remain separate from the total identity factor.
STUDY 1 We ascertained the fit of each individual model to the data
using the comparative fit index (CFI) and non-normed fit index
Method
(NNFI)—which compare the fit of the specified model to the fit
Sample and procedures. The sample for Study 1 consisted of a null model with no paths, covariances, or latent variables—
of 337 students (95 men, 241 women, 1 unidentified by gender; and the root mean square error of approximation (RMSEA) and
M age = 20.15 years, SD = 2.75) enrolled in a large, urban, standardized root mean square residual (SRMR), which specify
public university in Miami. In terms of ethnicity, 57 participants the extent to which the covariance structure specified in the
identified as non-Hispanic White, 82 as non-Hispanic Black, and model deviates from the covariance structure observed in the
198 as Hispanic. The majority of White (96%), Black (72%), data. The chi-square index, which tests the null hypothesis of
and Hispanic (70%) participants were born in the United States. perfect fit to the data, is often overpowered, and we therefore did
Foreign-born Blacks were primarily from Haiti and Jamaica, and not use it in interpretation (cf. Steiger, 2006). In general, CFI and
foreign-born Hispanics were primarily from Colombia, Cuba, NNFI values of .95 or higher, and RMSEA and SRMR values
Nicaragua, and Peru. Of participants reporting annual family of .06 or lower, represent excellent fit (Hancock & Freeman,
income (87% of the sample), 16% reported less than $30,000 2001), with .90 representing the lower bound for an acceptable
per year, 27% between $30,000 and $50,000, 29% between CFI or NNFI value and .08 representing the upper bound for an
$50,000 and $100,000, and 15% above $100,000. acceptable RMSEA or SRMR value (Kline, 2006; Quintana &
Data were collected in fall 2004 from introductory psychol- Maxwell, 1999).
ogy courses, and participants received course credit for taking Results indicated that the bifactor model, χ 2 (39) = 72.83,
part in the study. This data collection was part of a larger study p < .001; CFI = .97; NNFI = .95; RMSEA = .051 (90% CI =
(Schwartz, Zamboanga, Rodriguez, & Wang, 2007) on identity .032–.069); SRMR = .034; provided a better fit than either the
and cultural context. Participants completed the study measures two-factor model, χ 2 (51) = 107.79, p < .001; CFI = .93; NNFI
(which took approximately 1 hr to complete) at home and re- = .91; RMSEA = .058; SRMR = .049; or the one-factor model,
turned them to their instructor. The EPSI was the first measure χ 2 (53) = 132.02, p < .001; CFI = .90; NNFI = .87; RMSEA =
presented. .068; SRMR = .055. In all three models, the fit statistics reported
146 SCHWARTZ, ZAMBOANGA, WANG, OLTHUIS

Identity

.43*** .34** .46** .31* .00 .27* -.50*** -.77*** -.47*** -.21 -.25** -.41***

Item 2 Item 4 Item 5 Item 6 Item 8 Item 9 Item 1 Item 3 Item 7 Item 10 Item 11 Item 12

.34** .54*** .38** .61*** .56*** .57*** .32* .25 .34* .93*** .17* .33*

Synthesis Confusion

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

FIGURE 1.—Erikson Psychosocial Stage Inventory bifactor model.

here include two error covariances added based on modification suggests that the task of identity development in emerging adult-
indexes: Item 2 with Item 7 and Item 8 with Item 9. We retained hood consists both of creating a positive overall sense of self
these error covariances in Studies 2 and 3. and of balancing identity synthesis with identity confusion. This
To test whether the fit of the bifactor model was significantly suggests that developing a positive sense of self does not guar-
superior to the one- and two-factor models, we used the chi- antee a specific ratio between synthesis and confusion. Some
square difference () test (χ 2 ), the difference in CFI values individuals with a positive sense of identity may have more
(CFI), and the difference in NNFI values (NNFI). The null or less coherence and confusion than others. The emergence
hypothesis of equivalent model fit would be rejected if two of the of synthesis and confusion as separate method factors suggests
following three criteria were satisfied: χ 2 significant at p < that even for individuals who are quite sure of themselves and
.05 (Byrne, 2001), CFI ≥ .01 (Cheung & Rensvold, 2002), and of the direction they are taking in life, room still remains for
NNFI ≥ .02 (Vandenberg & Lance, 2000). Statistical tests in- “changing course” and for learning more about themselves. For
dicated that the bifactor model provided a significantly better fit example, taking a class can inspire a student to change her or his
to the data than either the two-factor model, χ 2 (12) = 34.96, area of academic study, and a relationship with someone from
p < .001; CFI = .041; NNFI = .042; or the one-factor another part of the world can change a person’s perspective on
model, χ 2 (13) = 59.19, p < .001; CFI = .070; NNFI = herself or himself.
.075. We therefore retained the bifactor model for subsequent From a measurement perspective, these results also suggest
analyses2 (see Figure 1). Cronbach’s alpha estimates for ob- that reasonably internally consistent subscales can be created
served scores were the following: total identity score = .83, for one’s overall sense of identity, for identity synthesis, and for
identity synthesis = .75, and identity confusion = .74. identity confusion. Using a total R 2 of .30, equivalent to a mul-
tiple R of .55 and corresponding to Cohen’s (1988) definition
Discussion of a medium effect size, the total-identity and method factors
These results suggest that a bifactor solution for the EPSI, explained at least 30% of variability in 9 of the 12 items. The
consisting of both an overall Identity factor and separate factors items not meeting this criterion were “I work to keep up a certain
for Identity Synthesis and Confusion, provides a better fit to the image when I’m with people” (Item 11), “I have a strong sense
data than either the one- or two-factor solutions. This finding of what it means to be male/female” (Item 8), and “I don’t really
feel involved” (Item 12). These items may need to be revised or
deleted in future versions of the EPSI.
2 We were not able to evaluate invariance of the bifactor solution across The superior fit of the bifactor model—with separate scales
gender or ethnicity because the ratio of participants to parameters estimated for overall identity and for positively and negatively worded
would have been only slightly greater than 1. items—allows for the possibility that overall identity, identity
FACTOR STRUCTURE OF IDENTITY 147

