ArticlePDF Available

When “Together“ Means “Too Close“: Agency Motives and Relationship Functioning in Coresident and Living-Apart-Together Couples

July 2015
Journal of Personality and Social Psychology

Authors:

Birk Hagemeyer

Friedrich Schiller University Jena

Franz J Neyer

Friedrich Schiller University Jena

Show all 5 authorsHide

The present research addresses the interplay between agency motives and objective dyadic closeness with regard to the functioning of intimate couple relationships. Applying a Person X Situation approach, we hypothesized (a) that partners’ implicit and explicit agency motives predict their selection of dyadic living arrangements characterized by high or low objective closeness (coresidence or living-aparttogether), (b) that agency motives have more negative effects on relationship functioning in coresident couples, (c) that agency motives predict agentic motivational states in coresident couples, and (d) that agentic states predict day-to-day changes in relationship satisfaction under conditions of high objective closeness. We found support for these between- and within-couple hypotheses in cross-sectional and prospective analyses of an age-heterogeneous sample of 548 heterosexual couples, and in a 2-week diary study with a subsample of 106 couples. Most notably, agentic motive dispositions and motivational states related to relationship functioning more negatively under conditions of high objective closeness. The overall positive effect of objective closeness on relationship functioning was diminished by strong agentic motivation. Perspectives for future research on agency motives in couple relationships are discussed.

Prediction of average daily agency appetence by explicit and implicit agency motives in COR and LAT relationships (Hypothesis 3). Results were plotted for predictor values one standard deviation below and above the grand mean. Panel A: Male effects of explicit agency desire on agency appetence in COR and LAT couples. Panel B: Female effects of explicit agency desire on agency appetence in COR and LAT couples. Panel C: Male effects of implicit agency motive on agency appetence in COR and LAT couples. Panel D: Female effects of implicit agency motive on agency appetence in COR and LAT couples. COR coresidence; LAT living-apart-together.

…

Prediction of day-to-day changes in relationship satisfaction by the interaction of daily hours spent together and daily agency appetence in COR and LAT relationships (Hypothesis 4). Results were plotted for deviations of 4 hours below and above the couple-specific average level of hours spent together and minimum (= 0) and maximum (= 3) values of agency appetence.

…

Figures - uploaded by Birk Hagemeyer

Content may be subject to copyright.

Content uploaded by Birk Hagemeyer

Content may be subject to copyright.

Content uploaded by Birk Hagemeyer

Content may be subject to copyright.

Journal of Personality and Social Psychology

When “Together” Means “Too Close”: Agency Motives

and Relationship Functioning in Coresident and

Living-Apart-Together Couples

Birk Hagemeyer, Felix D. Schönbrodt, Franz J. Neyer, Wiebke Neberich, and Jens B.

Asendorpf

Online First Publication, September 14, 2015. http://dx.doi.org/10.1037/pspi0000031

CITATION

Hagemeyer, B., Schönbrodt, F. D., Neyer, F. J., Neberich, W., & Asendorpf, J. B. (2015,

September 14). When “Together” Means “Too Close”: Agency Motives and Relationship

Functioning in Coresident and Living-Apart-Together Couples. Journal of Personality and

Social Psychology. Advance online publication. http://dx.doi.org/10.1037/pspi0000031

When “Together” Means “Too Close”: Agency Motives and Relationship

Functioning in Coresident and Living-Apart-Together Couples

Birk Hagemeyer

Friedrich-Schiller-Universität Jena

Felix D. Schönbrodt

Ludwig-Maximilians-Universität München

Franz J. Neyer

Friedrich-Schiller-Universität Jena

Wiebke Neberich

Humboldt-Universität zu Berlin, International Max Planck

Research School LIFE

Jens B. Asendorpf

Humboldt-Universität zu Berlin

The present research addresses the interplay between agency motives and objective dyadic closeness with

regard to the functioning of intimate couple relationships. Applying a Person ⫻ Situation approach, we

hypothesized (a) that partners’ implicit and explicit agency motives predict their selection of dyadic

living arrangements characterized by high or low objective closeness (coresidence or living-apart-

together), (b) that agency motives have more negative effects on relationship functioning in coresident

couples, (c) that agency motives predict agentic motivational states in coresident couples, and (d) that

agentic states predict day-to-day changes in relationship satisfaction under conditions of high objective

closeness. We found support for these between- and within-couple hypotheses in cross-sectional and

prospective analyses of an age-heterogeneous sample of 548 heterosexual couples, and in a 2-week diary

study with a subsample of 106 couples. Most notably, agentic motive dispositions and motivational states

related to relationship functioning more negatively under conditions of high objective closeness. The

overall positive effect of objective closeness on relationship functioning was diminished by strong

agentic motivation. Perspectives for future research on agency motives in couple relationships are

discussed.

Keywords: agency, implicit motives, intimate relationships, living-apart-together, relationship quality

Supplemental materials: http://dx.doi.org/10.1037/pspi0000031.supp

The most painful thing is losing yourself in the process of loving

someone too much, and forgetting that you are special too.

—Ernest Hemingway

Close relationships are among the most highly valued human

goals (Ben-Ari & Lavee, 2007; Caldwell & Peplau, 1982; Reis &

Shaver, 1988), and there is little debate among scholars that the

satisfaction of the need for closeness and communion is important

for relationship functioning as well as for individual well-being

and health (Baumeister & Leary, 1995; Hofer & Busch, 2011;

Holt-Lunstad, Smith, & Layton, 2010; McAdams & Vaillant,

1982; Patrick, Knee, Canevello, & Lonsbary, 2007; Prager &

Buhrmester, 1998). In adult life, romantic partnerships are mostly

viewed as the closest type of relationship, and couples’ relation-

ship quality and stability have been found to depend on the amount

of closeness that is actually experienced (Aron, Aron, & Smollan,

1992; Berscheid, Snyder, & Omoto, 1989b; Hassebrauck & Fehr,

2002; Smith, Snyder, Trull, & Monsma, 1988). However, humans

also possess individualistic or agentic needs for independence,

mastery, and self-sufficiency (Bakan, 1966; Hagemeyer & Neyer,

2012; McAdams, Hoffman, Day, & Mansfield, 1996; Prager &

Birk Hagemeyer, Institut für Psychologie, Friedrich-Schiller-Universität Jena;

Felix D. Schönbrodt, Department Psychologie, Ludwig-Maximilians-Universität

München; Franz J. Neyer, Institut für Psychologie, Friedrich-Schiller-Universität

Jena; Wiebke Neberich, Institut für Psychologie, Humboldt-Universität zu Berlin,

International Max Planck Research School LIFE; Jens B. Asendorpf, Institut für

Psychologie, Humboldt-Universität zu Berlin.

Wiebke Neberich is now at Affinitas GmbH, Berlin, Germany.

This research was funded by grants of the German Research Foun-

dation to Franz J. Neyer (NE 633/5-1), Jens B. Asendorpf (AS 59/16-1),

and Felix Schönbrodt (SCHO 1334/1-1). The authors thank Marcus

Mund as well as the members of the scientific network “Psychometrics

of Implicit Motives,” Holger Busch, Michael Dufner, Stefan Engeser,

Georg Hosoya, Gregor Kappler, Tobias Koch, Fridtjof Nussbeck, and

Elisabeth Zureck, for valuable comments and fruitful discussions about

the current research.

Correspondence concerning this article should be addressed to Birk

Hagemeyer, Friedrich-Schiller-Universität, Institut für Psychologie, Hum-

boldtstraße 11, D-07743 Jena, Germany. E-mail: birk.hagemeyer@

uni-jena.de

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

2015, Vol. 109, No. 10, 000 0022-3514/15/$12.00 http://dx.doi.org/10.1037/pspi0000031

Buhrmester, 1998; Wiggins, 1991), which may be understood as

antagonistic motivational forces in couple relationships: Whereas

communal needs strive for, and are fulfilled by, closeness, agentic

needs are at risk of being frustrated if closeness comes at the cost

of impaired individual freedom and independence (Baxter, 1990;

J. A. Feeney, 1999; Goldsmith, 1990; Hagemeyer & Neyer, 2012;

Hagemeyer, Neyer, Neberich, & Asendorpf, 2013; Kumashiro,

Rusbult, & Finkel, 2008). A major challenge for couples’ relation-

ship regulation therefore lies in the integration of the partners’

agentic needs, and, in fact, research has shown that too much

closeness can have negative effects on relational as well as indi-

vidual well-being (Frost & Forrester, 2013).

Applying a Person ⫻ Situation approach, the present investiga-

tion looks into the interplay of two factors that are assumed to play

important roles in the regulation of couples’ closeness, and thus for

optimal relationship functioning: the partners’ motive disposi-

tions—that is, individual differences in the strength of the need for

agency—and couples’ living arrangements (coresidence (COR) vs.

living-apart-together (LAT)), which provide differential contexts

for the regulation of dyadic closeness and distance. In two studies,

we investigated the relevance of agency motives for relationship

functioning in COR and LAT couples from between-couple and

within-couple perspectives. Study 1 focused on between-couple

differences in implicit and explicit agency motives in a large

age-heterogeneous sample of heterosexual couples. Study 2 was a

2-week diary study examining the day-to-day dynamics between

motivational states, dyadic closeness, and relationship quality

within COR and LAT couples.

Dyadic Closeness and Couples’ Living Arrangements

Closeness is a heterogeneous concept in relationship research

(see Ben-Ari & Lavee, 2007, and Berscheid, Snyder, & Omoto,

1989a, for overviews of different conceptualizations). In the pres-

ent research context, it is important to differentiate subjective

closeness, which is usually experienced as pleasant and fulfilling,

and thus includes an evaluation of relationship quality (e.g., Aron

et al., 1992), and more objectively defined and externally observ-

able types of closeness like spatial proximity, the amount of time

partners spend together, and dyadic interdependence (i.e., the

extent to which the two partners influence each other; H. H. Kelley

et al., 1983). Whereas the former type of closeness is an individual

variable (although partners may, to some extent, concur in their

subjective closeness ratings), the latter types can only be under-

stood as dyadic characteristics, that is, as variables that describe a

couple and are thus identical for both partners (Kenny, Kashy, &

Cook, 2006). Although the two kinds of closeness may not be

empirically independent (e.g., high subjective closeness without

some extent of objective closeness is hardly imaginable), they are

conceptually distinct. The focus of this article lies on closeness as

dyadic variables that are, in principle, accessible to external ob-

servation. To ensure a clear distinction from individual and sub-

jective sentiments of closeness, we will henceforward use the term

objective closeness for such variables.

Couples’ living arrangements provide an interesting context for

the study of agency motives and relationship functioning, because

they constitute differential relationship conditions that foster or

hinder objective closeness. We focus here on the differences

between two prototypical living arrangements. COR, in which

partners share a common household, is the conventional form of

couple life in Western cultures, and the majority of couples engage

in it (e.g., Asendorpf, 2008; Haskey, 2005). COR is characterized

by high objective closeness as manifested in high amounts of time

spent together, frequent routine interactions, and shared responsi-

bilities. LAT, on the other hand, in which the partners retain

separate households, is an unconventional living arrangement that

affords less objective closeness and more individual freedom,

independence, and privacy. LAT is also a heterogeneous type of

relationship. It includes couples in the early stages of their rela-

tionships who will eventually move in together (or separate) as

well as long-term LATs who live in close proximity to each other

or engage in long-distance relationships (Schneider & Ruckde-

schel, 2003). However, a common feature that distinguishes all

LAT couples from COR relationships is the greater opportunity for

individual freedom and privacy.

To date, LAT has received little attention in psychological

research, but sociologists have found some evidence that this type

of relationship becomes increasingly common and accepted in

Western societies (e.g., Duncan & Phillips, 2010; Haskey, 2005;

Milan & Peters, 2003; Schneider & Ruckdeschel, 2003; Strohm,

Seltzer, Cochran, & Mays, 2009). For the German population,

Asendorpf (2008) analyzed the data from the nationally represen-

tative German Socioeconomic Panel Study and found that the

relative proportion of LAT adults increased from 8.5% in 1992 to

10.9% in 2006. This historical trend was especially pronounced in

middle and late adulthood (age ⬎37 years), for which LAT in-

creased from 4.7% to 8.2%. Analyses by age showed that the

probability of LAT decreased until the age of about 40, and then

remained on a relatively constant level. Furthermore, the proba-

bility of moving in together within an assessment period of 6 years

was markedly lower for participants older than 37 years.

Qualitative interview studies have shed some light on the rea-

sons for having a LAT relationship (Holmes, 2006; Karlsson &

Borell, 2005; Levin, 2004; Milan & Peters, 2003; Roseneil, 2006;

Strohm et al., 2009). Overall, these reasons can be grouped into

external pressures and individual preferences. Among the most

frequently named external pressures are educational or work-

related requirements (especially for long-distance relationships),

care for children from a previous relationship or elderly relatives

living in the household, and the partner’s preference for living

apart. Individual preferences for LAT revolve around gaining

higher autonomy and privacy, and avoiding too much closeness,

intimacy, and daily hassles with the partner.

