Accommodating a new identity: Possible selves, identity change and well-being across two life-transitions

European Journal of Social Psychology

Eur. J. Soc. Psychol. (2009)

Published online in Wiley InterScience

(www.interscience.wiley.com) DOI: 10.1002/ejsp.669

*E

C

Research article

Accommodating a new identity: Possible selves, identity change andwell-being across two life-transitions

Correspondence to: Claudia Manzi, Departme-mail: [email protected]

opyright # 2009 John Wiley & Son

CLAUDIA MANZI1*, VIVIAN L. VIGNOLES2 ANDCAMILLO REGALIA1

1Catholic University of Milan, Italy2University of Sussex, UK

Abstract

We examined the role of desired, feared, and expected possible future identity structures in the restructuring of identity

after two life transitions. A longitudinal study was conducted on 86 young adults during the transition from school to

university and 143 adults during the transition to parenthood. In both samples, pre-transition desires and expectations

about the restructuring of identity predicted post-transition actual identity structures. Post-transition emotional well-

being was higher among those whose post-transition identity structures more closely matched their initial desires and less

closely matched their initial fears, and among those who reported a greater magnitude of identity change. We propose that

possible future identity structures play an important role in the identity accommodation process during life-transitions,

and that they have significant implications for well-being. Copyright # 2009 John Wiley & Sons, Ltd.

Identity changes during the life span are fundamental aspects of human development and aging (Cross & Markus, 1991).

Existing studies suggest that long-term self-concept change takes place especially in response to major changes in role and

situational demands (Banaji & Prentice, 1994). Such changes are often associated with periods of life-transition

(Kerpelman & Lamke, 1997). These crucial periods are instigated by a variety of factors, which may be related to family

life (e.g., marriage, childbirth, bereavement) or to educational and socio-economic activities (e.g., finishing school,

retirement, starting a new job). During a transition period, a person has to reorganize his/her social roles, and to cope with

temporal and structural changes in daily life, which will often be associated with changes in identity structure (Cigoli &

Scabini, 2006).

In this paper, we describe a longitudinal study of identity change over two different life-transitions: The transition to

parenthood and the transition from school to university. Our particular focus was on the role of possible selves in guiding

the processes of identity change. Possible selves (Markus & Nurius, 1986) are the component of identity that reflects the

perceived prospects of the individual. They include those selves that individuals expect to become (expected selves),

would like to become (desired selves) or are afraid of becoming (feared selves). Like the actual self, they consist of

organized self-views derived in part from representations of the self in the past and in part from perceived expectations

from others (Stryker & Serpe, 1994), but they are conceptualized in the future. Because of this future context, possible

selves represent the potential for change in the self-concept.

Previous research into possible selves and long-term change has typically focused on change in particular, single

identity elements, such as acquiring or losing a new social role (e.g., Dunkel, Kelts, & Coon, 2006; King & Raspin, 2004).

Thus, studies have examined the assimilation of new elements into identity, without considering the restructuring of

nt of Psychology, Universita Cattolica del Sacro Cuore, L.go Gemelli, 1, 20129 Milan, Italy.

s, Ltd.

Received 15 February 2007

Accepted 30 May 2009

Claudia Manzi et al.

existing identity content necessary to accommodate these new identity elements. In contrast, here we focused on the role of

possible selves in the restructuring of existing identity content when a new identity element is assimilated.

POSSIBLE SELVES AND ENDURING IDENTITY CHANGE

Based on Piaget’s (1936) theories of individuals’ cognitive development and adaptation, Breakwell (1986) describes

enduring change in identity as the result of cognitive processes of assimilation and accommodation. These terms describe

how the content and structure of an individual’s self-concept changes over time. Assimilation is the process by which

individuals incorporate new identity elements into the content and structure of their self-concept. Thus, a person who has

become a father absorbs this new identity content into his self-conception. Once assimilated, the presence of an important

new element of identity content is likely to change the overall structure of the self-concept to some degree.

Accommodation refers to the adjustment and reorganization of the existing identity content to ‘‘make space’’ for the new

information. When an individual assimilates a new identity element as a ‘‘father,’’ existing identity content will be re-

evaluated or reinterpreted in relation to the new identity element, and vice versa.

Existing studies into the role of possible selves in the process of identity change have focused mainly on their role in the

process of identity assimilation. In particular, researchers have identified two aspects of the assimilation process in which

possible selves are involved. First, a particular possible self is the potential new identity content that an individual desires/

fears/expects to assimilate into his/her identity. In this vein, the role of possible selves in the process of identity change has

been portrayed as a vital link between motivation and behavior. Possible selves are seen as the personalized, cognitive

manifestations of motives that drive individuals’ behavior (Markus & Nurius, 1987). Thus, a particular possible self gives

concrete form to abstract goals and both motivates and sustains behaviors in the direction of change. Second, once efforts

to change begin, in order to assimilate the new identity content, individuals seek evidence that the possible self is being

realized. In this process the most influential feedback comes from important others (Kerpelman & Pittman, 2001), and this

information may affect the stability of a particular possible self. In other words, without social validation of the ‘‘becoming

true’’ of a possible self, individuals more easily tend to change their choice of possible selves to pursue. On the other hand,

as validation continues over time, people increasingly think of themselves as actually being that which had been their

possible self. At this point, the possible self is assimilated into the identity structure, and becomes part of the actual self

(Wurf & Markus, 1991).

The vast majority of existing studies have analyzed this process by looking at possible selves as ‘‘goals’’ in the present

(i.e., studied at a single time point). For example, Yowell (2002) has shown that positive academic performance occurs

when a negative feared self acts in the present as a self-regulator. Recently, some studies have focused on the second part of

the process described: The validation of the possible self in the process of identity change. This literature has integrated

possible selves theory with self-verification theory (Swann, 1987) and identity control theory (Burke, 1991). Kerpelman

and Pittman (2001), for example, studied the possible self validation process during adolescence. This study shows that,

contrary to what happens with identity elements already set into the identity structure (cf. Swann, 1987), when possible

selves are important they tend to be unstable in the face of discrepant feedback.

These studies have begun to examine the role of possible selves in the assimilation process of a new element into the

identity structure. However, no previous research to our knowledge has considered the role of possible selves in the

accommodation process which accompanies the assimilation of a new identity element. Hence, we were interested to

examine whether possible selves might also play a role in the restructuring of existing identity content during significant

life changes.

POSSIBLE FUTURE IDENTITY STRUCTURES

In order to analyze identity accommodation here, we started from an holistic definition of identity as the subjective concept

of oneself as a person. Crucially, identity is composed of many aspects—or identity elements—on different levels of self-

representation (e.g., ‘‘shy,’’ ‘‘daughter,’’ ‘‘researcher’’), which vary considerably in their importance for self-definition

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. (2009)

DOI: 10.1002/ejsp

Possible selves and well-being

(Deaux, 1992). The concept of subjective identity structure (Vignoles, Regalia, Manzi, Golledge, & Scabini, 2006)

captures this multiplicity, referring to the perceived importance of each identity element within the self-image as a whole.

Thus, identity is viewed as a phenomenological structure in which each single element is evaluated as more or less central

and important for self-definition (Rosenberg & Gara, 1985).