synthesis, and identity confusion may show somewhat different well-being (Schwartz, Côté, & Arnett, 2005). The EIPQ as-
correlations to other identity measures and to measures of psy- sesses identity exploration and commitment—the dimensions
chosocial adjustment. We therefore proceeded to examine issues underlying the identity status model. The Exploration subscale
of convergent validity with other identity measures in Study 2. assesses the extent to which the individual has considered poten-
tial identity alternatives in the areas of political preference, reli-
STUDY 2 gious ideology, career choice, personal values, friendship ideals,
Method dating preferences, gender roles, and family relationships. Two
Sample and procedures. Participants in Study 2 were 194 items are included for each domain for a total of 16 exploration
students (46 men, 147 women, 1 unidentified by gender; M age items (possible scores range from 16–80). The Cronbach’s alpha
= 19.90 years, SD = 3.29) enrolled in introductory psychol- estimate for this scale was .65. Although this estimate is low, it
ogy courses at the same large, urban, public university in Mi- should be noted that exploration occurs unevenly across the var-
ami. In terms of ethnicity, 27 participants were White, 68 were ious domains surveyed (Pastorino, Dunham, Kidwell, Bacho, &
Black, and 99 were Hispanic. The majority of White (77%), Lamborn, 1997) and that although these domains were placed
Black (71%), and Hispanic (65%) participants were born in together within identity status theory, exploration in one domain
the United States. Foreign-born Whites were primarily from may not be related to exploration in other domains (Schwartz
Europe and Canada; foreign-born Blacks were primarily from et al., 2006). The 16-item Commitment subscale (α = .73) as-
Haiti and Jamaica; and foreign-born Hispanics were primar- sesses the extent to which the person has made commitments
ily from Cuba, Colombia, Nicaragua, and Peru. Of participants to one or more sets of goals, values, and beliefs in the same
reporting annual family income (82% of the sample), 21% re- eight domains. Possible scores on the commitment subscale
ported less than $30,000 per year, 23% between $30,000 and also range from 16 to 80. Sample items include “I have tried to
$50,000, 26% between $50,000 and $100,000, and 13% above learn about different occupational fields to find the best one for
$100,000. Procedures in Study 2 were the same as those used me” (exploration), and “I have firmly held views concerning my
in Study 1. role in my family” (commitment).
The EOM–EIS–II generates continuous scores for each of the
Measures four identity statuses (diffusion, foreclosure, moratorium, and
achievement) in eight content domains. Six of these domains
EPSI identity scales. We again used the EPSI Identity Syn-
are the same as those assessed by the EIPQ, and two domains
thesis and Confusion scales. Cronbach’s alpha coefficients for
differ between the two measures. The EOM–EIS–II assesses
scores on the subscales were .82 for the overall Identity scale,
lifestyle and recreational activities instead of personal values
.73 for Identity Synthesis, and .69 for Identity Confusion.
and family relationships. The EOM–EIS–II contains 64 items,
Comparison Eriksonian measures. We also used two ad- 16 measuring each of the four statuses. Possible scores for each
ditional Eriksonian measures—the Ego Identity Scale and the of the statuses range from 16 to 80. Sample items include “I
Identity Confusion Inventory (Côté, 1984; Schwartz, 2007; don’t give religion much thought, and it doesn’t bother me one
Schwartz, Zamboanga, Weisskirch, & Rodriguez, in press). way or the other” (diffusion); “My own views on a desirable life
These measures are intended to measure the same constructs style were taught to me by my parents, and I don’t see any need
as the EPSI identity subscales, and as such, they provide oppor- to question what they taught me” (foreclosure); “In finding an
tunities to assess the convergent validity of scores generated by acceptable viewpoint to life itself, I find myself engaging in a
the EPSI. A 5-point Likert scale ranging from 1 (strongly dis- lot of discussions with others and some self-exploration” (mora-
agree) to 5 (strongly agree) was used for each of the comparison torium); and “I’ve gone through a period of serious questions
Eriksonian measures. about faith and can now say I understand what I believe as an
The Ego Identity Scale consists of 12 items (α = .72) assess- individual” (achievement). Cronbach’s alpha estimates in this
ing the extent to which the individual has developed a clear and sample were diffusion, .71; foreclosure, .87; moratorium, .76;
consistent sense of self, goals, and plans for the future. Possible and achievement, .77. We note that only continuous identity sta-
scores range from 12 to 60. The Identity Confusion Inventory tus scores were used in this study, and that categorical identity
consists of 8 items (α = .70) assessing the extent to which the status classifications were not used.
individual feels conflicted, hesitant, or unsure about who she or
he is or about her or his goals and plans. Possible scores range Results
from 8 to 40. In past research, the Ego Identity Scale has been
found to pattern closely with other measures of identity consoli- We modeled the EPSI bifactor solution, as obtained in Study
dation such as identity commitment, endorsement of the identity 1, and allowed each of the three factors (overall Identity Scale,
achieved status, and measures of subjective adulthood and com- Identity Synthesis, and Identity Confusion) to correlate with ob-
munity integration (Schwartz, 2007). The Identity Confusion served scores on the Ego Identity Scale, the Identity Confusion
Inventory has been found to cluster with the identity confusion Inventory, the EIPQ Exploration and Commitment scales, and
items from the EPSI and with scores on the identity diffused the continuous EOM–EIS–II identity status scales. The resulting
status (Schwartz et al., in press). model provided an adequate fit to the data, χ 2 (118) = 221.50,
p < .001; CFI = .92; NNFI = .97; RMSEA = .068 (90% CI
Identity status measures. We utilized two continuous mea- = .054–.081); SRMR = .062. Correlations between the EPSI
sures of identity status—the EIPQ (Balistreri et al., 1995) and bifactor model scales and other identity measures are presented
the EOM–EIS–II (Bennion & Adams, 1986). Both of these in Table 1.
measures have been found to function in accordance with the- The total identity score was positively related to scores on the
oretical expectations vis-à-vis indexes of personal agency and Ego Identity Scale and on the Foreclosed and Achieved Status
148 SCHWARTZ, ZAMBOANGA, WANG, OLTHUIS

TABLE 1.—Correlations between EPSI subscales and comparison measures: compared to a model with all of the correlations set equal across
Study 2. gender. We held correlations among the other identity variables
EPSI Subscale
equal across gender in both models. We conducted a similar
procedure for ethnicity.
Identity Identity Identity The chi-square difference test between the constrained and
Measure/Subscale Totala Synthesisb Confusionb unconstrained models can be computed as the difference be-
tween the respective –2 log likelihood values. This difference
Ego Identity Scale .44∗∗ .45∗∗∗ −.47∗∗∗ is distributed as a chi-square with degrees of freedom equal to
Identity Confusion Inventory −.27∗ −.53∗∗∗ .41∗∗∗
EIPQ Identity Exploration .03 .01 .04
the difference in the numbers of parameters between the two
EIPQ Identity Commitment .27 .53∗∗∗ −.43∗∗∗ models. Following Kass and Wasserman (1995), we also report
EOM–EIS Identity Diffusion .03 −.19∗ .31∗∗∗ differences in the model information criteria—the Akaike in-
EOM–EIS Identity Foreclosure .21∗ −.01 .15 formation criterion (AIC), the Bayesian information criterion
EOM–EIS Identity Moratorium −.30∗ −.21 .41∗∗∗ (BIC), and the adjusted BIC (ABIC). The chi-square differ-
EOM–EIS Identity Achievement .27∗ .40∗∗∗ −.29∗∗
ence indicated that the correlations between the EPSI subscales
Note. EPSI = Erikson Psychosocial Stage Inventory; EIPQ = Ego Identity Process and the other identity variables were equivalent across gender,
Questionnaire; EOM–EIS = Extended Objective Measure of Ego Identity Status.
a
χ 2 (24) = 11.53, p = .98; AIC = 24.94, BIC = 103.00,
Latent variable attached to all of the EPSI items. b Latent variable attached only to the
corresponding EPSI items.
and ABIC = 26.98; and across ethnicity, χ 2 (48) = 47.55,
*p < .05. **p < .01. ***p < .001. p = .49; AIC = 0.90, BIC = 157.26, and ABIC = 5.21.

