Linking personality states, current social roles and major life goals
Transcript of Linking personality states, current social roles and major life goals
European Journal of Personality
Eur. J. Pers. 23: 509–530 (2009)
Published online 23 June 2009 in Wiley InterScience
(www.interscience.wiley.com) DOI: 10.1002/per.731
*2
C
Linking Personality States, Current Social Rolesand Major Life Goals
WIEBKE BLEIDORN*
Department of Psychology, Bielefeld University, Germany
Abstract
Employing an experience-sampling design, the interplay between personality states, social
roles and major life goals was examined as it unfolds in the stream of people’s daily lives.
Multilevel analyses revealed a considerable amount of both within- and between-person
variability in state expressions of personality traits justifying further examination of
predictors at both levels of analyses. Roles proved as predictors of current personality
states albeit effects differed significantly between individuals. Life goals accounted for
between-person differences in average personality states but were not effective in pre-
dicting differences in relations between personality states and roles. Altogether, findings
testify to the viability of the employed research strategy to analyse the interplay between
both dispositional and fluctuating influences on individuals’ trait expressions in behaviour.
Copyright # 2009 John Wiley & Sons, Ltd.
Key words: Five-Factor Model of Personality; multilevel analysis; social roles; goals;
experience-sampling
INTRODUCTION
The longstanding boundaries between trait and process approaches to personality have
blurred in the service of building up ‘an integrative science of the person’ (Mischel, 2004;
p. 1), since both sides have widely conceded that intra- and interindividual differences are
not mutually exclusive (e.g. Fleeson, 2004; Funder, 2001; Mischel, 2004). Structures and
processes should be considered as two sides of the same system that are mathematically
independent but probably interrelated to produce both consistency as well as variability in
behaviour (Fournier, Moskowitz, & Zuroff, 2008; Mischel & Shoda, 1998). According to
Funder (2001), a comprehensive study of personality should thus encompass all three
elements of the ‘personality triad’: The person, the situation and the behaviour, as
knowledge about any two of these should lead to an understanding of the third.
Correspondence to: Wiebke Bleidorn, Department of Psychology, Bielefeld University, Universitatsstrasse5, 33615 Bielefeld, Germany. E-mail: [email protected]
opyright # 2009 John Wiley & Sons, Ltd.
Received 26 November 2008
Revised 5 May 2009
Accepted 5 May 2009
510 W. Bleidorn
However, the post-war period of the person-situation debate is still characterized by an
imbalance in the study of this triad (Funder, 2001). On the one hand, a lot of research has
been done to identify variables defining stable differences between individuals in terms of
cross-situational consistencies. One of the most popular models among these attempts is
the Five-Factor Model of Personality (FFM; McCrae & John, 1992) assuming that the
structure of personality could be comprehensively described through five dimensions,
namely Neuroticism (N), Extraversion (E), Openness to experience (O), Agreeableness (A)
and Conscientiousness (C). On the other hand, comparative little is empirically known
about the dynamic interplay of personality with situations and behaviours, that is, the
differences within individuals in terms of cross-situational variability in personality
(Funder, 2001, Fournier et al., 2008).
The present study aimed to make a further contribution to the desired integration
between structure and process approaches to personality (e.g. Epstein, 1994; Fleeson,
2001; Mischel & Shoda, 1998). Specifically, employing an experience-sampling design
and multilevel modelling procedures (MLM; e.g. Raudenbush & Bryk, 2002), the
associations of FFM personality states with current social roles and major life goals were
examined as they unfold in the stream of people’s daily lives.
Assessing the five factors as states
A promising way to integrate structure and process approaches to personality has
been proposed by Fleeson (2001). In his density-distributions approach to personality,
behavioural manifestations of traits are described as frequency distributions of their
corresponding states. Emphasizing a certain degree of isomorphism between personality
states and traits, personality states are considered as trait-content manifestations in short-
term, continuous and concrete ways of acting, feeling and thinking that could be described
in the same way as traits (Fleeson, Malanos, & Achille, 2002). That is, just as individuals
can be described by means of their general level in N, E, O, A and C, similarly an
individual’s behaviour can be described as anywhere between low and high in N, E, O, A
and C. This approach does not describe specific actions or movements but is still an
assessment of how the person is behaving with regard to trait expression. Personality states
are thus characterized by both the relative consistency of the corresponding trait as well as
the variability of behavioural acts (Fleeson, 2001, 2007).
Empirical support for this approach has been obtained from a number of experience-
sampling studies in which participants repeatedly rated their current behaviour over a
couple of days by means of adjectives commonly used for describing Big Five traits
(Fleeson, 2001, 2007). Findings suggest that individuals express nearly all levels of all
traits in their everyday behaviour but are also characterized by unique behavioural
frequency distributions for each of the five domains. Despite the sizeable within-person
variability in personality states, the means and standard deviations of density distributions
turned out to be relative stable individual differences characteristics. Thus, by using
adjectives as descriptors of the trait content of behaviour, this approach accounts for both
processes within as well as structural differences between individuals (Fleeson, 2001). This
paves the way for further research into the mechanisms of personality functioning by
focusing on both varying factors that trigger within-person changes in personality states
and enduring dispositions accounting for stable between-person differences in personality
states.
Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Pers. 23: 509–530 (2009)
DOI: 10.1002/per
Personality states, social roles and major life goals 511
Current social roles and personality states
Reconfirming that intraindividual variability in trait-relevant behaviour is meaningful and
predictable, Fleeson (2007) has shown certain situational characteristics, such as task
orientation, to be significant predictors of the within-person variation in personality states.
These systematic ‘situation-based contingencies’ (Fleeson, 2007, p. 825) were not
universal but differed reliably between individuals. This finding is in line with the
Cognitive Affective Personality System approach (CAPS, Mischel & Shoda, 1995, 1998,
1999) assuming individuals to be characterized by distinct if. . .then situation-behaviour
profiles.
Fleeson’s study may be viewed as a first effort to examine the rather abstract hypothesis
that variation in personality states is due to variation in situations, since this work has been
exploratory with regard to the wide range of situational features studied. Further research is
needed to examine more theoretically driven hypotheses about specific contextual
characteristics that may play a functional role in the prediction of personality states. Such
hypotheses are difficult to deduce as there is no consensus regarding the most important
‘psychologically active features of situations’ (Shoda, Mischel, & Wright, 1994, p. 685).
Regarding the plethora of potential predictors, it seems reasonable to start off research
with the focus on one discrete class of contextual characteristics that potentially matter to
trait-relevant behaviour. Recent research related to the assessment of contextualized
personality suggests that the social role a person is fulfiling might be an interesting
candidate that matter to personality differences at both levels within and between persons
(Roberts, 2007; Heller, Watson, Komar, Min, & Perunovic, 2007; Roberts & Pomerantz,
2004; Wood, 2007; Wood & Roberts, 2006). A social role can be defined as a ‘set of
behavioural expectations attached to a position in an organized set of social relationships’
(Stryker, 2007; p. 1083). That is, social roles refer to positions in society that are associated
with specific expectations, goals and behaviours defining the way an individual relates to
his or her environment. A case in point might be the manager who is expected to act in an
assertive, tough-minded way but yet is expected to act very differently in her role as a
mother (Heller et al., 2007).
There are several features of social roles that qualify them as reasonable predictors of
trait-relevant behaviour. First, social roles can be considered as ‘conglomerations of
situations that all share a common thread of expectations and behavioural signatures’
(Roberts, 2007; p. 1073). Hence, in contrast to specific situational settings, such as ‘dorm’,
‘party’ or ‘lecture hall’, social roles are conceptualized at a breadth that is focused enough
to capture important aspects of the situation but not too narrowly defined as it would
diminish their predictive power. Second, since social roles are defined by certain
expectations, they offer a kind of behavioural guidance. As long as people confirm to role
expectations they will be accepted and rewarded by the relevant social group, while failing
to meet the given expectations might lead to negative sanctions (Roberts, 2007).
The links between social roles and personality have been subject to several studies in
which participants were explicitly asked to rate their personality separately across several
different roles by means of standard questionnaires (e.g. Roberts & Donahue, 1994; Wood,
2007; Wood & Roberts, 2006). Using these contextualized ratings, findings suggest that the
role-specific personality (e.g. personality at work) is a better predictor of role-specific
criteria (e.g. job satisfaction) than global personality (Heller et al., 2007).