In this study, we analyzed identity change at this structural level, focusing on changes in the perceived centrality of

multiple elements of identity before and after a life-transition. For example, if a woman’s identity before the transition to

parenthood includes the role of ‘‘daughter,’’ this identity element could gain or lose importance relative to other elements

in her subjective identity structure, after she becomes a mother. Extending this concept of subjective identity structure, we

developed the concept of possible future identity structures (see also Vignoles, Manzi, Regalia, Jemmolo, & Scabini,

2008), referring to the importance an individual gives to each of his/her current identity elements in his/her expected,

desired, and feared future selves. For example, if a woman perceives herself as ‘‘emotional,’’ and fears that she may

become even more emotional after becoming mother, this identity element can be said to assume a central position in her

feared possible future identity structure.

Vignoles et al. (2008) have shown that people can report their desires and fears about how existing content will be

restructured following assimilation of a new identity element, and that desires and fears about identity restructuring reflect

core identity motives as do desired and feared possible future selves. Clearly, this raises interesting questions about the role

of these possible future identity structures in the identity accommodation process. First, we should investigate if the

restructuring of existing identity content is purely reactive to the assimilation of new identity elements, or whether it is

meaningfully related also to these pre-existing possible future identity structures: Do expectations, desires and fears about

identity restructuring predict what actually happens to identity after the assimilation of a new identity element? Second,

does achieving or avoiding one’s pre-existing possible identity structures during a life-transition carry implications for

well-being?

We theorized that the restructuring of identity which follows the assimilation of a new identity element during a life

transition is directed at least partly by the possible future identity structures an individual has before the transition. In other

words, we propose that expected, desired, and feared possible future selves are involved in guiding not only the

assimilation of new identities but also the accommodation process involving existing identity elements. Hence, we

anticipated that the desired, feared, and expected importance of a particular identity element before a life-transition would

predict the actual importance of this content within subjective identity structure after the transition. We develop our

predictions concerning the role of possible future identity structures in shaping post-transition well-being in the next

section.

IDENTITY CHANGE AND EMOTIONALWELL-BEING

Changes in identity toward or away from a possible future self can be expected to affect psychological adjustment in

several ways. In particular, poor adjustment is likely to occur when the individual cannot achieve desired self-conceptions

or cannot avoid feared ones. According to self-discrepancy theory (Higgins, 1987), actual-ideal discrepancies in the self

typically lead to dejection-related emotions, including symptoms of depression. Similarly, Markus and Nurius (1986)

proposed that affect is derived from conflicts or discrepancies within the self-concept: Individuals will feel either

positively or negatively about themselves depending on the extent to which they achieve particular desired self-

conceptions or avoid particular feared ones.

Although research into the effects of actual-ideal self-discrepancies in the current self is widespread, less attention has

been paid to these issues in the possible selves literature. Existing studies generally focus on the compatibility of desired

and expected future selves. A key factor in these studies is not the distance between current and desired future selves, but

the degree to which desired future selves are viewed as likely, or attainable. Knox, Funk, Elliott, and Bush (1998) found

that the perceived likelihood of attaining hoped-for possible selves was a positive predictor of self-esteem among

adolescents of both sexes, and that the perceived likelihood of feared possible selves was a negative predictor of self-

esteem among female adolescents. Similarly, optimism is correlated with the positive evaluation of expected future selves

(Carver, Reynolds, & Scheier, 1994), and depression is associated with a greater expected likelihood of negative possible

future selves and a lower expected likelihood of positive possible future selves (Penland, Masten, Zelhart, Fournet, &

Callahan, 2000).

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. (2009)

DOI: 10.1002/ejsp

Claudia Manzi et al.

Less attention has been paid to the implications of actually attaining or avoiding one’s desired, or feared possible future

selves. An exception is Granberg (2006), who has studied the experience of validating or failing to validate a possible self

related to weight loss, and how people attempt to reconcile unvalidated possible selves. This qualitative study suggests

how efforts to validate a possible self can significantly complicate the integration of the new element of identity. In

particular, when social feedback does not match outcomes, an ‘‘identity interruption’’ is created and individuals have to

reformulate their previous possible self (e.g., ‘‘being thin and popular’’) in order to assimilate the new identity element

they have gained (‘‘being thin but not popular’’) into their identity structures.

In this study, we propose that not only the successful assimilation of a possible self into the identity structure but also a

successful accommodation process may have implications for individual adjustment after a transition. Thus, we theorized

that success or failure in achieving desired possible identity structures and avoiding feared ones has implications for

individual well-being. We hypothesized that the closer an individual’s post-transition subjective identity structure would

be to his/her pre-transition desired future identity structure, and the further from his/her pre-transition feared future

identity structure, the more positive emotion and the less negative emotion he/she would report.

THE CURRENT RESEARCH CONTEXT

The role of possible selves in guiding the restructuring of existing elements of identity, and the consequences of these

processes for individual well-being have almost never been tested directly. In the present study, we collected longitudinal

data across the transition to parenthood and across the transition from school to university. Although not providing the

causal certainty of an experimental design, we were able to test our hypotheses in these two natural settings.

Although entering university is primarily an educational transition and becoming a parent is a family transition, both

have huge repercussions beyond their primary domain, radically changing the structure of daily life and reorganizing

existing patterns of family and peer relationships (Cigoli & Scabini, 2006). Hence, it is not surprising that each should lead

to a complete restructuring of identity, rather than just some micro-changes (Cassidy & Trew, 2004; Cowan & Cowan,

1992; Ethier, 1996; Ethier & Deaux, 1994).

Yet, given that these two transitions are focused on different life-domains and typically occur at different points in the

lifespan, we were interested to explore for similarities and differences between them. In studying the transition to

parenthood, we were also especially aware of the possibility of heterogeneity in the experiences and reactions of different

individuals. In particular, there is good reason to suppose that the transition to parenthood might be experienced very

differently by new mothers and new fathers (Cowan & Cowan, 1992), as well as by those who were planning the birth and

those for whom the pregnancy was unexpected. Hence, we were also concerned to explore for possible heterogeneity in

effects among these various groups making the transition to parenthood.

SUMMARY OF OUR PREDICTIONS

If the process of identity change, and in particular the identity accommodation process after a life transition, is influenced

by the elaboration of possible future identity structures, then it should follow that identity elements perceived as more

central to desired and expected possible future identity structures before the transition will come to be perceived as more

central after the transition; in contrast, identity elements perceived as more central to feared possible future identity

structures will come to be perceived on the whole as less central after the transition. Thus, we hypothesized that

participants’ pre-transition desired and expected possible future identity structures would be positive prospective

predictors, whereas their feared possible future identity structures would be a negative prospective predictor of post-

transition residual change in their subjective identity structures.

It seemed unlikely that all participants would be equally successful in achieving their desired and in avoiding their

feared identity structures. We predicted that the closeness of participants’ post-transition subjective identity structures to

their pre-transition desired identity structures, and their distance from pre-transition feared identity structures, would be

positively associated with post-transition residual change in their emotional well-being.

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. (2009)

DOI: 10.1002/ejsp

Possible selves and well-being

METHOD

Participants and Procedure

Questionnaireswere distributed to high-school seniors and to expectingfirst-timeparents in Italy and theUK. Italian residents

responded in Italian, andUKresidents inEnglish.Thequestionnairewas translatedby themembers of the research team, all of

whom are fluent in both languages, using a ‘‘cultural decentering’’ approach (Werner & Campbell, 1973).