Sociological research thus provides some support for the as-

sumption that LAT is becoming a viable alternative to the more

conventional and close COR, especially for people who (a) are

around 40 or older, and (b) have a preference for agentic values

and goals. The increasing prevalence of LAT in middle and late

adulthood may be partially explained by a biologically induced

shifting of relationship priorities (Asendorpf, 2008). Around the

age of 40, women have to come to a final decision about having

(biological) children (Heckhausen, Wrosch, & Fleeson, 2001), and

after menopause, the wish to raise a new family is much less likely

to be a viable motivation for moving in with one’s partner. On the

contrary, if a new relationship is commenced at this age and there

are still children from a former relationship in the household, this

will favor the decision for LAT (Levin, 2004). In addition, LAT is

viewed by many who engage in this living arrangement as more

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

HAGEMEYER ET AL.

suitable than COR for the fulfillment of their agentic needs.

Especially if the partners’ households are in close proximity to

each other, LAT may be viewed as providing favorable conditions

for the integration of closeness with the partner and individualistic

commitments, interests, and goals (Karlsson & Borell, 2005;

Levin, 2004).

Agency Motives and Couple Relationships

Motive dispositions differentiate individuals with regard to their

preferences for the attainment of specific classes of incentives

(McClelland, 1985; Schultheiss, 2008). In the case of agency

motives, these incentives refer to self-oriented, individualistic ex-

periences and goals related to independence, self-reliance, domi-

nance, and mastery. According to Bakan (1966), agency entails a

tendency to form separations from others and a focus on the

individual self. Agency is thus reflected in self-protection, self-

assertion, and self-expansion. Drawing on Bakan’s original defi-

nition, Hagemeyer and Neyer (2012; see also Hagemeyer, Neyer,

et al., 2013) conceptualized agency motives in couple relationships

as relationship-specific dispositions that focus on the propensity to

form physical or psychological separations from the partner,

thereby seeking experiences that confirm the self as an indepen-

dent and capable individual. The proposed significance of

relationship-specific agency motives for individual well-being and

relationship functioning is twofold. People with strong agency

motives are assumed to seek out individual activities and distance

from their partners regularly. If, on the one hand, agentic strivings

are successful, this will lead to feelings of reward and satisfaction

on behalf of the actor, although the partner may feel deserted. On

the other hand, if agentic strivings are thwarted by too much

closeness (or other interferences), the resulting feelings of frustra-

tion and dissatisfaction will be more profound the stronger the

motive is. Such negative experiences will, in the long run, also

affect the partner’s satisfaction as they may spark quarrels and

conflicts (Hagemeyer & Neyer, 2012).

Implicit and Explicit Motives

We use the term agency motives in its plural form because we

differentiate between implicit and explicit motive dispositions.

According to McClelland’s dual motives theory, the two kinds of

motives pertain to distinct motivational systems and guide differ-

ent kinds of behavior (McClelland, 1980; McClelland, Koestner, &

Weinberger, 1989). On the one hand, implicit motives are assumed

to operate largely outside awareness and guide operant (or spon-

taneous) behavior. A key function of the implicit motivational

system is the regulation of affective processes during goal pursuit.

Thus, implicit motives not only direct and energize behavior but

also influence its emotional consequences. Affective experiences

of reward after successful goal attainment, but also frustration after

failure, increase with the strength of the corresponding implicit

motive (Brunstein, Schultheiss, & Grässmann, 1998; Schultheiss,

Jones, Davis, & Kley, 2008). Because they are not consciously

represented, implicit motives are measured in an indirect way,

usually with Picture Story Exercises (PSEs; e.g., Schultheiss &

Pang, 2007), which are research variants of the Thematic Apper-

ception Test (TAT; Morgan & Murray, 1935) (Morgan & Murray,

1935). Explicit motives, on the other hand, are assumed to guide

respondent (or controlled) behavior, such as deliberate choices

between a limited number of alternatives (Brunstein, 2008; Mc-

Clelland et al., 1989). In contrast to implicit motives, explicit

motives pertain to an individual’s cognitively elaborated self-

concept and guide behavior via conscious decision making and

goal selection. They reflect a person’s social values, and their

attainment provides a sense of meaning and coherence rather than

hedonic experiences (Cantor & Malley, 1991). Because they are

represented in a propositional format, explicit motives can be

assessed by self-reports, usually via standardized questionnaires

(e.g., Hagemeyer, Neyer, et al., 2013; Schönbrodt & Gerstenberg,

2012). Applied to relationship-specific agency motives, dual mo-

tives theory suggests that individuals are characterized by two

motive dispositions that are similar in the aspiration of individu-

alistic goal contents, but different in their functional underpin-

nings.

The theoretical distinction between implicit and explicit motives

originated from the repeated finding that indirect and direct motive

measures are largely uncorrelated (McClelland, 1980; McClelland

et al., 1989). A recent meta-analysis confirmed this for the most

frequently studied motives of achievement, power, and affiliation

using 56 independent samples from six decades of motive research

(Köllner & Schultheiss, 2014). The average correlations between

implicit and explicit motives were below .15 in all three domains.

According to this low convergence, the interindividual variation of

intrapersonal constellations of implicit and explicit motives is

nearly unrestricted. This raises the question about the conse-

quences of congruent or incongruent implicit– explicit motive con-

stellations (McClelland et al., 1989). Because of their common

goal content and complementary functional properties, the two

motivational systems may amplify each other’s effects, if implicit

and explicit motives are congruently strong in a person. In fact, a

recent study on communal motives suggests that motive (in)con-

gruence plays a significant role in relationship functioning, as a

strong implicit motive increased the positive actor effects of the

explicit motive on relationship satisfaction. Moreover, incongruent

intrapersonal motive constellations predicted a heightened risk of

relationship breakup over the time of 1 year (Hagemeyer, Ne-

berich, Asendorpf, & Neyer, 2013). These findings correspond to

other studies outside the area of couple research showing that

intrapersonal motive constellations are relevant for individual

well-being (e.g., Brunstein et al., 1998; Hofer, Chasiotis, & Cam-

pos, 2006). However, studies on the consequences of agentic

motive (in)congruence for couple relationships are missing.

Agency Motives and Relationship Functioning

Previous research has shown that both implicit and explicit

agency motives are related to various aspects of relationship func-

tioning. For the relationship-specific implicit agency motive,

Hagemeyer and Neyer (2012) found negative intra- and interper-

sonal effects of both partners’ motives on their accounts of rela-

tionship satisfaction. Notably, these effects were independent of

the complementary communal motive, thus corroborating the

unique significance of agentic motives for couple relationships. In

addition, a strong implicit need for power (a facet of agency) in

men has been related to impaired relationship satisfaction, a

heightened probability of separation, unrestricted sociosexuality,

and aggressive sexual behavior (Hofer et al., 2010; Stewart &

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

AGENCY MOTIVES AND RELATIONSHIP FUNCTIONING

Rubin, 1976; Zurbriggen, 2000). Explicit agentic motives and

goals depicting the motivation to establish independence, social

distance, and solitude are also associated with indicators of poor

relationship functioning, such as low relationship satisfaction,

avoidant romantic attachment, and low perceived support from

one’s partner (Craddock, 1997; Dewitte & De Houwer, 2008;

Hagemeyer, Neyer, et al., 2013). Thus, although, to date, the

literature on the consequences of agency motives in couple rela-

tionships is not very extensive, the existing studies are quite

consistent in identifying strong agency motives as potential prob-

lems for couple relationships in general. This applies to both

implicit and explicit motive dispositions.

In terms of Karney and Bradbury’s model of couples’ satisfac-

tion and stability, agency motives therefore qualify as enduring

vulnerabilities, that is, as dispositions that impair relationship

quality (a) by increasing the likelihood of stressful events (e.g.,

conflicts), and (b) by weakening the couple’s capacity to cope with

and adapt to such events (vulnerability-stress-adaptation model;

Karney & Bradbury, 1995). However, previous research falls short

of accounting for two important aspects of the interplay between

agency motives and relationship functioning. First, the role of

intrapersonal constellations of implicit and explicit agency motives

in relationship functioning has not been examined yet. Second, no

prior studies have considered between-couple differences in the

conditions of objective closeness provided by different living

arrangements as potential moderators.

The Present Research

We conducted two studies examining the relevance of implicit

and explicit agency motives for relationship functioning in hetero-

sexual COR and LAT couples from a between-couple (Study 1)

and a within-couple (Study 2) perspective. COR and LAT are

introduced to motive research as two prototypical living arrange-

ments that provide a great deal of the proximal relationship con-

texts in which the partners’ motive dispositions operate. In Study

1, we relied on a large age-heterogeneous sample of German

couples, and Study 2 employed a subsample of these couples for a

2-week diary study. COR couples were required to have a serious

romantic relationship and share the main household, irrespective of

their marital status (married or unmarried). LAT couples were

likewise required to see themselves as being in a serious relation-

ship, and, in addition, the partners’ separate households had to be

in relative proximity to each other, such that partners were able to

see each other every day if they wanted to. The former criterion

was used to exclude uncommitted dating couples, and the latter

was used to exclude long-distance relationships, which are often

living apart because of external pressures (e.g., job or educational

affordances) rather than individual preferences (Schneider &

Ruckdeschel, 2003). In addition, long-distance relationships differ

from proximal LATs and CORs in their ability to establish close-

ness on a daily basis and are therefore viewed here as a third and

distinct type of relationship that was not included in the current

research project.

Study 1 tested (a) whether individual differences in relationship-

specific agency motives predict the selection of living arrangement

(LAT vs. COR), and (b) whether cross-sectional and prospective

associations between the partners’ agency motives and indicators

of relationship functioning are moderated by couples’ living ar-

rangements. Study 2 complements these between-couple analyses

(a) by linking agentic motive dispositions to agentic motivational

states, and (b) by examining the consequences of day-to-day

variations of agentic motivational states and objective closeness

within couples for their daily relationship satisfaction, thereby

capturing potential differences in the motivational dynamics of

relationship functioning between COR and LAT couples.

Study 1

Hypothesis 1: Selection of Living Arrangement

People in LAT relationships often name agentic reasons for

having this living arrangement, such as career goals, greater per-

sonal freedom, and privacy (Holmes, 2006; Karlsson & Borell,

2005; Levin, 2004; Roseneil, 2006). Thus, individuals with strong

agentic motives who particularly value these qualities are expected

to prefer LAT over COR. Especially in the postreproductive life

phase (women’s age ⱖ40 years), when the wish to raise a family

together becomes less relevant for relationship decisions, agency

motives are expected to predict the choice of LAT. We tested

Hypothesis 1 with regard to explicit and implicit motive disposi-

tions, because we assume that the motivational processes under-

lying couples’ decisions on living arrangements involve both de-

liberate choices as well as more subtle and spontaneous

evaluations. The vast majority of intimate relationships start out

living apart. If the relationship seems to work out, most couples

come to a point when moving in together becomes an important

question, which is contemplated and debated among the partners,

and finally a decision is made. This decision reflects a choice

between two (or three) alternatives: staying LAT or moving in

together (or breaking up). According to dual motives theory, such

choices are expected to be guided more by explicit than implicit

motives (McClelland et al., 1989). These debates and decisions

involve explicit arguments exchanged in interactions with the

partner, but also gut feelings about preferences that are based on

less accessible, implicit motivations, and in some couples, a

change of living arrangement may not even become a topic at all.

Such more subtle, implicit selection processes are expected to

depend on implicit agency motives. In addition, we tested the

contribution of the interaction between implicit and explicit

agency motives to examine potentially amplifying effects of mo-

tive congruence.

Hypothesis 2: Moderation by Living Arrangement

We expected that living arrangement (LAT vs. COR) is a

moderator of the overall negative associations between

relationship-specific agency motives and relationship functioning

that were reported in prior studies (Hagemeyer & Neyer, 2012;

Hagemeyer, Neyer, et al., 2013). Specifically, we expected nega-

tive intra- and interpersonal effects of agency motives on the

partners’ subjective accounts of their relationship quality to be

stronger in COR couples. In addition, the amount of dyadic con-

flicts, a more behavioral indicator of relationship functioning,

should show stronger associations with agency motives in COR

couples. It is a fundamental assumption of motivational psychol-

ogy that the expression and the consequences of motive disposi-

tions depend on contextual variables (e.g., Atkinson, 1981). Al-

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

HAGEMEYER ET AL.

though we believe that the obligations that come with any serious

relationship present a challenge to individuals with strong agency

motives, this challenge should become a real problem under cir-

cumstances of obligatory high objective closeness as provided by

COR. Thus, irrespective of whether the living arrangement was

deliberately chosen or not, strong agency motives are at a higher

risk of being frustrated in COR relationships. To render a com-

prehensive account of the differential role of agentic motivational

dispositions, we tested Hypothesis 2 for implicit and explicit

motives as well as for the Implicit ⫻ Explicit interaction.