In a first attempt to detect associations between current social roles and personality
states, Theakston, Heller, Komar, and Lee (2006, as cited in Heller et al., 2007) have shown
Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Pers. 23: 509–530 (2009)
DOI: 10.1002/per
512 W. Bleidorn
that personality states vary across roles in predictable ways. By aggregating experience-
sampling records on personality states within role contexts, the authors have demonstrated
that undergraduates report higher degrees of E, A and O in the friend role compared to
higher degrees of C in the student role. N did not differ significantly between friend and
student roles, although there was a trend of being more neurotic in the student than in the
friend role. However, the employed aggregation technique to reveal mean-level differences
in personality across different roles was not suitable to examine the universality of these
effects. One purpose of the current study was thus to examine both the average within-
person associations of student and friend roles with personality states as well as the
individual differences in these within-person relationships.
Instead of exploring a wide range of contextual features likely to affect personality
states, the current study purposely focused on only two social roles as potential contextual
determinants of within-person variability. It was assumed that student and friend roles
differ in how adaptive or functional different personality states are. Individuals were
expected to adjust their behaviour when the role context changes in order to increase the
adaptiveness of their behaviour (Allport, 1937; Cantor & Fleeson, 1994; Fleeson, 2007;
Mischel, 2004). This is not to argue that social roles are the only or even the most powerful
contextual features that predict trait-relevant behaviour. There are several approaches that
have proposed quite different strategies to link personality with situational features in order
to provide a more comprehensive picture of the situation-dependent nature of personality
(e.g. Bem & Funder, 1978; Denissen & Penke, 2008; Fournier et al., 2008; Saucier,
Bel-Bahar, & Fernandez, 2007). The present study did not intend to sample the domain of
contextual characteristics that matter in terms of their relevance to trait-relevant behaviour
comprehensively. Rather, this research aimed to study specific hypotheses about the
within-person relationships between FFM personality states and the perceived role context
considered as one type of potentially relevant if–then contingencies.
Analyses were intended to examine both within-person effects holding for the average
person as well as between-person differences in these if–then contingencies. The latter
would suggest that individuals react differently to the same contextual demands; for
example, reflecting a differential responsivity to given role demands. Revealing such
Person� Situation interactions (e.g. Endler & Parker, 1992; Magnusson & Endler, 1977),
that is substantial variation in how people act while occupying social roles, would suggest
that personality state expressions in role contexts can be considered as an individual-
differences variable. Extending previous research, a further issue of the current study thus
concerns the exploration of person-based dispositions that may help predict between-
person differences in both average personality states as well as in within-person
relationships between social roles and personality states.
Major life goals and personality states
In search of relevant person-based influences on personality states, a promising approach to
elucidate the dynamics of personality functioning may be to treat persons as active,
dynamic and future-oriented agents who utilize situations to pursue personally relevant
goals (Fleeson & Cantor, 1995). Yielding a conception of the person as agentic and
proactive recognizes that though personality interacts with features of the social world, the
individual is not a helpless victim of the circumstances. Within the framework of the
CAPS, Mischel and Shoda (1995, 1998, 1999) account for the fact that individuals do not
always merely react to situations but rather select, construct and transform them more or
Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Pers. 23: 509–530 (2009)
DOI: 10.1002/per
Personality states, social roles and major life goals 513
less actively. Specifically, individuals are assumed to be characterized by a number of
relative stable dispositions (e.g. beliefs, competencies, goals) that interact with
psychologically relevant features of situations to generate patterns of behaviour.
Following this proactive approach, the current study aimed to examine the links between
long-term goals and personality states in different role contexts. As they provide a person’s
day-to-day activities with structure and meaning, goals were assumed to play a
fundamental role in both driving and guiding individuals’ behaviour (Austin & Vancouver,
1996; Maier & Brunstein, 2001; Pohlmann & Brunstein, 1997). It is generally agreed that
goals can be described along a hierarchical framework in which rather stable higher order
goals embrace several contextualized subgoals which in turn can be broken down into
specific goals relating to immediate actions (Austin & Vancouver, 1996).
Among the higher-order goal concepts, the class of major life goals was chosen for the
purpose of investigating the links between goals, personality states and social roles. Major
life goals can be defined as ‘a person’s aspirations to shape his or her life context and
establish general life structures such as having a career, a family and a certain kind of
lifestyle’ (Roberts, O’Donnell, & Robins, 2004, p. 542). In contrast to more contextualized
mid-level goal units, major life goals have greater generality and are relative stable over
time. The mental representation of the desired status and the actions that help achieve them
should guide behaviour while pursuing major life goals. Hence, individuals were assumed
to adapt their personality states more or less flexibly toward the achievement of their
leading life goals.
There are only a few studies that have aimed at the link between the content of goals and
the Big Five. Little, Lecci, and Watkinson (1992) have examined the links between the Big
Five and ideographically generated personal projects, which are conceptualized at a mid
level of the goal hierarchy. They have found personal projects in the academic domain to be
more related to C, while A and E were stronger associated with interpersonal projects.
More relevant to the present study are the findings reported by Roberts and Robins (2000),
since they have used normative importance ratings of different clusters of major life goals
in order to examine their relationships with the Big Five. They have found C and E to be
positively related to economic goals, such as goals for achievement or wealth. In contrast,
A has been strongly associated with aspirations to ‘get along’ as represented in social goals.
Subsequently, Roberts et al. (2004) have found meaningful links between long-term
changes in goal importance and personality traits over a period of 4 years. For example,
changes in social goals have been associated with changes in E, A and C.
The question arises whether the revealed associations of personality traits with major life
goals can be applied to the stable aspects of personality states, too. In that case, one would
expect different life goals to be associated with the average levels of specific personality
states. Major life goals may also moderate people’s behavioural reactions to different
contexts and, thus, help predict between-person differences in contingencies of personality
states on current social roles. That is, individuals striving strongly for a certain goal should
adapt their trait-relevant behaviour to the demands of a role context which they suppose to
be relevant for achieving this goal.
The present study
Employing an experience-sampling design in conjunction with MLM procedures, the
within- and between-person relationships among personality states, current social roles and
major life goals were examined. Four issues guided this research: First, the utility of the
Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Pers. 23: 509–530 (2009)
DOI: 10.1002/per
514 W. Bleidorn
FFM to describe temporal variation in personality states which is closely related to the
psychometric properties of the personality state measure. That is, FFM personality states
should reliably differentiate within and between persons. Furthermore, the intercorrela-
tional structure of the five states at the within-person level should at least approximate the
structure obtained at the between-person level.
Second, since social roles differ in their immanent expectations, different roles were
expected to provoke different sets of personality states. The current analyses focused on
two roles undergraduates frequently occupy: Student and friend. Without doubt, there are
further roles students take over (e.g. club member), but these are likely to be relevant for
only some subset of students. In contrast, the roles considered in this study were assumed to
encompass two important arenas in which virtually every student strives (Sheldon & Elliot,
2000). Furthermore, these roles represent prototypic exemplars of two broader dimensions
around which most social roles can be organized, namely affiliation and power which are
closely related to the fundamental human needs of belonging and status (Roberts, 2007).
Being a friend is clearly an affiliation role, while the student role is primarily associated to
achievement. Since academic achievement is usually associated with a successful career
and higher status in later life, this role context should be closer related to the dimension of
power. As these two roles have been also employed in other recent research related to the
assessment of contextualized personality (e.g. Roberts & Donahue, 1994; Theakston et al.,
2006, as cited in Heller et al., 2007; Wood & Roberts, 2006), it was possible to deduce
concrete hypotheses about their associations with different personality states: The friend
role was expected to be associated with higher values in E, A and O, while the student role
should be primarily related to higher values in C. Definite hypotheses about the relations
between the two role contexts and N were more difficult to deduce. However, according to
the results reported by Theakston et al. (2006), higher values in N were expected to be
positively related to the student role.
The third issue of this paper concerns the universality of these effects. According to the
CAPS approach (Mischel & Shoda, 1995, 1998, 1999), there should be considerable
differences in individuals’ if–then contingencies suggesting that individuals react
differently to the same role context. In line with previous findings (e.g. Fleeson, 2007),
participants were thus expected to differ substantially in their within-person links between
role contexts and personality states.