Regarding the transition from school to university, participants in the first wave were recruited through schools shortly

before taking their final school exams. We invited schools in the Lombardy region of Italy and in the southeast of England

to participate in the study. Schools were not sampled randomly but were selected to cover somewhat differing socio-

economic catchments. Pupils of three Italian schools and six UK schools participated. A member of the research team

explained the longitudinal design of the study and asked participants to leave their contact details on a separate sheet of

paper which was given the same code number as the anonymous questionnaire they completed. One hundred and ninety

two participants completed the questionnaire at the first wave. Three months after the beginning of their university courses

(eight months after the first wave of data collection), participants were contacted by phone or by email and were invited to

participate in the second wave of the study. They received the second questionnaire by mail or in organized meetings at

their new universities. Eighty six participants (44.9% of the first wave) completed the second wave questionnaire; the

attrition rate was quite high partly because many British participants could not be contacted as they were spending a year

abroad before starting university.

At time 1, students in the final sample were aged 17–20 years (mean: 18 years and 8 months; SD: 4 months); 64 were

resident in Italy and 22 in the UK; 42.2% of Italian participants and 77.3% of UK participants were female. As an index of

socio-economic background, parents of participants in both countries were educated on average to age 18 (Italy: Maturita;

UK: A-levels). In Italy, this was the mean, median, and modal level of education for both parents (40.7% of fathers and

38.3% of mothers). In the UK, there was greater diversity, with a larger proportion of parents educated to O-level/CSE (age

16: 25.3% of fathers and 33.0% of mothers) or to first-degree level (22.1% of fathers and 27.8% of mothers). Participants

who completed the second wave did not differ significantly from those who dropped out in age, gender ratio, or mother’s

and father’s education levels. Nor did they differ significantly in time 1 levels of anxiety, depressive symptoms, negative

affect, or positive affect.

Expecting first-time parents were recruited through health services and antenatal classes in the Lombardy region of

Italy and in England and Scotland. As with the preceding sample, we explained the longitudinal design of the study, asking

participants to provide their personal details on a separate sheet of paper, and we re-contacted them after the transition by

phone or email, to send their follow-up questionnaires. Participants responded on average in the 6th month of pregnancy

and 6months after the birth. Two hundred and forty six future parents participated in the first wave, of whom 143 (58.3% of

the first wave; 73 in Italy and 70 in the UK) participated in the second wave. Ages ranged between 21 and 53 years (mean:

32 years and 6 months; SD: 4 years and 6 months). Of the Italian sample, 70.0% were female and 87.1% were married; of

the UK sample, 71.2% were female and 89.0% were married. At time 1, most reported being in full-time employment

(UK: 87.5%; Italy: 79.9%). The modal level of education was first university degree in both countries (UK: 36.0%; Italy:

35.8%). A majority of parents (78.6%) reported that the pregnancy was planned, but 21.4% reported that it was

unexpected. Participants who completed the second wave did not differ significantly in average age, gender ratio, marital

status, occupational status, or education level from those who dropped out. Nor did they differ significantly in time 1 levels

of anxiety, depressive symptoms, negative affect, or positive affect.

Measures

Measures were included in a larger questionnaire about identity motives, subjective well-being, family processes, and

perceptions of the two life-transitions.1 Only those parts directly relevant to this paper are described here.

1Non-overlapping analyses of time 1 data only have been presented elsewhere, focusing on the motivational bases of subjective identity structures(Vignoles et al., 2006: Study 2), and of possible future identity structures (Vignoles et al., 2008: Study 2), and on the role of family processes ascontemporaneous predictors of psychological well-being (Manzi, Vignoles, Regalia, & Scabini, 2006).

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. (2009)

DOI: 10.1002/ejsp

Claudia Manzi et al.

Subjective Identity Structure and Possible Future Identity Structures Before the Transition

At time 1, participants were asked to specify freely 12 elements of identity content using a shortened version of the Twenty

Statements Test (Kuhn & McPartland, 1954). This test has been used in the past as a measure of self-esteem, and in that

case its validity has been widely criticized (see Wylie, 1974). However, we used it here only to ask participants to specify

identity content relevant for themselves. This was located at the beginning of the questionnaire and printed on a page

which folded out from the main answer booklet so that participants could see their responses when making the quantitative

ratings described below. Consistent with our holistic definition of identity, responses included a wide variety of identity

content spanning multiple levels of self-representation (e.g., ‘‘creative,’’ ‘‘strong,’’ ‘‘good friend,’’ ‘‘son,’’ ‘‘wife to Alan,’’

‘‘musician,’’ ‘‘human being’’).

Participants then were asked to rate each of their identity elements in relation to four questions. The first question was a

measure of the perceived centrality of each identity element within self-definition, used to assess current subjective identity

structure: ‘‘How much do you see each of the answers you have written as central or marginal to your identity?’’ (perceived

centrality at time 1). After some intervening measures, participants were asked to rate the expected, desired and feared

centrality of each identity element after the transition, to assess their possible future identity structures.Weused the following

threequestions: ‘‘Thinkingaboutyourself in this newphaseofyour life, howmuchdoyouexpect youwill seeeachansweryou

wrote at the beginning as central ormarginal to your identity at this time in the future?’’ (expected centrality at time 1); ‘‘Now,

please thinkabout thepersonyouwouldmost like tobecomeat thispoint in the future: Imagineyour ideal future self! If allyour

hopes were fulfilled, howmuchwould you see each of the answers you havewritten as central or marginal to your identity?’’

(desired centrality at time 1); ‘‘Now, please think about the person you are most afraid of becoming at this point in the future:

Imagine yourmost dreaded future self! If all your fears were realized, howmuchwould you see each of the answers you have

written as central or marginal to your identity?’’ (feared centrality at time 1). Each question was presented at the top of a new

page, with a block of 12 rating scales positioned underneath to refer to each identity element.We used 7-point scales with the

anchors: 1¼ very much marginal, 4¼ intermediate, 7¼ very much central.

Subjective Identity Structure After the Transition

At time2, eachparticipant received a personalized follow-up questionnaire including a copyof the list of identity elements he/

shehadwrittenbefore the transition.At the beginningof the questionnaire, participantswere asked to indicatewhether eachof

their responseswas still true,neededrevising,orwasno longer true inanyway; theywereasked to replaceany responseswhich

were no longer true and to update any which needed revising. Participants were then asked to rate the current perceived

centrality of each identity element, using the same procedure described above. We excluded from the analysis those identity

elements that participants had described as no longer true in any way (405 responses: �15% of the original responses).

Because our primary focus was on the accommodation processes among elements of existing identity content after the

assimilation of a new identity element, we excluded from the analyses 34 (�1%) of the original responses which referred

directly to the forthcoming new identity element (e.g., ‘‘I am a futuremother’’). We also excluded 208 (�8%) of the original

responses which referred to emotions or well-being, in order to avoid overlap with our outcome measures.