Method

Participants and procedure. We used data from a larger

research project on distance regulation in couples (see Hagemeyer,

Neberich, et al., 2013, Hagemeyer & Neyer, 2012, and Hage-

meyer, Neyer, et al., 2013, for previous publications partially

based on the current data). A sample of 714 German heterosexual

couples was recruited by a mass mailing to households in a

metropolitan area (Berlin-Charlottenburg) and in a rural area

(counties of Lower Saxony), and subsequent nationwide press

announcements and newspaper ads. Couples were invited to a

research project on different living arrangements and applied for

participation by phone. A brief telephone prescreening and under-

sampling of younger couples made sure (a) that the couples were

either COR or LAT, (b) that all participants spoke German flu-

ently, and (c) that the sample comprised similar numbers of cou-

ples from each age decade between 18 and 68 years (as defined by

women’s age). Participants could opt for a paper-and-pencil ver-

sion, which they received via mail, or an online questionnaire.

Participating couples were compensated with €20 and offered an

individual, confidential feedback for each of the two partners about

their personality and relationship. In the present study, we included

all couples with complete data on implicit and explicit agency

motives. The resulting sample comprised 548 couples (332 CORs,

216 LATs), with age ranging from 18 to 73 years (M ⫽ 40.4, SD ⫽

14.1). The partners’ reports on relationship duration were highly

correlated (r ⫽ .995), and thus averaged, ranging from 1 month to

53 years (M ⫽ 11.2 years, SD ⫽ 12.5). Participants reported 0 to

4 own children (62% had at least one child), and 42% were married

to their current partners. Fifty-eight percent had a German Abitur

(high school degree) or a higher educational degree.

One year after the first assessment (Time 1 [T1]), participants

were contacted again and asked to participate in a follow-up study

(Time 2 [T2]). As an incentive, participants were offered to take

part in two lotteries with prices amounting to a total of €3,700.

Response rate was 77% within the 548 T1 couples (at least one

partner of a couple responded). In the present longitudinal analy-

ses, we included T2 data from all stable couples who maintained

their T1 living arrangements and for which at least one partner

gave information on the current relationship quality. This was true

for 235 COR and 100 LAT couples. Missing data were treated with

the full information maximum likelihood procedure (Enders,

2010). A logistic regression predicting participation in the

follow-up study by all T1 variables of this investigation showed

that T2 couples were older (odds ratio [OR]⫽ 1.536, p ⬍ .001),

reported fewer conflicts (OR ⫽ 0.774, p ⫽ .002) and were less

likely to be LAT (OR ⫽ 0.393, p ⬍ .001) than T1 couples who

dropped out. This pattern is plausible, reflecting that dropout was

partially explained by relationship break-up and LAT couples

moving in together between T1 and T2.

Measures. The questionnaires administered at T1 and T2 in-

cluded a variety of measures of diverse aspects of the relationship,

as well as both partners’ motives and personality. In the following,

only the methods relevant for the present study are described.

Agency motives. Relationship-specific implicit and explicit

agency motives were assessed at T1. The implicit agency motive

was measured with the Partner-Related Agency and Communion

Test (PACT; Hagemeyer & Neyer, 2012), a relationship-specific

variant of the PSE, which is the most frequently applied approach

to the assessment of implicit motives (Schultheiss & Pang, 2007).

The PACT is comprised of eight picture cues showing interper-

sonal situations in the form of either line drawings or distorted

photographs. Participants are instructed to invent fantasy stories

about romantic relationships that describe the depicted scenes.

Three questions regarding the protagonist of the respective story

have to be answered in response to each picture cue: “What is

important to this person in this situation, and what is he/she

doing?,” “How is the person feeling in this situation, and how are

his/her feelings for his/her partner?,” and “Why is the person

feeling this way?” Participants’ answers are scored by trained

coders for the occurrence of agentic contents that pertain to one of

eight thematic categories referring to power, independence, and

individuality in couple relationships: Dominance (i.e., personal-

ized power), Help for Partner (i.e., socialized power),

Differen-

tiation (i.e., elaborated differences between partners), Indepen-

dence (i.e., pursuit of individualistic interests), Self-Growth (i.e.,

individualistic development), Alternative Contacts (i.e., social ac-

tivities in absence of the partner), Conflict (i.e., quarrels related to

differences between partners), and Fear of Dependency (i.e., neg-

ative affect related to overdependence). Motive raw scores are

derived from summing up the frequencies of all agentic categories

across all eight PACT tasks

(for details on PACT administration

and scoring, see Hagemeyer & Neyer, 2012).

In the present sample, PACT answers were scored by four trained

coders who worked on different cases that were randomly assigned to

them. Absolute agreement among coders as assessed in a random

subsample of 65 cases was acceptable (intraclass correlation ⫽ .76,

p ⬍ .001). Participants generated 0 to 11 agentic themes (M ⫽ 4.23,

SD ⫽ 2.09). Women had higher raw scores than men (see Table 1),

The Help for Partner category is scored for contents that indicate

support given out of a concern for one’s own power rather than out of

sympathy and compassion for the partner. However, one might argue that

this category introduces communal contents into the implicit agency mo-

tive score, which might lead to biased results. Therefore, all major analyses

of this article were repeated with a reduced score of the implicit agency

motive, omitting the Help for Partner category. The two scores correlated

at .95 and the results of all control analyses were very similar to the main

analyses.

Coefficient ␣ for the implicit agency motive score was modest (.51 for

women and .52 for men). However, internal consistency of TAT-like

motive measures such as the PACT is typically low. It is plausible that the

response processes in these tests are complex and do not follow the

assumptions of classical test theory (e.g., dynamic fluctuations of responses

across pictures, idiosyncratic instigating potentials of picture cues), which

makes measures of internal consistency questionable indicators of test

reliability. Thus, low internal consistency does not preclude validity of

TAT-like measures (see Lang, 2014, for an elaborated account of this

issue).

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

AGENCY MOTIVES AND RELATIONSHIP FUNCTIONING

but this difference was mainly attributable to women’s longer answers

and was not significant after word count was controlled, t(547) ⫽

1.632, p ⫽ .103, d= ⫽ 0.05. Because the raw motive scores were

correlated with the length of the answers, r ⫽ .28, p ⬍ .001, word

count was partialed out in the total sample before further analyses to

avoid confounding with verbal fluency. The correlation of these

corrected motive scores between partners was significant (see Table

1), indicating modest similarity of implicit agency motives within

couples.

The explicit agency motive was measured with the Desire for Being

Alone scale from the ABC of Social Desires questionnaire (Hage-

meyer, Neyer, et al., 2013). Participants reported, on a 7-point scale

(1 ⫽ never,4⫽ sometimes,7⫽ always), the frequency of appetitive

(four items; e.g., “I prefer being alone”) and aversive (four items; e.g.,

“I avoid being completely alone”) experiences regarding solitude and

self-reliance. Aversion items were reversed, and all eight items were

averaged to obtain explicit motive scores. Internal consistencies were

good (␣⫽.83 in men and .84 in women), and previous research

established high 1-year stabilities of the scale (r ⬎ .70; Hagemeyer,

Neyer, et al., 2013). As displayed in Table 1, women reported slightly

stronger explicit agency motives than men, t(547) ⫽ 5.201, p ⬍ .001,

d= ⫽ 0.16, and the between-partner correlation again indicated mod-

estly similar motives within couples. However, as expected from dual

motives theory, explicit and implicit agency motives were nearly

unrelated within individuals.

Although they do not seem to address identical contents, both the

direct and the indirect motive measure were developed to assess

individual differences in the propensity to create distance to one’s

partner, which is a core aspect of relationship-specific agency moti-

vation (Hagemeyer & Neyer, 2012; Hagemeyer, Neyer et al., 2013).

To assess the explicit motive, this propensity could be directly ad-

dressed in a self-report questionnaire. To assess the implicit motive

with the PACT, we identified, in a previous study, themes and

expressions in fantasy stories about romantic relationships that reflect

distance motivation (Hagemeyer & Neyer, 2012). These content cat-

egories comprised themes of power and independence in the relation-

ship. An experimental validation study showed that both power and

independence categories occurred more frequently in stories from

participants who were primed with the motivation for solitude and

distance to their partners compared with a control group (see Hage-

meyer & Neyer, 2012, for a detailed description of PACT develop-

ment). Thus, despite their obvious differences, both measures address

the same core aspect of relationship-specific agency motives, but on

different levels of representation (implicit vs. explicit).

Subjective relationship quality (SRQ). We view SRQ as a

hierarchically organized multifaceted construct. It entails various

evaluations of specific relationship aspects (e.g., commitment or sex-

uality), which can be diverging, but are all informed by a generalized

evaluation of the relationship (see Fletcher, Simpson, & Thomas,

2000, for empirical support and a comparison with other conceptual-

izations of SRQ). We thus opted for an extensive measurement

strategy and assessed various aspects of SRQ using single items and

brief scales at T1 and T2.

At T1, SRQ was assessed with five variables: General relationship

satisfaction and sexual satisfaction were measured with the items,

respectively, “How satisfied are you with your relationship in gen-

eral?” and “How satisfied are you with the sexuality in your relation-

ship?” Both items were answered on an 11-point scale (0 ⫽ not at all,

10 ⫽ a great deal). Sexual interest in the partner was assessed with

three items: “How often do you think of sex with your partner?,”

“How often do you feel sexual lust for your partner?,” both rated on

a 7-point frequency scale (1 ⫽ never,4⫽ about once a week,7⫽

several times a day), and “How strong is your lust for your current

partner compared with your lust for other potential sex partners?,”

rated on a 7-point scale (1 ⫽ more lust for others,7⫽ more lust for

partner). The three items were z-standardized and averaged. Commit-

ment to the partner was assessed with one item: “Could you imagine

that your current partner is the one for the rest of your life?,” rated on

a 5-point scale (1 ⫽ no, never;2⫽ rather no;3⫽ sometimes;4⫽

rather yes;5⫽ yes, fully so). Finally, perceived available support was

assessed with an item adapted from Asendorpf and Wilpers (1998):

“If I have problems, I would turn to this person to talk about my

problems.” This item was rated for six types of potentially supporting

persons (partner, mother, father, children, friends, colleagues) on a

5-point frequency scale (1 ⫽ never,5⫽ always). Only the ratings

concerning support from one’s partner were used.

At T2, general relationship satisfaction, sexual satisfaction, sexual

interest in the partner, and commitment were assessed in exactly the

same way as at T1. However, perceived available support was not

reassessed, but two additional aspects of relationship quality were

measured. First, emotional closeness to the partner was measured with

the item “How close (emotionally) to your partner are you currently

feeling?,” answered on a 7-point scale (1 ⫽ very distant, 7 ⫽ very

close). Second, COR and LAT couples reported their satisfaction with

Table 1

Descriptive Statistics and Correlations (Study 1)

Variables

Men Women Correlations

MSD M SD 12345

1. Implicit agency motive 4.01 2.04 4.45 2.12 .150

ⴱⴱ

.083 ⫺.132

ⴱⴱ

⫺.230

ⴱⴱ

.162

ⴱⴱ

2. Explicit agency desire 4.02 1.00 4.29 0.96 .083 .204

ⴱⴱ

⫺.284

ⴱⴱ

⫺.293

ⴱⴱ

.085

ⴱ

3. T1 SRQ 0.02 0.71 ⫺0.01 0.75 ⫺.289

ⴱⴱ

⫺.175

ⴱ

.548

ⴱⴱ

.734

ⴱⴱ

⫺.411

ⴱⴱ

4. T2 SRQ 0.01 0.76 ⫺0.03 0.80 ⫺.223

ⴱⴱ

⫺.092 .618

ⴱⴱ

.611

ⴱⴱ

⫺.386

ⴱⴱ

5. Dyadic conflicts

1.95 0.55 1.95 0.55 .157

ⴱⴱ

⫺.101

ⴱ

⫺.398

ⴱⴱ

⫺.369

ⴱⴱ

—

Note. N varies between 204 and 548 because of missing T2 data. Male correlations are above the diagonal. Female correlations are below the diagonal.

Correlations within couples are on the diagonal. Correlations of the implicit agency motive were controlled for word count. SRQ ⫽ subjective relationship

quality; T1 ⫽ Time 1; T2 ⫽ Time 2.

Because dyadic conflicts is a dyadic variable, M and SD are identical for men and women, and no within-couple correlation is reported.

ⴱ

p ⬍ .05.

ⴱⴱ

p ⬍ .01.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

HAGEMEYER ET AL.

the living arrangement responding to the item “How satisfied are you

with your current living arrangement?” on a 7-point scale (1 ⫽ not at

all, 7 ⫽ completely).

To obtain comparable SRQ indices at T1 and T2, the four variables

that were assessed at both measurement occasions— general relation-

ship satisfaction, sexual satisfaction, sexual interest in the partner, and

commitment—were z-standardized across men and women, and the

standardized scores were averaged within each time point. For the

T1-SRQ index, internal consistency was acceptable (␻⫽.73 for

women and .69 for men),

and there was no significant mean sex

difference, t(547) ⫽ 0.957, p ⫽ .339, d= ⫽ 0.03. The between-partner

correlation (see Table 1) indicated substantial similarity in SRQ, but

also individual variation within couples. The T2-SRQ index also had

acceptable internal consistency (␻⫽.81 for women and .77 for men),

showed no mean sex difference, t(203) ⫽ 0.750, p ⫽ .452, d= ⫽ 0.04,

and the between-partner correlation was similar to T1. The correla-

tions between the indices at T1 and T2 indicated high 1-year stability

of individual differences in SRQ for both women and men (see Table

1).