Finally, major life goals were expected to predict between-person differences in both
average personality states as well as within-person links between roles and personality
states. The present analyses focused on two different categories of major life goals
hypothesized to have a differential impact on personality states, namely achievement and
affiliation goals. These two goal categories were assumed to be relevant for the role
contexts considered in this study. Just as roles can be organized around the broader
dimensions of affiliation and power, similarly goals can be scaled along these dimensions
that have been also characterized as communion and agency (Bakan, 1966; Pohlmann &
Brunstein, 1997; Sheldon & Cooper, 2008). Achievement goals, representing agency-
orientation, contain aspirations to improve one’s competencies and to broaden one’s mind
(Pohlmann & Brunstein, 1997). In line with previous research into the relations between
personality traits and major life goals, this category was expected to be positively related to
O and C (e.g. Roberts & Robins, 2000). Affiliation goals comprise community-oriented
endeavours and refer to social ambitions, such as being with others or having a lot of friends
(Pohlmann & Brunstein, 1997). Positive associations were primarily expected to be found
with E and A (Roberts & Robins, 2000; Roberts et al., 2004). It should be noted that this is
Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Pers. 23: 509–530 (2009)
DOI: 10.1002/per
Personality states, social roles and major life goals 515
the first attempt to detect cross-level interactions between relative stable dispositions of the
person and within-person links between contextual variables and personality states. Thus,
with regard to the moderating influence of goal content, this study was rather explorative
and intended to examine the more abstract hypothesis that major life goals do moderate
within-person relationships between social roles and personality states.
METHOD
Participants
Participants were 52 undergraduate psychology students (33 females, 19 males) attending
Bielefeld University who participated in partial fulfilment of curricular requirements. All
participants completed the study and could be included in the analyses. Age of participants
ranged between 19 and 28 years (M¼ 22.40, SD¼ 2.44).
Procedure
The study started with an introductory session in which object and procedure of the
experience-sampling method was explained to the participants who also completed a set of
standard personality questionnaires. Each participant received a Palm Z22 handheld
computer programmed with version 4.0 of the Experience-Sampling Program (ESP, Barrett
& Feldman Barrett, 2000) in order to run a first practice trial (this was subsequently deleted
from the master data file).
During the subsequent experience-sampling period, participants were asked to rate their
behaviour and the degree to which they occupied different social roles during the previous
hour, six times per day for 10 consecutive days according to a regular schedule (12, 2, 4, 6,
8 and 10 pm). To ensure compliance and prevent loss of data, participants were instructed
to contact the research office immediately in case of any problems. After five days and after
the whole period of 10 days, all participants came to the research office for data upload,
where they also received feedback about their response rates. If these were below 90%,
participants were invited to extend the sampling period for one further day. At the end of
the study, participants answered a few questions about their participation. Using a 1 to 5
scale ranging from 1¼ not at all to 5¼ very much, participants reported that participating
in the study did not considerably changed their natural behaviour as pointed out by
M¼ 1.72 (SD¼ 0.90). Besides, participants indicated that they would again agree to
participate in such an experience-sampling study suggested by M¼ 4.04 (SD¼ 0.86).
Since reports were automatically date and time stamped, recordings not completed at
the six specified times could be excluded from analyses. The overall response rate was
quite satisfying. Across all participants and assessments, a total of 2962 of 3120 possible
recordings were collected. Across the 52 participants, the mean number of reports was
56.96 (SD¼ 2.04) of 60 possible (95%), with a range of 50 to 60 reports.
Measures
Experience-sampling reports
Following Fleeson’s density-distribution approach (2001), adjectives were used to measure
the five broad dimensions of the FFM as states. In the present study, personality states
Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Pers. 23: 509–530 (2009)
DOI: 10.1002/per
516 W. Bleidorn
should represent the five domains of the Revised NEO Personality Inventory (NEO-PI-R;
Costa & McCrae, 1992) instead of reverting to taxonomical based adjective lists. Initially, a
total of 90 bipolar adjectives were selected from the authorized German manual of the
NEO-PI-R (Ostendorf & Angleitner, 2004) which provides an extensive list of German
adjectives as exemplary descriptors of both opponent poles of the 30 facet scales. Only
those items were chosen that were easily used to describe concrete momentary behaviour.
To reduce the initial item pool to a manageable but reliable set of items suited for
experience-sampling designs, a pre-study was conducted on a sample of 166 undergraduate
students. Using the same instruction and rating scale as scheduled for the experience-
sampling, participants of the pre-study rated their behaviour by means of the 90 bipolar
adjectives in a standard one-occasion questionnaire. That is, respondents were instructed to
describe how they are at the moment rather than what they are like in general. Finally, those
30 bipolar adjectives were chosen that loaded most highly on the correct factor but
negligible on all other factors. As a result, each domain scale of the FFM state measure was
represented by six bipolar items which were standardized as T-scores (M¼ 50, SD¼ 10).
A complete list of the 30 adjective pairs is shown in the Appendix.
During the experience-sampling phase, participants were asked to rate their behaviour
during the previous hour on the basis of these 30 bipolar adjectives by means of a 5-point
bipolar rating scale (e.g. ‘Which of the two terms is better suited for describing your
behaviour during the previous hour: Responsible or irresponsible?’). Besides the 30
adjective pairs, participants also rated the degree to which different roles (student, friend,
romantic partner, employee, family member, club member) predominated their situational
setting during the previous hour by means of a 5-point Likert format ranging from 1¼ not
at all to 5¼ very much. The present analyses were exclusively based on the student and
friend roles which were also the two most frequently reported roles and showed the largest
within-person variation. A dimensional measurement approach was chosen since social
roles are not mutually exclusive, that is, an individual can be a friend at the same time he or
she is a student. Compared to the usual categorical classification of roles, there are further
advantages of the dimensional assessment of role contexts: ‘A dimensional approach
analyzes situations into their psychologically relevant characteristics, treats situations as
having degrees of characteristic, allows situations to have multiple characteristics at once,
studies characteristics independently, and produces situation-state contingencies with
coefficients that have direction and graded magnitude’ (Fleeson, 2007, pp. 829–830).
Major life goals
Achievement and affiliation goals were assessed by GOALS (Pohlmann & Brunstein,
1997), a standard questionnaire administered in the introductory session among other
inventories (not discussed here). GOALS consists of 24 life-goal descriptions pertaining to
six broad domains, namely Power, Variation, Intimacy, Altruism, Achievement and
Affiliation. All items start with ‘I want to. . .’ followed by the particular life goal.
Participants were asked to rate the importance of each of the 24 life goals according to its
relevance to their long-term or lifetime orientation on a 5-point Likert format ranging from
1¼ not important to 5¼ very important. The four items indicating the domain of
achievement were: ‘Broaden my horizons’, ‘develop my skills’, ‘improve my education
continuously’ and ‘continuously improve myself’. Affiliation goals were marked by ‘spend
a lot of time with other people’, ‘be friends with many people’, ‘engage in a lot of activities
with others’ and ‘have a large circle of friends’ (according to translations by Hofer, 2003;
Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Pers. 23: 509–530 (2009)
DOI: 10.1002/per
Personality states, social roles and major life goals 517
Hofer & Chasiotis, 2003). Cronbachs alpha was .83 for achievement and .88 for affiliation
goals, respectively.
Data analysis
As each measurement occasion can be considered nested in a person as higher order level,
MLM procedures were employed using the program HLM (Version 6; Raudenbush, Bryk,
Cheong, & Congdon, 2000). MLM can be understood as a series of nested regressions in
which the coefficients from one level of analysis become the dependent variables at the
next level of analysis (Raudenbush & Bryk, 2002). To address the issues of the current
study, a series of multivariate three-level models were estimated with the five personality
states as multiple dependent variables (Hox, 2002; Raudenbush & Bryk, 2002;
Raudenbush, Rowan, & Kang, 1991; Nezlek, 2007; Snijders & Bosker, 1999). Level 1
represented variation among the item scores within each measurement occasion, level 2
represented variation among occasions within each person and level 3 referred to the
variation among persons. While level 1 served exclusively as a measurement model, levels
2 and 3 may be viewed as a multivariate two-level model for the latent true scores.
This multivariate approach is associated with a number of advantages (e.g. Hox, 2002;
Raudenbush & Bryk, 2002; Snijders & Bosker, 1999). Of particular importance for the
present study, multivariate analyses (a) allow to investigate the structural relationships of
the latent state constructs with predictors at the within- and between-person levels of
analysis simultaneously, (b) avoid the danger of chance capitalization which would arise
when separate univariate analyses were performed, (c) exploit the associations between
personality states to provide more accurate standard errors and more powerful tests of the
within- and between-person effects and (d) allow to examine the psychometric properties
of the FFM state measure at both the within- and between-person level simultaneously.