Emotional well-being was assessed at both time points using three different measures. Anxiety about the transitionwas

measured using a six-item measure of state anxiety (Marteau & Bekker, 1992), in which participants rated to what extent

six mood adjectives described their feelings about the transition, on a scale from 0 (not at all) to 3 (very much). Symptoms

of depression were assessed using the Center for Epidemiological Studies Depression scale (CES-D; Radloff, 1977), in

which participants rate how often they have experienced each of 20 depressive symptoms during the past week, on a scale

from 0 (less than 1 day) to 3 (5–7 days)2. The Positive And Negative Affect Scale (Watson, Clark, & Tellegen, 1988) was

used to assess both positive and negative affect. Participants rated a series of positive and negative mood adjectives

2To calculate our index of depressive symptoms, we used the continuity approach instead of the discontinuity approach. In other words, we used the meanscore for the 20 items of the CES-D as a continuous variable, and not the recoded score in groups according to the traditional cut point of 16 (where onlyscores of 16 and above are considered diagnostic evidence of depression; Radloff, 1977). We followed this approach because our intent was not todiagnose depression in our participants but to assess changes in depressive symptoms. To assess the appropriateness of using continuous scores in thecurrent analyses, we compared the Fisher-transformed correlations of depressive symptoms with all other individual-level variables among groups ofparticipants scoring above (�) and below the clinical cut-point of 16 at each time point. Across 22 significance tests, just one comparison reached a .05level of statistical significance, as would be expected based on chance variation between the groups. This gave us confidence to use the continuousapproach, as it suggested that these different regions of the CES-D scale were not behaving differently in their relationships with other variables.

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. (2009)

DOI: 10.1002/ejsp

Table 1. Zero order correlations of perceived actual and possible centrality of identity elements (participant-mean centred) at time 1and time 2, and descriptive statistics of perceived actual and possible centrality of identity elements (row ratings) at time 1 and time 2

Variable 1 2 3 4 5

1. Perceived centrality time 2 —2. Perceived centrality time 1 .50�� —3. Expected centrality time 1 .53�� .51�� —4. Desired centrality time 1 .47�� .49�� .65�� —5. Feared centrality time 1 �.01 .00 �.03 �.16�� —

Total sample means 5.26 5.39 5.14 5.36 3.18Total sample standard deviations 1.55 1.54 1.75 1.91 2.20Students means 5.24 5.37 5.11 5.17 3.33Students standard deviations 1.59 1.54 1.80 2.10 2.24Parents means 5.28 5.40 5.16 5.50 3.07Parents standard deviations 1.53 1.54 1.72 1.75 2.16

Note: listwise n¼ 2721. ��p< .001.

Possible selves and well-being

according to how they felt ‘‘on the average,’’ using a scale from 1 (very slightly or not at all) to 5 (extremely). Across the

two time points, reliability coefficients ranged from .75 to .89 in the young adults sample and from .72 to .84 in the new

parents sample.

RESULTS

Predicting Identity Change Within Participants

Table 1 shows descriptive statistics and zero-order correlations between all identity ratings. In relation to perceived centrality

at time 1, it is notable that both expected and desired centrality showed similar mean values—thus, participants did not

typically eitherdesireor expect elementsof their existing identities tobecomeeithermore centralormoremarginalonaverage

across the two transitions; yet, the zero-order correlations of only around .5 indicate that they did expect and desire

considerable changes in the relative centrality of different elements within their identities after the transitions. In contrast,

feared centrality showed a much lower mean value, suggesting that participants tended to fear that elements of their existing

identities on average would be marginalized after the transitions, and a correlation of zero, suggesting that participants on

average feared that the existing structuring of their identity elements would be wiped-out by the transitions. In reality,

perceived centrality at time 2 showed a very similar mean value to time 1, indicating that no wholesale marginalization or

reinforcement of existing identity content had occurred during either transition. Yet the zero-order correlation of .50 between

perceived centrality at the two time points suggests that identity elements had shifted in their relative centrality quite

substantially over the course of the transition: Thus, very considerable identity restructuring had indeed taken place.

We now tested whether participants’ pre-transition expected, desired, and feared possible future identity structures

would be significant predictors of residual change in their subjective identity structures across the two transitions. Given

the nested data structure, with identity elements (level 1) clustered within participants (level 2), we used multilevel

regression analysis (Hox, 2002). Initially, we conducted analyses for the total sample and then we tested possible

interactions for type of transition. To avoid confounding between-participant covariance with the within-participant

relationships we were interested in here, we centered all predictors around participant means (cf. group mean centering:

Hofmann & Gavin, 1998; Kreft, de Leeuw & Aiken, 1995; Raudenbush, 1989).3

In a first step, we entered perceived centrality at time 1 as a predictor of perceived centrality at time 2. Not surprisingly,

participants rated as significantly more central after the transition those elements of identity which they perceived as more

3We used HLM 5.04 to conduct restricted maximum likelihood estimation with convergence criterion of 0.000001 (Raudenbush, Bryk, Cheong &Congdon, 2001). Models estimated using full maximum likelihood showed no substantive differences from those reported here. Since we were interestedto predict within-participant variance and not between-participant variance in perceived centrality, we estimated modeled variance within participants asR2W , based on comparison with a null model (Kreft & de Leeuw, 1998). Note that many authors are critical of the use of R2 estimates in multilevel

modeling warning that they should be treated with caution, or even avoided altogether. While recognizing these caveats, we include these estimates hereas we consider them to provide useful information about effect size, even if they are only approximations.

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. (2009)

DOI: 10.1002/ejsp

Claudia Manzi et al.

central before the transition (b¼ .48, SE¼ .02, p< .001). Time 1 perceived centrality accounted for an estimated 22.6% of

within-participants variance in time 2 perceived centrality.

In a second step, we added expected, desired, and feared possible future identity structures to the model. Time 1

perceived centrality remained a significant predictor (b¼ .26, SE¼ .02, p< .001), but both expected (b¼ .26, SE¼ .02,

p< .001) and desired (b¼ .10, SE¼ .02, p< .001) possible future identity structures made significant additional

contributions to the model. This model accounted for an estimated further 11.0% of within-participants variance in time

2 perceived centrality, compared to the preceding model. Thus, it appeared that pre-transition possible identity structures

were accounting for identity change among these participants: Individuals tended to approach their pre-existing expected

and desired possible identity structures. In contrast, the contribution of feared possible identity structure was close to zero

(b¼ .01, SE¼ .012, p¼ .60), and thus did not support our prediction that individuals would tend to move away from their

feared possible identity structures.

To test for heterogeneity in these findings between and within our two samples, we computed various models including

dummy variables for subgroups as level 2 predictors and as cross-level interaction terms. A significant cross-level

interaction would mean that the weight of time 2 perceived centrality on one of the predictors in our model differed

significantly across the two samples, or across two subgroups within a sample. Specifically, we tested for interaction

effects of type of transition across the entire dataset, for effects of country and gender within each sample, and for

differences between those who had planned the baby and those reporting unplanned pregnancies in the transition to

parenthood sample. None of these effects was significant. Thus, the results appeared to be consistent across the two

different life-transitions, and among these various subgroups within our samples.

Predicting Individual Differences in Emotional Well-being

Our second set of analyses was designed to test whether post-transition emotional well-being would be predicted by

individual differences in the extent to which participants’ post-transition identity structures matched their pre-transition

desires and expectations and diverged from their pre-transition fears. To do this, for each participant we computed distance

scores to represent the divergence of their post-transition subjective identity structures from their pre-transition desired,

feared, and expected future identity structures, and from their actual pre-transition subjective identity structures. These

scores were calculated, respectively, as the average absolute difference of the time 2 perceived centrality of a participant’s

identity elements from their time 1 desired, feared, and expected possible future centrality and from their time 1 perceived

centrality. Theoretically, these scores could range from 0 (if the time 1 ratings of every identity element named by that

participant exactly matched their time 2 perceived centrality ratings) to 6 (if every single identity element had moved from

one end-point of the perceived centrality scale to the other when comparing the two measures). Hence, to measure

achievement of expected, desired and feared identity structures, we reversed the relevant distance scores by subtracting

them from 6, so that a score of 6 would mean that that participant’s time 2 subjective identity structure exactly matched

their pre-transition expectations, desires or fears. We used the distance score from time 1 perceived centrality as a measure

of the overall magnitude of identity change.4

Note that, because these measures rely on absolute differences calculated at the level of each identity element, they

provide a more comprehensive operational definition of identity change and of convergence between actual and possible

identity structures than was possible in the preceding analyses. For example, the measure of identity change is affected not

only by differences in the rank ordering of identity elements for perceived centrality at time 1 and time 2, but also by any

change—upward or downward—in the perceived centrality of an identity element across the transition.