Dyadic conflicts. The amount of dyadic conflicts was assessed

at T1 in nine different domains derived from modified and amended

categories from Argyle and Furnham’s (1983) Sources of Conflict

scale: Finances, Communication, Planning of Visits and Activities,

Long-Term Life Planning, Other Persons (But Not Jealousy), Dealing

With Other Potential Partners, Shared Responsibilities, Personal Prob-

lems, and Living Arrangement. Participants rated how often they

experienced conflicts with their partners in each of these domains on

a 5-point scale (1 ⫽ not at all,5⫽ very much). The two partners’

ratings in the nine conflict domains correlated between .24 and .53 (all

ps ⬍.001), indicating substantial agreement between partners, but also

individual divergence because of subjective influences. Because we

were interested in the actual amount of conflicts that characterizes a

couple, we obtained a dyadic index by averaging both partners’

reports (internal consistency ␻⫽.84). As displayed in Table 1,

amount of conflicts showed only modest correlations with SRQ, and

is thus not statistically redundant as an additional indicator of rela-

tionship functioning.

Results and Brief Discussion

Analysis strategy. To test our two hypotheses, we employed

logistic regression (for Hypothesis 1) and linear dyadic path models

with multiple-group analysis (extensions of the actor–partner interde-

pendence model; Kenny et al., 2006; for Hypothesis 2) using SPSS

22 and Mplus 7 (Muthén & Muthén, 1998 –2012). All contin-

uous variables (predictors, covariates, criteria) were

z-standardized across men and women and across living arrange-

ments. Resulting (unstandardized) coefficients can thus be inter-

preted as standardized in the total sample.

Covariates. To control for potential confounding variables, age,

relationship duration, and the presence of additional cohabitants (e.g.,

children, parents, relatives, or friends who share the household with

the participants) were entered as covariates in all analyses. The

rationale for these controls is twofold. First, all three covariates

showed differences between living arrangements in the present sam-

ple: LATs were, on average, younger than CORs (M ⫽ 37.0 years,

SD ⫽ 13.2 vs. M ⫽ 42.6 years, SD ⫽ 13.8), t(1,094) ⫽ 6.548, p ⬍

.001, d ⫽ 0.41, reported shorter relationship durations (M ⫽ 4.3 years,

SD ⫽ 6.5 vs. M ⫽ 15.7 years, SD ⫽ 13.3), t(512.269) ⫽ 13.313, p ⬍

.001, d ⫽ 0.89, and were more likely to live with additional cohabi-

tants (57.9% vs. 43.4%), ␹

(1) ⫽ 11.004, p ⫽ .001, Cramer’s V ⫽

.142. Second, these variables showed associations with predictor

(agency motives) and criterion variables (SRQ, dyadic conflicts)

within the COR and LAT groups (r ranging from –.28 to .29). In order

to avoid confounds on both levels, we controlled for age, relationship

duration, and additional cohabitants in both the between-group and

the within-group analyses. Like relationship duration, age was highly

correlated between members of a couple (r ⫽ .94), and, for parsi-

mony, average age was used as covariate. Additional cohabitants was

entered as a dyadic dummy-coded variable (0 ⫽ no additional cohab-

itants, 1 ⫽ at least one additional cohabitant).

Sex differences. In order to determine whether men and women

had to be distinguished in the dyadic analyses, we conducted a priori

omnibus tests of sex differences as proposed by Kenny et al. (2006):

Using structural equation modeling, all predictors and criteria of a

given model are correlated with each other and each mean, each

variance, and each covariance is set equal between the sexes. If this

model fits the data well, further analyses can be conducted assuming

indistinguishable partners, which increases parsimony and statistical

power of the analyses. If the model does not fit the data, male and

female variables have to be distinguished in further analyses.

Hypothesis 1: Selection of living arrangement. Hypothesis 1

stated that strong agency motives predict a preference for LAT over

COR, especially in postreproductive couples. To test this hypothesis,

we conducted a hierarchical logistic regression analysis of couples’

living arrangement (COR vs. LAT) on their implicit and explicit

agency motives and (post)reproductive age group. Age group was

entered as a dyadic dummy variable (reproductive couples ⫽ wom-

en’s age ⬍40 were coded 0; postreproductive couples ⫽ women’s

age ⬎39 were coded 1; see Heckhausen et al., 2001, for a similar

demarcation of the developmental deadline for motherhood). As in-

dicated by the omnibus test of sex differences, male and female

motives had to be distinguished in the logistic regression model:

␹

(28) ⫽ 87.306, p ⬍ .001, comparative fit index (CFI) ⫽ 0.962, root

mean square error of approximation (RMSEA) ⫽ .062. In the first

step of the logistic regression, relationship duration and additional

cohabitants were entered as covariates. Next, age group and both

partners’ agency motives were entered. This block of predictors made

an additional significant contribution to the explanation of living

arrangement above and beyond the covariates, ␹

(5) ⫽ 49.875, p ⬍

.001. More specifically, the probability of being LAT was increased

for the postreproductive group and for couples for which women

reported a high explicit desire for agency (see Table 2). In a third step,

the product interactions between motives and age group were added

to the model, which rendered a further increase in predictive power,

␹

(4) ⫽ 16.310, p ⫽ .003. Two interaction effects became significant

(see Table 2). Women’s implicit agency motive predicted an in-

creased probability of LAT in postreproductive couples (b ⫽ 0.344,

p ⫽ .048, OR ⫽ 1.411), but not in the reproductive group

(b ⫽⫺0.126, p ⫽ .421, OR ⫽ 0.881). Similarly, men’s explicit

Note that for the indices of subjective relationship quality and dyadic

conflicts, we report coefficient ␻ instead of the more commonly used ␣.In

contrast to ␣, ␻ does not rely on the assumption of essential ␶-equivalence

(i.e., that all items measure the underlying construct equally well). Because

this assumption is likely untenable, especially in the case of heterogeneous

scales such as these indices, ␣ would tend to underestimate reliability, and

␻ is the more appropriate measure (K. Kelley & Cheng, 2012).

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

AGENCY MOTIVES AND RELATIONSHIP FUNCTIONING

agency motive predicted LAT in the postreproductive (b ⫽ 0.468,

p ⫽ .009, OR ⫽ 1.596), but not in the reproductive age group (b ⫽

0.154, p ⫽ .327, OR ⫽ 1.167). Finally, in a fourth step, we examined

whether the interactions between implicit and explicit motives and the

three-way interactions Implicit ⫻ Explicit ⫻ Age Group would make

incremental contributions. This was not the case, ␹

(4) ⫽ 1.244, p ⫽

.871. Thus, we dropped these interaction terms from the model. Steps

2 and 3 of the final logistic regression model, as depicted in Table 2,

rendered a substantial prediction of living arrangement (combined

change in Nagelkerke’s R

⫽ .115).

In summary, we found support for Hypothesis 1. After controlling

for relationship duration and additional cohabitants, postreproductive

couples were more than twice as likely to be LAT as reproductive

couples. In addition, strong agency motives were related to an in-

creased probability of being LAT. Women’s explicit agentic desire

was predictive irrespective of age, and their implicit motive was

associated with being LAT only in postreproductive age. Men’s

explicit desire was associated with being LAT, again, only in

postreproductive age. These findings corroborate the relevance of

(post)reproductive age and agency motives for couples’ selection of

living arrangements. Both partners’ motives contributed to the pre-

diction of living arrangement. Women’s explicit motive had a stron-

ger simple effect than men’s (Z ⫽ 3.193, p ⫽ .001), but no other sex

difference between corresponding effects reached significance (Z ⱕ

0.896, p ⱖ .370).

Hypothesis 2: Moderation by living arrangement. We ex-

pected that living arrangement (LAT vs. COR) moderates intra-

and interpersonal associations between agency motives and indi-

cators of relationship functioning such that strong motives have

more negative effects in COR relationships. To test this hypothe-

sis, we ran three dyadic path models with multiple-group analysis,

in which T1-SRQ, T2-SRQ, and amount of conflicts, respectively,

were regressed on agency motives, controlling for age, relationship

duration, and additional cohabitants.

Model 1: T1-SRQ. In Model 1, both partners’ accounts of

SRQ at T1 were regressed on both partners’ concurrently assessed

implicit and explicit agency motives and on the Implicit Motive ⫻

Explicit Motive interaction within persons. The a priori omnibus

test of sex differences indicated that men and women had to be

distinguished in the analysis, ␹

(20) ⫽ 88.516, p ⬍ .001, CFI ⫽

0.820, RMSEA ⫽ .079. A multiple-group analysis assuming all

corresponding actor and partner effects of the motive variables on

SRQ to be equal across COR and LAT couples did not fit the data

well, and thus indicated overall moderation by living arrangement,

␹

(12) ⫽ 29.756, p ⫽ .003, CFI ⫽ 0.945, RMSEA ⫽ .073. To

determine which single effects in the model were moderated, we

conducted further multiple-group analyses, each with one single

effect constrained to be equal across CORs and LATs. Table 3

displays the results of chi-square difference tests from these mod-

els. Aside from significant group differences in the intercepts of

men and women, these analyses showed that living arrangement

significantly moderated four effects: The simple actor and partner

effects of men’s implicit motive and the actor and partner effects

of men’s Implicit ⫻ Explicit interaction.

Inspection of the path coefficients from the unconstrained model

(see Table 3) revealed that the moderation effects for the male

Because relationship duration, age, and additional cohabitants were

associated with living arrangement, the moderating effect of living arrange-

ment could be attributable to these variables. To rule out such potential

confounds, we reran Model 1 using propensity score matching to balance

the COR and LAT groups with regard to the three covariates (Rosenbaum

& Rubin, 1983; Thoemmes, 2012). A weighted multiple-group model

assuming the four effects that were moderated by living arrangement in the

original analysis to be equal across CORs and LATs did not yield a good

fit, ␹

(4) ⫽ 17.485, p ⫽ .002, CFI ⫽ .914, RMSEA ⫽ .119, thus

confirming moderation by living arrangement in the balanced sample.

Inspection of the path coefficients from the unconstrained model showed

that the moderation was in the expected direction of stronger negative

effects of agency motives in CORs than in LATs (see Table S8 of the

online supplemental materials). Multiple-group tests of single-path mod-

eration revealed that three of four moderation effects remained substantial

in the balanced analysis, ␹

(1) ⱖ 3.128, p ⱕ .077. Only the moderation of

the actor effect of men’s Implicit Motive ⫻ Explicit Motive interaction was

reproduced as a nonsignificant trend, ␹

(1) ⫽ 1.289, p ⫽ .256. In sum,

men’s agency motives had stronger negative actor and partner effects on

SRQ in COR couples, even after group differences in relationship duration,

age, and additional cohabitants were controlled. This supports our inter-

pretation that the moderating effects of living arrangement are largely

attributable to group differences in objective closeness and not to other

group characteristics.

Table 2

Logistic Regression of Living Arrangement (COR vs. LAT) on (Post)Reproductive Age Group, and

Agency Motives (Hypothesis 1)

Predictors b (SE) pOR 95% CI

Constant ⴚ2.103 (0.396) ⬍.001 0.122 —

Age group 0.869 (0.307) .003 2.385 [1.387, 4.102]

Male implicit motive ⫺0.070 (0.127) .572 0.932 [0.717, 1.212]

Female implicit motive ⫺0.091 (0.131) .472 0.913 [0.704, 1.184]

Male explicit desire 0.094 (0.128) .437 1.099 [0.844, 1.430]

Female explicit desire 0.371 (0.124) .002 1.449 [1.117, 1.879]

Male Implicit ⫻ Age Group 0.206 (0.280) .433 1.229 [0.757, 1.997]

Female Implicit ⫻ Age Group 0.562 (0.282) .028 1.755 [1.106, 2.785]

Male Explicit ⫻ Age Group 0.524 (0.288) .048 1.689 [1.051, 2.716]

Female Explicit ⫻ Age Group 0.471 (0.335) .119 1.601 [0.948, 2.705]

Note. COR was coded 0. LAT was coded 1. Reproductive age (women’s age ⬍ 40 years) was coded 0.

Postreproductive age (women’s age ⬎ 39 years) was coded 1. Significant effects (p ⬍ .05) are printed in

boldface. Not displayed: covariates relationship duration and additional cohabitants. COR ⫽ coresidence;

LAT ⫽ living-apart-together; b ⫽ unstandardized regression weights; SE ⫽ bootstrapped standard errors of

regression weights (10,000 resamples); OR ⫽ odds ratio; 95% CI ⫽ 95% confidence intervals of odds ratios.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

HAGEMEYER ET AL.

agency motives were all in the expected direction of stronger

negative effects in COR couples. Specifically, simple actor and

partner effects of men’s implicit agency motive were significantly

stronger in CORs. The corresponding effects of men’s explicit

motive tended to be more negative in CORs, but the group differ-

ences were not significant. In addition, the simple actor and partner

effects of men’s motives were further qualified by significant

Implicit ⫻ Explicit interaction effects in COR couples. In LAT

couples, by contrast, a nonsignificant trend for a positive actor

interaction effect was observed. Figure 1 (Panels A and C) displays

this pattern of results for the male agency motives. In COR, but not

LAT couples, the overall negative actor effect of the implicit

agency motive was increased if accompanied by a strong (M ⫹ 1

SD) explicit motive (simple slope b ⫽⫺0.372, p ⬍ .001) and

decreased in the presence of a weak (M–1 SD) explicit motive

(b ⫽⫺0.078, p ⫽ .224; Panel A). The same was true for the

corresponding partner effect (b ⫽⫺0.363, p ⬍ .001, and

b ⫽⫺0.027, p ⫽ .674). Further, the average levels of SRQ as

reported by women and men were higher in COR couples (see the

intercepts in Table 3). However, in COR couples with men high in

both implicit and explicit agency motives (M ⫹ 1 SD), male as

well as female partners’ predicted SRQ was even slightly lower

than in the average LAT couple. The pattern of results emerging

for women looked rather different (Figure 1, Panels B and D).