subscales, and it was negatively related to scores on the morato- Discussion

rium status. Identity synthesis was positively related to scores on These results provide evidence for reasonable criterion valid-
the Ego Identity Scale and to identity commitment and achieve- ity for the EPSI subscales—both the total Identity score and the
ment scores. Identity synthesis scores were negatively related to independent effects of positively and negatively worded items
scores on the Identity Confusion Inventory and to moratorium (represented as Identity Synthesis and Confusion). Indeed, the
and diffusion scores. Identity confusion scores were positively total Identity score, representing the way in which the EPSI
related to scores on the Identity Confusion Inventory and to dif- has been traditionally scored, displayed fewer significant re-
fusion and moratorium scores and negatively related to scores on lationships with other identity measures than did the method
the Ego Identity Scale and on identity commitment and achieve- effects—the synthesis and confusion items. The synthesis and
ment. Identity exploration scores were unrelated to any of the confusion items were related to the comparison identity mea-
EPSI factors. Because the total Identity factor was specified as sures in theoretically expected ways—particularly with indexes
uncorrelated with the Identity Synthesis and Confusion factors of identity commitment, identity synthesis and confusion, and
(as is the assumption in bifactor models), the correlations of identity diffusion and achievement. These correlations strongly
the Identity Synthesis and Confusion scores with comparison suggest that the synthesis and confusion scores are substantively
measures can be regarded as independent from the correlations meaningful and are more than just method effects. Indeed, as
of the total Identity factor with these comparison measures. As has been found with the Rosenberg Self-Esteem Scale (Quilty
a result, it can be surmised that the effects of positively worded et al., 2006), the positively and negatively worded items carry
(synthesis) and negatively worded (confusion) items have sub- substantive meaning as separate factors beyond their inclusion
stantive importance beyond the effects of the EPSI Identity scale together within the larger Identity subscale.
as it has been traditionally scored. Continuous scores for identity commitment and for the
We then explored the extent to which these correlations would achieved status, as indexes of identity consolidation (Schwartz,
be consistent across gender and across ethnicity. In these analy- 2007), were positively related to identity synthesis and nega-
ses, we examined only heterogeneity of covariance (equivalence tively related to identity confusion as would be expected. These
of correlations across groups) to ensure that the model could results provide some evidence for construct validity, as much
be generalized across gender and across ethnic groups. Because of the convergence between Erikson’s (1950) and Marcia’s
only the relationships of the EPSI subscales to comparison iden- (1966) views of identity lies in the conceptualization of identity
tity measures—and not factor pattern coefficients, relationships achievement as an index of consolidating and synthesizing a
among the EPSI subscales, or relationships among the compari- sense of self (Schwartz, 2005). The association of moratorium
son identity measures—were tested across gender and ethnicity, with identity confusion is consistent with prior literature (Kid-
we estimated a finite mixture model (Muthén & Muthén, 2007). well et al., 1995; Meeus, 1996), suggesting that suspending com-
The KNOWNCLASS option in Mplus was used to specify gen- mitments and searching for a sense of self is associated with dis-
der or ethnicity as the categorical mixture variable (Muthén & orientation and distress. In fact, moratorium was more strongly
Muthén, 2007). This approach allows for specified parameters related than diffusion to identity confusion—calling into
to be estimated across levels of an observed categorical variable question the assumption that diffusion represents identity con-
and is useful in cases in which the sample sizes within each fusion within the identity status model (this is discussed more
group (e.g., genders or ethnic groups) are too small to reliably extensively in the General Discussion section following Study
estimate a full multigroup model—which would be required to 3). Interestingly, the EPSI overall identity scale was not signifi-
conduct a full invariance test (Vandenberg & Lance, 2000). In cantly related to identity commitment or to the diffused status. It
the invariance test by gender, only the correlations of the EPSI appears that the overall identity scale from the EPSI maps onto
subscales to the other identity variables were allowed to vary Erikson’s (1950, 1968) dimension of identity synthesis versus
across gender in the unconstrained model, and this model was confusion but less strongly onto identity status. The Synthesis
FACTOR STRUCTURE OF IDENTITY 149