Table 2 shows means and zero-order correlations of these measures and of our measures of emotional well-being at

each time point. Participants undergoing both transitions on average came slightly closer to their expected (M¼ 4.88) than

4We estimated the reliabilities of these aggregate scores using HLM. Note that these measures were derived from each participant’s ratings of identityaspects referring to a wide range of different content domains, levels of self-representation and social contexts, and so there was no a-priori reason toexpect internal consistency—to assume that people’s desires, fears, expectations or actual change processes would be similar across these many differentaspects of identity. Nevertheless, the reliability estimates ranged from .53 to .83. In theory, we might have increased these values by basing them on alarger number of identity aspects listed by each participant. Yet, previous studies using the Twenty Statements Test methodology have suggested that themeaningfulness of responses tends to tail off when more than around 10 responses are requested (e.g. Bochner, 1994). Hence, using a larger number ofidentity aspects might have increased the statistical reliability of these scores, but at the likely cost of reducing their conceptual validity as indicators ofwhat was happening on average across the main aspects of identity which were meaningful to our participants (see Cronbach, 1990, for an interestingdiscussion of trade-offs between reliability and validity in psychometric measures).

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. (2009)

DOI: 10.1002/ejsp

Table 2. Zero order correlations, means and standard deviations of individual-level variables

Variables 1 2 3 4 5 6 7 8 9 10 11 12 13 14

1. Anxiety t2 —

2. Depressive symptoms t2 .63�� —

3. Positive affect t2 �.23� �.34�� —

4. Negative affect t2 .60�� .55�� �.18�� —

5. Anxiety t1 .41�� .36�� �.20��� .34�� —

6. Depressive symptoms t1 .44�� .64�� �.28�� .51�� .59�� —

7. Positive affect t1 �.15� �.24�� .60�� �.18�� �.37�� �.34�� —

8. Negative affect t1 .36�� .39�� �.17� .60�� .49�� .65�� �.30� —

9. Expected achievement �.10 �.07 .22�� �.07 �.09 �.08 .09 �.07 —

10. Desired achievement �.11 �.31�� .27�� �.08 �.03 �.12 .16� �.09 .42�� —

11. Feared achievement .10 .12 .02 .18�� .01 .07 .05 .06 �.04 �.08 —

12. Identity change .01 .12 �.01 .08 .06 .09 �.07 .08 �.26�� �.39�� �.04 —

13. Country .17� .18�� .00 .12 .19�� .13 .15� .13� .03 .03 .21�� �.06 —

14. Gender .02 .08 .05 .02 .00 .12 .03 .02 .10 .09 �.05 .03 �.14� —

15. Type of transition �.25�� �.31�� .04 �.29�� �.36�� �.38�� .16� �.24�� .07 .19�� �.05 �.12 �.23�� �.19�

Total sample’s mean .93 .77 3.40 1.87 1.09 .79 3.48 2.03 4.88 4.71 3.25 1.05

Total sample’s SD .53 .27 .62 .59 .63 .25 .52 .69 .55 .59 1.27 .53

Adolescents’ mean 1.10 0.84 3.37 2.09 1.38 0.90 3.37 2.24 4.82 4.56 3.32 1.13

Adolescents’ SD .55 0.44 0.65 0.64 0.65 0.48 0.58 0.66 .48 .57 1.19 .48

Parents’ mean 0.83 0.56 3.42 1.74 0.91 0.56 3.55 1.91 4.91 4.80 3.20 .99

Parents’ SD .49 .39 .60 .52 .55 .31 .48 .67 .59 .59 1.31 .56

�p< .05; ��p< .001.

Possible selves and well-being

to their desired (M¼ 4.71) identity structures (t¼ 4.13, df¼ 226, p< .001). More strikingly, participants on average came

substantially closer to both expected and desired identity structures than to their feared identity structures (M¼ 3.25; both

t� 15.24, df¼ 224, p< .001). This probably reflects the fact, noted in the previous section, that participants generally

seemed to fear a wholesale marginalization of their existing identity content—which did not happen. It is also notable that

new parents on average came significantly closer to their desired identity structures (M¼ 4.80) than did new university

students (M¼ 4.56; t¼ 2.97, df¼ 225, p< .01).

Among our well-being measures, in the students samplewe observed a significant average decrease over time in anxiety

(t¼ 3.60, df¼ 85, p< .001; time 1 M¼ 1.38, time 2 M¼ 1.10) and in negative affect (t¼ 2.37, df¼ 85, p< .05; time 1

M¼ 2.24, time 2 M¼ 2.09). In the parents sample we found a significant decrease in both positive affect (t¼ 3.05,

df¼ 140, p< .01; time 1 M¼ 3.55, time 2 M¼ 3.42) and negative affect (t¼ 3.33, df¼ 140, p< .01; time 1 M¼ 1.91,

time 2 M¼ 1.74).

We now conducted a path analysis using the EQS software package, to test the role of our measures of the achievement

of expected, desired, and feared identity structures as predictors of our four measures of emotional well-being at time 2,

controlling for time 1 values of emotional well-being and for individual differences in the overall magnitude of identity

change. First, we screened the data for univariate and multivariate outlying cases and to check the variable distributions for

normality. Two variables, depressive symptoms at time 1 and depressive symptoms at time 2, showed mild positive skew.

This was corrected using square root transformation. We found no other problems of skew or kurtosis, and normalized

estimates of Mardia’s coefficient for multivariate kurtosis were acceptable.

As a first step, we analyzed a model in which each of the four measures of emotional well-being at time 2 was predicted

by the same emotional well-being measure at time 1 and by the four distance scores. The model showed optimal goodness

of fit indices: x2(28, N¼ 229)¼ 41.751, p¼ .05; NFI¼ .96; CFI¼ .98; RMSEA¼ .05 (90% confidence interval: .00–.08).

In this model, achievement of desired identity structure and achievement of expected identity structure were significant

predictors of positive affect (for achievement of desired identity structure b¼ .13, p< .01; for achievement of expected

identity structure b¼ .13, p< .01), whereas achievement of feared identity structure was a significant predictor of negative

affect (b¼ .14, p< .01).