Prior to testing the structural relationships between personality states, roles and goals, an
unconditional model (Model 1) was estimated (no predictors were specified at either level 2
or 3) which can be appropriately represented in three stages, starting at level 1:
Level 1 : Yijk ¼ d1ijk p1jk þ d2ijk p2jk þ d3ijk p3jk þ d4ijk p4jk þ d5ijk p5jk þ eijk; (1)
where Yijk is the score on item i at measurement occasion j for person k; dpijk is a dummy-
coded indicator variable taking on the value of 1 if item Yi belongs to scale p and 0
otherwise for the five personality state scales; ppjk is the latent true score for person k
at measurement occasion j, and eijk is a measurement error assumed to be normally
distributed with a mean of zero and a variance s2 (in the current study of about 100 due to
T-standardization). The error variance in each scale s2e results from s2 divided by the
number of items constituting each scale p (Raudenbush et al., 1991). Level 1 can be thus
compared to a restrictive confirmatory factor analysis with loadings constrained to be equal
for items belonging to the same scale and one common error variance (Raudenbush &
Bryk, 2002; Raudenbush et al., 1991). To account for the rather strong assumption of a
common error variance at level 1, items were rescaled to have approximately equal error
variances according to a procedure suggested by Raudenbush et al. (1991).1
1Prior to MLM analyses, items of the FFM state measure were rescaled to have approximately equal errorvariances according to the procedure described by Raudenbush et al. (1991, p. 304). For each scale, estimates ofthe pooled item variance were computed across the total number of measurement occasions. T-scores of itemsbelonging to the same scale were then divided by the square root of the pooled item variance of this scale to arriveat comparable error variances.
Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Pers. 23: 509–530 (2009)
DOI: 10.1002/per
518 W. Bleidorn
At level 2, the true scores of the five personality state scales pp were assumed to vary
across occasions within persons:
Level 2 : ppjk ¼ bp0k þ rpjk (2)
where bp0k is the true score mean in scale p of person k, and rpjk is a random effect on scale p
associated with occasion j in person k. For each measurement occasion, the five random
effects were assumed multivariate normal with means of zero and a 5� 5 variance–
covariance matrix Tp.
At level 3, the person mean scores on the five latent states vary around the respective
grand means gp00:
Level 3 : bp0k ¼ gp00 þ up0k (3)
For each person, the random effects up0k were assumed multivariate normal with means of
zero and a 5� 5 covariance matrix Tb. Please note that there were five equations on levels
2 and 3 a time, namely one for each of the five personality state scales.
As can be seen from Equations 1 to 3, the fixed part of this unconditional three-level
model contains p regression coefficients for the indicator variables, which are the five
overall means for the personality state scales. The random part contains two variance–
covariance matrices, Tp and Tb, and one level-1 variance s2. This model describes the
error-free variability in personality state scores that is to be explained at the within- and
between-person level in subsequent analyses providing baseline estimates that can be used
to estimate effect sizes (Raudenbush & Bryk, 2002). A well-known effect-size concept in
ordinary multiple regression analysis is R2 representing the proportional reduction in
observed outcome variance. In MLM, the interpretation of this concept can become more
complicated, since there is unexplained variance at several levels to account for (Hox,
2002; Snijders & Bosker, 1999). In the current model, there were two variance components
of interest, namely within- and between-persons. Thus, two Pseudo-R2 values were
estimated: The level-2 Pseudo-R2 as an approximate value for the within-person
variance modelled by level-2 predictors, while the level-3 Pseudo-R2 should approximate
the between-person variance modelled by level-3 predictors2.
The model also allows a simultaneous consideration of the level-specific intercorrela-
tions of scales, which can be easily estimated on the basis of the variance–covariance
matrices Tpand Tb. Furthermore, the reliability of the FFM state measure can be estimated
at both levels of analysis simultaneously3 (Raudenbush et al., 1991).
To examine the structural relationships among personality states, social roles and
major life goals, Model 1 was gradually extended by employing predictor variables at
2These values should be considered approximate, because some specifics can arise when calculating theseparameters in MLM (Hox, 2002; Roberts & Monaco, 2006; Snijders & Bosker, 1999). Specifically, due totechnical reasons implicit in the maximum likelihood procedure, it is possible that some of the variancecomponents increase rather than decrease when adding predictor variables. Thus, it is possible to obtain negativevalues for the Pseudo R2 which is not possible in ordinary regression analysis. Though, there are attempts toprovide alternative effect size statistics that will not produce negative values for two-level models, there are norecommendations for multivariate three-level models yet. Thus, in the present analysis the well-establishedPseudo R2 was employed and interpreted according to the introduction by Raudenbush & Bryk (2002).3The within-person internal consistency is given by app¼Tppp/(Tpppþ s2
e) and depends on the number of itemsper scale and the degree of intercorrelation among them within each measurement occasion. The between-personinternal consistency is given by abp¼Tbpp/(Tbppþ (Tpppþ s2
e)/Jk). Although affected somewhat by the number ofitems and the degree of intercorrelations among them, this reliability depends more heavily on the number ofmeasurement occasions sampled per person and the degree of intercorrelation among them (Raudenbush et al.,1991).
Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Pers. 23: 509–530 (2009)
DOI: 10.1002/per
Personality states, social roles and major life goals 519
levels 2 and 3. At level 2, each state was regressed on the student and friend role scores
represented by coefficients referred to as level-2 slopes (Model 2). Social role scores were
group-mean centred4 (Raudenbush & Bryk, 2002), while group refers to individual
participants as distinct level-3 units. Besides, number of measurement occasion was added
at level 2 in order to control for linear time trends in the expression of personality states5.
At level 3, the coefficients estimated at the within-person level (intercepts and slopes)
become the dependent variables and were regressed onto major life goals (grand-mean
centred) represented by regression coefficients referred to as level-3 slopes (Model 3).
At each step of this model building process, nonsignificant predictors were omitted from
the final model. Please note that predictors at both levels entered analyses as observed
variables. Hence, correction due to random error was only achieved on part of the criterion,
that is, the five personality states.
RESULTS
The state quality of FFM dimensions
Within- and between-person variation in personality states
Estimates of both within- and between-person variances for each of the five latent
personality state scores were significantly different from zero (Table 1). The substantial
within- and between-person variance in each of the five states justified subsequent analyses
including predictors at both levels 2 and 3, that is the shifting role contexts as well as the
enduring goals of the person.
Reliability of personality state scales
Table 1 also shows the internal consistencies at levels 2 and 3 representing average
reliabilities over measurement occasions and persons, respectively. At the between-person
level, internal consistencies ranged between .82 for E and .97 for O indicating that the
employed adjectives provide highly reliable measures of the five personality dimensions
(for the average person across measurement occasions). The internal consistencies at the
within-person level were also satisfying, particularly with regard to the fact that they are
each based on six items only. This indicates that the reliabilities were still substantial when
FFM state scores were computed for single measurement occasions from which the stable
between-person variation had already been removed.
4In order to detect and estimate the variation of level-2 slopes properly, Raudenbush and Bryk (2002, p. 143) haverecommended group-mean centering in cases where predictor means may vary across the higher-level units.Hence, the five intercepts represented the unadjusted means for person k and the variances described the variationamong the participants’ unit means across measurement occasions. Level-3 predictors, on the other hand, werecentred around their grand means. That is, with regard to the two major life goals, the five intercepts can beinterpreted as adjusted means for person k.5Intensive repeated assessment is at risk to result in reactivity caused bias over time. In order to check for time-related trends in the expressions of personality states, the unconditional model was gradually extended by twolevel-2 predictors: Number of measurement occasion (per person, counting from zero) was included as a singlepredictor to check for linear change, while both number of measurement occasion and squared number ofmeasurement occasion were included to test for quadratic time trends (Raudenbush & Bryk, 2002; Hox, 2002).There were no systematic time trends in the states of N, O and C. There was, however, a small linear decrease in Eand a small linear increase in A throughout the study interval. Although these effects were rather small (accountingfor less than 2% of the within-person variance in these states), number of measurement occasion was considered asa control variable in all further analyses.
Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Pers. 23: 509–530 (2009)
DOI: 10.1002/per
Table 1. Variance components and reliability coefficients at the within- and between-person level(Model1)
Scale
Within-person (level 2) Between-person (level 3)
Variance (%) Reliability Variance (%) Reliability
N: Neuroticism 31.84 (67) .66 15.84 (33) .95E: Extraversion 56.48 (90) .77 6.08 (10) .82O: Openness 28.46 (53) .63 24.86 (47) .97A: Agreeableness 30.88 (63) .65 18.00 (37) .96C: Conscientiousness 57.26 (78) .77 16.29 (22) .93
Note: N¼ 52 individuals, N¼ 2917 measurement occasions; percentages of within- and between-person variance
in the latent personality variables are in brackets.