First, women’s implicit and explicit motives also had negative

actor and partner effects on SRQ, but these effects were compa-

rable in size across COR and LAT couples (see Table 3). Second,

there were no significant Implicit ⫻ Explicit interaction effects of

women’s motives in neither COR nor LAT. Taken together, the

amounts of variance explained by male and female motive vari-

ables in the unconstrained model above and beyond the covariates

were markedly higher in COR than in LAT couples (17.7% in male

and 15.6% in female SRQ vs. 9.9% and 8.2%, respectively).

These findings corroborate Hypothesis 2 for male, but not for

female agency motives. For a further exploration of sex differ-

ences, we conducted additional multiple-group tests for all path

coefficients that were significant for one sex, but not for the other,

within the groups of COR and LAT. In CORs, this was the case for

three paths (see Table 3). Whereas the partner effect of women’s

explicit motive was not significantly different from the effect of

men’s motive, ␹

(1) ⫽ 1.427, p ⫽ .232, the actor and partner

effects of the Implicit ⫻ Explicit interaction term both showed

substantial sex differences, ␹

(1) ⱖ 3.694, p ⱕ .055. In the LAT

group, the actor and partner effects of the implicit agency motive,

as well as the actor effect of the explicit motive, showed sex

differences, all of which were at least marginally significant,

␹

(1) ⱖ 3.410, p ⱕ .065. These findings imply that sex differences

within both groups contributed to the sex-specific moderation

effects of living arrangement.

Model 2: T2-SRQ. In the next step, we examined whether the

cross-sectional results were replicable in a prospective analysis of

SRQ assessed 1 year after motive assessment. Therefore, Model 2

was identical with Model 1, except that T1-SRQ was replaced with

T2-SRQ. Again, the omnibus test showed significant sex differ-

ences, ␹

(20) ⫽ 79.423, p ⬍ .001, CFI ⫽ 0.729, RMSEA ⫽ .074,

and thus male and female motives were distinguished in Model 2.

The multiple-group analysis with all corresponding effects of the

motive variables set equal across COR and LAT couples indicated

significant group differences in the pattern of associations between

agency motives and SRQ, ␹

(12) ⫽ 30.671, p ⫽ .002, CFI ⫽

0.878, RMSEA ⫽ .096. Subsequent multiple-group tests of single-

effect moderation revealed the same four significant group differ-

ences that were found in Model 1 (see Table 4). In addition, the

moderation of the actor effect of men’s explicit motive was mar-

ginally significant.

Inspection of the path coefficients from the unconstrained model

(see Table 4) revealed a pattern that was very similar to the

cross-sectional results. Again, stronger negative actor and partner

effects of men’s implicit motives and the Implicit ⫻ Explicit

interaction were observed in COR compared with LAT couples.

Table 3

Cross-Sectional Actor and Partner Effects of Agency Motives on Subjective Relationship Quality in COR and LAT Couples

(Hypothesis 2, Model 1)

Predictors Effects

COR LAT

Group

differences

b (SE) p 95% CI b (SE) p 95% CI ␹

(1)

Intercept M 0.062 (0.071) .383 [⫺0.078, 0.201] ⴚ0.405 (0.128) .002 [⫺0.664, ⫺0.156] 10.533 .001

Intercept F ⫺0.006 (0.081) .942 [⫺0.172, 0.145] ⴚ0.385 (0.128) .003 [⫺0.647, ⫺0.141] 6.821 .009

Implicit motive M ¡ M ⴚ0.225 (0.050) ⬍.001 [⫺0.324, ⫺0.130] 0.062 (0.069) .371 [⫺0.077, 0.195] 12.649 ⬍.001

F ¡ F ⴚ0.235 (0.058) ⬍.001 [⫺0.350, ⫺0.122] ⴚ0.254 (0.072) ⬍.001 [⫺0.390, ⫺0.106] 0.052 .821

M ¡ F ⴚ0.195 (0.058) .001 [⫺0.310, ⫺0.083] 0.023 (0.060) .705 [⫺0.094, 0.143] 6.825 .009

F ¡ M ⴚ0.159 (0.049) .001 [⫺0.259, ⫺0.064] ⴚ0.163 (0.078) .037 [⫺0.316, ⫺0.013] 0.002 .964

Explicit desire M ¡ M ⴚ

0.232 (0.052) ⬍.001 [⫺0.337, ⫺0.133] ⴚ0.231 (0.061) ⬍.001 [⫺0.352, ⫺0.110] ⬍0.001 .985

F ¡ F ⴚ0.134 (0.060) .025 [⫺0.252, ⫺0.016] ⫺0.060 (0.062) .334 [⫺0.180, 0.065] 0.706 .401

M ¡ F ⴚ0.175 (0.056) .002 [⫺0.284, ⫺0.062] ⫺0.068 (0.072) .341 [⫺0.211, 0.073] 1.645 .200

F ¡ M ⫺0.078 (0.054) .148 [⫺0.186, 0.024] 0.054 (0.065) .411 [⫺0.079, 0.178] 2.371 .124

Implicit ⫻ Explicit M ¡ M ⴚ0.147 (0.046) .002 [⫺0.242, ⫺0.056] 0.109 (0.075) .149 [⫺0.053, 0.246] 9.626 .002

F ¡ F 0.001 (0.070) .984 [⫺0.135, 0.138]] 0.003 (0.071) .962 [⫺0.136, 0.141] 0.001 .982

M ¡ F ⴚ0.168 (0.054) .002 [⫺0.271, ⫺0.059] 0.074 (0.071) .298 [⫺0.078, 0.199] 8.139 .004

F ¡ M 0.008 (0.056) .892 [⫺0.107, 0.112] 0.007 (0.080) .934 [⫺0.138, 0.178] ⬍0.001 .992

Note. M ¡ F denotes the partner effect of male predictor on female outcome and vice versa. Significant coefficients (p ⬍ .05) are printed in boldface.

Not displayed: covariates age, relationship duration, and additional cohabitants. COR ⫽ coresident couples; LAT ⫽ living-apart-together couples; b ⫽

unstandardized path coefficients; SE ⫽ bootstrapped standard errors (10,000 resamples); 95% CI ⫽ 95% bias-corrected bootstrap confidence intervals; M ⫽

male; F ⫽ female.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

AGENCY MOTIVES AND RELATIONSHIP FUNCTIONING

Figure 1. Cross-sectional joint and interactional effects of implicit and explicit agency motives on SRQ and

amount of conflicts in COR and LAT couples (Hypothesis 2). Results are plotted for predictor and moderator

values one standard deviation below and above the grand mean. Panel A: Male actor effects on SRQ. Panel B:

Partner effects of female motives on male SRQ. Panel C: Partner effects of male motives on female SRQ. Panel

D: Female actor effects on SRQ. Panel E: Male effects on dyadic conflicts. Panel F: Female effects on dyadic

conflicts. SRQ ⫽ subjective relationship quality; COR ⫽ coresidence; LAT ⫽ living-apart-together.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

HAGEMEYER ET AL.

The additional trend for moderation of the male explicit actor

effect was also in line with Hypothesis 2, as men’s explicit agency

motive had a negative effect on SRQ in COR, but not LAT

couples. Women’s agency motives showed significant negative

effects on SRQ in COR couples that were similar to those observed

in Model 1. As also observed in Model 1, none of these effects

were moderated by living arrangement. With few exceptions, all

path coefficients in both groups fell within the boundaries of the

95% confidence intervals obtained in the cross-sectional Model 1.

Thus, the pattern of within-group associations between agency

motives and SRQ from Model 1 was very well reproduced despite

the time lag of 1 year. This highlights the stability of the observed

relations between agency motives and SRQ. The total amount of

variance in T2-SRQ explained by male and female agency motives

was again markedly higher in COR than in LAT couples (24.4% in

male and 18.0% in female SRQ vs. 10.6% and 6.4%, respectively).

Having established temporally stable group differences in

motive-SRQ associations, we turned to the prediction of differen-

tial change in SRQ. To this end, male and female accounts of SRQ

at T1 were added as covariates to Model 2. Significant change

effects would support our assumption of a causal pathway from

agency motives to SRQ. However, given the high stability of SRQ

(see Table 1), we did not expect to find large change effects.

Accordingly, the multiple-group test of overall moderation did not

reach significance, ␹

(12) ⫽ 18.122, p ⫽ .112, CFI ⫽ .985,

RMSEA ⫽ .055. Subsequent multiple group tests of single-effect

moderation revealed one significant group difference. In line with

Hypothesis 2, men’s explicit agency desire predicted decreases in

their own SRQ in COR couples (b ⫽⫺0.110, p ⫽ .023), but not

in LAT couples (b ⫽ 0.143, p ⫽ .242), ␹

(1) ⫽ 6.821, p ⫽ .009.

This moderation effect was not observed in the cross-sectional

Model 1, which might be explained by shared method variance

between the Explicit Desire scale and the SRQ index that would

enhance associations in both groups, thereby obscuring the true

group difference. No other path coefficient in neither COR nor

LAT couples reached the 5% level of significance, but two addi-

tional findings should be mentioned. First, and in line with Hy-

pothesis 2, men’s implicit agency motive tended to predict de-

creases in their own SRQ in COR couples (b ⫽⫺0.097, p ⫽ .073),

but not LAT couples (b ⫽ 0.085, p ⫽ .433), ␹

(1) ⫽ 3.372, p ⫽

.066. Second, there was another marginal trend for moderation

concerning the partner effect of women’s explicit motive, ␹

(1) ⫽

3.826, p ⫽ .051. However, this trend was inconclusive, because

the partner effect was not even marginally significant in either

group. A full account of the change results can be found in the

online supplemental materials (Table S1). In sum, men’s agency

motives tended to predict decreases in their own SRQ in COR, but

not LAT relationships, which was consistent with the cross-

sectional analyses and further corroborates Hypothesis 2.

Model 3: Dyadic conflicts. Next, we examined the cross-

sectional associations between agency motives and the amount of

dyadic conflicts in COR and LAT couples. Model 3 comprised the

same covariates and predictors as Model 1 and Model 2, and

amount of conflicts was entered as a dyadic criterion variable. The

omnibus test of sex differences was significant again, ␹

(15) ⫽

87.089, p ⬍ .001, CFI ⫽ 0.394, RMSEA ⫽ .094, and thus male

and female predictors had to be distinguished. The multiple-group

model with all corresponding effects of the motive variables set

equal between LAT and COR couples did not yield a good fit,

␹

(6) ⫽ 19.204, p ⫽ .004, CFI ⫽ 0.813, RMSEA ⫽ .090, indi

cating overall moderation by living arrangement. Further multiple-

group analyses showed that the effects of men’s implicit motive

and women’s explicit motive on dyadic conflicts were signifi-

cantly moderated by living arrangement (see Table 5).