As a second step, we performed three sets of multisample path analyses to test the equivalence of relationships among

variables, first in the different types of transition, then in the different countries, and finally betweenmales and females. We

started by computing a fully constrained model, fixing all structural paths between the variables to be equal for the parents

and the students samples. This model showed a significant lack of fit in the chi square index: x2(76, N¼ 229)¼ 101.098,

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. (2009)

DOI: 10.1002/ejsp

Claudia Manzi et al.

p< .01; NFI¼ .90; CFI¼ .97; RMSEA¼ .04 (90% confidence interval: .01–.06), suggesting the presence of significant

differences in the model between the two transitions. Modification indices revealed that there was just one significant

difference between the parents and the students samples: The path from identity change to anxiety at time 2 was negative

and significant in the students sample (b¼�.30, p< .01) but non-significant for the parents. The constraint on this path

was relaxed and the model was re-estimated. All fit indices were now acceptable: x2(75, N¼ 229)¼ 90.152, p> .10;

NFI¼ .91; CFI¼ .98; RMSEA¼ .03 (90% confidence interval: .00–.05). Then we performed a fully constrained model,

fixing all structural paths between the variables to be equal for the UK and Italian samples. This model showed optimal fit

indices: x2(76, N¼ 229)¼ 80.824, p> .05; NFI¼ .92; CFI¼ .99; RMSEA¼ .02 (90% confidence interval: .00–.04),

indicating that a single identical model fit the two samples. Finally we performed a fully constrained model to assess

differences between males and females. This model showed poor fit indices: x2(75, N¼ 229)¼ 116.324, p< .01;

NFI¼ .86; CFI¼ .94; RMSEA¼ .06 (90% confidence interval: .04–.08). Surprisingly, the modification indices did not

reveal any significant difference between the two samples, but the Lagrange Multiplier Test for adding parameters showed

the presence in the male sample of three new paths. Two of these were cross-over effects between well-being measures

over time: Time 1 negative affect predicted time 2 depressive symptoms and time 1 anxiety predicted time 2 positive

affect. The third was a significant path from time 1 depressive symptoms to identity change: Men who experienced greater

depressive symptoms at time 1 reported less identity change. These three paths were added in the model that now showed

optimal fit indices: x2(72, N¼ 229)¼ 92.517, p¼ .05; NFI¼ .89; CFI¼ .97; RMSEA¼ .04 (90% confidence interval:

.00–.08).

Finally, only within the parents sample we explored for differences between those who had planned pregnancies and

those who had unplanned pregnancies. The fully constrained multisample path analysis revealed poor fit indices: x2(76,N¼ 143)¼ 127.917, p< .01; NFI¼ .80; CFI¼ .90; RMSEA¼ .07 (90% confidence interval: .05–.09). Modification

indices revealed three significant differences between the samples: The path between the depressive symptoms at time 1

and depressive symptoms at time 2, the path between achievement of desired identity structure and depressive symptoms

at time 2, and the path between positive affect at time 1 and positive affect at time 2. Constraints on these paths were

relaxed and the model was re-estimated. All fit indices were now acceptable: x2(73, N¼ 143)¼ 101.627, p> .01;

NFI¼ .84; CFI¼ .94; RMSEA¼ .05 (90% confidence interval: .02–.07). This final model showed that the path between

achievement of desired identity structure and depressive symptoms was significant only for parents with unplanned

pregnancies. For the same participants the impact of positive affect and depression before the transition on positive affect

and depression after the transition was higher. We further investigated these results, analyzing whether there were

significant gender differences in the unplanned pregnancy parents. We used regression analysis because the number of

participants did not allow us to proceed with multisample path analyses. A significant 3-way interaction was found

between gender, planned pregnancy, and achievement of desired identity structure predicting depressive symptoms

(b¼ .16, p< .05). A split file post hoc regression revealed that the path between achievement of desired identity structure

and depressive symptoms was significant only for mothers who did not plan their pregnancy.

Figure 1 synthesizes the results of these path analyses. To simplify, only significant predictive paths are represented. In

all models, covariances between predictor variables were all significant except those with achievement of feared identity

structures, covariances between well-being variables at time 1 were all significant, covariances between well-being

variables at time 2 were all significant.

DISCUSSION

In this study, we sought to map out the relations between possible selves, identity change, and well-being in the context of

two naturally occurring life-transitions. In particular we aimed to analyze the role of possible selves in the restructuring of

existing identity content when a new identity element is assimilated. To our knowledge no previous studies have focused

on the role of possible selves in the accommodation which accompanies the assimilation of a new identity element. To

reach this purposewe developed the concept of possible future identity structure, i.e., the individual’s expectations, desires

and fears about the possible restructuring of his/her identity content will have after an important life transition.

Participants’ initial desired and expected possible future identity structures were reliable prospective predictors of residual

change in their subjective identity structures across both transitions. Individual differences in the extent to which

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. (2009)

DOI: 10.1002/ejsp

Figure 1. Results of the path analyses. Numbers are standardized regression weights. ��p< .01

Possible selves and well-being

participants achieved their desired and feared possible identity structures after the transition were significant predictions of

well-being.

Possible Selves and Identity Change

Our first aim was to investigate whether the restructuring of existing identity content is related to pre-existing possible

future identity structures. Consistent with the proposed role of expected and desired possible future identity structures in

guiding long-term identity change, those aspects of identity which participants initially expected would be more central

after the transition did indeed on average come to be perceived as more central at time 2; similarly, those aspects of identity

which participants initially desired would be more central after the transition on average came to be perceived as more

central at time 2.

Of course, a plausible and simple alternative explanation of the former result is that participants may have tended to

forecast accurately the changes which would occur in their identities as a result of each transition. To the extent that people

make accurate forecasts, expected possible identity structures by definition will be correlated with post-transition identity

structures. Yet, this commonsense account is not necessarily the full story, despite its intuitive appeal. Previous studies

have suggested that expected possible future selves are more than just neutral ‘‘predictions’’ of the future—that they have

a motivational function as goals for behavior, not unlike desired selves (e.g., Oyserman & Markus, 1990). Moreover,

expected selves may influence identity change through self-verification processes, by which people structure their social

interactions and their cognitive processing in such as way as to bring social reality in line with their expectations (Swann,

1987; see also Kerpelman & Pittman, 2001). Nevertheless, a strong test of the causal role of expected possible future

identity structures in identity change would require the use of controlled experimentation, where expectations were

manipulated rather than measured, as opposed to the naturalistic approach used here.

In contrast, it is harder to conceive of an alternative interpretation of the prospective effect of desired possible identity

structures on post-transition subjective identity structures. The results indicate that participants, in restructuring their

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. (2009)

DOI: 10.1002/ejsp

Claudia Manzi et al.

identities, typically approached their desired possible identity structures during both transitions. Although one can

imagine that participants may have moderated their desires to some extent to reflect what they believed would be likely or

possible after the forthcoming transitions, the significant effect of desired possible identity structures cannot reasonably be

explained in terms of accurate forecasting given that this effect was found whilst controlling for expectations. In contrast, it

seems more plausible to assume that it is because these structures were desired that they were approached—that they

functioned as goals for identity change.

We found no significant effects of feared future identity structures on the direction of identity change. Against our

prediction, those identity elements accorded greater centrality in participants’ initial feared future identity structures did

not come to be perceived as significantly less central in post-transition subjective identity structures. We should

acknowledge several potential interpretations of this null finding. One possibility is that participants, on average, did not

succeed in avoiding their feared selves to the same extent that they succeeded in achieving desired and expected selves.

Another possibility is that the null finding does represent at least partial success in avoiding the feared self—in the sense

that at least post-transition subjective identity structures bore no relation to participants’ initial fears. A further possibility,

discussed in more detail subsequently, is that our measurement of feared possible future selves may have suffered from the

focus on current identity content.

Achievement of Possible Future Selves and Well-being

Our second aim was to examine whether variation in the extent to which participants had approached their desired and

expected possible future identity structures and avoided their feared ones would be associated meaningfully with

emotional well-being after each transition. Results showed that the closeness of participants’ post-transition identity

structures to their initial possible future selves did predict residual change in emotional well-being, quite consistently

across the two transitions.