520 W. Bleidorn
Correlational structure at the within- and between-person level
The latent variable covariance matrices, Tp and Tb, supply estimates of the true-score
correlations among the five personality states at each level (Table 2). Correlations at the
between-person level (above the diagonal) varied in a range between �.21 for N and C to
.57 for E and A. At the within-person level (below the diagonal), correlations ranged
between .11 for E and C and .69 for E and O. Although the pattern of correlations obtained
at the within-person level largely mirrored the between-person pattern, there were some
notable differences, particularly regarding the dimension of E. While the between-person
relationship between E and O was only moderate, the respective within-person correlation
was considerably higher. In contrast, the correlation between E and C obtained at the
between-person level was moderate in size but rather negligible at the within-person level.
Personality states as a function of current social roles
In order to explain part of the within-person variation in personality states, the
unconditional model (Model 1) was extended by incorporating social roles as level-2
predictors whilst taking into account linear time trends in E and A. The correlation between
the two social roles predictors was rather low (r¼ .12). The first five columns of Table 3
summarize the results of this conditional model in which each of the five personality states
was simultaneously predicted from the student and friend role scores (Model 2). At this
step, variation in states was exclusively modelled at the within-person-level. That is, Model
2 was still unconditional at level 3.
Table 2. Correlations among Five-Factor Model personality states at the within- and between-person level (Model 1)
N E O A C
N: Neuroticism �.42 �.22 �.31 �.21E: Extraversion �.47 .27 .57 .31O: Openness �.36 .69 .42 .45A: Agreeableness �.53 .54 .46 .43C: Conscientiousness �.30 .11 .27 .26
Note: N¼ 52 individuals, N¼ 2917 measurement occasions. Between-person correlations are given above the
diagonal; within-person correlations are given below the diagonal.
Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Pers. 23: 509–530 (2009)
DOI: 10.1002/per
Table 3. Multivariate multilevel regressions of Five-Factor Model personality states on currentsocial roles (Model 2) and major life goals (Model 3) with control for time trends
Model parameter
Model 2
(conditional at level 2)
Model 3
(conditional at level 2 and 3)
Fixed (MLM
regression coefficients) N E O A C N E O A C
Intercept, gp00 50.01 50.65 49.99 49.10 49.99 50.01 50.65 49.99 49.11 49.99
Student role, gp10 0.39 �1.12 �0.48 �0.63 1.93 0.39 �1.12 �0.50 �0.66 1.90
Friend role, gp20 �0.45 2.62 0.88 1.01 — �0.45 2.61 0.88 1.00 —
Linear changea, gp30 — � 0.02 — 0.03 — — � 0.02 — 0.03 —
Achievement goals, gp01 — — — — — �3.65 1.47 3.37 1.95 3.59
Affiliation goals, gp02 — — — — — — 1.34 — 2.74 —
Random(variance components)
L2-Intercept, rpjk 29.86 33.47 24.48 24.38 42.80 29.86 33.47 24.48 24.38 42.80
L3-Intercept, up0k 15.88 9.36 24.94 15.65 16.57 11.85 6.42 19.99 11.46 11.08
L3-Student role, up1k 0.25 0.74 0.27 0.25 1.95 0.25 0.74 0.26 0.25 1.75
L3-Friend role, up2k 0.36 1.23 0.66 0.37 1.30 0.36 1.23 0.66 0.37 1.29
L3-Linear change, up3k — 0.01 — 0.01 — — 0.01 — 0.01 —
Modelled varianceb
L2-Pseudo R2 .06 .41 .14 .21 .25 .06 .41 .14 .21 .25
L3-Pseudo R2 — — — — — .25 .31 .20 .27 .33
Note: N¼ 52 individuals, N¼ 2917 measurement occasions. L2¼ level 2, L3¼ level 3. N, E, O, A, C¼Five-
Factor Model personality states (Neuroticism, Extraversion, Openness, Agreeableness, Conscientiousness).
MLM¼Multilevel modelling. Model 2¼Multivariate multilevel regressions of Five-Factor Model personality
states on current social roles with control for time trends (Model 2, only conditional at level 2). Model
3¼Multivariate multilevel regressions of Five-Factor Model personality states on current social roles and major
life goals with control for time trends (Model 3, fully conditional at level 2 and 3).Cross-level interactions
involving the level-2 and level-3 predictors were not presented as none of these effects was significant.aNumber of measurement occasion (per person, counting from zero). bProportional reductions in the variance
components of personality states (N, E, O, A, C). At the within-person level (L2-Pseudo R2), proportional
reductions in variance were estimated in comparison to the unconditional model (without any explanatory
variables, see Table 1); at the between-person level (L3-Pseudo R2), proportional reductions in variance were
estimated in comparison to Model 2 (only conditional at level 2). Values in boldface are significant at p< .01;
values in italics are significant at p< .05.
Personality states, social roles and major life goals 521
The fixed part of the model (see the upper part of Table 3) contains the unstandardized
multilevel regression coefficients which can be interpreted the same as standard regression
coefficients. The intercepts (gp00) represent the average levels in the dependent variables,
which are all about 50 due to T-score standardization. The slopes for student (gp10) and
friend role (gp20) reveal the direction and magnitude of associations between these level-2
predictors and the variation in FFM states for the average participant. As hypothesized,
there were significant relationships between the two roles and each of the five personality
states. Analyses revealed negative effects of the student role on E, O and A, but positive
effects on N and C. The friend role, on the other hand, showed fixed effects on all states,
except C. Being in the role of a friend was positively related to the states of E, A and O. For
instance, there was an increase of 2.62 T-score points in state E for every one-point-increase
in friend role scores (holding the student role constant).
The bottom part of Table 3 shows the random part of the model, that is, the variance
components. Comparing the variance components of Model 2 with those estimated in the
Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Pers. 23: 509–530 (2009)
DOI: 10.1002/per
522 W. Bleidorn
unconditional model (Table 1), it becomes apparent that entering the level-2 predictors has
consistently decreased the within-person variance in the five personality state scales (rpjk).
By comparing these values with the level-2 variances in the unconditional model, an index
for the proportion of latent variance explained at level 2 (L2-Pseudo R2) was estimated
(Raudenbush & Bryk, 2002). The largest amount of variance could be explained in state E
as about 41% of the within-person variance was due to changes in the degree the person
acted as a friend or a student. In contrast, the two social roles only accounted for 6% of the
latent variance in N indicating that state N did not vary effectively as a function of being in
the student or the friend role.
Variance in the level-3 intercepts (up0k) refers to the between-person differences in
mean-levels of personality states and was actually supposed to remain unchanged, since
there were no level-3 predictors in Model 2. However, the estimation and interpretation of
variance components in MLM is more complex than in ordinary regression analysis
(e.g. Hox, 2002). As can bee seen in Table 3, also the between-person variance in the five
personality states has changed. In fact, variances in all personality states, but A, increased
slightly to rather appreciably when level-2 predictors were added to the model. According
to the recommendation of Hox (2002), these conditional between-person variance
components (up0k) were used as the baseline model to evaluate the effects of level-3
predictors (major life goals) in subsequent analyses.
Level-3 variances in student role (up1k) and friend role (up2k) slopes indicate the amount
of variation in these within-person effects across participants. Effects of both student
and friend role differed significantly between participants. For example, while some
individuals acted increased conscientiously when occupying the student role, others only
marginally changed their level of state C in this role context. Since the fixed effect of the
slope for the relationship between friend role and state C was not significantly different
from zero, the significant variance of this slope point to between-person differences in the
direction of the effect. As shown in Figure 1, there were both individuals who increased and
individuals who decreased in state C when being in the role of a friend.
Figure 1. Predicted T-score values of state C for individuals who were one standard deviation above and onestandard deviation below the fixed within-person effect of friend role on state C.
Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Pers. 23: 509–530 (2009)
DOI: 10.1002/per
Personality states, social roles and major life goals 523
Personality states as a function of current social roles and major life goals
The final step of analyses sought to build up an explanatory model to account for the
between-person variability in both average personality states and the within-person links
between roles and states. For that purpose, the previous model was extended towards a full
three-level model (Raudenbush & Bryk, 2002) in which parameters at the within-person
level were treated as outcomes to examine how they vary as a function of measures at the
between-person level, namely achievement and affiliation goals (Model 3, see the last five
columns of Table 3). The correlation between the level-3 variables was moderate (r¼ .26).