The path coefficients from the unconstrained model displayed in

Table 5 indicated that these moderations were in the expected

direction. Men’s implicit agency motive predicted an increased

amount of conflicts in COR, but not LAT relationships. This

Table 4

Prospective Actor and Partner Effects of Agency Motives on Subjective Relationship Quality 1 Year After Motive Assessment in COR

and LAT Couples (Hypothesis 2, Model 2)

Predictors Effects

COR LAT

Group

differences

b (SE) p 95% CI b (SE) p 95% CI ␹

(1)

Intercept M ⫺0.041 (0.096) .667 [⫺0.231, 0.147] ⫺0.327 (0.273) .231 [⫺0.933, 0.151] 1.255 .263

Intercept F ⫺0.105 (0.104) .313 [⫺0.315, 0.095] ⫺0.201 (0.234) .390 [⫺0.696, 0.227] 0.163 .686

Implicit motive M ¡ M ⴚ0.310 (0.067) ⬍.001 [⫺0.443, ⫺0.180] 0.115 (0.129) .373 [⫺0.120, 0.395] 10.267 .001

F ¡ F ⴚ0.211 (0.068) .002 [⫺0.344, ⫺0.075] ⫺0.164 (0.154) .289 [⫺0.435, 0.169] 0.111 .739

M ¡ F ⴚ0.216 (0.066) .001 [⫺0.348, ⫺0.090] 0.109 (0.131) .406 [⫺0.124, 0.394] 5.940 .015

F ¡ M ⴚ0.139 (0.065) .031 [⫺0.268, ⫺0.013] ⫺0.049 (0.154) .748 [⫺0.335, 0.272] 0.367 .545

Explicit desire M ¡ M ⴚ0.290 (0.065) ⬍.001 [⫺0.420, ⫺0.166] ⫺0.044 (0.149) .767 [⫺0.327, 0.256] 3.731 .053

F ¡ F ⫺0.117 (0.072) .103 [⫺

0.260, 0.023] 0.036 (0.165) .829 [⫺0.276, 0.367] 1.080 .299

M ¡ F ⴚ0.183 (0.068) .007 [⫺0.317, ⫺0.053] ⫺0.104 (0.116) .370 [⫺0.338, 0.115] 0.403 .526

F ¡ M ⫺0.008 (0.068) .907 [⫺0.140, 0.125] ⫺0.131 (0.155) .398 [⫺0.411, 0.191] 0.632 .427

Implicit ⫻ Explicit M ¡ M ⴚ0.215 (0.065) .001 [⫺0.348, ⫺0.092] 0.190 (0.143) .182 [⫺0.105, 0.458] 9.655 .002

F ¡ F ⫺0.027 (0.071) .702 [⫺0.166, 0.116] ⫺0.170 (0.188) .365 [⫺0.507, 0.219] 1.006 .316

M ¡ F ⴚ0.194 (0.061) .002 [⫺0.313, ⫺0.071] 0.072 (0.122) .556 [⫺0.178, 0.311] 4.153 .042

F ¡ M ⫺0.021 (0.074) .778 [⫺0.152, 0.139] ⫺0.096 (0.168) .570 [⫺0.419, 0.251] 0.227 .634

Note. M ¡ F denotes the partner effect of male predictor on female outcome and vice versa. Significant coefficients (p ⬍ .05) are printed in boldface.

Not displayed: covariates age, relationship duration, and additional cohabitants. COR ⫽ coresident couples; LAT ⫽ living-apart-together couples; b ⫽

unstandardized path coefficients; SE ⫽ bootstrapped standard errors (10,000 resamples); 95% CI ⫽ 95% bias-corrected bootstrap confidence intervals; M ⫽

male; F ⫽ female.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

AGENCY MOTIVES AND RELATIONSHIP FUNCTIONING

association was further qualified by a significant Implicit ⫻ Ex-

plicit interaction in CORs. Figure 1 (Panel E) illustrates this

pattern of results. In COR couples, the overall positive effect of

men’s implicit agency motive on conflicts was stronger in the

presence of a strong (M ⫹ 1 SD) explicit motive (simple slope b ⫽

0.358, p ⬍ .001) than in the presence of a weak (M –1SD) explicit

motive (b ⫽ 0.104, p ⫽ .130). Further, in LAT couples, the

average amount of conflicts was higher (see the intercept in Table

5), and male agency motives were no significant predictors. How-

ever, in COR couples, the combination of high scores (M ⫹ 1 SD)

on implicit and explicit motives in men predicted amounts of

conflicts that slightly exceeded the average level of conflicts in

LAT couples. Regarding men’s motives, the pattern of results for

dyadic conflicts was thus very similar to the findings for SRQ. The

findings for women’s motives, again, looked different (see Table 5

and Figure 1, Panel F). In LAT couples, a strong explicit motive

was related to lower amounts of conflicts, whereas there was no

significant effect in CORs. Although this is plausible, no compa-

rable effects were found in the other models, and thus this effect

should be interpreted with caution. There was also a marginally

significant moderation of women’s Implicit ⫻ Explicit interaction

effect (see Table 5). However, because this interaction did not

reach significance within either group it is not interpreted. Overall,

male and female agency motives explained more variance in COR

than in LAT relationships (11.4% vs. 7.0%).

In summary, consistent with the findings for SRQ, men’s

agency motives tended to have stronger effects on dyadic conflicts

in COR couples than in LAT couples. Diverging from SRQ, the

female explicit motive predicted lower amounts of conflicts in

LATs, but not CORs. If this effect is reliable, it would point to the

possibility that agency motives might have even beneficial effects

for couples, if the contextual conditions are right. Finally, we

explored sex differences in the path coefficients employing the

same strategy as in Model 1. In the LAT group, the effect of

women’s explicit agency motive differed from the male motive,

␹

(1) ⫽ 7.310, p ⫽ .007. In the COR group, this was not the case,

␹

(1) ⫽ 2.709, p ⫽ .100, but the implicit ⫻ explicit interaction

effect showed a significant sex difference, ␹

(1) ⫽ 5.679, p ⫽

.017. Thus, as observed for SRQ, significant sex differences in

both groups contributed to the sex-specific moderation effects of

living arrangement.

Mediation analysis. In a final step, we specified a variant of

Model 1, in which dyadic conflicts were added as mediators of the

paths from agency motives to SRQ in the groups of COR and LAT.

The results for COR couples showed that eight of the nine signif-

icant effects observed in Model 1 (see Table 3) were mediated by

dyadic conflicts (indirect effects ab ranging from ⫺0.111

to ⫺0.053, p ⱕ .029; see Table S2 of the online supplemental

materials for detailed results). However, for six of these mediated

paths, the direct effects remained significant (b ranging

from ⫺0.180 to ⫺0.092, p ⱕ .022), indicating that the negative

effects of agency motives on SRQ cannot be fully explained by

increased dyadic conflicts. In the LAT group, by contrast, none of

the three significant effects from Model 1 were mediated by dyadic

conflicts (ab ranging from ⫺0.018 to ⫺0.005, p ⱖ .258). These

findings further highlight the differential role of couples’ living

arrangements for the expression of agency motives, as negative

associations with SRQ in CORs are partially grounded in different

processes than the corresponding effects in LATs.

Supplemental analyses. To further corroborate the robustness

and generalizability of the findings regarding Hypothesis 2, we

conducted a number of supplemental analyses. First, we reran

Model 1 and Model 2 using all information on SRQ that was

assessed at T1 and T2. In these analyses, the T1-SRQ index

comprised five measures (perceived available support was added),

and the T2-SRQ index comprised six measures (emotional close-

ness and satisfaction with the living arrangement were added). The

results were nearly identical to the original analyses. In both

models, all significant moderation effects were replicated (see

Tables S3 and S4 of the online supplemental materials for detailed

results). This highlights the generalizability of the moderating

function of living arrangement across different operationalizations

of SRQ.

Second, we tested variants of Models 1 through 3 in which

implicit and explicit agency motives were replaced with the com-

plementary communal motives as assessed with the PACT (Hage-

meyer & Neyer, 2012) and the ABC scales (Hagemeyer, Neyer, et

al., 2013), respectively. The multiple-group analyses of the cross-

sectional (Model 1) and prospective (Model 2) associations be-

tween communal motives and SRQ showed no significant overall

moderations by living arrangement, ␹

(12) ⫽ 9.703, p ⫽ .642,

CFI ⫽ 1.000, RMSEA ⫽ .000, and ␹

(12) ⫽ 10.625, p ⫽ .561,

Table 5

Cross-Sectional Effects of Agency Motives on the Amount of Dyadic Conflicts (Hypothesis 2, Model 3)

Predictors Effects

COR LAT

Group

differences

b (SE) p 95% CI b (SE) p 95% CI ␹

(1)

Intercept ⫺0.100 (0.070) .151 [⫺0.238, 0.035] 0.197 (0.139) .156 [⫺0.080, 0.456] 4.052 .044

Implicit motive M 0.231 (0.055) ⬍.001 [0.122, 0.337] 0.051 (0.059) .388 [⫺0.065, 0.167] 4.598 .032

F 0.171 (0.052) .001 [0.066, 0.270] 0.073 (0.059) .219 [⫺0.043, 0.190] 1.325 .250

Explicit desire M 0.122 (0.054) .025 [0.015, 0.227] 0.018 (0.066) .781 [⫺0.111, 0.144] 1.514 .219

F ⫺0.020 (0.056) .720 [⫺0.130, 0.092] ⴚ0.247 (0.077) .001 [⫺0.394, ⫺0.092] 6.438 .011

Implicit ⫻ Explicit M 0.127 (0.050) .011 [0.030, 0.227] ⫺0.009 (0.065) .889 [⫺0.139, 0.120] 2.538 .111

F ⫺0.061 (0.060) .315 [⫺0.172, 0.063] 0.094 (0.062) .132 [⫺0.040, 0.207] 2.852 .091

Note. Significant coefficients (p ⬍ .05) are printed in boldface. Not displayed: covariates age, relationship duration, and additional cohabitants. COR ⫽

coresident couples; LAT ⫽ living-apart-together couples; b ⫽ unstandardized path coefficients; SE ⫽ bootstrapped standard errors (10,000 resamples); 95%

CI ⫽ 95% bias-corrected bootstrap confidence intervals; M ⫽ male; F ⫽ female.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

HAGEMEYER ET AL.

CFI ⫽ 1.000, RMSEA ⫽ .000, respectively. Inspection of the

unconstrained paths in CORs and LATs mainly revealed positive

actor and partner effects of the explicit communal motive on SRQ

that were of comparable size in both groups (see Hagemeyer,

Neberich, et al., 2013, for related analyses of the present data on

communal motives). The analysis of dyadic conflicts (Model 3)

revealed significant overall moderation by living arrangement,

␹

(6) ⫽ 14.006, p ⫽ .003, CFI ⫽ 0.954, RMSEA ⫽ .070, but the

group-specific patterns of effects were different from the analysis

of agency motives (see Table S5 of the online supplemental

materials). These findings have two implications. First, the group-

specific associations found for agentic motives cannot be attrib-

uted to overlap with their communal counterparts. Second, the

group differences observed for agency motives are not attributable

to a general effect of higher interdependence in COR relationships

that would increase the effects of all relevant personality disposi-

tions.

Finally, we specified moderated actor–partner interdependence

models using multilevel regression as an alternative analysis strat-

egy to reproduce the cross-sectional and prospective findings from

Model 1 and Model 2 concerning SRQ (Kenny et al., 2006).

Complementing the multiple-group analyses, the multilevel ap-

proach provides (a) controls of between-group (but not within-

group) differences in the three covariates, and (b) direct tests of all

Predictor ⫻ Sex interactions. The results of the cross-sectional

multilevel model fully confirmed the findings from Model 1 (see

Table S6 of the online supplemental materials). Most notably, the

four-way interaction term Sex ⫻ Living Arrangement ⫻ Implicit

Motive ⫻ Explicit Motive showed significant actor and partner

effects (b ⫽ 0.066, p ⫽ .037, and b ⫽ 0.063, p ⫽ .043, respec-

tively). The results of the prospective model were similar, albeit

that the partner effect of the four-way interaction was marginally

significant (see Table S7 of the online supplemental materials).

Thus, the sex-specific patterns of moderation by living arrange-

ment were robust, irrespective of the methodological approach.

Summary. The findings concerning Hypothesis 2 can be

summed up as follows. First, both implicit and explicit agency

motives had independent and interactional effects on SRQ and

dyadic conflicts as indicators of relationship functioning. Second,

negative effects of agency motives on relationship functioning

were, as expected, more pronounced in COR couples than in LAT

couples. Third, this moderation was mainly found for men’s mo-

tives; women’s explicit motives overall had fewer negative effects

on relationship functioning than men’s, and only one effect of

women’s motives was moderated by living arrangement. Thus,

across different indicators of relationship functioning and cross-

sectional and prospective analyses, Hypothesis 2 was consistently

confirmed for men’s, but not for women’s, agency motives. The

benefits of the greater objective closeness provided by COR seems

to be diminished in the presence of congruently strong implicit and

explicit agency motives in men.

Study 2

Study 1 exclusively addressed the dispositional level of agency

motives and thus focused on between-couple differences. To get a

more comprehensive picture, we conducted a 2-week dyadic diary

study to examine the interplay between agentic motivational states,

objective closeness, and relationship functioning at the within-

couple level. It is noteworthy that the within-couple perspective of

Study 2 is complementary to the between-couple analyses of Study

1, and renders additional insights into relationship processes that

cannot be observed with the between-couple approach of Study 1

(Molenaar, 2004). In order to establish a link between the two

studies, however, the first research question of Study 2 addressed

the between-couple level, namely, whether typical agentic moti-

vational states in COR and LAT relationships could be predicted

by dispositional agency motives (Hypothesis 3). In particular,

we focused on a motivational state in which persons long for

more agency, which we will henceforward call agency appe-

tence. Second, we investigated, on the within-couple level,

whether the interaction between daily agency appetence and

objective closeness was predictive of daily relationship satis-

faction (Hypothesis 4).

Hypothesis 3: Agency Motives Predict Daily Agency

Appetence, Especially in COR Relationships

A defining feature of motive disposition is the propensity to be in

a specific motivational state (Atkinson, 1981; McAdams & Constan-

tian, 1983). Hence, dispositional agency motives were expected to

predict the mean level of daily agentic motivational states in our diary

study. However, dynamic theories of human motivation make specific

predictions when an association between a disposition and corre-

sponding motivational states should be expected (Atkinson, 1981;

Bischof, 1975, 1993). According to these theories, the current moti-

vational activation depends on the discrepancy between a set point

(the need or motive strength) and the current level of need satisfaction.