and Confusion subscales from the EPSI appear to map onto both We assessed self-esteem using the Coopersmith (1981) Self-
Erikson’s theory and onto identity status—again suggesting Esteem Scale (25 items, α = .87). This scale assesses partic-
that these two “smaller” subscales are necessary for scores on ipants’ overall self-worth. Sample items include “I’m a lot of
the EPSI to possess maximal construct validity. fun to be with,” and “I have a low opinion of myself” (re-
The null relationship of identity exploration to all three versed). We assessed purpose in life using the Purpose in Life
EPSI scales—Overall Identity, Identity Synthesis, and Identity Scale (Crumbaugh & Maholick, 1969; 12 items, α = .81). This
Confusion—supports the premise that there may be multiple scale assesses participants’ sense of life purpose and direction.
types of identity exploration. Some of these types may lead to Sample item includes “In life, I have very clear goals and aims
consolidating a sense of self in emerging adulthood, whereas for myself.” We assessed internal locus of control using Côté’s
others may not (Luyckx et al., 2008). The null relationships ob- (1997) adaptation of Rotter’s (1966) Locus of Control Scale,
tained here may result from aggregating these different types of with a 5-point Likert scale ranging from 1 (strongly disagree)
exploration into a single variable. to 5 (strongly agree) used in place of the traditional ipsative
The findings from Study 2 are consistent with a view of format. This adapted version consists of five items (α = .63).
achievement as an index of identity consolidation (cf. Schwartz, Sample items include “What happens to me is my own doing.”
2007) and of moratorium as inviting identity confusion We assessed ego strength using the Ego Strength Scale (Epstein,
(Schwartz et al., in press). The question that remains, however, 1983; 26 items, α = .81). This scale assesses the extent to which
is whether the overall Identity subscale, Identity Synthesis, and participants cope well, are resilient, and persevere in the face of
Identity Confusion contribute collaboratively and independently adversity. Sample items include “I don’t fall apart under stress as
to psychosocial functioning. We addressed this issue in Study 3. easily as most people do.” Each of these scales has been shown
to function equivalently across ethnic groups (Schwartz, Côté,
STUDY 3 et al., 2005).
Method
Internalizing symptoms. We operationalized internalizing
Sample and procedures. Participants in Study 3 were 523 symptoms in terms of anxiety and depression. To assess depres-
students (115 men, 407 women, 1 unidentified by gender; M sive symptoms, we utilized the Center for Epidemiologic Studies
age = 20.45, SD = 4.40) enrolled in introductory psychology Depression Scale (CES–D; Radloff, 1977; 20 items, α = .91).
courses at the same urban, public university in Miami. The We assessed anxiety symptoms using the Beck Anxiety Inven-
ethnic breakdown was 107 White, 57 Black, and 357 Hispanic. tory (Beck, Epstein, Brown, & Steer, 1988; 18 items, α = .92).
The majority of Whites (87%), Blacks (78%), and Hispanics Both of these instruments contain items referencing specific ex-
(68%) were born in the United States. Foreign-born Whites were periences during the week prior to assessment. Sample items
primarily from Europe and Canada; foreign-born Blacks were include “I have felt down and unhappy this week” (CES–D),
primarily from Haiti and Jamaica; and foreign-born Hispanics and “I have been worrying a lot this week” (Beck Anxiety In-
were primarily from Cuba, Colombia, Nicaragua, and Peru. ventory). The CES–D has been found to be especially sensitive
Of participants reporting annual family income (86% of the to severity of depressive symptoms (Santor, Zuroff, Ramsay,
sample), 16% reported less than $30,000 per year, 24% between Cervantes, & Palacios, 1995). The Beck Anxiety Inventory has
$30,000 and $50,000, 29% between $50,000 and $100,000, and been found to correlate well with other measures of anxiety
17% above $100,000. (Osman, Kopper, Barrios, Osman, & Wade, 1997).
We performed data collection for Study 3 over the Internet.
We uploaded measures onto an online platform, and we directed Results
participants to the study Web site through the psychology depart- We estimated a model in which the EPSI bifactor solution was
ment’s research participation system. Participants were required allowed to correlate with scores on the indexes of psychosocial
to check a box indicating that they consented to participate be- functioning. Due to the computational burden involved in es-
fore they could access the measures. The EPSI was administered timating a model with a bifactor solution and six other latent
first, followed by the internalizing measures and then the adap- variables (four for adaptive psychosocial functioning and two
tive functioning measures. for maladaptive psychosocial functioning), we entered the psy-
chosocial functioning measures as observed summed scores.
Measures All of these correlations were significant at p < .001 (see Ta-
EPSI Identity subscales. We again used the EPSI sub- ble 2). Both the identity total score and identity synthesis were
scales. Internal consistency reliability estimates in Study 3 were positively related to all four indexes of adaptive psychosocial
.85 for the total Identity scale, .75 for Identity Synthesis, and functioning and negatively related to anxiety and depression.
.75 for Identity Confusion. Identity confusion was negatively related to self-esteem and
purpose in life but positively related to internal locus of con-
Adaptive psychosocial functioning. We measured adap- trol and to ego strength. It was also significantly and positively
tive psychosocial functioning in terms of self-esteem, purpose related to anxiety.
in life, internal locus of control, and ego strength (resilience). To ascertain the extent to which overall identity, identity syn-
These four indexes not only reflect a sense of subjective well- thesis, and identity confusion subscales would independently
being (Sheldon et al., 2004), but they also reflect a sense of and collaboratively relate to adaptive psychosocial functioning
self-determination and agency (Côté, 2002; Schwartz, Côté, et and to internalizing problems, we utilized a multivariate mul-
al., 2005). These four scales have been shown to cluster closely tiple regression technique. This approach to multivariate mul-
together, which supports grouping them under the heading of tiple regression is a form of path analysis in which multiple
agency and well-being (Côté, 1997). independent and dependent variables are included. Similar to
150 SCHWARTZ, ZAMBOANGA, WANG, OLTHUIS

TABLE 2.—Correlations between EPSI subscales and psychosocial adjustment: tively related to internal locus of control; and the total Identity
Study 3. score and Identity Synthesis, but not Identity Confusion, were
EPSI Subscale
positively related to ego strength. These findings suggest that
there was enough independence between these two dimensions
Identity Identity Identity of identity to facilitate separate relationships to self-esteem,
Measure/Subscale Totala Synthesisb Confusionb purpose in life, ego strength, and depression. Moreover, these
relationships were strong enough to account for at least 10%,
Self-Esteem .54∗∗∗ .25∗∗∗ −.24∗∗ and in some cases more than 40%, of variability in the psy-
Purpose in Life .56∗∗∗ .17∗∗ −.26∗∗
Internal Locus of Control .27∗∗∗ .08 .25∗
chosocial adjustment indexes. Importantly, beyond the effects
Ego Strength .41∗∗∗ .23∗∗∗ .31∗∗∗ of the overall Identity score, both Identity Synthesis and Iden-
Anxiety −.29∗∗∗ −.21∗∗∗ .18∗∗∗ tity Confusion were positively related to adaptive psychosocial
Depression −.37∗∗∗ −.21∗∗∗ .18 functioning and negatively related to maladaptive psychosocial
functioning. This finding further emphasizes the importance of
Note. EPSI = Erikson Psychosocial Stage Inventory.
a
Latent variable attached to all of the EPSI items. b Latent variable attached only to the the synthesis and confusion subscales, beyond the total identity
corresponding EPSI items. subscale, in modeling the factor structure of the EPSI.
*p < .05. **p < .01. ***p < .001. To ascertain the extent to which these results would be gener-
alizable across gender and ethnicity, we estimated a path mixture
analysis in which the paths from the EPSI subscales to the in-
univariate multiple regression, directional paths are estimated dicators of psychosocial functioning would be allowed to vary
from each predictor to each outcome variable. We entered age, across gender and across ethnicity. We did not allow factor pat-
gender, and ethnicity as covariates, with ethnicity operational- tern coefficients, correlations between the EPSI subscales, or
ized as dummy-coded variables for Black and Hispanic. We correlations between the psychosocial functioning indexes to
allowed identity synthesis and identity confusion as well as all vary across gender or ethnicity. For both gender and ethnicity,
pairs of endogenous variables to intercorrelate. Allowing the we compared this unconstrained model to a constrained model
endogenous variables to correlate is consistent with the premise in which all parameters were held equal across groups. Results
that indicators of positive and negative psychosocial functioning indicated that the path coefficients were equivalent across gen-
should all be related to one another (Keyes, 2005). Correlations der, χ 2 (18) = 19.02, p = .39; AIC = 2.04, BIC = 74.28,
between the overall Identity scale and the Identity Synthesis and and ABIC = 17.15; and across ethnicity, χ 2 (36) = 25.73, p =
Confusion scales were constrained to zero as were correlations .90; AIC = 20.55, BIC = 173.27, and ABIC = 59.00.
between the covariates and the EPSI subscales.
The regression model fit the data well, χ 2 (149) = 352.51,
p < .001; CFI = .98; NNFI = .97; RMSEA = .051 (95% CI = GENERAL DISCUSSION
.044–.058); SRMR = .056. The vast majority (20 of 24) of the We conducted this set of studies to ascertain the extent to
paths from demographic covariates to psychosocial functioning which a one-factor, two-factor, or hybrid bifactor structure
indexes were nonsignificant. Three of the significant effects of would best characterize the EPSI as well as to evaluate the
demographic variables involved gender. Women scored signif- internal consistency reliability, cross-ethnic consistency, con-
icantly higher than men on purpose in life, anxiety, and de- current validity, and construct validity of the solution that we
pression. Additionally, Hispanics reported significantly lower retained. The issue of factor structure has important theoretical
self-esteem than Whites. Overall, the model accounted for more as well as methodological implications. From a theoretical per-
than 40% of variability in self-esteem and purpose in life; more spective, whereas Erikson (1950) proposed that identity would
than 20% of variability in ego strength, anxiety, and depression; best be represented as a single continuum ranging from syn-
and more than 10% of variability in internal locus of control thesis to confusion, Marcia (2002) called for a recasting of the
(see Table 3). Total identity, identity synthesis, and confusion identity stage as synthesis with confusion. A one-factor model
were all significantly related to all of the outcomes except in- would support Erikson’s (1950) conceptualization, a two-factor
ternal locus of control and ego strength. The total Identity score model would support Marcia’s (2002) conceptualization, and a
and Identity Confusion, but not Identity Synthesis, were posi- bifactor model would support both of these conceptualizations.