The intercepts (gp00) and the fixed effects of the level-2 predictors (gp10, gp20, gp30)
obtained from Model 3 closely resembled the results of Model 2. The last two rows of the
upper part of Table 3 refer to the fixed effects of achievement (gp01) and affiliation goals
(gp02), indicating the expected difference between mean-levels in personality states of two
persons who differ by one unit in their importance ratings of these goals. As expected,
the effects of achievement and affiliation goals differed among personality states.
Achievement goals were significantly related to each of the five states suggesting that
participants with high achievement goals acted on average in a more extraverted, open,
agreeable and conscientious but less neurotic way. Regarding affiliation, findings also
support the present hypotheses, as this goal category was significantly related to the states
of E and A. That is, participants striving stronger for affiliation acted on average in a more
extraverted and agreeable way compared to those scoring low in this goal category.
Cross-level interactions involving the role predictors at level 2 and the two goals at level
3 were not presented in Table 3 as none of these effects was significant. That is, the goal
categories considered did not account for a substantial amount of the between-person
variance in the within-person links between social roles and personality states.
The bottom part of Table 3 also shows the random effects of the final model. As this
model was exclusively expanded by incorporation of level-3 predictors, additional variance
was solely explained at the between-person level. Thus, solely the variance components of
the level-3 intercepts (up0k) had changed notably compared to the previous model (Model
2, see the left part of Table 3). By comparing the between-person variances across Models 2
and 3, an index of the variance explained at level 3 (L3-Pseudo R2) was estimated. Results
show a considerable reduction in between-person variance in all five personality states.
Specifically, participants’ long-term strivings for achievement accounted for 25 and 33% of
the between-person variance in persons’ average levels of state N and C, respectively. Both
affiliation and achievement goals explained 31% of the between-person variance in
persons’ mean-level of state E, 20% of the between-person variance in the mean-level of
state O, and 27% of the between-person variance in the mean-level of state A.
As there were no significant cross-level interactions, incorporation of level-3 predictors
only marginally reduced variances in slopes. In essence, though achievement and affiliation
goals proved to be effective predictors of average personality states, they did not help
considerably to account for the between-person differences in the within-person
associations between social roles and personality states.
DISCUSSION
A comprehensive understanding of personality functioning requires knowledge of both the
stable structures and the dynamic processes relevant to the system. This research aimed at
expanding our understanding of the stable and dynamic aspects of personality by focusing
Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Pers. 23: 509–530 (2009)
DOI: 10.1002/per
524 W. Bleidorn
on the interplay between FFM personality states, current social roles and major life goals as
it becomes apparent in the stream of people’s daily lives.
Using FFM dimensions to measure personality states
Providing further support for the density-distribution approach to personality, results
indicate that variability in trait-relevant behaviour can be reliably assessed by bipolar
adjective pairs relevant to the FFM as represented in the NEO-PI-R. Personality states
varied substantially within and between individuals. At least half of the total latent variance
was the result of variation within persons supporting the state quality of the five broad
dimensions (Fleeson, 2001).
With a few exceptions, the intercorrelational patterns among the five states obtained at
the within- and between-person levels closely resembled each other. That is, the structural
arrangement of personality states within persons largely parallels the organization of these
characteristics when they are conceptualized as generalized individual differences.
However, there were also some notable differences suggesting that some personality states
closely related at the within-person level (i.e. frequently expressed at the same time) are not
necessarily related to the same degree at the between-person level and vice versa. For
instance, acting extraverted at a given moment is strongly related to acting openly whereas
the between-person relationship between the average levels of these states is considerable
weaker. Thus, although personality variables relevant to within-person processes seem to
share many properties with their between-person counterparts, there are still not the same.
Integrated research into both structures and processes is necessary to explain personality
adequately (Fleeson, 2004; Mischel & Shoda, 1995; Nezlek, 2007).
Although this study mainly focused on the comparableness of the within- and between-
person structures of intercorrelations among personality states, the magnitude of the
correlation coefficients may demand some further comments. Regarding both within- and
between-person correlations, it becomes apparent that the five state dimensions are not
perfectly orthogonal. On the face of it, this finding seems to stand in contrast with the
conceptual assumptions of the FFM. However, it should be noted that orthogonality among
the five personality states was not predicted. Actually, also the FFM of personality traits
does not postulate orthogonality as a defining characteristic but rather as a theoretical
heuristic. There are a number of studies suggesting that the five broad domains assessed
by the NEO-PI-R are not completely independent but rather interrelated in a systematic
manner (e.g. McCrae, Zonderman, Costa, Bond, & Paunonen, 1996). For instance, McCrae
et al. (1996) reported substantial correlations among the five NEO-PI-R factors, partly
exceeding j.50j, when simple structure models were tested by means of confirmatory factor
analysis.
Within-person links between personality states and social roles
Student and friend roles turned out to be substantial predictors of different sets of
personality states supporting the assumption that the social role with its immanent demands
is a reasonable concept for contextualizing personality (Heller et al., 2007; Roberts, 2007;
Wood, 2007). Individuals show different personality states in different role contexts,
conceivably adapting their trait-relevant behaviour to the given role demands. However, the
social roles examined in this study did not explain a reasonable amount of within-person
variance in the state of N. There might be other role contexts for which changes in state N
Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Pers. 23: 509–530 (2009)
DOI: 10.1002/per
Personality states, social roles and major life goals 525
are more likely to arise. Referring to studies into long-term changes in trait N and social
relationships (Neyer & Asendorpf, 2001), being in the role of a romantic partner might
induce decreases in state N. Regrettably, most of the participants in the current study were
singles and consequently did not take over the role of a romantic partner. Hence, further
studies examining associations of personality states with other role contexts in different
samples are required to test this hypothesis systematically. Furthermore, future research is
desired in order to reveal the psychological components of roles that operate as proximal
determinants of personality states. Along these lines, Heller et al. (2007) considered roles
as higher order constructs and specified some subordinate features worthy of study, such as
short term goals pursued while occupying a role, expectations, characteristics of the people
with whom one enacts the role, or the ease of termination of a role.
It should be also noted that current findings did not allow causal conclusions. Although
theoretical arguments speak for the role context causing changes in personality states, the
reverse association is possible, too. For instance, it is reasonable to assume that individuals
who are generally high in C are more inclined to take over the student role, while occupying
the student role might in turn affect the trait-relevant behaviour shown in this role context.
These mutually influencing processes could be interpreted in line with the corresponsive
principle (Caspi, Roberts, & Shiner, 2005; Roberts & Wood, 2006) which was originally
formulated to explain the coexistence of stability and change of personality traits over the
life span in transactional theories of personality development. This principle proposes that
people select environments that are correlated with their personality, while the subsequent
environmental experiences should affect particularly those characteristics that have led
them to those experiences in the first place. That is, contextual demands do not impinge
themselves on people in a random fashion. At least in part, people seem to construct and
modify their contextual experiences themselves in accordance to their personalities
(Mischel & Shoda, 1995).
A further issue of the current study concerned the universality of the within-person
relationships between roles and personality states. Consistent with predictions, within-
person associations were not universal but differed significantly between individuals
(Fleeson, 2007; Mischel & Shoda, 1995). Considering the relationship between state C and
the friend role, there were even both individuals who increase in C and individuals who
decrease in C when being in the role of a friend. These differences might reflect the
differences in how individuals encode, interpret and evaluate a given role context. Even
two individuals who are similar in their average levels of personality states will display
distinctive and probably predictable patterns of states when they are occupying the same
role at a given moment. Thus, in line with interactional positions, findings emphasize that it
is necessary but not sufficient to know the contextual demands in order to predict an
individual’s trait-relevant behaviour adequately. How an individual will act in a given role
context is also a function of the person and the unique way how the person responds to the
perceived role demands (Allport, 1937; Magnusson & Endler, 1977; Mischel & Shoda,
1995).
Linking personality states, current social roles and major life goals
The final step of the present analyses was to incorporate person-based variables supposed
to predict the between-person differences in the within-person effects. Major life goals
were assumed to be important determinants of average personality states and the within-
person links between roles and states, since goals are conceptualized as motivational
Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Pers. 23: 509–530 (2009)
DOI: 10.1002/per
526 W. Bleidorn
constructs that interact with contextual features to give rise to trait-relevant behaviour
in the service of pursuing and achieving the intended aims. Individuals’ striving for
achievement and affiliation actually proved as reasonable predictors of between-person
differences in mean-levels of FFM states. As hypothesized, those individuals with high
affiliation goals acted in an increased extraverted and agreeable way. Individuals with high
achievement goals acted on average more openly and conscientiously. Moreover, they
exhibited higher mean-levels in the states of E and A and lower mean-levels in state N.