If the current level is below the set point, a motivational activation is

triggered with the goal to reduce the discrepancy. If the goal is

achieved, consummatory actions or experiences reduce the motiva-

tional activation (see also Kumashiro et al., 2008, for a similar

motivational mechanism applied to the regulation of personal vs.

relational concerns). Environmental factors play an important role in

this process, because they can foster or hinder immediate need satis-

faction. For example, a person with a strong dispositional agency

motive might rarely experience the corresponding state of agency

appetence when he or she is able to satisfy the need immediately. In

contrast, a person with a strong agency motive who lives in an

environment that hinders the satisfaction of that need will soon

experience a state of appetence.

Applied to our current study, the relationship environment pro-

vided by different living arrangements (COR vs. LAT) is expected

to interact with the strength of agency motives in the prediction of

motivational states. Specifically, we expected that agency appe-

tence would be better predicted by agency motives in COR cou-

ples, because living together provides more obligatory objective

closeness, which makes the satisfaction of agentic needs more

difficult. Comparable with Study 1, we tested Hypothesis 3 for

implicit and explicit agency motives and the Implicit ⫻ Explicit

interaction.

Hypothesis 4: Daily Relationship Satisfaction Is

Predicted by the Interaction of Objective Closeness

and Motivational States

The overarching assumption of the present investigation was that

relationship functioning depends on the match or mismatch between

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

AGENCY MOTIVES AND RELATIONSHIP FUNCTIONING

agency motives and objective closeness. Hypothesis 4 addresses this

assumption at the daily within-couple level. We expected relationship

functioning within couples to be poor when a mismatch between the

individual agency appetence and the actual objective closeness oc-

curred. More specifically, relationship satisfaction was expected to

decrease on days when high amounts of time spent with one’s partner

were counteracted by high agency motivation. In addition, we hy-

pothesized that this association is most pronounced in COR relation-

ships, because the average amount of time spent with the partner is

higher, and much of the time spent together in COR relationships is

likely obligatory proximity (e.g., routine actions like getting dressed,

having breakfast, watching one’s favorite TV show, etc., are usually

carried out with the partner nearby). This makes the regulation of

closeness in accordance with one’s needs more difficult, and thus

increases the probability of experiencing dissatisfaction and frustra-

tion.

Method

Participants and procedure. A subsample of 106 couples

from Study 1 participated in the 2-week diary study (48 COR, 58

LAT). Couples who had participated in Study 1 were invited to take

part in the diary study and selected for a balanced number of COR and

LAT couples. Participants were compensated with €40 per couple and

an individual, confidential feedback about the trajectory of their

relationship quality across the diary-study period. The average age

was 40.8 years (SD ⫽ 13.7, ranging from 18 to 72), and the average

relationship duration was 7.8 years (SD ⫽ 9.5 years, ranging from 5

months to 40 years). Thus, couples who took part in the diary study

had significantly shorter relationships compared with nonparticipating

couples from Study 1, t(546) ⫽ 3.219, p ⫽ .001, d ⫽ 0.336, probably

because of oversampling of LAT couples.

The diary data were completely assessed online. For each couple,

the assessment period started on a Monday and ended on Sunday 13

days later. The first and the last day were used for pre- and postdiary

questions that are not included in the current analyses. Thus, the diary

lasted for 12 days, from a Tuesday to a Saturday. Each partner

received an individual access code for each day of assessment by

e-mail that expired after answering the required questions. To ensure

confidentiality and some control about the assessment situation, par-

ticipants were instructed to answer all questions of the day at once

when the partner was not present. During the diary phase, the access

codes were sent at 6:00 p.m., and participants were instructed to

answer the questions between 8:00 p.m. and noon of the next day.

Analyses of the log-in times showed that 64% of the participants

reported on a day before 2:00 a.m. the next day, and 96% before noon

of the next day.

Missing data. The study design consisted of 106 couples ⫻ 2

persons ⫻ 12 days ⫽ 2,544 observations. Five days (0.39%) were

missing completely, which reduced the final data set to 2,534

observations. Other missing values were rare (⬍2.5% for each

variable) and were excluded on a per-analysis basis.

Background data. From the data of Study 1, the following

variables were used in the present analyses: implicit relationship-

specific agency motive (assessed with the PACT), explicit desire

for being alone (eight items, response scale 1 to 7), age, relation-

ship duration, and additional cohabitants (see Study 1 for details).

Daily measures. To assess agentic motivational states, partic-

ipants completed the item “The amount of time I had for myself

today was . . .,”

with a rating on a 7-point scale (⫺3 ⫽ too little,

0 ⫽ just right,3⫽ too much). Ratings ⬍0 indicated agency

appetence, and ratings ⬎0 indicated agency aversion. To obtain a

linear measure of agency appetence, ratings of 1 to 3 were set to

0, and ratings of ⫺3to⫺1 were reversed. Daily relationship

satisfaction was measured with the item “How satisfied with your

relationship are you today?,” rated on an 11-point scale (0 ⫽ not

at all,10⫽ a great deal). Finally, participants reported the amount

of time spent together with the partner on a given day in hours. It

was treated as a dyadic variable (i.e., we used the average of both

partners’ reports, which were nearly identical), and the maximum

was set to 18 hr (i.e., 24 hr excluding sleeping time).

Results and Brief Discussion

Mean differences. Table 6 shows descriptive statistics of the

daily measures. Before testing our Hypotheses, we examined using

multilevel models mean differences between living arrangements,

sex, or between specific cells of living arrangement and sex. For

the sake of brevity, we only report the results for group differences

at p ⬍ .10. As expected, COR couples spent on average more daily

hours together than LAT couples (M ⫽ 7.85, SD ⫽ 4.56 vs. M ⫽

4.40, SD ⫽ 4.69 hr), t(104.19) ⫽ 7.52, p ⬍ .001. Thus, COR

relationships were indeed characterized by higher objective close-

ness. Furthermore, average daily relationship satisfaction tended to

be higher in COR couples than in LAT couples (M ⫽ 7.90, SD ⫽

2.13 vs. M ⫽ 7.39, SD ⫽ 2.52), t(103.90) ⫽ 1.839, p ⫽ .069,

which was in line with the results from Study 1. Finally, there was

a significant sex difference in agency appetence. Irrespective of

living arrangement, women experienced agency appetence on

more days (see Table 6), t(2,380.30) ⫽ 5.504, p ⬍ .001, and made

higher appetence ratings, t(2,379.00) ⫽ 7.007, p ⬍ .001, than men.

Analysis strategy. The dyadic diary data were analyzed

with multilevel models (Bolger & Laurenceau, 2013) using the

lme4 package (Bates, Maechler, Bolker, & Walker, 2014) for

the R Statistical Environment (R Core Team, 2014). This sta-

tistical approach models the within-person causal process at

Level 1, treats couples as random factors at Level 2, accounts

for dyadic nonindependence of the couple members, and allows

for disentangling within-person and between-person processes.

The original German wording was “Die Zeit, die ich heute für mich

allein hatte, war mir . . .,” which implies a focus on the individual self and

privacy, thus reflecting a core aspect of relationship-specific agency mo-

tivation.

Table 6

Descriptive Statistics of the Daily Measures (Study 2)

Measures

Men Women

MSDMSD

Hours spent together (0–18)

5.96 4.94 5.96 4.94

Relationship satisfaction (0–10) 7.60 2.29 7.65 2.43

% of days with agency appetence

rating ⬎ 0 24.86 43.24 36.15 48.06

Agency appetence (0–3) 0.37 0.72 0.63 0.96

Because hours spent together is a dyadic variable, M and SD are identical

for men and women.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

HAGEMEYER ET AL.

It has been suggested that all within-person intercepts and

slopes should be allowed to vary randomly between persons,

which yields a conservative approach that preserves Type I

error rates (D. J. Barr, Levy, Scheepers, & Tily, 2013). This

approach, however, has also been criticized, as it involves

estimations from a large number of variances, which can lead to

degenerate variance– covariance matrices with small data sets

(Pinheiro & Bates, 2000). In our analyses, random slopes fre-

quently led to convergence problems. Thus, we fixed the slopes

and only included a random intercept for couples to control for

dyadic nonindependence.

To keep the multilevel models parsimonious, we fitted the

models to the data in a stepwise procedure, testing whether

additional blocks of potential predictors (e.g., interactions with

sex) increased the model fit. Model comparisons were con-

ducted via chi-square difference likelihood ratio tests and the

difference in Akaike Information Criterion (⌬AIC; Burnham,

Anderson, & Huyvaert, 2011). In general, the model with the

smaller AIC shows a better fit to the data. Rough guidelines

(e.g., Burnham & Anderson, 2002) suggest that an AIC differ-

ence ⱕ2 is regarded as marginal, and both models fit roughly

equally well. If 4 ⬍⌬AIC ⬍7, one would favor the model with

the smaller AIC, but not discard the other model yet. Models

with ⌬AIC ⬎10 have essentially no support.

Centering. Agency appetence, which is a central variable in

the analyses, has a natural zero point: Zero describes a day in

which no appetence was experienced, and positive values describe

an increasing appetence with a maximum value of 3. The number

of hours partners spent together was a dyadic variable and centered

at Level 1 within couples (Bolger & Laurenceau, 2013). Hence, the

Level 1 effect of hours spent together refers to deviations from a

couple’s typical level. The dependent variable daily relationship

satisfaction was kept in its original metric ranging from 0 to 10. All

variables on person level were z-standardized across men and

women.

Covariates. We included time as a covariate on Level 1 in all

models. Although we were not primarily interested in temporal

change, it is necessary to include this variable to control for

possibly spurious effects attributable to third variables (Bolger &

Laurenceau, 2013). For this purpose, time was centered at the

middle of the measurement period and rescaled such that one unit

expresses 1 week. Furthermore, in line with the analyses of Study

1, mean-centered age and relationship duration, as well as effect-

coded sex and additional cohabitants, were entered as covariates in

all models.

Sex differences. We made no a priori predictions about sex

differences in the hypothesized effects. Thus, the multilevel mod-

els were initially specified irrespective of sex. The contributions of

the interactions between predictors and sex were tested in a sub-

sequent step to explore sex differences. If these interaction terms

did not yield a better model fit, they were omitted for parsimony.

Hypothesis 3: Agency motives predict agency appetence,

especially in COR relationships. For the prediction of average

agency appetence, we fitted four nested models with increasing

numbers of predictors to the data. The baseline model included

the covariates time, sex, age, relationship duration, and addi-

tional cohabitants. Next, we added the focal predictors implicit

and explicit agency motives, living arrangement, and the Mo-

tive ⫻ Living Arrangement two-way interactions. This model

had a significantly better fit than the baseline model, ⌬␹

(5) ⫽

34.4, p ⬍ .001, ⌬AIC ⫽ 24.4. Next, the interactions between

sex and all predictor variables were added, which further in-

creased the model fit, ⌬␹

(5) ⫽ 55.5, p ⬍ .001, ⌬AIC ⫽ 45.5.

Thus, sex was retained in the model as a moderating variable.

Finally, we added the interaction between implicit and explicit

motives along with the three-way and four-way interactions

between the motives, living arrangement, and sex. This last

block of variables made no further significant contribution to

the prediction of agency appetence, ⌬␹

(4) ⫽ 2.1, p ⫽ .722,

⌬AIC ⫽⫺5.9, and was thus excluded from the model.

Table 7 shows the results of the final model. There were

significant effects of living arrangement and the explicit agency

motive. However, these simple effects were further qualified by

significant interaction effects. The explicit motive interacted with

sex, and living arrangement interacted with sex and the implicit

Table 7

Multilevel Model for the Prediction of Average Daily Agency Appetence by Implicit and Explicit

Agency Motive Dispositions, Living Arrangement, and Sex (Hypothesis 3)

Predictors b (SE) p 95% CI

Intercept 0.573 (0.060) ⬍.001 [0.460, 0.699]

Sex (M ⫽⫺1; F ⫽ 1) 0.097 (0.018) ⬍.001 [0.060, 0.131]

Living arrangement (LAT ⫽⫺1; COR ⫽ 1) 0.114 (0.046) .014 [0.024, 0.202]

Implicit motive 0.031 (0.023) .173 [⫺0.015, 0.076]

Explicit desire 0.117 (0.027) ⬍.001 [0.069, 0.159]

Living Arrangement ⫻ Implicit Motive 0.040 (0.029) .079 [⫺0.008, 0.089]

Living Arrangement ⫻ Explicit Motive 0.026 (0.028) .262 [⫺0.024, 0.071]

Sex ⫻ Living Arrangement ⫺0.023 (0.018) .185 [⫺0.061, 0.012]

Sex ⫻ Implicit Motive ⴚ0.113 (0.023) ⬍.001 [⫺0.161, ⫺0.068]

Sex ⫻ Explicit Motive 0.046 (0.023) .028 [0.005, 0.088]

Sex ⫻ Living Arrangement ⫻ Implicit Motive ⴚ0.128 (0.022) ⬍.001 [⫺0.172, ⫺0.085]

Sex ⫻ Living Arrangement ⫻ Explicit Motive 0.026 (0.021) .218 [⫺0.015, 0.066]

Note. Significant coefficients (p ⬍ .05) are printed in boldface. Not displayed: covariates time, age, relation-

ship duration, and additional cohabitants. b ⫽ unstandardized regression weights; SE ⫽ standard errors of

regression weights; 95% CI ⫽ 95% bootstrap confidence intervals; M ⫽ male; F ⫽ female; COR ⫽ coresident

couples; LAT ⫽ living-apart-together couples.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

AGENCY MOTIVES AND RELATIONSHIP FUNCTIONING

motive in the prediction of agency appetence.