In particular, we found that individuals whose achieved subjective identity structures—after the identity

accommodation process following a significant life transition—were closer to their initial desires and expectations

and further from their initial fears tended to report lower negative affect and higher positive affect after the transition.

These results were invariant across gender and countries considered. However, more detailed analyses showed important

variation within the new parents sample in the magnitude of these effects. We found that the negative path from

achievement of desired identity structures to depressive symptoms was significant only following unplanned

pregnancies—a situation which arguably could be expected to put people in a relatively self-protective frame of mind.

Further analysis showed that the effect on depressive symptoms was significant only among mothers who had not

planned to have children. A post-hoc comparison showed that mothers with unplanned pregnancy reported the highest

levels of these symptoms among the new parents. A 2 (country)� 2 (gender)� 2 (planning) ANCOVA of parents’ post-

transition depressive symptoms, controlling for pre-transition levels, showed a significant gender� planning interaction,

F(1, 134)¼ 4.87, p< .05. Following unplanned pregnancies, new mothers showed a mean CES-D score of 18.4 (rescaled

to fit the 0–60 scoring used in clinical settings), which exceeded the conventional cut-off of 16 for diagnosing mild to

moderate depression; scores were lower among new mothers following planned pregnancies (M¼ 10.5) and among new

fathers (unplanned M¼ 10.2; planned M¼ 9.5). Presumably, mothers with unplanned pregnancies would be especially

likely to experience a loss of control over their lives—a known antecedent of depression (Seligman, 1975); however,

participants in this group who managed to approach their desired possible selves during the transition did not appear to

show poorer emotional well-being than did mothers who had planned their pregnancies. Hence, in contrast with the

prevailing medical model of this condition, these findings are consistent with social psychological arguments that identity

changes associated with motherhood are an important antecedent of postpartum depression (e.g., Mauthner, 1999). This

suggests the possibility, albeit speculative, that incidence of postpartum depression might be reduced by helping expecting

mothers with unplanned pregnancies to formulate achievable desired possible future identity structures.

Magnitude of Identity Change

Banaji and Prentice (1994) argued that enduring identity change occurs especially—perhaps even only—during

significant life-transitions. Here, we found that subjective identity structures at time 1 were a relatively weak predictor of

subjective identity structures at time 2, suggesting that substantial change had occurred between times of measurement.

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. (2009)

DOI: 10.1002/ejsp

Possible selves and well-being

Vignoles et al. (2006, Study 4) reported somewhat greater stability of subjective identity structures over a 2-month time

period without intervening life-transitions. Hence, the extent of discrepancy here between times 1 and 2 subjective identity

structures cannot be explained by short-term contextual fluctuation, which also would have affected their results.

Nevertheless, future research should compare the extent of change across life-transitions with that experienced by people

without a life-transition over a similar length of time.

It is interesting to note that the magnitude of identity change appears to have been experienced slightly differently by

different groups, as evidenced by different relationships with our measures of post-transition well-being. Irrespective of its

relation to specific desires, fears or expectations, identity change was negatively related with anxiety only after the transi-

tion fromschool touniversity. Itmaybe that the experience of identity change is interpreted as indicating for theseparticipants

that one has achieved a new status—as young adult—irrespective of the specific changes which are made.

Limitations and Future Directions

In this study, data were collected at two time points only. A design including several time points both before and after each

transition, would have allowed us to investigate in greater detail the unfolding processes by which possible future selves

are formed and revised both in anticipation of, and following, the moment of the transition itself. Additionally, such a

design would have allowed us to highlight changes in subjective identity structure at the point of transition against the

background of more gradual changes which may be occurring normally. Nevertheless, we reiterate that this is the first

study ever to have examined the role of possible future identity structures in the restructuring of identity following a long-

term identity change; moreover, no previous study has examined longitudinally the emotional correlates of actually

achieving or avoiding one’s expectations, desires, and fears related to the identity accommodation process after a

significant life transition.

This study is unusual in the possible selves literature in focusing on the possible and actual futures of aspects of current

identity content. Thus, our study captures important forms of possible and actual identity change not typically accessed by

traditional approaches to possible selves (see Packard & Conway, 2006), but does not capture changes relating to the

possible or actual attainment of new identity content. This may especially have limited our measurement of feared future

identity structures. Perhaps, the defining feature of feared selves is the possibility of attaining new, feared identity

elements, rather than any particular reconfiguration of existing identity content. Indeed, the main feature of participants

ratings of feared possible identity structures appeared to be a wholesale marginalization of aspects of existing identity,

rather than any particular reconfiguration or reordering. Hence, it is perhaps unsurprising that our measure of feared

possible identity structures did not predict subjective identity structures after the transition. Nevertheless, it is interesting

to note that we did find some effects of feared self-distance on emotional well-being. This suggests that our measurement

of feared selves was not entirely invalid.

Caution must be exercised in drawing causal inferences from these data. We have already noted an alternative

interpretation of the path from pre-transition expected possible identity structures to post-transition actual identity

structures. Similarly, we cannot be certain about the causal processes underlying paths from identity measures to post-

transition well-being. Controlling for pre-transition levels of well-being rules out some alternative interpretations—in

particular, our results cannot be attributed to the influence of pre-existing levels of anxiety, depressive symptoms, or

positive and negative affectivity on the trajectories of identity change toward or away from desires, fears, or expectations.

Nevertheless, the possibility remains that some further unmeasured variable(s) could have influenced trajectories of both

identity and emotional well-being during these transitions.

Caution is also necessary in generalizing from these results, given that our samples were not representative of people

making the transition to university or the transition to parenthood. We explored for gender and national differences in our

analyses, suggesting that our findings were not biased by differences between the samples in these effects. Nevertheless, in

the absence of truly random sampling, we cannot rule out the possibility that our participants differed systematically in

some unknown way from the wider population of adolescents starting university and adults becoming parents.

Additionally, it would be valuable to examine these processes across a wider range of cultures.

Nevertheless, the current study provides an important step forward in the study of possible selves and long-term identity

change. In the real-world context of two significant life-transitions, we have shown not only that possible selves are

meaningfully related to enduring change in subjective identity structures but also that individual differences in these

relationships are significantly associated with emotional adjustment after transition. Of course, we have been treading new

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. (2009)

DOI: 10.1002/ejsp

Claudia Manzi et al.

theoretical and methodological ground in this study, and so the current findings provide just a ‘‘first look’’ at these

previously uncharted phenomena. The current study provides an overview of the domain, which should pave the way for

more fine-grained research—including controlled experimentation as well as further longitudinal studies in naturalistic

settings—to provide a more detailed picture of the causal influences and mediating social psychological processes

underlying the general pattern of findings we have observed here. In this way, we hope that subsequent research will build

on these findings and will shed further light on the causes and the consequences of enduring identity change.

REFERENCES

Banaji, M. R., & Prentice, D. A. (1994). The self in social contexts. Annual Review of Psychology, 45, 297–332.Bochner, S. (1994). Cross-cultural differences in the self-concept: A test of Hofstede’s individualism/collectivism distinction. Journal ofCross-Cultural Psychology, 25, 273–283.

Breakwell, G. (1986). Coping with threatened identities. London: Methuen & Co.Burke, P. J. (1991). Identity processes and social stress. American Sociological Review, 56, 836–849.Carver, C. S., Reynolds, S. L., & Scheier, M. F. (1994). The possible selves of optimists and pessimists. Journal of Research inPersonality, 28, 133–141.