TABLE 3.—Multivariate multiple regression: Psychosocial functioning on identity synthesis and confusion, Study 3.

Variable Self-Esteem Purpose in Life Internal Locus of Control Ego Strength Anxiety Depression

Age .04 −.01 .05 .05 −.02 −.02

Gender (female) −.01 .11∗ .06 −.08 .12∗∗ .11∗∗
Ethnicity (Black) −.07 −.04 −.01 −.02 −.04 −.01
Ethnicity (Hispanic) −.14∗∗ .02 .01 .02 .03 .04
Identity totala .53∗∗∗ .56∗∗∗ .26∗∗∗ .41∗∗∗ −.29∗∗∗ −.37∗∗∗
Identity synthesisb .29∗∗∗ .21∗∗∗ .09 .25∗∗∗ −.25∗∗∗ −.26∗∗∗
Identity confusionb .29∗ .33∗∗∗ .19∗∗∗ .22 −.27∗∗ −.30∗∗
Total R 2 .44∗∗∗ .46∗∗∗ .12∗∗ .28∗∗∗ .22∗∗∗ .28∗∗∗

Note. EPSI = Erikson Psychosocial Stage Inventory.

a
Latent variable attached to all of the EPSI items. b Latent variable attached only to the corresponding EPSI items.
*p < .05. **p < .01. ***p < .001.
FACTOR STRUCTURE OF IDENTITY 151

A Bifactor Model of Identity: Extent of Support in These (1950, 1968) identity stage vis-à-vis adaptive and maladaptive
Results psychosocial functioning.
These findings appear to support the bifactor model. This
model implies that the EPSI, and perhaps the structure of Erik- Convergent validity with other identity measures. The
son’s (1950) identity stage itself, may be characterized primarily finding that the EPSI subscales (primarily Identity Synthesis and
by a single dimension—but that identity synthesis and confu- Identity Confusion) correlated with identity status measures,
sion also function, to some extent, as separate variables. This again across gender and ethnicity, in theoretically consistent
suggests that although identity synthesis and confusion are in- ways carries both theoretical and methodological importance.
versely related at the bivariate level (r = −.83 in the two-factor Identity synthesis, the achieved status, and identity commitment
model), they are not mutually exclusive and can coexist within would be expected to interrelate as indexes of identity consoli-
an individual. The questions to be addressed would therefore be dation; and identity confusion would be expected to relate to the
both (a) how well developed is the person’s sense of self and diffused and moratorium statuses. Interestingly, the diffused sta-
(b) how much synthesis, and how much confusion, is present tus, which, in theory, represents Erikson’s (1950, 1968) notion
within the individual. of identity confusion, was less strongly correlated (r = .31) with
Methodologically, it is important to ascertain the extent to the EPSI Identity Confusion subscale than was the moratorium
which measurement instruments are able to accurately capture status (r = .41). Although the comparatively low Cronbach’s
the theoretical constructs of interest. In this study, we were able alpha for diffusion may partially explain this finding, the find-
to support both Erikson’s (1950) and Marcia’s (2002) positions. ing is more likely due to subtleties in the diffused status that
We were able to statistically reject a one-factor model that sup- are not captured by the EOM–EIS–II. For example, Luyckx,
ported only Erikson’s (1950) position as well as a two-factor Goossens, Soenens, Beyers, and Vansteenkiste (2005) empiri-
model that supported only Marcia’s (2002) position. Based on cally extracted two types of diffusion—a “carefree” type that
these results, we recommend that future research with this age did not appear confused or distressed and a “diffused diffusion”
group utilize the bifactor solution. To ascertain its appropri- status associated with significantly elevated scores on indexes of
ateness for different age groups, the bifactor solution should psychological distress. Both of these types of diffusion reflect a
also be replicated with middle school and high school students lack of systematic exploration coupled with an absence of com-
from diverse backgrounds to ascertain its viability with early mitment, which matches the definition of diffusion advanced
and middle adolescents in addition to emerging adults. It is also by Marcia (1966) and operationalized within the EOM–EIS–II.
important for further research to examine the factor structure However, aggregating across these two types of diffusion may
of the EPSI in ethnic and cultural groups not included in this have attenuated the relationship between the diffused status and
set of studies. Such studies will further examine the universal identity confusion. It is possible, for example, that diffused dif-
applicability of Erikson’s (1950) theory of identity. fusion is more consistent with identity confusion than is carefree
Moreover, the bifactor solution assumes a series of constraints diffusion.
that can be implemented only in a structural equation model- On the other hand, Luyckx et al. (2008) were only able to
ing (SEM) framework (e.g., each item loads separately on the identify one type of moratorium—one that is associated with
primary Identity factor and on either Identity Synthesis or Con- distress, confusion, and low self-worth. Similarly, Schwartz et
fusion; and the correlations between the primary Identity factor al. (in press) found that the moratorium status is related to psy-
and the other factors are constrained to zero). As a result, anal- chosocial functioning through identity confusion. These results
yses using the EPSI should be conducted using a latent variable suggest that active identity exploration, represented by scores
framework in which these constraints can be applied. Simpler on the moratorium status and by present-tense items, is strongly
analyses using only observed variables—such as bivariate corre- associated with identity confusion. The EIPQ Exploration scale,
lation, multiple regression, and analysis of variance—may not in which items are phrased in the past tense, was unrelated to
be appropriate for use with this measure (or with other Erik- any of the EPSI subscales. Identity exploration, in and of itself,
sonian identity measures). Latent-variable equivalents of these may both reflect open-mindedness and be associated with dis-
analyses, such as SEM and latent mean comparisons, should be tress and confusion. However, when exploration is ongoing, it
used instead. is likely to be coupled with the absence of commitment (i.e., the
moratorium status), which is then associated with distress and
poor psychosocial functioning (Luyckx et al., 2008; Schwartz
et al., in press).
Associations of the EPSI Identity Dimensions to It is also important to note that the Identity Synthesis and
Comparison Variables Identity Confusion subscales were more strongly and consis-
Three primary findings emerged from the correlations of the tently related to the comparison identity measures (with the
EPSI subscales with comparison variables. First, the three EPSI exception of the Ego Identity Scale and the Foreclosure sub-
subscales implied within the bifactor model—the overall Iden- scale from the EOM–EIS–II) than was the total Identity score.
tity scale, Identity Synthesis, and Identity Confusion—were re- This may signify that identity status, at least as measured by the
lated largely as expected to other Eriksonian and identity status EIPQ and the EOM–EIS–II, is more concerned with identity
measures. Second, the three EPSI subscales contributed inde- synthesis or confusion than with the individual’s overall sense
pendently and collaboratively to several indexes of psychosocial of self. To some extent, this pattern of findings is consistent
adjustment. Third, both of the aforementioned findings gener- with some criticisms of the identity status model (e.g., Côté &
alized across gender and across ethnic groups. This pattern of Levine, 1988; van Hoof, 1999), which have framed the status
findings speaks to the validity of the EPSI as an Eriksonian model as misrepresenting Erikson (1950, 1968). To the extent
measure of identity as well as to the relevance of Erikson’s to which Erikson’s theory of identity is captured within the total
152 SCHWARTZ, ZAMBOANGA, WANG, OLTHUIS