Figure 2 displays

this pattern of results. First, the explicit agency motive was positively

related to agency appetence, and this association was stronger in

women (simple slope b ⫽ 0.162, p ⬍ .001) than in men (b ⫽ 0.071,

p ⫽ .018; Panels A and B). Living arrangement was not a significant

moderator, but there was a nonsignificant trend indicating that the

stronger effect of the female explicit motive was largely attributable to

women in COR relationships. Second, the implicit agency motive

predicted higher agency appetence only for men in COR relationships

(b ⫽ 0.312, p ⬍ .001; Panel C). In COR women, unexpectedly, this

association showed a negative tendency (b ⫽⫺0.169, p ⫽ .001;

Panel D). In LATs, there was no significant association between the

implicit agency motive and agency appetence in neither women (b ⫽

0.006, p ⫽ .875) nor men (b ⫽⫺0.024, p ⫽ .595; Panels C and D).

In sum, this pattern of results lends partial support to Hypothesis 3.

As expected, dispositional agency motives were significantly linked

to agentic motivational states. However, the expected context sensi-

tivity was found only for the implicit motive disposition and only in

men.

Hypothesis 4: Daily relationship satisfaction is predicted by

the interaction of objective closeness and motivational states.

For the prediction of daily relationship satisfaction, we specified a

lag model using the agency appetence rating from yesterday (t

-1

)in

interaction with today’s (t

) hours spent with the partner to predict

today’s (t

) relationship satisfaction. This was the most reasonable

causal link, as the questionnaire ratings were given at the end of

each day. In addition, we controlled for yesterday’s (t

-1

) relation

ship satisfaction to model day-to-day changes in relationship sat-

isfaction.

To make sure that the moderating effect of living arrangement indicated

by this three-way interaction was not attributable to potential confounding of

living arrangement with the covariates age, relationship duration, or additional

cohabitants, we conducted three control analyses to test whether these cova-

riates had a similar moderating effect as living arrangement. In each of these

analyses, living arrangement was replaced by one of the covariates as a

moderator in the final multilevel model for Hypothesis 3. Age and relationship

duration had no significant moderating effects (p ⫽ .152 and p ⫽ .186,

respectively). However, the presence of additional cohabitants had a signifi-

cant moderating effect (b ⫽ 0.067, p ⫽ .002). Thus, we tested the moderations

by living arrangement and additional cohabitants simultaneously in a final

control analysis. The results showed that the effect of the three-way interaction

Sex ⫻ Implicit Motive ⫻ Living Arrangement was robust (b ⫽⫺0.099, p ⬍

.001) when the interactions with additional cohabitants were controlled. In

sum, the control analyses showed that the moderating effect of living arrange-

ment cannot be attributed to confounding with any of the covariates.

Figure 2. Prediction of average daily agency appetence by explicit and implicit agency motives in COR and LAT

relationships (Hypothesis 3). Results were plotted for predictor values one standard deviation below and above the

grand mean. Panel A: Male effects of explicit agency desire on agency appetence in COR and LAT couples. Panel

B: Female effects of explicit agency desire on agency appetence in COR and LAT couples. Panel C: Male effects of

implicit agency motive on agency appetence in COR and LAT couples. Panel D: Female effects of implicit agency

motive on agency appetence in COR and LAT couples. COR ⫽ coresidence; LAT ⫽ living-apart-together.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

HAGEMEYER ET AL.

Again, we fitted the model in a stepwise procedure. The

baseline model included the covariates yesterday’s relationship

satisfaction, time, sex, age, relationship duration, and additional

cohabitants. Next, we added the focal predictors living arrange-

ment, today’s hours spent together, and yesterday’s agency

appetence along with their two-way and three-way interaction

terms. This block of variables made a highly significant con-

tribution to the prediction of daily relationship satisfaction,

⌬␹

(7) ⫽ 124.04, p ⬍ .001, ⌬ AIC ⫽ 110.1. Finally, we entered

the interaction terms between sex and all predictor variables.

This step did not lead to a better model fit, ⌬␹

(7) ⫽ 4.2, p ⫽

.760, ⌬AIC ⫽⫺9.9. The interaction terms with sex were

therefore dropped from the model.

Table 8 presents the results of the final model. First, the

number of hours spent with the partner on a given day was

significantly associated with increased relationship satisfaction

on this day. This finding at the within-couple level underlines

the overall beneficial role of daily objective closeness and

complements previous studies that found positive associations

between objective closeness and relationship functioning at the

between-couple level (e.g., Berscheid et al., 1989b; Smith et al.,

1988). Second, there was a marginal trend for yesterday’s

agency appetence to predict decreases in today’s relationship

satisfaction. Finally, these two simple effects were further qual-

ified by a significant three-way interaction effect of Living

Arrangement ⫻ Hours Spent Together ⫻ Agency Appetence.

Figure 3 shows the pattern of this interaction. As expected, the

overall positive effect of hours spent together (simple slopes b

ranging from 0.103 to 0.142, all ps ⬍ .001) was diminished and

even reversed in sign, if participants in COR relationships

experienced high agency appetence (b ⫽⫺0.023, p ⫽ .645).

This effect of the mismatch between high objective closeness

and high agency appetence confirms Hypothesis 4.

To control for potential confounding of living arrangement with the

covariates in this three-way interaction effect, we employed the same

strategy as in the control analyses for Hypothesis 3 (see Footnote 6). Again,

neither age nor relationship duration had significant moderating effects

(p ⫽ .461 and p ⫽ .842), but the interaction effect of Additional Cohab-

itants ⫻ Hours ⫻ Agency Appetence was significant (b ⫽ 0.026, p ⫽

.028). However, a subsequent simultaneous analysis showed that the effect

of the interaction Living Arrangement ⫻ Hours ⫻ Agency Appetence

remained substantial (b ⫽⫺0.022, p ⫽ .078) when the interactions with

additional cohabitants were controlled. Thus, as in all previous analyses,

the moderating effect of living arrangement cannot be attributed to con-

founding with age, relationship duration, or additional cohabitants. We also

tested whether individual differences in dispositional agency motives mod-

erated the within-couple effect of the mismatch between agency appetence

and objective closeness in two additional models, in which either explicit

or implicit agency motives, along with all of their product-interaction terms

with all other predictors, were added to the final model for the test of

Hypothesis 4. In both models, the hypothesized three-way interaction

effect Living Arrangement ⫻ Hours Spent Together ⫻ Agency Appetence

remained significant (b ⱕ ⫺0.028, p ⱕ .026), and the four-way interaction

Living Arrangement ⫻ Hours Spent Together ⫻ Agency Appetence ⫻

Motive Disposition was not significant (p ⱖ .564).

Table 8

Lag Multilevel Model for the Prediction of Daily Changes in Relationship Satisfaction by Daily

Agency Appetence, Daily Hours Spent Together, and Living Arrangement (Hypothesis 4)

Predictors b (SE) p 95% CI

Intercept 5.209 (0.215) ⬍.001 [4.753, 5.646]

Relationship satisfaction (t–1) 0.318 (0.020) ⬍.001 [0.282, 0.358]

Living arrangement (LAT ⫽⫺1; COR ⫽ 1) 0.189 (0.118) .112 [⫺0.047, 0.443]

Hours together (t0) 0.104 (0.011) ⬍.001 [0.081, 0.127]

Agency appetence (t–1) ⫺0.090 (0.049) .067 [⫺0.188, 0.008]

Living Arrangement ⫻ Hours Together ⬍0.001 (0.011) .997 [⫺0.023, 0.021]

Living Arrangement ⫻ Agency Appetence ⫺0.065 (0.048) .177 [⫺0.150, 0.027]

Hours Together ⫻ Agency Appetence ⫺0.015 (0.012) .219 [⫺0.041, 0.007]

Living Arrangement ⫻ Hours ⫻ Appetence ⴚ0.027 (0.012) .022 [⫺0.049, ⫺0.006]

Note. Significant coefficients (p ⬍ .05) are printed in boldface. Not displayed: covariates sex, time, age,

relationship duration, and additional cohabitants. b ⫽ unstandardized regression weights; SE ⫽ standard errors

of regression weights; 95% CI ⫽ 95% bootstrap confidence intervals; COR ⫽ coresident couples; LAT ⫽

living-apart-together couples.

Figure 3. Prediction of day-to-day changes in relationship satisfaction by

the interaction of daily hours spent together and daily agency appetence in

COR and LAT relationships (Hypothesis 4). Results were plotted for

deviations of 4 hr below and above the couple-specific average level of

hours spent together and minimum (⫽ 0) and maximum (⫽ 3) values of

agency appetence. COR ⫽ coresidence; LAT ⫽ living-apart-together.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

AGENCY MOTIVES AND RELATIONSHIP FUNCTIONING

General Discussion

Applying a Person ⫻ Situation perspective, we investigated the

interplay between relationship-specific agency motives and objec-

tive dyadic closeness on the between-couple and the within-couple

levels. As an important environmental factor that affects objective

closeness, we introduced the distinction of two prototypical dyadic

living arrangements to the study of motive dispositions in couples:

living-apart-together (LAT) and coresidence (COR). Our analyses

showed, first, that implicit and explicit agency motives and

(post)reproductive age contributed to the prediction of couples’

living arrangements. Second, congruently strong implicit and ex-

plicit agency motives in men were associated with impaired rela-

tionship functioning in COR, but not LAT couples. Third, implicit

and explicit agentic motive dispositions contributed to the predic-

tion of daily agency appetence, and, fourth, concurrently high

agency appetence and objective closeness impaired relationship

satisfaction on a day-to-day level. In sum, these findings highlight

the benefits of a motivational Person ⫻ Situation approach to

couple research, and corroborate the assumption that relationship

functioning depends on the match or mismatch between the part-

ners’ agency motives and the objective closeness realized in the

relationship.

It is noteworthy that the present investigation relied on the

concept of relationship-specific agency motives as introduced by

Hagemeyer and Neyer (2012; see also Hagemeyer, Neyer, et al.,

2013). Consequently, the employed motive measures, especially

the self-report questionnaires, focused on the propensity to seek

distance from one’s partner, reflecting preferences for indepen-

dence, self-reliance, and privacy. In our opinion, these incentives

are core aspects of agentic motivation that are crucial for the

understanding of couples’ distance regulation and relationship

functioning, and the reported findings corroborate this assumption.

However, agency is a broad psychological construct, and other

measures emphasize different aspects, such as dominance and

assertion (e.g., Bartz & Lydon, 2004; Suh, Moskowitz, Fournier, &

Zuroff, 2004), achievement and competence (e.g., Sheldon &

Cooper, 2008), or self-esteem and self-validation (e.g., Greenberg

& Goldman, 2008; Prager & Buhrmester, 1998). Generalizations

of the current findings to other conceptualizations and operation-

alizations of agentic motives and goals should therefore be treated

with caution. In the following, we discuss further implications and

limitations of the two studies as well as perspectives for future

research related to the four hypotheses.

Selection of Living Arrangement

As expected, the selection of either COR or LAT depended

on (post)reproductive age and the partners’ agency motives.

Specifically, strong agency motives in women predicted being

in a LAT relationship. Whereas their explicit motive was pre-

dictive irrespective of age, women’s implicit motive was pre-

dictive only in postreproductive age. In addition, a strong

explicit agency motive in men was associated with being LAT

in postreproductive age. These findings support our assumption

that couples tend to select a relationship context that suits the

partners’ agentic needs with regard to objective closeness,

especially when otherwise predominant relationship goals like

having children together tend to lose relevance. The results are

also in line with previous findings from demographic and

sociological studies showing (a) that the historical trend for

more LAT relationships is most pronounced in postreproductive

age (Asendorpf, 2008), and (b) that the subjective reasons for

LAT are often agentic in nature (Holmes, 2006; Karlsson &

Borell, 2005; Levin, 2004; Roseneil, 2006). The present study

suggests that such reasons are rooted in dispositional implicit

and explicit motives, and thus in the partners’ personalities.

Overall, the explicit agency motive was a stronger predictor of

living arrangement than the implicit motive. This is in line with

the assumption of dual motives theory that deliberate choices

between limited numbers of alternatives are guided mainly by

explicit motives (McClelland et al., 1989). However, the find-

ing that the implicit motive in postreproductive women made an

additional prediction suggests that more subtle and affective

processes play a role in the selection of living arrangement, too.

However, because this was a cross-sectional and correlational

analysis, we cannot rule out the reverse causal relation, that is,

living arrangement affected the agency motives. In fact, a dynamic

transactional relationship between motives and couples’ environ-

ments is plausible, as intimate relationships have been shown to

provide important contexts for personality development (Mund,

Finn, Hagemeyer,

When “Together“ Means “Too Close“: Agency Motives and Relationship Functioning in Coresident and Living-Apart-Together Couples

Abstract and Figures