Though the latter associations were not predicted, it seems plausible to assume that
students are also more inclined to act in an extraverted, agreeable and emotional stable way
in order to accomplish their achievement goals in the academic environment. To sum up,
long-term aspirations to shape one’s life seem to channel an individual’s average trait-
relevant behaviour across different role contexts. This is line with the conception of persons
as future-oriented agents who not only react to environmental stimuli but also shape their
lives proactively (Mischel & Shoda, 1995, 1998, 1999).
The associations between goals and personality states largely mirrored the relations
previously revealed between goals and personality traits (Roberts & Robins, 2000; Roberts
et al., 2004). Hence, these findings may also shed further light on the processes through
which major life goals and personality traits are related at the between-person level. In
terms of the corresponsive principle, the associations between personality traits and major
life goals should reflect the fact that people choose goals that reinforce existing traits
(Roberts et al., 2004). This reinforcement of traits might be initiated by an increased
enactment of the respective states (Fleeson et al., 2002). In order to deepen our
understanding of the concrete mechanisms underlying these links, further research into the
long-term relations between micro- and macro-level structures and processes would be
desirable (Mroczek, Almeida, Spiro, & Pafford, 2006).
While achievement and affiliation goals proved to be effective predictors of average
personality states, they did not effectively help to account for the between-person
differences in the within-person links between personality states and current social roles.
Maybe, major life goals were defined at a level which is to abstract to explain individual
differences in varying within-person contingencies effectively. Goals which are situated at
a somewhat lower level of the goal hierarchy may be more effective predictors of context-
specific dynamics. Thus, future research might focus on mid-level units of the goal
hierarchy that are somewhat more specific and contextually embedded in one’s life
situation.
Furthermore, one should also consider methodical specifics with regard to the detection
of cross-level interactions. Since on average 57 records of daily behaviour were collected
from 52 participants, this study clearly exceeds common rules of thumb concerning
adequate sample sizes in multilevel models, when interest is mostly in the fixed parameters
(e.g. Hox, 2002). However, the number of persons (level-3 units) might have been still too
small to detect cross-level interactions with sufficient power. Thus, one should not
definitely rule out cross-level effects of major life goals on the within-person links between
social roles and personality states unless further studies have re-examined these effects
with larger sample sizes.
Limitations
The experience-sampling design provides an ecologically valid access to study personality
functioning in the ongoing stream of people’s daily behaviour (Heller et al., 2007).
Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Pers. 23: 509–530 (2009)
DOI: 10.1002/per
Personality states, social roles and major life goals 527
Combined with MLM procedures, it offers the opportunity to integrate nomothetic and
idiographic-oriented research into personality functioning (Fleeson et al., 2002). However,
despite this study exhibits a number of strengths, it is not without limitations. First, as
already discussed above, the present analyses identified associations only and cannot
support causal conclusions.
Second, this research purposely focused on social roles and major life goals
exemplifying only two types of variables that determine individuals’ personality states.
Trait-relevant behaviour is undoubtedly caused by several changing sources beyond social
roles, including other features of the context, the goals an individual has active at the
moment, or time-based processes such as cycles or internal physiological processes, to
name just a few examples (Fleeson, 2007; Mischel & Shoda, 1995). There are also other
stable units of the personality system, such as expectancies or competencies, which
probably interact as individuals select, encode, or modify their environment (Mischel &
Shoda, 1995). The current study might encourage continued research into further variables
underlying daily fluctuations in personality states in order to complete the overall picture
successively.
Regarding the rather strong restrictions of the multivariate multilevel regression
approach chosen to analyse the current experience-sampling data, it must be admitted that
multilevel structural equation modelling (e.g. Muthen, 1994) is more flexible and less
restrictive. However, multilevel structural equation modelling does not model raw scores
and therefore, does not produce the estimated scores on the latent variables, that is, the
fixed effects of social roles and major life goals on personality states (Hox, 2002).
Finally, future studies on more representative samples over longer time periods are
required to back up the robustness of the current findings. In this regard, it would be
desirable to implement additional methods beyond self-reports, such as other reports or
observational methods.
CONCLUSION
The present work demonstrates a flexible strategy to study relationships at the within- and
between-person levels of personality simultaneously. It thus presents a promising avenue to
proceed with integrated research into personality structures and processes as two
interrelated sides of the same behaviour-producing system. Pointing to the intraindividual
changeability of personality, present analyses revealed a substantial amount of within-
person variability that can be systematically related to varying role contexts. On the other
hand, long-term goals accounted for between-person differences in individuals’ average
levels of personality states. This study might encourage further research into the
relationships among personality structures and dynamics which will gradually enhance our
understanding of the complex mechanisms underlying personality functioning.
REFERENCES
Allport, G. W. (1937). Personality. A psychological interpretation. New York: Henry Holt andCompany.
Austin, J. T., & Vancouver, J. B. (1996). Goal constructs in psychology: Structure, process, andcontent. Psychological Bulletin, 120, 338–375.
Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Pers. 23: 509–530 (2009)
DOI: 10.1002/per
528 W. Bleidorn
Bakan, D. (1966). The duality of human existence: Isolation and communion in Western man. Boston:Beacon Press.
Barrett, D. J., & Feldman Barrett, L. (2000). The Experience Sampling Program (ESP, Version 4.0)[Computer software]. http://www2.bc.edu/�barretli/esp/
Bem, D. J., & Funder, D. C. (1978). Predicting more of the people more of the time: Assessing thepersonality of situations. Psychological Review, 85, 485–501.
Cantor, N., & Fleeson, W. (1994). Social intelligence and intelligent goal pursuit: A cognitive slice ofmotivation. In W. Spaulding (Ed.), Integrative views of motivation, cognition, and emotion(pp. 125–179). Lincoln, NE, US: University of Nebraska Press.
Caspi, A., Roberts, B. W., & Shiner, R. L. (2005). Personality development: Stability and change.Annual Review of Psychology, 56, 453–484.
Costa, P. T., Jr., & McCrae, R. R. (1992). Revised NEO Personality Inventory (NEO PI-R) and NEOFive Factor Inventory. Professional Manual. Odessa, Fl: Psychological Assessment Resources.
Denissen, J. J. A., & Penke, L. (2008). Motivational individual reaction norms underlying the Five-Factor model of personality: First steps towards a theory-based conceptual framework. Journal ofResearch in Personality, 42, 1285–1302.
Endler, N. S., & Parker, J. D. A. (1992). Interactionism revisited: Reflections on the continuing crisisin the personality area. European Journal of Personality, 6, 177–198.
Epstein, S. (1994). Trait theory as personality theory: Can a part be as great as the whole?Psychological Inquiry, 5, 120–122.
Fleeson, W. (2001). Toward a structure- and process-integrated view of personality: Traits as densitydistributions of states. Journal of Personality and Social Psychology, 80, 1011–1027.
Fleeson, W. (2004). Moving personality beyond the person-situation debate: The challenge and theopportunity of within-person variability. Current Directions in Psychological Science, 13, 83–87.
Fleeson, W. (2007). Situation-based contingencies underlying trait-content manifestation in beha-vior. Journal of Personality, 75, 825–862.
Fleeson, W., & Cantor, N. (1995). Goal relevance and the affective experience of daily life: Ruling outsituational explanations. Motivation and Emotion, 19, 25–57.
Fleeson, W., Malanos, A. B., & Achille, N. M. (2002). An intraindividual process approach to therelationship between extraversion and positive affect: Is acting extraverted as ‘good’ as beingextraverted? Journal of Personality and Social Psychology, 83, 1409–1422.
Fournier, M. A., Moskowitz, D. S., & Zuroff, D. C. (2008). Integrating dispositions, signatures, andthe interpersonal domain. Journal of Personality and Social Psychology, 94, 531–545.
Funder, D. C. (2001). Personality. Annual Review of Psychology, 52, 197–221.Heller, D., Watson, D., Komar, J., Min, J. A., & Perunovic, W. Q. E. (2007). Contextualized
personality: Traditional and new assessment procedures. Journal of Personality, 75, 1229–1254.Hofer, J. (2003). Gemeinschafts- und personzentrierte Lebensziele von maennlichen Jugendlichen in
Sambia: Methodische und inhaltliche Eroerterungen [Community- and ego-centered life goals ofmale adolescents in Zambia: Methodological and theoretical considerations]. Zeitschrift furEntwicklungspsychologie und Padagogische Psychologie, 35, 111–121.