Cassidy, C., & Trew, K. (2004). Identity change in Northern Ireland: A longitudinal study of students’ transition to university. Journal ofSocial Issues, 60, 523–540.

Cigoli, V., & Scabini, E. (2006). The family identity. Ties, symbols and transitions. Mahwah: Erlbaum.Cowan, C. P., & Cowan, P. A. (1992).When partners become parents: The big life change for couples. New York, NY, US: Basic BooksInc.

Cronbach, L. J. (1990). Psychological tests and personnel decisions. Urbana: University of Illinois Press.Cross, S. E., & Markus, H. R. (1991). Possible selves across the life span. Human Development, 34, 230–255.Deaux, K. (1992). Personalizing identity and socializing self. In G. M. Breakwell (Ed.), Social psychology of identity and the self-concept (pp. 9–34). London: Surrey University Press.

Dunkel, C. S., Kelts, D., & Coon, B. (2006). Possible selves as mechanisms of change in therapy. In C. Dunkel, & J. Kerpelman (Eds.),Possible selves: Theory, research and applications (pp. 187–204). Hauppage, NY: Nova Science Publishers.

Ethier, K. A. (1996). Becoming a mother: Identity acquisition during the transition to parenthood. Dissertational AbstractsInternational, 56, 5832.

Ethier, K. A., & Deaux, K. (1994). Negotiating social identity when contexts change: Maintaining identification and responding tothreat. Journal of Personality and Social Psychology, 67, 243–251.

Granberg, E. (2006). ‘‘Is that all there is?’’ Possible selves, self-change, and weight loss. Social Psychology Quarterly, 69, 109–126.Higgins, E. T. (1987). Self-discrepancy: A theory relating self and affect. Psychological Review, 94, 319–340.Hofmann, D. A., & Gavin, M. B. (1998). Centering decisions in hierarchical linear models: Implications for research in organizations.Journal of Management, 24, 623–641.

Hox, J. J. (2002). Multilevel analysis. Techniques and applications. Mahwah: Lawrence Erlbaum Associates.Kerpelman, J. L., & Lamke, L. K. (1997). Anticipation of future identities: A control theory approach to identity development within thecontext of serious dating relationships. Personal Relationships, 4, 47–62.

Kerpelman, J., & Pittman, J. F. (2001). The instability of possible selves: Identity processes within adolescents’ close peer relationships.Journal of Adolescence, 24, 491–512.

King, L., & Raspin, C. (2004). Lost and found possible selves, subjective well-being, and ego development in divorced women. Journalof Personality, 73, 603–632.

Knox, M., Funk, J., Elliott, R., & Bush, E. G. (1998). Adolescents’ possible selves and their relationship to self-esteem. Sex Roles, 39,61–80.

Kreft, I., & de Leeuw, J. (1998). Introduction multilevel modeling. London: Sage.Kreft, I., de Leeuw, J., & Aiken, L. S. (1995). The effect of different forms of centering in hierarchical linear models. MultivariateBehavioral Research, 30, 1–21.

Kuhn, M. H., & McPartland, T. S. (1954). An empirical investigation of self-attitudes. American Sociological Review, 19, 68–76.Manzi, C., Vignoles, V. L., Regalia, C., & Scabini, E. (2006). Cohesion and enmeshment revisited: Differentiation, identity, andpsychological well-being in two European cultures. Journal of Marriage and Family, 68, 673–689.

Markus, H., & Nurius, P. (1986). Possible selves. American Psychologist, 41, 954–969.Markus, H., Nurius, P., (1987) Possible selves: The interface between motivation and self-concept. In K. Yardley, T. Honess, (Eds.), Selfand identity: Psychosocial perspectives (pp. 157–172). Oxford: John Wiley& Sons.

Marteau, T. M., & Bekker, H. (1992). The development of a six-item shortform of the state scale of the Spielberger State-Trait AnxietyInventory (STAI). British Journal of Clinical Psychology, 31, 301–306.

Mauthner, N. S. (1999). ‘‘Feeling low and feeling really bad about feeling low’’: Women’s experiences of motherhood and postpartumdepression. Canadian Psychology, 40, 143–161.

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. (2009)

DOI: 10.1002/ejsp

Possible selves and well-being

Oyserman, D., & Markus, H. (1990). Possible selves in balance: Implications for delinquency. Journal of Social Issues, 46, 141–157.Packard, B. W.-L., & Conway, P. F. (2006). Methodological choice and its consequences for possible selves research. Identity, 6, 251–271.

Penland, E., Masten, W., Zelhart, P., Fournet, G. P., & Callahan, T. (2000). Possible selves, depression and coping skills in universitystudents. Personality and Individual Differences, 29, 963–969.

Piaget, J. (1936). The origins of intelligence in children. New York: Basic Books.Radloff, L. S. (1977). The CES-D Scale: A self-report depression scale for research in the general population. Applied PsychologicalMeasurement, 1, 385–401.

Raudenbush, S. (1989). ‘‘Centering’’ predictors in multilevel analysis: Choices and consequences. Multilevel Modelling Newsletter, 1,10–12 (Available from Multilevel Modelling Project, Institute of Education, 20 Bedford Way, London WC1H 0AL, UK.).

Raudenbush, S., Bryk, A., Cheong, Y. F., & Congdon, R. (2001). HLM 5: Hierarchical linear and nonlinear modeling. Lincolnwood:Scientific Software International.

Rosenberg, S., & Gara, M. A. (1985). The multiplicity of personal identity. Review of Personality and Social Psychology, 6, 87–113.Seligman, M. E. P. (1975). Helplessness: On development, depression and death. New York: W.H. Freeman & Co.Stryker, S., & Serpe, R. T. (1994). Identity salience and psychological centrality: Equivalent, overlapping, or complementary concepts?Social Psychology Quarterly, 5, 16–35.

Swann, W. (1987). Identity negotation: Where two roads meet. Journal of Personality and Social Psychology. Special Issue: Integratingpersonality and social psychology, 53, 1038–1051.

Vignoles, V. L., Manzi, C., Regalia, C., Jemmolo, S., & Scabini, E. (2008). Identity motives underlying desired and feared possiblefuture selves. Journal of Personality, 76, 1165–1200.

Vignoles, V. L., Regalia, C., Manzi, M., Golledge, J., & Scabini, E. (2006). Beyond self-esteem: Influence of multiple motives onidentity construction. Journal of Personality and Social Psychology, 90, 308–333.

Watson, D., Clark, L., & Tellegen, A. (1988). Development and validation of a brief measure of positive and negative affect: The PANASscale. Journal of Personality and Social Psychology, 56, 1063–1070.

Werner, O., & Campbell, D. T. (1973). Translating, working through interpreters, and the problem of decentering. In R. Naroll, & R.Cohen (Eds.), A handbook of method in cultural anthropology (pp. 398–420). Mahwah, NJ: Lawrence Erlbaum Associates.

Wurf, E., & Markus, H., (1991) Possible selves and the psychology of personal growth. In D. J. Ozer, J. M. Healy, A. J. Stewart, (Eds.),Perspectives in personality, Vol 3: Part A: Self and emotion; Part B: Approacches to understading lives (pp. 39–62). London: JessicaKingsley Publishers.

Wylie, R. C. (1974). The self-concept. Lincoln, NB: University of Nebraska.Yowell, C. M. (2002). Dreams of the future: The pursuit of education and career possible selves among ninth grade Latino youth.Applied Developmental Science, 6, 62–72.

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. (2009)

DOI: 10.1002/ejsp