identity score—which did not correlate consistently with iden- were not well represented. Because the numbers of Asian Amer-
tity status measures—these criticisms appear to possess some icans and Native Americans at the university where data were
merit. collected are extremely small, these groups were not included in
analysis. Moreover, the Hispanics in this sample are somewhat
Construct Validity With Measures of Psychosocial representative of the Miami Hispanic population, but they do
Functioning not represent the U.S. Hispanic population as a whole, which
is largely Mexican American and Puerto Rican (Ramirez & de
The total Identity score, Identity Synthesis, and Identity Con-
la Cruz, 2003). As a result, future studies should estimate the
fusion all appeared to make separate contributions to several
bifactor solution for the EPSI on ethnic groups other than those
important dimensions of emerging-adult psychosocial function-
well represented in these samples.
ing. The observed correlation between Identity Synthesis and
Fourth, although the psychosocial functioning measures used
Identity Confusion (r = −.13), along with constraining the cor-
in this study represent important indicators of well-being and
relations of the total Identity score with the other subscales to
distress in emerging adulthood (Côté, 2002; Schulenberg &
zero, suggested that the three EPSI subscales would be uncorre-
Zarrett, 2006), other prominent behavioral outcomes in emerg-
lated and that their contributions to the psychosocial function-
ing adulthood, such as drug and alcohol use (Arnett, 2005) and
ing indexes would be largely independent. Most intriguingly,
sexual behavior (Lefkowitz & Gillen, 2006), were not included.
Identity Synthesis and Confusion were both positively related
The possibility of method biases cannot be discounted; some
to indexes of adaptive psychosocial functioning and negatively
individuals may have been more likely to report both identity
related to indexes of maladaptive psychosocial functioning. Al-
confusion and adjustment difficulties or to report both iden-
though identity confusion has been found to be associated with
tity synthesis and high levels of well-being (cf. Podsakoff, Lee,
problematic outcomes in past research (e.g., Schwartz, Mason,
MacKenzie, & Podsakoff, 2003). As a result, it is essential for fu-
Pantin, & Szapocznik, 2008), within the bifactor model, it ap-
ture research to map the associations of the EPSI bifactor model
pears that—beyond the effects of possessing a positive sense of
to behavioral as well as psychosocial outcomes in emerging
self—some degree of both coherence and confusion is adaptive
adulthood.
in emerging adulthood. Such a conclusion suggests that at least
Despite these limitations, this study represents one of the
in emerging adulthood in the United States, it is adaptive to
first attempts to ascertain the dimensionality of Erikson’s (1950,
possess a synthesized sense of self and to feel that there is some
1968) identity stage, and to connect empirical measures of Erik-
“room left over” for more identity work. As a result, the most
son’s identity stage to specific statuses within Marcia’s (1966)
adaptive psychosocial functioning may be associated with hav-
model. Although Erikson’s theory of identity is indeed more
ing developed a positive sense of identity (cf. Reis & Youniss,
expansive than the identity status model (Schwartz, 2001), the
2004), feeling that one has a consistent self across time and
two theories do share some important themes in common (Wa-
across situations (cf. Dunkel, 2005), and leaving oneself open
terman, 1999)—most notably a focus on identity synthesis and
for further identity exploration (cf. Luyckx et al., 2008).
confusion. These results are supportive of such a link, specif-
In summary, these results suggest that a bifactor solution for
ically between the achieved status and identity synthesis, and
the EPSI, consisting of both an overall Identity scale and sep-
between the moratorium status and identity confusion. These
arate subscales for Synthesis and Confusion, provides scores
findings also suggest that EPSI scores have good psychometric
with reasonable validity and reliability and is consistent with
properties and can be used to index an overall sense of identity—
identity theory. All three subscales appear to be important in
as well as identity synthesis and confusion—and that can be used
accurately representing the factor structure of scores generated
to relate to or predict indexes of psychosocial functioning. Given
by the measure and in explaining variability in indexes of adap-
that identity status studies have dominated the personal identity
tive and maladaptive psychosocial functioning. This bifactor
literature and that studies based directly on Erikson’s theory of
solution may pave the way for more inclusive and integrative
identity have been less common, we hope that this study inspires
research on identity—from a perspective that brings together
more research drawing directly on Erikson’s work.
Erikson’s (1950, 1968) and Marcia’s (1966) viewpoints.
ACKNOWLEDGMENTS
Limitations
Preparation of this article was supported by National Institute
These results should be interpreted in light of some important on Drug Abuse grants 19409 (S. Schwartz, Principal Investigator
limitations. First, the use of a university student sample may in- [PI]) and 17262 (H. Pantin, PI), and by National Institute on
hibit generalizability. There are important differences between Mental Health grant 40859 (C. H. Brown, PI).
college and noncollege emerging adults in terms of financial,
emotional, and intellectual functioning (Halperin, 2001). As a REFERENCES
result, care should be taken when generalizing these results to
emerging adults who do not attend a university. Second, the Arnett, J. J. (2000). Emerging adulthood: A theory of development from the late
teens through the twenties. American Psychologist, 55, 469–480.
overrepresentation of women in all three samples is a con-
Arnett, J. J. (2005). The developmental context of substance use in emerging
cern. Although we found invariance across gender in Studies adulthood. Journal of Drug Issues, 35, 235–254.
2 and 3 (we did not test for invariance in Study 1), it is not Balistreri, E., Busch-Rossnagel, N. A., & Geisinger, K. F. (1995). Development
known whether the results would have been different had a and preliminary validation of the Ego Identity Process Questionnaire. Journal
more gender-balanced sample been used. Third, although the of Adolescence, 18, 179–190.
samples used in these studies were ethnically diverse, and al- Beck, A. T., Epstein, N., Brown, G., & Steer, R. A. (1988). An inventory for
though this represents an important advance in identity research measuring clinical anxiety: Psychometric properties. Journal of Consulting
(Schwartz, 2005; Sneed et al., 2006), a number of ethnic groups and Clinical Psychology, 56, 893–897.
FACTOR STRUCTURE OF IDENTITY 153