Hofer, J., & Chasiotis, A. (2003). Congruence of life goals and implicit motives as predictors of lifesatisfaction: Cross-cultural implications of a study of Zambian male adolescents. Motivation andEmotion, 27, 251–272.
Hox, J. J. (2002). Multilevel analyses: Techniques and applications. Mahwah, NJ: Erlbaum.Little, B. R., Lecci, L., & Watkinson, B. (1992). Personality and personal projects: Linking Big Five
and PAC units of analysis. Journal of Personality, 60, 501–525.Magnusson, D., & Endler, N. S. (1977). Personality at the crossroads: Current issues in interactional
psychology. Hillsdale, NJ: Erlbaum.Maier, G. W., & Brunstein, J. C. (2001). The role of personal work goals in newcomers’ job
satisfaction and organizational commitment: A longitudinal analysis. Journal of Applied Psychol-ogy, 86, 1034–1042.
McCrae, R. R., & John, O. P. (1992). An introduction to the five-factor model and its applications.Journal of Personality, 60, 175–215.
McCrae, R. R., Zonderman, A. B., Costa, P. T., Jr., Bond, M. H., & Paunonen, S. V. (1996).Evaluating replicability of factors in the Revised NEO Personality Inventory: Confirmatory factoranalysis versus Procrustes rotation. Journal of Personality and Social Psychology, 70, 552–566.
Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Pers. 23: 509–530 (2009)
DOI: 10.1002/per
Personality states, social roles and major life goals 529
Mischel, W. (2004). Toward an integrative science of the person. Annual Review of Psychology, 55,1–22.
Mischel, W., & Shoda, Y. (1995). A cognitive-affective system theory of personality: Reconceptua-lizing situations, dispositions, dynamics, and invariance in personality structure. PsychologicalReview, 102, 246–268.
Mischel, W., & Shoda, Y. (1998). Reconciling processing dynamics and personality dispositions.Annual Review of Psychology, 49, 229–258.
Mischel, W., & Shoda, Y. (1999). Integrating dispositions and processing dynamics within a unifiedtheory of personality: The cognitive-affective personality system. In L. Pervin, & O. P. John (Eds.),Handbook of personality: Theory and research (pp. 197–218). New York: Guilford Press.
Mroczek, D. K., Almeida, D. M., Spiro, A., & Pafford, C. (2006). Modeling intraindividual stabilityand change in personality. In D. K. Mroczek, & T. D. Little (Eds.), Handbook of personalitydevelopment (pp. 163–180). Mahwah, NJ, US: Lawrence Erlbaum Associates Publishers.
Muthen, B. O. (1994). Multilevel covariance structure analysis. Sociological Methods & Research,22, 376–398.
Neyer, F. J., & Asendorpf, J. B. (2001). Personality-relationship transaction in young adulthood.Journal of Personality and Social Psychology, 81, 1190–1204.
Nezlek, J. B. (2007). A multilevel framework for understanding relationships among traits, states,situations and behaviours. European Journal of Personality, 21, 789–810.
Ostendorf, F., & Angleitner, A. (2004). NEO-Personlichkeitsinventar nach Costa und McCrae.Revidierte Fassung (NEO-PI-R). Goettingen: Hogrefe.
Pohlmann, K., & Brunstein, J. C. (1997). GOALS: Ein Fragebogen zur Messung von Lebenszielen[GOALS: A questionnaire for assessing life goals]. Diagnostica, 43, 63–79.
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical Linear Models: Applications and dataanalysis methods (2nd ed.). Thousand Oaks: Sage Publications.
Raudenbush, S. W., Bryk, A. S., Cheong, Y. F., & Congdon, R. (2000). HLM (Version 6) [Computersoftware]. Chicago, IL: Scientific Software.
Raudenbush, S. W., Rowan, B., & Kang, S. J. (1991). A multilevel, multivariate model for studyingschool climate with estimation via the EM algorithm and application to U.S. high-school data.Journal of Educational Statistics, 16, 295–330.
Roberts, B. W. (2007). Contextualizing personality psychology. Journal of Personality, 75, 1071–1082.
Roberts, B. W., & Donahue, E. M. (1994). One personality, multiple selves: Integrating personalityand social roles. Journal of Personality, 6, 199–218.
Roberts, J. K., & Monaco, J. P. (2006). Effect size measures for the two-level linear multilevel model.Paper presented at the Annual Meeting of the American Educational Research Association, SanFrancisco, California.
Roberts, B. W., O’Donnell, M., & Robins, R. W. (2004). Goal and personality trait development inemerging adulthood. Journal of Personality and Social Psychology, 87, 541–550.
Roberts, B. W., & Pomerantz, E. M. (2004). On traits, situations, and their integration: Adevelopmental perspective. Personality and Social Psychology Review, 8, 402–416.
Roberts, B. W., & Robins, R. W. (2000). Broad dispositions, broad aspirations: The intersection ofpersonality traits and major life goals. Personality and Social Psychology Bulletin, 26, 1284–1296.
Roberts, B. W., & Wood, D. (2006). Personality development in the context of the Neo-SocioanalyticModel of Personality. In D. K. Mroczek, & T. D. Little (Eds.), Handbook of personalitydevelopment (pp. 11–39). Mahwah, NJ, US: Lawrence Erlbaum Associates Publishers.
Saucier, G., Bel-Bahar, T., & Fernandez, C. (2007). What modifies the expression of personalitytendencies? Defining basic domains of situation variables. Journal of Personality, 75, 479–504.
Sheldon, K. M., & Cooper, M. L. (2008). Goal striving within agentic and communal roles: Separatebut functionally similar pathways to enhanced well-being. Journal of Personality, 76, 415–447.
Sheldon, K. M., & Elliot, A. J. (2000). Personal goals in social roles: Divergences and convergencesacross roles and levels of analysis. Journal of Personality, 68, 51–84.
Shoda, Y., Mischel, W., & Wright, J. C. (1994). Intraindividual stability in the organization andpatterning of behavior: Incorporating psychological situations into the idiographic analysis ofpersonality. Journal of Personality and Social Psychology, 67, 674–687.
Snijders, T. A. B., & Bosker, R. J. (1999). Multilevel analysis: An introduction to basic and advancedmultilevel modeling. London: Sage.
Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Pers. 23: 509–530 (2009)
DOI: 10.1002/per
530 W. Bleidorn
Stryker, S. (2007). Identity theory and personality theory: Mutual relevance. Journal of Personality,75, 1083–1102.
Wood, D. (2007). Using the PRISM to compare the explanatory value of general and role-contextualized trait ratings. Journal of Personality, 75, 1103–1126.
Wood, D., & Roberts, B. W. (2006). Cross-sectional and longitudinal tests of the Personality and RoleIdentity Structural Model (PRISM). Journal of Personality, 74, 779–809.
APPENDIX ITEMS OF THE FFM STATE MEASURE
Scale
Copyright # 2009 John Wi
Item
N: Neuroticism
N1 Relaxedley & Sons, Ltd.
vs.
Eur. J. Pers. 23: 5
DO
Strained
N2 Even-tempered vs. Irritable N3 Confident vs. Unconfident N4 Stable vs. Labile N5 Unconcerned vs. Concerned N6 Self-controlled vs. ImpulsiveE: Extraversion
E1 Unhurried vs. Vivid E2 Restrained vs. Frolicsome E3 Reserved vs. Cordial E4 Silent vs. Talkative E5 Withdrawn vs. Sociable E6 Unventuresome vs. VenturesomeO: Openness
O1 Fanciless vs. Fanciful O2 Unpoetic vs. Poetic O3 Unimaginative vs. Imaginative O4 Lacking in ideas vs. Full of ideas O5 Inartistic vs. Artistic O6 Unfeeling vs. FeelingA: Agreeableness
A1 Ungentle vs. Gentle A2 Hard-hearted vs. Kind-hearted A3 Unforbearing vs. Forbearing A4 Uncooperative vs. Cooperative A5 Altruistic vs. Egoistic A6 Unfrank vs. FrankC: Conscientiousness
C1 Undisciplined vs. Disciplined C2 Unmotivated vs. Motivated C3 Unsystematic vs. Systematic C4 Ineffective vs. Effective C5 Thoughtless vs. Deliberate C6 Irresponsible vs. Responsible09–530 (2009)
I: 10.1002/per