Bennion, L. D., & Adams, G. R. (1986). A revision of the extended version of ing adults in America: Coming of age in the 21st century (pp. 235–256).
the Objective Measure of Ego Identity Status: An identity instrument for use Washington, DC: American Psychological Association.
with late adolescents. Journal of Adolescent Research, 1, 183–198. Luyckx, K., Goossens, L., & Soenens, B. (2006). A developmental contextual
Byrne, B. M. (2001). Structural equation modeling with AMOS: Basic concepts, perspective on identity construction in emerging adulthood: Change dynam-
applications, and programming. Mahwah, NJ: Lawrence Erlbaum Associates. ics in commitment formation and commitment evaluation. Developmental
Chen, F. F., West, S. G., & Sousa, K. H. (2006). A comparison of bifactor and Psychology, 42, 363–380.
second-order models of quality of life. Multivariate Behavioral Research, 41, Luyckx, K., Goossens, L., Soenens, B., Beyers, W., & Vansteenkiste, M. (2005).
189–225. Identity statuses based upon four rather than two identity dimensions: Ex-
Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes tending and refining Marcia’s paradigm. Journal of Youth and Adolescence,
for testing measurement invariance. Structural Equation Modeling, 9, 233– 34, 605–618.
255. Luyckx, K., Schwartz, S. J., Berzonsky, M. D., Soenens, B., Vansteenkiste,
Cohen, J. (1988). Statistical power analysis for the behavioral sciences. Hills- M., Smits, I., et al. (2008). Capturing ruminative exploration: Extending the
dale, NJ: Lawrence Erlbaum Associates. four-dimensional model of identity formation in late adolescence. Journal of
Coopersmith, S. (1981). The Self-Esteem Inventories. Palo Alto, CA: Consulting Research in Personality, 42, 58–82.
Psychologists Press. Marcia, J. E. (1966). Development and validation of ego identity status. Journal
Côté, J. E. (1984). The identity crisis: A formulation and empirical test of of Personality and Social Psychology, 5, 551–558.
Erikson’s theory of ego identity formation. Unpublished doctoral dissertation, Marcia, J. E. (2002). Identity and psychosocial development in adulthood. Iden-
York University, Toronto, Ontario, Canada. tity, 2, 7–28.
Côté, J. E. (1993). Foundations of a psychoanalytic social psychology: Neo- Markstrom, C. A., & Kalmanir, H. M. (2001). Linkages between the psychoso-
Eriksonian propositions regarding the relationship between psychic structure cial stages of identity and intimacy and the ego strengths of fidelity and love.
and social institutions. Developmental Review, 13, 31–53. Identity, 1, 179–196.
Côté, J. E. (1997). An empirical test of the identity capital model. Journal of Meeus, W. (1996). Studies on identity development in adolescence: An overview
Adolescence, 20, 421–437. of research and some new data. Journal of Youth and Adolescence, 25, 569–
Côté, J. E. (2002). The role of identity capital in the transition to adulthood: The 598.
individualization thesis examined. Journal of Youth Studies, 5, 117–134. Muthén, L. K., & Muthén, B. O. (2007). Mplus user’s guide. Los Angeles:
Côté, J. E., & Levine, C. (1988). A critical examination of the ego identity status Muthén and Muthén.
paradigm. Developmental Review, 8, 147–184. Osman, A., Kopper, B. A., Barrios, F. X., Osman, J. R., & Wade, T. (1997). The
Côté, J. E., & Schwartz, S. J. (2002). Comparing psychological and sociological Beck Anxiety Inventory: Reexamination of factor structure and psychometric
approaches to identity: Identity status, identity capital, and the individualiza- properties. Journal of Clinical Psychology, 53, 7–14.
tion process. Journal of Adolescence, 25, 571–586. Pastorino, E., Dunham, R. M., Kidwell, J. S., Bacho, R. A., & Lamborn, S. D.
Cross, S., Gore, J., & Morris, M. (2003). Relational-independent self-construal, (1997). Domain-specific gender comparisons in identity development among
self-concept consistency, and well-being. Journal of Personality and Social college youth: Ideology and relationships. Adolescence, 32, 559–577.
Psychology, 85, 933–944. Podsakoff, P. M., MacKenzie, S. B., Lee, J-Y., & Podsakoff, N. P. (2003). Com-
Crumbaugh, J. C., & Maholick, L. T. (1969). Manual of instructions for the mon method biases in behavioral research: A critical review of the literature
Purpose in Life Test. Munster, IN: Psychometric Affiliates. and recommended remedies. Journal of Applied Psychology, 88, 879–903.
Dunkel, C. S. (2005). The relationship between self-continuity and measures of Quilty, L. C., Oakman, J. M., & Risko, E. (2006). Correlates of the Rosenberg
identity. Identity, 5, 21–34. Self-Esteem Scale method effects. Structural Equation Modeling, 13, 99–
Epstein, S. (1983). The Ego Strength Scale. Princeton, NJ: Educational Testing 117.
Service. Quintana, S. M., & Maxwell, S. E. (1999). Implications of recent developments
Erikson, E. H. (1950). Childhood and society. New York: Norton. in structural equation modeling for counseling psychology. Counseling Psy-
Erikson, E. H. (1968). Identity: Youth and crisis. New York: Norton. chologist, 27, 485–527.
Greenberger, E., & Sorensen, A. B. (1974). Toward a concept of psychosocial Radloff, L. (1977). The CES-D scale: A self-report depression scale for research
maturity. Journal of Youth and Adolescence, 3, 329–358. in the general population. Applied Psychological Measurement, 1, 385–401.
Hall, S., & Brassard, M. R. (2008). Relational support as a predictor of identity Ramirez, R. R., & de la Cruz, G. P. (2003). The Hispanic population in the United
status in an ethnically diverse early adolescent sample. Journal of Early States, March 2002 [Current Population Report P20-545]. Washington, DC:
Adolescence, 27, 92–114. U.S. Census Bureau.
Halperin, S. (Ed.). (2001). The forgotten half revisited: American youth and Reis, O., & Youniss, J. (2004). Patterns in identity change and development in
young families, 1988–2008. Washington, DC: American Youth Policy Forum. relationships with mothers and friends. Journal of Adolescent Research, 19,
Hancock, G. R., & Freeman, M. J. (2001). Power and sample size for the root 31–44.
mean square error of approximation test of not close fit in structural equation Reise, S. P., Morizot, J., & Hays, R. D. (2007). The role of the bifactor model in
modeling. Educational and Psychological Measurement, 61, 741–758. resolving dimensionality issues in health outcomes measures. Quality of Life
van Hoof, A. (1999). The identity status field re-reviewed: An update of un- Research, 16, 19–31.
resolved and neglected issues with a view on some alternative approaches. Rosenthal, D. A., Gurney, R. M., & Moore, S. M. (1981). From trust to intimacy:
Developmental Review, 19, 497–556. A new inventory for examining Erikson’s stages of psychosocial development.
Kass, R. E., & Wasserman, L. A. (1995). A reference Bayesian test for nested Journal of Youth and Adolescence, 10, 525–537.
hypotheses and its relationship to the Schwarz criterion, Journal of the Amer- Rotter, J. B. (1966). Generalized expectancies for internal versus external locus
ican Statistical Association, 90, 928–934. of reinforcement. Psychological Monographs, 80(1, Whole No. 609).
Keyes, C. L. M. (2005). Mental illness and/or mental health: Investigating Santor, D. A., Zuroff, D. C., Ramsay, J. O., Cervantes, P., & Palacios, J. (1995).
axioms of the complete state model of health. Journal of Consulting and Examining scale discriminability of the BDI and CES-D as a function of
Clinical Psychology, 73, 539–548. depressive severity. Psychological Assessment, 7, 131–139.
Kidwell, J. S., Dunham, R. M., Bacho, R. A., Pastorino, E., & Portes, P. R. Schulenberg, J. E., & Zarrett, N. R. (2006). Mental health during emerging
(1995). Adolescent identity exploration: A test of Erikson’s theory of transi- adulthood: Continuity and discontinuity in courses, causes, and functions. In
tional crisis. Adolescence, 30, 785–793. J. J. Arnett & J. L. Tanner (Eds.), Emerging adults in America: Coming of age
Kline, R. B. (2006). Principles and practices of structural equation modeling in the 21st century (pp. 135–172). Washington, DC: American Psychological
(2nd ed.). New York: Guilford. Association.
Lefkowitz, E. S., & Gillen, M. M. (2006). “Sex is just a normal part of life”: Schwartz, S. J. (2001). The evolution of Eriksonian and neo-Eriksonian identity
Sexuality in emerging adulthood. In J. J. Arnett & J. L. Tanner (Eds.), Emerg- theory and research: A review and integration. Identity, 1, 7–58.
154 SCHWARTZ, ZAMBOANGA, WANG, OLTHUIS

Schwartz, S. J. (2005). A new identity for identity research: Recommendations Schwartz, S. J., Zamboanga, B. L., Weisskirch, R. S., & Rodriguez, L. (in press).
for expanding and refocusing the identity literature. Journal of Adolescent The relationships of personal and ethnic identity exploration to indices of
Research, 20, 293–308. adaptive and maladaptive psychosocial functioning. International Journal of
Schwartz, S. J. (2007). The structure of identity consolidation: Multiple corre- Behavioral Development.
lated constructs or one superordinate construct? Identity, 7, 27–49. Sheldon, K. M., Elliot, A. J., Ryan, R. M., Chirkov, V., Kim, Y., Wu, C., et al.
Schwartz, S. J., Adamson, L., Ferrer-Wreder, L., Dillon, F. R., & Berman, S. L. (2004). Self-concordance and subjective well-being in four cultures. Journal
(2006). Identity status measurement across contexts: Variations in measure- of Cross-Cultural Psychology, 35, 209–223.
ment structure and mean levels among White American, Hispanic American, Sneed, J. R., Schwartz, S. J., & Cross, W. E., Jr. (2006). A multicultural critique
and Swedish emerging adults. Journal of Personality Assessment, 86, 61–76. of Eriksonian-based research and theory: A call for integration. Identity, 6,
Schwartz, S. J., Côté, J. E., & Arnett, J. J. (2005). Identity and agency in emerging 61–84.
adulthood: Two developmental routes in the individualization process. Youth Steiger, J. H. (2006). Understanding the limits of global fit assessment in struc-
and Society, 37, 201–229. tural equation modeling. Personality and Individual Differences, 42, 893–
Schwartz, S. J., Mason, C. A., Pantin, H., & Szapocznik, J. (2008). Effects 898.
of family functioning and identity confusion on substance use and sexual Streitmatter, J. L. (1988). Ethnicity as a mediating variable of early adolescent
behavior in Hispanic immigrant early adolescents. Identity, 8, 107–124. identity development. Journal of Adolescence, 11, 335–346.
Schwartz, S. J., & Montgomery, M. J. (2002). Similarities or differences in Suh, E. M. (2002). Culture, identity consistency, and subjective well-being.
identity development? The impact of acculturation and gender on identity Journal of Personality and Social Psychology, 83, 1378–1391.
process and outcomes. Journal of Youth and Adolescence, 31, 359–372. Vandenberg, R. J., & Lance, C. E. (2000). A review and synthesis of the
Schwartz, S. J., Pantin, H., Prado, G., Sullivan, S., & Szapocznik, J. (2005). measurement invariance literature: Suggestions, practices, and recommen-
Family functioning, identity, and problem behavior in Hispanic immigrant dations for organizational research. Organizational Research Methods, 3, 4–
early adolescents. Journal of Early Adolescence, 25, 392–420. 70.
Schwartz, S. J., Zamboanga, B. L., Rodriguez, L., & Wang, S. C. (2007). The Waterman, A. S. (1999). Identity, the identity statuses, and identity status
structure of cultural identity in an ethnically diverse sample of emerging development: A contemporary statement. Developmental Review, 19, 591–
adults. Basic and Applied Social Psychology, 29, 159–173. 621.